Pathways and processes of phosphorus loss from pastures grazed by sheep

Alice Rowena Melland B.Agr.Sc. (Hons1), The University of Melbourne

A thesis submitted in total fulfilment of the requirement of the degree of Doctor of Philosophy

in the

School of Agriculture and Food Systems Institute of Land and Food Resources The University of Melbourne, Victoria, Australia

December 2003

ABSTRACT

Sheep producers in Victoria are applying more phosphorus (P) fertiliser and increasing stocking rates to increase production. At the same time, there is increasing awareness amongst research, community and producer groups that P-rich runoff water from agricultural land can contribute to the growth of undesirable algal blooms in surface waters. The loads, concentrations and forms of P in surface and subsurface hydrological flow pathways were estimated or measured directly on high and low P fertility hillslope plots in south-west Victoria to assess how this practice change could affect the P status of runoff and drainage water. Small- plot rainfall simulator studies were also conducted to investigate pasture management treatment effects. The spatial and temporal distribution of P loss from hillslope pastures, and the processes of P mobilisation in runoff identified in this study were used to identify appropriate management practises to help minimise P losses in runoff. Runoff P concentrations were then related to properties of pasture soils and runoff at a range of locations across Victoria to identify whether simple models and/or soil P tests could be used to predict P concentrations in runoff. Results from the study suggested that P concentrations in runoff from both low (8 kg P applied/ha) and high (25-40 kg P applied/ha) P fertility sheep pastures were unacceptably high relative to targets set for stream water quality in Victoria (annual mean concentrations >0.04 mg P/L), and that the dissolved reactive P fraction is the P fraction in runoff that increases the most as pasture P fertility increases. The volume of runoff was the primary factor influencing total P loads. Runoff was generated mainly by saturation excess flow during winter and spring from waterlogged soils that occupied <25% of the hillslope area. Subsurface hydrological pathways were only more important than surface pathways for P movement in dry years. Key management practises that are therefore recommended to minimise P losses from pastures in Victoria, are to avoid applying high rates of P fertiliser to landscape areas that become seasonally waterlogged, to consider retiring these areas from production and to maintain groundcover year-round to minimise erosion. An empirical model that uses a linear dissociation constant to describe the partitioning of P between runoff sediment and solution phases was reasonably well suited to predicting annual average DP concentrations in runoff from sheep pastures. The results also suggested, however, that factors such as runoff volume and fertiliser management practices (eg timing of application) should be considered in predictive models. The model was analogous to simple relationships developed between runoff P concentrations and soil test P. For the soils examined, which had a limited range of P sorption capacity (Pmax 862-1374 mg/kg), the POlsen soil test was a better predictor of runoff P concentrations than water extractable P or the degree of soil P

I saturation. Relationships between soil P status and hillslope runoff water quality need to be investigated for a wider range of soils, agricultural practices and catchments in Australia if environmentally critical soil P thresholds are to be set.

II DECLARATION

This is to certify that, (i) the thesis comprises only of my original work (ii) due acknowledgment has been made in the text to all other material used, (iii) the thesis is less than 100 000 words in length, exclusive of tables, figures and references

…………………………………… Alice Melland

Date

III Blank Page

IV ACKNOWLEDGMENTS

Firstly I would like to thank my supervisors Prof. David Chapman, Prof. Bob White and Dr. Malcolm McCaskill for the opportunity to undertake this degree, for your technical and scientific guidance and for the support and encouragement you offered throughout the project. From my interactions with you I have learnt the power of positive feedback and will endeavour to emulate this with colleagues in the future. I would also like to thank Dr. Jim Cox, Cyril Ciesiolka, Gavin Kearney and Dr. Tony Weatherley for specialised advice in hydrology, runoff measuring equipment, biometrics and soil chemistry respectively. Thanks also to Brendan Christy for guidance in using the daily soil water balance simulation spreadsheet which he, Bob White and others developed. For providing career development opportunities, encouragement and inspiration, I would like to thank Dr. George Riffkin, Prof. Nancy Millis, Geoff Saul and Dr. Cameron Gourley. I am also most grateful for the financial support of the DPI (then DNRE) Nancy Millis Postgraduate Award, the DPI Wool Program and in-kind support from SGS national experiment funding. Thanks also to Robert and Marilyn Lyons for allowing me to establish a peg and instrument farm on their sheep property and for their ongoing interest, support and enthusiasm for the research. To all the technical staff at DPI (then DNRE) Hamilton, you were an amazing team to work with and I have learned so much from you all. Thanks for your lateral thought, practical solutions, local wisdom, great company and above all your positive attitudes to work and life. In particular I would like to extend a big thank you to Jean Lamb – for providing the stamina and drive to complete 400 botanal readings in a day, for your competence, versatility, positive attitude and personal support, to Dion Borg – for managing the sheep and pastures at the Vasey field site and for teaching me how to pour concrete, use a shovel without breaking my back and for always finding a solution, and to Keiran Williams – for driving the drill rig on so many occasions, sharing philosophies on life and for the bentonite fights! Thanks also to Tim Plozza, Darren Gordon, Reto Zollinger, Paul Quinn, Belinda Gardener and the Farm staff for helping with soil sampling, fencing, rainfall simulating and lab support throughout the project. Thanks to all the staff at the Pastoral and Veterinary Institute in Hamilton for being so friendly and making me feel welcome, and for helping me out on matters of library searches, administration, finances, laboratory and sample processing techniques and for imparting aspects of your local knowledge with me. I would like to extend special thanks to Cassie Schefe for teaching me how to use the rainfall simulator, for giving me the opportunity to perform rainfall simulations at Maindample in conjunction with your Honours project, and for sharing your suspended sediment and runoff

V data with me. Thanks also to Andrew Smith from the Institute of Land and Food Resources (ILFR) at The University of Melbourne for performing the laboratory procedures for the soil water retention characteristics described in Chapter 3. Thanks also to Ron Teo at ILFR for laboratory and IT advice and assistance, and to Margaret O’Dowd and Catriona Ridland for trying to keep me on the administrative straight and narrow throughout the project. I am very grateful to George Croatto and the inorganic chemistry unit of the State Chemistry Laboratory in Werribee for instructing and assisting me in the use of soil grinding and analytical equipment (eg Milestone microwave digestion units, Carey UV-VIS spectrophotometer and the Skalar flow injection analyser). Thanks also to Ron Walsh (SCL) for performing the ICP-AES elemental analyses of my ammonium oxalate soil extracts. From the Department of Primary Industries Ellinbank I would like to thank David Nash for assisting with setting up the Vasey field site set up, providing access to the Ellinbank laboratory and a runoff field site in Gippsland, and historical data from that site as well as technical advice throughout my candidature, and David Halliwell for providing laboratory access at Ellinbank, and for instructing and assisting me in chemical methods and the use of analytical equipment (LaChat FI autoanalyser). Thanks also to Joanne Coventry and Megan Burns for laboratory assistance. Maindample event volume and P concentrations were kindly supplied for use in this thesis by Brendan Christy and Terry McClean at DPI Rutherglen. For camaraderie and peer support I am most grateful to my friends Raquel, Sam, Brendan, Leanne, buddies from the Western Outstationed Postgrad Student group and the Melbourne University Volleyball Club, and to Lucy, previously of La Trobe University, for being my travel companion to my first overseas conference, for hours of stimulating conservation about P (I mean it!) and for kindly sharing P sorption data for 90 agricultural soils across Australia with me. A big thanks also to fellow postgrads in ILFR Room 201, particularly to Li Yong for managing to teach me the basics of Linux computer language!, to Karen for your counsel and enthusiasm, and to Marlos, Arshad, Rodrigo, Najib and Lilanga for sharing some of your data, humour and cultures with me. I’ve already referred to over 50 people who have helped smooth out some of the bumps during the ride of my PhD journey. There are a few people, however, without whom I simply could not have reached my final destination. To my dear friends Megan, Bettina and Cheryl and to Mum, Rachel and Dad, I cannot thank you and your partners enough for your unconditional support, encouragement, patience and love. And likewise for Dave, – an extra special thank you for travelling so much of the journey with me. This is for you all…enjoy the ride!!☺

VI TABLE OF CONTENTS

ABSTRACT ...... I

DECLARATION...... III

ACKNOWLEDGMENTS...... V

TABLE OF CONTENTS...... VII

LIST OF TABLES...... XI

LIST OF FIGURES...... XIV

SYMBOLS AND ABBREVIATIONS ...... XIX

CHAPTER ONE...... 1

1INTRODUCTION ...... 1 1.1 Background, and problem statement...... 1 1.2 Aims ...... 3 1.3 Research approach and thesis outline ...... 3

CHAPTER TWO...... 5

2LITERATURE REVIEW...... 5 2.1 Phosphorus enrichment of waterways ...... 5 2.2 P cycle in pastures ...... 7 2.2.1 P pools in the soil-plant-animal system...... 7 2.2.2 Efficiency of P cycling in grazed pastures – ‘loss factors’...... 10 2.3 Forms of P in natural waters ...... 11 2.3.1 Inorganic and organic P fractions...... 11 2.3.2 Particulate and dissolved P fractions...... 15 2.4 Pathways of water movement in pasture systems...... 16 2.4.1 Surface Runoff...... 16 2.4.2 Subsurface flow ...... 17 2.4.3 Factors affecting hydrological pathways and runoff mechanisms...... 17 2.4.4 The location and extent of surface saturation zones...... 20 2.5 P mobilisation processes...... 22 2.6 Factors affecting P mobilisation from pastures...... 24 2.6.1 Land management and environmental conditions ...... 24 2.6.2 Pathways, concentrations and loads of P movement from pastures ...... 25 2.6.3 Mobility of pasture P pools...... 32 2.6.4 Amount and spatial and temporal distribution of P in grazed pastures ...... 34

VII 2.7 Experimental approaches to measuring P losses...... 37 2.8 Predicting P losses from pastures...... 37 2.9 Management strategies for minimising P losses in runoff ...... 40 2.10 Knowledge Gaps...... 41

CHAPTER THREE ...... 43

3CHARACTERISATION AND EXPERIMENTAL DESIGN OF THE VASEY FIELD SITE ...... 43 3.1 Introduction ...... 43 3.2 Climate, temperature and evaporation ...... 44 3.3 Landuse and management...... 44 3.4 Geology of the Dundas Tablelands...... 47 3.5 Soil description ...... 47 3.6 Soil physical properties...... 53 3.6.1 Bulk density ...... 53 3.6.2 Hydraulic conductivity...... 54 3.6.3 Soil water retention characteristics...... 57 3.7 Soil chemistry...... 59 3.8 The potential for movement of P in surplus water ...... 60 3.9 Experimental design of the hillslope runoff site...... 61

CHAPTER FOUR ...... 64

4HYDROLOGICAL CHARACTERISTICS OF THE VASEY RUNOFF SITE...... 64 4.1 Introduction ...... 64 4.2 Materials and methods...... 65 4.2.1 Rainfall...... 65 4.2.2 Evapotranspiration ...... 66 4.2.3 Surface runoff ...... 66 4.2.4 Surface soil water content ...... 68 4.2.5 Soil profile water content...... 68 4.2.6 Infiltration capacity...... 70 4.2.7 Water tables and subsurface flow...... 70 4.2.8 Soil water balance ...... 71 4.3 Results...... 72 4.3.1 Rainfall and evapotranspiration ...... 72 4.3.2 Surface runoff ...... 77 4.3.3 Surface waterlogging ...... 85 4.3.4 Soil profile water storage and water tables...... 88 4.3.5 Perched and groundwater tables...... 90 4.3.6 Subsurface vertical and lateral flows...... 94 4.3.7 Water balance...... 103 4.4 Discussion...... 105 4.4.1 Hillslope hydrological pathways and processes ...... 105 4.4.2 Surface runoff processes ...... 111

VIII 4.4.3 Spatial and temporal distribution of hydrological processes ...... 113 4.4.4 Modelling hydrological flow pathways...... 117 4.4.5 Implications for hydrological processes at a larger scale ...... 119 4.4.6 Implications for P movement via hydrological pathways ...... 120 4.5 Conclusion ...... 121

CHAPTER FIVE...... 122

5PHOSPHORUS, NITROGEN AND SEDIMENT IN RUNOFF AND DRAINAGE FROM PASTURES AT VASEY 122 5.1 Introduction ...... 122 5.2 Materials and methods...... 123 5.2.1 Sample collection...... 123 5.2.2 Chemical analyses...... 127 5.2.3 Methods used to calculate nutrient concentrations in hillslope runoff ...... 129 5.2.4 Statistical analysis...... 129 5.3 Results...... 130 5.3.1 Concentrations of P in surface and subsurface flows ...... 130 5.3.2 Concentrations of suspended sediment and N...... 134 5.3.3 Nutrient loads in runoff and subsurface flows...... 139 5.3.4 Factors influencing P losses ...... 142 5.4 Discussion...... 153 5.4.1 P concentrations in surface runoff and soil-water ...... 153 5.4.2 N concentrations and loads in runoff and soil-water...... 156 5.4.3 P loads in surface and subsurface flows and their environmental implications...... 156 5.4.4 Factors affecting P loss in runoff ...... 158 5.4.5 Experimental approaches: strengths and limitations ...... 163 5.5 Conclusions...... 165

CHAPTER SIX...... 167

6FORMS OF P AND PROCESSES OF P MOBILISATION IN RUNOFF ...... 167 6.1 Introduction ...... 167 6.1.1 Mobilisation of particulate and dissolved P fractions...... 168 6.1.2 Aims...... 173 6.2 Methods...... 173 6.2.1 Sample collection...... 173 6.2.2 Fractionation of runoff water ...... 174 6.2.3 Soil Olsen and total P...... 176 6.2.4 Statistical analysis...... 176 6.3 Results...... 177 6.3.1 Treatment effects on forms of P in surface runoff...... 177 6.3.2 Other factors affecting PP concentrations in runoff ...... 182 6.3.3 Other factors affecting DP concentrations in runoff...... 187 6.3.4 Effect of runoff volume on P concentrations in runoff...... 190 6.4 Discussion...... 191

IX 6.4.1 Mobilisation of particulate P...... 191 6.4.2 Mobilisation of dissolved P...... 194 6.4.3 Effect of flow volume on runoff P concentrations and fractions...... 199 6.4.4 Scale effects on P concentrations in runoff ...... 200 6.4.5 Mobilisation of organic P...... 202 6.4.6 Options for minimising P losses from pastures...... 202 6.5 Conclusion ...... 203

CHAPTER SEVEN ...... 205

7PREDICTING P CONCENTRATIONS IN RUNOFF USING SOIL P CHARACTERISTICS ...... 205 7.1 Introduction ...... 205 7.1.1 Soil P quantity and intensity tests ...... 205 7.1.2 Indices of soil P saturation ...... 207 7.1.3 Soil P thresholds for water quality targets...... 210 7.1.4 Aims...... 210 7.2 Materials and methods...... 211 7.2.1 Victorian pasture soils...... 211 7.2.2 Other soil P datasets...... 216 7.2.3 Runoff P concentrations...... 216 7.2.4 Statistical analysis...... 217 7.3 Results...... 217 7.3.1 Soil P characteristics ...... 217 7.3.2 Relationship between P concentration in runoff and soil P...... 225 7.4 Discussion...... 230 7.4.1 Soil P characteristics ...... 230 7.4.2 Relationship between the P concentration in runoff and soil P characteristics...... 234 7.5 Conclusions...... 242

CHAPTER EIGHT ...... 243

8GENERAL DISCUSSION AND CONCLUSIONS ...... 243 8.1 Hydrological pathways of P movement from sheep-grazed pastures ...... 243 8.2 The impact of increasing P fertiliser application and stocking rates on the loads and concentrations of P in surface runoff ...... 244 8.3 Important processes of P mobilisation in runoff...... 244 8.4 Recommendations for minimising P losses from pastures ...... 245 8.5 Using soil P tests and simple models to predict P concentrations in runoff...... 246 8.6 Further research ...... 247

REFERENCES ...... 249

APPENDIX A: TIPPING BUCKET FLOW RATE CALIBRATION EQUATIONS ...... I

APPENDIX B: PERCHED AND GROUNDWATER TABLES...... II

X APPENDIX C: STORAGE EFFECTS ON MRP CONCENTRATIONS...... VI

LIST OF TABLES

Table 2-1: Definitions and descriptions of P and sediments fractions of runoff water...... 13 Table 2-2: P concentrations, forms and loads in runoff and drainage...... 29 Table 3-1: Classification and description of the major soil types at Dundas Park, Vasey (Cox et al. 1998)...... 48 Table 3-2: Mean, standard error and number of soil bulk density (g/cm3) measurements for major soil horizons at Vasey...... 53 Table 3-3: Alpha values (1/cm) for the major soil horizons at Vasey...... 55

Table 3-4: Summary statistics for approximated values of Ksat (m/day) of major soil horizons. 56 Table 3-5: Volumetric water contents (cm3/cm3) of major soil horizons for a range of matric potentials (kPa) ...... 58 Table 3-6: P fertiliser application rates at the Vasey runoff site ...... 63 Table 4-1: Terminology and definitions of hydrological pathways and processes ...... 64 Table 4-2: Frequency distribution of rainfall and rainfall intensity during the whole year and over the runoff period ...... 75 Table 4-3: Maximum rainfall intensities (mm/h) over a 6 minute duration expected during each season at Rocklands Reservoir in rain events which recur once every 1 or 5 years ...... 76 Table 4-4: Volume and number of surface runoff events from the hillslope plots and flow in Dundas River in 1998 to 2000 ...... 77 Table 4-5: Distribution of runoff volume across levels of rainfall intensity (mm/h, 5-minute duration)...... 82 Table 4-6: A comparison of monthly runoff coefficients (runoff/rainfall, %) at a plot, hillslope and sub-catchment scale ...... 83 Table 4-7: The mean rates (mm/h) of rise and fall of water tables in the A horizons of plots 2, 3 and 4...... 95 Table 4-8: Range of vertical hydraulic head gradientsA (m/m) within piezometer nests...... 96 Table 4-9: The rangeA and directionB of groundwater and perched flow water rates (m/day x 10- 2) between upslope, midslope and lowerslope positions in each plot ...... 98 Table 4-10: Partitioning of water (mm) in the four runoff plots using a simulated water balance ...... 104 Table 5-1: Treatment descriptions for rainfall simulator plots at Vasey...... 124 Table 5-2: Treatment characteristics and TP concentrations in surface runoff and soil water of Vasey hillslope plots...... 132

XI Table 5-3: Runoff volume and nutrient concentrations in autumn runoff (14 April 2000)...... 132 Table 5-4: TP concentrations, pH and EC of rainfall and wetland flow sampled in 2000...... 133 Table 5-5: Treatment means, and standard errors in parentheses for P, N and SS concentrations in simulated runoff at VaseyA ...... 135 Table 5-6: Mean SS and N concentrations in hillslope surface runoff (FWMs) and soil water with standard errors in parentheses...... 138 Table 5-7: P loads in hillslope runoff events yielding at least 5 mm flow...... 141 Table 5-8: Estimates of total P flux in subsurface drainage from hillslope plotsA...... 142 Table 5-9: Treatment mean groundcover and runoff characteristics of rainfall simulator plotsA ...... 143 Table 5-10: Treatment mean soil, herbage and dung P in plots used for simulated rainfall with standard error in parenthesesA...... 146

Table 5-11: Mean POlsen in hillslope runoff plots on six occasions from 1998 to 2000, with standard errors in parenthesesA ...... 148

Table 5-12: Mean POlsen in stock camp and non-camp areas in October 2000 with standard errors in parenthesesA...... 150

Table 6-1: Pasture treatments, POlsen (0-10 cm), annual flow-weighted mean P concentrations and percentages of DP and DRP in hillslope runoff at Vasey ...... 177 Table 6-2: Mean P concentrations and percentages of DP and DRP in simulated runoff, and

POlsen (0-5 cm) for four pasture management treatments at Vasey...... 179

Table 6-3: POlsen (0-5 cm depth) and concentrations of P fractions in simulated runoff from high, medium and low P fertility pasture at MaindampleA ...... 181 Table 6-4: Sediment P concentration and PER for natural and simulated runoff at Vasey and Maindample ...... 185 Table 6-5: EC in runoff from 0.5 ha hillslope plots at Vasey ...... 188 Table 6-6: Arithmetic mean concentrations of P and the percentages of TP in PP, DP and DRP forms for small and large runoff events in plot 1 at Vasey...... 190 Table 6-7: Estimated and measured mean DP concentrations (mg/L) in hillslope and simulated runoff ...... 197 Table 7-1: Chemical description and extracting conditions for common soil P extractants and their relationships with runoff P concentrations ...... 206 Table 7-2: Selected properties of topsoil (0-10 cm) from five pasture sites in Victoria ...... 214

Table 7-3: POlsen, Pox and EPC of a) five pasture soils (0-5 cm) in Victoria. N=3 for all treatments and b) 90 soils (0-10 cm) from agricultural regions in Australia...... 218

Table 7-4: Pmax and Psat of a) five pasture soils (0-5 cm) in Victoria. N=3 for all treatments and b) 90 soils (0-10 cm) from agricultural regions in Australia ...... 221

XII Table 7-5: Correlation coefficients (r) between estimates of Pmax and Alox, Feox and Pox for five pasture soils (N=27)...... 223 Table 7-6: Regression models and variance accounted for (R2) for the relationships between TP

or DP in runoff (mg/L) and POlsen (mg/kg), Pwater (mg/kg) or PsatL (%) at Vasey (V) and Maindample (M) and Ruffy (R)...... 226

Table 7-7: Mean POlsen, PsatL, runoff TP and runoff volume for each plot year of runoff from hillslopes at Darnum, Maindample, Ruffy and Vasey...... 227 Table C-1: Statistics for TRP and DRP analysed within 24 h and after being stored frozen for 4 months...... vii Table C-2: Statistics for MRP concentrations measured using manual and automated methods ...... viii

XIII LIST OF FIGURES

Figure 3-1: Location of the Vasey field site in south-west Victoria, Australia (Maps from http://audit.ea.gov.au/ANRA/atlas_home.cfm and www.dpi.gov.au/vro)...... 43 Figure 3-2: Long term (1962-1990) mean monthly rainfall (solid line) and pan evaporation (broken line) at the Pastoral and Veterinary Institute, Hamilton (National Climate Centre, Bureau of Meteorology, 1999)...... 44 Figure 3-3: Vasey field site showing layout of the SGS experiment and the location of the runoff plots, soil pits (Pits 1-6) described by Cox et al. (1998) and soil pits used for measuring bulk density and hydraulic conductivity characteristics (Pits 7-11)...... 46 Figure 3-4: Soil profile of a Brown Chromosol in a midslope position at Vasey ...... 49 Figure 3-5: Approximate depth to the B and C horizons along the toposequence of the hillslope runoff plots at Vasey. Dashed lines represent B and C horizons...... 52 Figure 3-6: Layout of the runoff plots at Vasey showing elevation contours and positions of the neutron probe access tubes, nests of piezometers and ceramic cup samplers and permanent soil sampling points ...... 62 Figure 4-1: Hillslope runoff plots at Vasey...... 67 Figure 4-2: Tipping bucket flow meters and sample splitters used to measure and sub-sample surface runoff at Vasey ...... 67

Figure 4-3: Volumetric water content (θv, y axis) plotted against the count ratio (CR, x axis), 2 and the fitted calibration curve; lnθv = 1.50 lnCR - 0.52, R =0.775, P<0.001...... 69 Figure 4-4: Seasonal rainfall at the Vasey runoff site compared to the long term average...... 73

Figure 4-5: Monthly rainfall and ETp at the Vasey site over the experiment period...... 73 Figure 4-6: Rainfall and runoff from individual runoff plots in 1998...... 79 Figure 4-7: Rainfall and runoff from individual runoff plots in 2000...... 80 Figure 4-8: The frequency distribution of runoff events of increasing volume ...... 81 Figure 4-9: Runoff hydrographs and cumulative rainfall on a) 25 Sep, b) 2 Nov and c) 8 Oct, 2000 ...... 84 Figure 4-10: Mean volumetric water content of the surface soil (0-12 cm) over the lower 20 m of the wettest transect in each plot in winter and spring 2000...... 85 Figure 4-11: Patterns of surface soil water (0-12 cm) accumulation in the lower quarter of plot 1 on a) 7 September, b) 20 October and c) 3 November 2000 ...... 87 Figure 4-12: Mean and actual SWD (mm to 1.5 m soil depth) of the four topographic positions from 1 (upslope) to 4 (toeslope) in each of the plots over time. Maximum water storage (mm) at SWD = 0 indicated in parentheses ...... 89

XIV Figure 4-13: Degree of water saturation of the soil profile over time in midslope positions of a) plot 1 and b) plot 2, showing perched water tables...... 91 Figure 4-14: Degree of water saturation of the soil profile over time in the upper slope positions of a) plot 3 and b) plot 4 ...... 92 Figure 4-15: Rainfall, and shallow water table heights from August to November inclusive in 1998 and 2000 measured in dipwells at the base of all four plots ...... 93 Figure 4-16: Rate of rise in the A horizon water table (mm/h) against the peak antecedent rainfall intensity in 2000 ...... 95 Figure 4-17: Cross-sections of each plot showing surface, and interpolations of B and C soil horizons (dashed lines) and hydraulic head elevation above sea level (m) on 22 Sep 00 or 9 Sep 00 (black line) and 16 Oct 01 (grey line). Symbols indicate water levels in 2.9 m (■) and 1.4 m (▲) piezometers...... 97 Figure 4-18: Equi-head contours for 2.9 m piezometer water levels, and elevation contours (m) on 22 September 2000 ...... 99 Figure 4-19: Equi-head contours for 2.9 m piezometer water levels, and elevation contours (m) on 15 September 1999 ...... 100 Figure 4-20: Equi-head contours for 1.4 m piezometers, and elevation contours (m) on 15 September 2000 ...... 101 Figure 4-21: Equi-head contours for 1.4 m piezometers, and elevation contours (m) on 15 September 1999 ...... 102 Figure 4-22: Simulated and measured values of SWD (mm) in plot 2 ...... 103 Figure 4-23: Simplified representation of the hillslope hydrological model of Dahlhaus and MacEwan (1997) (not to scale)...... 106 Figure 5-1: Rainfall simulator and runoff plot ...... 125

Figure 5-2: Relationships between FWM TP or TPaut concentrations and mean POlsen for each 2 hillslope plot. Linear trends for TPaut (solid circles and line, R = 0.79, P<0.01) and TP (crosses and dashed line, R2 = 0.64, P<0.05) concentrations are shown...... 133 Figure 5-3: Relationship between TP concentration and the cumulative volume of soil water sampled within and below the rootzone across all plots and sample times ...... 135 2 Figure 5-4: Relationship between mean TP concentration in simulated runoff and plot POlsen, R = 0.42, P<0.001...... 136

Figure 5-5: Suspended sediment concentration in simulated runoff with increasing POlsen ...... 136 Figure 5-6: Equivalent loads (kg/ha) of P, N and SS in runoff from hillslope plots...... 140 Figure 5-7: Relationship between P load and volume of surface flow in runoff events from hillslope plots...... 141 Figure 5-8: Percentage bare ground across hillslope plots after storm in April 2000...... 144

XV Figure 5-9: Location of stock camps (dashed lines) and percentage dung cover across hillslope plots in April 2000. Elevation (m) contours also marked and labelled...... 145

Figure 5-10: Herbage P concentration (%) with increasing POlsen (0-5cm)(mg/kg). Trendline shows fitted regression curve where Herbage P concentration = 0.41-0.55*0.89POlsen P<0.001, R2=0.63...... 147

Figure 5-11: Dung P concentration (%) with increasing POlsen (0-5cm)(mg/kg). Trendline shows fitted regression curve where Dung P concentration = 1.10-1.14*0.92POlsen, P<0.001, R2=0.38 ...... 147 Figure 5-12: Relationship between pasture dry matter production (t/ha) and herbage P concentration (%) of quadrat samples from simulated rainfall plots at Vasey ...... 148

Figure 5-13: Change in POlsen with soil depth for high (solid line) and low (dashed line) fertility hillslope plots and depth-weighted 10 cm mean values for high (□) and low (о) fertility treatments. Error bars indicate ± standard error...... 149

Figure 5-14: Distribution of soil (0-10cm) POlsen (mg/kg) measured at 102 permanent sample points in hillslope plots in a) June 1998, b) September 1998, c) March 1999 and d) September 1999 ...... 151

Figure 5-15: Distribution of soil (0-10cm) POlsen (mg/kg) at 102 permanent sample points in hillslope plots in a) April 2000 and b) October 2000 ...... 152 Figure 6-1: Analytical fractionation of P in runoff samples...... 175 Figure 6-2: Flow weighted annual mean concentrations of P fractions in runoff from hillslope

plots at Vasey of contrasting mean POlsen (0-10cm). Solid line shows the fitted 2 regression curve for DRP, where DRP = 0.0147exp(0.208*POlsen) R = 0.47 P<0.05 178 Figure 6-3: Relationships between runoff concentrations of DRP, DUP, PRP and PUP fractions

(a-d) and plot POlsen for individual simulator runoff plots at Vasey with regression lines and statistics...... 179 Figure 6-4: Treatment mean TP, DP, PP (mg/L) and SS (g/L) concentrations in hillslope (●) and

small-plot simulator (□) runoff at Vasey with increasing POlsen (0-10cm depth). Bars indicate standard errors. Means from a hillslope plot that generated <0.1mm runoff, as indicated by the symbol x, were excluded from the analyses...... 180 Figure 6-5: Mean concentrations of P fractions for 3 runoff events from the low, and 6 events from the medium and high fertility catchments at Maindample in 2000...... 181 Figure 6-6: Mean concentrations of P fractions in simulated runoff from low, medium and high fertility treatments at Maindample in 1999. Line shows fitted regression curve for DRP from individual plot data, P<0.001, R2= 0.54...... 182 Figure 6-7: The relationship between annual FWM concentrations of PP or DP and SS concentrations in runoff from hillslope plots at Vasey...... 183

XVI Figure 6-8: Relationship between PP and SS concentrations in simulated runoff across four pasture treatments at Vasey. Line shows regression curve where PP = 2.10*SS-0.025, P<0.001, R2=0.43...... 183 Figure 6-9: Relationship between PP and TS in simulated runoff from low, medium and high fertility pasture treatments at Maindample. Regression lines shown for High (dashed line, PP=3.992*TS-0.527) and Low (solid line, PP=0.899*TS+0.105) treatments, P<0.001, model R2 = 0.76...... 184

Figure 6-10: Sediment P concentration in simulated runoff from individual plots vs POlsen (0- 5cm depth) at Vasey (solid circles) and Maindample (open circles). Line shows

regression for combined dataset where Sediment P concentration = 0.04* POlsen +0.91, P<0.001, R2=0.35...... 186 Figure 6-11: The relationship between the proportion of total runoff P in dissolved form and the flow-weighted sediment concentrations for annual runoff from hillslope plots at Vasey ...... 187 Figure 6-12: The relationship between the DP concentrations and the EC of runoff events from plot 1 at Vasey in a) 1998 where DP= -3.86*EC+0.30, R2= 0.90, P<0.001 and b) 2000 where DP = exp(-5.85*EC+0.41), R2=0.64, P<0.001 ...... 189 Figure 6-13: FWM annual TP concentrations in hillslope runoff from plots at Vasey with increasing annual runoff volume. Open circles represent plot 1 runoff...... 190 Figure 6-14: Quantity-intensity relationship between the sediment total P concentration (mg/kg) and DP (mg/L) in simulated runoff with lines showing regressions for Vasey (broken line, Sediment P concentration = exp(0.41lnDP + 1.20), P<0.001, R2=0.31) and Maindample (solid line, Sediment P concentration = 2.17*DP+0.84, P<0.001, R2=0.49) ...... 197

Figure 7-1: Relationships between Pwater and POlsen for Maindample (open circles and broken line, R2 = 0.90) and Vasey soils (0-5 cm depth) (solid circles and line, R2 = 0.71)..... 219

Figure 7-2: Relationships between Pwater and POlsen for nine SGS national experiment site soils (0-10 cm depth)...... 219 Figure 7-3: P sorption isotherms (symbols) for Vasey soil (0-5 cm) and fitted Langmuir

isotherms (lines) using POlsen (+, solid line) or Pox (o, dashed line) as values for Q..... 220

Figure 7-4: Langmuir P sorption maxima (Pmax) of soil at depths of 0-5, 5-10 cm and the B horizon at five pasture sites in Victoria ...... 222

Figure 7-5: Relationship between the Langmuir Pmax and the sum of Alox and Feox for soils (0-

5cm) from five pasture sites with the fitted model, Pmax =513ln(Alox+Feox)-3180 (P<0.001 R2=87.5) ...... 223 2 Figure 7-6: Relationship between Psatox(open circles, R =0.92, P<0.001) or PsatL (closed cirles, 2 R =0.99, P<0.001) and POlsen (0-5cm, mg/kg) across five soil types (n=27)...... 224

XVII Figure 7-7: Frequency distribution (bars) and cumulative percentage (line) of the PsatL (%) of 90 Australian soils. ‘More’ includes soils with very low P sorption capacities...... 225

Figure 7-8: Relationship between hillslope runoff TP concentrations and POlsen (0-5 cm depth) for Maindample, Ruffy and Vasey soils, with a line representing a single regression for 2 all three soils (TP=exp (0.073POlsen-2.4), R =0.57, P<0.001)...... 227 Figure 7-9: Relationships between the annual mean DP concentrations in runoff (∆) and the soil

EPC (●) and POlsen (0-5cm)(mg/kg) for four pasture soils in Victoria in 2000. DP was estimated for Ruffy and Maindample runoff as TP – 0.05mg/L (see text above). Fitted 2 logarithmic equations are indicated for EPC (POlsen = 15.4 lnEPC + 65.8 R = 0.77) and 2 DP (POlsen = 17.3 lnDP + 47.8 R = 0.73)...... 228 Figure 7-10: The relationship between the annual runoff volume and the mean annual TP concentration in runoff from four pasture sites...... 229 Figure 7-11: Relationships between DP (or TP for the combined Victorian sites) concentrations

in runoff and adjusted POlsen (0-5cm) for soils described in Table 7-1 and Table 7-6.

Vertical dashed lines indicate optimum POlsen range for pasture production in Victoria ...... 239 Figure B-1: Perched and groundwater table heights measured in a) lowerslope, b) midslope and c) upper slope positions in plot 1. Shaded regions indicate the depth of soil moisture accumulation indicated by neutron probe measurements...... ii Figure B-2: Perched and groundwater table heights in plot 2 in a) lowerslope and b) mid slope positions. Shaded regions indicate the depth of soil moisture accumulation indicated by neutron probe measurements...... iii Figure B-3: Perched and groundwater tables in plot 3 in a) lowerslope, b) midslope and c) upper slope positions. Shaded regions indicate the depth of soil moisture accumulation indicated by neutron probe measurements...... iv Figure B-4: Perched and groundwater tables in plot 4 in a) lowerslope, b) midslope and c) upper slope positions. Shaded regions indicate the depth of soil moisture accumulation indicated by neutron probe measurements...... v

XVIII SYMBOLS AND ABBREVIATIONS

Abbreviations

Alox Ammonium oxalate extractable aluminium ARI Average recurrence interval AWSC Available water storage capacity BD Bulk density CR Count ratio DP Dissolved phosphorus DRP Dissolved reactive phosphorus DUP Dissolved unreactive phosphorus EPC Equilibrium P concentration ER Enrichment ratio ESP Exchangeable sodium percentage

ETp Potential Evapotranspiration

Feox Ammonium oxalate extractable iron FWM Flow-weighted mean IFD Intensity-Frequency-Duration kDa Kilodalton, atomic mass unit

Ko Hydraulic conductivity

Ksat Saturated hydraulic conductivity

KsatD-infil Absolute infiltration rate from dipwell data

KsatD-rise Net infiltration rate from dipwell data

KsatD-fall Drainage rate from dipwell data MRP Molybdate reactive phosphorus N Nitrogen P Phosphorus PER Phosphorus enrichment ratio

Pmax Langmuir P sorption maximum

POlsen Soil Olsen P content

Pox Ammonium oxalate extractable phopshorus PP Particulate phosphorus PRP Particulate reactive phosphorus Psat Degree of P saturation

PsatL Langmuir P saturation index

Psatox Oxalate P saturation index

Psoil Total soil P content PUP Particulate unreactive phosphorus Q Quantity factor, amount of previously sorbed P RC Runoff coefficient

XIX SE Standard Error SGS Sustainable Grazing Systems SS Suspended sediment STP Soil test P SWD Soil water deficit TDS Total dissolved salts TP Total phosphorus Trt Treatment TS Total solids VSA Variable source area Symbols

θg Gravimetric soil water content

θsat Saturated soil water content

θv Volumetric soil water content ψ Soil matric potential

XX CHAPTER ONE

1 Introduction

1.1 Background, and problem statement

Livestock production systems throughout the high rainfall zone (>600 mm annual rainfall) of Australia are based on the use of pastures as the main source of feed for animals. Pastures commonly comprise a mixture of grasses and legumes, and maintenance of the botanical composition of these pastures depends on inputs of fertiliser to supplement the naturally low nutrient availability in Australian soils (Costin and Williams 1983; Wilson and Simpson 1993). Phosphorus (P) is the main nutrient supplied as fertiliser because it increases the nitrogen (N)-fixing potential of legumes, which supply N to the non-legume pasture species (Sale and Blair 1997). High soil P availability can also improve the nutritive value and drought tolerance of some pasture species (Lewis and Sale 1993) Significant economic gains can be made in sheep and cattle grazing systems which incorporate a combination of ‘sub’ (Trifolium subterranean), ‘super’ (superphophate fertiliser), and stocking rates that make optimal use of available pasture (Malcolm et al. 1996). In south-west Victoria, producers are encouraged to apply P fertiliser to achieve and maintain an agronomically optimum soil Olsen P level of around 14 mg/kg (Court et al. 1998; Cayley et al. 2002). Phosphorus fertiliser is normally applied to the surface as inorganic granular formulations between March and June. Prior to 1992, annual fluctuations in fertiliser applications were often related to wool prices (Patterson 1992). However, between 1993 and 1996 the district average rate of P application for sheep and beef farms increased from 3.7 to 7.3 kg P/ha (Cayley and Kearney 1999) despite static wool prices because of research data that linked profitability to fertiliser inputs. A pilot extension program known as the Grasslands Productivity Program (GPP) (Court et al. 1998) was instrumental in changing wool grower attitudes to fertiliser application. An evaluation of the program by Trompf (2001) showed that the average rate of P applied by 14 participants of the GPP program was 13.8 kg P/ha by 1997-98, which was significantly higher (P<0.01) than the 9.3 kg P/ha applied to 21 farms that were not involved in the program. By 1997-98, the stocking rates and net farm incomes on GPP farms were 15% and 46% higher respectively than on non-participant farms (Trompf 2001). Rotational grazing systems that optimise pasture growth and utilisation by sheep and hence allow for higher

1 stocking rates, have also been investigated and adopted throughout south-west Victoria (Waller et al. 2001a, b; Warn et al. 2001; Allan et al. 2003; Chapman et al. 2003). However, as adoption of more intensive grazing systems has increased, so has awareness amongst research, community and producer groups, of the potentially adverse impacts of farming practices on the wider environment (Andrew and Lodge 2003). Particular attention has been paid to the potential for P-rich runoff water from agricultural land to contribute to the growth of undesirable algal blooms in surface waters. Whilst most blooms reduce water quality and aesthetic appearance, up to two thirds of blooms also produce toxins harmful to stock and humans and they are costly to mitigate. Blooms associated with runoff from grazed pastures may also have an adverse impact on the marketability of agricultural produce due to increasing demands for clean and green products by consumers (Williams and Hook 1998). Already the cost of blue-green algal blooms to the entire Glenelg-Hopkins catchment in south-west Victoria is estimated at $1.9 million per year (Holmes 2000) of which over $1 million is borne by those affected by poor quality domestic and stock water. In the Glenelg-Hopkins catchment in south-west Victoria, cyanobacterial blooms of predominantly Anabaena and Microcystis spp. have been reported at 16 sites at varying frequencies, mainly in lakes and reservoirs (Holmes 2000). These species are known to produce harmful toxins such as hepatotoxins, neurotoxins and endotoxins which can lead to liver damage, muscle paralysis and skin irritation respectively (Kotak et al. 1993). Between 1991 and 1997, concentrations of total P and N at selected sites in rivers often exceeded guideline levels of 0.035 mg P/L and 1.0 mg N/L (Wagg 1997). Higher than desired P concentrations in some waterways and farm dams in the catchment (Wagg 1997; Sherwood et al. 1998; Turoczy 1999), and a weak trend for increasing TP and TN concentrations through time (Cottingham et al. 1995), highlighted that a reduction in nutrient input to waterways was required to prevent an increase in eutrophication problems (Wagg 1999b). The ratio of TN:TP in water samples collected along 260 km of the largest river in the catchment indicated P was the limiting nutrient for algal growth (Sherwood et al. 1998). Because P is also the main nutrient applied as fertiliser in the southwest of Victoria, managing the transfer of P from pastures to waterways was identified as having the potential to mitigate eutrophication problems. The environmental implications of pasture P management is therefore the focus of this research. Dryland pasture comprises 71% of the land use in the 2.6 million ha Glenelg-Hopkins catchment (Holmes 2000). Based on an estimated P load of 0.09 kg P/ha, a catchment scale nutrient export model attributed 53% of P input to waterways to exports from dryland pasture in this catchment (Wagg 1999a). The relative magnitude of loads from diffuse sources was approximately in proportion to the relative areal prominence in the catchment with the exception of native parks, which were a small nutrient contributor despite being the second

2 largest landuse (Wagg 1999a). That dryland pasture is the dominant landuse further demonstrates that the way pastures are managed has the potential to influence the quality of catchment surface waters. Combined, these factors led to concern from sheep producers and the wider community that increases in fertiliser use and stocking rates could increase P losses from pastures to waterway in south-west Victoria. This concern was reflected in the final report to members of the Grasslands Productivity Project, in which the implication of ‘P movement’ from pastures was listed as a priority issue requiring further research (Court et al. 1998). Information on the potential movement of P from grazed systems to waterways is needed to complement research on production responses from fertiliser (Cayley et al. 1998), P balances in pasture systems (McCaskill and Cayley 2000), and nutrient movement within stream networks (Wagg 1997; Sherwood et al. 1998) in south-west Victoria. The research described in this thesis was initiated in response to these concerns and needs.

1.2 Aims

The aims of this research were to identify the hydrological pathways of P movement from pastures grazed by sheep in south-west Victoria, to assess the impact of increasing P fertiliser application and stocking rates on the loads and concentrations of P in surface runoff, and to identify the important processes of P mobilisation in runoff. The findings from this research were used to develop recommendations for managing pasture systems in ways that minimise P losses. A final aim was to test whether simple empirical models and soil P characteristics could be used as a practical alternative to field experiments to predict P concentrations in runoff for a range of pasture soils in Victoria.

1.3 Research approach and thesis outline

In the following chapter of this thesis, (Chapter 2) a review of the pathways and processes of P losses from pastures both in Australia and overseas is presented. Chapter 3 then describes the location, physical characteristics and experimental design of a hillslope- scale experiment that was initiated to compare runoff quality and quantity from low and high P fertility sheep pastures in south-west Victoria. Implications of the field site characteristics for P movement are also discussed. The hydrology of catchments heavily influences nutrient exports so the measured hydrological characteristics of the hillslope study site are discussed in detail in Chapter 4. Chapter 5 describes the amounts and pathways of P loss from the hillslope pasture, as well as the design and results of a replicated rainfall simulator experiment. The simulator experiment was undertaken in the field at a micro-plot scale to further assess the differences

3 in P loss between the low and high P fertility pasture treatments and the processes of P loss in runoff. The physical and chemical forms of P in runoff from both the hillslope and small plot experimental scales are presented in Chapter 6, and the relevance of the small-plot simulator experimental results to processes of P mobilisation in runoff from the hillslope site is also discussed within Chapters 5 and 6. Chapter 6 also discusses the ability of existing models to predict the dissolved and particulate P concentrations measured in runoff. Further to this, the potential for predicting P concentrations in runoff from the soil P tests of a range of pastures across Victoria is discussed in Chapter 7. This multi-site comparison was undertaken to broaden the applicability of the field study results and place them into context with other grazed sites in Victoria. A general discussion in Chapter 8 describes the management implications of the experimental results for P movement from sheep pastures and highlights future research that would help further inform nutrient management decisions for pastures grazed by sheep.

4 CHAPTER TWO

2 Literature review

2.1 Phosphorus enrichment of waterways

Nutrients support the primary production of all terrestrial and aquatic flora and fauna. However, over-enrichment of nutrients, particularly phosphorus (P) and nitrogen (N) in fresh and saline waters (eutrophication) can cause a shift in the species structure of the aquatic ecosystems, which can lead to excessive growth of algae and cyanobacteria (commonly known as blue-green algae) (Harris 1994; Correll 1998). Excessive growth of algae can be problematic because some strains of cyanobacteria release toxins that threaten stock and human health (Falconer 1991), and because the algal biomass can physically disrupt infrastructure for pumping water, emit foul odours and deplete dissolved oxygen levels during decomposition (Anon 2000a). These problems can lead to stock deaths, human skin irritations and fish kills as well as reduced recreational amenity and closure of water reservoirs. P is often the limiting nutrient for algal growth because it is less readily exchanged between the atmosphere and the lithosphere than carbon and nitrogen (Correll 1998). Nutrient limitation in waterbodies is measured using the Redfield ratio of 105:15:1 which approximates the atomic ratio of C:N:P in algae (Redfield 1958). An atomic ratio greater than 15 N:P suggests the amount of P available will limit growth (Correll 1998). P, N and chlorophyll a concentrations are also currently used to determine the trophic state of rivers and streams (Anon 2000b). Guidelines for water quality in Australia and New Zealand suggest a set of ‘trigger levels’ for concentrations of total P and total N below which there is a low risk that adverse biological effects, in this case the growth of nuisance aquatic plants, will occur (Anon 2000b). In the absence of site-specific information, average trigger values of 0.037 mg TP/L and 1.60 mg TN/L were derived for lowland rivers and 0.050 mg TP/L and 0.44 mg TN/L for freshwater lakes. There are currently no recommended trigger levels for wetlands or farm dams although recent research suggests the age of farm dams and associated suspended sediment concentrations would influence aquatic primary production (Turoczy 1999). In Australian waters, P levels are sometimes very high (up to 4.5 mg P/L) without there being obvious signs of eutrophication (Williams and Wan 1972). This demonstrates that other factors such as high levels of turbidity and salinity, short water residence times, low sunlight penetration and low temperatures can also limit the growth of algae (Kirk 1977; Harris 1994). Low light availability reduces the energy source for algal photosynthesis. This is influenced by

5 shading provided by riparian vegetation, suspended sediment loads and self-shading by the algal biomass (Harris 1994). However, excess P loading can lead to algal blooms where flow residence times are long enough to allow particulate matter to settle thus increasing the light available for algal growth (Harris 1994). Conditions of low flow are often encountered in the water storages and in residual pools of regulated or ephemeral streams during the Mediterranean summers of southern Australia. Thermal stratification in calm waters also allows algal blooms to thrive undisturbed (Herath 1997). As well as this an anoxic layer can develop above bottom sediments, which creates reducing conditions that can then induce nutrient release from sediments which further fuels algal growth (Correll 1998; Anon 2000a). Blue-green algal blooms and deoxygenation of waterways are significant water quality problems arising from eutrophication across the globe (Kotak et al. 1993; Herath 1997; Gillingham and Thorrold 2000; Heathwaite and Dils 2000). In Australia, blooms recorded since the 1970s have occurred most frequently in Victoria and New South Wales (Herath 1997). Western Australia has also suffered from the effects of eutrophic inland and estuarine waters, with blooms on the Swan River, Peel Harvey estuary and Oyster Bay being precursors to an intense research effort into mitigative options (McComb and Davis 1993; Weaver and Prout 1993). The most infamous bloom in eastern Australia affected up to 1000 km of the Darling- Barwon river system and cost $10 million in lost recreation revenue (Herath 1997). The extensive impact of this bloom stimulated national and state government policy directives to improve the quality of catchment waters and reduce the incidence of these blooms (Anon 2000b; Holmes 2000; Environment Protection Authority Victoria 2001). A compilation of research aimed at identifying the source of P in Australian catchment waters revealed that the majority of P was from ‘diffuse’ or ‘non-point’ sources (Davis et al. 1998). Landuse that involves intensive animal production, high inputs of fertiliser, animal waste ponds and irrigation drains has the highest risk of P loss if not managed carefully (Davis et al. 1998). More critically, however, Davis et al (1998) highlighted that P sources varied between catchments and that local information on the most significant processes controlling P transport was required in order to identify the most appropriate mitigative management practices. Pastures are considered a rich source of P for mobilisation into waterways, particularly along surface runoff pathways because the interacting surface layer is enriched with P in soil, herbage, litter and dung material (Sharpley and Syers 1976; Sharpley 1996). P fertiliser is a key element in the management of clover-based pasture systems in Australia because high levels of plant available P are required by nitrogen-fixing clover species in order for them to supply adequate amounts of soil-N for grass species in the sward (Haynes and Williams 1993). Therefore in catchments where dryland grazing is a dominant land use, changes in fertiliser use and grazing intensity of pastures could potentially have a large impact on the water quality of the wider catchment.

6 2.2 P cycle in pastures

2.2.1 P pools in the soil-plant-animal system

Phosphorus plays a key role in the nutrition of all life forms. In plants, animals and micro-organisms, P is involved in energy transfers (eg ATP), structural components of cell wall membranes (eg phospholipids, nucleic acids, coenzymes) and phosphate derivatives are involved in many important metabolic pathways (Tate 1985).

Soil P

Grasslands have a very high turnover of plant shoot and root material (Bardgett and Cook 1998) resulting in high rates of nutrient cycling and a build up of labile organic matter at the soil surface over time (Russell 1960a, 1986; Frossard et al. 2000). As a result the cation exchange capacity and total P concentration of fertilised, legume-based pastures soils also increases, with inorganic P levels generally increasing at a greater rate than organic P levels (Haynes and Williams 1993). P occurs in a range of inorganic and organic forms in soil, many of which have not yet been characterised. Inorganic P comprises up to 45% of total soil P in fertilised, well- established pastures (Walker et al. 1959) and occurs in pools that are rapidly desorbable, only slowly exchangeable or almost entirely unavailable for plant uptake (Bieleski 1973). Inorganic P in soils primarily occurs adsorbed to the surfaces, or occluded in the crystal lattice, of oxides and hydrous oxides of Fe, Al and Ca, and also occurs as discrete Fe, Al and Ca precipitates, particularly in the highly acidic vicinity of fertiliser granules (Huffman and Taylor 1963; Ryden et al. 1973; White 1980; Frossard et al. 2000). The organic soil P pool is comprised of mono- and di- phosphate esters (eg nucleic acids and phospholipids), condensed phosphates (eg. ATP), and phosphonates (Ryden et al. 1973; Turner and Haygarth 2000b). Most organic P in soils is associated with humic and fulvic acid complexes in soil organic matter with inositol phosphate (a monoester-P compound) being the most abundant (60%) organic P species (Tate 1985). The partitioning of inorganic P between soil and solution phases through either sorption/desorption or dissolution/precipitation mechanisms is dependant on the concentrations of P in solution and the nature of the inorganic particulates (Ryden and Syers 1977; White 1980). Desorption and dissolution reactions have been studied extensively with regards to their ability to supply phosphate to plant roots (Beckett and White 1964; Cooke 1966; Barrow 1979a; White 1980; Holford 1989). Some organic P species such as inositol phosphate are retained by soils through precipitation reactions, whereas others are involved in sorption reactions similar to orthophosphate (Ryden et al. 1973). Up to 99% of inorganic P cannot be utilized by plants (Bieleski 1973), and whilst a large amount of soil organic P is insoluble (Bieleski 1973),

7 biological processes such as enzyme hydrolysis of organic P also influence P mobilisation on a seasonal basis. Organic forms of P are hydrolysed to phosphate by extracellular enzymes or other chemicals (Correll 1998) such that net mineralisation in spring makes a labile organic P pool available for plant uptake during the phase of rapid pasture growth (Tate 1985). Concentrations of both dissolved inorganic and organic P forms in the soil solution increase with increasing pH and P fertility status (Ron Vaz et al. 1993). The concentration of P 3- (as PO4 ) in grassland soil solution (0-5 cm) increased from 5 to 21 µM after 4.5 years of P fertiliser applications at 40 kg P/ha/yr (Wheeler and Edmeades 1995). Dissolved organic P is present in soil water at concentrations which remain fairly constant with depth but its relative contribution to the total P increases with depth because dissolved inorganic P becomes preferentially adsorbed (Chardon et al. 1997; Haygarth et al. 1998). Most Australian soils are deficient in the amount of P available for plant growth as a result of strong leaching and weathering during an extended geological history and because of the coarse textured nature of many parent materials (Costin and Williams 1983). Even some of the younger soils are derived from parent material that has been depleted of P. Exceptions are a few soils that are derived from deposits of basalt (such as Vertosols) or alluvium (such as

Dermosols), which may have native Olsen P (POlsen) levels adequate for crop and pasture growth (Moody and Bolland 1999). To maintain pasture growth, external inputs of P to the soil P pool are applied as organic and inorganic fertilisers. Superphosphate (9% P, 11% S) is the main P fertiliser applied to Australian pastures (Tate 1985; Gourley 2001), with average annual application rates varying from 7.3 kg P/ha for sheep and cattle pastures (Cayley and Kearney 1999) to 14 kg P/ha for dairy pastures (Gourley 2001). In contrast, there is a small external input of P to pasture soils from rainfall (Tabatabai and Laflen 1976). The mean deposition of P in rainfall at 30 sites across Victoria over 3 years was 0.4 kg/ ha/yr, with many sites recording inputs of less than 0.1 kg /ha/yr (Greenhill et al. 1983b).

Microbial P

Micro-organisms and invertebrates play an important role in the transformations of P between pools (Cosgrove 1976; Tate 1985; Perrott et al. 1990; Bolan 1991). Invertebrates such as earthworms help incorporate plant and organic matter residues into the soil (Haynes and Williams 1992), as well as with processing detritus. Bacteria immobilise inorganic P, and microbial-P is subsequently consumed and mineralised by amoebae and nematodes (Tate 1985). Plant roots and micro-organisms also exude phosphatase enzymes which enable organic P compounds to become available for plant and microbial uptake (Spier and Cowling 1991; Turner and Haygarth 2000b). Up to 29 kg P/ha was released from organic and microbial P in soil under grazed pasture in New Zealand and this P was considered to be a significant source for plant uptake during spring (Perrott et al. 1990).

8 Recycling of P by grazing animals

Grazing animals accelerate the cycling of nutrients through the soil-plant-animal system by consuming up to half the above ground and a quarter of the below ground annual net primary production (Bardgett and Cook 1998) and depositing between 60 and 99% of ingested nutrients in dung and urine (Barrow and Lambourne 1962; Haynes and Williams 1993). These nutrients are usually in a form more available for plant uptake than the ingested material (O'Hara 1996). Dung is also enriched with nutrients compared with pasture plants and soil due to preferential absorption of proteins and carbohydrates during digestion and additions of metabolic wastes such as bile (Dickinson and Craig 1990), whereas negligible concentrations of P are excreted in urine (Barrow and Lambourne 1962). Grains, hay and silage fed as supplements to pasture can represent a significant external P input to the P cycle in intensive dairy systems (Gourley 2001). The total P concentration of sheep dung ranges from 4.8 mg P/g on unfertilised pasture to 17 mg P/g dry matter on fertile pastures (Bromfield 1961; Williams and Haynes 1992), which far exceeds P concentrations in topsoils of grazed pastures (<0.6 mg/g, (McCaskill et al. 2003)). The total and inorganic P components of dung increase as the fertility of the ingested pasture increases (Rowarth et al. 1988; Nguyen and Goh 1992; O'Hara 1996), and whilst the water soluble organic component may remain low, between 0.5 and 1 mg P/g dry faeces, the total organic component may either remain stable, between 1 and 3 mg P/g faeces (Bromfield 1961; Barrow and Lambourne 1962), or increase up to 6.4 mg P/ g faeces (O'Hara 1996). Negligible mineralisation of the organic P in sheep dung is expected until the dung is incorporated into the soil (Rixon and Zorin 1978). Soluble P release from sheep manure increases as its moisture content increases (Rixon and Zorin 1978) but inorganic P in sheep faeces is predominantly dicalcium phosphate and is only moderately soluble in water (Bromfield 1961; Barrow 1975).

Herbage P

Pasture swards can contain up to 6 mg P/g of plant material (Gillingham et al. 1980; Williams and Haynes 1992; O'Hara 1996), which is also a much higher P concentration than in most soils. Between 41 and 84% of this P is inorganic and soluble in water (Bromfield and Jones 1972; Gillingham et al. 1980). The dissolved inorganic P represents the majority of the P in plants that is readily available for uptake by other plants once returned to the soil. Similar to dung, the total P concentration of pasture herbage, and the proportion that is soluble in water, tends to increase with increasing levels of plant available soil P (Bromfield and Jones 1972; Gillingham 1980; Rowarth et al. 1988; Williams and Haynes 1992).

9 2.2.2 Efficiency of P cycling in grazed pastures – ‘loss factors’

The majority of fertiliser P applied to pastures is retained within the topsoil (0-20 cm) through fixation to soil constituents (Williams and Haynes 1992). However, livestock also represent a small nutrient loss pathway with the amount of P removed from sheep pasture systems in body tissue and wool being approximately 1.5 kg P/ha/yr (Williams and Haynes 1992). P can also be accounted for in ungrazed herbage and plant roots (< 1 kg P/ha/yr)(Williams and Haynes 1992). Due to the high proportion of ingested nutrients that are returned to the pasture in the dung and urine of grazing animals, and the increased plant-availability of these nutrients, the spatial distribution of excreta can have a significant impact on the efficiency of nutrient cycling within a grazed area (Haynes and Williams 1993). Across a paddock, stock tend to congregate and rest in certain areas such as under trees, on ridges, hillcrests, irrigation levees and near water and feed troughs, and stock traffic also increases near gates, along laneways and in stock yards (Williams and Haynes 1992). The tendency for stock camping and dung accumulation in discrete flatter areas of a pasture tends to increase with increasing slope gradient (Gillingham 1980) such that on hill country sheep pastures in New Zealand, 60% of the dung and 55% of the urine was deposited over only 31% of the pasture area (Saggar et al. 1990). Elevated soil P levels in areas where stock congregate (stock camps) also represents an inefficiency of P cycling in pastures because much of this P is removed from the main grazing area (Gillingham and During 1973; Gillingham 1987; Williams and Haynes 1992; McCaskill and Cayley 2000). In New Zealand hill country, there was a net gain of 60-120 kg P/ha in camp areas at the top of slopes compared to a net loss of 10-20 kg P/ha on the steep grazed slopes (Gillingham et al. 1980). This occurred mainly because of higher dung returns to campsites than to the steep pasture slopes. On flatter dryland pasture in south-eastern Australia, McCaskill and Cayley (2000) found that 6.5% of P applied to perennial pastures over 17 years was transferred to sheep stock camps. Soil microbial activity is also higher in camp areas than non-camp soils (Haynes and Williams 1999). Whilst increased soil P levels can lead to higher pasture growth rates in the stock camp areas (Saunders 1984; Haynes and Williams 1993), treading damage can reduce effective pasture growth (Finlayson et al. 2002). Nutrient mass balances for sheep farming systems have been unable to account for 6- 35% of the fertiliser P applied to irrigated pastures (Nguyen and Goh 1992; Williams and Haynes 1992) and 6.6% of P applied to a flat dryland pasture (McCaskill and Cayley 2000). Unrecovered P is sometimes attributed to losses in runoff or leaching (Williams and Haynes 1992), however, McCaskill and Cayley (2000) suggested that much of the unaccounted P could be attributed to the large confidence intervals in measured components of the mass balance, such as total soil P. Losses of P from pasture systems via hydrological pathways have been

10 estimated as small compared with fertiliser inputs and losses in soil fixation (Saggar et al. 1990; Nguyen and Goh 1992; McCaskill and Cayley 2000) and have been largely ignored in P balance approaches to calculating fertiliser requirements (eg. Cornforth and Sinclair (1982)). For example loss of P in runoff from hill pastures in New Zealand were the equivalent of 2.9% of annual fertiliser P inputs and 3.1% of nutrients cycling annually though the plant pool (Lambert et al. 1985). Whilst measured P losses from dryland pastures are not usually financially significant (McColl et al. 1977; Sharpley and Syers 1979; Stevens et al. 1999), agricultural land has been identified as a significant source of nutrients reaching waterways, where even small concentrations of P can have an adverse impact on the ecology of freshwater systems (Harris 1994; Correll 1998; Davis et al. 1998; Anon 2000b). Losses of P via hydrological pathways therefore need more direct measurement techniques if they are to be accurately estimated. Consequently, the level of research and attention into P movement from grazed systems along hydrological pathways has increased over the last two decades (Ryden et al. 1973; Sharpley and Syers 1976; Lambert et al. 1985; Haygarth and Jarvis 1996; Tunney et al. 1997; Nash and Halliwell 1999).

2.3 Forms of P in natural waters

There is a continuum of forms of P in natural water samples ranging in size from orthophosphate ions to P complexed with large organic molecules and adsorbed to clay minerals. For the purposes of reproducibility, practicality and comparison, the forms of P in water samples are usually arbitrarily separated into analytically defined pools (Broberg and Persson 1988). Filtrate that passes through a membrane filter of 0.45µm nominal pore size is typically defined as dissolved P (DP) and the retained material as particulate P (PP)(Table 2-1). Chemical differentiation is also achieved between P fractions that are reactive with a molybdenum blue reagent (i.e. reactive P, RP)(Murphy and Riley 1962) and fractions that do not react (unreactive P, UP).

2.3.1 Inorganic and organic P fractions

Reactive P in water is largely composed of inorganic P, however some organic forms of P are readily hydrolysed in the acidic (pH<2) conditions of the molybdate reaction (Broberg and Persson 1988; Baldwin 1998). Organic P species in natural waters include DNA, RNA, c-AMP and inositol hexaphosphate (Baldwin 1998). Inorganic P occurs as orthophosphate and condensed phosphates as well as in association with minerals and colloidal precipitates of Al and Fe oxy-hydroxides and Ca carbonates (Broberg and Persson 1988). A measure of the fraction of P in water that is available for algal uptake can indicate the potential for P in runoff waters to accelerate algal growth in surface waters (Sonzogni et al. 1982; Sharpley 1993b). The majority of the P in the DRP fraction is considered immediately

11 bioavailable (Sonzogni et al. 1982; Sharpley 1993a, b), however, measures of DRP can give misleading estimates of the amount of bioavailable P as the turnover time of orthophosphate in natural waters can be just a few minutes (Haygarth et al. 1995). As well as this, a measure of the bioavailable P in water samples is likely to underestimate the total P available to algae because whilst ortho-phosphate is the only form of P that can be assimilated by algae (Bostrom et al. 1988; Correll 1998), algae can utilise P from suspended sediment, aquatic biota and sedimentary pools (Sharpley et al. 1991; Harris 1994). Of the commonly measured fractions of P, the TP content of a water sample may therefore be considered just as, if not more, useful than the DRP fraction for assessing the ecological P status of water bodies (Jansson et al. 1988; Harris 1994; Correll 1998). This is particularly the case in environments where there is a lag time between runoff and the low-flow, warm conditions that are conducive to the growth of algal blooms (Correll 1998). Reactive P is generally a larger fraction of the TP in runoff from pastures than UP regardless of whether inorganic or organic P amendments have been applied (Nash and Murdoch 1997; Smith et al. 2001a). However, in leachate and subsoil soil solution, DUP has been found to dominate due to its lower sorption affinity to soil particles than reactive P forms (Ron Vaz et al. 1993; Chardon et al. 1997). This suggests DUP may be an important form of P in drainage waters from agricultural lands. The enhanced mobility of DUP through soil compared to RP suggests also that in pasture soils where there is a greater pool of organic matter and potentially labile organic P than arable soils (Frossard et al. 2000), there will be more DUP available for mobilisation in runoff.

12 Table 2-1: Definitions and descriptions of P and sediments fractions of runoff water

Measured P fractions Abbreviation Analytical fraction Description TP Total P The total P in an unfiltered water sample measured using Mo colorimetric analysis after autoclave digestion with alkaline persulfate. Some occluded P may not be recovered using this digestion (O'Connor and Syers 1975). DP Dissolved P The total P in the filtrate of a water sample that has passed through a 0.45µm filter. This pool includes the soluble orthophosphate ion and organic P plus Pcomplexed with colloids and particles less than 0.45µm in size RP Reactive P P fractions which react with a Mo blue complex after reduction with ascorbic acid TRP Total reactive P The P in an unfiltered water sample that reacts with a Mo blue complex after reduction with ascorbic acid. DRP Dissolved reactive P The reactive P in a filtered water sample (<0.45µm). DRP is comprised largely of orthophosphate but may also include other P fractions such as readily hydrolysed polyphosphates (Broberg and Pettersson 1988; Gerke 1992; Ron Vaz et al. 1993; Baldwin 1998; Haygarth and Sharpley 2000) Calculated P fractions PP Particulate P PP represents the total P that is associated with particles >0.45 µm in size in a water sample. The particles include PP = TP-DP cells of plants, bacteria and animals, minerals and inorganic and organic-bound precipitates (Broberg and Persson 1988) PRP Particulate reactive P PRP is the reactive fraction of PP and represents the inorganic P associated with particles (>0.45µm) PRP=TRP-DRP (Table continued on next page)

13 Table 2-1 cont. Abbreviation Analytical fraction Description UP Unreactive P P fractions which do not react with a Mo blue complex after reduction with ascorbic acid TUP Total unreactive P TUP represents the unreactive P in an unfiltered water sample. This probably includes P that is tightly held within TUP = TP-TRP soil particles, organically bound P attached to soil particles, colloids and P in solution and microbial cell biomass (Hannapel et al. 1964). This pool includes inositol phosphates, nucleic acids, nucleotides, phospholipids, sugar phosphates and perhaps also condensed inorganic forms of P (Ron Vaz et al. 1993; Turner and Haygarth 2000b). DUP Dissolved unreactive P DUP represents the unreactive P in a filtered (<0.45µm) water sample, and is largely P associated with dissolved DUP=DP-DRP organic molecules but also includes some dissolved inorganic P (Broberg and Persson 1988) PUP Particulate unreactive P PUP is the unreactive P associated with particles >0.45µm in a water sample, and is assumed to be largely PUP= PP-PRP associated with soil, organic complexes and microbial biomass Sediment fractions SS Suspended sediment The particulate material retained on a 0.45µm filter after filtration of a water sample Sediment P Sediment P content = The total P content of the suspended sediment in a water sample content PP/SS TDS Total dissolved solids The total solid residue after evaporation of a filtered water sample

14 The degree to which the analytical fractions of P represent the chemical (inorganic, and organic) P pools is likely to vary depending on the source, pH and electrical conductivity of the water, and this can have implications for the mobilisation of P in the environment. For example, the inference that DRP is readily fixed to soil sorption sites may not be accurate if a substantial proportion of the DRP pool is made up of polyphosphates that are poorly retained by soil (Busman and Tabatabai 1985). Halliwell et al. (2000) compared direct measurements of orthophosphate with concentrations of DRP in runoff from irrigated dairy pastures and found that over 97% was orthophosphate, whereas Baldwin (1998) found that less than 20% of wetland water DRP was orthophosphate. The DUP fraction in waters draining a grazed catchment in South Australia was identified as being mainly dissolved organic P due to the high correlation with the size distribution of the dissolved organic C content (Nelson et al. 1996). In a nearby pasture catchment however, a high proportion of the DUP was larger (>100 kDa) and therefore was more likely composed of inorganic colloids.

2.3.2 Particulate and dissolved P fractions

Surface erosion of soil containing sorbed P (particulate P) is usually the dominant source and pathway of P loss in temperate environments if there is less than 75% soil cover when runoff occurs (Burwell et al. 1975; Lang 1979). Even where topsoil P concentrations are low, total P losses can be large where there is a high degree of erosion (Olness et al. 1975). Most of the erosion from Australian agricultural soils occurs during large infrequent rain events (Cullen 1990; Davis et al. 1998). Other factors affecting the amount of particulate material mobilised in surface runoff from pastures are the degree of ground cover, the slope angle, degree of soil disturbance and soil erodibility (Sharpley and Menzel 1987; Hairsine and Prosser 1997; Davis et al. 1998). Where pasture cover is greater than 75%, there is less of a link between groundcover, runoff and sediment losses (Costin 1980; Nash and Murdoch 1997) and the DP fraction in runoff becomes more significant. In temperate catchments a combination of good groundcover, lower intensity storms, efforts to stabilise streambank and gullies and larger areas of intensive agriculture suggests that DP exports may be of relatively greater importance than PP exports. In particular, runoff from forest and pasture on hillslopes may also contribute significant amounts of DP to waterways, primarily due to lower rates of soil erosion and higher desorbable P contents of soils (White and Sharpley 1996). For example, Nash and Murdoch (1997) found that greater than 90% of P in surface runoff from a dairy pasture in south eastern Australia was in the dissolved form. Significant amounts of DP are also lost in drainage through sandy agricultural soils, which have a low capacity to retain P against leaching (Ozanne et al. 1961; Holt et al. 1970; Lewis et al. 1981; Ritchie and Weaver 1993; Davis et al. 1998). Nelson et al. (1996) found 76% of the P in streamflow draining a grazed catchment in southern Australia was

15 DP. They suggested that as well as the mechanisms described above, dissolution of P from cow manure and the release of inorganic P from anoxic, waterlogged areas may also have lead to mobilisation of DP.

2.4 Pathways of water movement in pasture systems

Regardless of the chemical, physical or biological properties of the P in the environment, landscape hydrology is the dominant factor controlling its mobilisation (Burwell et al. 1975; Haygarth and Jarvis 1997; Kirkby et al. 1997; Stevens et al. 1999; Fleming and Cox 2001). Water that is not evaporated, transpired or stored in the soil-plant profile is redistributed as either surface runoff or subsurface vertical or lateral movement (Chorley 1978). These hydrologic pathways determine the total export of P from land by influencing the sources of P with which the water comes into contact, as well as the duration of contact. The following section therefore discusses the range of pathways and processes of water movement in pasture systems, and the spatial and temporal controls on hydrologic activity.

2.4.1 Surface Runoff

Surface runoff is defined here as water that flows over the soil surface. There are three main mechanisms of surface runoff generation; Hortonian flow (infiltration excess flow), saturation excess and return flow. Runoff that is produced when the lower limiting infiltration capacity of the soil is exceeded by the rainfall intensity is referred as infiltration excess runoff or Hortonian flow, after Robert Horton who developed the theory in the 1930s (Chorley 1978). Infiltration excess flow describes a mix of laminar and turbulent sheet flow across a slope of fairly spatially uniform initial infiltration capacity. In temperate regions, infiltration excess runoff rarely occurs, exceptions occurring during infrequent intense storms and on soils that have been compacted due to animal and vehicle traffic (Lambert et al. 1985; McColl et al. 1985; Haygarth et al. 2000), or in arid regions where bare soils form impermeable crusts during heavy precipitation (Ward 1984). However, some infiltration excess flow may occur over parts of a grassed hillslope during storms of even moderate intensity (Pilgrim et al. 1978). Horton’s model of surface runoff generation did not account for the process of subsurface flows re-emerging at the hillslope surface, known now as return flow. Return flow describes water that vertically infiltrates high in the landscape and is forced to return to the surface by artesian pressure where it meets local areas of saturation (Chorley 1978). A more generally applicable model of runoff generation was developed by Betson and Marius (1969), who acknowledged that the cumulative effects of upslope contributions of surface, subsurface and return flow leads to downslope zones of soil-water saturation within a catchment. Rain falling on these areas, even at intensities less than the local infiltration rate,

16 generates surface runoff in the form of saturation excess flow. A further development of runoff theory lead to the ‘variable source area’ concept which describes the expansion and retraction of the saturated zone both seasonally and during a storm event (Chorley 1978; Dunne 1978; O'Loughlin 1981; Ward 1984). The development, location and extent of variable source areas (VSA) are discussed shortly.

2.4.2 Subsurface flow

Water that does not leave a hillslope directly as surface runoff or evaporation percolates vertically as drainage, and downslope as subsurface lateral flow (Atkinson 1978). Water percolates either as matrix flow through the inter-granular pores and small structural voids or through larger voids referred to as macropores (Atkinson 1978). Macropores can be formed by worm and other faunal activity (including animal burrows), senesced tree and plant root cavities (Hatton et al. 2002), and in cracks between clay peds (Smettem et al. 1991). Artificial mole and tile drainage and natural springs are extreme examples of preferential flow pathways. Natural macropore networks are tortuous but in some cases can be continuous enough to make a significant contribution to the distribution of water within the soil profile (Jarvis and Leeds- Harrison 1987) such as providing pathways for storm rains to bypass topsoil horizons and otherwise impermeable subsoil boundary layers (Smettem et al. 1991). Preferential flow of water through macropores therefore increases the effective hydraulic conductivity of the soil matrix (Pilgrim et al. 1978; Clothier and Smettem 1990; Smettem et al. 1991) and in doing so can cause subsoil saturation and perched watertables even before the surrounding soil profile is completely wet (Chorley 1978). Whilst subsurface flow can contribute to the generation of surface runoff during storms by emerging as return flow (Dunne 1978), after the cessation of rains it is the main source of water for stream baseflow and for maintaining downslope wet areas where subsequent runoff is generated (Dunne and Black 1970b). The latter roles are attributable to subsurface lateral flows through both unsaturated and saturated soils (Chorley 1978; Anderson and Kneale 1982).

2.4.3 Factors affecting hydrological pathways and runoff mechanisms

Hydrological conditions are both spatially and temporally dynamic and often vary with scale (Haygarth et al. 2000; White et al. 2000; Cox and Pitman 2001). Where lower rainfall intensities, high surface soil infiltration rates and good vegetative cover dominate, infiltration excess flow is less likely (Smettem et al. 1991). The majority of storm flow is then attributed to return and saturation excess flow (Cooke and Dons 1988; Nash and Murdoch 1997) as well as subsurface flow (Sklash et al. 1986; Smettem et al. 1991; Cox and Ashley 2000) according to

17 the variable source area model. In any catchment, however, it is likely that more than one surface runoff mechanism will occur (Pilgrim et al. 1978).

Temporal factors

The partitioning of water between surface and subsurface pathways is highly variable from year to year as well as between and within storms because it is influenced by the amount and distribution of rainfall, rainfall intensity, soil moisture content and ground surface characteristics (White and Kookana 1998; Cox and Ashley 2000; Cox and Pitman 2001). The extreme inter-annual variability of the Australian climate is exemplified by the research of White et al. (2000) who studied flow pathways from 0.13 ha pastured plots over 4 years. Two of these years had no surplus water flow. In one year the majority of flow was partitioned as surface runoff, while in another year subsurface lateral flow dominated. Similarly, Costin (1980) found that in low rainfall years pasture catchments in New South Wales acted almost as closed hydrologic systems but in wet years, runoff occurred as saturation excess runoff in near- saturated soils. In Mediterranean climates, low antecedent water contents and high topsoil infiltration capacities often reduce the potential for surface runoff during summer (Costin 1980; McColl et al. 1985) except in some soils where the hydrophobic nature of surface organic matter reduces the infiltration rate of rainfall (Smettem et al. 1991). Across a range of environments, the majority of nutrient export from catchments occurs during storm events (Burwell et al. 1975; McColl et al. 1977; Cooke 1988; Nelson et al. 1996), particularly in environments like Australia where rainfall is highly episodic (Nash and Murdoch 1997). For example, rain falling onto wet soil in spring produced 28 mm of runoff from a texture contrast soil in New South Wales, which amounted to 20% of the total of five years of runoff (Costin 1980). An understanding of the dominant flow pathways and runoff mechanisms occurring during high flows will therefore help explain the main processes of P movement (Harris 1994). A combination of climatic variability and scale differences may also explain the differences in dominant flow paths measured in separate studies and years on a texture- contrast soil in South Australia’s Mt. Lofty Ranges. Smettem et al. (1991) and Kirkby et al. (1996), using a series of instrumented 20 x 3 m plots, measured the majority of flow as macropore – mediated subsurface lateral flow in 1987/88 and 1991 respectively whereas Stevens et al. (1999) measured predominantly surface runoff from two larger sub-catchments (4.2 and 3.6 ha) in 1997. As Stevens et al. (1999) point out, upscaling the results to conclude that the dominant catchment flow mechanism is surface runoff may underestimate the subsurface lateral flow contributions from upslope, which may contribute directly to surface waters in other parts of the landscape. The reverse may also be said for upscaling from the plot studies.

18 Spatial partitioning

Pathways of water movement within a landscape are influenced by slope, soil type and vegetation as well as anthropogenic factors such as artificial drainage, cultivation and grazing management (Pilgrim et al. 1978; Ward and Robinson 2000). The frequency and volume of surface runoff generally increases with slope gradient (eg Fleming and Cox 1998) although for pastures in southern Australia with less than 15% slope, factors such as rainfall and intensity had a larger influence on runoff volume than slope gradient (Greenhill et al. 1983a; Tham 1983). However, large flat catchments tend to generate runoff of higher ‘intensity’ (peak flow rate/volume) than steeper catchments (Fleming and Cox 1998). On steep convex slopes subsurface lateral flow may dominate over surface runoff (Pilgrim et al. 1978; Cox and Reynolds 1995), sometimes developing into springs which contribute surface runoff as return flow (Cooke and Dons 1988). Freeze (1972) suggested that subsurface flow is only a significant source of storm flow when hillslopes feed deeply incised valleys and the saturated hydraulic conductivity of the permeable horizon is large. Dunne (1978) concurred, observing that during storms of long duration, subsurface flow dominated the recession limb of the hydrograph and that for the deepest, well-drained soils, subsurface flow was the only form of stormflow. This is evident for deep permeable soils that dominate forested hillslopes in south-central Victoria and New Zealand (Mosley 1979; Finlayson and Wong 1982). Subsurface lateral flow may also dominate where highly permeable A horizons overlay heavy clay B horizons (Smettem et al. 1991; Stevens et al. 1999). Soils which exhibit a strong contrast in texture between the A and B soil horizons are known as duplex soils (Chittleborough 1992). When the hydraulic conductivity of the subsoil is low in absolute terms, as well as relative to the A horizon, the rate of vertical flow can become restricted in these soils causing perched watertables to develop above the limiting horizon. (Hammermeister et al. 1982; Graecen and Williams 1983). These conditions are typical of soils with sodic subsoils that are easily dispersed (Stevens et al. 1999). In flat areas this can lead to waterlogging and surface ponding, while on slopes the surplus water becomes surface and subsurface lateral flow (Costin 1980; Cox and McFarlane 1995). Duplex soils dominate the high rainfall agricultural zone of southern Australia (Chittleborough 1992). There have been a number of soil core (Cox et al. 1996a; Evans et al. 1996; Kirkby et al. 1997), plot (Smettem et al. 1991; Kirkby et al. 1996; White et al. 2000) and subcatchment (Fleming and Cox 1998; Cox and Ashley 2000) studies investigating the pathways of water and nutrient movement over and through these soils. As well as the temporal factors of higher annual (Stevens et al. 1999; White et al. 2000) or seasonal (Cooke and Dons 1988) rainfall, soil conditions that were found to favour a dominance of surface soil saturation

19 and surface runoff over subsurface lateral flow in duplex soils were; shallow and moderately permeable A horizons (Stevens et al. 1999), moderately permeable subsoils (Fleming and Cox 1998) and an accumulation of hydrophobic organic materials on the soil surface during summer (Smettem et al. 1991). In some duplex soils it is the C horizon (often comprised of saprolite material) which restricts vertical drainage as well as or instead of the B horizon (Hammermeister et al. 1982; Brouwer and Fitzpatrick 2002b). At the catchment scale, vegetation type has a major impact on the volume of runoff. Streamflow generally reduces in the order native grassland, sown pasture, native forest then exotic forest in New Zealand (McColl et al. 1977) and Australian catchments (Costin et al. 1984). The percentage of simulated rainfall as runoff on small pasture plots was generally greater for native vegetation compared with a phalaris/clover pasture (Costin 1980). For improved catchments, Lambert et al. (1985) found that pastures with higher soil fertility yielded less runoff than pastures with lower fertility. This was attributed to better infiltration due to enhanced earthworm activity and increased water use due to increased pasture growth. In contrast, Greenhill et al. (1983a) found no trend in the percentage of rainfall as runoff across four P application rates for three pasture locations in Eastern Victoria. Heavy grazing can cause soil compaction and reduce groundcover which reduces the infiltration capacity and removes barriers to the flow of runoff, leading to greater runoff volumes, discharge rates and frequencies (Alderfer and Robinson 1947; Sartz and Tolsted 1974; Chichester et al. 1979; Lambert et al. 1985). For example, runoff from compacted soils used for winter cattle feeding and summer grazing was 10 to 15% of rainfall compared with 2% from soils grazed in summer only (Chichester et al. 1979). In contrast, Greenhill et al. (1983a) found pugging caused by cattle treading increased the surface detention of water, which lowered runoff volumes. As well as this, where stocking rates are only moderate, compaction may be off-set by enhanced soil invertebrate activity which can flourish in fertile soils and lower the bulk density (Costin 1980). Human intervention such as vegetation clearance, artificial drainage, irrigation and surface dams also plays a role in modifying landscape hydrology (Costin et al. 1984; Gregory et al. 1992; Kaine and Niall 2001; Hatton et al. 2002). Whilst artificial surface and subsurface drainage decreases waterlogging (Cox and Reynolds 1995), the nutrient quality of drainage waters can be of concern (Cox and Pitman 2001).

2.4.4 The location and extent of surface saturation zones

Regardless of the partitioning of water between subsurface and surface pathways at the hillslope, zones of saturation within a catchment often account for the total storm flow (Betson and Marius 1969) or stream baseflow (McColl et al. 1985; Cooke and Dons 1988). The area of land from which surface runoff water is generated can vary from the entire catchment for infiltration excess flow to as little as 1 or 2% where subsurface flow dominates (Finlayson and

20 Wong 1982). The location and extent of the VSA is largely dependant on the antecedent soil water content and its distribution (Dunne 1978; Finlayson and Wong 1982; Barling et al. 1994). As the antecedent moisture content rises, the ability for any part of the soil profile to accept more water is reduced and runoff is enhanced (Tham 1983; Nelson et al. 1996; Gburek and Sharpley 1998). The soil water content is dependant on a range of topographic, soil, vegetation and climatic variables because of their influence on the soil water holding capacity, vertical and lateral drainage, direction of surface and subsurface flows and evapotranspiration. Factors that are important include the soil saturated hydraulic conductivity, depth and macroporosity, topographic features including catchment area, slope steepness (or hydraulic gradient), hillslope concavity, curvature and aspect, variation in surface vegetative cover and water use, and climatic variables such as rainfall, net radiation wind and temperature (Moore et al. 1991). Areas most prone to waterlogging and runoff generation are those close to streams, local depression lines, where slope decreases abruptly, at the base of hillslopes, in areas with thin topsoils and impermeable subsoils and at catchment outlets (Chorley 1978; Cooke and Dons 1988; Cox and McFarlane 1995; Gburek and Sharpley 1998). Areas of topographic convergence, such as hillslopes that are concave both parallel and perpendicular to the slope, also lend themselves to waterlogging due to the greater frequency of perched and groundwater tables intersecting the soil surface (Ward 1984; Smettem et al. 1991). Similarly, surface saturation develops where the water holding capacity and transmission capacity of the soil are exceeded by converging subsurface and surface flows (Ward and Robinson 2000). The extent of waterlogging usually increases as the degree of slope decreases (O'Loughlin 1981) and in wetter than average years (Cox and Reynolds 1995). Variable source areas can be remote from stream channels, however, so ultimately the level of hydrological connectivity between source areas determines their contribution to catchment runoff (Ward and Robinson 2000). Australia has few permanently saturated areas due to the ephemeral nature of many streams but where these do occur, the soils are often highly saline, sodic and bleached (Cox et al. 1996b). Waterlogging extends seasonally into soils which often display red stains and grey mottling caused by the redistribution of iron (Cox et al. 1996b), gleying (Cooke and Dons 1988), pugging due to animal treading (McColl et al. 1985) and which have thin topsoils with high gravel contents (and consequently low porosity). The infiltration rate in saturated zones is limited by the hydraulic conductivity of the least permeable saturated soil (Chorley 1978). Subsoil characteristics which are typical of saturated soils therefore include low hydraulic conductivity (less than about 0.03 m/day), poor structure (Cox and McFarlane 1995) and sodicity (Stevens et al. 1999). Hatton et al. (2002) found that soil hydraulics, site topography and presence of structures such as surface dams had a larger influence on the incidence of seasonal waterlogging of duplex soils in southwestern Australia than vegetation cover.

21 The size of the source area within a certain topography and soil type varies with the storm intensity and duration, initial soil moisture and soil-water storage capacity (Betson and Marius 1969; Dunne 1978). The area which becomes saturated during large storms in temperate climates can extend across large tracts of the catchment but this typically lasts for less than 48 hours (McColl et al. 1985). Expansion of the saturated area is greater where seasonal groundwater tables are near or discharging to the surface. During dry years this may be confined to the stream itself. A range of methods has been used to identify the location and extent of saturated areas of the landscape. Direct measurements of surface saturation have been made using light- emitting saturation sensors (Gburek and Sharpley 1998), logged tensiometers (Anderson and Burt 1978), logged electrical resistance sensors (Murphy and Lodge 2004), shallow wells (Betson and Marius 1969; Cox and Reynolds 1995), and visual or infrared mapping during storms (Dunne 1978; McColl et al. 1985; Cooke and Dons 1988). Srinivasan et al. (2002) used a combination of surface saturation sensors and mini- surface runoff weirs (Srinivasan et al. 2000) to identify where surface saturation coincided with runoff generation. A number of wetness indices and physical models based largely on topographic attributes have also been developed (O'Loughlin 1986; Beven 1989; Grayson et al. 1992; Barling et al. 1994) although rarely are these able to explain more than 50% of the distribution of soil moisture (Grayson and Western 2001). Soil geomorphological features (Cox et al. 1996b; Dahlhaus et al. 2000; Brouwer and Fitzpatrick 2002a) and types of vegetation (McColl et al. 1985) that indicate the frequency of waterlogging are also used to predict the location of seasonally saturated soils, but temporary saturation during storms alone is less likely to be evident form soil and vegetation characteristics (Cox and McFarlane 1995). There is a lack of information on the degree to which saturation excess runoff from VSAs occurs in south-west Victoria. The size and degree of expansion of saturated areas affects not only the volume of storm runoff, but also the quality of the water due to the variable nature of the parts of the landscape interacting with the surface flows (Dunne 1978). Management of waterlogged areas for water quality goals is therefore potentially more feasible than attempting to manage areas subject to less temporally but more spatially widespread Hortonian flow events (McColl et al. 1985; Cooke and Dons 1988).

2.5 P mobilisation processes

P is mobilised from pastures in surface runoff by a combination of chemical, physical and biological processes (Ryden et al. 1973; Haygarth and Jarvis 1999). In environments where soil loss via erosion is a key factor during runoff events, most of the nutrients lost will be attached to sediment (Burwell et al. 1975; Costin 1980; Hodgkinson 1996; White and Sharpley 1996). Even where topsoil P concentrations are low, total P losses can be large where there is a

22 high degree of erosion (Olness et al. 1975). A close link between concentrations of suspended solids and total P concentrations in runoff or stream water can indicate where P movement is dominated by erosion mechanisms (Tham 1983; Turoczy 1999). During water erosion, particles of soil, organic matter, faecal material or fertiliser undergo detachment, entrainment, transport and redeposition (Hudson 1995). P adsorbed to soil particles is also preferentially mobilised during erosion because smaller P-rich clay particles (<2 µm) are more readily entrained in runoff than larger particles that have fewer P sorbing constituents and surfaces (Ryden et al. 1973; Sharpley 1980). As a result, sediment in surface runoff is usually enriched with C, N, K and P compared with the soil from which it is derived (Sharpley 1980). The ratio of the nutrient concentration in sediment compared with the source soil is known as the enrichment ratio (ER) (Sharpley 1980). The proportion of total P mobilised as PP may be further enhanced by adsorption of DP to soil particles entrained in runoff (Sharpley et al. 1981c; Tham 1983). The mechanisms of P release from soil and pasture constituents into the soil solution and runoff water include desorption and slow diffusion from soil particles, dissolution of inorganic fertilisers, leaching from plant and organic matter, microbial mineralisation, enzymatic hydrolysis of humus, manure and plant residues, and lysis of microbial biomass through drying and re-wetting (Yli-Halla et al. 1995; Haygarth 1999). The amount of P desorbed from soil constituents by runoff or drainage will depend on a number of factors including the P buffering capacity, degree of soil-water mixing, temperature, solution ionic strength, pH and equilibration time and the equilibrium P concentration of the soil (White 1980). As the duration and area of contact between water and P sources increase, so does the amount of dissolution and desorption of P (Barrow 1979a; Sharpley et al. 1981a). High rainfall intensities, ratios of water to suspended sediment, and depth of soil wetting also increases the degree of interaction, which increases erosion and extraction or desorption of P from the pasture surface (Ahuja et al. 1981; Ahuja et al. 1982; Huang et al. 1999). Similarly, there is a decrease in dissolved P in soil-water as the duration and degree of mixing with sorbing soil constituents increases (White 1966; Barrow and Shaw 1975, 1979a; Sharpley et al. 1981c). For previously fertilised soils there is little hysteresis between adsorption and desorption of P because there is little reversion of freshly adsorbed P to more tightly bonded forms, however for P deficient soils, adsorption is likely to occur more readily than desorption from a P source (Barrow 1979a; White 1980). Empirical models describing the processes of PP and DP mobilisation are described in Chapter 6. Processes that act as P sinks along surface runoff pathways include chemical retention of P by adsorption and precipitation reactions in soils, physical trapping and retention of particles by soil and vegetation (Hairsine 1996) and the infiltration of runoff. P attenuation

23 along subsurface pathways occurs through physical entrapment along tortuous matrix pore networks (although some soils may release particles to drainage water through erosion of the walls of preferential pathways (Turner and Haygarth 2000a)), adsorption onto the soil matrix, and biological uptake of labile P by roots and micro-organisms. The sources of P and the processes of mobilisation and attenuation influence the forms of P that are in runoff from pastures and also influence the potential bioavailability of P in receiving waterways (Haygarth and Jarvis 1999). Net export or attenuation of P depends on the spatial and temporal degree of contact as well as the intensity of the P source.

2.6 Factors affecting P mobilisation from pastures

The extent of P loss from pastures is determined by the pathways of water movement in the landscape, the amount, form and mobility of P at the source, soil and geological properties, environmental conditions and pasture management practices (Holtan et al. 1988; Sharpley et al. 1994; White and Kookana 1998).

2.6.1 Land management and environmental conditions

In general, much of the P movement from agricultural land in Australia is associated with erosion of subsurface or surface soil during infrequent but large, and often high intensity, rainfall events (Hartley et al. 1984; Holford 1989; Cullen 1990; Hairsine and Prosser 1997; Davis et al. 1998). This erosion is exacerbated in the tropical environment of Northern Australia when a combination of cyclonic storms and minimal groundcover occurs following periods of drought (Davis et al. 1998). Within any dryland pasture catchment it is important to distinguish between gully and hillslope sources of P (Wallbrink et al. 1996; Wasson et al. 1996). Where gully erosion occurs, nutrient movement from eroded subsoil is likely be a more significant sediment source and transport mechanism for P losses than runoff from pastures even where the subsoil P levels are naturally low (Nelson et al. 1996; Davis et al. 1998). This was demonstrated in a nested catchment study in New Zealand which showed that while 85% of the P exported from the catchment was in particulate form, only 45% of P losses in surface runoff were attached to particles, the remainder of the sediment being mobilised from within the stream itself (Cooke 1988). As well as the erosivity of rainfall events (Hudson 1995; Huang et al. 1999), the main factors affecting the amount of particulate material mobilised in surface runoff from pastures are the degree of groundcover, slope gradient and length, degree of soil disturbance, and soil erodibility (Hudson 1995). Increasing slope angle increases the sediment transport capacity of runoff due to increased flow shear (Huang et al. 1999) and sediment, and hence P, loads tend to increase with higher velocity surface and subsurface flows (Heathwaite and Dils 2000).

24 Surface erosion is usually the dominant source and pathway of P loss in temperate environments if there is less than 75% soil cover when runoff occurs (Burwell et al. 1975; Lang 1979). Plant groundcover increases infiltration and decreases runoff by funnelling runoff down plant stems, providing macropores at the plant base for water to enter the soil, by absorbing raindrop energy thereby decreasing the rainfall erosivity, and by slowing the runoff down and reducing its sediment transport capacity (Greene et al. 1994). The potential for hillslope runoff and erosion is therefore increased when groundcover and/or soil strength is reduced by overgrazing (Costin 1980), waterlogging (Cooke 1988) or animal treading (Betteridge et al. 1998). A moderate degree of cattle pugging of the soil surface can however lower runoff volumes by increasing surface pooling of water (Greenhill et al. 1983a). Controlling grazing by rotating stock through a number of paddocks can enhance the survival of perennial pasture species (Warn et al. 2001; Chapman et al. 2003) except where other growth conditions are limiting (Waller et al. 2001b). Maintenance of groundcover and pasture density enhances rainfall infiltration, provides mechanical resistance to movement of particulate matter and increases surface roughness, which reduces runoff, erosion and PP losses (Olness et al. 1975; Costin 1980; Nash and Murdoch 1997). If not managed well, however, higher stocking rates supported by a rotationally grazed system can enhance soil disturbance and reduce ground cover on wet soils, which increases PP loss in runoff (Lambert et al. 1985). In contrast, runoff from forest and well-managed pasture hillslopes contributes significant amounts of DP to waterways, primarily due to low rates of soil erosion and high desorbable P contents of soils (White and Sharpley 1996). Nash and Murdoch (1997) found that greater than 90% of P in surface runoff from dairy pastures in south eastern Australia was in the dissolved form, and Lambert et al. (1985) also found the concentrations of dissolved inorganic

P in hillslope pasture runoff increased with increasing POlsen.

Seasonal variation

In environments where runoff occurs throughout the whole year, there is some evidence for higher P concentrations in runoff and stream flow during summer than in winter, which has been attributed to higher rates of mineralisation and lower pasture growth rates leading to accumulation of readily mobilised P between summer flow events, as well as dilution of P by winter flows (McColl et al. 1977; McColl and Gibson 1979; Heathwaite and Dils 2000).

2.6.2 Pathways, concentrations and loads of P movement from pastures

The hydrological pathway, its discharge capacity and contribution to overall movement of surplus water within a grassland catchment are some of the most important factors

25 controlling the amount of P exported from land to surface waters (Heathwaite and Dils 2000). Total nutrient loads (kg/m2.yr) being lost from pastures can be considered a better indicator of the risk of algal blooms in waterways in the long term than the concentrations in the flow because the total amount of P available determines the final biomass, rather than the growth rate, of aquatic plants and algae (Correll 1998). This suggests concern over even low concentrations of P where volumes of flow are large (Burwell et al. 1975). The pathway of water flow determines the P sources and sinks with which water comes into contact, as well as the duration and degree or area of contact (Ryden et al. 1973). This in turn influences the extent to which physical, chemical and biological processes of P mobilisation occur (Haygarth et al. 2000). The relative importance of each transport pathway and mobilisation process depends on the dominant hydrologic mechanisms of the hillslopes, which varies in time and space as discussed earlier. Where flows in active pathways are similar, however, the concentration of P becomes the determining factor. This was highlighted by Haygarth et al. (1998) who found that total P exports from drained pasture lysimeters were 30% lower than from undrained pasture lysimeters despite similar combined volumes of surface and drainage flow. The higher total P exports from the undrained pasture reflected the higher P concentrations in the surface pathways than subsurface flow. Pastures are usually enriched with nutrients within the top few centimetres of the soil surface due to applications of fertiliser, excretion of dung and urine and an accumulation of organic matter and cycled nutrients that is not disturbed through cultivation (Haynes and Williams 1993; Haygarth et al. 1998). Surface runoff is therefore usually the most important pathway for P movement from pastures because there can be a high energy of interaction between these surface P sources and runoff (Haygarth et al. 2000), and because surface runoff pathways usually provide a quicker flow route to catchment waterways than subsurface pathways. The nutrient concentrations of surface flows are also potentially more sensitive to single events such as heavy rain after fertiliser application than subsurface flows (Haygarth and Jarvis 1996). In studies where both surface and subsurface lateral flows occurred, surface runoff accounted for the majority of P exported from pastures, with loads in the range 0.1 to 2.6 kg P/ha from pasture catchments in northeast Victoria (Ridley et al. 2003), South Australia (Stevens et al. 1999; Fleming et al. 2001) and New Zealand (Cooke 1988) (Table 2-2). Concentrations of total P measured in surface runoff from sheep pastures in southern Australia range between 0.3 and 2.0 mg P/L and are similar to those measured in New Zealand and the United Kingdom (Table 2-2). Dissolved reactive P, total N and nitrate-N concentrations from sheep, cattle and dairy pastures ranged between 0.06 to 4.7, 1.68 to 6.4 and 0.1 to 1.79 mg/L respectively (Table 2-2). All these P and N concentrations are above desirable levels in streams and reservoirs (Anon 2000b). The variation in runoff P concentrations from pastures between locations reflects the range of soils, climates and land management practices studied

26 and highlights the need for regionally specific information on the concentrations of P that occur in runoff as well as the lack of data for pastures in south-west Victoria. Subsurface pathways tend to have fewer P sources and often more sinks than surface routes. The concentration of P in subsurface flows is therefore usually reflective of the P sorption capacity of the main soil horizon through which it flows (White 1980). Again the temporal degree of interaction is important because desorption and deposition decreases as the drainage rate increases (Cox and Pitman 2001). Movement of P via subsurface lateral flow pathways can be an important pathway of P loss on duplex soils (Kirkby et al. 1996; Stevens et al. 1999; Cox and Ashley 2000; Cox and Pitman 2001). Deep drainage as matrix flow is not normally an important pathway for P movement because subsoils are lower in P and organic matter than surface soils, and the clay content of subsoils usually increases with depth (Costin and Williams 1983; Heathwaite and Dils 2000). However, adsorption of soluble phosphate can be reduced in soils of high organic matter due to soluble organic matter coating P sorption sites (Pierzynski et al. (1994) in Kirkby et al. (1997) resulting in higher equilibrium soil water P concentrations (Ryden et al. 1973). In contrast, P in drainage can be significant in fertilized soils where drainage is high (Ridley et al. 2003) and in sandy soils where the capacity for adsorption of solution P is low due to the low clay and sesquioxide content (Russell 1960b; Ozanne et al. 1961; Holt et al. 1970; Lewis et al. 1981; Ritchie and Weaver 1993). Similarly, even leachate from a silty clay (22% clay to 30 cm depth), which is expected to fix most reactive P, can be moderately high in concentrations of dissolved reactive P (0.12 mg P/L) (Turner and Haygarth 2000a). Between 0.01 and 0.148 kg/ha.yr of P was lost via subsurface pathways that were measured at a range of scales in Australia (Table 2-2). Similar to loads, concentrations of P in subsurface flows (TP 0.1-0.9 mg/L, DP 0.01-1.40 mg/L) were generally lower than for surface flows. Some very high nitrate concentrations were measured in drainage under lucerne pasture (23 mg /L), however, and were attributed to low levels of nitrate adsorption in soils (Cox and Pitman 2001). Therefore in catchments where the main contribution to stream flow is from subsurface flow, P concentrations may bear little resemblance to surface soil and land management and instead reflect the underlying geology and native P concentrations (Thomas and Crutchfield 1974; Heathwaite and Dils 2000). In many situations, losses of P in groundwater from agricultural catchments are similar to those from forested catchments suggesting that the groundwater is unaffected by agricultural practices (Ryden et al. 1973). Where significant bypass flow occurs, however, adsorption of soluble P along macropores is restricted by the small contact area and time between the mobile phase solute and the surrounding soil surface (White and Kookana 1998) and because the inner walls of macropores are sometimes coated with an organic cutan layer (Chittleborough et al. 1996), which may further inhibit P sorption. This results in a non-equilibrium soluble P concentration in solution (White and Kookana 1998)

27 so the P concentrations in throughflow or drainage may be more reflective of surface P sources than that of the macropore surface or surrounding soil matrix (Kirkby et al. 1997). Particulate P can also be transported along and eroded within soil macropores (Heathwaite and Dils 2000; Turner and Haygarth 2000a). Macropores can therefore have essentially the same effect as artificial subsurface drainage in rapidly transporting ‘new’ water, and associated nutrients and chemicals, from landscapes (Kirkby et al. 1997; Heathwaite and Dils 2000). Artificial subsurface drainage can turn intractable land into highly productive sheep, cattle and dairy pastures (McFarlane and Cox 1992) and in many cases will reduce the volume and losses of P in surface runoff (Sharpley and Syers 1976; Haygarth et al. 1998). However, the losses of nutrients via surface and subsurface drainage can be of concern (Yeates 1993; Heathwaite and Dils 2000; Cox and Pitman 2001) as a result of an increased amount and rate of water movement in comparison to more tortuous flow routes of water and nutrients to stream networks in undrained landscapes (Simard et al. 2000). There is evidence, however, for low P concentrations in tile drainage where dairy manure amendments are incorporated and soils are fine textured, allowing for substantial P sorption to occur (Randall et al. 2000). The relative magnitude of surface and subsurface flow, and associated P transport, is not known for the dryland grazing environment of south-west Victoria and requires investigation. The ultimate significance of the various hydrologic pathways for P export within any catchment, however, will depend on the degree of connectivity of the pathway with the greater surface water network. For surface flows, P can enter reservoirs, channels and streams directly or become translocated downslope. P in subsurface flow may contribute to surface flows as return flow downslope, enter gullies directly as subsurface lateral flow, leach into groundwater or just become redistributed within the soil profile (Kirkby et al. 1997).

28 Table 2-2: P concentrations, forms and loads in runoff and drainage

Location and land use Flow pathway Average P or N Nutrient load Percentage of Author and notes fraction (mg P/L) (kg/ha.yr) TP as DRP, TRP or DP (%) UK Devon cattle pasture, Olsen P 1)Surface + lateral.subsurface 1)TP 0.15 ± 0.04 1) TP c. 3 1) 69 %DP (Haygarth et al. 1998) 6-8, 0- 25 kg P/ha ± slurry flow (0-30 cm) – undrained 2) Surface + lateral subsurface 2) TP 0.23 ± 0.05 2) TP c. 0.4 2) 69 % DP flow (0-30 cm) - drained UK Tile drained dairy pasture, Surface runoff TP 1.14 ± 0.08 43% DRP (Heathwaite and Dils Olsen P = 25 mg/kg, Fertiliser 26 Matrix flow to 20 cm TP 0.38 ± 0.19 67%DRP 2000) kg P/ha plus manure slurry Groundwater at 200 cm TP 0.36 ± 0.09 24% DRP Macropore drainflow at 45 cm TP 0.84 ± 0.07 23% DRP Dairy pasture, cattle slurry applied UK drainage TP 0.45 – 0.64 TP 3.6 – 5.0 36-41% TRP (Hooda et al. 1996) with or without fertiliser UK Brimstone 1)Drainage 1)TP 0.26-0.6 1)TP 0.37-0.91 1) 10 - 12 % TRP (Brookes et al. 1996) 2)Surface runoff 2) TP 1.5 2) TP 3.1 2) 13 – 27 % TRP UK Woburn Erosion Experiment Surface runoff TP max 13 (Brookes et al. 1996) UK Broadbalk, wheat, range of Pipe drainage at 65 cm TP 0.03 - 2.75 66 – 86% DRP (Heckrath et al. 1995) soil fertility levels England, Rosemaund, 8 sites, Surface and subsurface flows TP 1-2.6 (Hodgkinson 1996) variable management Shepparton, Australia, Irrigated Surface runoff (natural rainfall) TP 0.44-2.15 TP 0.14-1.6 64 – 76 % TRP (Nexhip and Austin dairy pasture, low, medium and TN 1.1-4.5 1998) high P fertility (much higher loads after irrigation & fertilisation) Taita, New Zealand. Sheep Surface runoff from 400 m2 TP 2.41 (min 0.2) (McColl and Gibson pasture, crash grazed @ 1250 runoff plots TKN 40.2 (min 1.01) 1979) sheep/ha, 51 kg P/ha Taita, New Zealand Streamflow (McColl et al. 1977) 1) Hill pasture, 380 kg super/ha 1) TP 0.143, NO3 0.384 1) TP 0.293, NO3 1.436 (nitrate is soil type not 2) Exotic forest 2) TP 0.062, NO3 0.014 2) TP 0.07, NO3 0.044 pasture effect) (Table continued on next page)

29 Table 2-2 cont. Location and land use Flow pathway Average P or N Nutrient load Percentage of Author and notes fraction (mg P/L) (kg/ha.yr) TP as DRP, TRP or DP (%) New Zealand sheep and cattle Surface runoff from High P: DRP 0.035 High P: TP 0.94, TN 9.9 High P: 15% DRP, (Lambert et al. 1985) pasture, 17-25o slope 0.13-1.53 ha plots InorgN 1.35 54% inorg.N High P; 48 –86 kgP/ha/yr, 9-13 Low P: DRP 0.017 Low P: TP 0.95, TN 9.7 Low P: 10% DRP, breeding units/ha InorgN 0.56 31% inorg.N Low P; 12-19 kgP/ha/yr, 6-10 breeding units/ha Victoria, Australia, Low intensity Surface runoff from TP 0.35-2.0 TP 0.22, 50% DRP (Tham 1983) sheep pasture 37-44 ha catchments DRP 0.06-1.05 TN 1.15 (patchier paddocks had - super 100 kg/ha, 0.2-2.0%slope TKN 3.67-6.40 NO3 0.014 lower DRP%) South Australia Streamflow 1)TP 0.35, TN 1.68, 1)TP 1.0, TN 9.0 1)14%DP, 67% (Nelson et al. 1996) 1)Sheep pasture, 5.7 kg P/ha SS 315 SS 924 TDN steep, loam over clay 1.3 km2 2) TP 0.4, TN 2.27, 2) TP 1.1, TN 6.6, 2)Dairy pasture, avg 11.4 kg P/ha SS 14 SS 40 2) 76%DP, 82% +some dairy effluent TDN flatter, sand over clay, 3 km2 Gippsland, Australia. Dairy Surface runoff from 3.6 ha TP 5.2 TP c. 0.95 93% DP (Nash and Murdoch ryegrass-clover pasture, 60 kgP/ha DRP ~ 4.7 89% DRP 1997) previous yr, Total suspended solids Olsen P 25-29 mg/kg 93-329 mg/L NSW, Australia, Surface runoff TP 0.12, TN 0.62 (Costin 1980) phalaris/clover pasture, 12 88 ha 3.7%slope SS 4-376 sheep/ha 12 kg P/ha Flaxley, South Australia paired 1)Surface runoff 1) & 2) combined 1)TP 0.3- 2.3, 1)50-60 % DP (Fleming and Cox - dairy pasture catchments 2.2/2.6 TP 0.3-3.0 NO3 0.12-0.43 1998) ha, 15 kg P/ha broadcast 2)Subsurface flow (A/B horizon DP 0.1-1.4 2)TP 0.02-0.03, 2) 40-50% DP 88-96% of flow was - interface) PP 0.1-1.9 NO3 0.09 surface runoff

(Table continued on next page)

30 Table 2-2 cont. Location and land use Flow pathway Average P or N Nutrient load Percentage of Author and notes fraction (mg P/L) (kg/ha.yr) TP as DRP, TRP or DP (%) Mt.Bold, South Australia Paired All flows combined (Stevens et al. 1999) cattle pasture catchments 4.2 & 1)surface runoff TP 0.37-1.02 1)TP c. 0.1-0.25 1) 50% DP 3.6 ha 8 kg P/ha A horizon Silty 2)A horizon throughflow DP 0.18-0.53 2)TP c. 0.01-0.07 2) 50%DP - clay loam 3)B horizon throughflow NO3 0.6 3) TP <0.04 B horizon med clay Myponga, South Australia Paired catchments Mainly flow path 2 Flow path 2 only (Stevens et al. 1999) Ungrazed paired pasture 1)surface runoff TP 0.134-0.31 TP c. 0.125, 2) 8%DP catchments 2.2 & 2.6 ha 2)A horizon throughflow DP 0.01-0.02 - A horizon coarse sand, B horizon 3)B horizon throughflow NO3 1.02-1.79 heavy dispersive clay Keynes, South Australia 1)surface drains 1) TP .0.3-1.4 1)TP 0.001-0.073, 1) 35% DP (Cox and Pitman 2001) - - 0.2 ha plots, 3 Pasture types: 2)A horizon throughflow drains NO3 0.1-0.3 NO3 0.001-0.043 (pasture and drain type standard, phalaris and lucerne, 2) TP 0.1-0.9 2)TP 0.001-0.148, 2) 70% DP had no significant - - NO3 0 -23.1(lucerne was NO3 0.012-0.782 effect) high)

31 2.6.3 Mobility of pasture P pools

It is possible for P in any of the fractions or pools found in the soil-plant-continuum to be transferred to surface water although the mobility via erosion/entrainment, dissolution, desorption or leaching of the various sources of P in soils and pastures varies.

Directly deposited dung

Dung is a potentially significant source of P for runoff and leaching losses from pastures because as dung decays the nutrient content of soil increases (Rixon and Zorin 1978; Williams and Haynes 1992) and because manure can be directly entrained in runoff water (Reddy et al. 1978; Withers et al. 2001) or act as a source of P for dissolution and desorption into runoff and drainage water (Sharpley and Syers 1976; Sharpley 1996). An extreme example of contributions of manure P to surface waters was documented by Holt (1970), where phosphate concentrations of up to 50 mg/L were measured in runoff from feedlots with concrete surfaces. Up to 85% of the P in ground samples of live and hayed-off pasture plants (Bromfield and Jones 1972; Gillingham et al. 1980) and 27% of P in sheep dung material (Bromfield 1961) is soluble in water and can potentially be leached into runoff water (Sharpley 1981; Sharpley and Moyer 2000).

Artificially applied manure

Nash and Murdoch (1996) found that the risk of soluble inorganic P leaching from cattle manure was less than that from inorganic fertilisers but that losses of total P were greater under manure. Similarly, Reddy et al. (1978) found that application of both cow manure and chemical fertiliser P increased the equilibrium P concentration of three soils, but that manure application had a greater effect. In contrast, P exports from pastures amended with liquid and solid cattle manure may be less than from inorganic fertiliser amendments when the water holding capacity of the organic materials reduces the volumes of runoff (Reddy et al. 1978; Joshua et al. 1998; Withers et al. 2001) and if P sorbing constituents in effluent such as carbon and iron significantly increase the P sorption capacity of soils (Holford et al. 1997).

Pasture herbage

Concentrations of up to 150 mg P/L in simulated pasture leachate were measured by Bromfield and Jones (1972). Whilst these concentrations may represent only a small contribution to the pool of available soil P when compared with fertiliser additions (Bromfield and Jones 1972), in leachate these concentrations would be considered a very high risk for losses in runoff and delivery to freshwater ecosystems. Direct canopy leachate from cotton,

32 sorghum and soybean yielded much lower, but still potentially significant, concentrations of soluble P (0.019 - 0.030 mg P/L) (Sharpley 1981). Depending on the amount of moisture and microbial activity, this inorganic P may then be partially immobilised during plant decomposition and become resistant to leaching (Bromfield and Jones 1972).

Soil

The forms of P that occur in soil vary in their availability for both plant uptake and mobilisation in runoff and leachate. Due to the surface enrichment with nutrients, soil solution P decreases with depth, even in unfertilised grassland soils. The high cation exchange capacity of pasture soils generally increases the retention of soil nutrients (Russell 1986). However, dissolved organic material, such as in soils beneath manure deposits, has a reduced sorption strength and the organic phosphates are more readily leached than inorganic phosphate complexes (Holford and Mattingly 1975; Holford 1989; Chardon et al. 1997). Furthermore, dissolved organic material can also decrease the chemical activity of aluminium that precipitates P (Bloom et al. 1979). Both these mechanisms enhance the release of P to the soil solution and therefore increase its availability to plants, as well as its potential mobility in runoff and leachate (Haynes 1999; Higgs et al. 2000). Similarly, dissolved organic P fractions are more readily leached than inorganic forms (Ron Vaz et al. 1993). Up to 90% of P in soil leachate may be in the dissolved organic form with the inorganic contribution decreasing with depth (Haygarth and Jarvis 1996; Chardon et al. 1997) reflecting preferential adsorption of DRP to the subsoil (Haygarth et al. 1998).

Fertiliser P

The mobility of surface applied fertiliser P is dependant on the solubility of the fertiliser compound and the timing of fertiliser application in relation to runoff or drainage events. Sparingly soluble fertilisers such as rock phosphate (Ca10(PO4)6(OH)2) remain intact longer than soluble compounds such as superphosphate (Ca(H2PO4)2.H2O) and ammonium phosphate (NH4H2PO4). While this reduces the risk of DP losses from sparingly soluble forms, it increases the risk of fertiliser being directly entrained in runoff (Sale and Blair 1989; Sharpley and Halvorson 1994; White and Sharpley 1996; Nash and Halliwell 1999). For sandy soils, however, more soluble formulations may present a higher risk of losses in leachate and runoff due to the low P sorption capacity of these high-rainfall zone soils (Yeates and Clarke 1993; Nelson et al. 1996). The rate and extent of precipitation/dissolution and desorption/adsorption reactions between soil constituents and fertilisers depend on soil and fertiliser chemical properties such as pH and the amount and species of cations present (Holford and Patrick Jnr 1979; White 1980; Nash and Halliwell 1999).

33 2.6.4 Amount and spatial and temporal distribution of P in grazed pastures

On-farm nutrient management

The spatial and temporal distribution of P in relation to active hydrological pathways within pasture systems has a large influence on the potential for P to be lost to waterways. The pasture surface, particularly the top 5 mm, is the layer with which surface runoff interacts (Ahuja et al. 1981), so the depth of fertiliser application can influence the risk of P mobilisation in runoff. Broadcast fertilisers are more readily available for direct wash off and dissolution in runoff when compared with incorporation or subsurface applied P (Holt et al. 1970; Withers et al. 2001). However, where deep banding involves cultivation, there may be an increased potential for particulate P losses where ground cover is concomitantly low and erosion is enhanced (Nelson et al. 1996). The timeliness of fertiliser and manure applications compared with antecedent moisture conditions and rainfall events can also have a large influence on total losses of P in runoff and drainage (Gilchrist and Gillingham 1970; Haygarth and Jarvis 1997; Preedy et al. 2001). For example, in studies where runoff initially occurred less than 12 days after fertiliser was applied, runoff P concentrations were strongly influenced by P application rates (Romkens and Nelson 1974) and time since fertiliser was applied (Gillingham et al. 1997; Nexhip and Austin 1998; Nash et al. 2000). Because sorption of fertiliser P to soil increases over time (Barrow and Shaw 1975), direct effects of fertiliser on runoff P concentrations are less likely in environments where application of fertiliser P occurs several months prior to runoff or drainage occurring (Sharpley and Syers 1976; McColl and Gibson 1979; Nelson et al. 1996). P may accumulate spatially in areas where fertilisers and feeds are stored, where farm machinery used in transporting nutrients travel and in livestock handling sheds and yards, including dairy shed effluent ponds (Anderson et al. 1992; Gourley 2001). Large amounts of P can be lost in runoff and drainage if these concentrated P sources are not isolated from drains and watercourses or if fertiliser drifts directly into watercourses during application or if P spills or leaks from machinery used for spreading and transport. When losses via these processes occur they usually far outweigh P input to streams from diffuse agricultural sources. For example, twenty per cent of annual P export from hill country catchments in New Zealand was accounted for by direct input from aerial fertiliser topdressing (Cooke 1988). At the farm scale, areas and operations that involve high concentrations of P need to be managed as point sources of P and kept spatially separated form hydrological pathways which connect with watercourses.

34 Nutrient balance and soil P test level

It is difficult to trace the original source of P in runoff (Nash and Halliwell 2000), however, in environments where there is a significant lag time between fertiliser application and runoff, and grazing intensities are low, soil and organic matter can have a greater influence than fertiliser and grazing on the P lost in rainfall-runoff (Nexhip and Austin 1998). Applying nutrient amendments to pastures can lead to systematic heterogeneity in the distribution, and hence availability to hydrologic pathways, of soil nutrients. Levels of soil P also accumulate when P applications exceed the demand for P by plants. Soil P surpluses can occur where animal manure is applied to crops and pastures for disposal purposes at rates that are based on plant N requirements alone and where nutrient amendments are applied at rates exceeding plant requirements (Sharpley et al. 1994; Bundy and Sturgul 2001; Gourley 2001). In the UK, despite little change in fertiliser and manure inputs over last 25 years there is still an overall annual P surplus of 10 kg P/ha which is mainly concentrated where pig and poultry manure is spread to arable areas, and to a lesser extent where dairy and cattle manure is applied to grassland (Withers 1996; Heathwaite et al. 2000). A recent survey, however, found that farmers were increasingly making an effort to utilise manure nutrients efficiently but often lacked the confidence in advice about their nutritive value (Smith et al. 2001b). The environmental consequence of applying P at rates above plant requirements is exacerbated when continual application of P amendments exceeds the soil’s capacity to adsorb P (Barrow et al. 1998), which in turn increases the risk of P loss through surface and subsurface flow (Ritchie and Weaver 1993; Redding 2001). P loss risk indices are being used in North America to identify specific paddocks that have a greater risk of P loss based on both soil P levels and the activity of various hydrological pathways (Sharpley et al. 1999). Whilst the majority of pasture soils in Australia are deficient in P for optimum production (MacLaren et al. 1996), fertiliser is sometimes applied in excess of plant requirements leading to a growing concern that the amount of nutrients ‘leaking’ into the wider catchment will increase (Reuter et al. 1996). A recent survey of P distribution in agricultural soils on the south coast of Western Australia revealed that more than half of the approximately

8000 sites tested had sufficient POlsen status to forego P application for at least one year and that correction of sulfur and potassium, rather than P, deficiencies would reduce the rate of increase of soil P (Weaver and Reed 1998). The concentration of P in the soil solution, which undergoes significant temporal variation (Friesen et al. 1985), generally ranges between 0.01 and 3.0 mg P/L in agricultural soils (Frossard et al. 2000) and ryegrass-clover pastures require soil solution P concentrations (0-5 cm) up to 1.33 mg P/L to reach 90% of maximum yield (Wheeler and Edmeades 1995). This upper concentration is more than 50 times greater than concentrations that can support

35 problem algal growth in aquatic systems (Correll 1998). The disparity has spawned particular focus on the relationship between soil P levels and runoff P levels and investigations into how to reconcile the contrasting goals of attaining high pasture growth rates whilst maintaining the quality of catchment waterways (Monaghan et al. 2003). In the USA and Europe, correlations between runoff and drainage P concentrations and soil test P have been used to define threshold soil P levels for environmentally acceptable runoff and drainage P concentrations (Breeuwsma and Silva 1992; Sims et al. 2000; Weld et al. 2001).

For example, in the UK the optimum POlsen for spring barley and winter wheat growth is 25 mg

P/kg (Johnston and Poulton 1997), and this was well below the POlsen level above which there is some evidence for drainage P concentrations exceeding desirable levels (i.e. above 60 mg P/kg) (Heckrath et al. 1995). The difference between 25 and 60 mg P/kg could be considered a good ‘buffer’ between increasing productivity and threatening water quality (Higgs et al. 2000). From a management point of view, these relationships are only relevant for soils that coincide with active hydrologic pathways for P transport. Soil testing and farm nutrient balances are encouraged in grazing enterprises in Australia to provide sufficient nutrients to support plant growth (Court et al. 1998) and to help avoid accumulation of surplus P in soils (Gourley 2001). However, the potential for threshold soil test P levels to be used to reduce the risk of P losses in runoff from pastures has not been investigated.

Grazing management

The uneven distribution of nutrients returned to the pasture in dung and urine deposits due to stock camping (Gillingham and During 1973; Gillingham 1987; McCaskill and Cayley 2000) and the effect of grazing events can influence the risk of P loss to waterways where these camps and grazing episodes coincide spatially and temporally with hydrological activity (Williams and Haynes 1992; Tate et al. 2000). The combined effects of faecal deposition, soil disturbance and defoliation of plants during a grazing event have been shown to increase losses of P in surface runoff by comparing runoff events from pastures grazed only intermittently or seasonally (McColl and Gibson 1979; Schepers and Francis 1982) (Chichester et al. 1979) to ‘non-grazed’ events, and by comparing runoff from grazed and non-grazed paired plots (Sharpley and Syers 1976). As well as P concentrations in runoff increasing by up to 40% due to grazing, increased concentrations of organic carbon and P have been measured indicating grazing increases the mobilisation of organic matter (Sharpley and Syers 1976; McColl and Gibson 1979; Schepers and Francis 1982). However the combined grazing effects are generally less than the effect of fertiliser applications (Sharpley and Syers 1976; McColl and Gibson 1979; Nexhip and Austin 1998; Nash et al. 2000). Preferential grazing and dung deposition on ridges and hillcrests may counter-act downslope movement of P in runoff (McColl and Gibson 1979). However, stock camps

36 sometimes develop in drainage lines (Johnston et al. 2003), on gentler slopes (Saggar et al. 1990), and in high traffic areas near water and feed troughs, gates and along laneways, and these areas can be more prone to surface runoff (Lambert et al. 1985; Haygarth et al. 2000). Stock camps are more distinct at low stocking rates (Morton and Baird 1990) so the practice of rotational grazing has the potential to reduce the effects of stock camping by increasing the area of dung distribution through increased stocking density and control of preferential grazing (Haynes and Williams 1993). Gillingham (1983), however, found dung distribution was influenced more by variations in residual pasture dry matter on offer than grazing management. A more even distribution of dung may improve the efficiency of nutrient cycling in pastures, however, there is little information on the potential for rotational grazing to be used as a tool for managing the proximity of stock camps to hydrological flow pathways within a pasture. Recent evidence found by Kuykendall et al. (1999) for beef pastures in the USA suggests that grazing method has no influence on the quality and quantity of runoff water. The risk of manure P transfers to runoff in extensive grazing systems is likely to be of greatest significance where stock have direct access to streams.

2.7 Experimental approaches to measuring P losses

A range of methods has been used to study the potential for P to be transferred from soil and other sources to water. These include investigations of soil and water interactions in laboratories, ex situ simulations of rainfall and runoff (Huang et al. 1999; McDowell and Sharpley 2001) and large soil core or lysimeter studies (Haygarth and Jarvis 1997; Condron et al. 2000; Turner and Haygarth 2000a). Sediment and nutrient movement has also been quantified in the field at a range of spatial and temporal scales including microplots (of 1 or 2 square metres), hillslope plots (Heathwaite et al. 1998; Pote et al. 1999b; Heng et al. 2001), subcatchments (Costin 1980; Tham 1983; Fleming and Cox 1998; Stevens et al. 1999) and stream gauging in large catchments (McColl et al. 1977; Nelson et al. 1996; Cox and Ashley 2000; Edwards et al. 2000). Another method that is used is nesting plots of a range of sizes within a single watershed (eg Costin (1980)). Each method of investigation has its merits and limitations and is chosen to best suit the purpose of the experiment. The merits and limitations of rainfed-hillslope and small-plot rainfall simulator studies are discussed in detail in Chapters 5 and 6.

2.8 Predicting P losses from pastures

Because of the funding, time and labour required for conducting field-based runoff studies, a tool that can be used to extend principles tested in field studies and to predict nutrient losses using easily measurable parameters, is sought after. Regression analysis of stream

37 nutrient concentrations and upstream landuse is a helpful indicator of major P sources within catchments (McFarland and Hauck 1999; Ekholm et al. 2000). Nutrient loss risk indexing approaches using geospatial information have also been developed and applied at farm and catchment scales (Gburek et al. 2000; Heathwaite et al. 2000; Caruso 2001; Weld et al. 2001). A range of empirical and physically-based models have been used to predict the concentrations and loads of the main forms of P in runoff from agricultural land (Knisel 1980; Sharpley et al. 1982; Williams et al. 1983; Cooper et al. 1992; Fleming and Cox 1998; Nash et al. 2000; Daly et al. 2002). Process models that predict the concentrations of particulate and dissolved P concentrations in runoff are described in Chapter 6. The simplest of these models are relationships between runoff and drainage P concentrations and soil test P (STP), and good correlations have been found for a range of soils and locations (Sharpley et al. 1977; Heckrath et al. 1995; Pote et al. 1996; Sibbesen and Sharpley 1997). STPs measure variable proportions of both the ‘quantity’ of soil P available for uptake by plants during a growth cycle and the ‘intensity’ (i.e. concentration) of soil solution P. Soil P extractants that primarily measure P intensity, such as CaCl2 (Kuo 1996) and water (Sissing 1971) have also been used instead of direct measurements of runoff P concentrations to estimate the risk of P release from soil to runoff or drainage water (Gartley and Sims 1994; Daly et al. 2001). Relationships between these intensity measurements and routine agronomic STPs (Oloya and Logan 1980; Hughes et al. 2000; Kleinman et al. 2000; Paulter and Sims 2000; McDowell et al. 2001) often reflect linear or curvilinear quantity-intensity plots typical of P sorption isotherms (Sposito 1989). The degree of curvilinearity of the plots depends on the range of soil P contents measured (Beckett and White 1964; McDowell et al. 2001). In some soils there is a sharp increase in the slope of the quantity-intensity plot above a certain STP level (Heckrath et al. 1995). This STP level has been referred to as a ‘change point’, and may be considered analogous to the break point that delineates the two linear sections of P sorption isotherms when fitting a two-surface Langmuir equation (Syers et al. 1973; Holford et al. 1974). For a wide range of soils from the United Kingdom, New Zealand and the USA the change point occurred at a POlsen of between 20 and 112 mg P/kg (McDowell et al. 2001). However, owing to differences in soil P sorption capacity and other soil properties (Sharpley 1995), relationships between runoff P concentration and soil P test levels are not geographically universal and usually need to be investigated on a soil-type basis (Pote et al. 1999a; McDowell and Sharpley 2001). In addition, soil tests that have been designed to assess the plant available P content of soils do not always reflect the magnitude of P concentrations in runoff. A range of new soil tests including the degree of soil P saturation (Beauchemin and Simard 1999) and water-extractions (Yli-Halla et al. 1995) have therefore been investigated for their ability to predict runoff P concentrations over a range of soil types (Sims et al. 2000).

38 The degree of soil P saturation (Psat), is a measure of the amount of sorbed P relative to either the total P sorption or P buffering capacity of the soil (Sharpley 1995; Sibbesen and Sharpley 1997). Soil P buffering capacity and the total P sorption capacity are key characteristics related to the partitioning of P between soil and solution. Addition of P to a soil with a higher P buffering capacity will cause less change in solution P concentration than a poorly buffered soil. Highly buffered soils therefore present a smaller risk of releasing dissolved P to the environment. To calculate Psat, the amount of sorbed P is usually measured using a routine STP, and the P sorption and buffering capacity are derived from adsorption isotherms (Fox and Kamprath 1970; Rayment and Higginson 1992), or estimated using correlated single point sorption indices or surrogate variables of the amount of sorbing material (Rayment and Higginson 1992; Mozafari and Sims 1994; Simard et al. 1994; Pote et al. 1999b; Paulter and Sims 2000; Burkitt et al. 2002). A range of methods for measuring Psat is reviewed by Beauchemin and Simard

(1999). Two of the commonly applied indices of Psat are the Langmuir (PsatL) and oxalate

(Psatox) indices.

The Langmuir Psat index (PsatL) is calculated utilising a user-defined STP to represent the quantity of labile P, expressed relative to the maximum quantity of P the soil can adsorb

(Pmax) (Sharpley 1995; Paulter and Sims 2000). However, levels of PsatL are rarely directly comparable because of differences in both the methods used to measure Pmax and the choice of extracting solutions used to measure the STP level. Estimates of Pmax depend on the range of P concentrations used and the closeness of fit of the Langmuir equation (Barrow 1978).

A second index used to measure Psat is the oxalate index (Psatox). The oxalate Psat index uses the sum of the amounts of Al and Fe (mg/kg) extracted by acidified ammonium oxalate (Feox and Alox) as a surrogate estimate of the soil’s P sorption capacity. Acid ammonium oxalate extracts amorphous minerals that form the bulk of the P sorption capacity of soils, namely Al and Fe in acidic soils (Toreu et al. 1988; Freese et al. 1992). Psatox is then calculated as the ratio of ammonium oxalate extractable P to the sum of Feox and Alox. The merits and limitations of these two Psat indices are discussed further in Chapter 7. Indices of P saturation have had varying success over the simple STP in predicting P in runoff from agricultural soils. Sharpley (1995) found a single linear relationship described the association between Psat and DP concentrations in simulated runoff from 10 Oklahoma soils, whereas relationships using STP as a predictor differed between soil types. In contrast, the routine Morgan’s STP was an adequate predictor of Psat (r = 0.88), for 59 New York soils of similar parent material, which highlighted the ability of Morgan’s solution to characterise both quantity and intensity factors in these soils (Kleinman et al. 1999). Using Morgan’s STP as a surrogate variable for Psat would allow soil testing laboratories to estimate environmental P loss without the need to perform P sorption tests (Kleinman et al. 1999). Pote et al. (1999b)

39 found that although Psat was significantly related to runoff DRP concentrations for each of three soil types, the linear regressions did not have the same slope. In their review of Psat indices, Beauchemin and Simard (1999) also concluded that if threshold levels of Psat above which there may be unacceptable runoff or drainage P concentrations were to be defined, soils may still need to be grouped according to their P sorption characteristics. Australian soils in their native form are usually deficient in P for optimum pasture production (Moody and Bolland 1999). Near maximum plant yields occur at a soil solution concentration of approximately 0.2 mg P/L (Fox and Kamprath 1970), which is markedly lower than the concentrations required to saturate sorption sites in most soils (Woodruff and Kamprath 1965). There is a large body of information regarding P sorption characteristics of Australian soils (Barrow 1980; Holford and Cullis 1985; Toreu et al. 1988; Singh and Gilkes 1991; Barbare et al. 1997), however, few researchers have reported levels of P saturation. It was estimated at a workshop on P that no more than one third of agricultural soils in southern Australia had reached ‘P saturation’ with respect to plant requirements (Costin and Williams 1983). Holford et al. (1997) found that the initiation of P leaching from sandy soils amended with diluted sewage effluent occurred at a Psat of 30%, which was equivalent to the percentage of the total P sorption capacity they earlier suggested was occupied by ‘high energy’ bonding sites (Holford et al. 1974). P sorbed to the remaining ‘low energy’ bonding sites is also described as the ‘reversibly’ adsorbed P which controls the supply of P in the soil solution, as opposed to the ‘irreversibly’ adsorbed P which is ‘fixed’ in the soil by slow diffusion mechanisms (Taylor and Schwertmann 1974; Breeuwsma and Silva 1992). There is little information on relationships between hillslope runoff P concentrations and the P status of Australian pasture soils, or the ability of more complex models to predict P concentrations in runoff from sheep-grazed pastures in southern Australia.

2.9 Management strategies for minimising P losses in runoff

A range of management practices has been advocated to mitigate P movement from pastures into waterways. For any set of climate, landscape and land management conditions, it is important to understand the mechanisms of P mobilisation and attenuation in order to define appropriate management strategies for minimising the loss of P from pastures. For intensive grazing industries, one option is to increase the P use-efficiency of the grazing animal. This may be achieved by minimising the amount of off-site supplementary feed in the animal’s diet and/or by lowering the dietary intake of P by dairy cattle. P excreted in faeces increases with dietary P concentration (Morse et al. 1992), however there is some evidence that there is no milk yield advantage when the dietary P concentration exceeds approximately 0.34% (Brodison et al. 1989).

40 A number of strategies can be used to minimise erosion and hence transport of PP from pastures. Mechanisms include maintaining groundcover above 75% (Lang 1979; Hairsine and Prosser 1997) and using sediment dams and/or vegetative buffer strips and riparian zones to trap PP before it moves into waterways (Heathwaite et al. 1998; Pearce et al. 1998; Clausen et al. 2000; Lee et al. 2000) (Cooper et al. 1995). Vegetative buffer strips are most effective in physically trapping particulate P, but may also reduce dissolved P by enhancing water infiltrating into the soil within the buffer (Hairsine 1997). Another important strategy for minimising both PP and DP movement from pastures is to control the availability, or source, of P in hydrologically active areas. Most importantly, to avoid a long term build-up of surplus P, inputs and outputs of P from the soil-plant-animal cycle in pastures need to be balanced using nutrient budgets that account for nutrients recycled in manure and spread as effluent to pastures (Gourley et al. 2001). In some regions in USA, soil P test levels that represent upper thresholds for maintaining acceptable concentrations of P in runoff and drainage have been identified and are used to inform decision making about pasture and crop nutrient requirements (Sharpley et al. 1996; Sibbesen and Sharpley 1997; Sims et al. 2000). A critical component of minimising losses of P in runoff is avoiding spreading fertiliser or effluent on areas that are prone to waterlogging and runoff (Gburek et al. 2000; Sharpley et al. 2001b) or just prior to runoff occurring (Nash et al. 2000). Similarly, in some cases, subsurface application of fertiliser or manure and periodic ploughing of no-till soils to remove P from the main zone of interaction with runoff and to redistribute nutrients that accumulate on the surface of pastures is also recommended (Sharpley and Rekolainen 1997). Demonstrated cost-effectiveness of any management practice recommendation is likely to markedly improve its adoption by farmers (Sharpley and Rekolainen 1997). Fertiliser distributors in south-west Victoria are offering a price incentive for early application of fertilisers (Montgomery 2000), which may help reduce fertiliser P losses by increasing the period between application and runoff occurring. In general terms, it is ideal to prevent an algal bloom occurring because the risk of recurrence after an initial bloom increases due to the seedbed of cysts and autospores left in sediments after senescence of a bloom (Harris 1994). Furthermore, the lag between reducing P load inputs and a reduction in waterway P concentrations occurring can be considerable (Smith and McCormick 2001) which places importance on anticipating management practices and regions which may pose a risk of P losses in the future.

2.10 Knowledge Gaps

Surface runoff, subsurface lateral and vertical matrix and macropore flow have all been identified as important hydrological flow pathways in soils and environments similar to the duplex soils and undulating pastures in south-west Victoria. However the relative importance of

41 each of these flow pathways for P transport, and the quantities of P loss from these pasture systems is not known and requires investigation in order to identify appropriate land management strategies for mitigating eutrophication of catchment waterways. The forms and concentrations of P in surface runoff and other hydrological flow pathways are influenced by climate, soil and land management practices that are spatially and temporally variable factors both within and between locations. It is therefore difficult to relate concentrations and forms of P measured under any one set of conditions to another space and time without validated conceptual, process or empirical models. There is no information on the concentrations and forms of P that occur in runoff from high or low P fertility sheep pastures in the high rainfall grazing region of south-west Victoria and there is no information on the potential for rotational grazing to be used as a tool for managing the proximity of stock camps to hydrological flow pathways. Similarly, the ability of empirical models or routine soil P tests to predict P concentrations in runoff from sheep pastures in southern Australia has not been tested and requires investigation. Simple predictive models would reduce the need for time- consuming field experiments to be conducted to answer land management questions.

42 CHAPTER THREE

3 Characterisation and experimental design of the Vasey field site

3.1 Introduction

An experimental runoff site used for this project was established on a commercial wool producing property located approximately 55 km north of Hamilton at Vasey (37o24’ S, 141o55’ E) on the Dundas Tablelands of south-west Victoria, Australia (Figure 3-1). The runoff site was set up adjacent to an existing replicated grazing experiment, which was part of the ‘Sustainable Grazing Systems’ (SGS) national experiment (Andrew and Lodge 2003; Chapman et al. 2003). This chapter describes the climate, landuse, geology and soil properties at Vasey, which were characterised using existing information and direct measurements. This information was used to identify the likely hydrological pathways for P movement from the hillslope pastures. The experimental design of the runoff site is also described.

• Vasey

Figure 3-1: Location of the Vasey field site in south-west Victoria, Australia (Maps from http://audit.ea.gov.au/ANRA/atlas_home.cfm and www.dpi.gov.au/vro)

43 3.2 Climate, temperature and evaporation

South-west Victoria has a cool, temperate climate and most of the region lies within a high rainfall zone (600-800 mm/yr) where approximately 60% of rainfall occurs during winter and spring, peaking normally in July and August (Gibbons and Downes 1964). The average rainfall at the field site is 625 mm. Mean monthly maximum temperatures range between 11.9 oC in July and 25.3 oC in January at the Pastoral and Veterinary Institute, 65 km south of the field site. Evaporation rates range from 36 to 214 mm/month respectively (data from the National Climate Centre at the Commonwealth Bureau of Meteorology, 1999). This leads to an overall moisture deficit during summer and surplus during winter (Figure 3-2). Streamflow is therefore strongly seasonal with two thirds of the average flow occurring between August and October (Wagg 1999b).

250

200

150

100 mm/month

50

0 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec

Figure 3-2: Long term (1962-1990) mean monthly rainfall (solid line) and pan evaporation (broken line) at the Pastoral and Veterinary Institute, Hamilton (National Climate Centre, Bureau of Meteorology, 1999)

3.3 Landuse and management

Dryland (i.e. rainfed) sheep and cattle pasture comprises 71% of the 2.6 million hectares of the Glenelg-Hopkins Catchment, which encompasses the south-west of Victoria (Figure 3-1). The Dundas Tablelands, where the field site was located, has a long history of Merino wool production (Gibbons and Downes 1964). Other important land uses in the Glenelg-Hopkins catchment are native forest (15%), dairy pasture (5%), remnant vegetation (3%) and cropping

44 (3%)(Holmes 2000). Softwood and Blue gum (Eucalyptus globulus) plantations are also of increasing importance, mainly in the south of the region (Wagg 1999a). The original vegetation of the Dundas Tablelands was savannah woodland consisting of sparsely scattered trees (predominantly Red Gum, Eucalyptus camaldulensis) and understorey grasses such as wallaby-grass (Danthonia spp.), kangaroo-grass (Themeda spp), spear-grasses (Stipa spp.) and tussocks (Poa australis) (Douglas and O'Brien 1971). In the 1940s, annual subterranean clover was introduced to improve the quality of fodder for livestock, and by 1965 60% of pastures were sown with this legume (Gibbons and Downes 1964). Phosphorus (P) fertiliser, as superphosphate, also became increasingly popular to overcome the P deficiencies of many soils. Increasingly, areas of native grasses receded to remnant bushland, and pastures became dominated by introduced annual grasses. The field site was established on grazing land which was sown with Australian phalaris (Phalaris aquatica cv. Australian) and subterranean clover (Trifolium subterranean cv. Trikkala) in 1994.

45 Figure 3-3: Vasey field site showing layout of the SGS experiment and the location of the runoff plots, soil pits (Pits 1-6) described by Cox et al. (1998) and soil pits used for measuring bulk density and hydraulic conductivity characteristics (Pits 7-11)

46 3.4 Geology of the Dundas Tablelands

The Dundas Tablelands are one of three distinct plateau regions in the south-west of Victoria. The tablelands are believed to have formed in the Paleozoic era 400 million years ago (Ma) on Silurian extrusions of granites and Rocklands Rhyolite lava. Grampians sedimentary sandstones also occur in the eastern tablelands (Simpson and Woodfull 1994; Dahlhaus and MacEwan 1997). Later, peneplanation, or erosion, occurred during the Mesozoic era resulting in dissection of the two main drainage lines – the Glenelg and Wannon Rivers (Gibbons and Downes 1964). Weathering continued into the Tertiary period to form the regolith in situ. Upon this weathered surface, accumulation and cementation of iron occurred in the Miocene and Pliocene periods (3 to 25 Ma), forming ferricrete (also termed laterite). In this period the rise and fall of watertables caused by nearby marine incursions provided weathering conditions conducive to ferruginisation (Douglas and O'Brien 1971; Dahlhaus and MacEwan 1997; Fawcett and Norton 2000). The final uplift of the Tablelands to their current elevation above sea level (approximately 240 m) occurred during the last 2-3 million years (Fawcett and Norton 2000). The weathering, erosion and wetting and drying processes which have dominated during the development of the Dundas soils led to the breakdown of much of the ferricrete to form the soil profile that exists today (Fawcett and Norton 2000). Wind erosion is uncommon for these soils, and water erosion is typically limited to some sheet erosion where groundcover is sparse.

3.5 Soil description

The major soil types at the Vasey field site were classified and described by Cox et al. (1998) according to methods described by Isbell (1996) and MacDonald et al. (1990). The soils on the tableland and midslope regions are ferric-sodic Yellow Sodosols and Brown Chromosols respectively. At the base of the slope, where there is greater watertable activity, a petroferric, salic Hydrosol was identified. Selected properties and descriptions for soils in these three topographic positions are summarized in Table 3-1. In brief, the existing soil profile consists of a sandy loam topsoil, a gravely loam A2 horizon, underlain by yellow, mottled light clay B2 and B22 horizons, and a red and white mottled, medium to heavy clay B23 which grades into a pallid, kaolin-rich horizon (B3) (Figure 3-4)(Gibbons and Downes 1964; Cox et al. 1998). An outcrop of the rhyolite parent material (C horizon) was evident at depth in the lower slope position.

47 Table 3-1: Classification and description of the major soil types at Dundas Park, Vasey (Cox et al. 1998)

DepthHorizon A TextureB Munsell Colour (moist) % Macropores Gravel Clay >2 mm cm Matrix Primary w/w % soil mottles weight Tableland Ferric, mottled-subnatric, Yellow Sodosol Pit 1C Sodic, non saline 0-5 A1 SL 5YR3/1 20 Common, <1 5 mm 5-25 A21 L 10YR4/2 10YR6/6 28 “ 15 25-50 B21 FSCL 10YR6/8 7.5YR5/6 59 “ 30 50-90 B22 CL 10YR4/6 10YR6/2 59 “ 60 90-140 B23 LMC 2.5YR3/6 10YR7/2 60 “ 60 7.5YR5/6 140-180 B23 MHC 10YR4/6 10YR6/2 63 “ 50 180-200 B23 HC 10YR4/6 10YR6/2 58 none 15 Midslope Ferric-sodic, eutrophic, Brown Chromosol Pit 2 Non sodic, non saline 0-5 A1 SL 10YR4/1 14 Many, <1mm 5 5-30 A21 L 10YR4/2 10YR6/6 16 Common, 30 <1mm 30-60 B21 CL 10YR5/6 10YR5/2 55 Few, <1 mm 25 60-130 B22 LMC 10YR5/6 10YR5/2 47 “ 20 130-180 B3 HC 10YR7/1 2.5YR4/4 46 None 10 180-200 B3 HC 10YR7/1 2.5YR4/4 33 None 5 Lowerslope Petroferric, salic, Hydrosol Pit 4 Sodic, saline 0-5 A1 SL 10YR3/2 12 Many, <1mm 4 5-20 A21 CS 10YR4/2 10 Many, 2-5mm 7 20-40 A22 CS 10YR4/2 23 Many,1-2mm 73 40-85 B21 LC 10YR5/4 10YR4/1 53 Few, <1mm 52 85-100 B22 HC 10YR5/4 2.5YR4/6 63 None 21 100-150 B23 HC 2.5YR4/5 10YR6/1 71 “ 8 150-200 B3 LC 2.5Y7/2 10YR5/8 45 “ 1 200-250 C CKS 10YR7/1 10YR5/8 “ 29 A Soil horizons were originally described by Cox et al. (1998) according to Soil Survey Staff (1996) nomenclature but were interpreted and reported here according to McDonald et al. (1990). B S=sand, SL=sandy loam, L=loam, C=clay, HC=heavy clay, MC=medium clay, LC= light clay, CS=clayey sand, CKS= coarse, clayey sand, FSCL=fine sandy clay loam C Soil pit locations are shown in Figure 3-3

48 Figure 3-4: Soil profile of a Brown Chromosol in a midslope position at Vasey

There are several geomorphic and physical features of the Vasey soils that indicate a potential for waterlogging and lateral throughflow and surface flow during winter. These include an accumulation of ironstone gravel at the top of the B horizon, the duplex profile, and the red mottling in the B horizons (Fawcett and Norton 2000). Soils of the Dundas Tablelands typically contain large amounts of ironstone gravel of between 6 and 20 mm diameter (Gibbons and Downes 1964). Both magnetic and non-magnetic types occur, with the cementing material being dominated by maghemite and hematite for the former and goethite for the latter (Brouwer and Anderson 2000). On the plateau area of the research site this gravel was prominent throughout the B horizon, but on the midslope, and particularly the lowerslope profiles, the gravel accumulated in the A22 (up to 70% of the soil weight) and B21 horizons (Table 3-1). The gravel is thought to be remnant of weathering of the previously more extensive ferricrete (Dahlhaus and MacEwan 1997; Fawcett and Norton 2000) and indicates a history of wetting and drying above and within the B horizon. The soils are duplex in nature as there is a marked increase in the clay content of the B horizon (35-70%) when compared with the A horizon (10-28%)(Table 3-1). The depletion of clay and concentration of ironstone gravel in the upper horizons is due to selective illuviation of smaller particles into lower horizons during very wet conditions (Fitzpatrick 1999; Fawcett and Norton 2000). Similarly, the A2 horizon is bleached grey when compared with the upper organic zone as it has been leached of organic matter and cations. This contrast in texture may

49 have the potential to restrict vertical drainage and facilitate the development of temporary perched watertables above the heavier clay B horizons, resulting in lateral subsurface and surface flows in sloped areas (Cox and McFarlane 1995; Cox et al. 1998). Also characteristic of frequent waterlogging are the predominantly prismatic clay peds that occurred in the B horizons of pits 2 and 4 (White 1997). Mottling occurred in the B2 horizons, indicating a history of non-uniform and frequent waterlogging whereby anoxic conditions have developed in structural cracks, causing reduction and redistribution of the red ferric oxides resulting in a beached appearance (Dahlhaus and MacEwan 1997). During unsaturated (oxidising) conditions, this iron was deposited in pockets of soil that have retained the colouration. The mottling was strongest in the B23 horizon. It is likely that seasonal perched watertables developed above the mottled, heavy clay B22 and B23 horizons. This is consistent with the interpretation of the geomorphology of a nearby soil by Brouwer and Fitzpatrick (2002b) where fresh water perched within, rather than above the B- horizon. The degree of mottling would have been limited by the duration of waterlogging (Brouwer and Fitzpatrick 1998), which can be strongly influenced by the presence of macropores that provide preferential flow pathways for drainage. In the Vasey soils, dark ribbons were observed and thought to have formed by translocation of clay along what more recently may have been old root channels (Fawcett and Norton 2000). This feature indicates there is the potential for preferential vertical flow in these soils. However, visual evidence of currently active macropores was limited to more than 5 very fine to fine (i.e. <2 mm in diameter) pores per cm2, and in the A2 horizon of the lowerslope profile, there were more than 5 medium width (i.e. 2-5 mm diameter) pores per 100 cm2 (Table 3-1). The potential for preferential vertical flow would therefore still be less than that for cracking clays, or duplex soils such as those in the Adelaide Hills where macropores of 12 mm diameter were observed (Smettem 1987). The presence of some vertical preferential flow pathways would be consistent with the conceptual model of water movement in these soils presented by Dahlhaus and MacEwan (1997) whereby rainfall quickly fills preferential pathways and perched watertables subsequently develop in the upper soil profile. The light grey coloration of the B3 horizon indicates that more uniform depletion of iron deposits has occurred at that depth and that prolonged waterlogging occurred above the C horizon, probably due to the seasonal rise of groundwater tables (Cox et al. 1998). Brouwer and Fitzpatrick (1998) found outcroppings of this kaolinised parent material near the soil surface in midslope positions of nearby soils, where it provided a layer that restricted vertical infiltration and caused perched watertables and surface waterlogging to develop. This demonstrated that surface waterlogging was not limited to the lowerslope regions of the toposequence.

50 A sketch of the variation in soil horizon depth along the toposequence of each runoff plot at Vasey was developed from soil pits and from soil cores taken for piezometer and neutron moisture meter access tube holes Figure 3-5). Plots 2, 3 and 4 show a reduction in the depth to the C horizon towards the base of the slope, whereas the depth of the A horizon tended to increase towards the base of the slope in each plot. Gully incision during the Pleistocene era may have created a shallower depth to the parent material downslope compared to upslope (Dahlhaus and MacEwan 1997), and the increasing depth of the topsoil toward the base of the slope in most plots would be explained by downslope accumulation of eroded particles during the more recent history of the landscape.

51 a) Plot 1 b) Plot 2 260 260

258 258

256 256 level (m)level 254 level (m) 254

252 252 Elevation/head sea above Elevation/head sea above 250 250 0 20 40 60 80 100 120 140 0 20 40 60 80 100 120 140 Horizontal slope distance (m) Horizontal slope distance (m)

c) Plot 3 d) Plot 4 260 260 258 258 256 256

254 254 sea (m) level sea (m) level

252 Elevation/head above 252 Elevation/head above

250 250 0 20 40 60 80 100 120 140 0 20 40 60 80 100 120 140 Horizontal slope distance (m) Horizontal slope distance (m)

Figure 3-5: Approximate depth to the B and C horizons along the toposequence of the hillslope runoff plots at Vasey. Dashed lines represent B and C horizons.

52 3.6 Soil physical properties

3.6.1 Bulk density

Measurement method

Bulk density (BD) was measured by weighing a dried (105oC, 24 hrs) known volume of soil. Soil cores were extracted from soil pit faces (Figure 3-3) using metal rings of diameter 75 mm and length 55 mm (volume 242 cm3). Samples were taken in October and November 1998. Cores taken during November from depths greater than 100 cm were not used for BD analyses as they were slightly shattered, which caused inaccurate measurements of the volume of soil. The average BD of each depth and horizon was calculated from a combination of these data and data reported by Cox et al. (1998). The latter were derived from soil cores of 50 mm diameter and length (volume 98.2 cm3).

Results and discussion

Table 3-2: Mean, standard error and number of soil bulk density (g/cm3) measurements for major soil horizons at Vasey

Depth Horizon BD SE n cm g/cm3 0 A1 1.15 0.04 15 10 A21 1.59 0.02 4 20 A21 1.54 0.03 13 30 - 50 A22 1.81 N/a 1 40 B21 1.48 0.05 12 60 B22 1.33 0.04 8 80 -120 B22 1.43 0.06 6 140 - 150 B3 1.63 0.05 5

The BD was lowest in the A1 horizon (Table 3-2). This was expected due to an abundance of plants roots and organic matter, which are less dense than soil minerals. The highest BD was in the A22 horizon (1.81g/cm3) that occurred only in Pit 11. The high BD was attributed to an accumulation of high-density gravel, which contributed up to 72% of the volume of the soil of the A22 horizon in this topographic position (Cox et al. 1998). Cox et al. (1998) also found evidence for BD increasing with gravel content in Vasey soils. Brouwer and Anderson (2000) found that the majority of gravel in a nearby soil at a depth comparable to the A22 horizon described above was a non-magnetic type (i.e. that dominated by goethite cementing materials) that had a particle density of between 2.3 and 2.9 g/cm3.

53 3.6.2 Hydraulic conductivity

Measurement method

The saturated hydraulic conductivity (Ksat) of the major soil horizons at the Vasey site was measured using a CSIRO 200 mm diameter disc permeameter (Perroux and White 1988). A supply potential of -20 mm was used to reduce the effects of macropores greater than 1.5 mm diameter, as per White and Ridley (1998). Measurements were taken on ledges within soil pits at two tableland positions (Pits 7 and 8, Figure 3-3) and one each at a topslope, midslope and toeslope position (Pits 9,10 and 11, Figure 3-3).

Ksat was initially calculated using a combined approximation (equation 3-1, (Smettem and Clothier 1989; White et al. 1989)) that was derived from the sorptivity estimation of Philip (1986) and Wooding’s (1968) estimation of steady state infiltration from a ponded disc;

2 2 Ko = q /πr − 4bSo /πr(θ f −θi ) (3-1)

2 where Ko is the hydraulic conductivity, q/πr is the water flux (steady state infiltration divided by the disc area), b is a capillary scale parameter approximated as 0.55 for field soils, So is the sorptivity, r is the radius of the supply disc, and θf and θi are the final and initial soil volumetric water contents. The sorptivity is a measure of the capillarity of the soil and is obtained from the slope of the relationship between cumulative infiltration and the square root of time using the early time data.

The value of Ko calculated using this method is sensitive to the difference between the initial and final volumetric water contents, and consequently many values of Ko were negative. This unrealistic response was caused by the field measurements being completed in October 1998 when the soils were quite wet, and was exacerbated by the difficulty in obtaining true samples of the saturated soil. This method is also sensitive to overestimated values of So, a parameter which is also difficult to measure accurately on wet soils.

An alternative method for deriving Ko was therefore investigated. Acknowledging the weakness of the previous technique when used for wet soils, Smettem and Clothier (1989) recommended a method that used only the steady state infiltration rates, from discs of two or more different radii (White et al. 1989) to solve for Ko and So. Their method may have been preferable because it does not rely on estimates of So, which contributes little to the Ko for wet soils. In the Smettem and Clothier (1989) method, soil moisture data are still required, but their contribution to Ko diminishes as the initial soil water content increases. Within the constraints of the data that were available from the single disc measurements, however, an approximation developed by R.E. White (pers.comm.) that is independent of soil water contents was used. A predetermined value of the macroscopic

54 capillary length λc (cm) was used to reduce equation 3-1 to Wooding’s (1968) steady state inflow equation;

Q /πr = ∆Ko (1+ 4λc /πr) (3-2)

Okom (1998) measured the Ksat of a range of soil textures in Victoria and derived λc and its inverse α (i.e. 1/λc (1/cm)) according to the following equation (White and Sully 1987),

2 λc = bSo /(θ f −θi )Ko (3-3)

Although λc is expected to increase with increasing clay content, Okom (1998) found no such relationship and instead found a single mean value for α of 0.2/cm was a satisfactory surrogate variable for all the soils. Alpha (α) was calculated from equation 3-3 for the Vasey soils using only the realistic values of Ko that were derived from equation 3-1. The range of α values found within and across texturally contrasting horizons is presented in Table 3-3.

Table 3-3: Alpha values (1/cm) for the major soil horizons at Vasey

Horizon α (1/cm) Mean SE A1 0.2 0.9 0.5 0.6 0.5 1.2 0.1 0.5 0.6 0.1 A21 0.6 0.6 N/a A22 <0.1 0.2 0.1 0.1 B21 <0.1 1.9 0.3 0.3 0.2 0.2 0.8 0.5 0.2 B22 0.3 0.6 0.3 0.4 0.1 C0.2 0.2N/a Pooled mean 0.5

Similar to the findings of Okom (1998), there was no clear relationship between clay content and α and there were no significant differences (P<0.05) between the horizon means.

Αn overall mean value of 0.5/cm was therefore substituted into equation 3-2 and Ksat at -20 mm supply potential was calculated for the entire dataset.

55 Results and discussion

Summary statistics for the approximated values of Ksat for five major soil horizons at

Vasey are shown in Table 3-4. There were insufficient data to test for differences in Ksat between topographic positions so the horizon data were pooled across the five soil pits.

Table 3-4: Summary statistics for approximated values of Ksat (m/day) of major soil horizons.

Horizon n MeanA SE Minimum Maximum A1 10 0.31a 0.05 0.15 0.55 A21 4 0.10bd 0.04 0.04 0.20 A22 2 0.30ac 0.12 0.18 0.42 B21 10 0.14bc 0.02 0.08 0.24 B22 4 0.13bc 0.02 0.10 0.19 C 2 0.06d 0.04 0.02 0.10 A Subscript notation denotes differences between horizon means at the 95% confidence level based on a multiple t-test.

The mean Ksat of the B21 and B22 horizons was about half that of the A1 horizon (Table 3-4). This may be explained by the higher clay content (Table 3-1) and BD of B horizon soil when compared to the A1 horizon. A higher clay content infers a greater frequency of smaller radius pores, which require more suction to be drained of water and therefore provide a greater resistance to the flow of water through the soil matrix (White 1997). However, the Ksat of the B21 horizon (mean, 0.14 m/day) was high relative to those measured on other duplex soils. For example, for B horizon soils near Keyneton, South Australia, J. Cox (pers.comm) found the mean Ksat at a potential of –10 mm was 0.069 m/day, and White et al. (2000) recorded a Ksat of

0.05 m/day for the B horizon of a duplex soil near Wagga Wagga, New South Wales. The Ksat of the A1 horizons were similar for all three locations and soil types (range 0.31 to 0.53 m/day). The -20 mm supply potential used in the current study should have eliminated the contribution of pores greater than about 1.5 mm in diameter to the permeability of the soil, so the relatively high Ksat for the B horizon at Vasey might be explained by the presence of a moderate network of macropores smaller than this diameter (Table 3-1).

The Ksat of the A21 horizon was significantly (P<0.05) lower than that of the horizons immediately above and below it. This is consistent with the higher BD (1.56 g/cm3) and massive structure of the A21 horizon compared to the A1 horizon. J. Cox (pers.comm.) also found evidence for Ko decreasing with BD.

An inverse relationship between BD and Ksat does not explain the high Ksat measured in the A22 horizon, as it also had a very high BD (1.8 g/cm3, Table 3-2). The magnitude of both properties may, however, be attributed to the abundance of ironstone gravel in this horizon.

Cracks may develop around the gravel and increase the Ksat (Dahlhaus and MacEwan 1997),

56 whereas, as previously mentioned, the high particle density of ironstone gravel confers an overall increase in the BD.

From the Ksat of soil horizons we can infer a propensity for waterlogging and the development of perched watertables. In sloped areas, perched watertables can contribute to the generation of surface runoff by causing waterlogging of the soil surface and the re-emergence of throughflow at the surface in other parts of the landscape. Ninety-six percent of the daily rainfall that has occurred in this region since 1962 was less than 40 mm/day, with the largest recorded daily total being 61.2 mm/day (Source: Commonwealth Bureau of Meteorology, Australia).

This daily volume is less than the Ksat of all except the C horizon in the current soil, which suggests that perched watertables would not be expected to develop in these soils. Transient watertables occurred in the A horizon of the previously mentioned Wagga Wagga soils where there was a 6-fold difference between Ksat of the A and B horizons (White et al. 2000). Similarly, Stevens et al. (1999) only found subsurface lateral flow to be the dominant pathway -5 for nutrient movement in soils where the subsoil Ksat was less than 3.3 x 10 m/day, and the topsoil was highly permeable (Ksat(-10mm) >2.6± 1.0 m/day). Perched watertables were also a feature in the upper soil profile of the Keyneton soil and were attributed to low Ksat in the B and C horizons in particular topographic positions (J.Cox, pers.comm.). In the latter study, the C horizon soils had a similar Ksat to that of the present soils, which suggests some perching of water may be expected to occur above this horizon. J. Cox (pers.comm.) suggested drainage in the C horizon was dependent on the degree of weathering of the saprolite material and that less weathered and hence more dense material is expected in upper slope rather than downslope positions. These results concur with the geomorphological interpretation that the A/B horizon interface would rarely restrict vertical drainage in these soils and that perched watertables and lateral flows might be expected to develop deeper in the profile during very wet conditions.

3.6.3 Soil water retention characteristics

Measurement method

Quadruplicate soil cores (55 mm length by 75 mm diameter) of the major soil horizons were collected from the vertical faces of five pits in late October and November 1998. Core rings were hammered into the soil parallel to the soil surface except for the surface cores, which were inserted perpendicularly and the grass trimmed flush with the soil. Some peds in the B horizons shattered during the collection process. This would have reduced the accuracy of BD and volumetric water content (θv) measurements. Cores were wrapped airtight and kept cool prior to analysis.

57 The soil water content of the cores was measured at matric potentials (Ψ) of -0.5, -1, -5, -10, -100, -1000, and –1500 kPa using porous ceramic plates, according to methods outlined by Cresswell and Smiles (1995). The soil core surface used to make contact with the ceramic plate was bound in mesh to prevent soil falling out, then the cores were soaked in water until saturated. Suctions up to 10 kPa were applied using a ‘hanging’ water column and greater suctions were applied in a pressure chamber using compressed air with a pressure regulator.

Suction at each level was applied until the gravimetric water content (θg) reached equilibrium. 3 3 The saturated volumetric water content (θsat, cm /cm ), where Ψ = 0, was assumed to be equal to the total soil pore space and was calculated from the BD assuming the matrix density was 2.65 g/cm3, equivalent to that of quartz (Cox et al. 1998);

BD θ = 1− (3-4) sat 2.65

Results and discussion

Table 3-5: Volumetric water contents (cm3/cm3) of major soil horizons for a range of matric potentials (kPa)

Horizon A1 A21 A22A B21 B22 & 23 B3 Depth (cm) 0-5 5-20 20-40 40-85 85-150 150+ BD (g/cm3) 1.15 1.57 1.81 1.48 1.38 1.63 3 3 Ψ (kPa) ------θv (cm /cm ) ------0 (θsat) 0.566 0.409 0.317 0.442 0.480 0.385 -0.5 0.383 0.348 0.414 0.399 0.440 0.416 -1 0.367 0.317 0.394 0.393 0.428 0.410 -5 0.335 0.287 0.379 0.375 0.415 0.403 -10 (FCB) 0.308 0.268 0.363 0.365 0.404 0.395 -100 0.281 0.232 0.298 0.313 0.392 0.385 -1000 0.247 0.174 0.203 0.276 0.354 0.375 -1500 0.229 0.156 0.182 0.247 0.327 0.358 AWSCC 0.079 0.111 0.180 0.118 0.076 0.036 Field FCD 0.34E 0.26 N/A 0.36 0.38 0.39 A An A22 horizon was not present in all soil profiles positions B FC; Field capacity C AWSC; Available water storage capacity: water retained between –10 kPa and –1500 kPa D Mean volumetric water content for wettest soil profiles as measured in the field using a neutron probe (see Chapter 4) E M. McCaskill (unpublished data) N/A; not available

The available water storage capacity (AWSC, mm/dm) increased from 7.9 in the sandy loam topsoil to 18 in the massive clay loam A22 horizon, then decreased again to 3.6 in the lower B horizon. The AWSC across a range of textures is usually between 12 and 16 mm/dm in Australian soils, with extremes being measured from 9.6 to 26.5 mm/dm (Graecen and Williams 1983). The values measured at the field site were therefore lower than expected.

58 Although soil structure, rather than texture, often has the greatest influence over the water holding capacity, medium textured soils (loams) are generally found to have the largest AWSC (Graecen and Williams 1983). The high AWSC measured in the clay loam A22 horizon may therefore have been attributed to the loam component, although its massive structure and high gravel content were likely to reduce the water holding capacity when compared with a well structured, gravel-free loam. Fine textured subsoils retain more water than sandy horizons at an equivalent suction due to a predominance of micropores (White 1997). This was reflected by the lower volumetric water contents of A horizon soils, particularly the A21 horizon, at all matric potentials when compared to the deeper horizons. The moderate AWSC of the B21 horizon again probably reflects its medium texture (light-medium clay), with the AWSC reducing as the proportion of clay increased with depth. The greater suctions required for extraction of water from clay than sandy textured soils means water is less available for plant growth. There was good agreement between the volumetric water content at field capacity (-10 kPa) derived from the moisture retention data at –10 kPa, and from neutron probe measurements of soil water content during the wettest times of the year (Table 3-5). This gives us confidence in drawing conclusions about the hydrological characteristics of the soil from data obtained by both methods.

3.7 Soil chemistry

Chemical characteristics of soil horizons at the Vasey field site are reported by Cox et al. (1998) and summarized here. The soils are generally acid with the pH (0.01M CaCl2) lowest in both the A horizon (5.5) and at greater than 1 m depth (4.1 - 6.1)(Cox et al. 1998). Total carbon was high in the A1 horizon (2.4 - 7.8%), variable in the A2 horizon (0.45- 3.1%) and very low in the B horizons (0.06 – 0.29%). A P sorption index (Rayment and Higginson 1992) indicated the relative P sorption capacities of the soil horizons. The index measured the percentage of P in an initial solution of 150 mg P/kg that was sorbed by the soil. The lowest P sorption index (6%) occurred in the A1 horizon of the lowerslope soil profile (Cox et al. 1998). Other A horizon soils adsorbed up to 44% of solution P. In the B2 and B3 horizons, up to 100% of solution P was adsorbed by the soils. This was not surprising considering the acidic nature and high kaolinitic clay and sesquioxide content of subsoils in this region (Cox et al. 1998). Kaolinite minerals have a large edge surface which becomes protonated and positively charged under acidic conditions. Sesquioxide minerals also have high variable charge surface areas that infer a high anion - 2- exchange capacity up to pH 8. Solution P (H2PO4 or HPO4 ) is specifically adsorbed to positive charge surfaces on these minerals and also forms precipitates with iron and aluminium oxyhydroxides.

59 The Olsen P (POlsen) of the surface (0-5 cm) soil ranged from 21 to 31 mg/kg and decreased to less than 4 mg/kg soil in the B horizon (Cox et al. 1998). This reflected a history of fertiliser applications to the soil surface and the increasing clay and sesquioxide content with depth. These characteristics suggested that most P in drainage water would be retained within the soil profile. The soils were typical of the highly weathered soils on the Dundas Tablelands where the high P sorption capacity and low natural abundance of P reduces the availability of P to plants. Fertiliser P is commonly applied to overcome P deficiencies for plant growth. Salts (particularly sodium salts) have accumulated in the rhyolite parent material due to marine incursions in the Miocene and Pliocene, extensive weathering and contributions of windblown and cyclic salt (Dahlhaus et al. 2000). Consequently the deep regolith contains up to 500 t/ha of salt (Brouwer and Fitzpatrick 2002b) and saline discharges are a natural occurrence in the region. It was originally perceived that exacerbation of surface salinity was due to the removal of deep-rooted perennial vegetation by European settlers leading to rising groundwater tables. More recently, however, it has been suggested that it is the duration of waterlogging that has increased due to lower plant water use, and consequently saline lateral flows near the soil surface have spread and increased the area of salt affected land (Dahlhaus and MacEwan 1997). The exchangeable sodium percentage (ESP), as a measure of the sodium concentration of the soil, increased with depth. However, the soil pits near the midslope positions of the hillslope runoff site were non-saline and non-sodic (Cox et al. 1998). The ESP in the B horizon of the lowerslope position, however, was very high (approximately 20%), probably due to a seasonal rise in the groundwater table. The sodic and dispersive nature of this soil horizon was likely to reduce its Ksat.

3.8 The potential for movement of P in surplus water

The topography, land use and soil type at the study site suggested that surface runoff, subsurface lateral flows and deep drainage were all potential hydrological pathways for P losses at Vasey. Surface runoff was likely to be the dominant pathway responsible for P loss because of the accumulation of soil P-rich sources such as fertiliser, faecal matter and plant litter at the soil surface. P concentrations in subsurface lateral and vertical flows were expected to be low, because the high clay and iron content of the acidic subsoil suggested most soil solution P would be adsorbed. These findings suggested the primary focus of the field experiment in this study should be to measure the quality and quantity of surface runoff water from pastures.

60 3.9 Experimental design of the hillslope runoff site

To measure runoff and nutrient movement that could be considered representative of the farm system, an experiment was set up to measure runoff at a hillslope scale. A 2 ha sloping area at the Vasey research site (Figure 3-3) was surveyed and fenced into four similarly shaped 0.5 ha paddocks Figure 3-6). The site had a generally north-easterly aspect with the area designated as plot 1 facing east. All plots had an average slope of 5%. Plots 1 and 4 were subdivided down the slope and fenced into four sections of equal area for a rotational grazing treatment. To investigate the effects of increasing fertiliser application and sheep stocking rates on P losses in surface runoff, each hillslope plot was allocated one of three treatments that were concurrently being studied at the adjacent SGS grazing experiment. Plots 2 and 3 were managed with set-stocked grazing using low and high P application rates respectively (treatments A and B in the SGS experiment), and plots 1 and 4 were managed with a high P application rate with a four-paddock rotational grazing system (treatment C in the SGS experiment). The treatments were randomly allocated to the four plots but due to the constraints on the size of the plots only one treatment was replicated. Treatment C was replicated as it reflected the most intensive grazing system of the three treatments, and rotational grazing practices are being increasingly adopted by producers in the region. The runoff plots were grazed by lambing Merino ewes. Sheep in the four-paddock rotation were moved between sections every two weeks, except in spring when they were moved weekly. Supplementary grain was fed in summer and autumn to maintain ewes above a condition score of 2.0. Stocking rates were adjusted 3-4 times per year to ensure pasture availability and stock live-weights were as close as possible to that of the equivalent treatments in the adjacent SGS grazing experiment. Stock numbers ranged from 7 breeding ewes in the low fertility set-stocked paddock to 11 breeding ewes in the higher fertility rotationally grazed plots. These stocking rates correspond to approximately 20 and 31 dry sheep equivalents/ha on an annual basis, assuming ewes carried single lambs (McClaren 1997).

61 Figure 3-6: Layout of the runoff plots at Vasey showing elevation contours and positions of the neutron probe access tubes, nests of piezometers and ceramic cup samplers and permanent soil sampling points

62 P fertiliser was applied as either single superphosphate (8.8%P, 11%S), double superphosphate (16.8%P, 4%S) or PasturephosK23® (8%P, 23.3%K, 3.8%S) as per equivalent treatments in the SGS experiment (Chapman et al. 2003). The rates of P applied to the four plots between 1997 and 2000 are shown in Table 3-6. The P application rates in the ‘high’ P treatments were aimed at maintaining POlsen in the range 14 – 16 mg/kg and the low P rate was aimed at maintaining POlsen at between 4 and 6 mg/kg (Chapman et al. 2003). The POlsen (0-10 cm) measured in November 1997 across the runoff site area ranged between 6 and 9 mg P/kg. In 1997, before the site was designated for the current research, urea (46%N) was applied to plot 4 and most of plot 3 at 100 kg N/ ha.

Table 3-6: P fertiliser application rates at the Vasey runoff site

Plot Treatment June 1997 Feb/June 1998 March 1999 March 2000 ------kg P/ ha ------1 High P, rotation 0 80 25 25 2 Low P, set stocked 0 8 8 6 3 High P, set stocked 30 50 25 25 4 High P, rotation 30 50 25 25

Surface runoff from each hillslope plot was measured from August 1998 until December 2000. The runoff plots spanned the entire slope length in order to allow landscape- scale soil movement and runoff processes to occur and to minimise any disturbance to the behaviour of livestock. Allowing grazing to occur over the entire slope length in set stocked plots was considered important as the development of stock camps on hillslopes can influence the transfer of nutrients and the amount of groundcover in different topographic positions (Gillingham 1983) and may in turn influence the quality of runoff. The plots were assumed to be large enough to represent the variability in soil and pasture characteristics likely to be encountered in paddocks of a commercial scale as well as cater for realistic stock numbers, yet small enough to allow the effects of contrasting treatments on the quality and quantity of runoff to be investigated across an area of relatively uniform soil type, topography and climate. The volume and pathways of surplus water movement measured at the field site are discussed in Chapter 4, and P losses via these pathways are discussed in Chapters 5 and 6.

63 CHAPTER FOUR

4 Hydrological characteristics of the Vasey runoff site

4.1 Introduction

Water plays a key role in the removal of phosphorus (P) and other nutrients from agricultural land, because it provides the energy and the carrier mechanism for transport processes (Haygarth et al. 2000; Heathwaite and Dils 2000). An understanding of the hydrological pathways and processes occurring in agricultural landscapes, and their spatial and temporal controls, is needed to assess the potential for P to move into the wider environment. On hillslopes, water that is not stored within the soil profile or evaporated is partitioned into surface and/or subsurface vertical and lateral flow pathways (Chorley 1978). For clarity, the terms that will be used in this thesis to describe hydrological pathways and processes are described in Table 4-1 and are consistent with definitions outlined by Haygarth et al. (2000). Most hydrological pathways may eventually connect agricultural land to surface waterbodies, however, unless large natural or artificial preferential flow pathways exist for removing excess water via subsurface pathways, surface pathways generally provide the quickest route. The pathways of water movement above and through the soil help identify the sources of P with which the water interacts and the rates of flow determine the amount and duration of contact between the P source and water. Together with the spatial and temporal distribution of water movement, these hydrological characteristics can help identify the areas and management practices with the most potential for P loss.

Table 4-1: Terminology and definitions of hydrological pathways and processes

Pathway/Process Definition Surface runoff Downslope movement of water exclusively over the soil surface Subsurface lateral flow Lateral flows below the soil surface Subsurface vertical flow Downward vertical flow under saturated or unsaturated conditions Preferential flow Vertical movement along larger subsoil pathways such as macropores, old root channels and structural cracks – often occurs in unsaturated conditions Return flow Where a subsurface pathway emerges at the soil surface

Three main mechanisms of surface runoff generation can be described. These are saturation excess flow, infiltration excess flow and subsurface baseflow. Saturation excess

64 runoff is produced from rain falling onto saturated land. These areas become waterlogged by discharge of perched or groundwater, or from return flow. Infiltration excess runoff occurs when the infiltration capacity of the soil is exceeded by the rainfall intensity and can occur across large tracts of land. Surface baseflow is defined in this thesis as surface runoff that has no hydrograph peak and occurs in the absence of rain. This type of flow occurs when subsurface lateral flow or groundwater discharge emerges at the soil surface and becomes return flow. The rainfall intensity, infiltration capacity and drainage properties of soils are likely to influence the processes by which runoff is produced, and the areal extent from which surface flows are generated. In any catchment it is likely that more than one runoff producing mechanism will occur. The area contributing to runoff varies throughout and between storms and is commonly referred to as the variable source area (VSA). Runoff source areas require particular attention regarding nutrient and erosion management. In combination with the knowledge of the potential for P to be mobilised, hydrological characteristics can be used to target where and when appropriate nutrient management strategies should be applied. The aim of the experimental work reported in this chapter was to identify the spatial and temporal behaviour of water movement and processes of runoff generation in the hillslope site at Vasey. The hydrological characteristics were then used to derive a conceptual model for the mechanisms controlling water movement at the field site and hence to identify the potential for P movement from the pastures.

4.2 Materials and methods

Surface runoff and other hydrological properties were measured from four 0.5 ha hillslope pasture plots at Vasey, in south-west Victoria between August 1998 and December 2000. The field site location, climate, soil type and experimental design were described in Chapter 3. In this section, the hydrological instrumentation and measurements are described.

4.2.1 Rainfall

Rainfall was measured using two automatic tipping bucket rain gauges, which were logged continuously from August 1998 until February 2001 by a CR10X Campbell Scientific® datalogger. One gauge measured rain in 0.1 mm increments and the other in 0.25 mm increments. Falls were logged each minute when they occurred, and hourly and daily totals were calculated and recorded. Rainfall was also measured manually using an incremented 10 cm diameter cylindrical rain gauge that was installed 30 cm above ground level in January 1999. Total rainfall was recorded whenever the runoff site was visited.

65 4.2.2 Evapotranspiration

Potential evapotranspiration (ETp) was calculated using the Priestley-Taylor equation (Priestley and Taylor 1972). Net radiation and wind speed were measured using a net radiometer and anemometer. Daily maximum and minimum air temperature and relative humidity were also measured at 10 s intervals and hourly means were logged. These data were recorded by an automated weather station, which was used for the SGS NE, and located 150 m from the runoff site.

4.2.3 Surface runoff

Polyethylene surface barriers (4-6 mm width) were inserted to a depth of 7 cm along the boundaries of four 0.5 ha pasture plots to prevent surface flow between plots and direct surface runoff into a collection area at the base of the hillslope (Figure 4-1). A reverse interceptor drain (Cox et al. 1994) was constructed across the top of the plots to prevent runoff from elsewhere in the landscape entering the plots. Steel tipping bucket gauges with 20 L capacity (Edwards et al. 1974) were installed beneath flow regulators at the base of each plot to measure the flow rate (Figure 4-2). The volume of surface runoff was measured using a magnetic counter and reed switch, which were triggered by each bucket tip. The reed switch sent a pulse signal to the Campbell Scientific® CR10X datalogger, which logged the number of tips per minute and calculated the volume per minute, and hourly and daily totals. The volume of flow collected by the buckets increased as the tip rate increased. A relationship between the tip rate and the volume per tip was supplied by the manufacturer ( Department of Natural Resources) for each bucket. The calibration was later adjusted with volumes measured at low flow rates in the field (Appendix A). The measured flow rate was not considered reliable by the manufacturer beyond a bucket tip rate of approximately 50 tips per minute. Hydrographs of flow from each plot were constructed using the volume per minute flow data. The magnetic counter was read manually and gave a useful backup measure of the cumulative volume of runoff.

66 Figure 4-1: Hillslope runoff plots at Vasey

Figure 4-2: Tipping bucket flow meters and sample splitters used to measure and sub-sample surface runoff at Vasey

67 4.2.4 Surface soil water content

The volumetric water content (θv) of the soil surface (0-12 cm) was measured along 3 transects in plot 1, and two transects in each of plots 2, 3 and 4, that extended upslope from the base of each plot. θv was measured every metre over the lower 10 m of the transect and every 2 m thereafter, for at least 30 m, using a Campbell Scientific Hydrosense™ portable frequency domain reflectometer (FDR). Measurements were made on several occasions between runoff events over the period 29 June - 29 November 2000. The FDR was calibrated against actual θv, calculated from the gravimetric water content (θg) and bulk density (BD) of 8 soil cores (12 cm deep x 3 cm diameter).

In addition to this, the θv of soil samples (0-10 cm) taken from a set of permanent sampling points in the autumn and spring of 1999 and 2000 was directly measured (θg × BD). The spatial distribution of surface soil water was then illustrated using Surfer® software, which used a kriging technique to interpolate gradients of θv between the sampling points.

4.2.5 Soil profile water content

The θv of soil profiles of the runoff plots was measured using the neutron moderation method (Gardner and Kirkham 1952). Four aluminium access tubes (internal diameter 41 mm) were installed along a central transect up the slope of each of the four plots (Figure 3-6). Ten tubes were installed to a depth of 1.8 m. The maximum depth of the remaining six tubes was restricted to between 0.6 and 1.6 m by a hard, saprolite layer. θv was measured at depths of 0.15, 0.25, 0.35, 0.45 0.6, 0.8, 1.0, 1.2, 1.4, 1.6 and 1.8 m at monthly intervals, increasing to fortnightly intervals during winter and spring. Neutron counts over a 16 second interval were recorded at each depth using a Boart Longyear CPN 503DR neutron-emitting probe.

Probe calibrations were performed using the measured θv of soil cores from 12 positions across the Vasey SGS and hillslope runoff experimental sites in October 1998 and February 1999. Soil cores were extracted using a combination of methods (Corbeels et al. 1999) including removal of cores during access tube installation and destructive sampling from around existing tubes using vertical cores (mechanically obtained) or short cores (obtained by hand) from a nearby pit face. Guidelines in the SGS NE Water Protocol were followed where possible (White and Ridley 1998). Vertical 41mm diameter soil cores were cut into 20 cm intervals starting at 10 cm depth and stored in airtight plastic bags. The top 10 cm was discarded. Cores were cut into further segments where sharp texture contrasts appeared. Pit face cores were 75 mm in diameter, 55 mm in length and were taken at the same depths as probe readings were taken. Soil moisture was measured using gravimetric analysis of soil core samples. Duplicate 16-second probe counts were read simultaneously with core extraction at 20 cm intervals from 20 cm to 260 cm depth where possible.

68 Regression analysis (Lawes Agricultural Trust 1997) demonstrated that a log-log linear model was appropriate to describe the relationship between θv, and a neutron count ratio (CR), for all soil depths combined (Figure 4-3). CR is the ratio of the actual neutron count relative to an average count read under standard conditions (40 cm depth in a drum of water) (Graecen et al. 1981). Standard counts were made each day the probe was used to account for drift in the neutron emission rate and detector sensitivity. Logarithmic transformation of both the independent and dependent data was required to normalise the distribution of residuals, which increased as θv increased. ) 3 0.60 /cm 3 0.50

0.40

0.30

0.20

0.10

0.00 Volumetric soil water content (cm 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 Count ratio

Figure 4-3: Volumetric water content (θv, y axis) plotted against the count ratio (CR, x axis), and 2 the fitted calibration curve; lnθv = 1.50 lnCR - 0.52, R =0.775, P<0.001

Degree of soil profile saturation

The θv of the soil at each measured depth was expressed as a percentage of θsat (defined as the total porosity and calculated from bulk density, see equation 3-4) for that depth. Surfer® mapping software (Anon. 1995) was then used to illustrate the degree of saturation of the soil at each depth in each tube position during the measurement period.

Soil water deficit

The soil water deficit (SWD) was defined as the difference between the actual stored water and the stored water at the wettest times of each year. The amount of water stored (mm) in the soil profile was calculated from the cumulative θv measured within the rootzone. The

69 maximum depth to which roots extracted water was estimated as 1.5 m because there was little change in water content over time beyond this depth. A field-derived field capacity was calculated for each tube from the mean of the four wettest profiles recorded over the three years of measurement. The mean encompassed some readings when the profile may not have been completely drained, and others when some evaporative losses had occurred. The SWD was assumed to be zero when soils were at this field-derived moisture level. The SWD for each neutron probe access tube position and the mean of the four positions in each hillslope plot were plotted against time. A linear mixed model with a cubic spline of time was used to test for differences in the time series of SWD between topographic positions and plots.

4.2.6 Infiltration capacity

The infiltration capacity was defined as the steady state rate of infiltration under ponded conditions. These conditions were met using a rainfall simulator, of the type described by Grierson and Oades (1977), to apply constant rainfall at a rate of 48 mm/h, to 43 bordered plots (0.64 m2) across the Vasey runoff and SGS national experiment sites. Runoff from the plots was measured volumetrically every minute and rainfall was applied for 1h (see Chapter 5 for more detailed materials and methods). The steady state infiltration rate was determined as the difference between the rates of rainfall and steady state runoff. Steady state runoff was defined as the rate at which the runoff rate first stabilised.

4.2.7 Water tables and subsurface flow

Piezometers were installed to depths of 2.9 m and 1.4 m in a lowerslope, midslope and upperslope position in each of the four hillslope runoff plots to measure water tables heights (Figure 3-6). A shallow piezometer (referred to here on as a dipwell) was also installed to the depth of the B horizon in the lowerslope position in each plot. The depths to the B horizon were 25, 45, 45, and 35 cm in plots 1 to 4 respectively. The piezometers and dipwells were constructed and installed based on methods described by Thompson et al. (1992). In brief, PVC tubing was perforated over the lower 0.7, 0.5 or 0.05 m of the 2.9 m and 1.4 m piezometers or shallow dipwells and covered with mesh. Tubes were lowered into augured holes and gravel was used to backfill around the perforated section. Bentonite was then used to form a seal above the gravel, and the original subsoil was subsequently used as backfilling material, followed by another bentonite seal at the soil surface. Water levels in the piezometers were measured using a fox whistle attached to a measuring tape at regular intervals between November 1998 and January 2001. Water levels in the four shallow dipwells were measured continuously by Dataflow® capacitive water level probes, and recorded each minute runoff or rainfall occurred, and hourly otherwise, by a Campbell Scientific® CR10X datalogger.

70 Perched water tables occur when a saturated zone develops above an unsaturated soil layer below. Perching was interpreted as occurring when water levels in shallower piezometers were higher than levels in the deeper piezometers in a landscape position, and also when the degree of saturation of the total soil porosity, calculated from neutron probe soil moisture measurements, was at least 10% greater than at the depth intervals above or below. The total hydraulic head (m) in each piezometer was calculated as the sum of elevation head (height of the bottom of each piezometer above that of the lowest piezometer in the hillslope) and the pressure head (water level in the piezometer). Contours of equal hydraulic head (equi-head contours) between piezometers of equal depth from the surface were interpolated and illustrated using Surfer® graphing software. Lateral head gradients were calculated as the change in hydraulic head divided by the horizontal distance between the piezometer nests. The vertical head gradient was calculated as the change in hydraulic head divided by the change in elevation of the bottom of the piezometers within each piezometer nest. The rate of subsurface saturated flow was calculated using Darcy’s equation (equation 4-1) where F is flow rate (m/day), Ksat (m/day, see Chapter 3) is assumed to be uniform at a

∆H particular depth between two piezometers, and ∆L is the lateral head gradient. ∆H F = K × (4-1) sat ∆L

4.2.8 Soil water balance

A daily soil water balance that used measured rainfall, runoff and profile soil water content and estimates of daily ETa to estimate subsurface lateral and vertical flows was used to calculate the annual partitioning of water above and below the ground surface for each runoff plot at Vasey. The partitioning of water was derived on a daily time-step using a method described by White et al. (2001). This approach uses the water balance equation;

P = ∆S + R + ETa + D (4-2)

where P is precipitation, ∆S is the change in the soil profile water storage over time (i.e.

SWDn -SWDn-1), R is surface runoff, ETa is actual evapotranspiration and D is the sum of subsurface vertical (drainage beyond the maximum rooting depth) and lateral flows. All variables were measured in mm/day. Measured daily rainfall and surface runoff data were used for the P and R terms. ETa was assumed to occur at the potential rate when the absolute SWD to 1.4 m was less than 25 mm and/or for the first 25 mm of a rainfall event. When these conditions were not met, ETa was calculated as a linear function of the SWD according to relationships developed by Scotter et al. (1979) and Heng et al. (2001) (equation 4-3),

71 ETa = a + bSWD (4-3)

where a and b are constants and the SWD is measured over the depth of the rootzone. The constants a and b were initially set at experimentally derived values (Heng et al. 2001) and used to calculate ETa and simulate changes in SWD and D. The soil water balance equation was then rewritten to solve for D (equation 4-4).

D = P − ETa − R − ∆S (4-4)

A measured value of the plot mean SWD (SWDn) when there was no subsurface flow (i.e D = 0) was used as the initial value to solve equation 4-4, so the equation was in the form of equation 4-5.

SWDn = P − ETa − R + SWDn−1 (4-5)

If SWDn was greater than or equal to zero, the surplus was assigned as D. The D value was then used to solve for the SWD for that day. The process was repeated to simulate daily changes in the SWD. Following solution of the water balance on a daily timestep, values of the coefficients a and b were optimised by successive iterations so that SWDn calculated in equation 4-5 was as close as possible to the corresponding measured value of SWD. Equation 4-4 was then solved using the optimised coefficients to estimate daily values of ETa and D, and daily values of each water balance component were summed to calculate the measured and estimated annual volumes. The degree of fit of the daily water balance model to the measured data was tested by regressing the simulated SWD (for absolute values greater than 25 mm) against the measured SWD, with the intercept set at the origin.

4.3 Results

4.3.1 Rainfall and evapotranspiration

Rainfall at the Vasey runoff site was 571, 565 and 577 mm in 1998, 1999 and 2000 respectively. These were all below the 12-year average of 623 mm that was measured 300 m from the field site, by a Bureau of Meteorology volunteer. Of particular note was the below average winter rainfall in each of the three years (Figure 4-4), and that between October 1996 and June 2000, rainfall in the Dundas was within the lowest 5% on record (Source; National Climate Centre, Bureau of Meteorology). In 1998 and 2000 however, spring rainfall was 44 and 34 mm above average (Figure 4-4).

The total ETp was 895 and 891 mm in the two complete years of measurement (1999 and 2000) and there was a surplus of rainfall over ETp during winter and part of spring over the measurement period (Figure 4-5).

72 300 Long term 250 average 200 1998 150 1999 100 Rainfall (mm) 50 2000 0 summer autumn winter spring

Figure 4-4: Seasonal rainfall at the Vasey runoff site compared to the long term average

Rainfall 160 ETp 140 120 100 80 (mm) 60 40 20 0

Rainfall or Potential Evapotranspiration Evapotranspiration Potential or Rainfall Jan-98 Jul-98 Feb-99 Aug-99 Mar-00 Oct-00 Apr-01

Figure 4-5: Monthly rainfall and ETp at the Vasey site over the experiment period

73 Daily rainfall and rainfall intensity

Table 4-2 shows that in 1998, 1999 and 2000, for the majority of days when rain occurred, there was ≤ 2 mm of rainfall. In the years that runoff occurred, a higher percentage of raindays had rainfall between 2 and 20 mm during the runoff season compared to the whole year. The majority of rain events also had peak intensities over a 5-minute period that were equivalent to 5 mm/h or less. Less than 10% of raindays had rainfall intensities exceeding 25 mm/h in each of the years of the experiment. According to Intensity-Frequency-Duration (IFD) curves generated for the study site location (Anon 1987) rainfall events with a maximum 6-minute intensity above 46 mm/h have an average recurrence interval (ARI) of more than 1 year at this location. Throughout the whole of 1999 there were no rain days with a maximum intensity greater than 46 mm/h.

74 Table 4-2: Frequency distribution of rainfall and rainfall intensity during the whole year and over the runoff period

------Whole year ------Runoff periodA ------1998 1999 2000 1998 2000 No. of raindays 227 239 85 % of raindays ≤ 2 mm/d 6572725768 % of raindays 2-20 mm/d 34 26 28 42 31 % of raindays > 20 mm/d 1B 2C 1D 21 % of raindays with peak intensityE ≤ 5 mm/hr 74F 79 75 58 60 % of raindays with peak intensity 5-25 mm/h 24 19 22 40 26 % of raindays with peak intensity > 25 mm/h 1 (2 events) 2 (4 events) 4 (8 events) 2 (97 mm/h) 6 (25, 41, 44 & 50 mm/h) A The runoff period was between 24 August and 13 November inclusive in 1998, and 1 September and 8 November inclusive and 14 April in 2000. No runoff occurred in 1999 B Includes one rainday of 42.5 mm C Includes four days with rainfall between 20 and 23 mm D Includes two raindays of 38.6 and 25.4 mm E Peak rainfall intensity over a 5- minute duration, equivalent to mm/h F Records of rainfall intensity available for May-Dec inclusive only in 1998

75 IFD data for individual seasons at the nearby Rocklands Reservoir (20 km north-east of the site) show the peak rainfall intensity that is expected to recur once every 1 or 5 years varies between seasons and is lower than the predictions based on annual data reflecting the lower probability of large storms in 3 months compared to 12 months (Table 4-3) (Hydrometeorological Advisory Service, Bureau of Meteorology). This shows that although the differences between short duration (6-minute) intensity peaks are not large, lower rainfall intensities are expected in winter compared to the other seasons.

Table 4-3: Maximum rainfall intensities (mm/h) over a 6 minute duration expected during each season at Rocklands Reservoir in rain events which recur once every 1 or 5 years

ARI Peak 6-minute Rainfall Intensity (mm/h) Summer Autumn Winter Spring 1 21.5 20.9 19.3 24.1 5 45.3 51.9 29.2 46.6

These seasonal IFD data refer to rainfall intensities over a 6-minute duration, whereas intensities calculated for the study site reflect 5-minute durations, so the ARI for the site may be slightly overstated. In 1998 there were two rain events during the spring runoff period with peak intensities with an ARI >1 year, according to the seasonal breakdown of IFD data for the Rocklands Reservoir. These included 23 September (97 mm/h, 1 in 10 yr ARI) and 27 November (36 mm/h) but the latter did not produce runoff. In the spring of 1999 one event had an ARI of 2-5 years (peak 5-minute intensity 30 mm/h), and in 2000 the ARI for spring rainfall was >1 year for two events (peak 5-minute intensities 25.2 and 40.8 mm/h on 11 Sept and 2 November respectively).

Rainfall duration

In 1998, the duration of rain events varied from one minute for light falls to long continuous showers lasting up to 24 hours. There was a higher proportion of shorter events during the runoff season when compared to the entire year. The duration of rain events that produced distinct runoff hydrograph peaks was from 45 to 1260 minutes in 1998 and from 5 to 1020 minutes in 2000.

76 4.3.2 Surface runoff

Runoff from the Vasey site occurred between August and November inclusive in both 1998 and 2000, with a single event also occurring in April 2000 (autumn). Similar total volumes of surface runoff were generated from the 2 ha hillslope site in both these years (Table 4-4). No runoff occurred in 1999.

Table 4-4: Volume and number of surface runoff events from the hillslope plots and flow in Dundas River in 1998 to 2000

Plot runoff or creek flow % of annual volume No. runoff eventsA volume (mm) from the four plots 1998 1999 2000 1998 2000 1998 2000 Plot 1 67.0 - 67.5 67 80 11 11 Plot 2 17.4 - 14.7 17.5 17.5 5 10 Plot 3 11.6 - 1.7 12 2 2 8 Plot 4 3.3 - 0.1 3.3 0.1 1 4 Total 99.3 - 84.0 Dundas CkB 31 6 22 A Includes events with distinct hydrograph peaks only. Surface baseflow also contributed to the total flow B Dundas River flow (DSE 2003)

Daily runoff that occurred from saturated soil, in the absence of rain and without a hydrographs peak is referred to as surface baseflow hereafter. Surface runoff and/or baseflow occurred almost continuously from plot 1 from mid August until late October. The last runoff event occurred on 13 November. Runoff from all plots in 1998 was dominated by a single large event on 23 September. This was the only rainfall event that produced runoff from plot 4 and one of only two producing flow from plot 3 (Figure 4-6). The peak flow rates and volumes for two events in plot 1, and one event in plots 3 and 4 were estimated from the existing portion of the hydrograph because the flow rate for part of the storm exceeded the throughput capacity of the tipping buckets. In 2000, runoff from plot 1 occurred on two occasions prior to 1 September, after which runoff occurred continuously until early October. The last runoff event occurred on 5 November. Runoff was initiated from plot 2 on five occasions throughout winter and spring, from plot 3 on four occasions and from plot 4 on only two occasions (Figure 4-7). Some runoff was generated from all plots during a storm in autumn. In both years, the frequency and magnitude of runoff events decreased in the order plot 1, 2, 3 and 4 (Table 4-4). Between 67 and 80% of surface runoff from the 2 ha study site was produced from plot 1, with as little as 0.1% being produced from plot 4. Each plot was 0.5 ha in area. The number of distinct runoff-producing rain events was similar between years although in 2000 there were more events in plots 2, 3 and 4 than in 1998 (Table 4-4). Despite this, the total amount of runoff was greatest in 1998, due to the large magnitude of events on 23 September

77 and 12 October 1998 (Figure 4-7). There were 13 and 11 mm of surface baseflow produced from plot 1 in 1998 and 2000 respectively. There would have been some degree of contribution of surface baseflow to hydrograph peaks (i.e. to runoff ‘events’), but this was not separately estimated. Flow data for Dundas River (DSE 2003), which drains the 211 ha subcatchment in which the field site was located, indicated that average flows since 1990 were greatest in July (9 mm), August (19 mm) and September (16 mm) and the average annual flow was 51 mm (range 6 - 124 mm). Of the four hillslope plots at Vasey, total annual flow in the Dundas River most closely matched the volume of runoff from plot 2 (Table 4-4).

78 50

40

30

20

Rainfall (mm) Rainfall 10

0 Jan Mar May Jul Sep Nov

30

20

10 Runoff, plot 1 (mm) 1 plot Runoff, 0

30

20

10 Runoff, plot 2 (mm) plot Runoff, 0

30

20

10 Runoff, plot 3 (mm) 0

30

20

10 Runoff, plot 4 (mm) 4 plot Runoff, 0

Figure 4-6: Rainfall and runoff from individual runoff plots in 1998

79 50

40

30

20

Rainfall (mm) Rainfall 10

0 Jan Mar May Jul Sep Nov

20

15

10

5 Runoff, plot 1 (mm) plot Runoff,

0

20

15

10

5 Runoff, plot 2 (mm) plot Runoff, 0

20

15

10

5 Runoff, plot 3 (mm) plot Runoff,

0

2

1 Runoff, plot 4 (mm) plot Runoff, 0

Figure 4-7: Rainfall and runoff from individual runoff plots in 2000

80 In both 1998 and 2000, larger runoff events dominated the annual runoff volume. Daily runoff >2 mm from the four plots combined accounted for 84% of annual runoff in 1998 and 81% in 2000. Between 62 and 77% of that flow was from plot 1. The dominance of the runoff volume from the larger events contrasted with the lower frequency of these events (Figure 4-8).

35 30 25 20 1998 15 2000 10 5

Frequency of events per year per events of Frequency 0 <1 <3 <5 <20 <60 Daily sum of runoff from four plots (mm)

Figure 4-8: The frequency distribution of runoff events of increasing volume

Table 4-5 shows that in 1998, the total hillslope runoff at Vasey was heavily influenced by a 1 in 100 year rain event, which contributed 56% of the total runoff from the site. In 2000, however, the volume of flow was more evenly distributed across a range of peak rainfall intensities, with 72% being produced by rain events with the highest frequency (once in 1 year, Table 4-5). In both years, runoff was only produced from all four plots when the peak 5-minute rainfall intensity was >25 mm/h (Table 4-5).

81 Table 4-5: Distribution of runoff volume across levels of rainfall intensity (mm/h, 5-minute duration)

Peak 5-minute rainfall intensity mm/h ≤ 6 ≤ 12 ≤ 25 ≤ 33 ≤ 47 ≤ 56 97+ ARIA year 1 1 1 1-2 2-5 5-10 100 1998 Max. no. plots with runoff - 1 3 3 4 Total runoff B mm 7.9 14.6 20.9 54.5 Percentage of annual flow in 1998 % 8 15 21 56 2000 Max. no. plots with runoff - 3 3 3 4 4 4 Total runoff B mm 19.7 40.6 5.8 9.9 6.5 1.4 Percentage of annual flow in 2000 % 24 48 7 12 8 2 A The ARI is based on rainfall intensity over a 6-minute duration and is therefore likely to slightly overstate the ARI for a rainfall intensity calculated over a 5-minute duration B The sum of runoff from all four plots

Runoff coefficients

The runoff coefficient (RC) is defined as the ratio of the volume of runoff per unit area to the volume of rain per unit area (i.e. mm runoff/ mm rainfall) and is described hereafter in percentage terms for monthly rainfall and runoff. The variability in runoff volume generated between runoff plots is further exemplified by the variability in monthly RC values across plots (Table 4-6). The RC for plot 2, and the mean monthly RCs for the 2 ha hillslope area were similar to the RC of the Dundas River sub-catchment (Table 4-6). The large difference between the RC for plot 3 in September 1998 and 2000 can be accounted for by a single large runoff event, as can the difference between the RCs for plot 1 in November of each year. The RCs for a 39 mm rain event, which occurred outside the main runoff season on the 14th April 2000, were 2.8, 0.2, 0.5 and 0.2 % for plots 1, 2, 3, and 4 respectively.

82 Table 4-6: A comparison of monthly runoff coefficients (runoff/rainfall, %) at a plot, hillslope and sub-catchment scale

1998 2000 Month PlotA HillB CreekC PlotA HillB CreekC 1234 1234 Aug----- 1100000 3 Sep431411318234912101613 Oct 37 5 0 0 11 4 13 1 0 0 3.5 5 Nov00000 1173005 6 A 0.5 ha hillslope runoff plots at the Vasey site B Mean of runoff coefficients from the four plots in the 2 ha hillslope area at the Vasey site C Calculated from streamflow and rainfall in the 211 km2 Dundas River sub-catchment (DSE 2003)

Hydrographs

The peak runoff rates from the four plots generally decreased in the order plot 1, 2, 3 and 4 (Figure 4-9). Hydrographs typical for runoff that followed both wet (25 September, Figure 4-9a) and dry (1 November, Figure 4-9b) antecedent soil water conditions displayed rapid and steep responses in runoff rate to episodes of rainfall, with only a single runoff peak for each rainfall peak. Less common were hydrographs with a more gradual rise and fall in flow rates (Figure 4-9c).

83 a) 25 September 2000

0.1 30

0.08 25 plot 1 20 0.06 plot 2 15 0.04 plot 3 10 Rainfall 0.02 5 Runoff rateRunoff (mm/min)

0 0 Cumulative Rainfall (mm) 10:0014:0018:0022:00 2:00 6:00 10:0014:00

b) 2 November 2000

0.1 10

0.08 8 plot 1 0.06 6 plot 2 plot 3 0.04 4 plot 4 0.02 2 Rainfall Runoff rateRunoff (mm/min)

0 0 Cumulative Rainfall (mm) 10:30 14:30 18:30 22:30

c) 8 October 2000

0.1 12

0.08 10 8 0.06 plot 1 6 0.04 plot 2 4 Rainfall 0.02 2 Runoff rateRunoff (mm/min)

0 0 Cumulative Rainfall (mm) 2:30 6:30 10:30

Figure 4-9: Runoff hydrographs and cumulative rainfall on a) 25 Sep, b) 2 Nov and c) 8 Oct, 2000

84 4.3.3 Surface waterlogging

From August until October 1998, the bottom half of plot 1 was noticeably waterlogged, particularly on the northern side, in the region of shade of a eucalyptus tree next to the plot. There was also a higher proportion of annual grass than the sown phalaris and clover species in the waterlogged area. Surface baseflow was generated from this area, even when there was no rain, at a rate of ~1 L/min. Plot 2 was also saturated to the surface at times in the lower half, particularly in the area adjacent to plot 1, but no surface baseflow occurred. No saturated areas were observed in plots 3 or 4. In 2000, a similar spatial distribution of surface waterlogging occurred. The mean volumetric water content measured using the 0-12 cm FDR probe over the lower 20 m of the wettest transect in each plot exceeded field capacity by early September in all four plots in 2000 (Figure 4-10). Plot 1 had a higher mean water content during spring and was wetter for a longer period of time than the other three plots (Figure 4-10). )

3 0.6 /cm 3 0.5 Plot 1

0.4 Plot 2

0.3 Plot 3 Plot 4 0.2 Field Capacity (-10kPa) 0.1

Volumetric water content (cm 0 Jul Aug Sep Oct Nov Dec

Figure 4-10: Mean volumetric water content of the surface soil (0-12 cm) over the lower 20 m of the wettest transect in each plot in winter and spring 2000.

In plot 1, water initially ponded on the surface between about 7 and 15 m upslope (Figure 4-11a), and later towards the base and across the slope (Figure 4-11b). Towards the end of the runoff season, the distribution of surface soil water reverted back to the original area of 3 3 ponding upslope from the base (Figure 4-11). A consistently dry (θv ≤ 0.26 cm /cm ) region between 30 and 37 m from the base of transect 1 in this plot coincided with a sandy rise (shown

in Figure 4-11b only). The saturated water contents (θsat) of the A1 (0-5 cm) and A21 (5-20 cm) 3 3 horizons were 0.57 and 0.41 cm /cm (see Table 2-5), so it was expected that θv larger than 0.41 cm3/cm3 in Figure 4-11 could reflect saturation of some of the 5-30 cm soil layer. The spatial

85 and temporal patterns of surface saturation reflected in Figure 4-10 and Figure 4-11 were consistent with visual observations of surface waterlogging. In plots 2, 3 and 4 there were no areas that remained saturated to the surface between runoff events near the base of the plot and 3 3 this is consistent with the mean θv in the lower 20 m remaining well below 0.41 cm /cm throughout the winter and spring. The size of the saturated area in plot 1 varied within and between storms. Evidence from the surface FDR soil water transects (Figure 4-10) and visual observations suggest the size of the saturated area between storms i.e. the semi-permanent area of saturation, varied from 0.9% of the 2 ha study site early in the runoff season to 1.7% later in October. However observations during a long, low intensity storm, indicated the saturated area expanded transiently to encompass all of plot 1 and 2.

86 c) Plot 1: 3 November 2000 a) Plot 1: 7 September 2000 b) Plot 1: 10 October 2000

35

30 30 30 0.46

0.44 ) 25 25 25 0.42 cm3/cm3 (

0.40 ) 20 0.38 20 20

0.36 cm 0-12 (

t

0.34 en t 15 15 15 0.32 er con 0.30 t Distance upslope from base of plot (m) ric wa Distance upslope(m)plotbase of from

10 0.28 t 10 10 Distance upslopefrom base of runoffplot (m) 0.26

0.24 Volume 5 5 5 0.22 0.20

0 5 10 15 0 5 10 15 0 5 10 15 Distance across slope (m) Distance across slope (m) Distance across slope (m)

Figure 4-11: Patterns of surface soil water (0-12 cm) accumulation in the lower quarter of plot 1 on a) 7 September, b) 20 October and c) 3 November 2000

87 4.3.4 Soil profile water storage and water tables

Soil water deficit

The SWD across topographic positions and plots ranged from approximately –120 mm to –60 mm during summer to slightly positive values in late spring 1998 (Figure 4-12). In contrast to 1998 and 2000, the SWD did not reach zero during spring in 1999 indicating that the soil did not fully wet up. The maximum average SWD in each plot over the three summers decreased in the order plot 1 (-94 mm), plot 4 (-91 mm), plot 3 (-85 mm) and plot 2 (-68 mm). The maximum deficit in 2000 may not have been measured due to cessation of regular field measurements after January 2001. There was a large degree of variation in the SWD of soil profiles within and between plots, particularly during summer (Figure 4-12). There was a consistent trend across all the plots for the lowerslope soil profiles (particularly position 4) to become drier in summer than the upperslope positions (1 and 2). This resulted in up to 100% variation in the maximum deficits between hillslope positions in any one plot (eg plot 3, position 2 and 4 Figure 4-12c). As well as this, the lower slope position 4 profiles appeared to reach their minimum annual SWD (i.e. wet up) later than profiles further up the slope in all plots. Linear mixed models including a cubic spline of time fitted for all plots and topographic positions indicated the overall mean SWD (linear effect) was smaller for the two upper landscape positions compared with the lower positions, except for in plot 1, where the uppermost position had the second largest SWD. The interaction between plot and position meant that there was no statistically significant overall difference (P>0.05) between plot mean SWDs. The average water storage (mm) at field capacity ranged from 183 to 573 mm. The maximum water storage in the four soil profiles that were <1.5 m (indicated by * in Figure 4-12), was less than in deeper soil profiles, however, the maximum SWDs were amongst the largest SWDs of all topographic positions (Figure 4-12).

88 a) Plot 1: High P rotation b) Plot 2: Low P set stocked 20 20 0 0 Jul- Nov - Mar- Jul- Nov- Mar - Jul- Nov - Mar- Jul- Jul- Nov- Mar - Jul- Nov- Mar - Jul- Nov - Mar - Jul- -20 98 98 99 99 99 00 00 00 01 01 -20 98 98 99 99 99 00 00 00 01 01 -40 1 (528) -40 1 (573)

-60 2 (519) -60 2 (553) 3 (474) 3 (493)

SWD (mm) (mm) SWD -80 -80 4 (476) SWD (mm) 4* (183) -100 -100 mean mean -120 -120

-140 -140

c) Plot 3: High P set stocked d) Plot 4: High P rotation 20 20

0 0 Jul- Nov - Mar- Jul- Nov- Mar - Jul- Nov - Mar- Jul- Jul- Nov- Mar - Jul- Nov - Mar - Jul- Nov- Mar - Jul- -20 98 98 99 99 99 00 00 00 01 01 -20 1 (512) 98 98 99 99 99 00 00 00 01 01 -40 -40 1 (531) 2 (511) -60 3* (368) -60 2 (529) 3 (517)

SWD (mm) SWD -80 4* (454)

SWD (mm) SWD -80 4* (377) mean -100 -100 mean -120 -120

-140 -140

Figure 4-12: Mean and actual SWD (mm to 1.5 m soil depth) of the four topographic positions from 1 (upslope) to 4 (toeslope) in each of the plots over time. Maximum water storage (mm) at SWD = 0 indicated in parentheses. * indicate SWD of profiles of <1.5m depth

89 4.3.5 Perched and groundwater tables

During late winter and spring, perched watertables were prevalent in the lower slope and midslope positions in 2000 (Figure B-1 to Figure B-4, Appendix B). Groundwater tables measured in the 2.9 m piezometers rose seasonally in most landscape positions but to a lesser extent in the drier year of 1999 than in 2000. Zones of higher soil saturation calculated from neutron probe soil moisture measurements indicated actual depths of water accumulation. For example a comparison of Figure 4-13 and Figure 4-14 shows that drainage was more uniform in the upper slope position of plots 3 and 4 compared with the midslope profiles of plots 1 and 2 where some degree of perching at 0.6 –1.0 m depth can be seen. The occurrence and depth of perched water tables were interpreted from the degree of soil saturation (shaded areas in Figure B-1 to Figure B-4, Appendix B) and from measured piezometer water levels. Water perched between 0.6 and 1.4 m and between 1.6 and 2.3 m in midslope landscape positions. This was likely to have been caused by restriction of drainage by the heavy clay B3 horizon (at about 1.3 m), and the C horizon (>2 m)(see Chapter 3), respectively. In the lower slope positions, drainage was restricted at about 0.6 m and again at between 1 and 2.3 m depth. In the lowerslope region these depths correspond with the B22 heavy clay horizon at about 0.85 m, and the C horizon at about 2 m. Figure B-1a and Figure B-4a (Appendix B) show the water levels in the A horizon at the base of plots 1 (initially only) and 4 were caused by water perched at <0.9 m, whereas the shallow water table in plots 2 and 3 developed within the A horizon only after the B horizon was saturated by groundwater (Figure B-2a and Figure B-3a, Appendix B). Water levels (from either perched or groundwater tables) reached within 0.70 m of the soil surface in the midslope and upper slope positions. At the base of each plot, the minimum depth to groundwater increased from 0.10 m in plot 1 to 1.14 m in plot 4. Groundwater levels fell below 2.9 m by May in 1999 and March in 2000. In the lowerslope position, perched water tables rose to within 17 cm in 1998 and 10 cm of the surface in 2000 in all four plots (Figure 4-15). In plot 1, the water level remained within 5 cm of the soil surface for most of spring, whereas water levels in the remaining plots were at least 12 cm below the ground surface on the day preceding runoff. In these plots, temporary water tables developed in response to rain events, with lag times between the peak runoff rates and peak water table height of 16 – 67 h (plot 2), 3 - 16 h (plot 3) and 0.5 – 16 h (plot 4). In plots 2, 3 and 4 these temporary water tables persisted in the A horizon for between 1 and 22 days. Water tables lasting more than 10 days were generally caused by perching in the lower B horizon, whereas the shorter duration water tables were perched within 0.9 m of the soil surface.

90 b) Plot 1, midslope

20

60

100 % soil Depth from surface (cm) surface from Depth water saturation 140 100

90 Jul 01 Jul 00 Jul 99 Jul 98 80

b) Plot 2, midslope 70

20 60

50 60 40

100 30 20 Depth from surface (cm) Depth

140 10

0

Jul 01 Jul 00 Jul 99 Jul 98

Figure 4-13: Degree of water saturation of the soil profile over time in midslope positions of a) plot 1 and b) plot 2, showing perched water tables

91 a) Plot 3, upper slope

20

60 % soil water 100 saturation Depth from surface (cm) surface Depth from

140 100

90 Jul 00 Jul 99 Jul 98 Jul 01 80

b) Plot 4, upper slope 70

60 20 50

60 40

100 30

20 140 Depth from surface (cm) surface from Depth

180

Jul 01 Jul 00 Jul 99 Jul 98

Figure 4-14: Degree of water saturation of the soil profile over time in the upper slope positions of a) plot 3 and b) plot 4

92 50 40 30 20

Rainfall (mm) 10 0

8 8 8 8 0 0 0 0

9 9 9 9 0 0 0 0 ------p v p v ct ct ug e o ug e o O O A S N A S N

10

0

-10

-20

-30

-40 Watertable Depth (cm) Depth Watertable -50

Plot 1 Plot 2 Plot 3 Plot 4

Figure 4-15: Rainfall, and shallow water table heights from August to November inclusive in 1998 and 2000 measured in dipwells at the base of all four plots

93 4.3.6 Subsurface vertical and lateral flows

Surface infiltration and water table drainage rates

The infiltration capacity of the soil surface, measured as the difference between the rainfall and steady state runoff rates using a rainfall simulator, ranged from 0 to 46 mm/h across 43 plots, and there were no significant differences (P>0.05) between paddocks or treatments. The mean (± se) infiltration capacity was 24 ± 2 mm/h. The rates of rise and fall of water tables from the logged dipwells in the A horizon of the grazing site are shown in Table 4-7. Water tables rose more quickly than they fell, and there was greater variability in the rates of rise than the rates of fall. It is possible that surface or subsurface water flowing directly into the dipwells via macropores and other preferential flow paths could have contributed to the high rates of rise (eg. 18.3 mm/h in the Plot 4 dipwell in 2000). There was a weak positive relationship between rate of rise and the peak antecedent rainfall intensity (Figure 4-16). Equivalent rates of drainage and infiltration through the soil matrix were calculated (Table 4-7) based on the assumption that for soils at field capacity, new infiltrating and draining water only had access to the unsaturated pore space fraction of the soil. The pore space fraction was estimated based on the difference between the volumetric water contents at saturation and field capacity. The rates of infiltration and drainage were then defined as KsatD (mm/h), i.e.

Ksat D = r(θ sat −θ FC ) (4-6) where r is the rate of rise or fall in the logged dipwell. The pore space fraction of the top 3 3 20 cm of soil at Vasey was estimated as 0.17 cm /cm (see Table 3-5, Chapter 3). KsatD during a rising phase is a measure of the net amount of water infiltrating into the A horizon. When the perched water table in the dipwell is falling, KsatD is a measure of the rate water can drain out of the A horizon into the subsoil. When the water level is rising, it can be assumed that water is leaving the A horizon at a rate equivalent to that during a falling period. Thus, the actual rate of infiltration into the A horizon should be the sum of that entering the A horizon, plus that leaving.

94 Table 4-7: The mean rates (mm/h) of rise and fall of water tables in the A horizons of plots 2, 3 and 4.

Plot Year r - rise KsatD_rise (Net r- fall (dipwell KsatD_fall KsatD_infil (dipwell rise infiltration rate fall rate) (drainage rate (absolute rate) into A horizon) out of A infiltration rate horizon) into A horizon)B 2 1998 2.1 (0.8A) 0.4 0.8 (0.2) 0.1 0.5 2000 5.6 (1.6) 0.9 1.9 (0.1) 0.3 1.2 3 1998 12.8 (1.3) 2.2 1.6 (0.5) 0.3 2.5 2000 13.1 (3.7) 2.2 2.3 (0.3) 0.4 2.6 4 1998 - - - - - 2000 18.3 (3.6) 3.1 4.7 (1.1) 0.8 3.9 A Standard error of the mean B KsatD_infil = (KsatD_rise) + (KsatD_ fall)

60.0

50.0

40.0 plot 2 plot 3 30.0 plot 4 20.0

10.0

Rate of rise of water table (mm/h) 0.0 0 1020304050 Peak rainfall intensity (mm/h)

Figure 4-16: Rate of rise in the A horizon water table (mm/h) against the peak antecedent rainfall intensity in 2000

Recharge and discharge conditions-vertical hydraulic head

The range of vertical head gradients of groundwater and perched water in each piezometer nest (difference between hydraulic head in deeper minus head in shallower piezometer) are shown in Table 4-8. Positive head gradients indicate net upward flow, or discharge zones, and negative gradients indicate net downward flow, or recharge (Cox and Reynolds 1995). Discharge conditions occurred in upper slope positions of plots 3 and 4 and the

95 lower slope position in plot 1. In the latter position, discharge was caused by upward pressure from both the perched and groundwater tables (Table 4-8). At the base of plot 1, this upward pressure was sufficient to cause surface discharge. It is likely that surface discharge occurred in other unmeasured positions in the landscape, particularly where surface waterlogging was observed, however, vertical head gradients within the top 1.4 m were only measured in the lowerslope positions. The cross sectional behaviour of the groundwater and perched watertables in each plot are illustrated in Figure 4-17a-d. These show that in plots 1 and 4 there was considerable distance between the perched watertable level and the groundwater in the midslope positions whereas in plots 2 and 3 the watertable surfaces were approximately parallel with the ground surface. There was usually greater head at the base of the slope in plot 1 compared with the other plots, at an equal surface elevation.

Table 4-8: Range of vertical hydraulic head gradientsA (m/m) within piezometer nests

Groundwater gradientB Perched water gradientC Piezometer nest Upslope Midslope Lowerslope Lowerslope Plot 1 - 0.02 - 0.74 - 0.41 – 1.78 +0.02 – 0.79 +0.05 – 0.95 Plot 2 - - 0.2 – 1.45 - 0.08 – 0.62 - 0.03 – 0.11 Plot 3 +0.18 – 0.52 - 0 – 1.06 - 0.17 – 0.85 - 0.02 – 1.40 Plot 4 +0.12 – 0.97 - 0.71 – 1.49 - 0.37 – 1.00 - 0.06 – 0.50 A + indicates upward head, - indicates downward head B Head gradient between water levels in the 2.9 m and shallowest piezometers C Head gradient between water levels in the 1.4 m and shallowest piezometers

96 a) Plot 1 b) Plot 2 260 260

258 258

256 256

254 254 sea level (m) sea level (m) level sea 252

Elevation/head above above Elevation/head 252 Elevation/head above above Elevation/head

250 250 0 20 40 60 80 100 120 140 0 20 40 60 80 100 120 140 Horizontal slope distance (m) Horizontal slope distance (m)

c) Plot 3 d) Plot 4 260 260

258 258

256 256

254 sea level (m) level sea 254 sea level (m) level sea Elevation/head above 252 Elevation/head above Elevation/head 252

250 250 0 20 40 60 80 100 120 140 0 20406080100120140 Horizontal slope distance (m) Horizontal slope distance (m)

Figure 4-17: Cross-sections of each plot showing surface, and interpolations of B and C soil horizons (dashed lines) and hydraulic head elevation above sea level (m) on 22 Sep 00 or 9 Sep 00 (black line) and 16 Oct 01 (grey line). Symbols indicate water levels in 2.9 m (■) and 1.4 m (▲) piezometers.

97 Direction and rate of subsurface lateral flow

Equi-head contours that were interpolated from water levels measured in the 2.9 m piezometers were assumed to represent the groundwater behaviour. During the wettest periods in 1999 and 2000 the head contours approximately followed the surface elevation contours (Figure 4-18). The direction of saturated subsurface flow is assumed to be normal to the head contours, indicated by arrows in Figure 4-18. There was some evidence for subsurface flow in plots 1 and 2 converging towards the base of plot 1 in 2000 (Figure 4-18), however the convergence was less evident during 1999, which was a drier year (Figure 4-19). The equi-head contours between the 1.4 m piezometer water levels were assumed to represent the subsurface lateral flow of perched water. In the wetter year of 2000, the direction of subsurface lateral flow closely resembled that of the groundwater flows (eg Figure 4-20). However, in the drier spring of 1999, localised fluctuations in hydraulic head (Figure 4-21) indicated there was less topographic control on the distribution of perched water tables. The rate of subsurface flow was calculated as being proportional to the head gradient between two piezometers, assuming the Ksat remained uniform (equation 4-1). Equi-head contours closest together therefore indicated a steeper head gradient and a more rapid flow rate. Figure 4-18 shows that the rate of downslope flow of groundwater in plot 1 decreased towards the base, owing to the rising groundwater table at the base of the slope. The ranges of groundwater and perched water flow rates between each piezometer nest are listed in Table 4-9.

Table 4-9: The rangeA and directionB of groundwater and perched flow water rates (m/day x 10-2) between upslope, midslope and lowerslope positions in each plot

GroundwaterC Perched waterD Plot Upslope / Midslope Midslope / Downslope Upslope / Midslope Midslope / Downslope 1 0.82 – 1.12 + 0.03 – 0.49 0.41 – 0.56 0.47 – 0.67 2 - 0.36 – 0.72 - 0.65 – 0.84 3 0.46 – 0.76 0.38 – 0.82 0.47 – 0.57 0.52 – 0.63 4 0.38 – 1.23 0.32 – 0.73 0.32 – 0.60 0.70 – 0.77 A Ksat assumed to be the B horizon Ksat of 0.13 m/day, (see Chapter 3) B + sign indicates net upslope flow, no sign indicates net downslope flow C Total flux measured between piezometers at 2.9 m depth D Total flux of perched water, measured between piezometers at 1.4 m depth,

98 11160

11140 6.75 Head (m) 2.41 11120 2.50 8.00 11100 1.47 7.00 3.95

11080 6.00 1.25

3.80 5.00 11060 1.03 Northing (m) Northing 4.00 11040 6.75 3.00

11020 2.00

1.00 11000 6.90

0.00 10980

960 980 1000 1020 1040 1060 1080 1100 1120 Easting (m) Figure 4-18: Equi-head contours for 2.9 m piezometer water levels, and elevation contours (m) on 22 September 2000

99 11160

Head (m) 11140 5.15

2.55 11120 1.86 8.00 11100 1.48 7.00 3.30

11080 6.00 1.48

2.95 5.00 11060 1.33 Northing (m) Northing 4.00 11040 5.03 3.00

11020 2.00

1.00 11000 5.21

0.00 10980

960 980 1000 1020 1040 1060 1080 1100 1120 Easting (m) Figure 4-19: Equi-head contours for 2.9 m piezometer water levels, and elevation contours (m) on 15 September 1999

100 11160

11140 7.10 Head (m) 4.93 11120 2.72 8.00 11100 2.66 7.00 4.91

11080 6.00 2.87

4.73 5.00 11060 2.92 Northing (m) Northing 4.00 11040 7.19 5.74 3.00

11020 2.00

1.00 11000 7.18

0.00 10980

960 980 1000 1020 1040 1060 1080 1100 1120 Easting (m) Figure 4-20: Equi-head contours for 1.4 m piezometers, and elevation contours (m) on 15 September 2000

101 11160

11140 6.25

4.51 Head (m) 11120 2.00

11100 8.00 0.00 4.48 7.00 11080 2.28 6.00 0.00 11060 0.00 Northing (m) Northing 5.00

0.00 11040 4.00 0.00 3.00 11020

2.00 11000 6.39 1.00

10980 0.00

960 980 1000 1020 1040 1060 1080 1100 1120 Easting (m) Figure 4-21: Equi-head contours for 1.4 m piezometers, and elevation contours (m) on 15 September 1999

102 4.3.7 Water balance

0

-20

-40

-60

SWD (mm) to 1.4 m -80

-100 Apr- Jul- Oct- Jan- Apr- Jul- Oct- Jan- Apr- Jul- Oct- Jan- Apr- 98 98 98 99 99 99 99 00 00 00 00 01 01

Actual SWD Estimated SWD

Figure 4-22: Simulated and measured values of SWD (mm) in plot 2

The relationship between the simulated daily SWD (equation 4-2) and measured SWD was similar for each of the four plots, so only that for plot 2 is shown in Figure 4-22. The simulated SWD explained only 55 to 78% of the variation in measured SWD values (Table 4-10). The low agreement between simulated and measured SWD was due mainly to overestimation of the SWD in summer 1999 and underestimation during the wetting up phases of 1999 and 2000. The latter may have resulted in an overestimation of the amount of subsurface vertical and lateral flow during winter in each of these years.

103 Table 4-10: Partitioning of water (mm) in the four runoff plots using a simulated water balance

Plot Year Rainfall Ea Surface Subsurface No. Partition Correlation (P) runoff (R) vertical and days ratioA coefficient lateral flow (D) (r2) ------mm ------1 1998 286 267 68 20 170 0.23 0.55 1999 565 473 0 66 365 1 2000 577 446 67 86 366 0.56 2 1998 286 286 17 40 170 0.71 0.78 1999 565 468 0 83 365 1 2000 577 444 15 134 366 0.90 3 1998 286 288 12 42 170 0.78 0.78 1999 565 470 0 82 365 1 2000 577 447 1.7 145 366 0.99 4 1998 286 296 3.3 46 170 0.93 0.69 1999 565 489 0 62 365 1 2000 577 468 0.1 127 366 0.99 A The partition ratio, (D/(D+R)) is defined as the ratio of subsurface flow to the sum of subsurface and surface flow A partition ratio (PR) was defined as the ratio of the amount of subsurface flow to amount of subsurface and surface flow. The PR is similar to that defined by White et al. (2001) who included subsurface lateral flow in the ‘R’ term rather than the ‘D’ term.

PR = D /(D + R) (4-7)

In years when runoff occurred, the PR increased from 0.23 – 0.55 in plot 1 to 0.70 – 0.99 in plots 2, 3 and 4 (Table 4-10). This increase was attributed to both more subsurface flow and less surface runoff in plots 2, 3 and 4. The amount of subsurface flow was estimated as markedly larger in 2000 compared to 1998 in all four plots. This occurred despite similar volumes of runoff in these years in plots 1 and 2.

104 4.4 Discussion

The hydrological behaviour of the hillslope at the study site was consistent with a generalised two dimensional model described by Chorley (1978). In Chorley’s model, water infiltrates downwards near the top of a hill, flows horizontally over the midslope region and vertically upwards at the base of the slope, the latter being caused by discharge or ‘artesian’ pressure developed by downward subsurface flows from higher in the landscape. Similarly, a conceptual model developed by Dahlhaus and MacEwan (1997), describing salinisation processes in the Dundas Tablelands, was well suited to the hydrological pathways and processes occurring at the Vasey runoff site. In their model, which considers the regolith to the depth of the parent material, lateral flow and surface saturation of lowerslope areas are key features. In the context of this latter model, the following discussion starts with a description of the pathways of water movement and mechanisms causing surface saturation at Vasey In the second part of the discussion, the processes of surface runoff generation at Vasey are discussed, after which the spatial and temporal variability of surface and subsurface flow pathways and surface saturation are addressed. Following this, the water balance approach to quantifying and partitioning of surplus water in each plot is discussed. The hydrology of the field site is then related to that of the wider environment and the chapter concludes with a discussion of the hydrological implications for P loss from the hillslope pastures.

4.4.1 Hillslope hydrological pathways and processes

In the hillslope hydrological model of Dahlhaus and MacEwan (1997)(Figure 4-23, numbers in parentheses refer to numbers in the figure), infiltrating rain rapidly backfills preferential vertical flow pathways (1) and causes ponding of water in the upper soil profile (2) that provides the head for downslope subsurface lateral flow (3). Return flow (4) is caused by thinner soils on erosional slopes as well as by lower hydraulic conductivity of downslope soils. Groundwater rises (5) in response to contributions of deep infiltration and slow moving subsurface lateral flow, and the lowerslope region subsequently becomes waterlogged (6) due to an accumulation of surface and subsurface lateral flow and groundwater discharge. Evidence for these processes and pathways at the Vasey runoff site is discussed sequentially below.

105 A

B21 (3) (2) (1) (4) B22,23 (2) (3) (4) (6)

B3 (6) (5)

C

Figure 4-23: Simplified representation of the hillslope hydrological model of Dahlhaus and MacEwan (1997) (not to scale)

106 Preferential vertical flow pathways (1)

Water table levels were near or above the ground surface in a number of positions in the landscape. However runoff during low intensity storms often occurred without the water levels in the piezometers at the base of the slopes reaching the soil surface. For saturation excess runoff to occur, it is hypothesized that during storms macropores fill quickly, providing a pseudo-saturated surface, after which the total pore volume may saturate more slowly. This slower saturation is assumed to be represented by the water table activity in the A horizon dipwells (Figure 4-15), and limited by the infiltration capacity of the soil. Such a mechanism of backfilling of macropores and preferred paths to cause surface saturation was also postulated for texture contrast soils in South Australia (Smettem et al. 1991) and for soils of the Dundas Tablelands (Dahlhaus and MacEwan 1997). In contrast to Vasey, Smettem et al. (1991) found that watertables did dissect the ground surface when runoff occurred on small plots (20 x 3 m). Structural macropores in the Vasey soils were smaller (up to 2 mm diameter, see Table 3-1) than the 12 mm diameter pores observed by Smettem et al. (1991), however, larger macropores such as root channels can also provide preferential pathways for the flow of water (Brouwer and Fitzpatrick 2002a). Because of the spatially variable distribution of macropores the transient pseudo-water tables were likely to be localised and not necessarily represented by the single dipwells at the base of each hillslope plot.

Perched water tables (2)

Perched watertables were detected in all four runoff plots at Vasey (see Figure B-1 to Figure B-4, Appendix B). In lowerslope positions, the soil water content profiles over time suggested that water was perched between 60-80 cm depth – which was above the heavy clay B22 horizon. Mottling was strongest in the B22 and B23 horizons, which is consistent with the second depth of perching that occurred above the C horizon. In midslope positions, vertical drainage was probably restricted by the heavy clay B3 and C horizons, causing water to perch above these horizons. Brouwer and Fitzpatrick (2002b) also found water perched within the B horizon and above a pallid kaolinised zone in a yellow duplex soil 2 km from Vasey. The distribution of perched water was consistent with the presence of a bleached A2 horizon and mottled subsoils, which are typical features in periodically waterlogged soils, and the significantly lower Ksat in the B (0.13 m/day) and C (0.06 m/day) horizons compared with the A1 (0.31 m/day) horizon (P<0.05, see Table 3-4). Mottling is often caused when iron oxides in waterlogged, anaerobic zones are reduced and dissolved whilst pockets of aerobic, oxidised material remains strongly coloured (Cox et al. 1996b). The mottling pattern in the Vasey soils

107 indicated perched water tables are a common and ancient feature in these soils (Dahlhaus and MacEwan 1997; Brouwer and Fitzpatrick 2002b). The rates of rise and fall of the water tables in the A horizons dipwells represent the rate of vertical flow under conditions of field capacity (Table 4-7). The dipwell infiltration and drainage rates are likely to be heavily influenced by both preferential flow through the unsaturated pore space fraction of the soil, as well as saturated flow through the soil matrix, i.e the Ksat. This contrasts to Ksat, which was measured at 20 mm suction using the disc permeameters to exclude the influence of macropores greater than 1.5 mm diameter (Smettem et al. 1991).

The absolute infiltration rates (KsatD_infil, Table 4-7), calculated to account for flow through the unfilled pore space, were less than the Ksat of the A1 (12 mm/h), A21 horizon (4 mm/h) and A22 (12 mm/h) horizons and were far less than the infiltration capacity (including macropore effects) that was measured for the topsoil (24 mm/h). Infiltration into the A horizon was therefore not limited by soil permeability and was more dependant on the intensity of rainfall (Figure 4-16). However, perched water tables did occur indicating that the drainage rate from the A horizon was limited by subsoil horizons. This was demonstrated by KsatD_fall values being less than KsatD_rise values measured in the dipwells of plots 2, 3 and 4 (Table 4-7).

However, the KsatD_fall rates, which accounts for preferential unsaturated flow, were markedly lower than Ksat measured using a disc permeameter at 20 mm suction for the B21 (5.8 mm/h), B22 (5.5 mm/h) and C (2.4 mm/h) horizons. This might have been because drainage beneath perched water tables is restricted by trapped air and low hydraulic gradients into the surrounding soil, whereas drainage beneath disc permeameters is restricted only by the supply potential. KsatD_fall rates are therefore likely to better represent the potential for perched water tables to occur than the disc permeameter Ksat values. This was exemplified by perched water tables occurring despite the Ksat of the B and C horizons being high compared with other texture contrast soils where perched water tables occur (Smettem et al. 1991; Cox et al. 1996b; White et al. 2000).

Subsurface lateral flow (3)

The perched and groundwater tables that developed across the hillslope provided positive head gradients for downslope lateral movement of water (Table 4-9). The rates of downslope movement, based on Darcy’s equation, ranged from 0 to 1.23 x 10-2 m/day, and in the lowerslope position of plot 1, a small upslope gradient developed for a short time. Lateral flows would have occurred at the depths the perched and groundwater tables occurred, which was above the C horizon and in places, above the B23 and B3 horizons. Subsurface lateral flow above B and C horizons is an important pathway for flow of surplus water in duplex soils in Australia (Smettem et al. 1991; Cox and McFarlane 1995; Stevens et al. 1999; White et al.

108 2000; Brouwer and Fitzpatrick 2002a; Hatton et al. 2002). In the model of Dahlhaus and MacEwan (1997), subsurface lateral flows above weathered parent material is considered a particularly dominant flow pathway. Dahlhaus and MacEwan (1997) also suggest that subsurface lateral flow, rather than rising groundwater, plays an important role in waterlogging and salinisation processes, and that removal of native vegetation since European settlement may have increased the volume of subsurface lateral flow and duration of waterlogging, thereby causing more widespread salinity problems.

Return flow (4)

Return flow occurred as slow seepage (referred to as surface baseflow) from a seasonally saturated area, near the base of plot 1, and was also assumed to contribute to surface flow when subsurface lateral flow met saturated soils at the base of the slope, as well as when perched or groundwater discharged at the ground surface. Initially, the seepage zone may have developed through perched subsurface lateral flows accumulating and emerging at the soil surface, a process described by Chorley (1978), which would also describe an ephemeral spring observed by Cooke and Dons (1988). The piezometric water levels were consistent with this hypothesis because surface saturation occurred upslope prior to the groundwater rising to the surface in the lower reaches of the plot. It was likely the initial runoff, which also occurred prior to the perched watertable at the base of the slope dissecting the ground surface, was generated by return and saturation excess flow from this saturated upslope seep area. Similar to observations of Brouwer and Fitzpatrick (2002b), the perched water tables developed in the hillside seep even when there was no apparent irregularity in surface topography. The finding that perched water tables and surface waterlogging developed earlier in plot 1 than the other plots (Figure 4-10 and Figure 4-15), initially appeared counter-intuitive, based on observations of the soil physical profile. During the installation of neutron probe access tubes, intrusions of the saprolite material to within 1.3 m of the soil surface were detected in some lowerslope areas of plots 2, 3 and 4. The saprolite material was hard enough to resist penetration by a hydraulic soil corer so it was assumed the saprolite would also restrict vertical drainage. In plot 1, however, there was no evidence of parent material at depth, nor was there any mechanical resistance to the corer, suggesting plot 1 had a more permeable soil profile than the remainder of the hillslope, and would be therefore be less susceptible to waterlogging. Lateral subsurface flow within the top 1.4 m of plot 1 was indicated, however, by the downslope head gradients measured between the midslope and lowerslope 1.4 m piezometer positions (Table 4-9, Figure 4-20). The presence of a perched water table within the upper B horizon in the region of the seep was also consistent with the strong mottling and bleaching at

109 about 60 cm depth in a nearby soil pit, which indicated that a perched watertable must seasonally saturate the soil (Cox et al. 1998). For the perched water table to occur, it was hypothesised that either a localised clay lens may have restricted drainage (White 1997), the soil horizon became thinner and therefore had a reduced transmission capacity (Ward and Robinson 2000) and/or a conduit of unusually high permeability within the upper soil profile could have provided a preferential flow pathway, which enabled the seep to develop. A combination of these explanations was likely because immediately upslope of the saturated seep area was a sandy rise through which water may have infiltrated at a more rapid rate than it could flow either laterally or vertically further downslope through the clay horizons. As winter progressed, the waterlogging in plot 1 was also associated with surface discharge of perched water and groundwater at the base of the slope. At the end of the runoff season (November), however, the distribution of saturated areas in plot 1 was similar to the initial pattern and the perched watertable at the base of the slope no longer dissected the ground surface. This suggests the hydrological behaviour reverted to the original process whereby waterlogging was caused by return flow alone, rather than in combination with discharge conditions at the base of the slope.

Groundwater (5)

Consistent with the model of Dahlhaus and McEwan (1997), slowly rising groundwater was important at Vasey, with the proximity of the groundwater to the soil surface at the base of the slope decreasing in the order plot 1, 2, 3 and 4. Rising groundwater in the lower slope region of hillsides is considered an important component of storm flow in adjacent streams in the conceptual framework presented by Ward (1984), both directly as subsurface flow and indirectly through an increase in the area of surface saturation (see ‘Surface waterlogging (6)’). The groundwater also rose to within 1 m of the soil surface in upper slope positions across the entire runoff site and midslope positions in plots 2 and 3. Upslope groundwater activity was not indicated in the model of Dahlhaus and McEwan (1997), probably because it was based on much deeper soil profiles. However, the seasonal groundwater rise was consistent with the pallid B3 horizon, which indicated uniform depletion of iron oxides deposits through prolonged episodes of waterlogging (Cooke and Dons 1988; Cox et al. 1996b).

Surface waterlogging (6)

The seasonally saturated area that developed in plot 1 is commonly referred to as a variable source area (VSA) for surface runoff (Betson and Marius 1969). The VSA in plot 1 contrasts with the temporary surface saturation due to the back-filling of soil macropores that probably occurred across a larger area of the hillslope during storms.

110 Initial development of the VSA in plot 1 was attributed to subsurface lateral flow emerging at the surface, part way up the slope (see ‘return flow’), however a change in the pattern and area of saturation in plot 1 in late September (Figure 4-11a-c) indicated a second mechanism was then involved. The saturation was then also attributed to discharge (upward flow) of perched and groundwater, due to accumulation of surface and subsurface lateral flows at the base of the slope. Evidence for this mechanism was water perching above 60 cm depth and groundwater rising to near the soil surface at the base of the slope as the area of surface saturation expanded downslope (Figure B-1a, Appendix B) and a shallow watertable dissected the ground surface at the base of plot 1 for the latter part of the runoff periods in 1998 and 2000 (Figure 4-15). The saturated areas expanded and contracted in response to fluctuations in the perched and groundwater tables, and from variable inputs from rain, surface runoff and return flow. Ward (1984) describes an increase in pressure potential with depth (i.e. discharge conditions) in lower slopes as facilitating rapid saturation of surface soil layers after even small inputs of water. Surface waterlogging at the base of the hillslope by these mechanisms was also consistent with that found by Dunne and Black (1970a). Similarly, Hammermeister et al. (1982) found that prolonged perching of water in lowerslope positions in Oregon, USA was attributable to both lower vertical and horizontal hydraulic gradients than upslope regions, and to the supply of subsurface water from upslope. Vertical discharge did not occur in midslope positions to the extent that it did in upper or lowerslope positions. This may have been because greater lateral drainage occurred in the steeper sloped midslope region. At the base of the hillslope, saturated conditions were likely to have been enhanced by heavier textured B horizons in this position compared with midslope and tableland profiles (see Table 3-1). Soil profiles are typically less permeable soils in lowerslope landscape positions (Hammermeister et al. 1982; Cox et al. 1996b; Dahlhaus et al. 2000) probably due to greater soil weathering through more frequent waterlogging (White 1997).

4.4.2 Surface runoff processes

Processes of surface runoff generation can indicate the area of hillslope that is likely to contribute runoff, as well as the likelihood of runoff occurring. These are important considerations for investigations of runoff water quality and quantity.

Saturation excess flow

Of the three potential mechanisms of surface runoff generation, that is infiltration excess, saturation excess flow and surface baseflow, most runoff at Vasey occurred when rain fell on saturated soil. Saturation excess flow and return flow are widely considered to be the common mechanisms producing storm flow in temperate regions (Ward 1984; Cooke and Dons

111 1988; Nash and Murdoch 1997). Subsurface lateral flow, a component that was not measured directly in this study, may also contribute to storm flow (Ward 1984; Sklash et al. 1986), particularly in duplex soils (Smettem et al. 1991; Stevens et al. 1999; Cox and Ashley 2000). Visual evidence of a seasonally saturated area in plot 1 and the occurrence of runoff long after rainfall ceased suggested most of the runoff from this plot was produced by saturation excess and return flow mechanisms. Saturation excess runoff also occurred in plots 2 and 3 during large rain events, presumably by transient saturation of the soil surface by either pseudo (saturation of macropores) or temporary (matrix saturation) water tables. A rapid response to rain by surface runoff may be indicative of saturation excess and return flow from already saturated areas (eg Finlayson and Wong (1982) and Dunne (1978)) or of infiltration excess flow (Pilgrim et al. 1978). This was illustrated by the similarly steep slopes of the rising limbs of the hydrographs of both a saturation excess event (25 September 2000, Figure 4-9a), and an infiltration excess event (1 November 2000, Figure 4-9b). It was therefore not possible to identify the runoff processes using hydrographs alone. The flatter rising hydrograph limb of 8 October (Figure 4-9c) was likely to reflect a steady distribution of rainfall and drier antecedent soil water conditions. None of the hydrographs displayed a second, delayed peak, indicative of large contributions of slower moving return flow (O'Loughlin 1981), which suggests that return flow during runoff events was either operating as ‘quickflow’ in a similar fashion to saturation excess runoff, or was a small contribution, the rate of which was controlled by the rate of subsurface lateral flow. Surface baseflow refers to surface runoff that occurred in the absence of rain and with no hydrograph peak, which occurred as return flow from the saturated zone in plot 1. In both years runoff occurred this comprised only 13% of the total runoff from the site and was therefore a less important contribution to total flow than the distinct runoff events.

Infiltration excess flow

In temperate regions, infiltration excess runoff (also known as Hortonian overland flow) is usually less important than saturation excess flow, and often only occurs during infrequent intense storms on soils that have been compacted due to animal and vehicle traffic (Lambert et al. 1985; McColl et al. 1985; Haygarth et al. 2000) or where soils form impermeable crusts during heavy precipitation (Ward 1984). On three occasions in 2000 (including the April event) and one occasion in 1998, it is possible that some infiltration excess flow occurred because the 5-minute peak rainfall intensity exceeded the measured infiltration capacity of the topsoil. In other temperate catchments, up to 68% of total storm flow from pastures damaged by treading was attributed to infiltration excess runoff during large, infrequent storms (McColl et al. 1985). The mean infiltration capacity of the surface soil at Vasey (24 mm/h) was lower than for pasture topsoil in New South Wales (50-

112 75 mm/h, Costin (1980)), higher than for heavily grazed permanent grass in the UK (<1 mm/h, Haygarth et al. (2000)) and comparable to a Sodosol in Book Book, New South Wales (14 mm/h, White et al. (2000)). The infiltration capacity is enhanced by greater vegetative cover, as plant bases provide cracks and water entry sites, and by high organic matter, biological activity and low bulk density, all of which improve soil porosity. Structural cracks and faunal activity can also provide preferential flow pathways for infiltrating water. The range of measured infiltration capacities was large at Vasey, probably reflecting the heterogeneity in these topsoil characteristics that might be expected at the surface of established pastures. Runoff from plot 4 was not produced unless peak 5-minute rainfall intensities were greater than 25 mm/h (Table 4-5), which suggested that the infiltration capacity of the soil matrix and macropore network was exceeded and that infiltration excess flow mechanisms may have contributed to flow from all plots on these occasions. Smettem et al. (1991), used the comparison between storm rainfall intensity, and the hydraulic conductivity (>345 mm/h) at -5 mm supply potential of the 0-5 cm depth of topsoil to suggest that infiltration excess mechanisms were unlikely to generate runoff in a texture contrast soil which had good vegetative cover and macropores up to 12 mm in diameter. In contrast, no macropores greater than 2 mm diameter were observed in the soil at Vasey, so some contribution of infiltration excess flow was expected. It was notable that the volumes of infiltration excess runoff during the high intensity storm event following summer in 2000 were negligible (Figure 4-7), which demonstrates the soil infiltration rates were high prior to ponding at the soil surface. Nelson et al. (1996), also measured low catchment discharge rates despite high autumn rainfall intensities.

4.4.3 Spatial and temporal distribution of hydrological processes

Spatial distribution of surface and subsurface runoff

Both spatially and temporally, the most important runoff producing area was plot 1. Between 67 and 80% of surface runoff from the 2 ha study site was produced from plot 1, with as little as 0.1% being produced from plot 4 (Table 4-4). The saturated area in plot 1 conferred a higher propensity for runoff than non-saturated areas, resulting in double the number of runoff events occurring from plot 1 compared with plot 2. Even when a combination of infiltration excess and saturation excess runoff from widespread perched water tables may have occurred across almost all the hillslope during the largest of the events, the total runoff from plots 2, 3 and 4 was still only 50% of the runoff from plot 1 (Figure 4-6). There was little evidence for any direct pasture treatment effects on the field site hydrology. Rotational grazing management and higher soil P fertility may have played a role in increasing plant water use, leading to larger summer SWDs (94 to 91 mm) than the low fertility

113 set stocked plot (68 mm) Figure 4-12). The interaction between plot and slope position meant that differences between plot means were not significant. Similarly, White et al. (2003) found only marginal differences between maximum SWD across grazing management and soil fertility pasture treatments at SGS national experiment sites. It was more likely that differences in runoff volumes between plots were attributable to soil and topographic controls on the distribution of soil moisture. Similary, Hatton et al. (2002) found spatial variation in perched water levels and waterlogging on duplex soils in Western Australia were controlled by topographic, soil and drainage management rather than by water use of legume crops.

Influence of topography on surface saturation and runoff

During the autumn storm event in 2000, when the soil in all plots was uniformly dry, plot 1 still produced between 5 and 14 times as much runoff as the other three plots. The soil profile was at its driest at this time, and the peak rainfall intensity exceeded the mean soil infiltration capacity, so infiltration excess, rather than saturation excess mechanisms were important during this event. Therefore even without the development of a semi-permanent saturated zone at the hillslope base, the region isolated by plot 1 had a higher propensity for runoff production. The main factors controlling variation in the amount of flow between plots during this storm were likely to be the amount of groundcover, and the plot surface topography. There was 45-65% groundcover at the base of plot 1 and 65 to 95 % at the base of plots 2, 3 and 4. Reduced groundcover probably reduced detention storage, allowing less time for infiltration, thus causing increased runoff from plot 1 (Lang 1979; Hairsine and Prosser 1997). On hillslopes, gravity and soil properties influence the development of lateral flows. The degree of slope and its shape affects the magnitude, speed and direction of lateral flows as well as the distribution of zones of saturation (O'Loughlin 1981; Ward and Robinson 2000). Convergence of both subsurface and surface flow increases the frequency of water tables dissecting the soil surface during wet periods (Smettem et al. 1991), which leads to a greater frequency of saturation excess runoff when compared to slopes with more planar lateral drainage. Convergence of flow occurs where soils are thinner, where the hydraulic gradient decreases at the base of slopes, and in hillslope concavities (Ward and Robinson 2000). Saturated areas then expand and contract in response to variable inputs of water from rain and upslope flows (Barling et al. 1994). At Vasey, the overall slope of each plot was similar (Figure 3-5). In plot 1, however, there was a slight gully line along the slope length, illustrated by the convergence of the elevation contours in this part of the hillslope. In contrast, the other three plots were relatively planar, with plot 2 sloping towards plot 1. The development of the VSA at the base of plot 1 was therefore consistent with the hypothesis that the surface topography played a role in

114 controlling the distribution of soil moisture. Patterns of soil surface debris that were observed after the autumn storm in 2000 confirmed that surface runoff flowed perpendicular to the elevation contours. Similarly, the head contours in Figure 4-18 to Figure 4-20 show the direction of subsurface flow also approximately followed the topography of the land, enabling better lateral drainage of both perched and groundwater from the more planar and less convergent hillslopes in plots 2, 3 and 4 than plot 1. The equi-head contours do not define the slight gully in the topography between plots 1 and 2 but this is most likely due to the low intensity of piezometer nests contributing to the interpolation function within the mapping software. A decreasing head gradient between mid slope and lowerslope positions in plot 1 (Table 4-9) further reduced the lateral drainage of water leading to a sustained perched water table in the upper horizons. A decrease in the hydraulic head gradient from downslope convergence of flow is indicative of hillslopes that are concave both in cross-section and in plan (Ward and Robinson 2000). Higher peak runoff rates and faster initiation of runoff in plot 1 than the other plots (Figure 4-9) was also consistent with the development of a VSA and the convergent topography of plot 1 because flow that becomes concentrated along a gully line gathers speed and because rain falling on already saturated ground immediately runs off (Hudson 1995). Topographic controls did therefore appear to strongly influence the direction and distribution of surface and subsurface flow in this soil type and landscape. However, Grayson and Western (2001) suggested that in catchments where vertical fluxes dominate over lateral, there may be little association between the distribution of soil water in the rootzone of the hillslope and the amount of water which accumulates in low lying gullies, because much of the water drains to groundwater and is conveyed at depth to gully regions. The variation in volume and frequency of runoff and the position of the VSA at Vasey suggested the winter SWD would have been smallest in plot 1 and in downslope positions compared with other plots and positions. Figure 4-12 shows, however, that SWDs were larger in downslope positions than upslope during summer. In agreement with Grayson and Western (2001), this indicated that water moved above or below the measured rootzone to accumulate at the base of the hillslope in plot 1 (eg via groundwater discharge). However the downslope hydraulic head gradients of the perched water table did provide some evidence that there was subsurface flow within the rootzone and therefore that there was still a high degree of connectivity between the hillslope, as a source of soil water, and the location and size of the VSA in plot 1 through topographically controlled lateral flows.

Influence of soil properties on soil saturation

As well as terrain characteristics and plant water use, soil properties also influence the distribution of soil wetness in landscapes (Moore et al. 1991; Grayson and Western 2001). For

115 example, Betson and Marius (1969) found that the location of VSAs was associated with the depth of A horizon soil in a small catchment in USA. At Vasey, variability in the depth, texture, macropore density and localised depressions of the soil surface would have contributed to the variability in surface soil water content (eg. Figure 4-11), particularly during the wetting phase when lateral flows had not yet developed. Freer et al. (1997) and others (Cox and McFarlane (1995), Dunne and Black (1970b)) have also shown that the depth of bedrock or another impeding layer can also influence the development and direction of subsurface lateral flow resulting in surface waterlogging. At Vasey, interpolated cross-sections and equi-head contours maps of the plots showed that perched water tables approximately followed the surface topography (Figure 4-17 and Figure 4-20). However perched water tables often developed in regions where there was a shallow outcropping of the C horizon (eg. see midslope plots 1 and 4, Figure 4-17), and this sometimes lead to groundwater levels deviating from the surface contours.

Size of the variable source area

At Vasey, the size of the VSA (the semi-permanently saturated region) ranged from 0.9 to 1.7% of the 2 ha hillslope during the runoff season in 2000. Although expansion of the saturated area during storms was not measured directly, up to about 50% of the site was observed as waterlogged at the surface shortly after a storm. The VSA was smaller than that measured in overseas catchments of varying size, however, the degree of expansion in response to rain was not dissimilar to that noted by McColl et al. (1985). They found the saturated area of a New Zealand pasture catchment varied seasonally from 8 to 14% and expanded to up to 100% of the catchment for a few hours during the largest storms. In another study, the saturated area of a 16 ha hill pasture catchment expanded during winter and spring from permanently saturated stream and seepage zones to comprise 13% of the total catchment. Further increase in the area of saturation occurred as temporary headward expansion of the stream source (Cooke and Dons 1988). Gburek and Sharpley (1998) found that even during large storms (124 mm) the maximum extent of the VSA, in a 26 ha cropped catchment in the USA, was 30 m upslope of a stream channel. It was not surprising that the VSA at the Vasey site was quite small compared to those in New Zealand, as the Australian environment has fewer permanently saturated areas - exemplified by an abundance of ephemeral streams.

Temporal variation in rainfall and surface runoff

Rainfall was below average in each of the years of the study but higher than average in spring in 1998 and 2000 (Figure 4-4). This was consistent with runoff being generated from the site in these two years only. As for other duplex soils, perched water tables were less prevalent and most surplus water drained via subsurface vertical or lateral pathways in the drier year of

116 1999 (Cox and Ashley 2000; Cox and Pitman 2001). The highest rainfall intensity in 1999 was 38.4 mm/h (5-minute), which occurred in summer. Only one storm exceeded the average infiltration capacity of the topsoil during spring but infiltration excess runoff did not occur. In the wetter years of 1998 and 2000, groundwater discharge and subsurface lateral flows contributed to surface waterlogging and surface runoff occurred, highlighting the inter-annual variation in flow pathways also found for other duplex soils (Cooke and Dons 1988; Cox and Reynolds 1995; Stevens et al. 1999). Winter and spring rainfall dominated the annual totals, the majority of rain days yielded 2 mm or less rain in each year and rain events lasted for up to 21 h. These rainfall characteristics were typical of the weather patterns of the south-west of Victoria in which most rainfall occurs in the winter half of the year when temperatures are low and high intensity storms are rare. Higher intensity rainfall is often associated with more localised convective systems, which tend to occur in the warmer months. The seasonal IFD estimates in Table 4-3 reflect this pattern, showing that although the seasonal trend is not strong, high intensity storms are expected less frequently during winter than in spring or autumn. Large runoff events (>2mm) dominated the annual flow (81 to 84% of total site flow) despite their lower frequency when compared to smaller events. Infrequent, large events commonly dominate annual catchment flows in Australia (Nelson et al. 1996). For example, rain falling onto wet soil in spring produced 28 mm of runoff from a texture contrast soil in New South Wales, which amounted to 20% of the total of five years of runoff (Costin 1980).

4.4.4 Modelling hydrological flow pathways

Daily water balance model

A water balance for each runoff plot was derived using measured quantities of rainfall, surface runoff, ETp and the soil water storage in 1998, 1999 and 2000. The relatively poor fit of relationships between predicted and measured SWD across the four plots (Table 4-10) was likely to have been influenced by the large topographic variation in SWD about the plot mean values used in the daily water balance. White et al. (2001) found a much better correlation (r2 = 0.84) using the same approach for a perennial pasture in north-east Victoria. Underestimation of the SWD during the wetting up phases in each year may be attributed to infiltrating water bypassing the rootzone through cracks and fissures that developed during the dry summer. Hydrographs from 10 m deep piezometers at Vasey showed that much of the rise in water table occurred in early winter (M. McCaskill, unpublished data), much earlier than when the simple water balance model would indicate the soil profile was at field capacity. Bypass flow of this nature was also suggested to occur in soils in the region by Dahlhaus and MacEwan (1997), and is a feature of clayey soils in the UK (Haygarth et al.

117 2000). The simple soil water balance did not account for drainage when the soil had a large SWD and therefore overestimated the amount of water stored within the rootzone during the wetting up phase.

Partitioning of surplus water in the hillslope landscape

The annual partitioning of water that was predicted using the simulated daily water balance (Table 4-10) showed that 46 and 77% of surplus water was partitioned into surface runoff in plot 1 in 1998 and 2000. Subsurface vertical and lateral flow was estimated to account for between 71 and 99% of the surplus water in plots 2, 3 and 4 in these years, and was the only pathway of surplus water movement in the driest year, 1999, across the entire hillslope. In north east Victoria, White et al. (2001) found that deep drainage comprised 41 to 56% of surplus water in a strongly duplex Sodosol and up to 97% in a Kurosol with less textural contrast between soil horizons. At the current site, where the soil type was relatively uniform across the hillslope area, the contrast in partitioning across the runoff plots was probably attributable to the topographic controls discussed earlier. It was not surprising that surface runoff did not always account for the majority of surplus water, as geomorphological studies performed by Cox et al. (1998) identified features in the hillslope soil profiles typical of perched watertables, and these would facilitate some subsurface lateral flow. However, this subsurface lateral flow may not have been as important as for other duplex soils, as there was only about a two-fold difference in Ksat between horizons. 6 This contrasts markedly to the 3 x 10 fold decrease in Ksat from the loamy sand topsoil to the heavy clay B horizon of a Chromosol at Myponga, South Australia where subsurface lateral flow was found to dominate the partitioning of surplus water (Stevens et al. 1999). It is therefore expected that deep drainage was an important component of the subsurface water partitioning. This is supported by the gradual rise in the groundwater table during the runoff season. A lack of runoff in 1999 also highlighted that even in soils with low Ksat at depth, in low rainfall years, runoff will probably be confined to small saturated areas near watercourses and subsurface flow may be the main contributor to catchment stream flows (Cox and Ashley 2000). The estimated annual amounts of subsurface lateral and vertical flow combined at Vasey were similar to drainage values alone (i.e. no subsurface lateral flow component) derived for pastures in north east Victoria (range 78 – 269 mm) and much larger than the 12 – 33 mm average drainage estimated for phalaris pastures in NSW (Heng et al. 2001). The large inter- annual variation in the volume of subsurface flow values that occurred at Vasey was also found in these studies. The estimated annual subsurface flow was four-fold greater in 2000 than 1998 despite similar volumes of runoff in plots 1 and 2. This reflected the higher percentage of raindays during the 2000 runoff season with falls of 2 mm or less, compared with 1998 when rain fell in less frequent, but larger doses, allowing more water to be evaporated between events.

118 Variation between years in the partitioning of water is to be expected in the variable climate of south eastern Australia. For example, White et al. (2000) found surplus water occurred in only two of four years of monitoring in pasture plots in NSW.

4.4.5 Implications for hydrological processes at a larger scale

The variability in runoff volumes from the four plots highlights the spatial heterogeneity of runoff generation, which may not only be expected across small areas such as the study site, but also at the catchment scale (Dunne and Black 1970b; Pilgrim et al. 1978). Uniform generation of runoff across landscapes is a phenomenon restricted mainly to arid-to-subhumid climates with thin or bare vegetative cover and limited soil variation (Dunne 1978). The variation in monthly runoff coefficients between plots at Vasey further illustrated the variation in the propensity for runoff across the Vasey hillslope area. The large runoff coefficients in plot 1 were probably partly due to contributions of return flow to the runoff component. This flow component may have originated from areas outside the plot surface boundary and therefore is an input of water that is not represented by the runoff coefficient. The differences in magnitude of runoff between years at the site reflected the variation in flow at the sub-catchment scale in the Dundas River. The Vasey hillslope produced on- average a similar amount of monthly runoff per unit area and per unit rainfall (represented by the runoff coefficient) as the sub-catchment in 1998 and 2000 (Table 4-6) indicating the variation and range of hillslope hydrological processes were not dissimilar to those occurring at larger landscape scales. The runoff coefficient of plot 2 also closely matched that of the sub- catchment which reflected the intermediary drainage and runoff characteristics of this plot compared with the saturated and well-drained extremes of plots 1 and 4 respectively. The variability of hydrological processes at different spatial scales may explain the differences in dominant flow paths measured in separate studies and years on a duplex soil in South Australia’s Mt. Lofty Ranges. Smettem et al. (1991) and Kirkby et al. (1996), using a series of instrumented 20 x 3 m plots, measured the majority of flow as macropore –mediated throughflow in 1987/88 and 1991 respectively, whereas Stevens et al. (1999) measured predominantly surface runoff from two larger sub-catchments (4.2 and 3.6 ha) in 1997 at the same location. As Stevens et al. (1999) pointed out, upscaling the results to conclude that the dominant catchment flow mechanism is surface runoff may underestimate the return flow contributions of subsurface lateral flows from upslope. The reverse may also be said (i.e. an underestimation of the importance of return and surface flow) when upscaling from the plot studies, which highlights the importance of studying hydrological processes at the landscape scale.

119 Sources of error in hillslope runoff plot experiments

Surface flows from almost an entire hillslope length were isolated in order to investigate runoff generated at the farm scale. Nutrient and soil loss measured at this scale is an accurate representation of those resources lost from the pasture system (Hudson 1993), although it will rarely also represent sediment and nutrient delivery to receiving waterways. To reduce interference to the normal hydrological behaviour of the study site, no obstruction to subsurface flows was installed at either the external or internal plot boundaries. Some artificial interruption to surface flow would have been encountered, however, at the surface barriers at the boundaries between plots. Some leakage at the runoff outlet of plot 1 occurred but was negligible relative to the total flow.

4.4.6 Implications for P movement via hydrological pathways

The fate of surplus water can have a large influence on the potential for P to be transported to receiving surface waterbodies. The dominant runoff producing areas, and the extent of expansion of the saturation zone are important for understanding water quality because they determine the land area that interacts with surplus water (Dunne 1978). Movement of nutrients in runoff from the VSA in plot 1 is therefore likely to be the most important process governing nutrient losses from the field site. In low rainfall years, surface runoff is unlikely so the quality of subsurface flows is more important for controlling P losses. In wetter years, even when the amount of surface flow is approximately equal to the amount of subsurface flow, it is likely that surface runoff will present a greater risk of P movement from pastures in this environment because of the enrichment of nutrients at the soil surface, the high P-sorbing clay content of the subsoils and the direct and fast flow pathway. Consistent with other studies, surface flow was dominated by large storms and these events are usually the most important hydrologic feature in controlling nutrient losses (Burwell et al. 1975; McColl et al. 1977; Cooke 1988; Nash and Murdoch 1997). The following chapters set about testing the propositions concerning spatial and temporal likelihood of P movement along the surplus water pathways by investigating the processes and amounts of P loss, and the interaction between P sources and flow pathways.

120 4.5 Conclusion

The pathways and processes of water movement at the field site were largely consistent with a conceptual hydrological model presented by Dahlhaus and MacEwan (1997). Saturation excess flow from a variably sized saturated source area was the most important runoff- producing mechanism, although during three very intense storms it was likely a small degree of infiltration excess runoff occurred. The dominant runoff producing area comprised only about 2% of the 2 ha study site. During infrequent, large storms, this area may have increased to 50% of the site. However, even during these large storms, the majority of flow was produced from plot 1 which comprised 25% of the area. The saturated zone presented the greatest potential for surface water movement in the landscape. The remainder of the hillslope presented a lower risk for runoff, because less runoff was produced less often. Subsurface pathways were more important than surface pathways in drier years. The implications of these hydrological characteristics on P movement from the sheep pasture are that most P is likely to be lost via saturation excess surface runoff during winter and spring from the VSA.

121 CHAPTER FIVE

5 Phosphorus, nitrogen and sediment in runoff and drainage from pastures at Vasey

5.1 Introduction

P transported via surface and subsurface pathways from agricultural land makes an important contribution to eutrophication of surface water bodies in Australia (Davis et al. 1998). Concentrations of P in runoff and drainage from pastures can be high enough to support nuisance aquatic algal growth (Correll 1998), and therefore may be of concern where there is a large degree of hydrological connectivity with a water body in which P is the limiting factor for algal growth. Whilst P loss from pastures can be substantial in terms of risk of eutrophication in the wider environment, losses are rarely a financial burden from the point of view of lost production (Sharpley and Halvorson 1994; Stevens et al. 1999). The major pathway for P movement is usually determined by the dominant pathway of water movement. P losses from pastures have been measured in a variety of environments, but owing to variations in land management, soil type, climate and topography, the processes that are most important usually vary between locations. For example P is readily leached from poorly buffered sandy pasture soils (Lewis et al. 1981; Ritchie and Weaver 1993), is entrained in surface runoff from irrigated (Austin et al. 1996) and dryland (Tham 1983; Nash and Murdoch 1997; Fleming and Cox 1998) pastures, and can move in lateral subsurface flow pathways in duplex soils (Stevens et al. 1999; Cox and Pitman 2001). There is concern within the sheep industry in south-west Victoria that increasing P application rates to boost pasture production and stocking rates could lead to a deterioration in the quality of surface runoff (Court et al. 1998). Little information exists on the potential for P losses in runoff from the soils under different land uses and climates in south-west Victoria so the field experiment described in this study was undertaken to investigate water and P movement from pastures. The pathways of water movement at Vasey were investigated in Chapter 4, and this chapter sets about quantifying the movement of P along these pathways as well as identifying the types and distribution of sources of P in pastures and how these interact with flow pathways.

122 The aims of the experiments reported in this chapter were to: 1. Identify the amount of P lost in runoff from sheep pastures and whether losses were related to P fertility 2. Identify pasture characteristics that affect runoff, erosion and P losses 3. Identify whether subsurface drainage and lateral flow were significant pathways for P loss from sheep pastures These aims were met by firstly measuring the concentrations and loads of P and sediment in runoff from high and low fertility sheep pastures. Natural runoff from hillslope plots and simulated rainfall-runoff from small (0.64 m2) plots were investigated. The surface characteristics of the pastures, the amount of P in herbage, dung and soil and the distribution of these P sources were also measured and related to the surface water quality. To assess the potential for off-site P movement via subsurface flow pathways, the concentrations of P in the soil water within and below the pasture rootzone were also measured.

5.2 Materials and methods

5.2.1 Sample collection

Hillslope runoff

Surface runoff from four hillslope plots (0.5 ha each) grazed by sheep at Vasey was measured using surface interceptor barriers and calibrated tipping buckets as described in Chapter 4. The pasture management treatments, described in more detail in Chapter 3, were; Treatment A: Low P application, set stocked Plot 2 Treatment B: High P application, set stocked Plot 3 Treatment C: High P application, 4-paddock rotationally grazed Plots 1 and 4 The volume, temporal distribution and mechanisms of flow in each plot were described in Chapter 4. Flow-weighted subsamples of runoff were collected from the tipping buckets using a sample splitting device (Figure 4-2). The sample splitter consisted of a 3 mm by 145 mm plastic slot, which was positioned such that the bottom of the slot was level with, and directly adjacent to, the fully ‘tipped’ bucket. An average of 0.12% (20 mL), 0.09% (14 mL), 0.13% (24 mL) and 0.24% (37 mL) of every alternate bucket tip was collected by the sample splitters in plots 1 to 4 respectively. Plastic tubing connected the sample splitter to three collection vessels, which filled successively under gravity. The first vessel was a 500mL high density polyethylene bottle (vessel A), and the second and third (vessels B and C) were 15L plastic containers. All collection vessels were washed in dilute HCl and rinsed repeatedly with deionised water prior to installation in the field.

123 After runoff events, or on a daily basis during low flow conditions, the total sample volume in each vessel was recorded, after which vessel A was collected and swapped for an empty vessel, and two thoroughly mixed 500 mL subsamples were collected from each of vessels B and C. Samples were insulated in an ice box whilst being transported to the laboratory. Unfiltered samples were stored at –15oC prior to analysis for total P (TP).

Simulated runoff

Artificial rainfall was used to produce runoff on forty-three 0.64 m2 plots during August and September 1999. The plots were located on paddocks included in the Vasey SGS national experiment and the hillslope runoff plots. The pasture treatments studied included those represented in the hillslope runoff study as well as a low P fertility pasture that had not been sown to productive pasture species. Approximately equal numbers of simulator plots were assigned to each of the improved pasture treatments, and these were distributed across replicate paddocks of each treatment (Table 5-1). Rainfall at an average intensity of 48 mm/h was produced using a pressure regulated spinning disc simulator, which was mounted on a trailor (Grierson and Oades 1977)(Figure 5-1). The water outlet was elevated to at least 2m above the ground to ensure droplets reached a terminal velocity similar to natural rainfall. Runoff was isolated by a metal frame and collected in a tray at its lower boundary. The frame was installed to a depth of 4 cm and provided an above ground barrier 6cm in height. Plots were positioned parallel to the slope and where the lower boundary was as level as possible with the metal frame to minimise bare earth being exposed to runoff. Rainfall was applied for 60 minutes and runoff was pumped into calibrated measuring cylinders from which the height was manually recorded every minute to determine the runoff rate. A 70 mL subsample was collected every 5 minutes and stored in an ice box. In the laboratory, composite samples (approximately 120 mL) representing the first, second and third 20-minute periods of runoff were prepared by mixing flow-weighted proportions of the original samples. Chemical analyses were performed on composite samples. Unfiltered samples were stored at –15oC prior to analysis for TP and SS.

Table 5-1: Treatment descriptions for rainfall simulator plots at Vasey

Trt Treatment description PaddockA No. plots B High P, set stocked, improved pasture A3, V5, V14 3+5+5=13 A Low P, set stocked, improved pasture A2, V3, V11 3+5+5=13 D Low P, unimproved pasture (dominated by onion grass) V18 5 C High P, Four-paddock rotation grazed, 8 days since A1/1, A4/2 3+3=6 grazing C High P, Four-paddock rotation, 35 days since grazing A1/3, A4/4 3+3=6 Total 11 43 A ‘A’ denotes paddocks that were also used in the hillslope runoff experiment, ‘V’ denotes paddocks located on the Vasey SGS national experiment site

124 Figure 5-1: Rainfall simulator and runoff plot

Soil water

Soil water within and below the rootzone of the four 0.5 ha runoff plots was sampled using porous ceramic cups (Magesan et al. 1994; Ridley et al. 2001). The cups, which were 40 mm in diameter and 80 mm in length, were joined to PVC pipes and then pre-washed with dilute HCl and rinsed with deionized water. The pipes were stoppered with a rubber bung at the atmosphere outlet. Narrow tubing, which was inserted through the bung to reach the length of the pipe, was connected to a battery powered Dynavac® pump and used as a conduit to apply suction and evacuate subsequent water samples. The tubing was clamped to maintain between 70 to 80 kPa suction between sample times. Cups were inserted on a 45o angle to a vertical depth of 30cm, and vertically to a depth of 120 cm to sample soil water from within and below the rootzone. Excavated soil was repacked around the cups to ensure adequate contact between the soil and ceramic. Pairs of cups were installed in an upper slope, midslope and lowerslope position in each of the runoff plots. Cumulative samples were collected every 7 to 14 days during winter and spring in 1999 and 2000 and analysed for TP and TN - and NO3 -N. There was little seasonal change in TP or TN concentrations in soil water so the average

125 annual concentrations of nutrients were calculated from the arithmetic mean of the cumulative samples. The total nutrient flux in drainage waters was calculated using the annual average nutrient concentrations and the total subsurface lateral and vertical flux for each plot that was estimated from the water balance approach described in Chapter 4.

Pasture phosphorus sources and surface characteristics

Herbage

Hillslope runoff plots: Representative pasture herbage samples were collected from each of four subplot strata in the 0.5 ha runoff plots for TP analysis. Twenty random handfuls of pasture, cut at ground level along a ‘W’-shaped transect, in each subplot were combined into a bulk sample. The percentage groundcover was visually estimated every 7 –14 days during winter and spring. Groundcover at each permanent soil sampling point was visually estimated four days after a storm that occurred in April 2000. Simulated rainfall plots: Samples of pasture herbage were also collected from the simulated rainfall plots after the rainfall. These were cut at ground level from within a randomly placed 0.0625 m2 quadrat in each plot. All herbage samples were washed to remove contaminants, dried at 60oC and then ground for TP analysis. For each simulated rainfall plot, the total herbage mass was estimated by weighing the dried quadrat samples. The average herbage density was visually ranked on a 5 unit scale ranging from 1 (many bare spaces between plants) to 5 (plants formed a thick mat and completely covered the soil surface). The average herbage height in each plot was calculated from 10 random measurements and the percentage groundcover was visually estimated.

Dung

Hillslope runoff plots: The percentage dung cover in 0.0625 m2 (25 x 25 cm) quadrats placed near each permanent soil sampling point in the hillslope runoff plots was visually estimated four days after the autumn storm. The location of stock camps (bare ground and accumulation of dung) was also visually assessed at this time. Simulated rainfall plots: All the dung within the rainfall simulator plot boundaries was collected after the application of artificial rainfall, dried at 60oC, weighed and ground for TP analysis.

Soil

Hillslope runoff plots: Surface soil samples (0-10 cm depth) were obtained for each hillslope plot in September 2000 by collecting 10 cores (2.5 cm diameter) in each of four subplot regions per plot. All soil samples were air-dried (40oC) and sieved (<2 mm) prior to chemical analysis.

126 In the hillslope runoff experiment, the spatial distribution of extractable soil P was investigated by collecting soil cores (0-10 cm depth, 2 cm diameter) near 102 permanent points before fertiliser application (early to late autumn) and in spring in 1998, 1999 and 2000. The sampling sites were spaced approximately every 15 x 15 m within the plots and every 4 m along the boundaries of each plot (Figure 3-6) giving 27, 26, 26 and 23 sample points in plots 1, 2, 3 and 4 respectively. Ten soil samples were taken randomly within a 1m radius of each sampling point and bulked. Bulked samples were transported in airtight plastic bags from the field site to the laboratory at the Pastoral and Veterinary Institute, Hamilton. The bulked samples were then thoroughly mixed and a sub-sample (approximately 100 g) was weighed, dried at 105oC for 24 h, and reweighed to determine its volumetric water content. The remainder of the bulked sample was air-dried (40oC), stored at 4oC and sieved to less than 2 mm prior to analysis of soil Olsen P (POlsen). Forty-three random samples collected in spring 1998 and representing each hillslope plot, were also ground using a ring rock grinder and used for total P analysis. The northing, easting and elevation co-ordinates of each point were measured by surveyors, and Surfer® mapping software (Anon. 1995) was used to graphically represent the spatial distribution of POlsen at and between the sampled points. Kriging, as an interpolation procedure, was built into the software and it was assumed there was linear variation in

POlsen between the sampled points. The vertical distribution of P in the soil surface 15 cm was measured by collecting fifteen 3 cm diameter cores randomly along a ‘W’-shaped transect in each of the four subplot strata in each 0.5 ha plot in October 2000. Samples were not collected near fence-lines where fertiliser applications were likely to be less uniform. Each core was cut into section of 0-2, 2-5 and 5-10 cm depths. Sections representing each depth interval were combined to create a bulk sample for each subplot. Samples were air-dried (40oC), plant and root material was removed, and sieved (<2 mm) samples o were stored at 4 C prior to POlsen analysis. Simulated rainfall plots: During the rainfall simulation experiment, 10 soil cores (0-5 cm depth, 2.5 cm diameter) were taken from just outside the perimeter of each 0.64 m2 plot. Cores were not taken from inside the plot after the cessation of rainfall due to the difficulty in removing saturated soil cores and to avoid any changes to labile P concentrations that may have been caused by the rainfall itself.

5.2.2 Chemical analyses

Phosphorus, nitrogen and suspended sediment in water

In the laboratory, a known volume of sample was passed through a 47mm diameter glass fibre Bonnet® prefilter and a 0.45µm pore size cellulose acetate membrane (Gelman® or Millipore®)

127 filter, both held in a polysulfone filter holder. Filtered (F) and unfiltered (UF) samples were stored at –15oC prior to analysis for P, N and SS. TP and DP were measured in water samples that had been frozen for up to 5 months. Potassium persulfate was used to oxidise or hydrolyse organic- and poly- phosphates to - orthophosphate and N compounds to NO3 based on Method 4500-P B No.5 by Clesceri et al. (1998). Alkaline K persulfate was prepared by dissolving 26.6 g K persulphate and 6.0 g NaOH in 400 mL of high purity water. Sample aliquots (8 mL) of thawed runoff water samples were digested with 2 mL of the oxidising reagant in an autoclave for 60 minutes at 115 kPa (121oC)(Hosomi and Sudo 1986; Nash 1997). P in digested and cooled samples was measured colorimetrically at 882nm by the Mo-blue method of Murphy and Riley (1962), with ascorbic acid as the reductant, using a Lachat flow - injection system (Huberty and Diamond 1996). To measure TN, NO3 in the digested samples was - - reduced to NO2 in a copperized Cd column, and the NO2 was measured colorimetrically at 520 nm using a Lachat Flow injection system (Nash 1997). Nitrate-N concentrations in surface runoff were measured in 1998 and 2000 for a proportion of the runoff events. Ammonium-N was measured in 1998 only. Subsamples for these analyses were stored at -15oC and analysed by a commercial laboratory (Incitec®). Suspended sediment (SS) concentrations in hillslope runoff samples from 2000 were measured by weighing the residue on the glass fibre and 0.45 µm pore filters. All the filters used to filter a known volume of runoff sample were dried at 37oC in pre-weighed petri dishes, cooled in a desiccator then weighed. Average weights of dried, unused filters were used to calculate the sediment weight by difference. SS concentrations in simulated runoff samples, and hillslope runoff samples from 1998, were measured using the same method, having been previously air-dried and stored for 4 and 14 months respectively. An evaporative method was also used to measure SS concentrations, but a statistical comparison of methods (Jorgensen 1985) revealed greater variance using the evaporative method so results reported here were measured using the filter method only.

Dung and herbage total P

Dung and herbage samples (0.500 g ± 0.010) were digested in 4mL of 69% (w/v) HNO3 and

1.5 mL H2O for 20 minutes at 4100-4800 kPa in pressurised Teflon® vessels using a Milestone® MDR-1000 microwave digestion system. Digests were analysed for total P (% dry matter) using ICP- AES (Anon 1998a). These analyses were conducted at the State Chemistry Laboratory in Werribee, Victoria.

128 Soil Olsen and total P

Soil samples representing the hillslope and rainfall simulator plots, and soil collected for the study of the surface and vertical spatial distribution of P were analysed for plant available P using the

Olsen (NaHCO3 extraction) method (Olsen et al. 1954; Rayment and Higginson 1992). A 2.5 g sample was weighed and mixed with 50 mL of 0.5M NaHCO3 (pH 8.5) for 30 minutes on an end- over-end shaker. Shaken samples were centrifuged for 5 minutes at 3300 rpm and the supernatant was decanted. P in the extracts was measured colorimetrically at 882nm using the Mo-blue method of Murphy and Riley (1962), either by a manual method (Rayment and Higginson 1992) or after dialysis separation of reactive P into a neutral solution using a Skalar® flow injection autoanalyser at the State Chemistry Laboratory in Werribee, Victoria (Anon 1998b). TP was measured on a subset of 43 ground samples using a modified method of Piper (1950) by digesting 0.5 g of soil with 1 mL of 18M H2SO4 and at least 5 mL conc. HNO3 in a heat block at 150oC for 18 hours. Cooled digest solution was separated from residual soil material using a vortex then decanted and made up to 50 mL with purified water. P was measured colorimetrically at 400 nm using a vanadomolybdophosphoric acid method (Piper 1950; Black et al. 1965; Clesceri et al. 1998).

5.2.3 Methods used to calculate nutrient concentrations in hillslope runoff

A weighted concentration for each event was calculated based on the proportions of subsample collected in each of the three collection vessels (A, B and C). An event was defined as any flow occurring between the times that samples were collected from the field. The event concentration was multiplied by the total measured flow to derive the nutrient load (mass) in each event. The annual flow-weighted mean (FWM) concentration of nutrients in the hillslope runoff was calculated as the sum of the event nutrient loads divided by the total flow volume. This represented an areal and temporal average concentration for runoff from each of the 0.5 ha hillslope plots (White and Kookana 1998).

5.2.4 Statistical analysis

Hillslope runoff study: Bootstrapping (MathSoft Inc 1999) was used to estimate the FWM and SE of the FWM P concentrations for plots 1 and 2. Bootstrapping is a method that resamples multiple sets of observed data, and population mean and distribution are estimated from the resampled data sets. Bootstrapping is useful in giving confidence to means derived from datasets that are not Normally distributed. In this case, bootstrapping was used to simulate the error of the FWM, which could not be calculated in a classical sense. This enabled a paired t-test between FWM P concentrations in runoff from plots 1 and 2 but it was not justified for plots 3 and 4 where there were

129 only a few runoff events. This is because an assumption of bootstrapping is that the observed data is representative of the underlying population and this was not necessarily fulfilled for low sample numbers. Time series analysis for comparing mean runoff P concentrations between plots was no more useful than bootstrapping due to the small number of runoff events in plots 2, 3 and 4. Simulated rainfall study: A Restricted Maximum Likelihood analysis (REML, Genstat® 5v.4.1(Lawes Agricultural Trust 1997)), which copes well with unbalanced data, was used to compare treatment effects in the replicated simulated rainfall experiment.

5.3 Results

5.3.1 Concentrations of P in surface and subsurface flows

Hillslope runoff and soil water

Autumn runoff

The annual FWM concentrations of TP in runoff (TPaut) from the hillslope plots ranged from 0.19 to 1.26 mg P/L (Table 5-2). TP concentrations in a single autumn event (14 April 2000)(Table 5-3) were the highest recorded for all plots. This rain event yielded 39 mm at a maximum intensity of 50 mm/h for 5 minutes, and occurred prior to the start of the growing season (and before fertiliser was applied in that year). Groundcover measured shortly after the storm ranged between 20 and 95 %. Although the maximum flow produced on any plot was 1.1 mm, runoff from the event accounted for 11 and 64% of the annual flow from plots 3 and 4 respectively. The FWM P concentrations in these plots were therefore heavily influenced by the autumn runoff event.

Winter-spring runoff

The FWM TP concentrations from the winter/spring period only (TP) ranged from 0.19 to

0.83 mg P/L (Table 5-2) and were lower than the TPaut means. There was greater than 90% groundcover on all four hillslope runoff plots throughout the winter/spring runoff period in both 1998 and 2000. TP concentrations were 1.7 to 4 times greater in runoff from paddocks receiving the high P fertiliser rate (plots 1, 3 and 4) compared with the low rate (plot 2). All TP concentrations were greater than the current objective 75th percentile concentration of 0.04 mg P/L for rivers and streams in south-west Victoria (Environment Protection Authority Victoria 2001). Above this concentration, there is a risk that adverse ecological effects will occur.

130 The FWM concentrations sometimes differed quite markedly from the arithmetic means (Table 5-2). The arithmetic mean P concentration in runoff from plot 1 reflected the generally lower P concentrations in the frequent surface baseflow (those with <3 mm and no hydrograph peak) events (Table 6-6). Surface baseflow comprised over 80% of the sampled events but contributed only 14% to the total flow from this plot. The standard errors of the arithmetic mean concentrations were very large or absent for plots 3 and 4 (Table 5-2), which reflected the small number of events that occurred on those plots in each year (refer to Chapter 4). The influence of runoff volume on P concentrations in runoff is discussed further in Chapter 6.

FWM runoff TP concentrations increased with the plot mean Olsen P (POlsen, measured in spring)(Table 5-2). The relationship between soil runoff TP and POlsen was stronger when the concentrations in the autumn runoff event were included than when they were not (Figure 5-2). Relationships between runoff P and soil P tests are discussed further in Chapter 7. The concentration of TP measured in the soil solution was low compared with surface runoff and was only slightly higher within the rootzone than below (Table 5-2). Soil-water P concentrations were similar to those measured in rainfall and in a nearby wetland area (Table 5-4). There was little difference between concentrations from the four plots, except for plot 3, which yielded a higher mean concentration within the rootzone. The soil moisture content or cup intake rate (Hansen and Harris 1975), as reflected by the sampled volume, may explain some of the variation in TP concentration because TP concentrations decreased as cumulative volumes of sampled soil water increased (Figure 5-3). However, there was no trend between mean soil water P concentrations and the annual drainage volume, which in 2000 was considerably lower in plot 1 than the other 3 plots (Table 4-10). In addition to measurements of surface flow from the hillslope plots, samples of water that flowed through a low-lying wetland area 0.5 km below the hillslope were collected on 5 occasions, and rainfall samples were collected on 9 occasions in 2000. The mean EC, pH and TP concentration of these samples are described in Table 5-4.

131 Table 5-2: Treatment characteristics and TP concentrations in surface runoff and soil water of Vasey hillslope plots

Plot Treatment Applied P POlsen ------Surface runoff ------Soil water ------FWM FWM TPB Mean TPB TPB range Below rootzone Within rootzone A TPaut Mean TP Mean TP kg/ha mg/kg mg P/L mg/L 1998 1998 1 C 80 10 - 0.34 (0.06)C 0.18 (0.03) 0.05-0.62 N/a N/a 2 A 8 7 - 0.19 (0.02) 0.16 (0.01) 0.15-0.20 N/a N/a 3 B 56 11 - 0.41 0.69 (0.29) 0.40-0.99 N/a N/a 4 C 50 16 - 0.83 0.83 (n/a) 0.83 N/a N/a 2000 1999 & 2000 1 C 25 13 0.39 0.37 (0.07) 0.26 (0.04) 0.06-1.27 0.02 (0.002) 0.03 (0.007) 2 A 6 6 0.23 0.22 (0.03) 0.27 (0.03) 0.14-0.45 0.02 (0.002) 0.03 (0.005) 3 B 25 13 0.98 0.81 1.12 (0.21) 0.23-1.94 0.02 (0.002) 0.09 (0.029) 4 C 25 20 1.26 0.72 0.74 (0.29) 0.58-0.90 0.02 (0.004) 0.03 (0.003) A Includes the runoff event in autumn 2000 B Excludes the autumn 2000 event C Standard errors in parentheses. FWM standard errors were calculated for plots with >6 sampled events using bootstrap analysis Table 5-3: Runoff volume and nutrient concentrations in autumn runoff (14 April 2000)

Plot Runoff Percentage of TP SS total runoff in 2000 mm % mg/L g/L 1 1.08 1.5 1.37 0.38 2 0.07 <1 1.08 0.35 3 0.20 11 2.31 0.14 4 0.06 64 1.60 0.75

132 1.40

1.20

1.00

0.80

0.60

0.40 TP in(mgP/L) runoff 0.20

0.00 0 5 10 15 20 25 Soil Olsen P (0-10cm) (mgP/kg)

Figure 5-2: Relationships between FWM TP or TPaut concentrations and mean POlsen for each hillslope 2 2 plot. Linear trends for TPaut (solid circles and line, R = 0.79, P<0.01) and TP (crosses and dashed line, R = 0.64, P<0.05) concentrations are shown.

Table 5-4: TP concentrations, pH and EC of rainfall and wetland flow sampled in 2000

Rainfall Wetland flow TP pH EC TP pH EC mg/L dS/m mg/L dS/m Mean 0.01 6.37 1.18 0.02 5.22 9.0 Standard Error 0.004 0.22 0.88 0.009 0.61 1.85 No. sample times977544

133 Simulated runoff

Mean TP concentrations in simulated runoff from small plots were similar to those measured in natural runoff from the hillslope plots (compare Table 5-5 and Table 5-2). Concentrations in runoff from individual plots decreased by about 50% within the 60 min period that rainfall was applied (data not shown). A multiple t-test between pairs of treatments was performed to test for treatment differences. Concentration data were log transformed to normalise the distribution for the treatment comparisons and standard errors were predicted using a REML analysis. Similar to the hillslope runoff, TP concentrations increased with increasing POlsen (Figure 5-4) but only TP in runoff from the low fertility onion grass treatment (D) was significantly (P<0.05) lower than from the improved pasture treatments (Table 5-5). Differences in TP greater than 0.4 mg/L were generally required in order to establish significant differences between treatments (Table 5-5).

5.3.2 Concentrations of suspended sediment and N

Mean SS concentrations in simulated runoff ranged from 0.14 to 0.24 g/L across treatments

(Table 5-5). There was no relationship between SS and treatment POlsen (Figure 5-5) and the differences between treatments were not significant (P>0.05)(Table 5-5). The highest SS concentration of 0.96 g/L from an individual plot was measured in treatment C, and can be explained by a higher proportion of exposed bare ground and short grass facilitating higher than average erosion from one of the six plots. This value was not used in the calculation of the treatment mean. In hillslope runoff, the concentration of SS was generally 0.1 g/L or less across all plots in both years, except for in plot 4 in 2000 where it increased to 0.27 g/L (Table 5-6). In 1998, increasing flow and rainfall intensity tended to increase SS concentrations, however, the relationship did not hold well in 2000. Except for plot 4 where hillslope results depended on only one or two events, the concentrations were approximately half those in simulated runoff. Concentrations of SS in the autumn runoff event were larger than winter-spring FWM concentrations (compare Table 5-3 to Table 5-6).

134 1400

1200

1000

800

600

400 Sampled (mL) volume 200

0 0 0.05 0.1 0.15 0.2 TP concentration in soil water (mg/L)

Figure 5-3: Relationship between TP concentration and the cumulative volume of soil water sampled within and below the rootzone across all plots and sample times

Table 5-5: Treatment means, and standard errors in parentheses for P, N and SS concentrations in simulated runoff at VaseyA

Treatment TP TN SS ------mg /L ------g/L

A0.48a (0.07) 2.87 (0.18) 0.16 (0.03)

B0.92a (0.15) 3.68 (0.50) 0.24 (0.02)

C0.77a (0.12) 3.24 (0.48) 0.19 (0.03)

D0.28b (0.09) 2.09 (0.26) 0.14 (0.01) A Subscript notation denotes differences at the 95% confidence level

135 1.8 1.6 1.4 1.2 1 0.8 0.6 0.4 0.2 TP in simulated (mg/L) runoff 0 0204060 Plot Olsen P (0-5cm)(mg/kg)

2 Figure 5-4: Relationship between mean TP concentration in simulated runoff and plot POlsen, R = 0.42, P<0.001

1.2

1

0.8

0.6

runoff (g/L)runoff 0.4

0.2

0 Suspsended sediment in simulated simulated in sediment Suspsended 020406080 Plot Olsen P (0-5 cm)(mg/kg)

Figure 5-5: Suspended sediment concentration in simulated runoff with increasing POlsen

136 Average TN concentrations in simulated runoff ranged between 2 and 3.5 mg N/L with the lowest concentration being measured from the unimproved treatment. There was a trend for concentrations of TN increasing as the treatment mean POlsen increased but the differences between treatments were not significant (P>0.05)(Table 5-5). TN concentrations in runoff from hillslope plots ranged from 1.1 to 4.2 mg/L and were generally higher in 2000 than in 1998 (Table 5-6). Annual - FWM NO3 -N concentrations were less than 1 mg/L across all plots (Table 5-6) and there was no clear relationship with soil P fertility. In the soil water beneath the hillslope plots, TN concentrations were similar to runoff TN concentrations (Table 5-6), despite the absence of a particulate fraction in the sampled soil-water, because of the extraction through porous ceramic. TN concentrations were higher within the rootzone than below, a similar trend to the TP concentrations. Higher TN concentrations were measured beneath the higher fertility plots (plot 1, 3 and 4) than the low fertility plot (plot 2)(Table 5-6). These plots had not received any N fertiliser since 1997. Nitrate-N concentrations were only available for 9 - samples during 2000. In contrast to TP and TN, NO3 concentrations were higher at depth than within the rootzone (Table 5-6). The variability between sampling dates was large with concentrations from - below the rootzone, across all plots, ranging between 0 and 7.5 mg NO3 -N/L. Within the rootzone, - the range was from 0 to 1.5 mg NO3 -N/L.

137 Table 5-6: Mean SS and N concentrations in hillslope surface runoff (FWMs) and soil water with standard errors in parentheses

Plot Surface runoff Soil water – Soil water – within rootzone below rootzone - - - TN NO3 -N SS TN NO3 -N TN NO3 -N ------mg/L ------g/L ------mg/L ------1998 ------1 2.3 (0.3)A 0.12 0.07 (0.02) 2 2.4 0.20 0.08 3 3.0 0.10 0.06 4 1.1 0.10 0.03 ------2000 ------1999 & 2000 ------1 4.2 (0.4) 0.84 0.10 (0.01) 6.1 (2.2) 1.0 (n/a) 4.3 (0.5) 2.2 (0.4) 2 2.9 (0.3) 0.09 0.10 (0.02) 3.8 (0.8) 0.1 (0.1) 3.6 (0.4) 1.7 (0.6) 3 3.6 0.19 0.08 6.1 (1.0) (n/a) 4.4 (0.6) 3.6 (0.9) 4 4.2 N/a 0.27 9.2 (4.7) 0.6 (0.5) 5.4 (0.9) 1.3 (0.8) A Standard errors in parentheses. FWM standard errors were calculated for plots with >6 sampled events using bootstrap analysis

138 5.3.3 Nutrient loads in runoff and subsurface flows

The mean export of P from the site in the years that runoff occurred were 0.08 kg/ha in 1998 and 0.085 kg/ha in 2000. In 1999 no runoff occurred. This was less than 1% of the amount of P applied in fertiliser and represented less than $1 per hectare in financial loss from the system, assuming the total cost of spreading P was $2.35/kg (Vickery Bros. pers. comm.) P loss in runoff from the hillslope plots was strongly influenced by the volume of runoff, and so the largest P loads (equivalent to 0.23-0.26 kg P/ha/yr) were in runoff from plot 1, where the majority off runoff occurred (Figure 5-6). TN (0.01-2.84 kg/ha) and SS (0.5-59.9 kg/ha) losses from each plot were similarly influenced by the amount of flow (Figure 5-6). The majority of P loss in surface runoff occurred during infrequent but large runoff events (Figure 5-7). Storms yielding at least 5 mm of runoff were responsible for 68 and 55% of the total runoff from the site and 74 and 58% of the TP loss in 1998 and 2000 respectively (Table 5-7). These storms accounted for only 4% and 10% of the number of days when runoff occurred in 1998 and 2000 respectively.

139 80 0.30 70 0.25 60 1998 1998 50 0.20 2000 2000 40 0.15 30 0.10 Runoff (mm) 20 10 0.05

0 (kg/ha) load Phosphorus 0.00 High P (1) Low P (2) High P (3) High P (4) High P (1) Low P (2) High P (3) High P (4) Field plot number and treatment Field plot number and treatment

3.5 70 3.0 60 1998 2.5 50 1998 2.0 2000 40 2000 1.5 30 (kg/ha) 1.0 20 0.5 in runoff(kg/ha) 10 Nitrogen load in runoff 0.0 0 Suspended sediment load load sediment Suspended High P (1) Low P (2) High P (3) High P (4) High P (1) Low P (2) High P (3) High P (4) Field plot number and treatment Field plot number and treatment

Figure 5-6: Equivalent loads (kg/ha) of P, N and SS in runoff from hillslope plots

140 60

50 plot1 40 plot2 30 plot3 plot4 20

10 Phosphorus load (g/event) (g/event) load Phosphorus

0 0 5 10 15 20 25 30 Runoff (mm/event)

Figure 5-7: Relationship between P load and volume of surface flow in runoff events from hillslope plots

Table 5-7: P loads in hillslope runoff events yielding at least 5 mm flow

Plot No. events Flow Percentage P load Percentage mm of total kg of TP loss flow 1998 1 2 42 62 0.09 78 2 1 14 82 0.01 85 3 1 11 98 0.02 94 4 A Site total 2 67 68 0.13 74 2000 1 4 37 55 0.08 63 2 1 9 60 0.01 50 3 B 4 C Site total 5 46 55 0.09 58 A One event (3.3 mm) accounted for 100% of TP loss (15 g) B Maximum event size was 1.2 mm accounting for 71% of flow and 58% of TP loss (4.8 g) C Maximum event size was 0.06 mm accounting for 64% of flow and 80% of TP loss (0.5 g)

141 Estimates of the total subsurface drainage plus lateral flow for each plot in 1999 and 2000 (see Chapter 4) were used to estimate the loss of P below the rootzone of each plot using the arithmetic mean soil water P concentration for each year. Between 13 and 27 g P/ha was estimated to have drained from the plots (Table 5-8). Smaller estimated drainage volumes from plot 1, in which a saturated area developed seasonally, resulted in lower subsurface P losses than in the better drained plots 2, 3 and 4.

Table 5-8: Estimates of total P flux in subsurface drainage from hillslope plotsA

Plot Drainage + subsurface Mean soil water P TP (g/ha) lateral flow (mm) concentration (mg/L) 1999 2000 1999 2000 1999 2000 1 66 86 0.02 0.02 13 17 2 83 134 0.03 0.02 25 27 3 82 145 0.03 0.01 25 15 4 62 127 0.03 0.02 19 25 A Soil water P concentrations were measured in 1999 and 2000 only

5.3.4 Factors influencing P losses

Pasture treatment characteristics that can influence the volume of runoff, degree of erosion and concentration of P in surface runoff were quantified in the hillslope and simulated runoff plots to assess whether there was an increased risk for P loss in runoff from high versus low fertility pastures. The characteristics included pasture height, density and groundcover, and pasture and deposited dung P contents. The spatial distribution of stock camps, deposited animal dung and POlsen was also assessed to identify areas of high P fertility that coincided with runoff source areas.

Ground surface conditions

Pasture height, mass and groundcover on rainfall simulator plots were measured to reveal any differences between pasture treatments that may explain differences in runoff volume and erosion. However, there were no significant differences in runoff volume (range 17 – 24 mm) or SS concentrations between treatments and this was consistent with the high degree of groundcover (>90%) across treatments (Table 5-9). The high groundcover reflected a nearly or completely closed pasture canopy and/or the ground surface being covered with attached or detached plant or faecal material. Some significant differences in pasture mass and height between treatments occurred, mainly associated with accumulation of poor quality onion grass in the unimproved treatment D (Table 5-9), but this caused no measurable effect on runoff volume or erosion.

142 Table 5-9: Treatment mean groundcover and runoff characteristics of rainfall simulator plotsA

Treatment Units A B C D

Dry Matter t/ha 1.5a 1.7ab 2.0b 3.4c Sward height cm 2.0a 3.0a 3.8a 10.1b Groundcover % 96 98 99 100 RunoffB mm 21 24 19 17 A Subscripts denote significant differences using a multiple t-test, P<0.05 B Rainfall applied was 48 mm

There was at least 85% groundcover in all four hillslope runoff plots throughout the winter/spring runoff period (data not shown). However, at the autumn 2000 rain event, groundcover was as low as 25% in some areas (Figure 5-8). Groundcover levels after the autumn storm were overestimated because clover germination had occurred in the 5 days between the storm and the bare ground assessment, and because stock had been removed from the plots for 5 weeks for joining. During this autumn storm, detached surface material was washed into ‘debris-dams’. The presence or absence of debris-dams was visually assessed at each permanent soil nutrient sample point 5 days after the storm. No dams were observed in areas with complete groundcover, whereas dams were observed in all but one measured location which had greater than 30% bare ground, indicating surface water movement was enhanced at the lower levels of groundcover. The pattern of groundcover in autumn 2000 was influenced by the grazing behaviour of the sheep, with 80 to 100% cover in the lower half of the two set stocked plots and as little as 25% in camp areas at the top of plot 3. The rotationally grazed plot 1 had between 25 and 35% bare ground across its entire slope length and plot 4 had quite variable cover, with about half the plot having between 25 and 65% bare ground (Figure 5-8).

Spatial distribution of stock camps

Stock camps were observed at the top of plots 2 and 3, in the upslope corners of the subsections of plot 4 along the fence shared with plot 3 and in the upper three subsections of plot 1, again along the fence shared with plot 2. The camp area in plot 3 was the largest, spanning nearly one quarter of the whole plot. The set stocked plot tended to have a larger camp area with less intense distribution of dung at the top of the hillslope whereas smaller camp areas developed in the rotationally grazed subplots. The camp areas were reasonably well matched by the distribution of dung measured in April 2000 (Figure 5-9).

143 11160

11140 % bare ground

11120 75

11100 65

55 11080

45 11060 northing (m) 35

11040 25

15 11020

5 11000

10980

960 980 1000 1020 1040 1060 1080 1100 1120 easting (m)

Figure 5-8: Percentage bare ground across hillslope plots after storm in April 2000

144 11160

11140

11120 % dung cover

11100 100

90 11080 80

11060 70 northing (m)

60 11040 50

11020 40

30 11000 20

10 10980 0 960 980 1000 1020 1040 1060 1080 1100 1120 easting (m)

Figure 5-9: Location of stock camps (dashed lines) and percentage dung cover across hillslope plots in April 2000. Elevation (m) contours also marked and labelled

145 Herbage, dung and soil P content

The TP concentration in dung and pasture herbage after rainfall was measured in samples collected from the 45 simulated rainfall plots. TP concentrations (%) in both dung and herbage samples were higher in the high fertility treatments (B and C) when compared with the two low fertility treatments (A and D)(Table 5-10). The unimproved pasture (treatment D) had significantly (P<0.05) lower dung and herbage P levels than the high fertility treatments. The P concentration in herbage and dung tended to increase with POlsen (Figure 5-10 and Figure 5-11). The average P concentration in herbage for the high fertility treatments was generally above the range of P concentrations of 0.18 - 0.32% that are critical for maintaining growth of phalaris and subterranean clover (Reuter and Robinson 1997) (Table 5-10, Figure 5-12), and there was little increase in herbage P concentration beyond a POlsen (0-5 cm) of about 27 mg/kg (Figure 5-10). Figure 5-12 shows that at the time of sampling (late winter), herbage mass was within the spring target range of 1200-2500 kg/ha (Chapman et al. 2003) in about half of the improved pasture plots.

The POlsen (0-5 cm) of the rainfall simulator plots was significantly higher (P<0.05) in the high fertility treatments B and C than the low fertility treatments A and D (Table 5-10). A difference of at least 10 mg of NaHCO3 extractable P per kg soil was required for significant treatment differences. Treatment C was applied to two paddocks whose mean POlsen (0-5 cm) were 26 and 36 mg/kg. This large contrast was not reflected in the mean TP concentrations in runoff from those paddocks (0.66 and 0.69 mg/L respectively), however, resulting in less difference between treatments C and A in runoff P concentrations (Table 5-5) compared to

POlsen. The target POlsen (0-10 cm) for sheep pastures is about 15 mg/kg (Cayley et al. 1998), which corresponds to about 24 mg/kg over the 5 cm depth. The mean POlsen in the higher fertility treatments B and C met this target (Table 5-10).

Table 5-10: Treatment mean soil, herbage and dung P in plots used for simulated rainfall with standard error in parenthesesA

Treatment A B C D B POlsen (0-5cm) mg/kg 14a (1) 34b (3) 31b (4) 8a (1) Herbage P % 0.30a (0.01) 0.37ab (0.02) 0.41b (0.01) 0.16d (0.02) Dung P % 0.69ab (0.05) 0.94a (0.09) 1.06a (0.07) 0.26b (0.14) A Treatment means compared using a multiple t-test. Subscript notation denotes differences at the 95% confidence interval. SE of differences estimated using REML analysis, Genstat® B The variance in POlsen increased with the value of the mean so data were loge transformed for treatment comparisons to normalise the distribution of residuals

146 0.6

0.5

0.4

0.3

0.2

0.1 Herbage P concentrationHerbage (%) 0 0 10203040506070

Plot P Olsen (0-5 cm)(mg/kg)

Figure 5-10: Herbage P concentration (%) with increasing POlsen (0-5cm)(mg/kg). Trendline shows fitted regression curve where Herbage P concentration = 0.41-0.55*0.89POlsen P<0.001, R2=0.63

1.6 1.4 1.2 1.0 0.8 0.6 0.4

Dung PDung concentration (%) 0.2 0.0 0 10203040506070

Plot P Olsen (0-5cm)(mg/kg)

Figure 5-11: Dung P concentration (%) with increasing POlsen (0-5cm)(mg/kg). Trendline shows fitted regression curve where Dung P concentration = 1.10-1.14*0.92POlsen, P<0.001, R2=0.38

147 16 14 Treatment A 12 Treatment B 10 Treatment C 8 Treatment D 6 Min and max DM targets 4 Range of critical herbage P concentrations 2 Pasture Dry Matter (t/ha) Matter Dry Pasture 0 0 0.2 0.4 0.6 0.8 Herbage P concentration (%)

Figure 5-12: Relationship between pasture dry matter production (t/ha) and herbage P concentration (%) of quadrat samples from simulated rainfall plots at Vasey

Across all sample times, hillslope plots 1 and 3 had similar mean POlsen (Table 5-11).

The mean POlsen in plot 2 was significantly lower (P<0.05) and plot 4 significantly higher (P<0.05), than plots 1 and 3 for most of the sample times (Table 5-11). There was no or little change in the mean POlsen in plot 2 over the six sample times (Table 5-11), whereas the high fertility plots increased by 2 to 6 units.

Table 5-11: Mean POlsen in hillslope runoff plots on six occasions from 1998 to 2000, with standard errors in parenthesesA

Sample time Plot 1 Plot 2 Plot 3 Plot 4

June 1998 5.2a 6.1a 9.0b 17.8c (0.2) (0.3) (0.6) (1.4) September 1998 9.0a 6.0b 10.8a 15.5c (0.9) (0.2) (1.0) (1.0) March 1999 7.3a 5.4b 9.3c 14.0d (0.6) (0.2) (0.5) (1.1) September 1999 10.1a 6.2b 11.4a 19.2c (0.7) (0.3) (1.0) (1.3) April 2000 10.5a 5.8b 14.5c 16.4c (0.9) (0.4) (0.9) (1.2) October 2000 13.0a 6.1b 13.1a 19.8c (1.2) (0.3) (1.0) (2.1) A Data were transformed using the natural logarithm to normalise the distribution of residuals. Plot means were compared using a multiple paired t-test. Standard errors of differences were estimated using REML, Genstat. Subscripts indicate significant differences between plots at each sample time (P<0.05)

Spatial distribution of extractable soil P

Vertical distribution

148 The POlsen decreased in value from the surface to a depth of 10 cm in all the hillslope runoff plots (Figure 5-13). Olsen P values for more than one depth interval were calculated by weighting the values obtained for each individual depth interval. The POlsen measured in the top 2 cm of field soil was on average 2.5 times greater than the depth-weighted 0-10 cm values (Figure 5-13). There was slightly less vertical contrast in the low fertility plot (2.38 times) and slightly more in the high fertility plots (2.62 times). Weighted 0-5 cm values were approximately 1.6 times greater than the weighted 0-10 cm values.

Olsen P (mg/kg) 0 102030405060 0

1

2

3

4

5

6 Depth from surface (cm) surface from Depth 7

8

Figure 5-13: Change in POlsen with soil depth for high (solid line) and low (dashed line) fertility hillslope plots and depth-weighted 10 cm mean values for high (□) and low (о) fertility treatments. Error bars indicate ± standard error

Horizontal distribution

Spatial distribution maps developed by interpolation procedure for POlsen data collected at 102 permanent sampling points in the hillslope plots illustrated changes before and after fertiliser was applied, and between the years 1998, 1999 and 2000 (Figure 5-14 and Figure 5-15). There was greater variation in soil P levels in plots receiving the high rates of P application (plots 1, 3 and 4) than that receiving the low rate of 8 kg P/ha (plot 2).

The spatial distribution of POlsen (0-10 cm) across the four hillslope runoff plots reflected the timing and rate of P fertiliser application (Table 3-6). There was relatively uniform

149 accumulation of soil P across each plot, which reflected broadcast P applications. By October 2000, however, some grazing effects were also recognised. Consistently higher P levels in the upper half of the low fertility, set stocked plot 2 when compared to the lower half coincide with the upslope stock camp area. Some variation in the plot POlsen was due to individual ‘hotspots’. Hotspots refer to points where POlsen levels were much greater than the plot mean, such as at a single sample point in the lower region of plot 3 (Figure 5-15b). The hotspots sometimes occurred where stock camps had developed and near the base of the tree between plots 3 and 4. To further investigate the accumulation of nutrients in stock camp areas, the average

POlsen of sample points within camp areas in October 2000 was compared using a t-test to points in the remainder of the plot. In each plot the POlsen in camps were higher than that in non-camp areas, and the differences were significant (P<0.05) for plots 2 and 3 (Table 5-12). Some camp areas in the rotationally grazed plots were too small to encompass any of the arbitrarily spaced sample points (eg plot 4). This would have reduced the power of this sampling program to pick up differences in nutrient status.

Table 5-12: Mean POlsen in stock camp and non-camp areas in October 2000 with standard errors in parenthesesA

POlsen (mg/kg) Plot Camp Non-camp 1 14.3 (2.1) 12.7 (1.3) 27.3a (0.7) 5.6b (0.3) 3 16.8a (2.2) 11.4b (0.9) 4 28.2 (6.7) 19.0 (2.1) A Subscripts denote significant differences (P<0.05) for each plot

150 a) June 1998 b) September 1998

11160 11160

11140 11140

11120 11120

11100 11100

11080 11080

11060 11060 Northing (m) Northing (m)Northing

11040 11040

11020 11020

11000 11000

10980 10980

960 980 1000 1020 1040 1060 1080 1100 1120 960 980 1000 1020 1040 1060 1080 1100 1120 Easting (m) Easting (m)

c) March 1999 d) September 1999

11160 11160

11140 11140

11120 11120

11100 11100

11080 11080

11060 11060 Northing (m) Northing (m) Northing

11040 11040

11020 11020

11000 11000

10980 10980

960 980 1000 1020 1040 1060 1080 1100 1120 960 980 1000 1020 1040 1060 1080 1100 1120 Easting (m) Easting (m) Olsen P (mg/kg) 0510 15 20 25 30 35

Figure 5-14: Distribution of soil (0-10cm) POlsen (mg/kg) measured at 102 permanent sample points in hillslope plots in a) June 1998, b) September 1998, c) March 1999 and d) September 1999

151 a) April 2000 b) October 2000

11160 11160

11140 11140

11120 11120

11100 11100

11080 11080

11060 11060 Northing (m) Northing latitude north (degrees) north latitude 11040 11040

11020 11020

11000 11000

10980 10980

960 980 1000 1020 1040 1060 1080 1100 1120 960 980 1000 1020 1040 1060 1080 1100 1120 Easting (m) longitude east (degrees)

Olsen P (mg/kg) 051015 20 25 30 35

Figure 5-15: Distribution of soil (0-10cm) POlsen (mg/kg) at 102 permanent sample points in hillslope plots in a) April 2000 and b) October 2000

152 5.4 Discussion

5.4.1 P concentrations in surface runoff and soil-water

P concentrations and their environmental implications

The concentration of P in runoff, or drainage water, from pastures can determine the rate of algal growth in receiving waterbodies where these flows contribute a significant fraction of the surface water supply. The FWM annual concentrations of TP in runoff from the Vasey hillslope plots ranged from 0.19 – 0.83 mg P/L and exceeded the desirable upper limit of 0.04 mg P/L for freshwater streams in the Western Plains (Environment Protection Authority Victoria 2001). When other factors such as light, temperature and flow residence time are non- limiting, concentrations as low as 0.02 mg P/L can lead to algal growth (Correll 1998). The runoff P concentrations could therefore be problematic if the runoff reached a water body. FWM TP concentrations in runoff at Vasey were also higher than the mean TP concentration measured in rainfall of 0.01 mg/L (Table 5-4), confirming that rainfall was enriched with P after coming into contact with the pasture surface. This was not surprising as soil, dung and plants all represented potential sources of P at the soil surface. The FWM more accurately represented the P concentrations of the dominant runoff events than the arithmetic mean, which was strongly influenced by frequent but less significant low flow events. The FWM TP concentrations in winter/spring runoff from the hillslope plots were similar to, or lower than those found in runoff from sheep (0.45 – 2.0 mg P/L), cattle (0.37-1.02 mg P/L) and dairy (0.3-5.3 mg P/L) pastures in southern Australia (Tham 1983; Nash and Murdoch 1997; Fleming and Cox 1998; Stevens et al. 1999). In contrast to environments where runoff occurs throughout most of the year (McColl and Gibson 1979; Cooke 1988; Heathwaite and Dils 2000; Tunney et al. 2000a), seasonal trends in P concentrations in runoff were not important at Vasey, due to the lack of runoff and drainage during summer and autumn. In the one, small runoff event (1.4 mm) that occurred in autumn, up to 95% of some areas of the ground surface was bare, and both SS and TP concentrations in the runoff were higher than during winter/spring (Table 5-3). Higher TP concentrations were attributed to higher levels of both particulate and dissolved forms of P (data not shown). Elevated dissolved P levels may have been caused by a build up of decaying organic debris, dry soil conditions and a lack of plant uptake of P over summer, which would have increased the availability of labile P at the soil surface for mobilisation in runoff (McColl and Gibson 1979; Sharpley 1981; Pote et al. 1999a). The higher SS and particulate P levels suggested there was increased erosion from this event. The low runoff volumes may have also concentrated the P.

153 Average P concentrations in the soil solution both within and below the rootzone of all the hillslope plots (<0.09 mg P/L, Table 5-2) were lower than in surface runoff, reflecting the high P sorption capacity (Langmuir Pmax >1500 mg/kg) and the low P status (POlsen <4 mg/kg, (Cox et al. 1998)) of soil below a depth of 25 cm. The soil solution concentrations were at the low end of the range found in the soil solution in other agricultural soils (0.01 - 3.0 mg P/L (Cox and Ashley 2000; Frossard et al. 2000; Heathwaite and Dils 2000; McDowell and Condron 2000). It is possible that the P concentrations measured using the ceramic cups underestimated the soil solution P concentration due to adsorption of P to the ceramic material (Hansen and Harris 1975; Nagpal 1982). The cups were pre-washed with HCl and de-ionised water, however, which Grover and Lamborn (1970) found led to between 70 and 100% recovery of P from solutions of either 0.25 and 1 mg P/L. The equilibrium P concentration of the B horizon at Vasey (0.001 mg/L, data not shown) was substantially lower than the mean soil solution P concentration, suggesting some macropore flow from surface soil layers may have delivered higher P concentrations to the subsoil. Because the average P concentration in the soil solution below the rootzone was less than 0.04 mg P/L (Table 5-2) it was unlikely that drainage from the high or low fertility pasture that contributes to surface water would contribute to eutrophication, a finding similar to that of Cooke (1988) for a New Zealand soil. The nutrient concentration of subsurface flows in naturally drained soils are also less sensitive to single events such as heavy rain after fertiliser application than surface flows (Haygarth and Jarvis 1996). The concentrations of P in subsurface flows at Vasey therefore presented a lesser risk of eutrophication than surface runoff concentrations. Surface water from a wetland area downslope of the hillslope site was also low in P, and the strongly acidic and highly saline nature of this water (Table 5-4) indicated a strong groundwater contribution to this flow. Discharging groundwater that has a low P concentration could therefore partially dilute any eutrophic surface runoff reaching streams. However, streamflow for the Dundas Creek, recorded 20 km from the Vasey site (DSE 2003), and runoff recorded at Vasey (Table 5-7) were dominated by large surface runoff events rather than baseflow from groundwater. Concentrations of P in runoff from hillslopes surrounding streams would therefore have a stronger influence on stream P concentrations than groundwater. High TP concentrations in surface runoff from pastures in the Glenelg-Hopkins catchment could be of concern because 67% of the catchment is used for dryland sheep and cattle farming. Results from this study indicate that dilution of streamflow with runoff from areas of low fertility and low erosion potential would be important for maintaining the quality of water in the catchment. However, an increase in pasture improvement, fertiliser application and stocking rates to boost wool production and improve groundcover could reduce the proportion

154 of well-managed low fertility hillslope areas within the Glenelg Hopkins pastoral region, leading to an overall reduction in surface water quality. Perhaps of more immediate concern regarding high runoff P concentrations is its effect on the P status of farm dams, which may not receive groundwater or other runoff to dilute surface runoff contributions. In a runoff study at Werribee, Tham (1983) found similar P concentrations in farm dams and in the runoff from sheep-grazed pastures adjacent to the dams. The P concentrations were above EPA guidelines (Environment Protection Authority Victoria 2001), which adds weight to the importance of managing P concentrations in runoff from pastures.

P concentrations in runoff from high and low fertility pastures

P concentrations measured in simulated runoff increased significantly with POlsen (P<0.05, also see Chapter 6), and there was a significant difference (P<0.05) in the concentration of TP in runoff from the unimproved (treatment D) and improved pasture treatments (treatments A, B and C) (Table 5-5). The treatment effect was attributed to a combination of increased soil, dung and herbage P contents in the high fertility pastures, associated with the higher P fertiliser and stocking rates compared with the low fertility pasture. Differences in TP concentrations in runoff from the hillslope runoff plots were consistent with these findings, with the high P fertility treatments having higher annual FWM TP concentrations in runoff (0.34 – 0.83 mg P/L) than the low P fertility treatment (0.19 – 0.22 mg P/L). The difference in mean runoff TP concentrations between the low and high fertility hillslope plots was small considering the 10-fold difference in initial P application rates. However, direct entrainment or dissolution of fertiliser granules was assumed to be unimportant for these pastures because fertiliser was applied 3 to 4 months prior to runoff occurring which would have allowed sufficient time for fertiliser to equilibrate with the soil (Barrow and Shaw 1975). A difference in TP concentration of at least 0.4 mg/L in simulated runoff between treatments was significant (P<0.05), which added weight to the hypothesis that concentration differences in the hillslope runoff were also attributable to treatment differences. A t-test between the bootstrap FWM concentrations of P in runoff from plots 1 and 2 indicated, however, the difference between these concentrations was not significant at the 95% confidence level. It is likely the treatment effect for plot 1 was masked by the large annual runoff volume from this plot diluting the P concentrations. The effect of flow volume on P mobilisation in runoff is discussed further in Chapters 6 and 7.

155 5.4.2 N concentrations and loads in runoff and soil-water

TN loads (0.01-2.84 kg N/ha/yr) in runoff from the Vasey hillslope plots were similar to or lower than loads (0.62 – 9 kg N/ha/yr) measured in runoff from other pastures in south- - eastern Australia (Costin 1980; Tham 1983; Nelson et al. 1996). TN and NO3 -N concentrations in runoff were within the ranges (up to 6.4 mg TN/L and 1.79 mg nitrate-N/ L) measured from other pastures (Lambert et al. 1985; Stevens et al. 1999; Cox and Pitman 2001). Nitrate-N comprised up to 21% of the TN lost in runoff over the four plots on an annual basis. Consistent - with the findings of Ridley et al. (2001) for perennial pasture in north-east Victoria, NO3 -N concentrations were lower in surface runoff than either shallow or deep subsurface drainage (see Table 5-6). Nitrate concentrations measured in surface runoff and drainage were well below the - maximum acceptable level for potable water of 11.3 mg NO3 -N/L (World Health Organisation - 1996), however, NO3 and TN concentrations were above the guideline trigger levels of 0.01 and 0.44 mg N/L respectively, for freshwater lowland reservoirs (Anon 2000b). The concentrations of N in runoff at Vasey would therefore be of concern if not attenuated by dilution, immobilisation or denitrification prior to reaching a stream or reservoir where algal growth was limited by N, rather than P.

5.4.3 P loads in surface and subsurface flows and their environmental implications

P loads in runoff are an important measure of the total amount of P that may become available for algal growth in the long term in receiving water bodies (Correll 1998). An average P load in surface runoff from the Vasey site of 0.08 kg P/ha/yr was substantially lower than loads measured from dairy pastures in Gippsland (annual equivalents of 1.9, 5.7 and 9.7 kg P/ha between 1994 and 1996 (Nash et al. 2000)) reflecting both the lower P concentration and annual volume of runoff at the Vasey site. Loads from the individual hillslope plots (<0.26 kg P/ha) were also at the low end of the range measured in runoff from other pasture sites in south- eastern Australia (0.1 to 2.6 kg/ha) (Costin 1980; Nash and Murdoch 1997; Fleming and Cox 1998; Nexhip and Austin 1998; Stevens et al. 1999). The TP lost in runoff was equivalent to less than 1% of the P applied as fertiliser, which represented a negligible financial burden on the production system. This is commonly the case for runoff P losses from extensive grazing systems (McColl et al. 1977; Stevens et al. 1999). In Ireland, losses of less than 0.25 kg P /ha/yr were estimated as being acceptable for good water quality (Tunney et al. 2001). However, acceptable limits for loads will vary between environments depending on the vulnerability of receiving waterways to eutrophication, as well as the potential uses for the water. Water quality guidelines based on total P and N loads have

156 not been widely developed in Australia. This is partly because the data required for predicting algal biomass from nutrient load data using an approach such as the Vollenweider model is not always available (Anon 1999). Furthermore these models may not be appropriate for Australian conditions because the assumptions of rainfall exceeding evaporation and lake outflow being equal to lake inflow are often not met (Harris 1994). Guidelines tend, therefore, to focus on acceptable nutrient concentrations rather than loads (Environment Protection Authority Victoria 2001). The mean P loss from the 2 ha site of 0.08 kg P/ha was close to the estimate used (0.09 kg P/ha) by the Glenelg Hopkins Catchment Management Authority and the Department of Natural Resources and Environment in a Catchment Management Support System (CMSS) model that calculated the contribution of runoff from dryland sheep and cattle areas to P loads in rivers (Wagg 1999a). The objective of their study was to identify priority regions for implementation of nutrient management action plans. The average N load at Vasey was 0.71 kg/ha, which was lower than the CMSS estimate of 1.7 kg/ha. The variation in nutrient loads in runoff across the hillslope at Vasey highlights, however, that the CMSS model underestimated potential losses from discrete, hydrologically active areas of the landscape and overestimated losses from well-drained and flat tableland areas. Without some assessment of landscape hydrology, the priority areas for nutrient management were unlikely to have been well defined. When comparing P loads reported by different studies it is important to note the variable runoff volumes (Nelson et al. 1996) and also the scale of measurement. The impact of a particular management practice, such as spreading fertiliser, on runoff water quality may be large for a small hillslope but depending on the percentage of land affected, the impact on the larger scale catchment may be less significant (Prairie and Kalff 1986; Lennox et al. 1997). In this study, P loads represented losses from the hillslope but did not necessarily reflect the amount of P that actually reached surface water. This is because the load may have been altered by processes such as infiltration of runoff water, further entrainment of P or adsorption of P to subsoil and channel sediments beyond the paddock boundary. In any catchment the impact of edge-of-field P losses would increase with proximity and hydrological connectivity of the pasture to a stream or dam where eutrophic effects are realised. The limitations and merits of experimental methods used in this study are discussed further in the section ‘Experimental approaches: strengths and limitations’. The amount of P loss estimated in combined drainage and subsurface flow at Vasey of 0.01- 0.03 kg P/ha/yr was small and at the low end of the range measured from subsurface pathways under pasture in South Australia (0.01 to 0.148 kg P/ha)(Fleming and Cox 1998; Stevens et al. 1999; Cox and Pitman 2001). In the years when surface runoff occurred, more P was lost in surface than subsurface pathways. However, for plots 3 and 4, which were well drained, and in 1999 when no runoff occurred, subsurface drainage and lateral flow were the

157 important pathways for P loss. Concentrations of P in the soil solution, and hence loads of P in drainage waters, may have been affected by sorption of P to the ceramic cup material and cups are unlikely to accurately reflect the movement of P along macropore pathways (Kung and Donohue 1991; Ryan and Noonan 1995). In general, however, surface runoff was considered the most important pathway for off-site P movement because of the higher P loads and concentrations and because surface pathways have the potential to contribute flows to streams and reservoirs more quickly than more tortuous subsurface pathways. This suggests that reducing the size of the VSA through subsurface drainage may reduce total P losses. However, the potential for adverse effects such as increasing the spread of saline groundwater in this environment would need to be considered along with the costs of drainage installation.

5.4.4 Factors affecting P loss in runoff

The amount of P lost in runoff is influenced by soil properties, landscape hydrology, environmental conditions, pasture management and by the amount and form of P at the runoff source (White and Kookana 1998). In this section, the influence of volume and distribution of runoff at Vasey on P loads is discussed. The spatial distribution of soil P, dung and groundcover at the soil surface and the P content of pasture plants and deposited dung are also discussed in relation to the risk of mobilisation of P in runoff from the high and low fertility sheep pastures. The forms of P identified in runoff are discussed in Chapter 6.

Volume and temporal distribution of runoff

The volume of runoff, rather than its P concentration, was the primary determinant of the total loss of P in runoff from the Vasey site. This is because flow volumes varied by orders of magnitude more than concentrations. Therefore, when runoff volumes are high, large loads of P can be lost from pastures even when the concentrations of P in runoff are low. This trend was reflected by the larger total loss of P from plot 1 where the mean annual TP concentration was approximately 0.35 mg/L compared with plot 4, where the mean TP concentration reached 0.83 mg P/L but runoff was minimal. The dominant influence of the magnitude and frequency of flow on total P loss is consistent with findings from other studies (Burwell et al. 1975; Haygarth and Jarvis 1997; Stevens et al. 1999; Fleming and Cox 2001). The loads and concentrations of P in runoff from sheep pastures at Vasey were very similar to those in Werribee, Victoria (Tham 1983), reflecting the similar rainfall, landscape and landuse. The influence of site hydrology also meant that in 1999, when no runoff occurred, there was no P loss via this pathway. Only 6 mm of flow occurred in 1999 in the nearby Dundas Ck, which drains an area of approximately 211 km2, whereas in the years when runoff occurred at Vasey, there was at least 20 mm of flow in the creek. Creek flow was less than 20 mm in 3 of the last 12 years. It is therefore unlikely that surface runoff occurred at Vasey in these 3 years either.

158 The within-year pattern of P losses in runoff at Vasey reflected the episodic nature of rainfall in the temperate climate of south-west Victoria. At least 58% of the TP load was lost in no more than 5 runoff events in both 1998 and 2000. Runoff from these events comprised at least 55% of the total annual flow. In Gippsland, in eastern Victoria, the pattern of P loss was similar with 72% of the total P exported from a rain-fed dairy pasture resulting from 8 storms, which accounted for 56% of the total flow over three years (Nash et al. 2000). In New Zealand, 31% of P and 48% of nitrate was lost from a hill pasture in a single flood with less than 20% of P losses from forested and pasture catchments being exported during low flows (McColl et al. 1977). In another pasture catchment, approximately 40% of the 1.3 kg P/ha exported in one year was lost during four storms (Cooke 1988). Similarly, in an erosion study, infrequent, large events contributed the majority of flow from pastures with good ground cover (Lang 1979). The impact of this episodic pattern of P and N loss on catchment waterways is not well understood. Ryden et al. (1973) suggest eutrophication problems may be exacerbated when large amounts of P are exported during short periods, particularly if the algal biomass is in a growth phase. Nutrient loadings from storms can also continue to have ecological impacts for up to 10 years (Harris 1994). However McColl et al. (1977) suggest that a supply of P well in excess of algal population requirements may be sequestered elsewhere into immediately unavailable forms or flushed out of the river system, which may shift the nutrient enrichment problem downstream (Peter Davies, pers. comm.). Surface flows in summer and autumn are uncommon in south-west Victoria and therefore P loss in runoff is negligible during this period. At Vasey, less than 11 g of P was lost in the only runoff that occurred outside the winter/spring period. However, P that accumulates in stream and reservoir sediments during winter/spring flows can be sufficient to support the growth of nuisance weeds and blue-green algal blooms during the warmer months, when other predisposing conditions such as water residence time, turbidity and temperature are non- limiting.

Groundcover and suspended sediment

Groundcover reduces runoff and soil loss by intercepting and absorbing the energy of rainfall impact, impeding the flow and resisting the erosive force of runoff and by increasing infiltration. Where groundcover cover is less than 70%, there is a strong link between groundcover, runoff and sediment losses, however the relationship weakens when the range is cover is small, and at extremely high rainfall intensities (Lang 1979; Costin 1980; McIvor et al. 1995; Nash and Murdoch 1997). A study of cropped and grazed catchments in the USA showed that regardless of landuse, there was negligible erosion when groundcover was rated as ‘good’, whereas when there was little or no groundcover up to 50% of annual sediment loss occurred in a few major storms (Olness et al. 1975).

159 Groundcover on the Vasey hillslope plots was greater than 90% during the winter/spring runoff period. Consequently the FWM concentration of SS in runoff was similar for all four plots (Table 5-6). The SS concentrations (0.03-0.27 g/L) were higher than those measured in runoff from a dairy pasture on a Dermosol in eastern Victoria (Nash et al. 2000) and lower than the total solids measured in runoff from tropical pastures in northern Australia (0.3-0.9 g/L, McIvor et al. (1995)), and cultivated fields overseas (0.3 – 1.3 g/L, Yli-Halla et al. (1995)). Groundcover was also >90% in the simulated rainfall plots and despite there being some differences in pasture height and dry matter (Table 5-9), there was no significant difference in runoff volume or SS concentrations (range 0.14 - 0.24 g/L) between treatments (Table 5-5). After an autumn storm that occurred when groundcover was as low as 25% in some areas, there was visual evidence of surface water movement and erosion by way of debris dams. Debris dams always occurred in areas with at least 30% bare ground, supporting Lang’s (1979) findings that greater than about 70% groundcover is needed to minimise erosion. Whilst runoff was negligible (1.4 mm) from this rainfall event (39 mm), due to the dry antecedent soil conditions, SS concentrations were greater than in winter/spring runoff (Table 5-3 and Table 5-6). This demonstrates the importance of controlling grazing to maintain groundcover throughout the year. The pattern of groundcover also showed that under a moderate grazing pressure (14 ewes/ha), set-stocked sheep preferentially grazed upslope during summer and autumn, providing a convenient pasture buffer strip at the base of slope for intercepting runoff and sediment. However, in other studies, set-stocked improved pastures often have less groundcover during summer compared with rotationally grazed pastures due to poor recovery of the perennial grass component between grazing events (Waller et al. 2001a; Warn et al. 2001). The average SS load of 0.016 t/ha.year at Vasey was low compared with erosion from other well-managed perennial pastures (0-0.375 t/ha/yr, (Adamson 1976; Costin 1980)) and cropped soils (1-8 t/ha, (Edwards 1987)) in south-eastern Australia. The low SS load probably also reflected the generally low intensity of rainfall (<25 mm/h) in south-west Victoria. Close to 100% groundcover during the winter/spring runoff season reflected well-managed pastures at the commercial scale and was considered optimal for minimising runoff and sediment loss and hence P loads.

P in dung and herbage

Pastures are considered a rich source of P for mobilisation by surface runoff because nutrient-rich soil, herbage, litter and dung material accumulate at the surface (Sharpley and Syers 1976; Sharpley 1996). Up to 85% of the P in ground samples of live and hayed-off pasture plants (Bromfield and Jones 1972; Gillingham et al. 1980) and 27% of P in dung

160 material (Bromfield 1961) is soluble in water and can potentially be leached into runoff water (Sharpley 1981; Sharpley and Moyer 2000). P concentrations in leachate from intact plant material are rarely measured directly but concentrations of up to 150 mg P/L have been measured in leachate from hay (Bromfield and Jones 1972) and between 0.019 - 0.030 mg P/L from cotton, sorghum and soybean (Sharpley 1981). The P concentration of herbage and dung, and the proportion that is soluble in water, tends to increase with increasing levels of plant available soil P (Bromfield and Jones 1972; Gillingham 1980; Rowarth et al. 1988). At Vasey, the P concentration in herbage (treatment averages 0.16 – 0.41 %) and deposited dung (treatment averages 0.26 – 1.06 %) was in excess of all but the most fertile soils. It is likely therefore that P leached from herbage and dung material would have contributed to P in runoff. An increase in dung and herbage P content with increasing POlsen (Figure 5-10 and Figure 5-11) suggested there was greater pool of potentially mobile P in high fertility pastures than low (Holt et al. 1970). At the time of sampling, the high fertility treatments B and C had adequate soil P levels for maintaining pasture growth (POlsen (0-10 cm) > 15 mg/kg, (Cayley et al. 2002)) and herbage P concentrations were within or above the target range for these pasture species (Figure 5-12). This suggests that increasing the available P content of the soil further would increase the risk of P mobility in runoff without necessarily increasing pasture production. The high dry matter produced at the lower soil and herbage P contents reflected a large component of unpalatable annual grasses in the unimproved pasture treatment (treatment D), which is undesirable for livestock production. Plant and dung material larger than 3 mm diameter that floated in the runoff did not pass through the runoff sample splitter and was therefore not measured in the SS fraction of runoff. Pastures readily trap debris such as this, and it was estimated from field observations that floating debris contributed less than 2% of the SS load in runoff at Vasey. Similarly the floating debris fraction was negligible in runoff from a large (88 ha) catchment in New South Wales (Costin 1980). However, the P and N contents of floating debris can be high (0.6 %P, 1.6%N) (Costin 1980), so where this material does get washed into surface waterways it may be a significant source of nutrients.

Spatial distribution of soil P and stock camps

Soil P

Mean POlsen values (0-10 cm) in the high fertility treatments were significantly higher than the low fertility treatments in both hillslope and rainfall simulator plots indicating that there was a greater pool of labile P that could be desorbed from the soil by runoff. This was

161 consistent with the trend for higher TP concentrations in runoff from the high fertility treatments compared with the low. Established pastures are typically enriched with plant-available P in the top few centimetres of soil (Haynes and Williams 1993; Haygarth et al. 1998), reflecting an accumulation of P from plant residues, dung and fertiliser and a buildup in surface organic matter. This is also the layer with which surface runoff interacts so the source of P available for mobilisation is larger for surface runoff than subsurface flow (Ahuja et al. 1981). Routine soil samples taken from the top 7.5 - 15 cm to assess the P requirements of soils for plant production may therefore markedly underestimate the amount of P interacting with runoff (Sharpley et al. 1996; Haygarth et al. 1998). At Vasey, the top 2 and 5 cm of soil was enriched with bicarbonate extractable P by factors of 2.5 and 1.6 when compared to the standard sampling depth for Victoria of 10 cm. A slightly lower degree of enrichment (1.9) was found for the top 2 cm of pasture soil in Ireland (Culleton et al. 2000). There was little variation in the degree of enrichment with increasing soil available P at Vasey which suggested routine soil sample depths can provide adequate estimates of the P content at shallower depths. Routine agronomic soil sample depths are also favoured because there are extensive existing data sets and because at the standard depths there is less influence of surface organic matter causing variation in soil nutrient content and bulk density than at shallower depths.

Stock camps

At the hillslope or larger scale, the temporal and spatial deposition of animal manure relative to the occurrence of storm events and proximity to waterbodies determines much of the risk a grazed pasture presents to water quality (Tate et al. 2000). The disproportionate deposition of manure in stock camps relative to the rest of the pasture (Gillingham 1987) increases the risk of P losses from manure in camp areas (McColl and Gibson 1979; Williams and Haynes 1992). At Vasey, sheep tended to preferentially graze in upslope regions in the set- stocked plots 2 and 3 leading to an increase in dung deposition and significantly higher POlsen levels in these camp areas compared to the remainder of the plot. Similarly, Gillingham et al.

(1987) found that the POlsen status in soil under sheep stock camps was at least twice as high as that of the general paddock soil. In contrast, stock camps in the rotationally grazed plots (1 and

4) were smaller and more discrete than in the set stocked plots and changes in POlsen due to stock camping were not detected. This may have been because the sampling density and time for nutrient accumulation were insufficient during the 3-year study. None of the stock camps in the hillslope plots coincided spatially with the seasonally waterlogged runoff source area in the lower quarter of plot 1 (see Chapter 4), which meant that accumulation of P in camp areas did not increase the risk of P loss in runoff.

162 Sheep often choose to camp in drier, upslope regions, which naturally lessens the risk of increased nutrient losses from manure sources in runoff (McColl and Gibson 1979). Grazing management may however be a useful tool for reducing the risk of P losses from manure in runoff through discouraging the accumulation of dung in areas prone to surface water movement. Animals also tend to gather near water sources (West et al. 1989), so preventing stock access to natural watercourses by reticulating water is also important for reducing the contribution of P in livestock manure to eutrophication of surface waters.

5.4.5 Experimental approaches: strengths and limitations

A variety of experimental scales have been used to characterise nutrient losses from land to water, including laboratory studies (Romkens and Nelson 1974; McDowell et al. 2001), small plot rainfall simulations (Costin 1980; Sharpley and Moyer 2000), hillslope (Heathwaite and Dils 2000; Nash et al. 2000) and catchment studies (Sharpley and Syers 1979; Pionke et al. 1996; Stevens et al. 1999). Each approach has advantages and limitations relating to repeatability, practicality, the degree to which landscape processes are represented and the ability to test for treatment effects. Two experimental scales were used in the current study to measure the effects of increasing soil P fertility and sheep stocking rates on the P status of runoff water at Vasey. Hillslope plots (0.5 ha) that were used to characterise nutrient losses under natural rainfall conditions, allowed realistic hydrological processes and the effects of grazing behaviour to be investigated. Runoff measured at the hillslope scale also allowed the cumulative influence of small scale spatial (<1 m2) variability in pasture characteristics, grazing, topography and P mobilisation processes to be integrated by each runoff event. However, the limitations of hillslope and catchment scale studies must be acknowledged. Apart from being cost and labour intensive to set-up, the main limitation of this scale of study was that treatments could not be replicated, which prevented a rigorous statistical analysis of treatment effects. It was also not possible to control discrete components of the pasture system in order to identify, for example, the relative importance of soil, herbage and dung P sources. It was possible, however, to draw sensible conclusions about the processes that occurred based on quantifiable and qualitative differences between treatments and plots. The potentially inconclusive nature of catchment scale studies was demonstrated by McColl et al. (1977), who found high nitrate concentrations in streamflow from both pastoral and native forest catchments when compared to other native and exotic forest catchments. This demonstrated that factors other than vegetation type influenced the chemical nature of runoff and that drawing conclusions based on landuse alone can be misleading. A second major limitation of the hillslope experiment was the reliance on natural runoff for data, particularly as the cumulative rainfall during the period October 1996 – July 2001 in

163 the south-west of Victoria was the lowest 5 yr total on record for this region (Wright and Jones 2003). The small number of runoff events that occurred over the study period, particularly in plots 3 and 4, made testing differences in runoff nutrient concentrations between plots difficult. The bootstrap technique enabled paired t-tests to be performed between FWM concentrations derived from at least 6 runoff events but there was little confidence in this technique for plots 3 and 4, where runoff was less frequent. The small number of events in plots 3 and 4 meant time- series analysis was not appropriate either. In contrast, small plot studies using artificial rainfall offer the advantages of independence from climatic constraints, replication, fewer infrastructure requirements, and standardisation of key factors such as rainfall, pasture composition, soil characteristics, slope gradient and the area of interaction between runoff and the pasture surface (Costin 1980). The simulated runoff experiment was used to test for treatment effects on runoff, erosion and nutrient mobilisation. A single rainfall intensity and duration was applied to plots of uniform size (0.64 m2) in order to reduce variations in nutrient concentrations caused by hydrological variation. However, even with standard conditions and treatment replication, there were some difficulties in detecting treatment effects. Hudson (1993) suggested that small plot studies are powerful methods for identifying treatment effects where the treatments are markedly different from each other. Treatments in the current study were chosen to reflect the range of current farmer practices within the sheep industry, rather than extremes of soil fertility so this may have limited the potential for the influence of soil P fertility on runoff water quality to be quantitatively tested. Certainly relationships between runoff P concentrations and soil P fertility or fertiliser applications are most distinct when the range of soil P fertility is large, such as when heavy manure applications are made or runoff is tested shortly after fertilisers are applied (Romkens and Nelson 1974; Sharpley 1995; Pote et al. 1996). A limitation of the small plot approach is that runoff and nutrient movement generated using simulated rainfall rarely reflect actual processes occurring under natural rainfall at larger scales. In the current study, there were no significant differences in the runoff volumes between treatments under simulated rainfall whereas at the hillslope scale, more than 66% of runoff was produced from less than 25% of the 2 ha hillslope area. This demonstrated that nutrient and sediment loads measured at small plot scale did not reflect losses at the hillslope scale due to the spatial heterogeneity in runoff generation processes. Greenhill et al. (1983a) acknowledged runoff volumes measured in 40 m2 plots did not account for subsurface return flow, which is often a feature at the catchment scale (Pionke et al. 1996; Stevens et al. 1999; Dahlhaus et al. 2000). The influence of experimental scale on P mobilisation processes is discussed in Chapter 6. Another limitation of any plot-scale runoff study is that the fate of nutrients beyond plot boundaries is not known (Johnes and Hodgkinson 1998) and consequently, nutrient loads

164 entering receiving waters can be markedly different from those generated at the hillslope (Lennox et al. 1997).

5.5 Conclusions

Runoff measured from pasture hillslope plots at Vasey demonstrated that the volume of runoff, rather than stocking rate or level of soil P fertility was the primary factor that determined the amount of P loss in runoff. Two-thirds of the runoff from the 2 ha hillslope site was produced from a seasonally waterlogged runoff source area which occupied <25 % of the total site (see Chapter 4). As a result, 70% of the TP loss from the field site occurred from this same area. However, for a given volume of runoff, higher fertility pasture systems led to increased P loss due to elevated average P concentrations in the runoff. This finding was supported by simulated runoff from small plots suggesting that fertiliser application rates and stocking rate play an important role in determining P concentrations in runoff. Average annual concentrations of P in runoff measured over two winter/spring seasons were at least five times higher than desirable concentrations in freshwater streams (0.04 mg P/L). These concentrations could cause water quality problems where runoff dominates dam and stream water supply. The elevated concentrations in runoff from the high P fertility pastures were likely to have been caused by interaction of the runoff with the higher concentrations of potentially labile P in soil, plant and dung material in these pastures. Dung and herbage are potentially large sources of P in comparison to soil because the high groundcover across all the pasture treatments led to minimal erosion, and because P fixing components in the soil can buffer labile P concentrations. The similarity in runoff SS concentrations between plots suggests that dissolved, rather than particulate, forms of P are largely responsible for the higher concentrations of TP in runoff from the high soil P fertility pastures. The influence of erosion and labile soil P pools on runoff P concentrations is investigated in the following chapter. TP concentrations in soil water within and below the rootzone (representing the subsurface flow pathways) were generally below 0.05 mg P/L and total P losses in deep drainage, estimated using an average P concentration and estimated annual drainage from a soil water balance model, were less than 30 g/ha. This was less than P loads measured in runoff from the two hillslope plots that produced the majority of runoff, but comparable to losses in runoff from the two well-drained hillslope plots. The risk of P loss was therefore greater for surface runoff than subsurface flow pathways in this environment due to a combination of higher concentrations of P and the potential for large volumes of runoff to be shed during infrequent storms. Pasture conditions such as herbage mass, groundcover in spring and sward height differed little between pasture treatments and therefore were unlikely to have an impact on runoff volume or water quality during the winter spring period. Phosphorus accumulated in

165 sheep camps in the set-stocked pastures, however, because the location of these camps did not coincide with runoff source areas, the grazing behaviour of sheep did not increase the risk of P losses from pastures. However, overgrazing on and adjacent to sheep camps can increase the erodibility of soil through summer months. Both rotational and set-stocked grazing methods may be useful tools for controlling the spatial distribution of groundcover and stock camps in such a way that nutrient hotspots and bare ground are eliminated near runoff source areas. This would reduce the risk of erosion and nutrient losses in runoff. The forms of P in runoff are investigated in Chapter 6 in order to identify the main processes of P mobilisation via surface flow pathways that need to be targeted to further minimise P movement.

166 CHAPTER SIX

6 Forms of P and processes of P mobilisation in runoff

6.1 Introduction

P is mobilised from pastures in surface runoff by a combination of chemical, physical and biological processes (Haygarth and Jarvis 1999). These processes include detachment and entrainment of soil, organic matter, faecal or fertiliser particles that carry P, desorption or dissolution of soil P into solution, leaching of P from organic material and mixing of runoff water with existing soil-water to which P has been released through physico-chemical or biological mechanisms. The sources of P and the processes of mobilisation will influence the forms of P that are in runoff from pastures and will also influence the potential bioavailability of P in receiving waterways (Haygarth and Jarvis 1999). The main operationally defined fractions of P in runoff are particulate and dissolved, and reactive and unreactive forms, and these were described in detail in Chapter 2. It is important to understand the mechanisms of P mobilisation in order to define appropriate management strategies for minimising the loss of P from pastures. For example, in New Zealand, 85% of P in runoff from a pasture catchment was in the particulate form, which suggested erosion control was a priority (Cooke 1988). In Chapter 5 it was postulated that the dissolved forms of P are likely to have dominated the P loads in runoff from the fertilised pastures, and that particulate P losses are likely to be low due to low levels of erosion. It is also expected that unreactive P would be an important fraction of the total P because of the typically high organic matter levels in temperate pasture soils and because the P in this fraction has a low propensity to adsorb onto soil particles. This chapter sets about exploring these propositions and attempts to confirm the processes of P mobilisation in runoff. A range of empirical and physically-based models have been used to predict the concentrations and loads of the main forms of P in runoff from agricultural land (Knisel 1980; Sharpley et al. 1982; Williams et al. 1983; Cooper et al. 1992; Fleming and Cox 1998; Nash et al. 2000; Daly et al. 2002). However, the applicability of existing models to runoff from sheep pastures in southern Australia has not been tested. Sound modelling is an increasingly important tool for land mangers to predict the impact of farming practices and land use on water quality without having to invest in costly and time-consuming measurements. A selection of existing models that predict particulate and dissolved concentrations of P in runoff are described below.

167 6.1.1 Mobilisation of particulate and dissolved P fractions

Fleming et al. (2001) and others (Sharpley and Halvorson 1994; Haygarth and Jarvis 1999; Nash et al. 2000) suggest that the processes controlling DP and PP mobilisation in runoff are distinct from each other and for this reason the two fractions are discussed separately below.

Mobilisation of particulate P in runoff

During water erosion, soil particles undergo detachment, entrainment, transport and redeposition. Clay-sized particles (<2µm) with low settling velocities are preferentially entrained in runoff, particularly as the flow rate decreases. These particles include clay and sesquioxide minerals, which have a high chemical association with nutrients, and offer a higher specific surface area for adsorption and transport of P in runoff compared to coarser particles in the bulk soil (Rhoton 1979; Sharpley 1985b). As a result, sediment in surface runoff is usually enriched with C, N, K and P compared with the soil from which it is derived (Sharpley 1980). The ratio of the nutrient concentration in sediment compared with the source soil is known as the enrichment ratio (ER) (Sharpley 1980). As the kinetic energy of the surface flow increases, less suspended sediment settles out and the size distribution of the sediment more closely resembles that of the source soil. Increased slope can cause an increase in the runoff energy and lead to a decrease in the ER. Similarly, the increased kinetic energy of raindrop impact on uncovered, compared with covered, soil or under higher rainfall intensity is expected to reduce the ER of runoff sediment (Sharpley 1980). Cooke (1988) also found the sediment ER for P (PER) was higher for summer runoff compared with winter and attributed this to more coarse material being eroded in high winter flows. As well as rainfall/runoff factors, the PER is increased by fertiliser amendments and increased soil P status (Sharpley 1980), again due to the strong chemical association between the clay fraction and added P. Sharpley (1980) also suggested that soil P, as opposed to freshly added fertiliser P, may be strongly associated with clay particles that are aggregated by organic matter into larger (>2µm), less erodible particles and therefore have less influence on the PER. The proportion of TP in runoff that is attached to particles may increase due to adsorption of orthophosphate to entrained soil particles (Sharpley et al. 1981c; Tham 1983). Entrained sediment is therefore either a potential sink or reservoir of P, depending on the equilibrium between P in the water column and P in the sediments (Holt et al. 1970). Given the above physical processes and relationships, Sharpley et al. (1980) found that the concentration of PP (mg/L) in runoff can be estimated from a combination of the TP in the source soil (Psoil in mg/kg), the PER of the suspended sediment and the sediment concentration (SS in kg/L) in the runoff, according to equation 6-1.

PP = Psoil × PER × SS (6-1)

168 Sharpley et al. (1982) later found good agreement between the concentrations of PP predicted using equation 6-1 when compared with mean annual PP concentrations in runoff from cropped and grassed catchments of 1.1 to 52.8 ha in size.

Mobilisation of dissolved P in runoff

P is released from soil and pasture constituents into solution via a range of mechanisms including desorption and slow diffusion from soil particles, dissolution of inorganic fertilisers, leaching from plant and organic matter, microbial mineralisation, enzymatic hydrolysis of humus, manure and plant residues, and lysis of microbial biomass through drying and re-wetting (White 1980; Yli-Halla et al. 1995; Haygarth 1999). Desorption and dissolution reactions have been studied extensively with regards to their ability to supply phosphate to plant roots (Beckett and White 1964; Cooke 1966; Barrow 1979a; White 1980; Holford 1989). Principles developed through these studies were then extended to explain the release of P from soil to runoff and drainage water (Sharpley et al. 1981a; Sharpley et al. 1985; Brookes et al. 1996; Chapman et al. 1997). Sharpley et al. (1981a) described desorption of P from soil by runoff as a combination of release to both an infinite and a finite sink. The infinite sink is represented by the very large water to soil ratios and by continuous removal of the desorbed P during runoff. Under these conditions, the amount of P released does not influence further release whereas under conditions of pooled water (a finite sink), desorption of P abates the rate of further release. The large water to soil ratio during runoff, even without continuous removal, would be expected to enhance the amount of P desorption per gram of soil but lower the soil-water P concentration due to dilution (White 1966; Barrow and Shaw 1979a). There is also a shorter period of soil-water contact during runoff events compared with the release of P to the soil solution (Sharpley et al. 1981b). The effect of time on P sorption has been explored using quantity-intensity curves. When the soil solution P concentration is less than the equilibrium P concentration (EPC), the amount of P sorbed decreases with time. Conversely, when the solution P concentration is greater than the EPC, the amount of P adsorbed increases with time (Beckett and White 1964; Barrow and Shaw 1975). Therefore more P would be desorbed in pooled water than in runoff, albeit at a diminished rate. The initial rate of P desorption from soil was also found to increase with the quantity of labile P in the soil (Cooke 1966). Other conditions which can enhance desorption of P from soil are higher temperatures (Fordham 1963; Barrow 1979b), mechanical breakdown of soil aggregates (eg. through increasing the effective soil surface area to unit volume of water by increasing the water to soil ratio (Barrow and Shaw 1979a)), variation in soil pH (White 1980; Barrow 1984), and decreased ionic strength of weakly acid or alkaline solutions (Barrow and Ellis 1986). The rate of P desorption from soils is also reduced by high concentrations of salts in solution, and in

169 solutions dominated by Na+ or Mg2+ rather than Ca2+ (Ryden et al. 1977; Barrow and Shaw 1979b). Anoxic conditions, which may develop in waterlogged soils, can lead to the reduction of ferric minerals to ferrous ions, with a co-release of P associated with the iron minerals. However an increase in surface area caused by the reduction of oxyferric hydroxides may lead to a subsequent increase in P sorption under reduced conditions (Khalid et al. 1977), so the net release of phosphate depends on the distribution and availability of sorption sites. Sharpley et al. (1981b) found that an empirical power relationship (equation 6-2) described the influence of some of these mechanisms on the mass of P desorbed (Pd, mg) by runoff. Equation 6-2 combines the effects of the initial amount of desorbable P in the soil (Po, mg/kg), the water to sediment ratio (W, cm3/g) and the duration of contact (t, min) between soil and solution for a given soil (with equation constants K, α and β)(Sharpley et al. 1981b).

α β Pd = KPot W (6-2) The constants in equation 6-2, for 60 major soil types in USA, were well correlated with the ratio of clay to organic carbon contents, where P sorption decreased with higher organic carbon (Sharpley 1983). This was attributed to competition between organic acids and phosphate for sorption sites. In these experiments the release of P from soil was simulated by shaking homogenised (<2 mm), air-dried soil samples with distilled water. The model was also tested by simulating runoff on repacked soil in boxes of no more than 1 m in length (Sharpley et al. 1981a). P was found to desorb throughout the period simulated runoff was applied, however the rate of P release, and hence the runoff P concentration, decreased over time. Equation 6-2 was modified to describe the average concentration of DP ( c ) in a runoff event (equation 6-3, (Sharpley et al. 1981a).

α β c = [KPoSt W ]/ R (6-3) In this equation, R is the total rainfall during the event (mm) and S, which was positively correlated with Pd, represents the mass of soil interacting with rainfall and runoff (Sharpley et al. 1981a). The concentration of DP decreased as the total rainfall increased due to dilution, despite the increased amount of P desorbed per gram of soil by the larger water to soil ratio (Yli-Halla et al. 1995) and kinetic energy of rainfall (Sharpley et al. 1981a; Ahuja et al. 1982). The effective depth of interaction (EDI, mm) between the soil and runoff was calculated from S in equation 6-3 using the soil bulk density, and from 32P studies for five soils, and ranged from 1.3 to 3.2 mm for 60 mm/h rainfall and 4% slope angle and increased up to 37 mm under 160 mm/h rainfall at 20% slope (Ahuja et al. 1981; Sharpley et al. 1981a; Sharpley 1985a). Higher values of S and EDI for high soil slope angle, rainfall intensity and soil loss, was attributed to more turbulent mixing as a result of increased runoff and rainfall energy (Sharpley et al. 1981a; Sharpley 1985a).

170 The effects of slope length, slope angle, and soil cover on c in runoff were also investigated by Ahuja et al. (1982) using artificial rainfall on repacked soil in boxes of 1 m length. Under these conditions, it was found that an increase in the degree of slope and decrease in soil cover increased c by increasing S and that greater slope length increased c by increasing the depth and velocity of flow. Importantly, however it was noted that on a long slope, desorbed P may become adsorbed further down the slope (Sharpley et al. 1981a). Daniel et al. (1993) found a slightly modified version of equation 6-3 (i.e substituting mm runoff volume for R) provided a reasonable prediction (R2 = 0.43) of DP concentrations in simulated runoff events from 9 m2 pasture plots. Wallach and Shabtai (1993) later developed a physically based model to describe the distribution of P within the soil profile and runoff concentration hydrographs. However, for predicting annual flow-weighted mean DP concentrations in runoff,

Po alone gave good estimates (Sharpley et al. 1982). Another model, used to predict DP yields in surface runoff, is that used in the Erosion Productivity Impact Calculator (EPIC)(Williams et al. 1983). As well as a model proposed by Storm et al. (1988), EPIC accounts for P desorption and sorption processes. In the EPIC model, it is assumed that P in runoff is mostly associated with the sediment phase, and the DP concentration is described using a partitioning coefficient (Kd) for P between the sediment and solution phases (equation 6-4).

YSP = (0.01× CLP × Q) / Kd (6-4)

where YSP is the yield of DP in kg/ha lost in runoff volume Q in mm, CLP is the concentration of available P in the topsoil in g/t and Kd is the P concentration of the sediment (i.e the sediment P concentration in g/t) in the runoff divided by the P concentration in the water 3 3 (i.e. the DP concentration in g/m ) in m /t. The value of Kd used in EPIC is 175. CLP was assumed to be equal to POlsen, (0-10 cm depth) because the critical CLP for plant growth used in the EPIC plant growth sub-model was 20 g/t, which is close to the critical POlsen level for Australian temperate sheep pastures of 12-14 g/t (Cayley et al. 2002). The DP concentration

(accounting for units) can therefore be derived either from Kd (equation 6-5),

P DP = sed (6-5) Kd

where Psed is the P concentration in the sediment, or from CLP (equation 6-6);

C DP = LP (6-6) Kd

171 DP is also released into surface runoff by dissolution of fertiliser granules, leaching from vegetation and faecal material and through biological processes. Dissolved fertiliser compounds can contribute large amounts of the DP in runoff, particularly in environments where the time between fertiliser application and runoff occurring is short (Sharpley and Syers 1976; Nash et al. 2000). Nash et al. (2000) derived an empirical model for runoff event TP concentrations in Gippsland dairy pastures that incorporated the effects of fertiliser dissolution (equation 6-7). TP in this model closely represents DP concentrations because >90% of the TP in dairy pasture runoff was found to be in the dissolved form (Nash and Murdoch 1997). The back-transformed model (equation 6-7, k, ai, b and c are constants) shows inverse power relationships between TP and days since fertiliser was applied (DF), days since last grazed (DG) and total event flow (TF).

−a −b −c TP = ki × DF × DG ×TF (6-7)

The dilution of TP concentrations by larger runoff event volumes was similar to the effect described by the model of Sharpley et al. (1981a)(equation 6-3), however, TF was only significant (P<0.05) if fitted before year (ki)(Nash et al. 2000). P that is leached directly from vegetation and faecal material at the soil surface can also make an important contribution to runoff DP levels (Holt et al. 1970; Timmons et al. 1970; White 1973; Gburek and Heald 1974; Burwell et al. 1975; Sharpley 1981; Preedy et al. 2001; Withers et al. 2001), however these factors have not been explicitly included in these models of DP concentration in runoff. Biological mechanisms may play a significant role in determining the size of the DP pool through mineralisation and hydrolysis of soil organic P (Haygarth 1999). Biological mechanisms were not assessed directly in this study, but it was assumed that the net effect of the biological transformations which involve P, and the ensuing physico-chemical modification of the biological products would be reflected in the final fractionation of P in runoff waters. Regardless of the mechanism of P release, DP is subject to adsorption and precipitation processes commensurate with the soil and organic matter with which it interacts. Considering these processes of PP and DP mobilisation in runoff, PP losses were expected to occur from sheep pastures in temperate Victoria, largely in association with erosion during large runoff events. Concentrations of PP were likely to be lower than from cultivated soils or degraded pastures provided >75% groundcover was maintained during the winter-spring runoff season. Some DP losses were also likely to occur and both PP and DP concentrations in runoff were expected to increase with greater fertiliser applications and stocking rates due to greater enrichment of the surface soil with nutrients.

172 6.1.2 Aims

The primary aim of the work reported in this chapter was to identify the main processes of P mobilisation from sheep pastures in temperate Victoria, in order to identify management practices that should be targeted in order to minimise P losses in runoff. This was achieved by measuring the main fractions of P in runoff from low and high fertility pastures under natural and simulated rainfall conditions. The ability of existing empirical models to predict PP and DP concentrations in runoff was also tested using data from two pasture sites in Victoria (Vasey and Maindample) to determine whether models could be used instead of expensive field studies to estimate the effects of pasture management on runoff water quality.

6.2 Methods

In order to identify the main fractions of P in runoff from sheep pastures, operationally defined P fractions were measured in natural rainfall runoff, and in simulated runoff from small plots at two pasture sites in Victoria.

6.2.1 Sample collection

Vasey

Runoff generated through natural rainfall from four 0.5 ha hillslope plots on a commercial sheep property at Vasey in South-West Victoria was measured and analysed for P. Details of the site, and materials and methods were described in Chapters 3, 4 and 5. Runoff from small plots (0.64 m2) was generated using simulated rainfall at an intensity of 48 mm/h for 60 minutes on four pasture treatments at the same location in 1999. Details of this experiment were described in Chapter 5. An extra 250 mL sample was collected after 25 minutes of runoff for analysing the effects of sample storage time on RP concentrations (Appendix C).

Maindample

Simulated rainfall was also generated from pastures at Maindample in north-east Victoria in July 1999. The materials and methods were the same as those used at Vasey. Details of soil properties and treatments at the Maindample site are described in Chapter 7. In brief, the experimental site was part of the Sustainable Grazing Systems (SGS) national experiment and consisted of high, medium and low P input pasture catchments, which were 4.9, 13.7 and 11.5 ha in size respectively. The soil is a Brown Sodosol, consisting of a sandy loam topsoil overlying a sodic clay B horizon at approximately 40 cm depth.

173 Natural runoff was measured at the Maindample site from each of the three catchments from 1998 to 2001 as part of the SGS national experiment. Details of these measures can be found in Ridley et al. (2003). The TRP, DRP and DP fractions in runoff were measured for 3 runoff events from the low fertility, and 6 events from the medium and high fertility catchments in 2000. Simulated runoff was generated from three plots in five locations across each of the three catchments (45 plots in total) using rainwater supplied for 1h at an intensity of 48 mm/h. A 70 mL subsample was collected every 5 minutes and stored in an ice box. In the laboratory, composite samples (approximately 120 mL) representing the first, second and third 20 minute periods of runoff were prepared by mixing flow-weighted proportions of the original samples. Chemical analyses were performed on composite water samples. Runoff volume and total soluble salt content, pasture dry matter, plot slope, soil antecedent and saturated water content (0-5 cm), and the depth to the wetting front were also measured (C. Schefe pers. comm.).

6.2.2 Fractionation of runoff water

The nomenclature and analytical scheme used to identify the forms of P in runoff in the current study are based on the analytically defined physical and chemical fractions described in Chapter 2 (Table 2-1). P in runoff samples was therefore physically and chemically separated into fractions that represented P dissolved in solution (DP), P attached to particles (PP), inorganic P (reactive P, RP) and organically bound P (unreactive, UP) (Figure 6-1). In the laboratory, a known volume of runoff sample (approximately 250 mL or half the collected sample) was passed through a 47 mm diameter glass fibre prefilter and a 0.45 µm cellulose acetate membrane filter using a syringe and filter holder. Filtered and unfiltered samples were analysed within 24 h or stored at –15oC prior to analysis of P. TP and DP were measured according to methods described in Chapter 5, Section 5.2.2 ‘Chemical analyses’. For hillslope and simulated runoff samples collected at Vasey in 1999 and 2000, DRP and TRP were analysed manually within 24h using method H2a of Rayment and Higginson (1992). Samples from Vasey in 1998, and simulated runoff samples from Maindample in 1999, were stored at-15oC for up to 5 months prior to analysis of DRP and TRP using a Lachat flow injection system (Huberty and Diamond 1996). TRP concentrations in samples that had been stored frozen for 3-4 months prior to analysis were approximately 15% lower than concentrations measured within 24 hrs of collection, whereas there was no significant (P>0.05) difference in DRP concentration (see Appendix C). Rather than making any adjustments to the TRP concentrations measured after frozen storage, the reported concentrations are considered conservative estimates.

174 Runoff sample

Filtration through 0.45µm membrane filter

Alkaline K persulphate autoclave digest

Mo-blue colorimetric analysis at 882 nm (after reduction with ascorbic acid)

DRP DP TP TRP

DUP = DP-DRP PP = TP-DP PRP = TRP-DRP PUP = PP-PRP

Figure 6-1: Analytical fractionation of P in runoff samples

175 6.2.3 Soil Olsen and total P

Surface (0-10 cm) soil samples were collected from the 0.5 ha hillslope plots at Vasey as described in Chapter 5. At Maindample, fifteen soil cores (2.5 cm diameter) of 0-5 and 5-10 cm depths were collected and bulked across three different landscape strata in each catchment in 2000. At both Vasey and Maindample, 10 soil cores (0-5 cm depth, 2.5 cm diameter) were collected from just outside the perimeter of each 0.64 m2 plot used for rainfall and runoff simulations in 1999. The cores were bulked and mixed. All soil was air dried (40oC) and sieved to 2 mm prior to chemical analysis. Soil Olsen P (POlsen, mg/kg) and total P in soil (Psoil, mg/kg) were measured as described in Chapter 5, Section 5.2.2 ‘Chemical analyses’.

6.2.4 Statistical analysis

Standard errors and t-tests were calculated for the annual flow-weighted mean P concentrations of runoff from the hillslope plots using a bootstrap method described in Chapter 5. Differences between treatment means for the variables measured during the rainfall simulator experiments at Maindample and Vasey were analysed using a multiple t-test, using runoff plots as units and the three levels of fertiliser and stocking rate as treatments. Without replication of the treatments at Maindample, treatments effects could not be tested by statistical analysis, but the large size of the catchments meant that a great deal of natural variation was accounted for within each treatment, which increased the likelihood that any observed differences were due to the treatments. The natural logarithms of all plot mean data except flow volume were used to compare treatment means because this normalised the distribution and standardised the variance of the residuals of the original data. Standard errors of the differences between means were predicted using a restricted maximum likelihood method (REML) using Genstat5® statistical software (Lawes Agricultural Trust 1997). The influence of the spatial and temporal experimental scale on P and SS concentrations was tested by comparing the annual mean concentrations of P and SS from the hillslope plots to treatment mean concentrations in runoff from the small simulator plots using linear regression and ANOVA. Data from one hillslope plot-year combination was excluded from the analyses because the volume of runoff (<0.1 mm) was negligible.

176 6.3 Results

6.3.1 Treatment effects on forms of P in surface runoff

Vasey

In runoff from the 0.5 ha hillslope runoff plots at Vasey, between 46 and 99% of the TP was in the dissolved form, with flow-weighted mean annual DP concentrations ranging from 0.09 – 0.82 mg/L (Table 6-1). DRP comprised 6 and 31% of TP in runoff from the low fertility pasture in 1998 and 2000 respectively, and comprised 31-89% of TP from the higher fertility pastures (Table 6-1). DRP concentrations were at least 2.8 times, and up to 73 times, lower in runoff from the low fertility plot 2 compared to the high fertility plots 1, 3 and 4 (Table 6-1). DP concentrations followed a similar trend. Except for plot 4 in 2000, where negligible flow occurred, PP concentrations were less than 0.2 mg/L. The DRP fraction was responsive to an increase in POlsen whereas concentrations of DUP, PRP and PUP generally remained below 0.20 mg P/L with increasing levels of POlsen (Figure 6-2).

Table 6-1: Pasture treatments, POlsen (0-10 cm), annual flow-weighted mean P concentrations and percentages of DP and DRP in hillslope runoff at Vasey

A Plot Trt POlsen TP DP DRP PP DP/TP DRP/TP mg/kg ------mg/L ------% % 1998 1 C 10 0.34 (0.06)B 0.23 0.11 0.11 67 32 2 A 7 0.19 (0.02) 0.09 0.01 0.12 46 6 3 B 11 0.41 0.36 0.30 0.05 88 72 4 C 16 0.83 0.82 0.73 0.01 99 89 2000 1 C 13 0.37 (0.07) 0.28 (0.07) 0.19 (0.07) 0.10 (0.02) 74 52 2 A 6 0.22 (0.03) 0.14 (0.02) 0.07 (0.02) 0.09 (0.02) 61 31 3 B 13 0.81 0.64 0.58 0.17 79 71 4 C 20 0.72 0.34 0.28 0.38 47 39 A Treatments: A; Low P, set stocked, B; High P, set stocked, C; High P, rotationally grazed, B bootstrap standard error, calculated only for means with ≥ 6 runoff event samples

177 1.00 DRP PRP 0.80 DUP PUP 0.60

0.40 P form in runoff (mgP/L) 0.20

0.00 0 5 10 15 20 Soil Olsen P (mg/kg)

Figure 6-2: Flow weighted annual mean concentrations of P fractions in runoff from hillslope plots at Vasey of contrasting mean POlsen (0-10cm). Solid line shows the fitted regression curve for DRP, 2 where DRP = 0.0147exp(0.208*POlsen) R = 0.47 P<0.05

In the simulated runoff, DRP concentrations were again responsive to increases in

POlsen, but in contrast to the hillslope runoff, so too were the PRP and PUP fractions (Figure

6-3). There was little increase in DUP concentration as POlsen increased (Figure 6-3), a result also observed in the hillslope runoff plots. The concentrations of all measured P fractions, except DUP, increased in the treatment order unimproved (D), low P fertility set stocked (A), high P fertility rotationally grazed (C) then high P fertility set stocked (B), which corresponded to increasing levels of the treatment mean POlsen (0-5 cm depth), however, treatment differences were not always significant at the 95% confidence level (Table 6-2).

178 Table 6-2: Mean P concentrations and percentages of DP and DRP in simulated runoff, and POlsen (0-5 cm) for four pasture management treatments at Vasey

Treatment POlsen DP TRP DRP PP PRP DUP DP/TP DRP/TP mg/kg ------mg/L ------% % A A14a 0.24 0.31ab 0.20 0.25ac 0.12ac 0.04 47 36 B B34b 0.42 0.59a 0.35 0.51b 0.23b 0.06 45 38

C31b 0.34 0.56a 0.30 0.44bc 0.25bc 0.05 46 38

D8a 0.16 0.15b 0.10 0.12a 0.05a 0.06 53 32 A Subscript notation denotes differences in treatment means at the 95% confidence level B Significant differences between treatments (P<0.05) were calculated using the log transformed mean, which in contrast to the original mean, was greater for treatment B than C.

a) DRP b) DUP 1.2 1.2 P <0.001, R2=0.49 P <0.05, R2=0.07 1 1 0.8 0.8 0.6 0.6 0.4 0.4 DUP (mg/L) DUP

DRP (mg/L) 0.2 0.2 0 0 -0.2 020406080 0 20406080

P Olsen (0-5cm)(mg/kg) P Olsen (0-5cm)(mg/kg)

c) PRP 1.2 d) PUP 2 1.2 1 P <0.001, R = 0.23 2 1 P <0.01 R = 0.15 0.8 0.8 0.6 0.6 0.4 0.4 PRP (mg/L) PUP (mg/L) 0.2 0.2 0 0 0 20406080 020406080

P Olsen (0-5cm)(mg/kg) P Olsen (0-5cm)(mg/kg)

Figure 6-3: Relationships between runoff concentrations of DRP, DUP, PRP and PUP fractions (a- d) and plot POlsen for individual simulator runoff plots at Vasey with regression lines and statistics

Effect of experimental scale on P concentrations and forms in runoff

A comparison of the effect of experimental scale on concentrations of TP, DP, PP and

SS concentrations shows that TP and DP increased with POlsen at both the hillslope and simulator plot scales, whereas PP only increased at the simulator plot scale (P<0.05 Figure 6-4 a,b&d). Differences between the two experimental scales were not significant (P>0.05, R2=0.75) for TP concentrations. In contrast, DP concentrations in the simulated runoff were lower at high POlsen levels than in hillslope runoff (P<0.05, R2=0.71)(Figure 6-4 b). Mean PP and SS concentrations were significantly higher (P<0.05) in the small-plot simulator runoff than in hillslope runoff (Figure 6-4 d&c).

179 a) TP b) DP

1.2 1.2 1.0 1.0 0.8 0.8 0.6 0.6 0.4 0.4 0.2 0.2 TP inTP (mg/L) runoff

0.0 DP in(mg/L) runoff 0.0 0 5 10 15 20 25 0 5 10 15 20 25 P (0-10 cm)(mg/kg) Olsen P Olsen (0-10 cm)(mg/kg)

c) SS d) PP

0.30 1.2 0.25 1.0 0.20 0.8 0.15 0.6 0.10 0.4 0.05 0.2 SS in runoff (g/L) runoff in SS 0.00 PP in(mg/L) runoff 0.0 0 5 10 15 20 25 0 5 10 15 20 25

P Olsen (0-10 cm)(mg/kg) P Olsen (0-10 cm)(mg/kg)

Figure 6-4: Treatment mean TP, DP, PP (mg/L) and SS (g/L) concentrations in hillslope (●) and

small-plot simulator (□) runoff at Vasey with increasing POlsen (0-10cm depth). Bars indicate standard errors. Means from a hillslope plot that generated <0.1mm runoff, as indicated by the symbol x, were excluded from the analyses.

Maindample

At Maindample, the mean POlsen (0-5 cm depth) of the low and medium fertility catchments was similar and low (7 and 11 mg/kg), compared to the high fertility catchment (27 mg/kg)(Figure 6-5). There were limited data available for the fractions of P in natural runoff. Data from 3 runoff events from the high and low fertility catchments and 6 runoff events from the medium fertility catchment in 2000 showed the mean DRP and DUP concentrations

increased at the highest POlsen level (B. Christy, pers. comm.) (Figure 6-5). The mean runoff TP concentrations for these events were less than 0.38 mg P/L, which is at the lower end of the range in annual FWM TP concentrations from hillslope plots at Vasey. The proportion of the TP in the dissolved fraction was at least 67%. Whereas 55% of the TP was in the DRP form in the high fertility catchment, only 13 and 17% of TP was DRP in runoff from the low and medium fertility catchments.

180 0.25

0.20 DRP

0.15 DUP

0.10 PP

0.05

P form in field (mg/L) runoff 0.00 0102030 Catchment Olsen P (0-5cm) (mg/kg)

Figure 6-5: Mean concentrations of P fractions for 3 runoff events from the low, and 6 events from the medium and high fertility catchments at Maindample in 2000

Data from the simulated runoff at Maindample showed many of the same features as at

Vasey. DRP concentrations in runoff were similar and increased with POlsen, as did the percentage DRP/TP whereas the DUP fraction did not increase (Figure 6-6, Table 6-3). In contrast to Vasey, however, the PUP fraction was larger than the DRP fraction, and the PRP fraction remained constant with increasing soil P fertility (Figure 6-6).

Table 6-3: POlsen (0-5 cm depth) and concentrations of P fractions in simulated runoff from high, medium and low P fertility pasture at MaindampleA

Treatment POlsen TP DP PP DRP DP/TP DRP/TP mg/kg ------mg/L ------% % Low 9a 0.72a 0.18a 0.54a 0.07a 25 10 (0.3)B (0.08) (0.02) (0.07) (0.01) Medium 14b 0.52a 0.23a 0.29b 0.13b 44 25 (0.9) (0.04) (0.03) (0.03) (0.03) High 35c 1.49b 0.61b 0.88c 0.48c 41 32 (2.6) (0.18) (0.07) (0.13) (0.06) A Subscript notation denotes differences in treatment means at the 95% confidence level B Standard errors of the means are in parentheses

181 Maindample 1.00 DRP PRP ) 0.80 PUP DUP

0.60

0.40

0.20 P in simulated runoff (mg/L 0.00 0 10203040

P Olsen (0-5cm) (mg/kg)

Figure 6-6: Mean concentrations of P fractions in simulated runoff from low, medium and high fertility treatments at Maindample in 1999. Line shows fitted regression curve for DRP from individual plot data, P<0.001, R2= 0.54

6.3.2 Other factors affecting PP concentrations in runoff

Effects of suspended sediment

Suspended sediment (SS) concentrations in both hillslope and simulated runoff from the Vasey site were described in detail in Chapter 5. In brief, concentrations of SS were similar and less than 0.3 g/L across all four 0.5 ha runoff plots. SS concentrations in runoff from simulated rainfall events on the small plots were approximately double those in the hillslope runoff and were not related to the pasture treatment or POlsen. In runoff from the 0.5 ha hillslope plots, there was a trend towards the annual FMW concentrations of PP increasing as SS concentrations increased (Figure 6-7) but the relationship did not hold for individual runoff events on each plot. A positive relationship between PP and SS concentrations was also found for runoff events under simulated rainfall at Vasey (Figure 6-8).

182 1.00

0.80 PP DP 0.60

0.40

0.20 P in field (mgP/L) runoff

0.00 0.00 0.10 0.20 0.30 Suspended sediment (g/L)

Figure 6-7: The relationship between annual FWM concentrations of PP or DP and SS concentrations in runoff from hillslope plots at Vasey

1.4

1.2

1

0.8

0.6

0.4

0.2 PP in simulated runoff (mg/L)PP in runoff simulated 0 0.00 0.10 0.20 0.30 0.40

Suspended sediment (g/L)

Figure 6-8: Relationship between PP and SS concentrations in simulated runoff across four pasture treatments at Vasey. Line shows regression curve where PP = 2.10*SS-0.025, P<0.001, R2=0.43

Concentrations of total solids (TS, where TS = SS + total dissolved salts (TDS)) in simulated runoff on high, medium and low fertility pastures at Maindample ranged from 0.14 to 1.03 g/L for individual plots (C.Schefe pers. comm.). In contrast to Vasey, where there were no significant treatment differences in SS concentrations, the mean TS for low (0.48 ± 0.07 g/L)

183 and high (0.35 ± 0.03 g/L) fertility catchments at Maindample were significantly (P<0.05) higher than that from the medium fertility catchment (0.25 ± 0.01 g/L) (Figure 6-9). Assuming the TDS in simulated rainfall runoff was minimal, TS concentrations at Maindample were generally higher than at Vasey (compare Figure 6-8 and Figure 6-9). PP concentrations in the simulated runoff at Maindample were strongly related to TS concentrations in the high and low fertility catchments (Figure 6-9) but not in the medium fertility catchment where the PP and TS concentration ranges were narrower (Figure 6-9).

High ) 2.50 Medium 2.00 Low

1.50

1.00

0.50 PP in simulated runoff (mg/L 0.00 0.00 0.25 0.50 0.75 1.00 1.25 Total Solids (g/L)

Figure 6-9: Relationship between PP and TS in simulated runoff from low, medium and high fertility pasture treatments at Maindample. Regression lines shown for High (dashed line, PP=3.992*TS-0.527) and Low (solid line, PP=0.899*TS+0.105) treatments, P<0.001, model R2 = 0.76

Enrichment of suspended sediment with P

Some of the noise in the relationships between SS and PP may be explained by the entrained soil particles being enriched with P to variable degrees. The PP fraction in runoff was assumed to consist of P attached to sediment (>0.45µm) suspended in the runoff and so the amount of PP per gram of SS (sediment P concentration, mg/g) was calculated as the ratio of PP to SS concentrations (i.e. PP/SS, the slopes of relationships in Figure 6-8 and Figure 6-9, Table 6-4). The mean annual sediment P concentration of natural runoff from 0.5 ha plots at Vasey ranged from 0.32 to 2.11 mg P/g sediment (Table 6-4). There was no systematic change in either PP or SS concentrations in runoff with increasing POlsen, which resulted in a mean sediment P concentration (PP/SS) of 1.2 g/kg in runoff from all hillslope plots at Vasey. The SS

184 was therefore enriched with P compared to the average total P content for the top 10 cm of hillslope soil (Psoil, equivalent range of 0.50 - 0.73 mg P/g, Table 6-4). The ratio of the sediment

P concentration to Psoil is referred to as the P enrichment ratio (PER). The PER calculated using this method reflects firstly the selective entrainment of clay-sized particles, which are enriched with adsorbed P compared with larger soil fractions. Secondly, this PER incorporates the tendency for Psoil to be greater in the upper layers of the soil. In hindsight, it would have been better to measure Psoil from 0-2 cm soil samples, because these would be more representative of the source material for SS. In contrast to the 0-2 cm layer, the average Psoil over 10 cm soil depth was not expected to change markedly over time, so values of Psoil measured in 1998 were used to calculate the PER for each plot in both 1998 and 2000. P enrichment ratios ranged from

0.4 to 3.0 across plots and years and also had no apparent relationship with POlsen (Table 6-4). In contrast, in simulated runoff at both Vasey and Maindample sediment P concentrations showed a strong relationship with POlsen (Figure 6-10 and Table 6-4). Treatment mean PERs at Vasey were similar where measured and were generally higher than in hillslope runoff. At Maindample, the Psoil contents were considerably lower than at Vasey and the PERs of 4.0, 4.9 and 8 were therefore much higher (Table 6-4).

Table 6-4: Sediment P concentration and PER for natural and simulated runoff at Vasey and Maindample

Hillslope runoff Simulated runoff Trt Psoil Plot Sediment P PER Trt Sediment P PER 0-10 cm concentration concentration mg/kg mg P/g SS mg P/g SSA Vasey 1998 2000 1998 2000 A 499 2 1.38 0.88 2.8 1.8 A 1.68 (0.26) 3.1 (1.8) B 694 3 0.83 2.11 1.2 3.0 B 2.45 (0.29) 3.7 (0.9) C 730 4 0.32 1.43 0.4 2.0 C 2.35 (0.36) 3.8 (1.1) C 521 1 1.56 0.99 3.0 1.9 - - - D ------D 0.94 (0.28) N/a Maindample L 235 1.15a (0.09) 4.9 M 295 1.18a (0.12) 4.0 H 310 2.40b (0.21) 8 A Subscript notation denotes differences in treatment means at the 95% confidence level

185 6

5

4

3

2

1 Sediment P concentration (mg P/g) 0 0 10203040506070

P Olsen (0-5cm)(mg/kg)

Figure 6-10: Sediment P concentration in simulated runoff from individual plots vs POlsen (0-5cm depth) at Vasey (solid circles) and Maindample (open circles). Line shows regression for combined 2 dataset where Sediment P concentration = 0.04* POlsen +0.91, P<0.001, R =0.35

186 6.3.3 Other factors affecting DP concentrations in runoff

Effects of suspended sediment

In hillslope runoff at Vasey there was a trend towards FMW concentrations of DP, and the proportion of TP that was dissolved, decreasing as annual FWM SS concentrations increased (Figure 6-7 and Figure 6-11). In simulated runoff at Vasey, however, the individual plot SS concentrations did not systematically affect the plot DP concentrations in runoff (P>0.05).

100 1998 2000 80

60

40

20 DP/TP in hillslope runoff (%)

0 0.00 0.05 0.10 0.15 0.20 0.25 0.30 Suspended Sediment (g/L)

Figure 6-11: The relationship between the proportion of total runoff P in dissolved form and the flow-weighted sediment concentrations for annual runoff from hillslope plots at Vasey

187 Influence of ionic strength of runoff on DP concentrations

The EC in runoff was used as a surrogate variable for ionic strength according to the relationship;

TDS = EC × 0.64 (6-8)

where TDS is total dissolved salts (g/L) and EC is electrical conductivity (dS/m)(White 1997). In the most hydrologically active hillslope plot (plot 1) at Vasey, the FWM EC of runoff water ranged from 0.11 to 0.78 dS/m in 1998 and from 0.03 to 0.66 dS/m in 2000. The large range of conductivity was attributed to variable contributions of saline groundwater discharge to surface flows (see Chapter 4) and a diluting effect of flow volume. Except for plot 2 in 1998, the range and maximum EC of runoff from plots 2, 3 and 4 was not as large as for plot 1 (Table 6-5). For plot 1 runoff, there was a strong inverse relationship between the EC and the DP concentrations in runoff in both years runoff was measured (Figure 6-12a and b).

Table 6-5: EC in runoff from 0.5 ha hillslope plots at Vasey

------EC (dS/m) ------Hillslope Plot 1234 1998 Range 0.11- 0.78 0.10 - 0.94 0.11 0.03 Mean 0.46 0.40 0.11 0.03 Standard Error 0.03 0.16 N/a N/a 2000 Range 0.03 - 0.66 0.18 - 0.22 0.12 - 0.17 0.04 Mean 0.41 0.20 0.15 0.04 Standard Error 0.03 0.01 0.01 N/a

188 a) Plot 1 1998 1.00

0.80

0.60

0.40

0.20 DP (mg/L) in runoff

0.00 0.00 0.25 0.50 0.75 1.00 EC of runoff (dS/m)

b) Plot 1 2000 1.00

0.80

0.60

0.40

DP in (mg/L) runoff 0.20

0.00 0.00 0.25 0.50 0.75 1.00 EC of runoff (dS/m)

Figure 6-12: The relationship between the DP concentrations and the EC of runoff events from plot 1 at Vasey in a) 1998 where DP= -3.86*EC+0.30, R2= 0.90, P<0.001 and b) 2000 where DP = exp(- 5.85*EC+0.41), R2=0.64, P<0.001

189 6.3.4 Effect of runoff volume on P concentrations in runoff

At the Vasey field site, there was a trend for the annual FWM concentration of TP in runoff to decrease as the total volume of runoff increased in plots 2, 3 and 4. However, plot 1 had a higher FWM TP concentration than would be expected based on the trend for the other 3 plots, as illustrated in Figure 6-13.

1.00

0.80

0.60

0.40

concentration (mg/L) (mg/L) concentration 0.20

Flow weighted annual mean TP 0.00 0 20406080 Annual runoff (mm)

Figure 6-13: FWM annual TP concentrations in hillslope runoff from plots at Vasey with increasing annual runoff volume. Open circles represent plot 1 runoff

There were sufficient runoff events in plot 1 to allow analysis of P forms for large events (≥ 3 mm) in comparison to small events (< 3 mm). Large events had over twice the concentration of TP, DP and DRP as small events (Table 6-6). As a percentage of TP, DRP was the dominant form in large events, whereas the percentage of PP did not vary greatly. It was not possible to conduct a similar analysis on data from plots 2-4 because there were insufficient runoff events.

Table 6-6: Arithmetic mean concentrations of P and the percentages of TP in PP, DP and DRP forms for small and large runoff events in plot 1 at Vasey

% total % sampled TP DP DRP PP DP DRP flow events ------mg/L ------% ---- A Runoff events ≥ 3 72 14 0.41a 0.31a 0.21a 28 76 62 mm (0.07)B (0.07) (0.07) Runoff events < 3 28 86 0.20b 0.14b 0.08b 38 70 28 mm (0.03) (0.02) (0.02) A Differences between the mean concentrations were tested using an unpaired t-test. Subscript notation denotes differences in flow volume category means at the 95% confidence level B Standard error of the mean

190 6.4 Discussion

6.4.1 Mobilisation of particulate P

An important process of P mobilisation in runoff is through the erosion of P-rich particles. There was evidence from pastures at both Vasey and Maindample that concentrations of PP in runoff increased with increasing SS concentrations in runoff (Figure 6-7 and Figure 6-8), indicating that even in environments where soil loss was minimal from all the pasture treatments (see Chapter 5), erosion still played a role in determining the overall TP concentration in runoff. The PP fraction in runoff was lower than values reported for New Zealand pastures (Bargh 1978, 76%; Lambert et al. 1985, 85%; Cooke 1988, 85%). The mean percentage of TP that was in the particulate form in hillslope runoff at Vasey was greater in flow from the low P fertility plot (range 39-60%, mean 45%) compared to the high P fertility plots (range 1-53%, mean 24%). This supports evidence that particulate P in rivers draining agricultural catchments in Australia is largely attached to particles eroded from streambanks, cultivated soil and overgrazed pastures (Nelson et al. 1996; Davis et al. 1998), rather than from well-managed fertile pastures. Higher DP contributions can, however, be expected from well-managed pastures due to reduced soil erosion and increased desorbable P contents (Tham 1983; White and Sharpley 1996). Nash and Murdoch (1997) found no relationship between soil cover, which ranged from 79 to 100%, and P loss, suggesting that for Darnum soils and climate in Gippsland, the management of the dairy pasture was sufficient to minimise erosion. Suspended sediment concentrations are influenced by soil texture, degree of soil aggregation, vegetative cover, slope gradient and length, as well as rainfall event characteristics. The sediment transport capacity of individual runoff events, which is influenced by flow velocity and depth, and the pool of readily detached particulate and colloid material at the soil surface, would have determined the size distribution of SS entrained in runoff. The proportion of P-enriched clay fraction in SS would therefore vary between events and would contribute to the noise in the relationship between SS and PP (Romkens and Nelson 1974). At Vasey there were no clear relationships between rainfall or runoff characteristics and the sediment concentrations in hillslope runoff for individual events. Similarly in New Zealand, storm characteristics provided little explanation for the variation in PP in runoff from pastures, whereas the extent of the saturated area (i.e. the area available for entrainment of soil-derived P) was a reasonable predictor (Cooke 1988). At Maindample, the TS concentrations were greater than the SS concentrations measured at Vasey. The pasture cover was high and similar for all treatments at Maindample

191 (Schefe 1999), however, TS was significantly greater from plots in the low and high input catchments compared with the medium. The differences could have been related to a range of soil and pasture conditions such as soil structure, organic matter content, micro-topography and pasture density. Without replication of the pasture catchments, it was not possible to test whether the differences in soil loss were directly related to the treatments (i.e. degree of P input). Interestingly, C.Schefe (pers. comm.) found that in summer, when the rainfall simulations on the same 3 pastures were conducted, the low fertility catchment yielded a mean TS that was lower than that of the medium and high fertility catchments. This was attributed to uneven groundcover in the higher fertility pastures due to higher grazing pressures. Particulate P concentrations in runoff are influenced by both the amount of entrained soil, and the P content of the entrained soil. It was perhaps surprising, therefore, that PP concentrations in hillslope runoff at Vasey did not increase with increasing POlsen because it would be expected that soils with a high POlsen value would have greater amounts of P sorbed to soil particles, particularly in the soil layer interacting with runoff. In a New Zealand pasture catchment, event mean PP concentrations in streamflow increased after aerial topdressing and PP concentrations in subsequent events took longer (13 weeks) to return to pre-fertiliser levels than DRP concentrations (9 weeks), indicating a sustained influence of fertiliser P on PP concentrations (Sharpley and Syers 1979). It is possible, however, that because PP is usually associated with the clay fraction in sediment, the effect of POlsen status on runoff PP concentrations may have been masked by variations in the percentage of suspended sediment in the clay fraction between runoff events and plots (Romkens and Nelson 1974; Quinton et al. 2001).

Particulate P concentrations did increase with POlsen in the simulated runoff, however, possibly reflecting a more consistent particle size distribution in the suspended sediment due to the uniform rainfall energy and path length of the small plots. Gillingham et al. (1997) also found that in simulated runoff the response of PP to POlsen was similar to that of DRP. Similarly, in simulated runoff on bare soils in the USA, positive relationships were found between sediment extractable-P concentrations and fertiliser rates, (Romkens and Nelson 1974) and between PP in runoff and soil test P (Sharpley 1995).

P enrichment ratio

In runoff at both Vasey and Maindample, SS in hillslope runoff was enriched with P compared to the top 10 cm of the bulk soil from which it was derived, which is consistent with other studies (Sharpley 1980, 1985b; Gillingham and Thorrold 2000). A higher P concentration in sediment was expected when compared to the 0 – 10 cm depth of topsoil firstly due to selective removal of the nutrient-rich clay fraction of the top 1-2 cm of fertilised soil (Uusitalo

192 et al. 2001). Secondly, the clay fraction has a greater potential for P sorption because of its larger specific surface area, and through specific adsorption of P to sesquioxide minerals. Broadcast fertilisers tend to accumulate in the top 2 cm of pasture soils relative to the 2- 10 cm depth so higher fertility treatments were expected to have a higher PER than low fertility soils, as shown for the simulated runoff at Maindample and Vasey, and Maindample catchment runoff (Figure 6-10 and Table 6-4). Similarly, Sharpley (1980) found an approximately two-fold increase in the PER for soils amended with 50 kg P/ha compared to the 0.5 cm layer of soils receiving no amendments. A PER measured using Psoil in the top 2 cm is therefore likely to be smaller than the PERs reported here and would take better account of fertiliser treatment differences. For example, assuming that prior to the experiment, the 0-2 cm layer had a 30% higher Psoil than the 0-10 cm layer (eg. McCaskill and Cayley (2000) and McCaskill et al.

(2003)), and that 70% of added P remained in the top 2 cm, estimates of Psoil in the 0-2 cm layer of soils from the low, medium and high fertility treatments at Maindample lead to PERs for these pastures of 3.5, 2.7 and 4.9 respectively. These values are 61 – 71% of the PER values calculated from Psoil for 0-10 cm (Table 6-4). The original PERs (Table 6-4) were similar to other studies that considered a range of erosion conditions and soil types (Romkens and Nelson 1974; Sharpley 1985b; Cooke 1988). A PER of 5 was measured on stream sediment assumed to originate from the A horizon of Typic dystrochept and Aeric haplaquept soils in a pasture catchment in New Zealand (Cooke 1988) whilst a lower average ratio of about 1.5 was measured for the 0-10 cm layer of six repacked soils in Oklahoma and Texas, USA (Sharpley 1985b).

Predicting particulate P concentrations

Models that accurately reflect the concentration of P in runoff under a defined set of conditions allow results from intensive field studies to be extrapolated to broader geographical regions and can offer a cheap and quick method of estimating the impacts of land management practices on runoff water quality. The PER term in equation 6-1 is derived from measured values of PP so estimates of PP calculated using equation 6-1 could not be tested independently against the values of PP measured at Vasey or Maindample. Sharpley (1980) and Edwards et al (1996) found an inverse relationship between the sediment yield in runoff from pastures and the event PER which enabled an independent estimation of the PER, however the relationship did not hold for the sediment yields in the simulated runoff in the current study. The model is therefore only useful for estimating mean annual PP concentrations in runoff from pastures in environments similar to Maindample and Vasey using the sediment P concentrations measured in this study, and an estimate of sediment loss.

193 At Vasey, single values for the sediment P concentration (2 mg/g) and PER (2) over the range of levels of soil fertility investigated would be justified to derive the PER because of the lack of treatment differences. For Maindample soils, where treatment differences in the PER occurred, a sediment P concentration of 2.4 mg/g would be appropriate for pastures receiving upwards of 22 kg P/ha/yr and a value of 1.2 mg/g would be suitable for pastures receiving inputs of between 5.5 and 11 kg P/ha/yr.

6.4.2 Mobilisation of dissolved P

There was evidence at both Vasey and Maindample for a greater concentration of DRP, the form of P most immediately available for algal uptake (Sonzogni et al. 1982; Sharpley and Menzel 1987), in runoff from high P fertility pasture treatments when compared to low P fertility treatments. This is consistent with findings from other fertile dairy pastures in

Gippsland (POlsen 29 mg/kg) where 89% of the TP was in the DRP fraction (Nash and Murdoch

1997), and in UK pastures (POlsen 25 mg/kg) where 70% of TP in surface runoff was DRP (Heathwaite and Dils 2000). Others have also found positive relationships between DRP concentrations in runoff and the concentration of soil labile P (Sharpley et al. 1977; Yli-Halla et al. 1995; Pote et al. 1996). The increase in the concentration of DRP with increasing soil P fertility suggested that desorption of largely inorganic labile P from soil and other sources such as plant, faecal and organic matter and residual fertiliser increased as the quantity of this pool (represented by the

POlsen) increased. In Chapter 5 it was shown that the total P content in dung and pasture increased with POlsen and other research has shown that an increase in the total P content in these materials is largely due to an accumulation of inorganic P (RP)(Bromfield 1961; Barrow and Lambourne 1962; Bromfield and Jones 1972; Gillingham et al. 1980). Others have also proposed vegetation and faecal deposits on the pasture surface as potential sources of DP in runoff (McColl et al. 1977; Sharpley et al. 1994)(also see Chapter 5). The original source of P in runoff water is difficult to trace (Nash and Halliwell 2000), but may include rainfall, soil water, fertiliser, plant and dung material and soil particles. Efforts to determine the relative importance of soil, fertiliser and grazing events have shown that for winter runoff from a dairy pasture in northern Victoria, soil and organic matter were the main contributing sources of P (Nexhip and Austin 1998). Cooke (1988) found that whilst PP concentrations in runoff from New Zealand pasture were related to soil P sources, DRP concentrations were related to fertiliser inputs. Nash et al. (2000) also found P in runoff from a dairy pasture, which was dominated by DP fractions, was primarily influenced by time since fertiliser was applied. Others have similarly attributed increases in the DRP fraction in runoff and streamflow to dissolution of fertilisers applied to the

194 soil surface (Olness et al. 1975; Sharpley and Syers 1979; Lambert et al. 1985; Gillingham et al. 1997). The effects of fertiliser on runoff P concentrations tend to diminish over time, due to dissolution, plant uptake and sorption to surface soil components. For example, DRP concentrations in runoff were found to decay until 20 days after fertiliser was applied to a Victorian rainfed dairy pasture (Nash et al. 2000), and streamflow DRP concentrations took 9 weeks to return to pre-fertiliser levels after aerial topdressing of a pasture catchment in New Zealand (Sharpley and Syers 1979). DRP concentrations in runoff from pastures at Maindample and Vasey were less likely to have been affected by dissolution of residual fertiliser because fertiliser was applied more than 60 days in advance of runoff occurring. Instead, the increase in runoff DRP concentrations with increasing POlsen was likely to have been related to the labile P content of surface soil, plant and dung material.

Effects of suspended sediment on dissolved P fractions in runoff

Regardless of the mechanism of release into solution or runoff, DP is subject to physicochemical sorption processes. Although erosion was minimal at Vasey, the reduction in DP concentrations in hillslope runoff as SS concentrations increased suggested that DP was adsorbed to SS during runoff, transit and storage. Adsorption of DP to SS also occurred in catchments in the USA where for individual events there was an exponential decrease in DP concentrations in runoff with increasing erosion over similar and wider ranges of SS concentrations to those measured at Vasey (Sharpley et al. 1981c). Tham (1983) similarly found both the proportion of TP as DRP, and concentration of DRP in runoff was less for sheep pastures in Victoria with only patchy cover compared to ‘good’ cover. In contrast to the hillslope runoff, the sediment P concentration in the simulated runoff at Vasey also increased with increasing SS and POlsen concentrations (Figure 6-10, Table 6-4). Dissolved P concentrations tended to increase as sediment P concentrations increased (see Figure 6-14) which suggests that more P was desorbed as the amount of previously sorbed P in the sediment (likely to be proportional to the sediment P concentration) increased (see Figure 6-14 below). Similarly, Ekholm et al. (2001) attributed a positive relationahip between DRP and SS concentrations in runoff from cropped catchments to desorption of P from eroded soil. The results suggest that in high fertility pastures, a low level of erosion allows DP to adsorb onto SS (Nelson et al. 1996). The SS may settle out in calm conditions such as a farm dam, leaving the water column lower in DP and SS. Conversely, if erosion is inhibited, higher DP concentrations would remain in the water column (McKergow et al. 2003). Erosion of P- rich particles will also lead to higher equilibrium P concentrations in runoff and receiving waters.

195 Predicting dissolved P concentrations in runoff

In simulated rainfall-runoff events conducted on 0.64 m2 plots at Vasey, a positive relationship between DP concentrations in runoff and the POlsen status occurred, consistent with the model in equation 6-3, and with other laboratory and field studies (Sharpley et al. 1977; Pote et al. 1996). The ability and utility of using soil tests alone to predict P concentrations in runoff from pastures in Victoria is discussed further in Chapter 7. R, t and S from equation 6-3 could be assumed reasonably constant in these simulated runoff events but would vary between events under natural rainfall conditions. Differences in other parameters such as organic matter and clay content, which are related to the constants K, α and β (Sharpley 1983), could also significantly influence the accuracy of predicted DP concentrations (Sharpley and Smith 1989; Daniel et al. 1993). For equation 6-3, Sharpley et al. (1981a) assumed that W and R were proportional to each other for a given rainfall energy per unit volume, so doubling R led to a decrease in c by a factor of 2 β 2 . At Vasey, however, where the sediment to water ratio (1/W) was assumed to be reflected by observed SS concentrations (g/L), W and R or runoff volume were not directly related in either the simulated nor natural runoff. Without this dependency, an increase in W alone would be expected to dilute the DP concentration independent of R, despite an increase in the amount of P desorbed per gram of soil (Sharpley et al. 1981a). W would therefore need to appear in the denominator of equation 6-3 to reflect this dilution. Consistent with this expected dilution effect, a decrease in SS concentration (i.e. an increase in W) in simulated runoff at Vasey led to a decrease in DP concentrations. However, the decrease in the FWM DP concentrations with increasing SS concentrations in natural runoff from the 0.5 ha hillslope plots at Vasey suggested that further modification of equation 6-3, to describe the sorption processes between sediment and runoff P, would also be essential in order to predict event DP concentrations. The discrepancies between the kinetic model proposed by Sharpley et al. (1981a) and results from the current study suggest equation 6-3 is not appropriate for predicting DP concentrations in simulated or hillslope runoff events from pastures in Victoria. Storm et al. (1988) used parameters from a Langmuir sorption isotherm that reflected the equilibrium between soluble and sorbed P to modify predictions of DP calculated from equation 6-3. By adjusting for sorption processes during transport and storage and given sufficient time for these processes to occur, the runoff DP concentrations should in theory be no greater than the equilibrium P concentration (EPC) of the suspended sediment. An investigation of the influence of the EPC and other P sorption characteristics of a range of pasture soils in Victoria on runoff P concentrations is discussed in Chapter 7. The DP concentrations in runoff at Vasey and Maindample derived using the EPIC model (equation 6-6, Kd = 175) were between 4.2 to 5.7 times lower on average than the

196 measured values in hillslope and simulated runoff at both Vasey and Maindample (Table 6-7).

Kd values of from 31 to 42 therefore led to more accurate and reasonable predictions of the DP concentrations (Table 6-7). The appropriateness of using a single partition coefficient (Kd) to describe the relationship between DP concentrations and sediment P content in runoff was assessed using data from simulated runoff events at Vasey and Maindample (Figure 6-14).

Table 6-7: Estimated and measured mean DP concentrations (mg/L) in hillslope and simulated runoff

Hillslope runoff 1998 Simulated runoff Plot DPEPIC DPEq6-10 FWM Treatment DPEPIC DPEq 6-10 Mean DP DP Vasey Kd=175 Kd=31 Kd=175 Kd=42 1 0.05 0.29 0.11 0.23 A 0.05 0.21 0.11 0.24 2 0.03 0.19 0.09 0.09 B 0.12 0.50 0.49 0.42 3 0.06 0.35 0.04 0.36 C 0.11 0.48 0.37 0.34 4 0.09 0.48 0.02 0.82 D 0.03 0.12 N/a 0.16 Maindample Kd=175 Kd=34 L 0.03 0.18 0.16 0.18 M 0.05 0.26 0.17 0.23 H 0.13 0.64 0.77 0.61 )

6.00

5.00

4.00 Maindample 3.00 Vasey (mg/g) 2.00

1.00

Sediment P concentration (PP/ SS (PP/ P concentration Sediment 0.00 0.00 0.50 1.00 1.50 DP (mg/L)

Figure 6-14: Quantity-intensity relationship between the sediment total P concentration (mg/kg) and DP (mg/L) in simulated runoff with lines showing regressions for Vasey (broken line, Sediment P concentration = exp(0.41lnDP + 1.20), P<0.001, R2=0.31) and Maindample (solid line, Sediment P concentration = 2.17*DP+0.84, P<0.001, R2=0.49)

197 The relationship at each site was highly scattered because the data were drawn from independent simulated runoff events and plots but there was evidence that the DP in simulated runoff was approaching equilibrium with P in the solid phase. The relationships were described well by logarithmic curves of the form in equation 6-9, where the sediment P concentration is sedimentP, rather than a simple linear association and single Kd value as assumed in the EPIC model.

sedimentP−c DP = e a (6-9)

The coefficients a and c, estimated from data in Figure 6-14, are 671 and 3051 for Vasey soil (R2 = 0.26) and, 863 and 2711 for Maindample soil (R2 = 0.52), respectively. In the absence of measured values, the sediment P concentration can be estimated using Psoil and PER according to equation 6-1, giving the overall approximation for DP in equation 6-10.

(Psoil xPER−c) DP = e a (6-10)

Dissolved P concentrations estimated using equation 6-10 (DPEq6-10) are given in Table

6-7. The match between estimated (DPEq6-10) and measured DP concentrations was poor for runoff from the 0.5 ha hillslope plots at Vasey, particularly for higher DP concentrations. There was a better match between estimated (DPEq6-10) and measured DP concentrations for simulated runoff, presumably because the equation constants were derived from the simulated runoff data but possibly also because DP in simulated runoff was more strongly influenced by sorption to the SS than was DP in the hillslope runoff. A potential deficiency of the EPIC model described in equation 6-4 is its failure to describe the nutrient enrichment of SS compared to the source soil. A comparison of equations 6-6 and 6-7 suggests the EPIC model assumes the sediment P content is equal to the soil labile P content. This is contrary to the principal that sediment is usually enriched with P compared with the topsoil, a finding confirmed by data from Vasey and Maindample. The PERs described for Vasey and Maindample pastures (range 1.8 to 8, Table 6-4) described the differences in total P content in soil and suspended sediment. An even greater degree of enrichment would therefore be expected when comparing a measure of the labile P content of the topsoil (i.e. CLP) to the total P content of the suspended sediment. For example, over a range of treatment mean POlsen (corrected to 0-10 cm depth) at Vasey of 5 to 21 mg/kg, the corresponding sediment total P content of simulated runoff was 940 to 2450 mg/kg, suggesting enrichment ratios of between 116 and 188. Overall, the best predictor of annual DP concentrations was equation 6-6 using a linear value of between 30 and 40 for Kd, rather than the default value of 175, or the non-linear approximations derived from the experimental data. Using the linear Kd parameter, equation 6-6

198 is analogous to the simple relationships developed between runoff and soil test P by Sharpley et al. (1982) and Edwards et al. (1996), and is reasonably well suited to predicting annual average DP concentrations in runoff from pastures.

Effect of ionic strength of runoff water on dissolved P concentrations

At Vasey, the range of EC values measured in plot 1 was larger than would be expected from runoff dominated by surface flows. Because the groundwater in the Vasey soil was highly saline, the high ionic strength of surface flows in this plot added weight to the conclusion in Chapter 4 that discharging groundwater contributed to surface flow in plot 1, probably as direct discharge and/or return flow. The influence of a rising groundwater on the EC of near-surface water was also found on duplex soils in South Australia (J.Cox, pers. comm.). At Vasey, there was an inverse relationship between DP concentration and EC for hillslope runoff in plot 1 in both 1998 and 2000 (Figure 6-12). Two explanations for this relationship can be offered. Firstly, considering EC as a surrogate variable for the ionic strength of the runoff, the relationship was consistent with findings that less P is desorbed from soil in solutions of pH greater than about 5 with high ionic strength, than in weaker solutions (Ryden et al. 1977; Barrow and Ellis 1986). McDowell and Sharpley (2001) also found more soil P was desorbed in water than in CaCl2 solutions, and Yli-Halla et al. (1995) attributed some of the temporal variation in DRP concentrations in runoff in Finland to variations in ionic strength. Secondly, the inverse relationship could be attributed to the source of the runoff water. The groundwater at Vasey was highly saline and had low P concentrations (see Chapter 5) and in Chapter 4 it was shown that return flow of subsurface water contributed to surface runoff in this plot. Return flow of low P groundwater was therefore likely to have diluted the P-enriched surface flows. The ionic strength of water was not included in any of the models discussed here. However, considering its strong influence on runoff DP concentrations at Vasey, solution ionic strength is a factor that may need to be incorporated to better predict P concentrations in surface flows occurring on saline soils or influenced by saline groundwater discharge.

6.4.3 Effect of flow volume on runoff P concentrations and fractions

There was a trend for the annual FWM concentrations of TP in runoff from the four pastures plots at Vasey to decrease as the annual runoff volume increased. If in equation 6-7, annual flow was substituted for the event flow parameter TF, and DG and DF were ignored over the annual time scale, the model of Nash et al. (2000) would describe the lower annual mean TP concentrations in runoff from the hillslope plot 1 at Vasey compared to plot 3, which had a similar level of POlsen. Grazing and fertiliser events were less likely to influence annual P concentrations in runoff from sheep pastures compared to dairy pastures because fertiliser is

199 normally applied well in advance of the runoff period and because grazing pressure is more uniform over time. Despite a dilution effect occurring for annual mean TP concentrations, there was a significant (P<0.05) trend for lower concentrations of TP in runoff events less than 3 mm compared to larger, less frequent events in plot 1. This may have been due to less soil disturbance during low flows, as many of the low flows were generated by subsurface return flow, surface discharge and saturation excess flow after only a few millimetres of rainfall. The event-flow results contrasted with the dilution effect described by equations 6-3 and 6-9, but were in agreement with findings of Tunney et al. (2000b), Haygarth et al. (1998) and McColl et al. (1977). Haygarth and Jarvis (1999) suggest high magnitude events are likely to be associated with large episodic losses of P (eg.McColl et al. (1977), Cooke (1988)), particularly if occurring soon after spreading of fertiliser or manure (eg. Preedy et al. (2001)), and are also likely to be associated with physical P transfer processes such as erosion, rather than dissolution processes (eg. Salvia-Castellvi et al. (2001)). Interestingly, downstream of the Vasey field site, TP and TN concentrations in streamflow were positively correlated with flow, turbidity and suspended sediments (Wagg 1997), however, the elevated TP concentrations in the hillslope runoff appeared to be due more to increased DRP mobilisation than to enhanced erosive processes (i.e. a larger increase in %DRP compared to %PP, Table 6-6). An increase in the effective depth of interaction between the runoff water and the soil with increasing flow volume may have increased the amount and concentration of desorbed P in runoff. Similarly, Haygarth et al. (1998) found DP concentrations increased more than the PP fraction with increasing runoff discharge rate from grazed pastures in the UK.

6.4.4 Scale effects on P concentrations in runoff

Some of the merits and limitations of the two experimental scales used in this study were discussed in Chapter 5. In this chapter the effect of scale on the concentrations and forms of P and SS in runoff from pastures at Vasey was investigated. The processes of P and SS mobilisation occurring in the small-plot simulator study were not expected to match those in hillslope runoff. The delivery of sediment in runoff per unit area usually declines as the size of agricultural catchments increases due to greater redeposition of entrained particles, disruptions (such as contour drains) to the hydrological connectivity of flow pathways, a reduction in drainage density and flatter topography (Ciesiolka and Freebairn 1982; Prairie and Kalff 1986). Ekholm et al. (2000), however found no scale effect for catchments greater than 20 km2 in Finland, with the percentage of cultivated fields and land slope having greater influence on sediment loads. In other cultivated soils, small plot simulators only generate sheet erosion and therefore underestimate rill and gully erosion at the landscape scale (Truman et al. 2001; Hamed

200 et al. 2002). However, landscape-scale erosion is usually overestimated using simulated rainfall on small plots because the variable spatial distribution of runoff and sediment generation, and storage of sediment within the catchment is not accounted for, and higher than normal rainfall intensities provide greater energy for erosion (Scoging 1989; Hudson 1995). Consistent with these latter principles, SS concentrations in simulated runoff in the current study were significantly higher than those measured in hillslope runoff. The rainfall intensity of 48 mm/h, maintained over an hour, represented a 1 in 100 yr rain event at the field site location (Anon 1987). In contrast, the range of TP concentrations in simulated runoff were similar to those measured in hillslope runoff. This suggests that small plot rainfall simulators may be useful for comparing pasture treatment effects on runoff water quality and may provide reasonable predictions of concentrations of TP in hillslope runoff from pastures. However, concentrations of PP were higher, and the maximum treatment mean dissolved P (DP) concentrations were lower in simulated runoff than in hillslope runoff indicating that the processes of mobilisation differed between the spatial and temporal scales. This is consistent with other studies where PP concentrations and loads decreased with increasing path length (McColl and Gibson 1979; Prairie and Kalff 1986), and where an increase in the duration of contact between runoff and soil increased DP concentrations (Sharpley et al. 1981b; Gillingham et al. 1997; Gascho et al. 1998). Ekholm et al. (2000) similarly attributed an increase in the concentration of DP with catchment size (>20 km2) to release of P from suspended river sediment over time. Alternatively, higher SS concentrations in the simulated runoff may have lowered the DP concentrations by providing a greater surface area for adsorption of P during transit. In contrast Cornish et al. (2002) found that variation in the area of land contributing to runoff influenced DP concentrations in runoff, but that DP concentrations in simulated runoff from small plots (1 m2) and rainfall-runoff from a 4 ha dairy pasture were similar. Further evidence for P mobilisation processes differing between experimental scales was that the electrical conductivity (EC) and volume of runoff events had an influence on P concentrations at the landscape scale. Higher annual runoff volumes diluted P concentrations in runoff from hillslope pastures of similar POlsen, and P concentrations decreased with increasing runoff event EC in a hillslope plot where groundwater discharge contributed to surface runoff. Transformations of P associated with these hillslope-scale processes were not observed at the small-plot experimental scale because these factors were relatively uniform. Whilst simulator runoff from small plots did not reliably reflect the relative importance of the P mobilisation and transformation processes that occurred at the hillslope – annual timestep scale, it is likely to provide upper estimates of SS and P concentrations that may be expected under extreme rainfall conditions.

201 6.4.5 Mobilisation of organic P

At Vasey, there was little change in the concentration of DUP (<0.20 mg/L) in hillslope and simulated runoff as POlsen increased (Figure 6-2 and Figure 6-3). Similarly DUP was <5% of DP in simulated runoff from four repacked pasture soils of varying fertility in the USA (Sharpley et al. 1982). The percentage contribution of total unreactive P (DUP + PUP) to runoff P in the low fertility plot was 99 and 55% in 1998 and 2000 respectively. This was a greater contribution than that measured for surface and A horizon throughflow (45%) from pasture plots of similar POlsen status in the UK (Haygarth et al. 1998). The organic P component of faecal (Bromfield 1961) and plant material (Bromfield and Jones 1972; Gillingham 1980) remains fairly stable with increasing pasture, and Psoil content (Floate 1970). Because P that is leached from both these sources has the potential to contribute to the DUP component of runoff, this is consistent with the findings that the DUP fraction generally remained stable with increasing POlsen. At Maindample the DUP fraction in hillslope runoff was larger than at Vasey, and was the largest fraction for the low and medium fertility catchments but like at Vasey, the mean concentration was less than 0.2 mg/L for all treatments. DUP has been found to exceed levels of DRP in soil solution in temperate grasslands (Ron Vaz et al. 1993), and in lake water (Broberg and Persson 1988). Based on a study on temperate grassland soils in Scotland, Chapman et al. (1997) suggested that the supply and amount of DUP in soil water was controlled by biological mechanisms rather than sorption processes. In their study, the largest P fraction in simulated runoff was PUP, reflecting higher SS concentrations in runoff compared to Vasey. The PUP fraction in runoff includes tightly bound soil P as well as particulate plant and dung material and is the form of P that typically dominates the total organic P pool in natural waters (Broberg and Persson 1988). The large contribution of the PUP fractions of TP in runoff from the Maindample pasture indicated that the sources of particulate organic P in pastures at Maindample may require management in order to reduce the TP losses in runoff.

6.4.6 Options for minimising P losses from pastures

Both PP and DP fractions were important in surface runoff from the hillslope plots at Vasey, with DP increasing in importance in runoff from the more fertile pasture treatments. It may be difficult to reduce erosion in pasture systems like these where near complete groundcover is maintained throughout the runoff season. However PP may still contribute up to 50% of TP losses, which highlights the importance of managing pastures in a way that minimises soil erosion at all times. Groundcover can be maintained above the critical level of 70% through controlled grazing, supplying adequate soil nutrition for plant growth and by minimising soil disturbance such as pugging in runoff source areas (McDowell et al. 2003),

202 particularly along stream banks. The success of restricting stock access to seasonally saturated areas in reducing P losses depends on how well the remaining vegetation can trap and assimilate P in runoff. The use of vegetative buffer strips was recommended by Cooke (1988) for a region in New Zealand where more than 85% of transported P was in particulate form. At Vasey, however, there was a lesser PP contribution, so the use of buffer strips to control erosion is unlikely to be as effective in reducing DP, and hence TP losses.

The positive relationship between DRP concentrations in runoff and POlsen means that it may be necessary to reduce the labile P pool of pastures in areas where the majority of runoff occurs in order to reduce the TP concentrations in runoff. Soil can be depleted of P by lowering P inputs, increasing plant uptake and removing plant and animal products. Alum, coal combustion by-products and other P sorbing amendments may also be applied where excessive amounts of labile P have built up (Stout et al. 1998), provided these do not adversely affect the environment themselves (Summers et al. 1996; Frossard et al. 2000). Depleting the soil P fertility of a runoff source area and allowing it to revert to water-tolerant species such as sedges and rushes that are unlikely to invade well-managed neighbouring pastures could improve the quality of runoff. These runoff areas would need to be carefully managed to avoid the incursion of pasture weeds that are capable of growing at low soil nutrient levels (Snaydon 1981).

6.5 Conclusion

The main analytical fractions of P in runoff under natural and simulated rainfall conditions from pastures in Victoria were measured to identify the important P mobilisation processes occurring. Both PP and DP were important P fractions in runoff from sheep pastures at Maindample and Vasey. Concentrations of DRP in runoff increased with treatment POlsen suggesting desorption, extraction and dissolution of predominantly inorganic P from surface materials are the most important processes of P mobilisation in runoff as pasture soil P fertility increases. In both hillslope and simulated runoff there was evidence that adsorption relationships between DP and SS influenced DP concentrations in runoff. Therefore a model that incorporates the effects of both desorption and dissolution of P from surface materials, and adsorption onto entrained sediment may provide a better prediction of DP concentrations than models that focus on either desorption or sorption processes. DP concentrations in natural runoff from hillslope plots were more strongly influenced by the POlsen (being representative of the P availability in soil, plant, residual fertiliser and organic matter) than by sorption equilibria with eroded soil particles (i.e the sediment P concentration or mass of eroded particles). Particulate and organic P contributed a reasonably consistent amount to runoff P at Vasey from both high and low P fertility pasture. In simulated runoff, the PP fraction was more important, probably because there was greater soil loss due to the high artificial rainfall intensity and lack of opportunity for redeposition of SS in the small plots. Despite low SS (<0.30 g/L)

203 and PP concentrations (<0.20 mg P/L) in runoff from all hillslope plots, there was a positive relationship between PP and SS concentrations, highlighting the need to minimise erosion from pastures. Suspended sediment was enriched with P compared to the top 10 cm of surface soil by factors of between 2 and 8. However, concentrations of PP in hillslope runoff did not increase with soil P status. This suggested PP losses were more heavily influenced by factors that influenced the concentration and size distribution of eroded particles, than by the amount of P adsorbed onto those particles. In contrast, PP concentrations in the simulated runoff did increase with increasing POlsen, however, probably reflecting a more uniform particle size distribution due to the standardised rainfall and runoff conditions. It was also shown that PP losses are difficult to predict without information regarding erosion rates and sediment P enrichment ratios. An investigation of the effect of experimental scale on the forms of P measured in runoff suggested rainfall simulations on small plots are useful for identifying treatment effects and TP concentrations in runoff from pastures but may not reliably reflect the relative importance of P mobilisation processes that need to be targeted for mitigation of nutrient losses. In general, erosion was minimal from the well managed pastures, which suggested that for pastures with close to 100% groundcover, reducing the amount of inorganic P available at the soil surface in areas prone to the production of runoff may reduce the TP concentration in runoff and hence an opportunity to reduce total P losses and improve the quality of runoff.

204 CHAPTER SEVEN

7 Predicting P concentrations in runoff using soil P characteristics

7.1 Introduction

The pathways and amounts of P movement from sheep pastures at Vasey, Victoria were reported in Chapters 4, 5 and 6, where it was shown that the main factor controlling the transport of both soluble and particulate P from agricultural land is the volume of water movement along surface or subsurface pathways. However, P loads in surface runoff were also found to be influenced by the availability of P for desorption, dissolution and entrainment into moving water (Romkens and Nelson 1974; Olness et al. 1975; Sharpley et al. 1982). The buildup of surplus P in soils that are prone to runoff and erosion has also been identified as the main cause of deteriorating water quality in the USA (Sharpley et al. 2001a). Soils that present a risk of degrading water quality when runoff or drainage occurs have been identified using empirical relationships between dissolved P (DP) concentrations in runoff and either soil test P (STP) levels (Weld et al. 2001) or indices of P sorption and desorption from soils (Breeuwsma and Silva 1992; Bramley et al. 1998). In both the USA and Australia, this information has been used to develop indices that identify parts of an agricultural catchment with the greatest risk of off-site P movement (Sharpley et al. 1999; Bramley and Wood 2000; Eghball and Gilley 2001; Sharpley et al. 2001b). The index approaches take account of sources of P as well as transport mechanisms and flow pathways (Lemunyon and Gilbert 1993; Gburek et al. 2000). This chapter describes some soil P characteristics that can be used to predict P concentrations in runoff, and investigates relationships between soil P tests and runoff P at Vasey and a range of other pasture sites in Victoria.

7.1.1 Soil P quantity and intensity tests

Soil P measured using traditional agronomic soil P tests has shown strong, positive associations with direct measures of DP concentrations in natural or simulated runoff from a range of soils (Sibbesen and Sharpley 1997). Table 7-1 summarizes the soils, extractants used and runoff conditions for some of these empirically derived relationships. Relationships between runoff P concentrations and STP are often developed based on concentrations of DP, rather than TP, and so do not account for concentrations of P adsorbed to suspended particulate

205 matter that is potentially available to aquatic organisms through desorption and redox mechanisms (Daniel et al. 1994; Baldwin 1996; Sharpley et al. 1996).

Table 7-1: Chemical description and extracting conditions for common soil P extractants and their relationships with runoff P concentrations

STP Extracting conditions Estimate of runoff P concentration R2 Ref. A Olsen 0.5M NaHCO3, pH 8.5, Sim runoff in situ, grassland USA 0.72 1 1:20 soil: solution, 30 Sim runoff in situ, grassland UK 0.95 2 min Hillslope runoff grass ley, Finland 0.96 3

Bray-1 0.03M NH4F + 0.025M Sim runoff grassland, USA 0.91- 4 HCl 0.95 Morgan 1M NaOAC, pH 4.8 Drainage and runoff, Denmark, 0.92 5 Ireland, UK, USA

Mehlich 0.02N CH3COOH + Sim runoff ex situ, mixed landuse 0.67 6

III 0.025N NH4NO3+ USA

0.015N NH4F + 0.013N

HNO3 +0.001M EDTA, pH 2.5 soil:solution 1:10, 5 min NaCl 0.1M NaCl, soil:solution Rainfall runoff, fertilised and 0.97- 7 1:20, 40 h, 0-1 cm depth unfertilised grassland, New Zealand 0.99B A Simulated rainfall-runoff B r, correlation coefficient 1: Pote et al. (1996); 2: McDowell and Sharpley (2001) 3: Turtola and Yli Halla (1999); 4: Pote et al. (1999b); 5: Tunney et al. (2001); 6: Weld et al. (2001); 7: Sharpley et al. (1978)

STPs measure variable proportions of both the ‘quantity’ of soil P available for uptake by plants during a growth cycle and the ‘intensity’ (i.e. concentration) of soil solution P. The soil P ‘buffering capacity’, defined as the change in quantity required for a change in intensity, governs the partitioning of P between soil and solution (Moody and Bolland 1999). The intensity factor has been used as an indicator of whether P is likely to be adsorbed to or desorbed from soil that is in contact with runoff, drainage and stream waters (Taylor and Kunishi 1971; Kelly 1999; Edis et al. 2002). For example, the equilibrium P concentration (EPC), derived from P sorption isotherms, represents the solution P concentration at which P is neither adsorbed to, nor desorbed from soil (Beckett and White 1964). Soil P extractants that primarily measure P intensity, such as CaCl2 (Kuo 1996) and water (Sissing 1971) have also been used instead of direct measurements of runoff P concentrations to estimate the risk of P release from soil to water (Gartley and Sims 1994; Daly et al. 2001).

206 There is little information on the relationships between soil P and runoff P for Australian agricultural soils. One study, on irrigated dairy pastures in Victoria, found a linear association between the average TP concentrations in natural winter runoff and Olsen P (POlsen) over a POlsen range of 8 to 46 mg/kg (Nexhip and Austin 1998). Relationships between runoff and STP often break down when comparing many soil types and land uses (Pote et al. 1999a; McDowell and Sharpley 2001) due to variation in chemical and physical properties of soil and surface material that influence the release of P to water (Sharpley 1995). For example, Simard et al. (1994) found that the water extractable P content of five soils could only be described by a single relationship by including parameters for soil pH, organic matter and Ca content in addition to STP. Therefore soil P indices that reflect the current soil P status relative to the P sorbing characteristics of soils may be more effective measures of the risk of P loss from soil to water than STP alone, among variable soil types. The degree of soil P saturation (Psat), is a measure of the amount of sorbed P relative to either the total P sorption or P buffering capacity of the soil (Sharpley 1995; Sibbesen and Sharpley 1997). The following section discusses two of the commonly applied indices of Psat, the

Langmuir (PsatL) and oxalate (Psatox) indices.

7.1.2 Indices of soil P saturation

The Langmuir Psat index (PsatL)

The Langmuir Psat index (PsatL) is calculated utilising a user-defined STP to represent the quantity of labile P, expressed relative to the maximum quantity of P the soil can adsorb

(Pmax) (Sharpley 1995; Paulter and Sims 2000). Pmax is calculated from fitting a Langmuir equation to a P sorption curve (Sposito 1989).

The percentage PsatL may therefore be defined as:

STP Psat L = ×100 (7-1) Langmuir Pmax

Levels of PsatL are rarely directly comparable because of differences in both the methods used to measure Pmax and the choice of extracting solutions used to measure the STP level. Estimates of Pmax depend on the range of P concentrations used, and the estimate of previously sorbed P used to generate the P sorption curves, as well as goodness of fit of the Langmuir equation (Veith and Sposito 1977; Barrow 1978). There is little standardisation of methods used to measure P sorption curves because the range and number of points required depends on the purpose of the study. For example, to investigate the P buffering capacity of the soil at solution P concentrations optimal for pasture plant growth (approximately 0.3 mg/L, (Ozanne and Shaw 1967)), solution P concentrations

207 beyond 5 mg/L would be inappropriate, whereas to assess the P loading potential of a soil to which sewage sludge is to be applied, much higher solution P concentrations would be necessary. The standard method published by Rayment and Higginson (1992) in Australia recommends concentrations up to 25 mg P/L, whilst the Southern Extension/Research Activity - Information Exchange Group in the United States suggests concentrations up to 100 mg P/L (Pierzynski 2000). An estimate of sorbed P in the soil prior to P additions during sorption experiments is referred to as the quantity, Q (mg/kg). Q is added to the calculated quantity of P adsorbed from solution to calculate the total amount of P adsorbed at each P addition, and the total sorbed is subsequently plotted to derive a sorption curve. Q can be estimated using least squares optimisation of an equation fitted to a P sorption curve (Barrow 1978; Barrow et al. 1998; Nair et al. 1998) or can be measured directly. Direct methods used to measure Q include resin anion exchange (Fitter and Sutton 1975; Kelly 1999; Edis et al. 2002), isotopic exchange (Olsen and

Watanabe 1957; Holford et al. 1974; White and Taylor 1977b) and extractions using CaCl2,

H20, NaHCO3 and NH4 oxalate as extractants (Sharpley et al. 1981b; Van Der Zee and Van Riemsdijk 1988; Freese et al. 1992; Yuan and Lavkulich 1994; Burkitt et al. 2002). The most appropriate estimate of Q can be chosen based on practicality (such as availability of routine soil test data), and on applicability, such as attempting to represent the field conditions in which desorption is expected to occur. For example, Sharpley et al. (1981b) used two 1-hour water or NaHCO3 extractions to estimate Q and acknowledged that the estimates were not absolute measures of the desorbable fraction of soil P, but represented only the P that could be released to water within a 3 hour period of rainfall and runoff. NaHCO3 extracts contain a labile fraction of both inorganic and organic P (Perrott and Sarathchandra 1989), with the molybdate reactive P pool being almost entirely orthophosphate (Coventry et al.

2001). In contrast, Freese et al. (1992) used ammonium oxalate extractable P (Pox) as an estimate of the total phosphate originally sorbed by amorphous iron and aluminium in their assessment of the total P sorption capacity of acid soils in Germany. Oxalate extractable P was approximately four times greater than the estimate of the reversibly adsorbed P measured using Fe-oxide strips. Regardless of the method chosen, estimates of Q are arbitrary because in reality there is a continuous spectrum of adsorption affinities for P in heterogenous soil (White and Taylor 1977a), and any laboratory desorption method is likely to differ from the processes occurring in situ due to the drying and grinding of soil during sample preparation, the wide soil to solution ratios used, and the variable soil-solution contact times.

208 The oxalate Psat index (Psatox)

The oxalate index (Psatox) is calculated as;

Pox Psatox = ×100 (7-2) α(Alox + Feox )

where Pox, Feox and Alox are the amounts of P, Al and Fe (mg/kg) extracted by acidified ammonium oxalate, and α is an experimentally derived coefficient representing the fraction of

Feox and Alox that is involved in P sorption (Breeuwsma and Silva 1992).

The oxalate index uses the sum of Feox and Alox as a surrogate estimate of the soil’s P sorption capacity. Acid ammonium oxalate extracts amorphous minerals that form the bulk of the P sorption capacity of soils, namely Al and Fe in acidic soils (Toreu et al. 1988; Freese et al.

1992). Feox + Alox was well correlated with the total P sorption capacity of soils in the Netherlands (Van Der Zee and Van Riemsdijk 1988), acid soils in Germany (Freese et al. 1992), soils in Denmark and Quebec (Borggaard et al. 1990; Simard et al. 1994), acid soils with low organic matter in Mid-Atlantic USA (Paulter and Sims 2000), and soils in Queensland, Australia (Toreu et al. 1988; Kelly 1999). The α coefficient in equation 7-2 is arbitrary because it depends on the length of equilibration time used to experimentally derive the total P sorption capacity of a soil (Yuan and Lavkulich 1994). It has been determined experimentally as 0.48 for Dutch soils equilibrated for 40 hours (Van Der Zee and Van Riemsdijk 1988) and as 0.38 for USA and Dutch soils equilibrated for a similar length of time (Paulter and Sims 2000). Pote et al. (1999b) simplified

Psatox by assuming α was equal to one.

Although traditionally used for acid soils only, the Psatox index may also have some relevance to alkaline soils, because Ryan et al. (1985) suggest amorphous Fe may be a significant component of the sorption capacity in calcareous soils. Beauchemin et al. (1996) also found (Alox and Feox) were better estimates of the P sorption capacity than exchangeable Ca, clay, organic matter contents and other soil properties for neutral to alkaline Gleysolic soils in Quebec. Indices of P saturation have provided single relationships with runoff P concentrations for some soils (Sharpley 1995), however, others have found that routine STPs have similarly uniform correlations with runoff P concentrations across soil types (Kleinman et al. 1999)](Pote et al. 1999b). Most Australian soils are deficient in the amount of P available for plant growth as a result of strong leaching and weathering during an extended geological history and the coarse textured nature of many parent materials (Costin and Williams 1983) and many have a high capacity to adsorb P. Whilst there is a large body of information regarding P sorption characteristics of Australian soils (Barrow 1980; Holford and Cullis 1985; Toreu et al. 1988;

209 Singh and Gilkes 1991; Barbare et al. 1997) there is little reported information on levels of P saturation or its relationship with runoff P concentrations.

7.1.3 Soil P thresholds for water quality targets

Relationships developed between runoff P and STP or Psat have been used in Europe and the USA to establish threshold STP levels above which the potential for deterioration of water quality is unacceptable (Breeuwsma and Silva 1992; Sibbesen and Sharpley 1997; Sims et al. 2000). Imposing upper limits on allowable levels of soil P can then be used within a strategy to achieve water quality targets. For example, the threshold level of the Mehlich-3 STP (Pierzynski 2000) determined for soils in north-eastern USA to maintain runoff P concentrations below 1 mg/L was 450 mg/kg (Weld et al. 2001). This is approximately equivalent to a POlsen value of 165 mg/kg (Sibbesen and Sharpley 1997), a level which rarely occurs in agricultural soils in Australia, and which satisfies a water quality standard 20 times higher than guideline concentration of 0.04 mg TP/L for Victorian lowland rivers and streams (Anon 2000a). Acceptable concentrations of P in runoff have not been set in Australia, but will partly depend on the degree of dilution and attenuation of P concentrations in hillslope runoff between the ‘edge of field’ and receiving streams and reservoirs as well as the sensitivity of the receiving waters to eutrophication (Williams and Wan 1972; Harris 1995). Threshold soil P levels developed using the Psat measure are expected to have broader applicability than thresholds based on STP. Psatox greater than 25% (equation 7-2, where α = 0.5) was found to increase the risk of leaching and runoff loss of P in the Netherlands

(Breeuwsma and Silva 1992). Others have determined Psatox (α = 0.5) threshold levels between 11 and 50% for a range of water quality criteria (Beauchemin and Simard 1999; Pote et al. 1999b; Turtola and Yli-Halla 1999; Hooda et al. 2000; Kleinman et al. 2000). Because both STP and Psat thresholds were derived from empirical relationships, their relevance to Australian soils will be limited, highlighting a need for threshold soil P levels to be determined for Australian soils based on local water quality criteria.

7.1.4 Aims

The primary aim of work discussed in this chapter was to test the utility of soil P tests as indices of potential runoff P concentrations in an Australian context. Olsen P, water extractable P, EPC and Psat were measured for a range of soils from grazed pastures in Victoria and related to P concentrations in runoff. Advantages and limitations of each soil P test as predictive measurements, and the concept of threshold soil P levels are also discussed. The degree of P saturation of 90 agricultural soils from across Australia was then compared with environmental threshold levels for soils overseas in order to set the potential for P movement from pastures in southern Australia in a more global context.

210 7.2 Materials and methods

7.2.1 Victorian pasture soils

Soil sample collection

Soils were sampled from five grazed pasture sites across Victoria. The sites included a commercial sheep property at Vasey (SGS national experiment site) and the Long Term Phosphate Experiment at the Pastoral and Veterinary Institute at Hamilton, both in southwest Victoria, sheep and cattle properties at Maindample and Ruffy in north-east Victoria (SGS NE sites), and a dairy property at Darnum in Gippsland, eastern Victoria. Site characteristics and soil sampling procedures are described below, and selected soil properties are documented in Table 7-2.

Vasey (37o24’ S, 141o55’ E)

The Vasey SGS national experiment site was adjacent to the runoff experiment described in Chapters 3 - 6. The location, climate, geology and soils were therefore the same as those described in Chapter 3 for the hillslope runoff experiment. Other soil properties were described by Cox et al. (1998). Soil was sampled from three replicate paddocks of the set-stocked low P fertility pasture treatment of the SGS national experiment (Chapman et al. 2003). The pastures were grazed by Merino ewes and sown to phalaris and subterranean clover in 1994. The SGS experiment began in 1997 and fertiliser P was applied at 8 kg P/ha in 1998 and 1999. No P was applied in 1997 or 2000. Twenty topsoil cores (2.5 cm diameter) of 0-5 and 5-10 cm depths were collected randomly and bulked in each paddock on 23 October 2000. Four cores (3 cm diameter) of B horizon soil material were collected and bulked for each of the paddocks. Soil was also sampled as described in Chapter 5 to determine the POlsen and water extractable P of hillslope and rainfall simulator plots.

Hamilton (37o49’S, 142o04’E)

The Hamilton Long Term Phosphate Experiment was established in 1977 and consisted of 18 paddocks originally sown to perennial ryegrass, phalaris and subterranean clover which were grazed by sheep. Average annual rainfall at the site is 700 mm. The duplex soil at the Hamilton site is classified as a mottled-sodic, eutrophic Brown Chromosol (Isbell 1996) derived from basalt. Further details of this soil type are described by McCaskill and Cayley (2000). Soil was sampled on 6 November 2000 from three low fertility paddocks, which had received virtually no fertiliser since their establishment. Fifteen soil cores (2.5 cm diameter) of

211 0-5 and 5-10 cm depths were collected randomly and bulked in each paddock. The B horizon began at an average depth of 55 cm and was of medium clay texture. Four cores (3 cm diameter) of B horizon soil material were collected and bulked for each of the paddocks.

Maindample (36o59’S, 145o59’E)

The Maindample SGS national experiment site was located near Mansfield in north- eastern Victoria. Site and treatment details are reported by Ridley et al. (2003). The soil is duplex, derived from sedimentary sandstones and siltstones of Silurian-Lower Devonian origin and classified as a Brown Sodosol (Isbell 1996). The soil profile has a sandy loam topsoil overlaying a sodic, medium clay B horizon with the sharp texture contrast occurring at an average depth of 40 cm. Average rainfall is 720 mm/yr. Soil samples were collected on 31 October 2000 from low, medium and high soil fertility pasture catchments. The catchments were 4.9, 13.7 and 11.5 ha in size respectively and were instrumented to measure runoff. High, medium or low P fertiliser rates of 22, 11 and 5.5 kg/ha/yr respectively, had been applied to the pastures since 1992. The high and medium input catchments had been sown in 1972 with phalaris and subterranean clover. Fifteen soil cores (2.5 cm diameter) of 0-5 and 5-10 cm depths were collected and bulked from upslope, midslope and toeslope or gully landscape strata in each catchment. Ten soil cores (0-5 cm), from just outside the perimeter of each 0.64 m2 plot used for the rainfall simulation experiment described in Chapter 6, were also sampled and bulked in 1999.

Ruffy (37o00’S, 145o27’E)

The Ruffy SGS national experiment site was located near Seymour in NE Victoria. Site and treatment details are reported by Ridley et al. (2003). The soil at Ruffy is a bleached- mottled magnesic Yellow Kurosol (Isbell 1996) (a duplex soil) that developed on granite of the late Devonian era. The topsoil is a sandy loam and a strongly acid, sandy clay subsoil occurs at about 20 cm depth. Average annual rainfall is 800 mm. Soil samples were collected on 31 October 2000 from low, medium and high soil fertility pasture catchments. The catchments were 1.8, 3.1 and 3.9 ha in size respectively and were instrumented to measure runoff. High, medium or low P fertiliser rates of 22, 11 and 5.5 kg P/ha/yr had been applied to the pastures since 1992. The high and medium input catchments were sown to cocksfoot and white or strawberry clover in 1991 and the low input pasture was unimproved. Pastures were grazed by both sheep and cattle. Ten soil cores (2.5 cm diameter) of 0-5 and 5-10 cm depths were collected and bulked from upslope, midslope and toeslope or gully landscape strata in each catchment. Material from the soil B horizon (90-110 cm depth) from both Maindample and Ruffy was obtained in duplicate cores (4 cm diameter) collected from at least 7 positions in each of the catchments in

212 1999. Composite samples of B horizon (material from 90-110 cm) soil were prepared for each catchment by mixing subsamples of equal volume from each core sample.

Darnum (38o10’S 146o3’E)

Soil was collected on 2 November 2000 from a runoff catchment (1.9 ha) at Darnum that had been established on a commercial dairy in 1994 (Nash and Murdoch 1997). Rainfall at the site is 1000 mm/yr on average. The perennial ryegrass pasture had received P fertiliser at rates of between 60 and 110 kg P/ha/yr between 1994 and 1997 and no fertiliser was applied in 1999 or 2000. The soil at Darnum developed on late tertiary alluvial sediments and is a bleached-acidic, dystrophic, Grey Dermosol (Isbell 1996). The topsoil is a fine sandy clay loam that graduates to a light medium clay B horizon at between 50 and 130 cm. Ten soil cores (2.5 cm diameter) of 0-5 and 5-10 cm depths were collected and bulked across each of upslope, midslope and gully landscape strata within the catchment. Material from the soil B horizon from four 7 cm diameter hand-augered holes was collected and bulked for analysis. Soil Olsen P values for the years 1994 to 1999 were an average for the whole catchment (D.Nash, unpublished data). Soil from all sites was air-dried (40oC), stored at 4oC and sieved through a 2 mm sieve prior to analysis for POlsen, oxalate extractable P, Fe and Al, and P sorption characteristics. Water extractable P was also measured for soils collected from around the rainfall simulator plots at Maindample and Vasey.

213 Table 7-2: Selected properties of topsoil (0-10 cm) from five pasture sites in Victoria

A Site pH EC 1:5 Organic Bulk ESP Psoil Texture (0.01M soil:water carbon density 3 CaCl2) dS/m % g/cm % mg/kg VaseyB 5.1 0.12 7.8 1.1 1.3 220 sandy loam Hamilton C 4.8 - 4.6 1.0 ~1.7 ~400 clay loam MaindampleD High 4.9 0.10 1.2 310 sandy loam Medium 4.6 0.09 1.1 295 sandy loam Low 4.2 0.08 6.2 1.0 235 sandy loam RuffyD 1.4 High 4.9 0.13 260 sandy loam Medium 4.5 0.10 340 sandy loam Low 4.4 0.07 6.8 255 sandy loam DarnumE 5.1 0.13 8.9 1.0 <1 530 fine sandy clay loam A Exchangeable sodium percentage B Cox et al. (1998) C McCaskill et al. (2000) D A.Islam (pers. comm.), M.McCaskill (unpublished data) McCaskill et al. (2003), Andrew and Lodge (2003) E Nash et al. (2000) and D.Nash (unpublished data)

Soil P tests

Soil Olsen P (POlsen, mg/kg) was measured as described in Chapter 5, Section 5.2.2

‘Chemical analyses’. Water extractable P (Pwater) was determined by shaking 5 g of air-dried soil (<2 mm) in 50 mL purified (MilliQ®) water end over end for 1 h as per Kuo (1996). Suspensions were centrifuged at 3000 rpm for 5 min and the supernatant decanted and analysed colorimetrically for MRP (after Murphy and Riley (1962)) using a Cary® UV VIS Spectrophotometer at 882nm. Oxalate extracts were prepared by shaking 0.5g finely ground (<0.5 mm) air-dry soil in 50 mL of acidified ammonium oxalate (pH 3) end-over-end for 4h at 25oC in the dark as per Method 13A1 in Rayment and Higginson (1992). Suspensions were centrifuged at 3000 rpm for

5 min and 30 mL of the supernatant extract was analysed using ICP-AES for Fe Al and P (Feox,

Alox and Pox)(%w/w). The 0.2M ammonium oxalate extractant was prepared by adding 16.2 g of ammonium oxalate and 10.8 g oxalic acid to 1L of water.

P sorption curves

P sorption curves were measured for topsoils (0-5 cm) and B horizons of five grazing sites in Victoria. P sorption curves were measured for each soil sample using a modification of

214 method 9J2 in Rayment and Higginson (1992). Air-dry soil (4g, <2 mm) was shaken end-over- end in 40 mL of P solution for 16 h, after addition of 3 drops of chloroform to suppress microbial activity. P solutions were prepared with KH2PO4 in 0.01M CaCl2 at initial concentrations of 5, 10, 25, 50, 75, 100, 200, 300 and 400 mg P/L. Suspensions were centrifuged at 3000 rpm for 5 min and the supernatant was decanted into 10 mL vials. The final concentration c (mg/L) of molybdate-reactive P was determined by the colorimetric method of Murphy and Riley (1962) using a Skalar® flow-injection autoanalyser at the State Chemistry Laboratory. The change in the amount of P sorbed x (mg/kg) was calculated from the difference between the initial and final solution P concentrations. Olsen P was used as the estimate of the amount of previously adsorbed P (Q, mg/kg). The value of Q was added to the quantity of solution P adsorbed during equilibration to calculate the total amount of P adsorbed by the soil (x + Q). The lowest equilibrium P solution concentration was omitted from the isotherm because for some soils the concentration was below the detection limit of the autoanalyser. The highest final P concentration (after addition of 400 mg P/L in the initial solution) caused a sharp increase in the amount of P sorbed relative to the preceding rate of sorption for most samples. As precipitation, rather than adsorption, of solution P was the more likely process occurring at the highest P addition, the final data point was omitted from the P sorption curves to enable a more accurate assessment of the P sorption characteristics of the soils. A Langmuir model of the P sorption curve was derived by plotting c/(x+Q) on the y- axis against c on the x-axis. A linear regression was fitted and the P sorption capacity (Pmax, mg/kg) calculated as the inverse of the regression gradient, according to the linear form of the one-surface Langmuir model;

c 1 c = + (7-3) (x + Q) aPmax Pmax where a is a coefficient representing the bonding energy for P (White 1980).

Indices of soil P saturation

The Langmuir index (PsatL) and the oxalate index (Psatox) were calculated for the five

Victorian pasture soils. PsatL was calculated using equation 7-1, where POlsen was used as the

STP and Pmax was calculated from equation 7-3 where Q= POlsen. Psatox was calculated according to equation 7-2, where α was equal to 1.

Equilibrium P concentration

The EPC was calculated for the five Victorian pasture soils for which P sorption curves were measured. Using the P sorption data from the initial concentration range of 5 to 50 mg/L

215 only, a straight line was fitted to the plot of x (y-axis) against the natural logarithm of c (ln c)(x- axis). The EPC was calculated as the antilog of ln c after extrapolating the line back to x equal to zero, in units of mg/L.

7.2.2 Other soil P datasets

SGS national experiment sites

A separate set of soil samples (0-10cm, <2mm) collected in spring 1998 from nine sites across Australia for the SGS NE (Andrew and Lodge 2003) was analysed for water and Olsen extractable P. The sites and soil types were Maindample, Ruffy, and Vasey as described above, as well as Albany (Petroferric Brown Sodosol) and Esperance (Subnatric Sodosol) in Western Australia, and Wagga Wagga (Red Chromosol/ leptic Tenosol) and the north west slopes of New South Wales (Barraba (Red Chromosol), Manilla (Red Chromosol/ Brown Vertosol) and Nundle (Brown Chromosol/ sodic Brown Sodosol)). Further details of the soil chemical properties are given by McCaskill et al. (2003).

Ninety Australian agricultural soils

The Langmuir Pmax (equation 7-3, Q = Colwell P) and PsatL (equation 7-1, STP= POlsen) were derived for 90 agricultural topsoils (0-10 cm) whose P sorption isotherms, EPC, POlsen and Colwell P had been analysed and described previously (Burkitt et al. 2002). In brief, the soils represented those used for pasture, crops and horticulture from Victoria (41), New South Wales (15), Queensland (13), Western Australia (11), South Australia (5) and Tasmania (5). Soil textures ranged from sand to clay. The P sorption isotherms for all soils were derived using a range of initial solution P concentrations from 0.5 to 100 mg P/L. Further details of the original sources of the data and selected soil properties are described in Burkitt et al. (2002). The values of PsatL for this range of soils were compared to threshold values for water quality targets reported for soils overseas.

7.2.3 Runoff P concentrations

Annual mean runoff P concentration data from Vasey, Maindample, Ruffy and Darnum were related to spatially averaged POlsen data for the same year. Runoff P data referred to in this chapter are: flow-weighted mean (FWM) TP and DP concentrations in simulated and natural- rainfall runoff from Vasey, which were measured according to methods described in Chapters 5 and 6; TP and DP concentrations in simulated runoff from the Maindample site, which were determined according to methods described in Chapter 6; mean annual concentrations of TP in hillslope runoff from Maindample and Ruffy measured in 1998, 1999 and 2000 and the FWM concentrations of TP and DP in runoff from the Darnum site for the years 1994-2001.

216 7.2.4 Statistical analysis

Linear regression of relationships between runoff P and soil P values was performed using Genstat® 5v.4.1 (Lawes Agricultural Trust 1997). Original data were transformed using the natural logarithm to normalise the distribution of residuals where necessary. Regressions were not forced through the origin because it was considered possible that P in runoff could originate from sources other than the pool of soil P estimated by STP or Psat index.

7.3 Results

7.3.1 Soil P characteristics

POlsen, Pwater and EPC

The range of POlsen (bicarbonate extraction, (Olsen et al. 1954)) values measured for five topsoils (0-5cm) from Victorian pastures was 7-82 mg P/kg and the average EPC ranged from 0.01 to 0.92 mg P/L, increasing with POlsen (Table 7-3). A survey of 90 soils from

Australian agricultural regions revealed a moderate range of POlsen levels and a wide range of EPC (Table 7-3).

217 Table 7-3: POlsen, Pox and EPC of a) five pasture soils (0-5 cm) in Victoria. N=3 for all treatments and b) 90 soils (0-10 cm) from agricultural regions in Australia

Site P Treatment POlsen Pox EPC mg/kg mg/L a) Treatment means for Victorian pasture soils Vasey Low 8 129 0.05 Hamilton Low 4 105 0.01 Ruffy High 22 216 0.18 Med 13 177 0.05 Low 8 163 0.02 Site average 15 185 0.08 Maindample High 27 231 0.10 Med 11 140 0.01 Low 7 139 0.02 Site average 15 170 0.05 Darnum 82 584 0.92 b) 90 Australian agricultural soils Range 0-164 0.013-65A Mean 17 2.9 Median 80.13 A(Burkitt et al. 2002)

The amount of P extracted from a range of pasture soils in Australia by the POlsen test was compared to that extracted by water to identify how the STP levels were correlated. At the Vasey and Maindample sites, there was a significant (P<0.001) linear relationship between

POlsen and Pwater (Figure 7-1). Olsen P values were about four times greater than Pwater values. Figure 7-2 shows the relationships between the two STPs were similar for Vasey, Ruffy and Maindample soils but relationships varied for groups of soils from some of the other SGS national experiment sites.

218 18 16 14 12 10 8 6 4 2 Water-extractable P (mg/kg) Water-extractable 0 0 20406080

Olsen P (0-5cm) (mg/kg)

2 Figure 7-1: Relationships between Pwater and POlsen for Maindample (open circles and broken line, R = 0.90) and Vasey soils (0-5 cm depth) (solid circles and line, R2 = 0.71)

14

12 Barraba Manilla 10 Nundle 8 Wagga Maindample 6 Ruffy Vasey 4 Esperence 2 Albany Water Extractable P (mg/kg) P (mg/kg) Extractable Water

0 0 1020304050 Olsen P (0-10cm) (mg/kg)

Figure 7-2: Relationships between Pwater and POlsen for nine SGS national experiment site soils (0-10 cm depth)

219 P sorption indices

Linear regressions of the 8-point Langmuir isotherm were significant (P<0.01, R2>0.85) for all soils from the five pasture sites, although P sorption in the middle range of solution P concentrations was generally overestimated by the model (Figure 7-3). Using Pox, instead of POlsen, as Q had the effect of increasing the total sorbed P represented by the Langmuir isotherm by approximately the value of oxalate extractable P (Figure 7-3).

1400

1200

1000

800

600

400

200 Total P sorbed + Q) (mg/kg) (x

0 0 50 100 150 200 Final solution P concentration (mg/L)

Figure 7-3: P sorption isotherms (symbols) for Vasey soil (0-5 cm) and fitted Langmuir isotherms

(lines) using POlsen (+, solid line) or Pox (o, dashed line) as values for Q

P sorption capacity, Pmax

The average Pmax (Q = POlsen) for soils at the five sites ranged between 862 and 1374 mg/kg and decreased in the order Hamilton, Vasey, Maindample, Ruffy and Darnum (Table

7-4). The Pmax for the 5-10 cm depth of soil was similar to (for Hamilton and Ruffy), or less than that of the top 5 cm (Figure 7-4). At all the sites the B horizon had a larger Pmax than the top

5cm (Figure 7-4). The B horizon at Vasey and Hamilton had the largest Pmax of all five soil types and depths (Figure 7-4).

Pmax derived from isotherms where the initial solutions contained between 5 and 75 mg/L of P (5 data points) were 42-57% of the estimates of Pmax derived from isotherms using

220 initial solutions concentrations of between 5 and 300 mg/L (8 data points) (Table 7-4). The difference between Pmax and Pmax 5-75 was attributed to limited ability of the simple Langmuir equation to describe P sorption isotherms over a large range of concentrations and confirms the need to use standard methods to measure P sorption isotherms for calculating the Langmuir

Pmax. Pmax in this chapter refers to values calculated from 8 point isotherms, where Q = POlsen, unless otherwise stated.

Table 7-4: Pmax and Psat of a) five pasture soils (0-5 cm) in Victoria. N=3 for all treatments and b) 90 soils (0-10 cm) from agricultural regions in Australia

A Site P Treatment Pmax Pmax5-75 PsatL Psatox mg/kg % % a) Means for pasture soils in Victoria Vasey Low 1169 557 0.7 2.7 Hamilton Low 1374 618 0.3 1.4 Ruffy High 844 420 2.9 8.7 Med 985 451 1.4 5.8 Low 1115 537 0.8 4.2 Site average 981 470 1.7 6.3 Maindample High 844 426 3.3 8.4 Med 930 480 1.2 4.6 Low 1199 584 0.6 3.2 Site average 991 497 1.7 5.4 Darnum 862 487 9.5 17.7 b) 90 Australian agricultural soils Range 0-3.8x103 0.1-2.5x103 Mean 542 41.8 Median 437 2.3 A Pmax calculated from P sorption curves derived using five initial P solutions with concentrations from 5- 75 mg P/L, Q = POlsen

221 2500

2000

1500 0-5cm 5-10cm 1000 B horizon Pmax (mg/kg) Pmax

500

0 Darnum Hamilton M'dample Ruffy Vasey

Figure 7-4: Langmuir P sorption maxima (Pmax) of soil at depths of 0-5, 5-10 cm and the B horizon at five pasture sites in Victoria

A single curvilinear relationship explained Pmax as a function of the sum of Alox and Feox for the combined dataset of soils from the five pasture sites (Figure 7-5). This suggested that P sorption was dependent on the amorphous Fe and Al mineral components of the soil. The sum of Alox and Feox was a better predictor of Pmax than either Alox or Feox alone (Table 7-5) and was also better correlated to Pmax (r = 0.92) than Pmax derived when using Pox as the estimate of Q

(Pmax(ox), r = 0.83)(Table 7-5).

222 1500

Vasey 1000 Darnum M'dample

(mg/kg) Ruffy

max Hamilton

P 500 Fitted model

0 0 2000 4000 6000 8000 10000

Alox + Feox (mg/kg)

Figure 7-5: Relationship between the Langmuir Pmax and the sum of Alox and Feox for soils (0-5cm) 2 from five pasture sites with the fitted model, Pmax =513ln(Alox+Feox)-3180 (P<0.001 R =87.5)

Table 7-5: Correlation coefficients (r) between estimates of Pmax and Alox, Feox and Pox for five pasture soils (N=27)

Feox+Alox Alox Feox Pox Pmax(ox)

Alox 0.787

Feox 0.983 0.663

Pmax(ox) 0.829 0.651 0.816 0.105

Pmax 0.917 0.720 0.903 -0.450 0.836

The degree of P saturation (Psat)

The degree of P saturation for topsoils (0-5cm) from the five grazing sites ranged from

0.3 to 9.5% using the Langmuir Psat index (PsatL) and from 1.4 to 17.7% using the oxalate Psat index (Psatox)(Table 7-4). Both indices of Psat showed a high degree of correlation with POlsen (Figure 7-6). The two Psat indices were highly correlated with each other (R2=0.97, P<0.001, equation 7-4), reflecting the strong correlation between the two measures of P sorption capacity (Figure 7-5).

0.686 Psatox = 4.08Psat L (7-4)

223 The method for measuring Psatox is more standardised than for measuring PsatL, so the ability to predict Psatox from PsatL over the measured range is useful for comparing levels of Psatox to values reported in other studies.

25

20

15

10

5 Degree of soil P saturation (%) P soil saturation Degree of 0 0 20406080100 Olsen P (mg/kg)

2 Figure 7-6: Relationship between Psatox(open circles, R =0.92, P<0.001) or PsatL (closed cirles, 2 R =0.99, P<0.001) and POlsen (0-5cm, mg/kg) across five soil types (n=27)

The range of Pmax and PsatL of 90 soils from Australian agricultural regions was large

(Table 7-4). Extreme maximum EPC and PsatL, and minimum Pmax, values were attributed to sandy soils with very low P buffering capacities, rather than to heavily fertilised soils. The median PsatL, however, was only 2.3% (Table 7-4). Eighty-two percent of the soils were less than 10% saturated with P (Figure 7-7), which according to equation 7-4 was equivalent to the upper threshold Psatox level of 20% reported for drainage waters in the Netherlands

(Breeuwsma and Silva 1992). A Queensland horticultural soil, which had the highest POlsen (163 mg/kg) of all the soils, had a PsatL of 47%.

224 45 100% 40 90% 35 80% 70% 30 60% 25 50% 20 40% 15 30% 10 20%

Frequency (number of soils) 5 10% Cumulative percentage of soils 0 0% 2 5 10 15 30 50 80 100 More

PsatL (%)

Figure 7-7: Frequency distribution (bars) and cumulative percentage (line) of the PsatL (%) of 90 Australian soils. ‘More’ includes soils with very low P sorption capacities.

7.3.2 Relationship between P concentration in runoff and soil P

Runoff P concentrations from the Vasey and Maindample sites were compared with the mean POlsen (0-5 cm) using linear regression analysis. Runoff P was represented by the flow weighted mean concentrations (FWM) of TP and DP in runoff from the Vasey hillslope runoff plots (see Chapter 5 and 6), annual arithmetic mean TP concentrations in runoff from the pasture catchments at Maindample, and mean concentrations in simulated runoff from small plots at both Vasey and Maindample, and related to the POlsen from rainfall simulator plots or pasture catchments (see also Chapters 5 and 6 for methods). The regression statistics are shown in Table 7-6. Where POlsen was analysed on the top 10 cm of soil only, the soil test P values were multiplied by a factor of 1.6 to account for the degree of enrichment of the surface 5 cm compared to the 5-10 cm depth (see Chapter 5).

All linear regressions of TP against POlsen were significant (P<0.05) and the slope of the regression varied only slightly (range 0.02 – 0.05) between sites and experimental scales, however most of the relationships were poor (R2<0.50). The variance accounted for by the STP was greater for TP than DP, where measured (Table 7-6). The strongest relationship (R2= 0.64 and 0.68 for Vasey and Maindample soils, respectively) occurred for the annual mean P concentrations in natural rainfall runoff. Because these data points were annual means, they were a closer reflection of the population mean TP concentration than the individual simulated

225 rainfall runoff events. The regression models were not improved when TP in runoff was related to the Pwate, rather than the POlsen content of the topsoil (Table 7-6).

Table 7-6: Regression models and variance accounted for (R2) for the relationships between TP or

DP in runoff (mg/L) and POlsen (mg/kg), Pwater (mg/kg) or PsatL (%) at Vasey (V) and Maindample (M) and Ruffy (R)

Soil P testB Dependent Regression Runoff type Intercept N R2 (0-5cm) variable slope Vasey * * Hillslope POlsen TP 0.03 -0.04 8 0.64 A * ** * POlsen ln DP 0.07 -2.56 80.45 *** * *** Simulated POlsen TP 0.02 0.24 43 0.42 *** *** *** POlsen ln DP 0.03 -2.21 43 0.28 C *** *** *** Pwater TP 0.05 0.35 43 0.30 Maindample ** ** Hillslope POlsen TP 0.05 -0.35 9 0.68 *** * *** Simulated POlsen TP 0.03 0.31 43 0.43 *** *** POlsen DP 0.01 0.08 43 0.43 ** *** ** Pwater TP 0.09 0.59 43 0.20 Maindample, Ruffy and Vasey (MRV) *** *** *** Hillslope POlsen ln TP 0.073 -2.408 25 0.57 *** *** *** PsatL ln TP 0.525 -2.027 25 0.46 *, ** and *** represent significance at P=0.05, 0.01 and 0.001 respectively, as determined by linear regression analysis (Lawes Agricultural Trust 1997) A Natural logarithms of the concentration of DP or TP in runoff were used to normalise the distribution of residuals in the regression analyses B Units are TP (mg/L), POlsen and Pwater (mg P/kg), and PsatL (%) C One outlier removed

Mean values for annual runoff TP concentrations in hillslope runoff and POlsen from a single sample time in each plot year (adjusted to represent 0-5cm depth where necessary) from the four pasture sites from which runoff P concentrations were measured are presented in Table

7-7. Mean TP concentrations in runoff and POlsen at Darnum were significantly (P<0.05) higher than at Vasey, Maindample and Ruffy (Table 7-7) and the two variables were not correlated with each other over the six years of measurement at Darnum (r2=0.06). This suggested the Darnum data represented a different population to the sheep grazed sites so data from this site were excluded from a multi-soil relationship. An increase in mean annual runoff TP concentration with increasing POlsen for Maindample, Ruffy and Vasey soils was explained reasonably well by a single exponential curve (Figure 7-8, Table 7-6). A distinct change point was not evident for these data. PsatL was a poorer predictor of mean annual runoff TP concentrations than POlsen for the three sites combined (Table 7-6).

226 Table 7-7: Mean POlsen, PsatL, runoff TP and runoff volume for each plot year of runoff from hillslopes at Darnum, Maindample, Ruffy and Vasey

Site POlsen (0-5cm) PsatL(0-5cm) Runoff TP Flow volume No. treatment mg/kg % mg/L mm years C AB B Darnum 72a 8.4a 13a 34b 7 (46-94) (5.4-11) (3.3–29.6) (3-84)

Maindample 19b 2.0b 0.50b 110a 9 (7-29) (0.6-3.4) (0.10-1.19) (53-205)

Ruffy 20b 2.1b 0.37b 14b 8 (11-35) (1.0-4.1) (0.16-1.02) (4-22)

Vasey 19b 1.6b 0.49b 23b 8 (10-32) (0.8-2.7) (0.20-0.83) (0.1-67.5) A Subscripts denote significant differences (P<0.05) between site means B Mean of 6 plot years C Data from Nash et al. (2000) and D.Nash (unpublished data)

1.40

1.20

1.00 Maindample 0.80 Ruffy 0.60 Vasey MRV Model 0.40

TP in hillslope runoff (mg/L) runoff hillslope in TP 0.20

0.00 0 10203040 Olsen P (0-5cm) (mg/kg)

Figure 7-8: Relationship between hillslope runoff TP concentrations and POlsen (0-5 cm depth) for Maindample, Ruffy and Vasey soils, with a line representing a single regression for all three soils 2 (TP=exp (0.073POlsen-2.4), R =0.57, P<0.001)

227 Equilibrium P concentrations

A comparison of annual mean DP concentrations in runoff and soil EPC was made to identify whether EPC could be used as a surrogate measure for DP. Figure 7-9 shows the relationship between POlsen, and DP and EPC for the four pasture sites where runoff was measured. Annual mean DP concentrations for Maindample and Ruffy catchments were approximated by subtracting 0.05 mg P/L, which was the maximum PP concentration measured from a small number of events at Maindample (see Chapter 6). Both the relationships represent the quantity/intensity relationship expected for soil P sorption characteristics, and EPC values were generally lower than measured or estimated DP concentrations in runoff in the year EPC values were measured. By assuming there were no PP contributions to the P in runoff, an extrapolation of the logarithmic relationship between POlsen and DP concentrations in Figure 7-9 suggests a negative

POlsen would be required to achieve a guideline DP concentration of 0.04 mg/L in lowland rivers.

In contrast, the relationship between POlsen and EPC suggests the critical POlsen (0-5cm) for the guideline DP concentration would be about 16 mg/kg. This is equivalent to a 0-10 cm POlsen of 10 mg/kg.

3.5 3.0

2.5 2.0 Runoff DP EPC 1.5 EPC 1.0 Runoff DP 0.5

Soil EPC or DP in runoff (mg/L) 0.0 0 20 40 60 80 100

P Olsen (0-5cm) (mg/kg)

Figure 7-9: Relationships between the annual mean DP concentrations in runoff (∆) and the soil

EPC (●) and POlsen (0-5cm)(mg/kg) for four pasture soils in Victoria in 2000. DP was estimated for Ruffy and Maindample runoff as TP – 0.05mg/L (see text above). Fitted logarithmic equations are 2 2 indicated for EPC (POlsen = 15.4 lnEPC + 65.8 R = 0.77) and DP (POlsen = 17.3 lnDP + 47.8 R = 0.73).

228 Effect of runoff volume

The effect of the annual runoff volume on runoff TP concentrations was also assessed for the four grazing sites where the runoff volume was measured (Vasey, Maindample, Ruffy and Darnum). There appeared to be an inverse relationship when the data were combined with perhaps 3 groups (Figure 7-10). One group, which included Ruffy and plots 2, 3 and 4 at Vasey had equivalent TP concentrations in runoff at a lower volume of runoff than a second group, which included Maindample and plot 1 at Vasey. The Darnum data were different again reflecting the significantly higher mean TP concentrations in runoff (Table 7-7).

100

Vasey 10 Darnum Maindample Ruffy 1 TP in hillslope runoff (mg/L) runoff hillslope in TP

0 0 50 100 150 200 250 Annual runoff volume (mm)

Figure 7-10: The relationship between the annual runoff volume and the mean annual TP concentration in runoff from four pasture sites

229 7.4 Discussion

7.4.1 Soil P characteristics

Soil P tests: POlsen and Pwater

Olsen P is routinely measured in Victoria to quantify the pool of soil P available for plant uptake over the growing season and therefore indicate the need for P application. Phosphorus extracted from the soil in the bicarbonate solution includes soluble and readily desorbed orthophosphate (i.e. intensity and quantity fractions) as well as variable amounts of organic or condensed P compounds (Coventry et al. 2001). Olsen P levels are therefore affected by the P buffering capacity of the soil (Moody and Bolland 1999). The POlsen recommended for maintaining productive and profitable pasture systems in perennial grass/clover pastures in Victoria is between 7 and 15 mg P/kg for sheep pastures (Cayley et al. 2002) and for dairy pastures is approximately 20 mg/kg (Gourley et al. 2001). The median POlsen of the 90 agricultural soils from across Australia was only 8 mg/kg (Table 7-3), which suggests that over half of these soils would be considered deficient in P for maximum growth of pasture plants. This reflects the generally low P content of the old and highly weathered Australian soils and only a modest increase in the total P fertiliser applied annually over the last 30 years (Smith 1983; Fertilizer Industry Federation of Australia Inc 2003).

Water extractable P was closely related to POlsen at Maindample and Vasey but the relationship was not consistent across a range of high rainfall pasture soils in Australia (Figure 7-2). The variability in the relationships between the two STPs probably reflects variation in characteristics related to P desorption such as pH, organic matter and exchangeable cation composition (Curtin et al. 1992; Simard et al. 1994) across these soils. Water-extractable P has been found to closely estimate actual DRP concentrations in runoff (Yli-Halla et al. 1995; Pote et al. 1996). However, at neither Maindample nor Vasey was Pwater a better predictor of runoff

TP concentrations than the routinely measured POlsen, suggesting there would be no benefit in performing the non-routine water extraction for these soils. Considering the variable relationships between Pwater and POlsen across soil types, a comparison of water and Olsen extracts as predictors of runoff P concentrations would need to be made for individual soils to identify a preferred test.

P sorption indices

P sorption curves for soils from the five grazed pastures in Victoria were measured to calculate Pmax and PsatL. Values of Pmax were high across the soils tested, ranging from 862 mg/kg to 1374 mg/kg. Differences in the ability to sorb P were consistent with the different

230 mineralogy of the soils’ parent material. The Hamilton soil was formed on basalt, the Vasey on acidic rhyolite, the Darnum soil on alluvium, Ruffy on granite and Maindample on sedimentary material. Toreu et al. (1988) found the P sorption capacity of Queensland soils was greatest for those derived from basalt, followed by alluvial and metamorphic material and then granitic soils. Consistent with findings of Barrow (1998) that previous fertiliser additions decreased the

P buffering capacity of some soils, the Pmax of the low fertility soils at both the Maindample and Ruffy sites was higher than for the medium and high fertility soils.

The high degree of correlation between the Psat and the simple POlsen test values (Figure 7-6) for the Victorian pasture soils was indicative of the reasonably similar P sorption capacities of the five soils. Olsen P may not have been such a reasonable substitute for the degree of P sorption saturation of a sandy soil of low P sorption capacity. The high correlation between the two estimates of Psat was not surprising because there were linear relationships between the two measures of the P sorption capacity (Pmax and the sum of Alox and Feox) for each soil type. The linear correlation would be useful if it could be extended beyond these soil types as the range of PsatL indices currently available could then be described in terms of the Psatox index, which is more easily compared to current literature (see Beauchemin and Simard (1999)).

The range of PsatL (0.3 and 9.5%) and Psatox (1.4 to 17.7%) of soils from the five

Victorian grazing sites was similar to the Psatox of grazed fields with a long history of manure application in Indiana, USA (3-20%) and to intensively cropped soils in Quebec where mineral fertilisers were used (5-35%)(Beauchemin and Simard 1999). Much higher degrees of P saturation (up to 150%) have been measured in soils that have received poultry manure amendments or been used for intensive animal production in Europe and the USA (Beauchemin and Simard 1999). The PsatL estimated for 90 agricultural soils across Australia ranged from 0 to 50% (equivalent to approximately 0-60% Psatox), which showed that some soils had reached levels of Psat as high as those reported overseas. Many of the soils with high PsatL were used for horticultural production.

Limitations and merits of methods used to estimate P sorption indices

The method used to measure the P sorption curves influenced the estimates of Pmax and hence the ability for PsatL for soils in this study to be compared to levels measured in other soils. For example, sorption isotherms were measured using solutions with a wide range of initial P concentrations because the purpose was to evaluate the maximum P sorption capacity of soils from an environmental viewpoint. It was considered that high solution P concentrations would provide a more realistic and accurate estimate of this soil property than the smaller ranges (eg. 0-25 mg P/L) typically used to identify the ability of soils to buffer the plant- available soil solution P.

231 The final solution P concentration, after addition of the highest initial P concentration (400 mg/L), ranged from 171 to 317 mg P/L between the varying soil samples. These were extremely high concentrations in environmental terms and would only ever be experienced immediately surrounding a fertiliser granule. It is likely that at such high concentrations, P could have formed precipitation products with soil sesquioxides and complex compounds with other cations such as Ca, K and H (Taylor et al. 1964), which could explain the sharp increase in P sorption that was observed for most soils at this highest P concentration. It is unlikely, however, that precipitation of solution P surrounding fertiliser granules would lead to a net reduction in the soil solution P concentration because a simultaneous drop in pH may lead to re- dissolution of some of the precipitation products (Huffman and Taylor 1963; White and Taylor 1977b).

Another consequence of using a wide range of initial P concentrations to derive Pmax was that the estimates of Pmax could not be easily compared with those derived from a smaller range of initial P concentrations (eg Singh and Gilkes (1991)). Most P sorption curves are generated from a range of initial solution P concentrations not exceeding 100 mg/L (Holford et al. 1974; Barrow 1978; Pierzynski 2000), in order to gain precision in the estimation of sorption mechanisms occurring near the normal soil solution P concentration (up to approximately 3 mg/L (Frossard et al. 2000)). For the soils used in this study, the total Pmax calculated over the range 5-300 mg P/L was approximately 50% greater than that calculated over a lower concentration range (0 to 75 mg P/L initial concentration). The precision of the latter estimate of

Pmax was hampered by there being only four equilibration points in this range, but estimates for the Maindample and Ruffy low fertility catchments were similar to those obtained from a 7 point isotherm over the same concentration range (A. Islam pers. comm.). Whilst it is widely acknowledged that the magnitude of the Langmuir Pmax is dependent on the concentration range from which the isotherm is derived (Barrow 1978), the lower Pmax from a smaller concentration range presented a false indication of the maximum P sorption potential of the soil.

Either POlsen or oxalate extractable P would have been satisfactory estimates of Q for use in the Langmuir isotherms (Figure 7-3). However, the more conservative estimate, POlsen, was considered to most closely represent the sorbed P that would influence desorption processes during runoff events, which in the temperate environment were usually of the order of 2 – 8 hours (Figure 4-9). It is likely the oxalate extractable fraction of soil P, which represented the P adsorbed to amorphous Fe and Al minerals, contained P which was specifically adsorbed and occluded as well as labile or reversibly adsorbed P (Lookman et al. 1995). Van Der Zee and

Van Riemsdijk (1988) found Pox was a good estimate of total soil P and it therefore may be an overestimate of the P involved in sorption reactions. Perhaps a better estimate would be somewhere between the extracting strengths of the Olsen and oxalate extractions as Burkitt et

232 al. (2002) found that the even Colwell P, which extracts more P than Olsen due to the longer extraction time, underestimated Q for some soils with low buffering capacities.

Further error involved in the determination of the Pmax occurs considering the commonly poor fit of the simple Langmuir model when the range of initial P concentrations is large (Holford et al. 1974). The poor fit is understood to occur because the single surface Langmuir model assumes that the bonding energy in soil for P in solution remains constant as the amount of adsorbed P increases, i.e. a linear relationship between the final P concentration, c, and the ratio of c to the amount of adsorbed P whereas the resulting isotherm is frequently curvilinear (Holford et al. 1974; Singh and Gilkes 1991). Curvilinear deviation of the residuals of the actual measures from the fitted isotherm was found for all soils in this study and most likely represented a continuous spectrum of bonding affinities for P, where the adsorption energy decreased as the amount of sorbed P increased (White and Taylor 1977a). This decrease in energy is described as either linear by the Tempkin model or exponential by the Freundlich equation (Barrow 1978). The simple Langmuir equation from which Pmax in this study was derived has been modified in some cases to more closely describe the curvilinear nature of the isotherm (Holford et al. 1974; Barrow 1978), however, in this study the simple Langmuir equation provided a significant model (P<0.05, R2>0.89) for all soils so more complex models were not required.

The variation in the methods used to measure P sorption curves and estimate Pmax makes comparisons of Pmax between soils difficult. This detracts from the suitability of PsatL as an index for indicating the potential release of P to runoff and drainage. Attempts to standardise the P sorption isotherm methods have been made in USA (eg Pierzynski (2000)) and to some extent in Australia, although variations between soil testing laboratories do occur.

Pmax can also be estimated using surrogate variables including clay (Ballard and Fiskell 1974; Scheinost and Schwertmann 1995), kaolinite (Kelly 1999) and organic C content (Lopez- Hernandez and Burnham 1974; Simard et al. 1994), pH and amorphous and crystalline Fe and Al fractions in acidic soils (Saunders 1965; Singh and Gilkes 1991). In the current study, the

Langmuir Pmax had a high correlation with the sum of Alox and Feox (r = 0.92), which represents the amorphous fractions of these minerals, for the five soils combined and was less strongly correlated with either Alox (r = 0.72) or Feox (r = 0.90) alone. A similarly strong correlation was found for forest Podzol soils with a range in clay contents of 4 to 14% by Yuan and Lavkulich (1994) and for acid soils of low organic matter from the USA and the Netherlands (Paulter and

Sims 2000), but a lack of significant association was found between Pmax and Alox + Feox for three differing soil groups (clay less than 10%) in Quebec (Beauchemin and Simard 1999). Considering these good correlations and standard extraction conditions, the oxalate extraction offers advantages over measuring Pmax in order to estimate the P sorption capacity of these soils. The oxalate method is also markedly less time consuming, although using Inductively Coupled

233 Plasma Atomic Emission Spectroscopy for analysing total extracted Al and Fe is not necessarily cheaper than digestion followed by colorimetric analysis.

7.4.2 Relationship between the P concentration in runoff and soil P characteristics

STP indices for individual soils

Soil test P can be a useful indicator of the risk of P transfer to runoff water where there is a strong relationship between runoff P concentrations and STP. A positive relationship between runoff TP concentrations and POlsen for the hillslope and simulator runoff plots at Vasey and Maindample occurred (Table 7-6), however, the correlations were not strong (R2<0.50), which reduced the ability of the STP levels alone to predict runoff P concentrations with a high degree of precision. Positive correlations between DP concentrations in runoff and STP under the relatively uniform conditions of the simulated rainfall at both Maindample and Vasey (R2= 0.43 and R2=0.28 respectively) were also weak when compared to other small-plot, simulated runoff studies (r2 = 0.72-0.95, Table 7-1) The correlations for Vasey and Maindample runoff simulations were likely to have been weakened by the low and narrow range of POlsen levels, when compared to the large range of STPs measured in soils of studies in international literature, most of which had received swine, cattle or poultry manure amendments (Sharpley 1995; Pote et al. 1996; Pote et al. 1999a; McDowell and Sharpley 2001; Weld et al. 2001). Similarly for drainage P concentrations at Rothamsted UK, a strong relationship with STP occurred above POlsen levels of 60 mg/kg, whereas concentrations at POlsen less than 60 mg/kg were generally below 0.2 mg/L and not related to the POlsen status (Brookes et al. 1996). All

POlsen levels (0-5 cm) of the Vasey rainfall simulator plots were below 60 mg/kg. For Maindample and Vasey soils water extractable P was a poorer predictor of runoff P concentrations than the POlsen test. This contrasted to other studies where Pwater provided the strongest correlation with DP concentrations in runoff, when compared to other STPs such as Olsen, Mehlich 3 and Morgan’s (Pote et al. 1996) and was able to simulate actual runoff DRP concentrations (Yli-Halla et al. 1995). Water-extractable P may be more strongly influenced by

P sorption properties of soils than the alkaline POlsen extracts. The poor relationship with runoff

P concentrations and variable relationship to the POlsen value across soil types also suggested water was a less suitable extractant than the widely used bicarbonate, for addressing both pasture requirements and environmental concerns. The majority of studies of the relationships between runoff and soil P concentrations have reported the behaviour of DP or DRP only. Presumably this is because of the equilibrium

234 that develops between solution and soil P concentrations. However, the operationally defined fractions of runoff P may not accurately represent the desorbable pool of soil P because P can occur in a continuum of particulate and colloidal sizes in soil water and runoff (Haygarth 1999) and the DP fraction arbitrarily only represents P that is less than 0.45µm in size. In the current study, the TP concentration revealed the strongest correlations with extractable soil P at Vasey, which suggested the quantity of P adsorbed to soil particles and organic matter entrained in runoff may also have reflected the soil P status. However, there was no clear relationship between PP concentrations in natural runoff and STP (see Figure 6-4d, Chapter 6). The TP in runoff may therefore represent the sum of all the adsorption and desorption processes occurring before and during mobilisation of P from soils, which may be related to the initial P loading of the topsoil. Even where there was a positive relationship between PP and POlsen, as for simulated runoff from Vasey (see Figure 6-4d, Chapter 6), the variation in the concentration and P content of sediment in runoff would have provided noise around the final relationship between TP and STP. As discussed in Chapter 6, variation in TP concentrations in runoff were also influenced by total volume of flow, ionic strength and source of surface flow so it was not surprising that STP alone was a relatively poor predictor of runoff P concentration.

Soil P indices for multiple soil types

POlsen and PsatL Relationships between runoff P and soil P are rarely consistent across soil types due to variable P sorption characteristics and land management (Sibbesen and Sharpley 1997). The variable influence of land use on runoff P concentrations was exemplified by the higher TP concentrations in runoff and POlsen for Darnum soils compared to Maindample, Ruffy and Vasey. Soils at Darnum were used for intensively grazed dairy pasture, where fertiliser was applied at rates much higher than the sheep grazed sites. Fertiliser was also often applied during the runoff season rather than a few months prior to runoff, as occurred at Maindample, Ruffy and Vasey. The fertiliser management practices at Darnum therefore left shorter periods of time for the fertiliser to equilibrate with the surface soil and greater opportunity for direct entrainment of fertiliser granules in the runoff (Nash et al. 2000). At the sheep grazed sites these direct fertiliser effects were not as important. Despite the variation among soil types from the three sheep-grazed sites, a single log- linear model described the relationship between runoff TP concentration and POlsen (Figure 7-8) or PsatL (Table 7-6). The curvilinear nature of the relationship was expected based on known principles of quantity/intensity P sorption relationships, and was consistent with findings of others (Turtola and Yli-Halla 1999; Paulter and Sims 2000). The relatively uniform runoff P – STP relationships between sites may partially be attributed to similarities between some of the

235 soil properties that influence POlsen. For example, all three topsoils were acidic and had high organic carbon contents, typical of established grass-clover pastures (Russell 1986; Haynes and Williams 1993), and the textures ranged only from sandy loam to clay loam. Other similarities included the small range in total soil P (200 - 400 mg/kg where measured) and all the sites were grazed and had almost complete groundcover when runoff occurred.

PsatL was not a better predictor of runoff P concentrations than POlsen (Table 7-6).

Similar to the uniform runoff P – POlsen relationship, this may be partially explained by the fact that all four topsoils were acidic and had medium texture, which are key properties affecting P sorption characteristics. Others have also found that soils with similar pH and texture show strong and uniform relationships between Psat and P solubility (Beauchemin and Simard 1999).

Similarly Kleinman et al. (1999) found Morgan’s STP was an adequate substitute for Psatox for 59 USA soils with similar parent material. The low coefficient of determination (R2= 0.46,

Table 7-6) for the relationship between runoff P concentrations and PsatL across the three sheep-grazed sites may again be partially attributed to the narrow range of PsatL occurring in the soils (0.3 to 9.5%). A stronger relationship (r2 = 0.85) was found by Sharpley (1995) for soils (0-1cm) that had been amended with poultry manure and were the equivalent of 0 to 27% saturated with P. In all cases, the utility of runoff P – soil P relationships depends upon the degree of precision required. For the purpose of maintaining runoff P concentrations below levels deemed unacceptable in surface waters, the level of precision required would be high, considering concentrations as low as 0.02 mg P/L can lead to eutrophication (Correll 1998). Considering the extra time and expense required to measure P sorption properties and the variability in methodologies already discussed, measuring Psat would be unjustified for predicting P concentrations in runoff for soils with similar P sorption properties because it offers little advantage over the routine POlsen test. Beauchemin and Simard (1999) suggested that the power of the Psat method reduces with an increase in variation between soils and that rather than replacing the STP as a measure of the soil’s propensity to release P to solution, a Psat index should be used as a complementary tool which takes some account of the differences in the P sorption capacity between soils. Alternatively, Simard et al. (1994) found that estimates of soil pH, organic matter and Ca content were needed in addition to STP in order to develop a single relationship with desorbable P across five soil series. However, if Psat or other soil tests are better indicators of the risk for P loss than STP alone then performing the extra tests would be preferable to the expensive and time consuming task of measuring concentrations of P in hillslope runoff for a wide range of individual soil types.

The effect of runoff volume on P concentration

An important factor influencing concentrations of P in runoff from the Victorian sites was the annual volume of runoff. The relationships in Figure 7-10 suggest that greater rainfall

236 and runoff may result in an overriding dilution of the P mobilised into runoff. Evidence for this process occurring across different areas of a hillslope was presented in Chapter 6 (Figure 6-13) where it was shown that, although mean POlsen levels for plots 1 and 3 at the Vasey runoff site were similar in 2000, there was a two-fold increase in the concentration of P in the runoff from plot 3 and a 40-fold difference in runoff volume. This suggests that flow not only has a profound influence on annual loads of nutrients in runoff but also on the background level of dilution of the available P supply. The total flow and the flow-weighted runoff P concentrations are not independent variables because the concentration is derived from the total P load divided by the total flow. However, the inverse relationship that occurred between the annual surface flow and the mean annual runoff P concentration for each of the four grazing sites (Figure 7-10) adds weight to the suggestion that P concentration in runoff is affected by both the availability of P at the pasture surface (i.e. the strength of the P source) and the volume of water passing through the source on an annual basis. The effect of flow volume is likely to vary between environments. For example, Pote et al. (1999b) found lower DRP concentrations in simulated rainfall-runoff from well-drained plots, and higher DRP concentrations from plots yielding higher volumes of runoff. They attributed the difference to rapid infiltration of DRP into the well-drained soils which reduced the pool of DRP available for transfer into surface flows. In their study, however, there was an overall trend for higher DRP concentrations from runoff plots with higher runoff, which was contrary to the current findings. Daly et al. (2002) also found a contrast in the P loss behaviour between two major soil types. They found that high molybdate reactive P concentrations in Irish rivers were associated with catchments that had predominantly poorly drained peat and gley soils (and higher runoff) and that lower river P concentrations were associated with catchments with well drained soils. This effect was attributed to the ‘wet’ peaty soils having a lower capacity to trap and chemcially adsorb P in runoff from upland agricultural and urban areas. In summary, the results from the Victorian pasture sites suggested that for a given soil, hydrological behaviour and land use, greater annual runoff will lead to an overall dilution of runoff P concentrations, which exceeds differences caused by variation in soil P status. Relationships between flow volume and runoff P concentrations are also likely to vary between temporal and spatial scales. Pote et al. (1999b) found that standardising DRP concentrations in simulated plot runoff by expressing concentrations relative to the total event flow volume removed some of the soil type differences found in the runoff P-STP relationships, however, no such improvement in the multisite relationship was found after adjusting the annual field runoff data in the current study (data not shown). Tunney et al. (2000b) also found some evidence for P concentrations in surface runoff from pastures increasing with event flow rate. It is possible that expansion of the VSA as flow increased at the Vasey site increased the amount of contact between the runoff and pasture P sources, which could explain the higher P concentrations in

237 storm flow than in surface baseflow from plot 1 (see Table 6-6). Further investigation into these mechanisms may be warranted.

Equilibrium P Concentration

The EPC values for the five Victorian pasture soils and the median value for the Australian soil dataset (0.13 mg P/L) were high compared with values measured on sugarcane (<0.001-0.06 mg/L, (Edis et al. 2002)) and soybean (0.004-0.017 mg/L, (Moody et al. 1983)) soils in Queensland but comparable to the mean EPC of ten English soils (0.05-1.08 mgP/L, (Hughes et al. 2000)) and fertilised USA soils (0.2-0.3 mg/L, (Taylor and Kunishi 1971)). Of the five pasture sites studied in detail, the EPC of the soils from Darnum and the high fertility catchments at Ruffy and Maindample were highest, reflecting their higher POlsen contents. The plot of P sorbed versus the natural logarithm of the solution P concentration may have overestimated the EPC values, particularly as the accuracy and precision of the P sorption isotherms at low solution P concentrations was reduced because of the small number of data points. Nonetheless, the EPC values for these high P fertility soils exceed the trigger levels for concern for eutrophication problems in freshwater streams (0.04 mg/L) in Victoria (Environment Protection Authority Victoria 2001). Most of the lower fertility soils, however, would be expected to act as sinks for P near these environmentally sensitive solution P concentrations. Despite the potential overestimation of EPC values by the extrapolation method used, DP concentrations in runoff were not adequately predicted by soil EPC values (Figure 7-9). The disparity between the two measures of solution P concentration was not surprising for a number of reasons. Firstly, the EPC of suspended sediment (SS) in the runoff was likely to have been higher than the EPC values measured on surface (0-5cm) soil material, because SS is generally enriched with P compared to surface soils (see ‘P enrichment ratio’ in Chapter 6). Secondly the ionic strength of laboratory CaCl2 solutions (~0.78 g soluble salt/L) used to measure P sorption isotherms was higher than in runoff water (0.04-0.14 g/L) and was therefore likely to have suppressed P desorption. DP in runoff was also not expected to be in equilibrium with the soil P because DP concentrations were also likely to have been influenced by dissolution and desorption of P from a range of other sources of P at the surface of pastures such as plant and faecal material. The high P concentrations in runoff at Darnum demonstrated that non-soil related factors such as fertiliser applications can also strongly influence runoff P concentrations. These factors may also partially explain why an extrapolation of the multisite runoff P-STP model to a POlsen of zero suggests concentrations of TP in runoff remain greater than zero. This remaining P probably includes the plant and faecal fractions of DP, as well as PP contributions and desorption of soil P that is not extracted by the bicarbonate solution. None of these fractions are necessarily directly related to the POlsen.

238 Comparison of the STP v runoff P relationship with other studies

The DP- POlsen relationships described in Table 7-6 for Vasey and Maindample soils were similar to or steeper than those found for the range of land uses and soils in USA, Finland, the UK and Ireland presented in Table 7-1 (Figure 7-11). The runoff TP-STP model for the three sheep-grazed sites (MRV POlsen model, Table 7-6, Figure 7-11) was steeper again reflecting the contribution of the PP component in runoff and the non-soil related factors influencing DP concentrations. The range of relationships between runoff P and STP between soils was not surprising considering their empirical nature and the variation in scale, soil type, timing and magnitude of P amendment, groundcover, environment and runoff and soil sampling techniques used in their derivation. This reinforces the need for relationships between runoff and soil P concentrations to be derived on a site basis rather than applying empirically derived thresholds from elsewhere.

Vasey Hillslope Vasey Sim Maindample Sim Pote et al 1999b US Sharpley 1995 US Weld et al 2001 US Pote et al 1996 US Turtola et al 1999 Finland McDowell & Sharpley 2001 UK Tunney et al 2001 TP MRV model

1.2

) 1

0.8

0.6

0.4

DP or TP in runoff (mg/L 0.2

0 0 20406080100

Adjusted P Olsen 0-5cm (mg/kg)

Figure 7-11: Relationships between DP (or TP for the combined Victorian sites) concentrations in runoff and adjusted POlsen (0-5cm) for soils described in Table 7-1 and Table 7-6. Vertical dashed lines indicate optimum POlsen range for pasture production in Victoria

239 The multi-site relationship was also slightly steeper than that found for winter runoff from irrigated dairy pastures in Shepparton, Victoria (equation 7-5, linear model approximated from data in Nexhip and Austin (1998)).

lnTP = 0.02POlsen(0−5cm) − 0.59 (7-5)

In South Australia, Colwell P differences weren’t significant between two runoff catchments and yet there were significant differences in P concentrations in water flows (Stevens et al. 1999). From this, it was concluded Colwell P was not a good environmental indicator for those soils. The flows in this study were a combination of surface runoff and throughflow. Contributions of throughflow P would not be expected to reflect the surface soil P status unless there was substantial preferential bypass flow that has had limited interaction with the soil matrix. Even so, this again highlights that the soil P status is one of many factors that may influence the P concentrations in runoff. The runoff TP – STP relationship for the three Victorian pasture soils was developed using STP values that were either measured or adjusted for a 0-5 cm depth, rather than the 0-10 cm depth that is routinely sampled for measuring plant available P. The shallower depth was sampled in order to more accurately represent the depth of soil that interacts with runoff water (Sharpley et al. 1978). However in Chapter 5 it was shown that for both low and high fertility pastures at Vasey, the POlsen in the top 5 cm of soil was directly related to that in the 0-10 cm depth which suggests the routine soil test depth of 10 cm should be adequate for measuring relationships with runoff P concentrations. Considering the potential contribution of P desorbed, entrained or leached from other surface materials such as plant, organic matter and faecal material, however, a sampling and measurement procedure that accounts for non-soil sources of P is likely to explain more of the variation in runoff P concentrations than standard soil tests alone.

Threshold levels of soil P

Water quality targets of up to 1 mg DP/L have been used to set critical STP levels in some states in the USA (Sharpley et al. 1996) and these critical STPs are often 2-6 times higher than agronomically optimum STP levels (Gartley and Sims 1994; Sibbesen and Sharpley 1997; Weld et al. 2001). In these systems, restricting fertiliser applications above the optimum level required for plant growth may be a sufficient management tool for maintaining acceptable concentrations of P in runoff. Figure 7-11 shows the optimum POlsen (0-5cm) range for pasture production in Victoria. The slope of the runoff TP-STP-relationship for Maindample, Vasey and Ruffy soils suggests there is less of a buffer between soil P levels required for optimum pasture production and threshold limits for maintaining target runoff P concentrations in southern Australia than for the USA soils reported by Gartley and Sims (1994). For example, using the

240 MRV model in Table 7-6, the POlsen required to meet a runoff P concentration target of 1 mg

TP/L was 21 mg/kg, which is just above the optimum POlsen for productive wool pastures in southwest Victoria of about 15 mg/kg. However, to meet the target stream P concentrations in Victoria of 0.04 mg TP/L (Environment Protection Authority Victoria 2001), runoff would need to be diluted by a factor of 25. At the optimum POlsen of 15 mg/kg, the predicted runoff P concentration for the sheep pasture sites was 0.52 mg/L (95% confidence interval 0.19 – 1.37 mg/L). Whilst the confidence interval of the estimated concentration is large, even the lower limit exceeds concentrations capable of supporting algal blooms (0.02 mg P/L).

The range of Psatox required for maintaining groundwater, EPC or runoff DRP concentrations below 0.2 mg/L for soils in the Netherlands and the USA was 12.5- 25% (Breeuwsma and Silva 1992; Beauchemin and Simard 1999; Pote et al. 1999a). Using the regression parameters in Table 7-6, a PsatL of 0.8% in the Victorian pasture soils would lead to a TP concentration in runoff of 0.2 mg/L. According to equation 7-4, this level of PsatL is equivalent to 3.5% Psatox, which is well below the thresholds for the overseas soils. Only twenty-two per cent of the 90 soils that represented a range of agricultural and horticultural regions in Australia had PsatL ≤ 0.8%. This is not surprising considering STP levels optimal for pasture production range from an POlsen (0-10cm depth) of 15 to 20 mg/kg, which was approximately equivalent to 2.7 - 3.7% PsatL (Figure 7-6). A threshold of 0.8% would therefore not be feasible as a regulatory soil P level for maintaining water quality because soils may be deficient in P, which favours the dominance of pasture weeds rather than good quality forage or native species (Snaydon 1981), and can lead to enhanced erosion due to poor plant growth and soil protection. In contrast, threshold STP levels for Irish soils to achieve a water quality target of 0.035 mg P/L in receiving waters, a level which is similar to the Victorian stream guideline level, are at the lower end of the range for optimum pasture production (Tunney et al. 2001). Clearly threshold STPs will vary between environments, land management and soil type depending on the relationship between runoff P and STP and the water quality target. As well as this, Australian freshwater systems are generally shallower, have higher water temperatures and are more turbid than those in the northern hemisphere and are often ephemeral with episodic flows. It is therefore likely that the response of these freshwater systems to excess nutrient loadings will differ from models developed for deep, clear, perennial lakes in Northern America (Harris 1994). The contrast in environments supports the need for localised information on relationships between runoff and soil P because those developed in contrasting environments and landuse are unlikely to be of relevance to Australian soils and conditions.

241 7.5 Conclusions

The POlsen test, which is routinely performed in Victoria for assessing fertiliser requirements of pastures, is considered adequate in providing information on both the plant- available soil P and the propensity for a soil to release P to runoff. The results suggest an extra test such as a water extraction is unlikely to add to this information and is therefore not necessary. Whilst Psat was no better in predicting runoff P concentrations than POlsen for the four soils studied in detail in this study, it could be a helpful measure when comparing the potential for P to be released from soils with strongly contrasting P sorption characteristics.

Psatox was found to be the more robust and feasible method for estimating the degree of P sorption saturation of soils, considering the lack of standardisation in the determination of sorption isotherms to derive the PsatL. The Pmax of five pasture soils in Victoria was well correlated to the sum of Alox and Feox suggesting P was largely sorbed to amorphous Al and Fe mineral components, which was consistent with the expected influence of the soil parent material. The relatively weak relationship between runoff P concentrations and STP across the four Victorian pasture soils was in part due to the variation in hydrology at the sites. The annual volume of runoff was an important factor in determining not only P loads in runoff but also annual P concentrations by diluting the P sources. The influence of hydrology, soil type, land use and management was also likely to influence any comparison of these empirical data to other studies, however, runoff P concentrations were generally higher at the Victorian sites for an equivalent STP range than for many European and USA soils.

For the Victorian soils, optimum POlsen levels of between 15 and 20 mg/kg were predicted to lead to TP concentrations in runoff of approximately 0.5 –1.0 mg/L, which are an order of magnitude greater than the concentration considered to be environmentally acceptable.

Soils with a P status below these optimum POlsen levels may be unsustainable from both a production and erosion viewpoint. Considering threshold STPs set for soils and water quality targets overseas are sometimes 2 to 6 times greater than the optimum STP for plant production, regulations imposed on the soil P status of overseas soils are unlikely to be relevant for Australian soils, hydrology and water quality targets. This also highlights the need for research into the relationships between runoff and drainage water quality and soil P status for a wider range of soils and agricultural practices in order to establish threshold soil P levels.

242 CHAPTER EIGHT

8 General Discussion and Conclusions

This chapter discusses the implications of the main findings of this study in terms of identifying the processes governing hydrological movement of P from sheep pastures and how P losses can be predicted and managed. It also suggests priorities for further research.

8.1 Hydrological pathways of P movement from sheep-grazed pastures

A combination of duplex soils, undulating topography and winter-spring dominant rainfall led to saturation excess runoff from variable source areas being the major source of water movement from the Vasey field site. A comparison with river flow downstream in the catchment area suggested the average runoff activity across the 2 ha field site was typical of the region. It is likely that the spatial and temporal variability in surface runoff generation across the hillslope also reflected the hydrological behaviour of the wider catchment, considering duplex soils and undulating topography are typical landscape features in south-west Victoria. The risk of P loss was greater for surface runoff than subsurface flow pathways in this environment due to a combination of higher concentrations of P at the pasture surface and the potential for large volumes of runoff to be shed during infrequent storms. The implications of the hydrological behaviour in terms of the potential for nutrient movement from pastures are that areas in the landscape that are prone to waterlogging and surface runoff during the winter should be targeted for minimising P movement. The concept of managing critical runoff source areas to control nutrient movement has also been well documented for other environments such as in the USA (Gburek and Sharpley 1998) and New Zealand (McColl et al. 1985). The corollary of this is that areas of the property that are better drained or not hydrologically connected to the receiving waterways will be less likely to cause P enrichment of streams, which offers opportunities for intensifying pasture grazing management on these areas. In well-drained pastures and in drier years subsurface water movement, as a combination of deep drainage and lateral flows, is likely to dominate over surface flows. However, P movement along subsurface pathways is of lesser concern because of the high P- fixing potential of the iron and clay-rich soils and because of the more tortuous pathway water travels through, compared with surface runoff. In contrast, nitrate leaching to groundwater may need to be considered because nitrate concentrations in drainage below the rootzone measured - in this study were higher than ecological water guideline levels of 0.01 mg NO3 -N/L.

243 8.2 The impact of increasing P fertiliser application and stocking rates on the loads and concentrations of P in surface runoff

The study showed that the amount of P loss along surface pathways was primarily dependant on the volume of runoff, rather than the combined influence of stocking rate and level of soil P fertility. For a given volume of either hillslope or simulated runoff, however, more P was lost from higher fertility pasture systems due to higher average TP concentrations in the runoff. This suggests that as long as P fertiliser application rates continue to rise (as advocated) in south-west Victoria, it is likely that the runoff of P from pasture to waterways will increase. Across both the low and high P fertility pasture treatments the average annual concentrations of TP in runoff were at least five times higher than desirable concentrations in freshwater streams (0.04 mg P/L). It is therefore unlikely that concentrations of P in runoff from even the low fertility pastures could be kept below target stream concentrations. This highlights a challenge associated with managing nutrients in areas where pasture runoff dominates dam and stream water supply. It is therefore important that close attention be paid to nutrient management of sheep pastures to avoid deterioration of surface water quality, particularly where P fertiliser and stocking rates are increased to boost production and there is a high degree of hydrological connectivity between the pasture and receiving waterways.

8.3 Important processes of P mobilisation in runoff

Identification of the main processes of P mobilisation in runoff at Vasey highlighted the most appropriate strategies for minimising P losses from sheep-grazed pastures in this environment. Whilst both particulate P (PP) and dissolved P (DP) were identified as important P fractions in runoff at Maindample and Vasey, dissolved reactive P (DRP), the form of P most readily available for algal uptake, is the P fraction in runoff of most concern as soil P fertility increases. An increase in the concentration of DRP with increasing treatment POlsen suggested that desorption, extraction and dissolution of predominantly inorganic P from surface materials were the most important processes of P mobilisation in runoff as soil P fertility increased. Dissolved P concentrations in natural runoff from hillslope plots also appeared to be more strongly influenced by POlsen (being representative of the P availability in soil, plant, residual fertiliser and organic matter) than by sorption equilibria between DP and eroded soil particles. A small amount of erosion of P deficient soil might therefore reduce the concentration of P that is in the form immediately availability to algae (DRP) at low POlsen, but it is unlikely to significantly reduce runoff DRP concentrations at higher POlsen .

244 Low levels of erosion and PP loss in the hillslope runoff suggested that PP movement is not a major concern in well-managed sheep pastures. However, despite the low level of erosion, PP concentrations increased with increasing soil loss in both hillslope and simulated runoff. Particulate P losses were also more heavily influenced by factors that influenced the SS loads and size distribution of particles, than by the amount of P adsorbed to the soil particles. This suggested that PP losses can be minimised by maintaining near-complete groundcover in this environment. However, the implications of the dominance of DRP forms in runoff for managing P losses are that erosion control measures such as vegetative buffer strips at the edge of paddocks, streams and dams are not likely to be completely effective in controlling P movement from fertile pastures. Under the high rainfall intensity of the simulated rainfall, PP concentrations increased with increasing POlsen. This suggests that during high intensity storms the soil P status may affect both the PP and DP concentrations in runoff, reinforcing the need to restrict the P fertility of pastures in runoff source areas. Concentrations of P were higher in larger volume runoff events than small, and lower in surface runoff that mixed with low P, saline discharge, which again demonstrated the importance of hydrology, as well as chemistry and biology, in the mobilisation of P in runoff. The potential mixing of surface flows from pastures with groundwater discharge, perched water return flow and surface flows from conservation areas therefore needs to be considered in an holistic approach to managing catchment water quality.

8.4 Recommendations for minimising P losses from pastures

From a pasture production view-point alone there is little financial incentive to manage nutrient losses in runoff considering total amounts of P lost in runoff were equivalent to less than 1% of the P applied as fertiliser. However there are costs associated with eutrophication such as the physical nuisance and risks to animal health of blue-green algal blooms in farm dams, and the potential for degraded water quality to harm the clean and green reputation of the wool and meat industries. These costs need to be factored into decisions about nutrient and water management. The probability of runoff and therefore P loss occurring from different parts of grazing properties should therefore be considered along with other soil and water issues such as nitrate leaching, salinity and erosion control when establishing whole farm plans. In order to reduce total P losses and improve the quality of runoff from pastures with close to 100% groundcover, the amount of inorganic P available at the soil surface in areas prone to runoff production should be decreased. It is likely this management would significantly increase the amount of good quality water yielded from undulating pasture catchments in the south-west of Victoria. Reducing the source of inorganic P may be achieved by reducing fertiliser applications and stocking rates in runoff source areas during winter and spring and by retiring these areas from production. To minimise erosion and associated PP movement,

245 stocking rates should be adjusted in both low and high fertility pastures, and rotation lengths managed carefully, to ensure pasture cover remains above 70% all year-round and that pugging is minimised on wet soils. Grazing management strategies should also be used to reduce the extent and location of stock camps and grazing events in relation to where and when runoff occurs. High fertiliser and stocking rates should be used on well-drained soils (providing there is a low risk of nitrate leaching to groundwater) or where surface flows are likely to reinfiltrate downslope rather than in seasonally waterlogged, sloping pastures. If wet areas of pasture are retired from production, plants with low nutrient demands such as native species, grasses and shrubs should be sown to avoid groundcover decline and erosion. The influence of converting pasture to forestry on the quality of surface runoff should also be investigated because this is becoming common practice in the south-west of Victoria.

8.5 Using soil P tests and simple models to predict P concentrations in runoff

The processes of DP mobilisation identified in this study indicated that an empirical model that incorporates the effects of both desorption and dissolution of P from surface materials, as well as adsorption to entrained sediment may provide a better prediction of DP concentrations than models that focus on either desorption or sorption processes alone. For the soils studied, the Erosion Productivity Impact Calculator (EPIC)(Williams et al. 1983) model, using an adjusted linear dissociation constant (Kd) to describe the partitioning of P between sediment and solution phases, was reasonably well suited to predicting annual average DP concentrations in runoff. By using a linear Kd parameter, however, the EPIC model is analogous to simple relationships developed between concentrations of P in runoff and soil P tests. Given the reasonable ability for a single soil P dissociation constant to describe DP concentrations in runoff, the utility of soil P tests as indices of potential runoff P concentrations was tested for a range of Victorian pasture soils and soil tests. For runoff from sheep pastures,

POlsen (0-5 cm) was a better predictor of mean annual TP concentrations than the degree of P saturation (Psat) or water-extractable P (Pwater). This suggests that for the soil types studied, a soil test that is used for routine agronomic purposes can provide an indication of TP concentrations in runoff from pastures. There is little incentive therefore to develop a new soil test that describes the availability of P to runoff rather than to plants for these soils, as has been the case for some soils overseas (Gartley and Sims 1994). However, because Psat describes aspects of soil P sorption characteristics, it may be a useful test in combination with a routine soil P test should a larger range of soil types be evaluated for the risk of P losses. Factors other than the soil P status, such as runoff volume and the timing of fertiliser application in relation to runoff, should also be considered in estimates of runoff P concentrations. For example, there

246 was an inverse relationship between the annual TP concentrations and volume of runoff at each site, which partly explains the relatively poor predictive ability of soil P tests alone. Differences in hydrology, soil type, land use and management are likely to influence any comparison of these empirical runoff P-soil P relationships with other studies. However, runoff P concentrations were generally higher at the Victorian sites for an equivalent STP range than for many European and USA soils. For the Victorian soils, agronomically optimum POlsen levels of between 15 and 20 mg/kg (approximately equivalent to 2.7 - 3.7% PsatL) were predicted to lead to TP concentrations in runoff of approximately 0.5 – 1.0 mg/L, which are an order of magnitude greater than the concentration considered to be environmentally acceptable.

Soils with a P status below these optimum POlsen levels may be unsustainable from both a production and erosion viewpoint. Considering threshold STPs set for soils and water quality targets overseas are sometimes 2 to 6 times greater than the optimum STP for plant production, regulations imposed on the soil P status of overseas soils are unlikely to be relevant for Australian soils, hydrology and water quality targets. As well as runoff P- soil P relationships, dilution and attenuation of P beyond field boundaries should be investigated and considered in establishing threshold soil P levels.

8.6 Further research

The national water quality management strategy (Anon 2000a) outlines a framework for estimating the risk of negative biological impacts due to changes in water quality parameters. With respect to nutrient concentrations in water, proposed trigger levels below which there is a low risk of such deleterious conditions arising must be assessed for each location. P and N in runoff from pastures during winter and spring can degrade the quality of surface waters if these flows are not diluted by low nutrient runoff and groundwater from other catchment areas. However, high evaporation, water extraction for stock and domestic supplies and P release from bottom sediments during the hot summers of the Mediterranean environment of Victoria increase the concentrations of nutrients in farm dams (Turoczy 1999) and waterways, which may partially negate any prior runoff dilution effects. This ‘summer concentration’ effect suggests threshold P concentrations for paddock runoff should be lower than stream and reservoir threshold levels. However, detailed research on the relative contributions of flow from different parts of the catchment landscape is required to identify the degree of dilution of pasture runoff that occurs. The results from the current study suggest that runoff from low P fertility pastures is unlikely to adequately dilute P-rich runoff waters, and this presents a challenge for defining runoff water quality targets that are both acceptable and achievable. Further research into the relationships between hillslope runoff water quality and soil P status for a wider range of soils, agricultural practices and catchments in

247 Australia is required in order to establish threshold soil P levels. Dilution and attenuation of P beyond field boundaries should be investigated and considered in establishing thresholds. Environmentally critical soil P levels may be useful as part of a P loss risk indexing scheme (Gburek et al. 2000), and could be used in combination with soil type and agronomic soil P requirements to better inform nutrient management decisions. Nutrient management strategies should be aimed at reducing the amount of P available in important runoff source areas rather than at reducing runoff yields because this would have a further detrimental impact on catchment water quality and ecosystem health as well as jeopardising the water supplies. Managing P mobility and availability in VSAs is also likely to be more effective than increasing the reliance on buffer strips to remove nutrients from runoff because particulate P is not the clearly dominant source of P. Modelling, indexing and observation systems for identifying runoff source areas within catchments and on farms should be developed further and integrated into farm nutrient management decision frameworks. The medium to long-term effectiveness of reducing the amount of P available in runoff source areas by reducing P application and stocking rates or fencing off and planting native, low nutrient demanding vegetation should then also be investigated under field conditions. A cost- benefit analysis of managing runoff source areas at either low-P, low stocking rate or retiring these areas from production and managing as ‘water-producing’ areas should also be performed to better inform farmers of the long-term cost-effectiveness of these nutrient management options.

248 REFERENCES

Adamson, C. M. (1976) Some effects of soil conservation treatment on the hydrology of a small rural catchment at Wagga. J. Soil Conserv. Serv. NSW 32, 230-249. Ahuja, L. R., Sharpley, A. N., Yamamoto, M. and Gburek, W. (1981) The depth of rainfall- runoff interaction as determined by 32P. Water Resour. Res. 17, 969-974. Ahuja, L. R., Sharpley, A. N. and Lehman, O. R. (1982) Effect of soil slope and rainfall characteristics on phosphorus in runoff. J. Environ. Qual. 11, 9-13. Alderfer, R. B. and Robinson, R. R. (1947) Runoff from pastures in relation to grazing intensity and soil compaction. J. Am. Soc. Agron. 39, 948-958. Allan, C. J., Mason, W. K., Reeve, I. J. and Hooper, S. (2003) Evaluation of the impact of SGS on livestock producers and their practices. Aust. J. Exp. Agric. 43, 1031-1040. Anderson, D. L., Hanlon, E. A., Miller, O. P., Hoge, V. R. and Diaz, O. A. (1992) Soil sampling and nutrient variability in dairy animal holding areas. Soil Sci. 153, 314-321. Anderson, M. G. and Burt, T. P. (1978) The role of topography in controlling throughflow generation. Earth Surf. Process. Landforms 3, 331-344. Anderson, M. G. and Kneale, P. E. (1982) The influence of low-angled topography on hillslope soil-water convergence and stream discharge. J. Hydrol. 57, 65-80. Andrew, M. H. and Lodge, G. M. (2003) The Sustainable Grazing Systems National Experiment 1: Introduction and methods. Aust. J. Exp. Agric. 43, 695-709. Anon (1987) "Design Rainfall Intensity,". Hydrometeorological Service, Commonwealth of Australia Bureau of Meteorology, Melbourne. Anon (1998a) "Method 020: Determination of Ca, Mg, K, S, P, Fe, Cu, Zn, Mn, Al and B in plant matter by microwave digestion and ICP emission spectroscopy,". State Chemistry Laboratory, Werribee, Victoria. Anon (1998b) "Method 054: Available Soil Phosphate by Olsen's Bicarbonate Method,". State Chemistry Laboratry, Werribee, Victoria. Anon (1999) "National Water Quality Management Strategy: Australian and New Zealand Guidelines for Fresh and Marine Water Quality. Draft,". Australia and New Zealand Environment and Conservation Council and Agriculture and Resource Management Council of Australia and New Zealand. Anon (2000a) "National Water Quality Management Strategy: 9 Rural landuses and water quality,". Australia and New Zealand Environment and Conservation Council and Agriculture and Resource Management Council of Australia and New Zealand. Anon (2000b) "Australian and New Zealand Guidelines for Fresh and Marine Water Quality Paper No. 4 Volume 1 The Guidelines - Aquatic Ecosystems,". Australia and New

249 Zealand Environment and Conservation Council and Agriculture and Resource Management Council of Australia and New Zealand. Anon. (1995) Surfer (Win32) Surface Mapping System. Golden Software. Atkinson, T. C. (1978) Techniques for measuring subsurface flow on hillslopes. In "Hillslope Hydrology" (M. Kirkby, ed.), pp. 73-120. John Wiley and Sons, GB. Austin, N. R., Prendergast, J. B. and Collins, M. D. (1996) Phosphorus losses in irrigation runoff from fertilised pasture. J. Environ. Qual. 25, 63-68. Baldwin, D. S. (1996) Effects of exposure to air and subsequent drying on the phosphate sorption characteristics of sediments from a eutrophic reservoir. Limnol. Oceanogr. 41, 1725-1732. Baldwin, D. S. (1998) Reactive 'organic' phosphorus revisited. Water Res. 32, 2265-2270. Ballard, R. and Fiskell, J. G. A. (1974) Phosphorus retention in coastal plain forest soils: I. Relationship to soil properties. Soil Sci. Soc. Am. Proc. 38, 250-255. Barbare, A. M., Gilkes, R. J. and Sale, P. W. G. (1997) The effect of phosphate buffering capacity and other soil properties on North Carolina phosphate rock dissolution, availability of dissolved phosphorus and relative agronomic effectiveness. Aust. J. Exp. Agric. 37, 1037-1049. Bardgett, R. D. and Cook, R. (1998) Functional aspects of soil animal diversity in agricultural grasslands. Appl. Soil Ecol. 10, 263-276. Bargh, B. J. (1978) Output of water, suspended sediment, and phosphorus and nitrogen forms from a small agricultural catchment. N. Z. J. Agric. Res. 21, 29-38. Barling, R. D., Moore, I. D. and Grayson, R. B. (1994) A quasi-dynamic wetness index for characterizing the spatial distribution of zones of surface saturation and soil water content. Water Resour. Res. 30, 1029-1044. Barrow, N. J. and Lambourne, L. J. (1962) Partition of excreted nitrogen, sulphur, and phosphorus between the faeces and urine of sheep being fed pasture. Aust. J. Agric. Res. 13, 461-471. Barrow, N. J. (1975) Chemical form of inorganic phosphate in sheep faeces. Aust. J. Soil Res. 13, 63-7. Barrow, N. J. and Shaw, T. C. (1975) The slow reactions between soil and anions: 2. Effects of time and temperature on the decrease in phosphate concentration in the soil solution. Soil Sci. 119, 167-77. Barrow, N. J. (1978) The description of phosphate adsorption curves. J. Soil Sci. 1978, 447-462. Barrow, N. J. (1979a) The description of desorption of phosphate from soil. J. Soil Sci. 30, 259- 270. Barrow, N. J. (1979b) Three effects of temperature on the reactions between inorganic phosphate and soil. J. Soil Sci. 30, 271-279.

250 Barrow, N. J. and Shaw, T. C. (1979a) Effects of solution:soil ratio and vigour of shaking on the rate of phosphate adsorption by soil. J. Soil Sci. 30, 67-76. Barrow, N. J. and Shaw, T. C. (1979b) Effects of ionic strength and nature of the cation on desorption of phosphate from soil. J. Soil Sci. 30, 53-65. Barrow, N. J. (1980) Differences amongst a wide-ranging collection of soils in the rate of reaction with phosphate. Aust. J. Soil Res. 18, 215-224. Barrow, N. J. (1984) Modelling the effects of pH on phosphate sorption by soils. J. Soil Sci. 35, 283-297. Barrow, N. J. and Ellis, A. S. (1986) Testing a mechanistic model. V. The points of zero salt effect for phosphate retention, for zinc retention and for acid/alkali titration of a soil. J. Soil Sci. 37, 303-310. Barrow, N. J., Bolland, M. D. A. and Allen, D. G. (1998) Effect of previous additions of superphosphate on sorption of phosphate. Aust. J. Soil Res. 36, 359-72. Beauchemin, S., Simard, R. R. and Cluis, D. (1996) Phosphorus sorption-desorption kinetics of soil under contrasting land uses. J. Environ. Qual. 25, 1317-1325. Beauchemin, S. and Simard, R. R. (1999) Soil phosphorus saturation degree: Review of some indices and their suitability for P management in Quebec, Canada. Can. J. Soil Sci. 79, 615-625. Beckett, P. H. T. and White, R. E. (1964) Studies on the phosphate potentials of soils Part III. The pool of labile inorganic phosphate. Plant Soil 21, 253-282. Betson, R. P. and Marius, J. B. (1969) Source areas of storm runoff. Water Resour. Res. 5, 574- 582. Betteridge, K., Mackay, A. D., Russell, J. R., Costall, D. A. and Budding, P. J. (1998) Effect of cattle treading on the soil and pasture resource and the wider environment. In "Animal production systems and the environment; An International Conference on Odor, Water Quality, Nutrient Management and Socioeconomic Issues", pp. 29-34, Des Moines, Iowa, USA. Beven, K. (1989) Changing ideas in hydrology - the case of physically-based models. J. Hydrol. 105, 157-172. Bieleski, R. L. (1973) Phosphate pools, phosphate transport, and phosphate availability. Ann. Rev. Plant Physiol. 24, 225-52. Black, C. A., Evans, D. D., Ensminger, L. E., White, J. L. and Clark, F. E., eds. (1965) "Methods of soil analysis. Part 2. Chemical and Microbiological Properties" First edn American Society of Agronomy, Inc., Madison, Wisconsin, USA. Bloom, P. R., McBride, M. B. and Weaver, R. M. (1979) Aluminium organic matter in acid soils: Buffering and solution aluminium activity. Soil Sci. Soc. Am. J. 43, 488-493.

251 Bolan, N. S. (1991) A critical review on the role of mycorrhizal fungi in the uptake of phosphorus by plants. Plant Soil 134, 189-207. Borggaard, O. K., Jorgensen, S. S., Moberg, J. P. and Raben-Lange, B. (1990) Influence of organic matter on phosphate adsorption by aluminium and iron oxides in sandy soils. J. Soil Sci. 41, 443-449. Bostrom, B., Persson, G. and Broberg, B. (1988) Bioavailability of different phosphorus forms in freshwater systems. Hydrobiologia 170, 133-155. Bramley, R. G. V., Edis, R. B., White, R. E. and Wood, A. W. (1998) "Environmentally sound phosphorus management for sugarcane soils. Final Report on SRDC Project No. CSS3S,". Sugar Research and Development Corporation. Bramley, R. G. V. and Wood, A. W. (2000) "Risk assessment of phosphorus (P) loss and guidelines for P use in lower Herbert soils. Final Report on SRDC Project No. CLW010,". Sugar Research and Development Corporation. Breeuwsma, A. and Silva, S. (1992) "Phosphorus fertilisation and environmental effects in The Netherlands and the Po region (Italy)," Rep. No. 57. The Winard Staring Centre, Wageningen, The Netherlands. Broberg, O. and Pettersson, K. (1988) Analytical determination of orthophosphate in water. Hydrobiologia 170, 45-59. Broberg, O. and Persson, G. (1988) Particulate and dissolved phosphorus forms in freshwater: composition and analysis. Hydrobiologia 170, 61-90. Brodison, J. A., Goodall, E. A., Armstrong, J. D., Givens, D. I., Gordon, F. J., McCaughey, W. J. and Todd, J. R. (1989) Influence of dietary phosphorus on the performance of lactating dairy cattle. J. Agric. Sci. 112, 303-306. Bromfield, S. M. (1961) Sheep faeces in relation to the phosphorus cycle under pastures. Aust. J. Agric. Res. 12, 111-123. Bromfield, S. M. and Jones, O. L. (1972) The initial leaching of hayed-off pasture plants in relation to the recycling of phosphorus. Aust. J. Agric. Res. 23, 811-24. Brookes, P. C., Catt, J. A., Farina, R., Heckrath, G., Rose, K. R. and Thomas, D. (1996) The movement of phosphorus from soil to water. Soil Use Manage. 12, 224. Brouwer, J. and Fitzpatrick, R. W. (1998) Relations between soil macro-morphology and current soil hydrology in a toposequence in SE Australia. In "Proceedings of the International Soil Science Society Congress. Symposium 15", pp. CD-ROM, Montpellier, France. Brouwer, J. and Anderson, H. (2000) Water holding capacity of ironstone gravel in a typic plinthoxeralf in southeast Australia. Soil Sci. Soc. Am. J. 34, 1603-1608.

252 Brouwer, J. and Fitzpatrick, R. W. (2002a) Restricting layers, flow paths, and correlation between duration of soil saturation and soil morphological features along a hillslope with an altered soil water regime in western Victoria. Aust. J. Soil Res. 40, 927-946. Brouwer, J. and Fitzpatrick, R. W. (2002b) Interpretation of morphological features in a salt- affected duplex soil toposequence with an altered soil water regime in western Victoria. Aust. J. Soil Res. 40, 903-926. Bundy, L. G. and Sturgul, S. J. (2001) A phosphorus budget for Wisconsin cropland. J. Soil Water Conserv. 56, 243-249. Burkitt, L. L., Moody, P. W., Gourley, C. J. P. and Hannah, M. C. (2002) A simple phosphorus buffering index for Australian soils. Aust. J. Soil Res. 40, 497-513. Burwell, R. E., Timmons, D. R. and Holt, R. F. (1975) Nutrient transport in surface runoff as influenced by soil cover and seasonal periods. Soil Sci. Soc. Am. J. 39, 523-528. Busman, L. M. and Tabatabai, M. A. (1985) Hydrolysis of trimetaphosphate in soils. Soil Sci. Soc. Am. J. 49, 630-636. Caruso, B. S. (2001) Risk-based targeting of diffuse contaminant sources at variable spatial scales in a New Zealand high country catchment. J. Environ. Manage. 63, 249-268. Cayley, J. W. D., Hannah, M. C., Kearney, G. A. and Clark, S. G. (1998) Effects of phosphorus fertiliser and rate of stocking on the seasonal pasture production of perennial ryegrass- subterranean clover pasture. Aust. J. Agric. Res. 49, 233-48. Cayley, J. W. D. and Kearney, G. A. (1999) Changes in bicarbonate-extractable phosphorus of a basalt-derived duplex soil associated with applications of superphosphate to pasture grazed by sheep. Aust. J. Agric. Res. 50, 546-54. Cayley, J. W. D., McCaskill, M. R. and Kearney, G. A. (2002) Available phosphorus, sulfur, potassium and other cations in a long-term grazing experiment in south western Victoria. Aust. J. Agric. Res. 53, 1349-1360. Chapman, D. F., McCaskill, M. R., Quigley, P. E., Thompson, A. N., Graham, J. F., Borg, D., Lamb, J., Kearney, G. A., Saul, G. R. and Clark, S. C. (2003) Effects of grazing method and fertiliser inputs on the productivity and sustainability of phalaris-based pastures in Western Victoria. Aust. J. Exp. Agric. 43, 785-798. Chapman, P. J., Edwards, A. C. and Shand, C. A. (1997) The phosphorus composition of soil solutions and soil leachates: Influence of soil: solution ratio. Eur. J. Soil Sci. 48, 703- 710. Chardon, W. J., Oenema, O., delCastilho, P., Vriesema, R., Japenga, J. and Blaauw, D. (1997) Organic phosphorus in solutions and leachate from soils treated with animal slurries. J. Environ. Qual. 26, 372-378.

253 Chichester, F. W., Keuren, R. W. V. and McGuinness, J. L. (1979) Hydrology and chemical quality of flow from small pastured watersheds: II Chemical quality. J. Environ. Qual. 8, 167-171. Chittleborough, D. J. (1992) Formation and pedology of duplex soils. Aust. J. Exp. Agric. 32, 815-25. Chittleborough, D. J., Varcoe, J. and Cox, J. W. (1996) Origin and development of cutans in a texture contrast soil at Keyneton, South Australia. In "Australian and New Zealand National Soils Conference: Soil Science - Raising the Profile", pp. 43. Australian Soil Science Society Inc., Melbourne. Chorley, R. J. (1978) The Hillslope Hydrological Cycle. In "Hillslope Hydrology" (M. Kirkby, ed.). John Wiley and Sons, GB. Ciesiolka, C. A. A. and Freebairn, D. M. (1982) The influence of scale on runoff and erosion. In "Agricultural Engineering Conference", pp. 203-206. The Institution of Engineers, Australia, Armidale, New South Wales, Australia. Clausen, J. C., Guillard, K., Sigmund, C. M. and Dors, K. M. (2000) Water quality changes from riparian buffer restoration in Connecticut. J. Environ. Qual. 29, 1751-1761. Clesceri, L. S., Greenberg, A. E. and Eaton, A. D., eds. (1998) "Standard Methods for the Examination of Water and Wastewater" 20th edn American Public Health Association, USA. Clothier, B. E. and Smettem, K. R. J. (1990) Combining laboratory and field measurements to define the hydraulic properties of soil. Soil Sci. Soc. Am. J. 54, 299-304. Condron, L. M., Toor, G. S., Di, H. J., Cameron, K. C., Hendry, T. and McLenaghen, R. D. (2000) Phosphorus loss from soil under irrigated dairy pasture: assessment and management. In "New Zealand Fertiliser Manufacturer's Research Association 26th Technical Conference", Lincoln University. Cooke, I. J. (1966) A kinetic approach to the description of soil phosphate status. J. Soil Sci. 17, 56-64. Cooke, J. G. (1988) Sources and sinks of nutrients in a New Zealand hill pasture catchment II Phosphorus. Hydrol. Proc. 2, 123-133. Cooke, J. G. and Dons, T. (1988) Sources and sinks of nutrients in a New Zealand hill pasture catchment I. Stormflow generation. Hydrol. Proc. 2, 109-122. Cooper, A. B., Smith, C. M. and Bottcher, A. B. (1992) Predicting runoff of water, sediment and nutrients from a New Zealand grazed pasture using CREAMS. Trans. ASAE 35, 105-112. Cooper, A. B., Smith, C. M. and Smith, M. J. (1995) Effects of riparian set-aside on soil characteristics in an agricultural landscape: Implications for nutrient transport and retention. Agric. Ecosyt. Environ. 55, 61-67.

254 Corbeels, M., Hartmann, R., Hofman, G. and Van Cleemput, O. (1999) Field calibration of a neutron moisture meter in vertisols. Soil Sci. Soc. Am. J. 63, 11-18. Cornforth, I. S. and Sinclair, A. G. (1982) Model for calculating maintenance phosphate requirements for grazed pastures. N. Z. J. Exp. Agric. 10, 53-61. Cornish, P. S., Hallissey, R. and Hollinger, E. (2002) Is a rainfall simulator useful for estimating phosphorus runoff from pastures - a question of scale-dependency? Aust. J. Exp. Agric. 42, 953-959. Correll, D. L. (1998) The role of phosphorus in the eutrophication of receiving waters: A review. J. Environ. Qual. 27, 261-266. Cosgrove, D. J. (1976) Microbial transformations in the phosphorus cycle. In "Advances in Microbial Ecology" (M. Alexander, ed.), pp. 95-134. Plenum Press, New York. Costin, A. B. (1980) Runoff and soil and nutrient losses from an improved pasture at Ginninderra, Southern Tablelands, New South Wales. Aust. J. Agric. Res. 31, 533-46. Costin, A. B. and Williams, C. H. (1983) "Phosphorus in Australia," Centre for Resource and Environmental Studies, Australian National University, Canberra. Costin, A. B., Greenaway, M. A. and Wright, L. G. (1984) "Harvesting water from land,". Centre for Resource and Environmental Studies, Canberra. Cottingham, P., Bennison, G., Dunn, R., Lidson, J. and Robinson, D. (1995) "Algal blooms and nutrient status of Victorian Inland Waters,". Government of Victoria. Court, J., Barrett, A., Hill, R., Saul, G. and Shovelton, J. (1998) "Grassland's Productivity Program 1993-1996 Final Report to Members,". Grassland Society of Victoria, Victoria. Coventry, J. L., Halliwell, D. J. and Nash, D. M. (2001) The orthophosphate content of bicarbonate soil extracts. Aust. J. Soil Res. 39, 415-421. Cox, J., Kirkby, C. A., Smythe, L. J. and Chittleborough, D. J. (1996a) Soil properties which indicate phosphorus movement through a duplex soil toposequence. In "Australian and New Zealand National Soils Conference: Soil Science - Raising the Profile", pp. 55-56. Australian Soil Science Society Inc., Melbourne, Australia. Cox, J. W., McFarlane, D. J. and Skaggs, R. W. (1994) Field evaluation of DRAINMOD for predicting waterlogging intensity and drain performance in south-western Australia. Aust. J. Soil Res. 32, 653-71. Cox, J. W. and McFarlane, D. J. (1995) The causes of waterlogging in shallow soils and their drainage in southwestern Australia. J. Hydrol. 167, 175-194. Cox, J. W. and Reynolds, M. B. (1995) "Hydrogeological results from the Keynes catchment, South Australia (Stage 2)," Rep. No. 5. Co-operative Research Centre for Soil and Land Management.

255 Cox, J. W., Fritsch, E. and Fitzpatrick, R. W. (1996b) Interpretation of soil features produced by ancient and modern processes in degraded landscapes VII Water duration. Aust. J. Soil Res. 34, 803-24. Cox, J. W., Fitzpatrick, R. W., Merry, R. H., McCaskill, M. R. and Mao, R. (1998) "Characterisation of six soil profiles at the MLA SGSKP site at Vasey, Vic.," Rep. No. 38/98. CSIRO Land and Water. Cox, J. W. and Ashley, R. (2000) Water quality of gully drainage for texture-contrast soils in the Adelaide Hills in low rainfall years. Aust. J. Soil Res. 38, 959-72. Cox, J. W. and Pitman, A. (2001) Chemical concentrations of overland flow and throughflow from pastures on sloping texture-contrast soils. Aust. J. Agric. Res. 52, 211-220. Cresswell, H. P. and Smiles, D. E. (1995) Pore space and water retention, Chapter 3. In "Soil Physical Measurement and Interpretation for Land Evaluation" (K. Coughlan, N. McKenzie, W. McDonald and H. Cresswell, eds.) Australian Soil and Land Survey Handbook Series, Vol. 5, Penultimate Draft. Australian Collaborative Land Evaluation Program. Cullen, P. (1990) Land use and declining water quality. Aust. J. Soil Water Conserv. 4, 4-8. Culleton, N., Liebhardt, B., Murphy, W. E., Cullen, P. and Cuddihy, A. (2000) "Thirty years of phosphorus fertiliser on Irish pastures: Animal-soil-water relationships,". TEAGASC, Johnstown Castle Research Centre, Wexford. Curtin, D., Syers, J. K. and Bolan, N. S. (1992) Phosphate sorption by soil in relation to exchangeable cation composition and pH. Aust. J. Soil Res. 31, 137-149. Dahlhaus, P. G. and MacEwan, R. J. (1997) Dryland salinity in south west Victoria challenging the myth. In "Collected Case Studies in Engineering Geology, Hydrogeology and Environmental Geology. Third Series" (G. McNally, ed.), pp. 165-181. Environmental, Engineering and Hydrology Specialist Group (EEHSG). Dahlhaus, P. G., MacEwan, R. J., Nathan, E. L. and Morand, V. J. (2000) Salinity on the southeastern Dundas Tableland, Victoria. Aust. J. Earth Sci. 47, 3-11. Daly, K., Jeffrey, D. and Tunney, H. (2001) The effect of soil type on phosphorus sorption capacity and desorption dynamics in Irish grassland soils. Soil Use Manage. 17, 12-20. Daly, K., Mills, P., Coulter, B. and McGarrigle, M. (2002) Modeling phosphorus concentrations in Irish rivers using land use, soil type, and soil phosphorus data. J. Environ. Qual. 31, 590-599. Daniel, T. C., Edwards, D. R. and Sharpley, A. N. (1993) Effect of extractable soil surface phosphorus on runoff water quality. Trans. ASAE 36, 1079-1085. Daniel, T. C., Sharpley, A. N., Edwards, D. R., Wedepohl, R. and Lemunyon, J. L. (1994) Minimising surface water eutrophication from agriculture by phosphorus management. J. Soil Water Conserv. 49, 30-38.

256 Davis, R., Hamblin, A., O'Loughlin, E., Austin, N., Banens, R., Cornish, P., Cox, J., Hairsine, P., McCulloch, M., Moody, P., Olley, J., Prove, B., Smalls, I. and Weaver, D. (1998) "Phosphorus in the landscape: Diffuse sources to surface waters," Rep. No. 16/98. Land and Water Resrouces Research and Development Corporation. Dickinson, C. H. and Craig, G. (1990) Effects of water on the decomposition and release of nutrient from cow pats. New Phythol. 115, 139-147. Douglas, M. H. and O'Brien, L., eds. (1971) "The Natural History of Western Victoria" Australian Institute of Agricultural Science, Hamilton. DSE (2003) "Electronic data source for Victorian streams". Department of Sustainability and Environment, www.vicwaterdata.net Dunne, T. and Black, R. D. (1970a) An experimental investigation of runoff production in permeable soils. Water Resour. Res. 6, 478-490. Dunne, T. and Black, R. D. (1970b) Partial area contributions to storm runoff in a small New England watershed. Water Resour. Res. 6, 1296-1311. Dunne, T. (1978) Field studies of hillslope processes. In "Hillslope Hydrology" (M. Kirkby, ed.), pp. 227-294. John Wiley and Sons, GB. Edis, R. B., Bramley, R. G. V., White, R. E. and Wood, A. W. (2002) Desorption of P from sugarcane soils into simulated natural waters. Mar. Fresh. Res. 53, 961-970. Edwards, A. C., Cook, Y., Smart, R. and Wade, A. J. (2000) Concentrations of nitrogen and phosphorus in streams draining the mixed land-use Dee Catchment, north-east Scotland. J. Appl. Ecol. 37, 1-13. Edwards, D. R., Hann, C. T., Sharpley, A. N., Murdoch, J. F., Daniel, T. C. and Moore Jr., P. A. (1996) Application of simplified phosphorus transport models to pasture fields in northwest Arkansas. Trans. ASAE 39, 489-496. Edwards, I. J., Jackson, W. D. and Fleming, P. M. (1974) Tipping bucket gauges for measuring run-off from experimental plots. Agric. Meteorol. 13, 189-201. Edwards, K. (1987) "Runoff and soil loss studies in New South Wales,". Soil Conservation Service of NSW and Macquarie University. Eghball, B. and Gilley, J. E. (2001) Phosphorus risk assessment index evaluation using runoff measurements. J. Soil Water Conserv. 56, 202-206. Ekholm, P., Kallio, K., Salo, S., Pietilainen, O. P., Rekolainen, S., Laine, Y. and Joukola, M. (2000) Relationship between catchment characteristics and nutrient concentrations in an agricultural river system. Water Res. 34, 3709-3716. Ekholm, P. (2001) Relationships between erosion and algal-available phosphorus in agricultural runoff. In "Connecting Phosphorus Transfer from Agriculture to Impacts in Surface Waters" (P. M. Haygarth, L. M. Condron, P. J. Butler and J. S. Chisholm, eds.). Institute for Grassland and Environmental Research, Plymouth University, England.

257 Environment Protection Authority Victoria (2001) "Nutrient objectives for rivers and streams - ecosystem protection. Draft State Environment Protection Policy (Waters of Victoria),". EPA Victoria. Evans, L., Stagnitti, F. and Sherwood, J. E. (1996) Nutrient transport through basaltic agricultural soils near Warrnambool: Evidence of preferential flow. In "Australian and New Zealand National Soils Conference: Soil Science - Raising the Profile", pp. 77-78. Australian Soil Science Society Inc., Melbourne. Falconer, I. R. (1991) Tumor promotion and liver injury caused by oral consumption of cyanobacteria. Environ. Toxicol. Water Qual. 6, 177-184. Fawcett, J. and Norton, R. (2000) A new look at land degradation on the Eastern Dundas Tablelands. Agric. Sci. 13, 28-30. Fertilizer Industry Federation of Australia Inc (2003) "Industry Statistics". FIFA www.fifa.asn.au Finlayson, B. L. and Wong, N. R. (1982) Storm runoff and water quality in an undisturbed forested catchment in Victoria. Aust. For. Res. 12, 303-15. Finlayson, B. L., Betteridge, K., Mackay, A., Thorrold, B., Singleton, P. and Costall, D. A. (2002) A simulation model of the effects of cattle treading on pasture production on North Island, New Zealand, hill land. N. Z. J. Agric. Res. 45, 255-272. Fitter, A. H. and Sutton, C. D. (1975) The use of the Freundlich isotherm for soil phosphate sorption data. J. Soil Sci. 26, 241-246. Fitzpatrick, R. W. (1999) Nature and significance of minerals formed in Australian mediterranean soils during land use changes. In "6th International Meeting on soils with Mediterranean type of climate", Barcelona, Spain. Fleming, N. K. and Cox, J. W. (1998) Chemical losses off dairy catchments located on a texture-contrast soil: carbon, phosphorus, sulfur and other chemicals. Aust. J. Soil Res. 36, 979-95. Fleming, N. K. and Cox, J. W. (2001) Carbon and phosphorus losses from dairy pasture in South Australia. Aust. J. Soil Res. 39, 969-978. Fleming, N. K., Cox, J. W., Chittleborough, D. J. and Dyson, C. B. (2001) An analysis of chemical loads and forms in overland flow from dairy pasture in South Australia. Hydrol. Proc. 15, 393-405. Floate, M. J. S. (1970) Decomposition of organic materials from hill soils and pastures II. Comparative studies on the mineralization of carbon, nitrogen and phosphorus from plant materials and sheep faeces. Soil Biol. Biochem. 2, 173-185. Fordham, A. W. (1963) The measurement of chemical potential of phosphate in soil suspensions. Aust. J. Soil Res. 1, 144-156.

258 Fox, R. L. and Kamprath, E. J. (1970) Phosphate sorption isotherms for evaluating the phosphate requirements of soils. Soil Sci. Soc. Am. J. 34, 902-907. Freer, J., McDonnel, J., Beven, K. J., Brammer, D., Burns, D., Hooper, R. P. and Kendal, C. (1997) Topographic controls on subsurface storm flow at the hillslope scale for two hydrologically distinct small catchments. Hydrol. Proc. 11, 1347-1352. Freese, D., Van Der Zee, S. E. A. T. M. and Van Riemsdijk, W. H. (1992) Comparison of different models for phosphate sorption as a function of the iron and aluminium oxides of soils. J. Soil Sci. 43, 729-738. Freeze, R. A. (1972) Role of subsurface flow in generating surface runoff 1. Base flow contributions to channel flow. Water Resour. Res. 8, 609-23. Friesen, D. K., Blair, G. J. and Duncan, M. (1985) Temporal fluctuations in soil test values under permanent pasture in New England, N.S.W. Aust. J. Soil Res. 23, 181-93. Frossard, E., Condron, L. M., Oberson, A., Sinaj, S. and Fardeau, J. C. (2000) Processes govering phosphorus availability in temperate soils. J. Environ. Qual. 29, 15-23. Gardner, W. and Kirkham, D. (1952) Determination of soil moisture by neutron scattering. Soil Sci. 73, 391-401. Gartley, K. L. and Sims, J. T. (1994) Phosphorus soil testing: Environmental uses and implications. Commun. Soil Sci. Plant Anal. 25, 1565-1582. Gascho, G. J., Wauchope, R. D., Davis, J. G., Truman, C. C., Dowler, C. C., Hook, J. E., Sumner, H. R. and Johnson, A. W. (1998) Nitrate-nitrogen, soluble, and bioavailable phosphorus runoff from simulated rainfall after fertiliser application. Soil Sci. Soc. Am. J. 62, 1711-1718. Gburek, W. J. and Heald, W. R. (1974) Soluble phosphate output of an agricultural watershed in Pennsylvania. Water Resour. Res. 10, 113-118. Gburek, W. J. and Sharpley, A. N. (1998) Hydrologic controls on phosphorus loss from upland agricultural watersheds. J. Environ. Qual. 27, 267-277. Gburek, W. J., Sharpley, A. N., Heathwaite, L. and Folmar, G. J. (2000) Phosphorus management at the watershed scale: a modification of the phosphorus index. J. Environ. Qual. 29, 130-144. Gerke, J. (1992) Orthophosphate and organic phosphate in the soil solution of four sandy soils in relation to pH-evidence for humic-Fe-(Al-) phosphate complexes. Commun. Soil Sci. Plant Anal. 23, 601-612. Gibbons, F. R. and Downes, R. G. (1964) "A Study of the Land in South-Western Victoria," Soil Conservation Authority, Victoria. Gilchrist, A. N. and Gillingham, A. G. (1970) Phosphate movement in surface run-off water. N. Z. J. Agric. Res. 13, 225-31.

259 Gillingham, A. G. and During, C. (1973) Pasture production and transfer of fertility within a long-established hill pasture. N. Z. J. Exp. Agric. 1, 227-32. Gillingham, A. G. (1980) Phosphorus uptake and return in grazed, steep hill pastures I. Pasture production and dung and litter accumulation. N. Z. J. Agric. Res. 23, 313-21. Gillingham, A. G., Syers, J. K. and Gregg, P. E. H. (1980) Phosphorus uptake and return in grazed, steep hill pastures II. Above-ground components of the phosphorus cycle. N. Z. J. Agric. Res. 23, 323-30. Gillingham, A. G. (1983) The role of the grazing animal in nutrient cycling in hill pastures and implications for management. In "Foothills for Food and Forests, Symposium Series No.2" (D. B. Hannaway, ed.). Timber Press, Oregon State University, College of Agricultural Sciences. Gillingham, A. G. (1987) Phosphorus cycling in managed grasslands. In "Managed Grasslands, B. Analytical Studies" (R. Snaydon, ed.). Elsevier Science Publishers B.V., Amsterdam. Gillingham, A. G., Thorrold, B. S., Wheeler, D. M., Power, I. L., Gray, M. H. and Blennerhassett, J. D. (1997) Factors influencing phosphate losses in surface runoff water. In "The New Zealand Fertiliser Manufacturer's Research Association 24th Technical Conference" (H. Furness, ed.), pp. 144-153. NZFMRA, Auckland, New Zealand, Invercargill, New Zealand. Gillingham, A. G. and Thorrold, B. S. (2000) A review of New Zealand research measuring phosphorus in runoff from pasture. J. Environ. Qual. 29, 88-96. Gourley, C. (2001) What is Australia doing in terms of dairy nutrient management research and education? In "Babcock Institute's 3rd Technical Workshop. Nutrient Management Challenges in Livestock Operations: International and National Perspectives" (K. Kanwar and R. Baggett, eds.), pp. 71-81. Board of Regents of the University of Wisconsin, Madison, Wisconsin, USA. Gourley, C. J. P., Awty, I., O'Doherty, M. and Aarons, S. R. (2001) Phosphorus fertiliser requirements for temperate dairy pastures and milk production in south eastern Australia. In "Proceedings of the XIX International Grasslands Congress", São Pedro, Brazil. Graecen, E. L., Correll, R. L., Cunningham, R. B., Johns, G. G. and Nicholls, K. D. (1981) Calibration. In "Soil Water Assessment by the Neutron Method." (E.L.Graecen, ed.), pp. 50-72. CSIRO, Melbourne. Graecen, E. L. and Williams, J. (1983) Physical properties and water relations. In "Soils: An Australian Viewpoint" pp. 499-530. Division of Soils, CSIRO, (CSIRO:Melbourne/Academic Press: London). Grayson, R. and Western, A. (2001) Terrain and the distribution of soil moisture. Hydrol. Proc. 15, 2689-2690.

260 Grayson, R. B., Moore, I. D. and McMahon, T. A. (1992) Physically based hydrologic modeling 1. A terrain-based model for investigative purposes. Water Resour. Res. 28, 2639-2658. Greene, R. S. B., Kinnell, P. I. A. and Wood, J. T. (1994) Role of plant cover and stock trampling on runoff and soil erosion from semi-arid wooded rangelands. Aust. J. Soil Res. 32, 953-973. Greenhill, N. B., Peverill, K. I. and Douglas, L. A. (1983a) Surface runoff from sloping, fertilised perennial pastures in Victoria, Australia. N. Z. J. Agric. Res. 26, 227-231. Greenhill, N. B., Fung, K. H., Peverill, K. I. and Briner, G. P. (1983b) Nutrient content of rainwater in Victoria and its agricultural significance. Search 14, 46-7. Gregory, P. J., Tennant, D., Hamblin, A. P. and Eastham, J. (1992) Components of the water balance on duplex soils in Western Australia. Aust. J. Exp. Agric. 32, 845-55. Grierson, I. T. and Oades, J. M. (1977) A rainfall simulator for field studies of run-off and soil erosion. J. Agric. Eng. Res. 22, 37-44. Grover, B. L. and Lamborn, R. E. (1970) Preparation of porous ceramic cups to be used for extraction of soil water having low solute concentrations. Soil Sci. Soc. Am. J. 34, 706-. Hairsine, P. (1996) Comparing grass filter strips and near-natural riparian forests for buffering intense hillslope sediment sources. In "First National Conference on Stream Management in Australia", pp. 203-206, Merrijig, Australia. Hairsine, P. (1997) "Controlling sediment and nutrient movement within catchments," Rep. No. 97/9. Cooperative Research Centre for Catchment Hydrology. Hairsine, P. and Prosser, I. (1997) Reducing erosion and nutrient loss with perennial grasses. Aust. J. Soil Water Conserv. 10, 8-14. Halliwell, D., Coventry, J. L. and Nash, D. (2000) Inorganic monophosphate determination in overland flow from irrigated grazing systems. Intern. J. Environ. Anal. Chem. 76, 77- 87. Hamed, Y., Albergel, J., Pepin, Y., Asseline, J., Nasri, S., Zante, P., Berndtsson, R., El-Niazy, M. and Balah, M. (2002) Comparison between rainfall simulator erosion and observed reservoir sedimentation in an erosion-sensitive semiarid catchment. Catena 50, 1-16. Hammermeister, D. P., Kling, G. F. and Vomocil, J. A. (1982) Perched water tables on hillsides in Western Oregon: I. Some factors affecting their development and longevity. Soil Sci. Soc. Am. J. 46, 811-818. Hannapel, R. J., Fuller, W. H. and Fox, R. H. (1964) Phosphorus movement in a calcareous soil: II Soil microbial activity and organic phosphorus movement. Soil Sci. 97, 421-427. Hansen, E. A. and Harris, A. R. (1975) Validity of soil-water samples collected with porous ceramic cups. Soil Sci. Soc. Am. Proc. 39, 528-836.

261 Harris, G. (1995) Eutrophication - Are Australian Waters Different From Those Overseas? Water May/June, 9-12. Harris, G. P. (1994) "Nutrient loadings and algal blooms in Australian waters - a discussion paper.,". Occasional Paper Series12/94, Land and Water Resources Research and Development Corporation, Canberra. Hartley, R. E., Maschmedt, D. J. and Chittleborough, D. J. (1984) Land management - key to water quality control. Water, 18-24. Hatton, T. J., Bartle, G. A., Silberstein, R. P., Salama, R. B., Hodgson, G., Ward, P. R., Lambert, P. and Williamson, D. R. (2002) Predicting and controlling water logging and groundwater flow in sloping duplex soils in western Australia. Agric. Water Management 53, 57-81. Haygarth, P. M., Ashby, C. D. and Jarvis, S. C. (1995) Short-term changes in the molybdate reactive phosphorus of stored soil waters. J. Environ. Qual. 24, 1133-1140. Haygarth, P. M. and Jarvis, S. C., eds. (1996) "Pathways and forms of phosphorus losses from grazed grassland hillslopes" Advances in Hillslope Processes Vol. 1, pp. 1-John Wiley & Sons Ltd. Haygarth, P. M. and Jarvis, S. C. (1997) Soil derived phosphorus in surface runoff from grazed grasslands lysimeters. Water Res. 31, 140-148. Haygarth, P. M., Hepworth, L. and Jarvis, S. C. (1998) Forms of phosphorus transfer in hydrological pathways from soil under grazed grassland. Eur. J. Soil Sci. 49, 65-72. Haygarth, P. M. and Jarvis, S. C. (1999) Transfer of phosphorus from agricultural soils. Adv. Agron. 66, 195-249. Haygarth, P. M. (1999) Mechanisms and mitigation of phosphorus transfers from soil to water. In "Scientific basis to mitigate the nutrient dispersion into the environment" (A. Sapek, ed.). Institute for land reclamation and grassland farming, Falenty/Nadarzyn near Warsaw. Haygarth, P. M. and Sharpley, A. N. (2000) Terminology for Phosphorus transfer. J. Environ. Qual. 29, 10-15. Haygarth, P. M., Heathwaite, A. L., Jarvis, S. C. and Harrod, T. R. (2000) Hydrological factors for phosphorus transfer from agricultural soils. Adv. Agron. 69, 154-178. Haynes, R. J. and Williams, P. H. (1992) Long-term effect of superphosphate on accumulation of soil phosphorus and exchangeable cations on a grazed, irrigated pasture site. Plant Soil 142, 123-133. Haynes, R. J. and Williams, P. H. (1993) Nutrient cycling and soil fertility in the grazed pasture ecosystem. In "Adv. Agron." (D. Sparks, ed.), Vol. 49. Academic Press, USA. Haynes, R. J. (1999) Size and activity of the soil microbial biomass under grass and arable management. Biol. Fertil. Soils 30, 210-216.

262 Haynes, R. J. and Williams, P. H. (1999) Influence of stock camping behaviour on the soil microbiological and biochemical properties of grazed pastoral soils. Biol. Fertil. Soils 28, 253-258. Heathwaite, A. L., Griffiths, P. and Parkinson, R. J. (1998) Nitrogen and phosphorus in runoff from grassland with buffer strips following application of fertilizers and manures. Soil Use Manage. 14, 142-148. Heathwaite, A. L. and Dils, R. M. (2000) Characterising phosphorus loss in surface and subsurface hydrological pathways. Sci. Total. Env. 251/252, 523-538. Heathwaite, L., Sharpley, A. and Gburek, W. (2000) A conceptual approach for integrating phosphorus and nitrogen management at watershed scales. J. Environ. Qual. 29, 158- 166. Heckrath, G., Brookes, P. C., Poulton, P. R. and Goulding, K. W. T. (1995) Phosphorus leaching from soils containing different phosphorus concentrations in the Broadbalk experiment. J. Environ. Qual. 24, 904-910. Heng, L. K., White, R. E., Helyar, K. R., Fisher, R. and Chen, D. (2001) Seasonal differences in the soil water balance under perennial and annual pastures on an acid Sodosol in southeastern Australia. Eur. J. Soil Sci. 52, 227-236. Herath, G. (1997) Freshwater algal blooms and their control: Comparison of the European and Australian experience. J. Env. Manage. 51, 217-227. Higgs, B., Johnston, A. E., Salter, J. L. and Dawson, C. J. (2000) Some aspects of achieving sustainable phosphorus use in agriculture. J. Environ. Qual. 29, 80-87. Hodgkinson, R. A. (1996) Phosphorus loss to surface waters from two lowland agricultural catchments. Soil Use Manage. 12, 226-7. Holford, I. (1989) Phosphate behaviour in soils. Agric. Sci. 2, 15-21. Holford, I. C. R., Wedderburn, R. W. M. and Mattingly, G. E. G. (1974) A Langmuir two- surface equation as a model for phosphate adsorption by soils. J. Soil Sci. 25, 242-255. Holford, I. C. R. and Mattingly, G. E. G. (1975) The high- and low-energy phosphate adsorbing surfaces in calcareous soils. J. Soil Sci. 26, 407-417. Holford, I. C. R. and Patrick Jnr, W. H. (1979) Effects of reduction and pH changes on phosphate sorption and mobility in an acid soil. Soil Sci. Soc. Am. J. 43, 292-297. Holford, I. C. R. and Cullis, B. R. (1985) Effects of phosphate buffer capacity on yield response curvature and fertilizer requirements of wheat in relation to soil phosphate tests. Aust. J. Soil Res. 23, 417-427. Holford, I. C. R., Hird, C. and Lawrie, R. (1997) Effects of animal effluents on the phosphorus sorption characteristics of soils. Aust. J. Soil Res. 35, 365-73. Holmes, G. (2000) "Glenelg-Hopkins Nutrient Management Plan,". Glenelg-Hopkins Catchment Management Authority.

263 Holt, R. F., Timmons, D. R. and Latterell, J. J. (1970) Accumulation of phosphates in water. J. Agric. Food Chem. 18, 781-784. Holtan, H., Kamp-Nielsen, L. and Stuanes, A. O. (1988) Phosphorus in soil, water and sediment: an overview. Hydrobiologia 170, 19-34. Hooda, P. S., Moynagh, M. and Svoboda, I. F. (1996) A comparison of phosphate losses in drainage water from two different grassland systems. Soil Use Manage. 12, 224. Hooda, P. S., Rendell, A. R., Edwards, A. C., Withers, P. J. A., Aitken, M. N. and Truesdale, V. W. (2000) Relating soil phosphorus indices to potential phosphorus release to water. J. Environ. Qual. 29, 1166-1171. Hosomi, M. and Sudo, R. (1986) Simultaneous determination of total nitrogen and total phosphorus in freshwater samples using persulphate digestion. Int. J. Environ. Stud. 27, 267-275. Huang, C., Wells, L. K. and Norton, L. D. (1999) Sediment transport capacity and erosion processes: model concepts and reality. Earth Surf. Process. Landforms 24, 503-516. Huberty, A. and Diamond, D. (1996) "Determination of Phosphorus by Flow Injection Analysis Colorimetry QuickChem Method 31-115-01-3-A,". LaChat Instruments Inc., USA. Hudson, N. (1995) "Soil Conservation," New/Ed. Batsford, London. Hudson, N. W. (1993) "Field Measurement of Soil Erosion and Runoff," Food and Agriculture Organisation, Rome. Huffman, E. O. and Taylor, A. W. (1963) The behaviour of water-soluble phosphate in soils. Agric. Food Chem. 11, 182-187. Hughes, S., Reynolds, B., Bell, S. A. and Gardner, C. (2000) Simple phosphorus saturation index to estimate risk of dissolved P in runoff from arable soils. Soil Use Manage. 16, 206-210. Isbell, R. F. (1996) "The Australian Soils Classification," CSIRO Publishing, Australia. Jansson, M., Olsson, H. and Pettersson, K. (1988) Phosphates; origin, characteristcs and function in lakes. Hydrobiologia 170, 157-175. Jarvis, N. J. and Leeds-Harrison, P. B. (1987) Modelling water movement in drained clay soil. I. Description of the model, sample output and sensitivity analysis. J. Soil Sci. 38, 487- 498. Johnes, P. J. and Hodgkinson, R. A. (1998) Phosphorus loss from agricultural catchments: pathways and implications for management. Soil Use Manage. 14, 175-185. Johnston, A. E. and Poulton, P. R. (1997) Defining critical levels of available soil phosphorus for agricultural crops. In "Phosphorus Loss from Soil to Water" (H. Tunney, O. T. Carton, P. C. Brookes and A. E. Johnston, eds.), pp. 441-445. CAB Int. Press, England. Johnston, W. H., Garden, D. L., Rancic, A., Koen, T. B., Dassanayake, K. B., Langford, C. M., Ellis, N. J. S., Rab, M. A., Tuteja, N. K., Mitchell, M., Wadsworth, J., Dight, D.,

264 Holbrook, K., LeLievre, R. and McGeoch, S. M. (2003) The impact of pasture development and grazing on water-yielding catchments in the Murray-Darling Basin in south-eastern Australia. Aust. J. Exp. Agric. 43, 817-841. Jorgensen, M. A. (1985) Comparison of routine analyses by different laboratories. The New Zealand Statistician 20, 35-43. Joshua, W. D., Michalk, D. L., Curtis, I. H., Salt, M. and Osborne, G. J. (1998) The potential for contamination of soil and surface waters from sewage sludge (biosolids) in a sheep grazing study, Australia. Geoderma 84, 135-156. Kaine, G. and Niall, E. (2001) The adoption of sub-surface and on-off grazing by Victorian dairy farmers. In "Proceedings of the 10th Australian Agronomy Conference", Hobart, Australia. Kelly, G. L. A. (1999) Phosphorus Sorption in Herbert River Soils and the Implications for Water Quality. Master of Agricultural Science Thesis, The Univerisity of Melbourne, Melbourne. Khalid, R. A., Patrick, W. H. J. and DeLaune, R. D. (1977) Phosphorus sorption characteristics of flooded soils. Soil Sci. Soc. Am. J. 41, 305-310. Kirk, J. T. O. (1977) Attenuation of light in natural waters. Aust. J. Mar. Freshwater Res. 28, 497-508. Kirkby, C. A., Chittleborough, D. J., Smettem, K. R. J. and Cox, J. W. (1996) Water, phosphate, clay and DOC movement through a texture contrast soil. In "Australian and New Zealand National Soils Conference: Soil Science - Raising the Profile", pp. 141- 142. Australian Soil Science Society Inc., Melbourne, Australia. Kirkby, C. A., Smythe, L. J., Cox, J. W. and Chittleborough, D. J. (1997) Phosphorus movement down a toposequence from a landscape with texture contrast soils. Aust. J. Soil Res. 35, 399-417. Kleinman, P. J. A., Bryant, R. B. and Reid, W. S. (1999) Development of pedotransfer functions to quantify phosphorus saturation of agricultural soils. J. Environ. Qual. 28, 2026-2030. Kleinman, P. J. A., Bryant, R. B., Reid, W. S., Sharpley, A. N. and Pimental, D. (2000) Using soil phosphorus behaviour to identify environmental thresholds. Soil Sci. 165, 943-950. Knisel, W. G. (1980) "CREAMS: A Field Scale Model for Chemicals Runoff, and Erosion from Agricultural Management Systems. USDA Conservation Report No. 26,", Washington D.C. Kotak, B. G., Kenefick, S. L., Fritz, D. L., Rousseaux, C. G., Prepas, E. E. and Hrudey, S. T. (1993) Occurrence and toxicological evaluation of cyanobacterial toxins in Alberta Lakes and farm dugouts. Water Res. 27, 495-506.

265 Kung, K. J. S. and Donohue, S. V. (1991) Improved solute-sampling protocol in a sandy vadose zone ground-penetrating radar. Soil Sci. Soc. Am. J. 55, 1543-1545. Kuo, S. (1996) Phosphorus. In "Methods of Soil Analysis: Chemical Methods Part 3" (D. L. Sparks, ed.) SSSA Book Series No.5 pp. 891-893. Soil Science Society of America, Inc., Madison, USA. Kuykendall, H. A., Cabrera, M. L., Hoveland, C. S., McCann, M. A. and West, L. T. (1999) Stocking method effects on nutrient runoff from pastures fertilized with broiler litter. J. Environ. Qual. 28, 1886-1890. Lambert, M. G., Devantier, B. P., Nes, P. and Penny, P. E. (1985) Losses of nitrogen, phosphorus, and sediment in runoff from hill country under different fertiliser and grazing management regimes. N. Z. J. Agric. Res. 28, 371-379. Lang, R. D. (1979) The effect of ground cover on surface runoff from experimental plots. J. Soil Conserv. NSW 35, 108-114. Lawes Agricultural Trust (1997) Genstat 5 Release 4.1. IACR Rothamsted. Lee, K., Isenhart, T. M., Shultz, R. C. and Mickelson, S. T. (2000) Multispecies riparian buffers trap sediment and nutrients during rainfall simulations. J. Environ. Qual. 29, 1200- 1205. Lemunyon, J. L. and Gilbert, R. G. (1993) The concept and need for a phosphorus assessment tool. J. Prod. Agric. 6, 483-486. Lennox, S. D., Foy, R. H., Smith, R. V. and Jordan, C. (1997) Estimating the contribution from agriculture to the phosphorus load in surface water. In "Phosphorus Loss from Soil to Water" (H. Tunney, O. Carton, P. Brookes and A. Johnston, eds.), pp. 467. CAB International, UK. Lewis, D. C., Clarke, A. L. and Hall, W. B. (1981) Factors affecting the retention of phosphorus applied as superphosphate to the sandy soils in South-eastern Australia. Aust. J. Soil Res. 19, 167-74. Lewis, D. C. and Sale, P. W. G. (1993) Management of nutrients for pastures. In "Pasture Management Technology for the 21st Century" (D. R. Kemp, ed.), pp. 38-50. CSIRO Australia. Lookman, R., Jansen, K., Merckx, R. and Vlassak, K. (1995) Geostatistical assessment of the regional distribution of phosphate sorption capacity parameters (Feox and Alox) in northern Belgium. Geoderma 66, 285-296. Lopez-Hernandez, D. and Burnham, C. P. (1974) The covariance of phosphate sorption with other soil properties in some British and tropical soils. J. Soil Sci. 25, 196-206. MacLaren, G. S., Crawford, D. M., Brown, A. J. and Maheswaran, J. (1996) Temporal and spatial changes in soil chemistry across Victoria: (I) Available K and P. In "Australian

266 and New Zealand National Soils Conference: Soil Science - Raising the Profile", Vol. 3, pp. 149. Australian Soil Science Society Inc., Melbourne. Magesan, G. N., White, R. E., Scotter, D. R. and Bolan, N. S. (1994) Estimating leaching losses from sub-surface drained soils. Soil Use Manage. 10, 87-93. Malcolm, L. R., Sale, P. and Egan, A. (1996) "Agriculture in Australia: An Introduction," Oxford University Press, Melbourne. MathSoft Inc (1999) S-PLUS 2000 User Guide. Data Analysis Products Division Mathsoft, Seattle, WA. McCaskill, M. R. and Cayley, J. W. D. (2000) Soil audit of a long-term phosphate experiment in south-western Victoria: total phosphorus, sulfur, nitrogen, and major cations. Aust. J. Agric. Res. 51, 737-48. McCaskill, M. R., A. M. Ridley, A. Okom, R. E. White, M. H. Andrew, D. L. Michalk, A. Melland, Johnston, W. H. and Murphy, S. R. (2003) SGS Nutrient Theme: Environmental assessment of nutrient application to extensive pastures in the high rainfall zone of southern Australia. Aust. J. Exp. Agric. 43, 927-944. McClaren, C. (1997) Dry Sheep Equivalents for comparing different classes of livestock. In "Agriculture Notes, September 1997, AG0590", pp. 1-4. McColl, R. H. S., White, E. and Gibson, A. R. (1977) Phosphorus and nitrate run-off in hill pasture and forest catchments, Taita, New Zealand. N. Z. J. Mar. Fresh. Res. 11, 729- 44. McColl, R. H. S. and Gibson, A. R. (1979) Downslope movement of nutrients in hill pasture, Taita, New Zealand II. Effects of season, sheep grazing, and fertiliser. N. Z. J. Agric. Res. 22, 151-61. McColl, R. H. S., McQueen, D. J., Gibson, A. R. and Heine, J. C. (1985) Source areas of storm runoff in a pasture. J. Hydrol. 24, 1-19. McComb, A. J. and Davis, J. A. (1993) Eutrophic waters of southwestern Australia. Fert. Res. 36, 105-114. McDonald, R. C., Isbell, R. F., Speight, J. G., Walker, J. and Hopkins, M. S. (1990) "Australian Soil and Land Survey Handbook," 2nd Edition/Ed. Inkata Press, Melbourne, Australia. McDowell, R. W. and Condron, L. M. (2000) Chemical nature and potential mobility of phosphorus in fertilized grassland soils. Nutr. Cycl. Agroecosyst. 57, 225-233. McDowell, R. W. and Sharpley, A. N. (2001) Approximating phosphorus release from soils to surface runoff and subsurface drainage. J. Environ. Qual. 30, 508-520. McDowell, R. W., Sharpley, A., Brookes, P. C. and Poulton, P. (2001) Relationship between soil test phosphorus and phosphorus release to solution. Soil Sci. 166, 137-149.

267 McDowell, R. W., Drewry, J. J., Paton, R. J., Carey, P. L. and Monaghan, R. M. (2003) Influence of soil treading on sediment and phosphorus loss in overland flow. Aust. J. Soil Res. 41. McFarland, A. M. S. and Hauck, L. M. (1999) Relating agricultural land uses to in-stream stormwater quality. J. Environ. Qual. 28, 836-844. McFarlane, D. J. and Cox, J. W. (1992) Management of excess water in duplex soils. Aust. J. Exp. Agric. 32, 857-64. McIvor, J. G., Williams, J. and Gardener, C. J. (1995) Pasture management influences runoff and soil movement in the semi-arid tropics. Aust. J. Exp. Agric. 35, 55-65. McKergow, L. A., Weaver, D. M., Prosser, I. P., Grayson, R. B. and Reed, A. E. G. (2003) Before and after riparian management: sediment and nutrient exports from a small agricultural catchment, Western Australia. J. Hydrol. 270, 253-272. Monaghan, R. M., Morton, J. D., McDowell, R. W., Drewry, J. J. and Thorrold, B. S. (2003) The development of environmental best practices for intensive dairying. Proc. N. Z. Soc. Anim. Prod. 63, 3-6. Montgomery, J. (2000) Fertiliser industry incentives to keep nutrients on farm. In "Smart nutrient management for pastures". Western District Branch Grasslands Society of Victoria, Pastoral and Veterinary Institute, Hamilton, Victoria, Australia. Moody, P. W., Haydon, G. F. and Dickson, T. (1983) Mineral nutrition of soybeans grown in the region of south-eastern Queensland 2. Prediction of grain yield response to phosphorus with soil tests. Aust. J. Exp. Agric. 23, 38-42. Moody, P. W. and Bolland, M. D. A. (1999) Phosphorus. In "Soil Analysis; An Interpretation Manual" (K. I. Peverill, L. A. Sparrow and D. J. Reuter, eds.). CSIRO, Australia. Moore, I. D., Grayson, R. B. and Ladson, A. R. (1991) Digital terrain analysis modelling: A review of hydrological, geomorphological and biological applications. Hydrol. Proc. 5, 3-30. Morse, D., Head, H. H., Wilcox, C. J., van Hern, H. H., Hissem, C. D. and Harris Jr, B. (1992) Effects of concentration of dietary phosphorus on amount and route of excretion. J. Dairy Sci. 75, 3039-3045. Morton, J. D. and Baird, D. B. (1990) Spatial distribution of dung patches under sheep grazing. N. Z. J. Agric. Res. 33, 285-294. Mosley, M. P. (1979) Streamflow generation in a forested watershed, New Zealand. Water Resour. Res. 15, 795-806. Mozafari, M. and Sims, J. T. (1994) Phosphorus availability and sorption in an Atlantic coastal plain watershed dominated by animal-based agriculture. Soil Sci. 157, 97-107. Murphy, J. and Riley, J. P. (1962) A modified single solution method for the determination of phosphate in natural waters. Anal. Chim. Acta 27, 31-36.

268 Murphy, S. R. and Lodge, G. M. (2004) Surface soil water dynamics in pastures in northern New South Wales. 1. Use of electrical resistance sensors. Aust. J. Exp. Agric. 44, in press. Nagpal, N. K. (1982) Comparison among and evaluation of ceramic porous cup soil water samplers for nutrient transport studies. Can. J. Soil Sci. 62, 685-694. Nair, V. D., Graetz, D. A. and Reddy, K. R. (1998) Dairy manure influences on phosphorus retention capacity of Spodosols. J. Environ. Qual. 27, 522-527. Nash, D. and Murdoch, C. (1996) Phosphorus leaching from cattle dung and fertiliser in a kraznozem. In "Australian and New Zealand National Soils Conference: Soil Science - Raising the Profile", Vol. 3, pp. 187-188. Australian Society of Soil Science Incorporated, Melbourne, Australia. Nash, D. (1997) "Alkali Persulfate Digestion for the Lachat Flow Injection Analytical System used for TP and TN analyses,". Ellinbank Dairy Research Institute. Nash, D. and Murdoch, C. (1997) Phosphorus in runoff from a fertile dairy pasture. Aust. J. Soil Res. 35, 419-429. Nash, D. and Halliwell, D. (1999) Fertilizers and phosphorus loss from productive grazing systems. Aust. J. Soil Res. 37, 403-429. Nash, D., Hannah, M., Halliwell, D. and Murdoch, C. (2000) Factors affecting phosphorus export from a pasture-based grazing system. J. Environ. Qual. 29, 1160-1166. Nash, D. M. and Halliwell, D. J. (2000) Tracing phosphorus transferred from grazing land to water. Water Res. 34, 1975-1985. Nelson, P. N., Cotsaris, E. and Oades, J. M. (1996) Nitrogen, phosphorus and organic carbon in streams draining two grazed catchments. J. Environ. Qual. 25, 1221-1229. Nexhip, K. J. and Austin, N. R. (1998) Defining key nutrient reduction practices for irrigated dairy pastures. In "Irrigation Association of Australia National Conference", Brisbane, Australia. Nguyen, M. L. and Goh, K. M. (1992) Nutrient cycling and losses based on a mass-balance model in grazed pastures receiving long-term superphosphate applications in New Zealand 1. Phosphorus. J. Agric. Sci. 119, 89-106. O'Connor, P. W. and Syers, J. K. (1975) Comparison of methods for the determination of total phosphorus in water containing particulate material. J. Environ. Qual. 4, 347-350. O'Hara, C. (1996) The cycling of phosphorus in a grazed pasture. Honours thesis for the Bachelor of Science, Australian National University, Canberra. Okom, A. E. A. (1998) Estimating soil hydraulic properties and the solute transport volume using surrogate variables. PhD, the University of Melbourne. Olness, A. E., Smith, S. J., Rhoades, E. D. and Menzel, R. G. (1975) Nutrient and sediment discharge from agricultural watersheds in Oklahoma. J. Environ. Qual. 4, 331-336.

269 O'Loughlin, E. M. (1981) Saturation regions in catchments and their relations to soil and topographic properties. J. Hydrol. 53, 229-246. O'Loughlin, E. M. (1986) Prediction of surface saturation zones in natural catchments by topographic analysis. Water Resour. Res. 22, 794-804. Oloya, T. O. and Logan, T. J. (1980) Phosphate desorption from soils and sediments with varying levels of extractable phosphate. J. Environ. Qual. 9, 526-531. Olsen, S. R., Cole, C. V., Watanabe, F. S. and Dean, L. A. (1954) "Estimation of available phosphorus in soils by extraction with sodium bicarbonate,". US Government Printing Office, Washington. Olsen, S. R. and Watanabe, F. S. (1957) A method to determine a phosphorus adsorption maximum of soils as measured by the Langmuir isotherm. Soil Sci. Soc. Am. Proc. 21, 144-149. Ozanne, P. G., Kirton, D. J. and Shaw, T. C. (1961) The loss of phosphorus from sandy soils. Aust. J. Agric. Res. 12, 409-423. Ozanne, P. G. and Shaw, T. C. (1967) Phosphate sorption by soils as a measure of the phosphate requirement for pasture growth. Aust. J. Agric. Res. 18, 601-12. Patterson, A. (1992) The Economic Significance of Pastures - Findings from the South West Victorian Farm Monitor Farm project. In "Realising the potential of our grazing land - Proceedings of 1992 Pasture Specialists Conference" (G. Ward, K. Reed and K. Bishop, eds.), Hamilton. Paulter, M. C. and Sims, J. T. (2000) Relationship between soil test phosphorus, soluble phosphorus, and phosphorus saturation in Delaware soils. Soil Sci. Soc. Am. J. 64, 765- 773. Pearce, R. A., Trlica, M. J., Leininger, W. C., Mergen, D. E. and Frasier, G. (1998) Sediment movement through riparian vegetation under simulated rainfall and overland flow. J. Range Manage. 51, 301-308. Perrott, K. W. and Sarathchandra, S. U. (1989) Phosphorus in the microbial biomass of New Zealand soils under established pasture. N. Z. J. Agric. Res. 32, 409-413. Perrott, K. W., Sarathchandra, S. U. and Waller, J. E. (1990) Seasonal storage and release of phosphorus and potassium by organic matter and microbial biomass in a high-producing pastoral soil. Soil Biol. Biochem. 28, 593-608. Perroux, K. M. and White, I. (1988) Designs of the disc permeameter. Soil Sci. Soc. Am. J. 52, 1205-1215. Philip, J. R. (1986) Linearized unsteady multidimensional infiltration. Water Resour. Res. 22, 1717-1727.

270 Pierzynski, G. M., ed. (2000) "Methods of Phosphorus Analysis for Soils, Sediments, Residuals and Waters" Southern Cooperative Series Bulletin No.#396 North Carolina State University. Pilgrim, D. H., Huff, D. D. and Doak-Steele, T. (1978) A field evaluation of subsurface and surface runoff II Runoff processes. J. Hydrol. 38, 319-341. Pionke, H. B., Gburek, W. J., Sharpley, A. N. and Schnabel, R. R. (1996) Flow and nutrient export patterns for an agricultural hill-land watershed. Water Resour. Res. 32, 1795- 1804. Piper, C. S. (1950) Wet digestion with sulfuric acid and nitric acids. In "Soil and Plant Analysis. A laboratory manual of methods for the examination of soils and the determination of inorganic constituents of plants" pp. 368. A monograph from The Waite Agricultural Research Institute. The University of Adelaide, Adelaide, Australia. Pote, D. H., Daniel, T. C., Sharpley, A. N., Moore, P. A. J., Edwards, A. C. and Nichols, D. J. (1996) Relating extractable soil phosphorus to phosphorus losses in runoff. Soil Sci. Soc. Am. J. 60, 855-859. Pote, D. H., Daniel, T. C., Nichols, D. J., P.A. Moore, J., Miller, D. M. and Edwards, D. R. (1999a) Seasonal and soil-drying effects on runoff phosphorus relationships to soil phosphorus. Soil Sci. Soc. Am. J. 63, 1006-1012. Pote, D. H., Daniel, T. C., Nichols, D. J., Sharpley, A. N., Moore, P. A. J., Miller, D. M. and Edwards, D. R. (1999b) Relationship between phosphorus levels in three Ultisols and phosphorus concentrations in runoff. J. Environ. Qual. 28, 170-175. Prairie, Y. T. and Kalff, J. (1986) Effect of catchment size on phosphorus export. Water Resour. Bull. 22, 465-70. Preedy, N., McTiernan, K., Matthews, R., Heathwaite, L. and Haygarth, P. (2001) Rapid incidental phosphorus transfers from grassland. J. Environ. Qual. 30, 2105-2112. Priestley, C. H. B. and Taylor, R. J. (1972) On the assessment of surface heat flux and evaporation using large-scale parameters. Monthly Weather Review 100, 81-92. Quinton, J. N., Catt, J. A. and Hess, T. M. (2001) The selective removal of phosphorus from soil: Is event size important? J. Environ. Qual. 30, 538-545. Randall, G. W., Iragavarapu, T. K. and Schmitt, M. Z. (2000) Nutrient losses in subsurface drainage water from dairy manure and urea applied for corn. J. Environ. Qual. 29, 1244-1252. Rayment, G. E. and Higginson, F. R. (1992) "Australian Laboratory Handbook of Soil and Water Chemical Methods," Inkata Press, Melbourne, Sydney. Redding, M. R. (2001) Pig effluent-P application can increase the risk of P transport: two case studies. Aust. J. Soil Res. 39, 161-174.

271 Reddy, G. Y., McLean, E. O., Hoyt, G. D. and Logan, T. J. (1978) Effects of soil, cover crop, and nutrient source on amounts and forms of phosphorus movement under simulated rainfall conditions. J. Environ. Qual. 7, 50-54. Redfield, A. C. (1958) The biological control of chemical factors in the environment. Am. Sci. 46, 1-221. Reuter, D., Peverill, K., Cox, J. and Williams, J. (1996) Nutrients and the environment. In "Proceedings of the Victorian Grasslands Society Conference", pp. 26-. Grasslands Society of Victoria, Churchill Campus, Murdoch University. Reuter, D. J. and Robinson, J. B., eds. (1997) "Plant Analysis; an Interpretation Manual" 2nd edn CSIRO Publishing, Australia. Rhoton (1979) J. Environ. Qual. 8, 424-427. Ridley, A. M., White, R. E., Helyar, K. R., Morrison, G. R., Heng, L. K. and Fisher, R. (2001) Nitrate leaching loss under annual and perennial pastures with and without lime on a duplex (texture contrast) soil in humid southeastern Australia. Eur. J. Soil Sci. 52, 237- 252. Ridley, A. M., Christy, B. P., White, R. E., McLean, J. and Green, R. (2003) North-East Victoria SGS National Experiment Site: sustainability and production implications of grazing systems at different levels of intensity. Aust. J. Exp. Agric. 43, 799-815. Ritchie, G. S. P. and Weaver, D. M. (1993) Phosphorus retention and release from sandy soils of the Peel-Harvey catchment. Fert. Res. 36, 115-122. Rixon, A. J. and Zorin, M. (1978) Transformations of nitrogen and phosphorus in sheep faeces located in saltbush rangeland and on irrigated pasture. Soil Biol. Biochem. 10, 347-354. Romkens, M. J. M. and Nelson, D. W. (1974) Phosphorus relationships in runoff from fertilized soils. J. Environ. Qual. 3, 10-13. Ron Vaz, M. d., Edwards, A. C., Shand, C. A. and Cresser, M. S. (1993) Phosphorus fractions in soil solution: Influence of soil acidity and fertiliser additions. Plant Soil 148, 175- 183. Rowarth, J. S., Gillingham, A. G., Tillman, R. W. and Syers, J. K. (1988) Effects of season and fertiliser rate on phosphorus concentrations in pasture and sheep faeces in hill country. N. Z. J. Agric. Res. 31, 187-193. Russell, J. S. (1960a) Soil fertility changes in the long term experimental plots at Kybybolite, South Australia. I. Changes in pH, total nitrogen, organic carbon and bulk density. Aust. J. Agric. Res. 11, 902-26. Russell, J. S. (1960b) Soil fertility changes in the long-term experimental plots at Kybybolite, South Australia II. Changes in phosphorus. Aust. J. Agric. Res. 11, 927-947. Russell, J. S. (1986) Improved pastures. In "Australian Soils: The Human Impact" (J. Russell and R. Isbell, eds.), pp. 374-96. University of Queensland Press: St Lucia.

272 Ryan, J., Curtin, D. and Cheema, M. A. (1985) Significance of iron oxides and calcium carbonate particle size in phosphate sorption by calcareous soils. Soil Sci. Soc. Am. J. 48, 74-76. Ryan, M. and Noonan, D. (1995) Using dyes to measure flow patterns of water in soil. Farm and Food 5, 9-11. Ryden, J. C., Syers, J. K. and Harris, R. F. (1973) Phosphorus in runoff and streams. Adv. Agron. 25, 1-45. Ryden, J. C. and Syers, J. K. (1977) Desorption and isotopic exchange relationships of phosphate sorbed by soils and hydrous ferric oxide gel. J. Soil Sci. 28, 596-609. Ryden, J. C., Syers, J. K. and McLaughlin, J. R. (1977) Effects of ionic strength on chemisorption and potential-determining sorption of phosphate by soils. J. Soil Sci. 28, 62-71. Saggar, S., Mackay, A. D., Hedley, M. J., Lambert, M. G. and Clark, D. A. (1990) A nutrient- transfer model to explain the fate of phosphorus and sulphur in a grazed hill-country pasture. Agric. Ecosyt. Environ. 30, 295-315. Sale, P. W. G. and Blair, G. J. (1989) Low solubility phosphate fertilisers for pastures. Agric. Sci. May, 34-39. Sale, P. W. G. and Blair, G. J. (1997) Fertilisers and pasture nutrition. In "Pasture Production and Management" (J. V. Lovett and J. M. Scott, eds.). Inkata Press, Australia. Salvia-Castellvi, M., Mosig, J., Iffly, J. F., Vander Borght, P. and Hoffmann, L. (2001) Dynamics of dissolved and particulate phosphorus during storm events in two rural river basins and intwo headwater streams. In "Connecting Phosphorus Transfer from Agriculture to Impacts in Surface Waters" (P. M. Haygarth, L. M. Condron, P. J. Butler and J. S. Chisholm, eds.), pp. 83. Institute of Grasslands and Environmental Research, Plymouth University, England. Sartz, R. S. and Tolsted, D. N. (1974) Effect of grazing on runoff from two small watersheds in southwestern Wisconsin. Water Resour. Res. 10, 354-356. Saunders, W. M. H. (1965) Phosphate retention by New Zealand soils and its relationship to free sesquioxides, organic matter and other soil properties. N. Z. J. Agric. Res. 8, 30-57. Saunders, W. M. H. (1984) Mineral composition of soil and pasture from areas of grazed paddocks, affected and unaffected by dung and urine. N. Z. J. Agric. Res. 27, 405-412. Schefe, C. R. (1999) "Runoff and sediment removal under different pasture management regimes,". Institute of Land and Food Resources, The University of Melbourne. Scheinost, A. C. and Schwertmann, U. (1995) Predicting phosphate adsorption-desorption in a soilscape. Soil Sci. Soc. Am. J. 59, 1575-1580. Schepers, J. S. and Francis, D. D. (1982) Chemical water quality of runoff from grazing land in Nebraska: 1. Influence of grazing livestock. J. Environ. Qual. 11, 351-354.

273 Scoging, H. (1989) Runoff generation and sediment mobilisation by water. In "Arid Zone Geomorphology" (D. Thomas, ed.), pp. 87-116. Belhaven Press, London. Scotter, D. R., Clothier, B. E. and Turner, M. A. (1979) The soil water balance in a fragiaqualf and its effect on pasture growth in Central New Zealand. Aust. J. Soil Res. 17, 455-465. Sharpley, A. N. and Syers, J. K. (1976) Phosphorus transport in surface run-off as influenced by fertiliser and grazing cattle. N. Z. J. Sci. 19, 277-282. Sharpley, A. N., Tillman, R. W. and Syers, J. K. (1977) Use of laboratory extraction data to predict losses of dissolved inorganic phosphate in surface runoff and tile drainage. J. Environ. Qual. 6, 33-36. Sharpley, A. N., Syers, J. K. and Tillman, R. W. (1978) An improved soil-sampling procedure for the prediction of dissolved inorganic phosphate concentrations in surface runoff from pasture. J. Environ. Qual. 7, 455-456. Sharpley, A. N. and Syers, J. K. (1979) Effect of aerial topdressing with superphosphate on the loss of phosphate from a pasture catchment. N. Z. J. Agric. Res. 22, 273-7. Sharpley, A. N. (1980) The enrichment of soil phosphorus in runoff sediments. J. Environ. Qual. 9, 521-526. Sharpley, A. N. (1981) The contribution of phosphorus leached from crop canopy to losses in surface runoff. J. Environ. Qual. 10, 160-165. Sharpley, A. N., Ahuja, L. R. and Menzel, R. G. (1981a) The release of soil phosphorus to runoff in relation to the kinetics of desorption. J. Environ. Qual. 10, 387-391. Sharpley, A. N., Ahuja, L. R., Yamamoto, M. and Menzel, R. G. (1981b) The kinetics of phosphorus desorption from soil. Soil Sci. Soc. Am. J. 45, 493-496. Sharpley, A. N., Menzel, R. G., Smith, S. J., Rhoades, E. D. and Olness, A. E. (1981c) The sorption of soluble phosphorus by soil material during transport in runoff from cropped and grassed watersheds. J. Environ. Qual. 10, 211-215. Sharpley, A. N., Smith, S. J. and Menzel, R. G. (1982) Prediction of Phosphorus losses in Runoff from Southern Plains Watersheds. J. Environ. Qual. 11, 247-251. Sharpley, A. N. (1983) Effect of soil properties on the kinetics of phosphorus desorption. Soil Sci. Soc. Am. J. 47, 462-467. Sharpley, A. N. (1985a) Depth of surface soil-runoff interaction as affected by rainfall, soil slope, and management. Soil Sci. Soc. Am. J. 49, 1010-1015. Sharpley, A. N. (1985b) The selective erosion of plant nutrients in runoff. Soil Sci. Soc. Am. J. 49, 1527-1534. Sharpley, A. N., Smith, S. J., Berg, W. A. and Williams, J. R. (1985) Nutrient runoff losses as predicted by annual and monthly soil sampling. J. Environ. Qual. 14, 354-360. Sharpley, A. N. and Menzel, R. G. (1987) The impact of soil and fertilizer phosphorus on the environment. Adv. Agron. 41, 297-324.

274 Sharpley, A. N. and Smith, S. J. (1989) Prediction of soluble phosphorus transport in agricultural runoff. J. Environ. Qual. 18, 313-316. Sharpley, A. N., Troeger, W. W. and Smith, S. J. (1991) The measurement of bioavailable phosphorus in agricultural runoff. J. Environ. Qual. 20, 235-238. Sharpley, A. N. (1993a) An innovative approach to estimate bioavailable phosphorus in agricultural runoff using iron oxide-impregnated paper. J. Environ. Qual. 22, 597-601. Sharpley, A. N. (1993b) Assessing phosphorus bioavailability in agricultural soils and runoff. Fert. Res. 36, 259-272. Sharpley, A. N., Chapra, S. C., Wedepohl, R., Sims, J. T., Daniel, T. C. and Reddy, K. R. (1994) Managing agricultural phosphorus for protection of surface waters: Issues and options. J. Environ. Qual. 23, 437-451. Sharpley, A. N. and Halvorson, A. D. (1994) Management of Soil Phosphorus. In "Soil Processes and Water Quality" (R. Lal and B. Stewart, eds.) Advances in Soil Science. CRC Press. Sharpley, A. N. (1995) Dependence of runoff phosphorus on extractable soil phosphorus. J. Environ. Qual. 24, 920-926. Sharpley, A. N. (1996) Availability of residual phosphorus in manured soils. Soil Sci. Soc. Am. J. 60, 1459-1466. Sharpley, A. N., Daniel, T. C., Sims, J. T. and Pote, D. H. (1996) Determining environmentally sound soil phosphorus levels. J. Soil Water Conserv. 51, 160-166. Sharpley, A. N. and Rekolainen, S. (1997) Phosphorus in agriculture and its environmental implications. In "Phosphorus Loss from Soil to Water" (H. Tunney, O. T. Carton, P. C. Brookes and A. E. Johnston, eds.), pp. 1-54. CAB International, England. Sharpley, A. N., Daniel, T., Sims, T., Lemunyon, J., Stevens, R. and Parry, R. (1999) "Agricultural phosphorus and eutrophication,". USDA Agricultural Research Service. Sharpley, A. N. and Moyer, B. (2000) Phosphorus forms in manure and compost and their release during the simulated rainfall. J. Env. Qual. 29, 1462-1469. Sharpley, A. N., McDowell, R. W. and Kleinman, P. J. A. (2001a) Phosphorus loss from land to water: integrating agricultural and environmental management. Plant Soil 237, 287-307. Sharpley, A. N., McDowell, R. W., Weld, J. L. and Kleinman, J. A. (2001b) Assessing site vulnerability to phosphorus loss in an agricultural watershed. J. Environ. Qual. 30, 2026-2036. Sherwood, J., Magilton, C. and Rouse, A. (1998) "The Glenelg River: Nutrients and Estuarine Hydrodynamics,". Department of Natural Resources and Environment, Victoria. Sibbesen, E. and Sharpley, A. N. (1997) Setting and justifying upper critical limits for P in soils. In "Phosphorus Loss from Soil to Water" (H. Tunney, O. Carton, P. Brookes and A. Johnston, eds.), pp. 151-176. CAB International, England.

275 Simard, R. R., Cluis, D., Gangbazo, G. and Pesant, A. R. (1994) Phosphorus sorption and desorption indices in soil. Commun. Soil Sci. Plant Anal. 25, 1483-1494. Simard, R. R., Beauchemin, S. and Haygarth, P. M. (2000) Potential for preferential pathways of phosphorus transport. J. Environ. Qual. 29, 97-105. Simpson, C. J. and Woodfull, C. J. (1994) New field evidence resolving the relationship between the Grampians Group and the Rocklands Rhyolite, western Victoria. Aust. J. Earth Sci. 41, 621-624. Sims, J. T., Edwards, A. C., Schoumans, O. F. and Simard, R. R. (2000) Integrating soil phosphorus testing in environmentally based agricultural management practices. J. Environ. Qual. 29, 60-71. Singh, B. and Gilkes, R. J. (1991) Phosphorus sorption in relation to soil properties for the major soil types of South-western Australia. Aust. J. Soil Res. 29, 603-18. Sissing, H. A. (1971) Analytical procedure of the Pw-method, used for the assessment of phosphate status of arable soils in the Netherlands. Plant Soil 34, 483-486. Sklash, M. G., Stewart, M. K. and Pearce, A. J. (1986) Storm runoff generation in humid headwater catchments 2. A case study of hillslope and low-order stream response. Water Resour. Res. 22, 1273-1282. Smettem, K. R. J. (1987) Characterisation of water entry into a soil with a contrasting textural class: Spatial variability of infiltration parameters and influence of macroporosity. Soil Sci. 144, 167-174. Smettem, K. R. J. and Clothier, B. E. (1989) Measuring unsaturated sorptivity and hydraulic conductivity using multiple disc permeameters. J. Soil Sci. 40, 563-568. Smettem, K. R. J., Chittleborough, D. J., Richards, B. G. and Leaney, F. W. (1991) The influence of macropores on runoff generation from a hillslope soil with contrasting textural class. J. Hydrol. 122, 235-252. Smith, A. N. (1983) Australian dependence on phosphorus. In "Phosphorus in Australia" (A. B. Costin and C. H. Williams, eds.), Vol. Monograph 8 pp. 70-91. Centre for Resource and Environmental Studies, Australian National University, Canberra, ACT. Smith, E. P. and McCormick, P. V. (2001) Long-term relationship between phosphorus inputs and wetland phosphorus concentrations in a northern everglades marsh. Environ. Monit. Assess. 68, 153-176. Smith, K. A., Jackson, D. R. and Withers, P. J. A. (2001a) Nutrient losses by surface run-off following the application of organic manures to arable land. 2. Phosphorus. Environ. Pollut. 112, 53-60. Smith, K. A., Brewer, A. J., Crabb, J. and Dauven, A. (2001b) A survey of the production and use of animal manures in England and Wales. III. Cattle manures. Soil Use Manage. 17, 77-87.

276 Snaydon, R. W. (1981) The ecology of grazed pastures. In "Grazing Animals; World Animal Science" (F. Morley, ed.), pp. 13-31. Elsevier, The Netherlands. Soil Survey Staff (1996) "Keys to Soil Taxonomy," 7th edition/Ed. USDA, US Government Printing Office, Washington DC, USA. Sonzogni, W. C., Chapra, S. C., Armstrong, D. E. and Logan, T. J. (1982) Bioavailability of phosphorus inputs to lakes. J. Environ. Qual. 11, 555-563. Spier, T. W. and Cowling, J. C. (1991) Phosphatase activities of pasture plants and soils: relationship with plant productivity and soil P fertility indices. Biol. Fertil. Soils 12, 189-194. Sposito, G. (1989) "The Chemistry of Soils," Oxford Uni Press, Oxford. Srinivasan, M. S., Wittman, M. A., Hamlett, J. M. and Gburek, W. J. (2000) Surface and subsurface sensors to record variable runoff generation areas. Trans. ASAE 43, 651-660. Srinivasan, M. S., Gburek, W. J. and Hamlett, J. M. (2002) Dynamics of stormflow generation - A hillslope-scale field study in east-central Pennsylvania, USA. Hydrol. Proc. 16, 649- 665. Stevens, D. P., Cox, J. W. and Chittleborough, D. J. (1997) "Analysis and storage of water samples for determining molybdate reactive phosphorus," Rep. No. 2/97. Co-operative Research Centre for Soil and Land Management, Adelaide. Stevens, D. P., Cox, J. W. and Chittleborough, D. J. (1999) Pathways of phosphorus, nitrogen and carbon movement over and through texturally differentiated soils, South Australia. Aust. J. Soil Res. 37, 679-93. Storm, D. E., Dillaha III, T. A., Mostaghimi, S. and Shanholtz, V. O. (1988) Modeling phosphorus transport in surface runoff. Trans. ASAE 31, 117-127. Stout, W. L., Sharpley, A. N. and Pionke, H. (1998) Reducing soil phosphorus solubility with coal combustion by-products. J. Environ. Qual. 27, 111-118. Summers, R. N., Smirk, D. D. and Karafilis, D. (1996) Phosphorus retention and leachates from sandy soil amended with bauxite residue (red mud). Aust. J. Soil Res. 34, 555-567. Syers, J. K., Browman, M. G., Smillie, G. W. and Corey, R. B. (1973) Phosphate sorption by soils evaluated by the Langmuir adsorption equation. Soil Sci. Soc. Am. J. 37, 358-363. Tabatabai, M. A. and Laflen, J. M. (1976) Nitrogen and sulfur content and pH of precipitation in Iowa. J. Environ. Qual. 5, 108-112. Tate, K. R. (1985) Soil Phosphorus. In "Soil Organic Matter and Biological Activity" (D. Vaughan and R. Malcolm, eds.) Developments in Plant and Soil Science, Vol. 16 pp. 330-377. Martinus Nijhoff/Dr W Junk Publishers, The Netherlands. Tate, K. W., Atwill, E. R., McDougald, N. K., George, M. R. and Witt, D. (2000) A method for estimating cattle fecal loading on rangeland watersheds. J. Range Manage. 53, 506-510.

277 Taylor, A. W., Gurney, E. L. and Moreno, E. C. (1964) Precipitation of phoshpate from calcium phosphate solutions by iron oxide and aluminium hydroxide. Soil Sci. Soc. Am. Proc. 28, 49-52. Taylor, A. W. and Kunishi, H. M. (1971) Phosphate equilibria on stream sediment and soil in a watershed draining an agricultural region. J. Agric. Food Chem. 19, 827-831. Taylor, R. M. and Schwertmann, U. (1974) The association of phosphorus with iron in ferruginous soil concretions. Aust. J. Soil Res. 12, 133-45. Tham, S. H. (1983) "Nutrient runoff from agricultural lands,". Animal Research Institute, Werribee. Thomas, G. W. and Crutchfield, J. D. (1974) Nitrate-nitrogen and phosphorus contents of streams draining small agricultural watersheds in Kentucky. J. Environ. Qual. 3, 46-49. Thompson, J. M., Williamson, D. R., Fitzpatrick, R. W. and Davies, P. J. (1992) "Piezometer design, installation and monitoring techniques to study ground and perched water tables," Rep. No. 28/1992. Division of Soils, CSIRO. Timmons, D. R., Holt, R. F. and Latterell, J. J. (1970) Leaching of crop residues as a source of nutrients in surface runoff water. Water Resour. Res. 6, 1367-1375. Toreu, B. N., Thomas, F. G. and Gillman, G. P. (1988) Phosphate-sorption characteristics of soils of the North Queenslnad coastal region. Aust. J. Soil Res. 26, 465-77. Trompf, J. (2001) Farm profitability and participation in the Grassland's Productivity Program. In "Farm Monitor Project; Summary of Results 2000-2001" (L. Beattie and J. Hamilton, eds.). Department of Natural Resources and Environment, Hamilton. Truman, C. C., Wauchope, R. D., Summer, H. R., Davis, J. G., Gascho, G. J., Hook, J. E., Chandler, L. D. and Johnson, A. W. (2001) Slope length effects on runoff and sediment delivery. J. Soil Water Conserv. 56, 249-256. Tunney, H., Carton, O. T., Brookes, P. C. and Johnston, A. E., eds. (1997) "Phosphorus Loss from Soil to Water" 467. CAB International, England. Tunney, H., Kurz, I., Morgan, G., Kiely, G., Coxon, C. and Daly, K. (2000a) Soil test phosphorus levels and phosphorus loss to water from grassland. In "Agricultural Research Forum", pp. 123-124, Ireland. Tunney, H., Coulter, B., Daly, K., Kurz, I., Coxon, C., Jeffrey, D., Mills, P., Kiely, G. and Morgan, G. (2000b) "Quantification of phosphorus loss from soil to water. Final Report and Literature Review R&D Report Series No.6,". Envrionment Protection Agency, Johnstown Castle, Ireland. Tunney, H., Foy, R. H. and Haygarth, P. M. (2001) Soil test phosphorus and measured concentrations of phosphorus in water from grassland. In "Connecting Phosphorus Transfers from Agriculture to Impacts in Surface Waters" (P. Haygarth, L. Condron, P.

278 Butler and J. Chisholm, eds.), pp. 61. Institute of Grasslands and Environmental Research, Plymouth Univeristy, UK. Turner, B. L. and Haygarth, P. M. (2000a) Phosphorus forms and concentrations in leachate under four grassland soil types. Soil Sci. Soc. Am. J. 64, 1090-1099. Turner, B. L. and Haygarth, P. M. (2000b) Organic phosphorus characterisation by phosphatase hydrolysable phosphorus techniques: Application to soil extracts and runoff waters. In "Phosphatases in the Environment" (B. A. Whitton and I. Hernandez, eds.). Kleuwer Academic Press, The Netherlands. Turoczy, N. (1999) "A preliminary assessment of the nutrient status of selected farm dams in the Glenelg River catchment,". Report for the Department of Natural Resources and Environment, Victoria. Turtola, E. and Yli-Halla, M. (1999) Fate of phosphorus applied in slurry and mineral fertiliser: accumulation in soil and release into surface runoff water. Nutr. Cycl. Agroecosyst. 55, 165-174. Uusitalo, R., Turtola, E., Kauppila, T. and Lilja, T. (2001) Particulate phosphorus and sediment in surface runoff and drainflow from clayey soils. J. Environ. Qual. 30, 589-595. Van Der Zee, S. E. A. T. M. and Van Riemsdijk, W. H. (1988) Model for long-term phosphate reaction kinetics in soil. J. Environ. Qual. 17, 35-41. Veith, J. A. and Sposito, G. (1977) On the use of the Langmuir equation in the interpretation of 'adsorption' phenomena. Soil Sci. Soc. Am. J. 41, 697-702. Wagg, C. (1997) "A summary of water quality in the Glenelg catchment,". State Government of Victoria. Wagg, C. (1999a) "Development of the Catchment Management Support System (CMSS) in the Glenelg Hopkins CMA,". Department of Natural Resources and Environment, Victoria. Wagg, C. (1999b) "Draft Glenelg Nutrient Management Plan,". Department of Natural Resources and Environment, Victoria. Walker, T. W., Tharpa, B. K. and Adams, A. F. R. (1959) Studies on soil organic matter: 3 Accumulation of carbon, sulfur, organic and total phosphorus in improved grassland soils. Soil Sci. 87, 135-140. Wallach, R. and Shabtai, R. (1993) Surface runoff contamination by chemicals initially incorporated below the soil surface. Water Resour. Res. 29, 697-704. Wallbrink, P. J., Olley, J. M., Murray, A. S. and Olive, L. J. (1996) The contribution of channel banks and gully walls to total phosphorus loads in the Murrumbidgee River. In "First National Conference on Stream Management in Australia", pp. 1-6. Cooperative Research Centre for Catchment Hydrology, Merrijig, Australia. Waller, R. A., Sale, P. W. G., Saul, G. R. and Kearney, G. A. (2001a) Tactical versus continuous stocking in perennial ryegrass-subterranean clover pastures grazed by sheep

279 in south-western Victoria 1. Stocking rates and herbage production. Aust. J. Exp. Agric. 41, 1099-1108. Waller, R. A., Sale, P. W. G., Saul, G. R. and Kearney, G. A. (2001b) Tactical versus continuous stocking in perennial ryegrass-subterranean clover pastures grazed by sheep in south-western Victoria 2. Ryegrass persistence and botanical composition. Aust. J. Agric. Res. 41, 1109-1120. Ward, R. C. (1984) On the response to precipitation of headwater streams in humid areas. J. Hydrol. 74, 171-189. Ward, R. C. and Robinson, M. (2000) "Principles of Hydrology," Fourth/Ed. McGraw Hill, UK. Warn, L., McLarty, G. and Frame, H. (2001) Improving pasture and wool production with rotational grazing. In "42nd Annual Conference Proceedings. Systems making cents", pp. 168-169. Grasslands Society of Victoria Inc., Mount Gambier. Wasson, R. J., Donnelly, T. H. and Murray, A. S. (1996) Imports can be dangerous - appropriate approaches to Australian rivers and catchments. In "First National Conference on Stream Management in Australia", pp. 313-319. Cooperative Research Centre for Catchment Hydrology, Merrijig, Australia. Weaver, D. M. and Prout, A. L. (1993) Changing farm practice to meet environmental objectives of nutrient loss to Oyster Harbour. Fert. Res. 36, 177-184. Weaver, D. M. and Reed, A. E. G. (1998) Patterns of nutrient status and fertiliser practice on soils of the south coast of Western Australia. Agric. Ecosyt. Environ. 67, 37-53. Weld, J. L., Sharpley, A. N., Beegle, D. B. and Gburek, W. J. (2001) Identifying critical sources of phosphorus export from agricultural watersheds. Nutr. Cycl. Agroecosyst. 59, 29-38. West, C. P., Mallarine, A. P., Wedin, W. F. and Marx, D. B. (1989) Spatial variability of soil chemical properties in grazed pastures. Soil Sci. Soc. Am. J. 53, 784-789. Wheeler, D. M. and Edmeades, D. C. (1995) Effect of depth and lime or phosphorus fertiliser applications on the soil solution chemistry of some New Zealand pastoral soils. Aust. J. Soil Res. 33, 461-76. White, E. M. (1973) Water-leachable nutrients from frozen or dried prairie vegetation. J. Environ. Qual. 2, 104-107. White, I. and Sully, M. J. (1987) Macropscopic and microscopic capillary length and time scales from field infiltration. Water Resour. Res. 23, 1514-1522. White, I., Sully, M. J. and Perroux, K. M. (1989) The disk permeameter. Water Resour. Res. White, R. E. (1966) Studies on the phosphate potentials of soils IV. The mechanism of the "soil/solution ratio effect". Aust. J. Soil Res. 4, 77-85. White, R. E. and Taylor, A. W. (1977a) Reactions of soluble phosphate with acid soils: The interpretation of adsorption-desorption isotherms. J. Soil Sci. 28, 314-28.

280 White, R. E. and Taylor, A. W. (1977b) Effect of pH on phosphate adsorption and isotopic exchange in acid soils at low and high additions of soluble phosphate. J. Soil Sci. 28, 48-61. White, R. E. (1980) Retention and release of phosphate by soil and soil constituents. In "Soils and Agriculture: Critical Reports on Applied Chemistry" (P. Tinker, ed.), Vol. 2. Soc. Chem. Industry Blackwell, London. White, R. E. and Sharpley, A. N. (1996) The fate of non-metal contaminants in the soil environment. In "Contaminants in the Soil Environment in the Australasia-Pacific Region: Proceedings of the First Australasia-Pacific Conference on contaminants and soil environment in the Australasia-Pacific Region, held in Adelaide, Australia, 18-23 February 1996" (R. Naidu, R. Kookana, D. Oliver, S. Rogers and M. McLaughlin, eds.). Kluwer Academic Publishers, Dordrecht/Boston/London. White, R. E. (1997) "Principles and Practice of Soil Science," Third/Ed. Blackwell Science. White, R. E. and Kookana, R. S. (1998) Measuring nutrient and pesticide movement in soils: benefits for catchment management. Aust. J. Exp. Agric. 38, 725-43. White, R. E. and Ridley, A. M. (1998) Water theme protocol. In "Themes and Experimental Protocols for Sustainable Grazing Systems" (G. Lodge, ed.) Occasional Paper No 13/98. Land and Water Resources Research and Development Corporation Occasional Paper No 13/98. White, R. E., Helyar, K. R., Ridley, A. M., Chen, D., Heng, L. K., Evans, J., Fisher, R., Hirth, J. R., Mele, P. M., Morrison, G. R., Cresswell, H. P., Paydar, Z., Dunin, F. X., Dove, H. and Simpson, R. J. (2000) Soil factors affecting the sustainability and productivity of perennial and annual pastures in the high rainfall zone of south-eastern Australia. Aust. J. Exp. Agric. 40, 267-283. White, R. E., Christy, B., Ridley, A. M. and Heng, L. K. (2001) Estimating deep drainage under pasture. In "MODSIM 2001; International Congress on Modelling and Simulation" (F. Ghassemi, D. Post, M. Sivapalan and R. Vertessy, eds.), Vol. 1: Natural Systems, pp. 475-480. Modelling and Simulation Society of Australia and New Zealand Inc., The Australian National University, Canberra, Australia. White, R. E., Christy B.C., Ridley A.M., Okom A.E., Murphy S.R., Johnston W.H., Michalk D.L., Sandford P, McCaskill M.M., Johnson I.R., D.L., G., Hall D.J.M. and Andrew, M. H. (2003) SGS Water Theme: influence of soil, pasture type and management on water use in grazing systems of the high rainfall zone of southern Australia. Aust. J. Exp. Agric. 43, 907-926. Williams, J. and Hook, R. A. (1998) Livestock management for catchment care: what role quality assurance? Anim. Prod. Aust. 22, 27-37.

281 Williams, J. R., Dyke, P. T. and Jones, C. A. (1983) EPIC- a model for assessing the effects of erosion on soil productivity. In "Analysis of Ecological Systems: State-of-the-art in Ecological Modelling" (W. K. Lauenroth, G. V. Skogerboe and M. Flug, eds.), pp. 553- 572. Elsevier Scientific Publ. Co., Amsterdam. Williams, P. H. and Haynes, R. J. (1992) Balance sheet of phosphorus, sulphur and potassium in a long-term grazed pasture supplied with superphosphate. Fert. Res. 31, 51-60. Williams, W. D. and Wan, H. F. (1972) Some distinctive features of Australian inland waters. Water Res. 6, 829-836. Wilson, A. D. and Simpson, R. J. (1993) The pasture resource base: status and issues. In "Pasture Management: Technology for the 21st Century" (D. R. Kemp and M. D.L., eds.), pp. 1-25. CSIRO, Australia. Withers, P. J. A. (1996) Phosphorus cycling in UK agriculture and implications for water quality. Soil Use Manage. 12, 221. Withers, P. J. A., Clay, S. D. and Breeze, V. G. (2001) Phosphorus transfer in runoff following application of fertilizer, manure, and sewage sludge. J. Environ. Qual. 30, 180-188. Wooding, R. A. (1968) Steady infiltration from a shallow circular pond. Water Resour. Res. 4, 1259-1273. Woodruff, J. R. and Kamprath, E. J. (1965) Phosphorus adsorption as measured by the Langmuir isotherm and its relationship to phosphorus availability. Soil Sci. Soc. Am. Proc. 29, 148-150. World Health Organisation (1996) "Guidelines for Drinking Water; Health criteria and other supporting information Vol. 2," 2nd/Ed., Geneva. Wright, W. J. and Jones, D. A. (2003) Long-term rainfall declines in southern Australia. In "Proceedings National Drought Forum", pp. 12, Brisbane, Australia. Yeates, J. S. (1993) Soils and fertiliser use in southwestern Australia. Fert. Res. 36. Yeates, J. S. and Clarke, M. F. (1993) Developing alternatives to phosphate fertilizers of high water solubility. Fert. Res. 36, 141-150. Yli-Halla, M., Hartikainen, H., Ekholm, P., Turtola, E., Puustinen, M. and Kallio, K. (1995) Assessment of soluble phosphorus load in surface runoff by soil analyses. Agric. Ecosyt. Environ. 56, 53-62. Yuan, G. and Lavkulich, L. M. (1994) Phosphate sorption in relation to extractable iron and aluminium in Spodosols. Soil Sci. Soc. Am. J. 58, 343-346.

282 APPENDIX A: TIPPING BUCKET FLOW RATE CALIBRATION

EQUATIONS

y = L/tip, x = tips/min Plot 1 y = 14.615 + 0.3037* x − 0.0105* x2 + 0.0001* x3

Plot 2 y = 17.529 + 0.2657 * x − 0.0086* x2 + 0.0001* x3

Plot 3 y = 17.334 + 0.0978* x − 0.0012* x2 + 0.00008* x3

Plot 4 y = 17.858 + 0.3342* x − 0.0088* x2 + 0.0001* x3

i APPENDIX B: PERCHED AND GROUNDWATER TABLES

a) Plot 1 Lower slope

-0.5 Mar-99 Jun-99 Aug-99 Oct-99 Nov-99 Feb-00 May-00 Jul-00 Sep-00 Sep-00 Oct-00 Jan-01 0 0.5 1 1.5 2

Watertable Depth (m) Depth Watertable 2.5 3 A horizon piezo 1.4 m piezo 2.9 m piezo

b) Plot 1 Mid slope Mar-99 Jun-99 Aug-99 Oct-99 Nov-99 Feb-00 May-00 Jul-00 Sep-00 Sep-00 Oct-00 Jan-01 0 0.5 1 1.5 2 2.5 Watertable Depth (m) Depth Watertable 3 1.4 m piezo 2.9 m piezo

c) Plot 1 Upper slope Mar-99 Jun-99 Aug-99 Oct-99 Nov-99 Feb-00 May-00 Jul-00 Sep-00 Sep-00 Oct-00 Jan-01 0 0.5 1 1.5 2 2.5 Watertable Depth (m) Watertable 3 1.4 m piezo 2.9 m piezo

Figure B-1: Perched and groundwater table heights measured in a) lowerslope, b) midslope and c) upper slope positions in plot 1. Shaded regions indicate the depth of soil moisture accumulation indicated by neutron probe measurements.

ii a) Plot 2 Lower slope Mar-99 Jun-99 Aug-99 Oct-99 Nov-99 Feb-00 May-00 Jul-00 Sep-00 Sep-00 Oct-00 Jan-01 0 0.5 1 1.5 2 2.5 Watertable Depth (m) Depth Watertable 3 A horizon piezo 1.4 m piezo 2.9 m piezo

b) Plot 2 Mid slope Mar-99 Jun-99 Aug-99 Oct-99 Nov-99 Feb-00 May-00 Jul-00 Sep-00 Sep-00 Oct-00 Jan-01 0 0.5 1 1.5 2 2.5 Watertable Depth (m) Depth Watertable 3 1.4 m piezo 2.9 m piezo

Figure B-2: Perched and groundwater table heights in plot 2 in a) lowerslope and b) mid slope positions. Shaded regions indicate the depth of soil moisture accumulation indicated by neutron probe measurements.

iii a) Plot 3 Lower slope Mar-99 Jun-99 Aug-99 Oct-99 Nov-99 Feb-00 May-00 Jul-00 Sep-00 Sep-00 Oct-00 Jan-01 0 0.5 1 1.5 2 2.5 Watertable Depth (m) 3 A horizon piezo 1.4 m piezo 2.9 m piezo

b) Plot 3 Mid slope Mar-99 Jun-99 Aug-99 Oct-99 Nov-99 Feb-00 May-00 Jul-00 Sep-00 Sep-00 Oct-00 Jan-01 0 0.5 1 1.5 2 2.5 Watertable Depth (m) Watertable 3 1.4 m piezo 2.9 m piezo

c) Plot 3 Upper slope Mar-99 Jun-99 Aug-99 Oct-99 Nov-99 Feb-00 May-00 Jul-00 Sep-00 Sep-00 Oct-00 Jan-01 0 0.5 1 1.5 2 2.5 Watertable Depth (m) Depth Watertable 3 1.4 m piezo 2.9 m piezo

Figure B-3: Perched and groundwater tables in plot 3 in a) lowerslope, b) midslope and c) upper slope positions. Shaded regions indicate the depth of soil moisture accumulation indicated by neutron probe measurements.

iv a) Plot 4 Lower slope Mar-99 Jun-99 Aug-99 Oct-99 Nov-99 Feb-00 May-00 Jul-00 Sep-00 Sep-00 Oct-00 Jan-01 0 0.5 1 1.5 2 2.5 Watertable Depth (m) Depth Watertable 3 A horizon piezo 1.4 m piezo 2.9 m piezo b) Plot 4 Mid slope Mar-99 Jun-99 Aug-99 Oct-99 Nov-99 Feb-00 May-00 Jul-00 Sep-00 Sep-00 Oct-00 Jan-01 0 0.5 1 1.5 2 2.5 Watertable Depth (m) Depth Watertable 3 1.4 m piezo 2.9 m piezo c) Plot 4 Upper slope Mar-99 Jun-99 Aug-99 Oct-99 Nov-99 Feb-00 May-00 Jul-00 Sep-00 Sep-00 Oct-00 Jan-01 0 0.5 1 1.5 2 2.5 Watertable Depth (m) Depth Watertable 3 1.4 m piezo 2.9 m piezo

Figure B-4: Perched and groundwater tables in plot 4 in a) lowerslope, b) midslope and c) upper slope positions. Shaded regions indicate the depth of soil moisture accumulation indicated by neutron probe measurements.

v APPENDIX C: STORAGE EFFECTS ON MRP

CONCENTRATIONS

Introduction

Molybdate reactive P (MRP) concentrations in natural waters often fluctuate over time after collection due to uptake by micro-organisms, cell lysis and decomposition and adsorption to suspended sediment and collection container walls. However, the effect of storage conditions and time on measured MRP concentrations varies between waters. Haygarth et al (1995) found that over 32 days, the least changes in MRP concentrations in soil leachate occurred in samples stored in ambient laboratory conditions and under refrigeration. In contrast, Stevens et al (1997) measured MRP concentrations in runoff and through flow samples from the Adelaide Hills, South Australia and concluded that should storage of samples be unavoidable, concentrations were less likely to change if frozen rather than at 4 or 20oC. To avoid changes occurring, it is widely advised that samples be stored in acid-washed containers and measured within 24 – 48 hrs. For runoff samples from Vasey and Maindample, MRP analysis was not always possible within 24h of sample collection (see Chapter 5). Therefore, the effect of long term storage of filtered and unfiltered samples at –15oC was investigated to identify whether adjustments needed to be made to DRP and TRP concentrations measured on stored samples.

Materials and methods

In 1999, twenty-seven simulated runoff sub-samples collected at Vasey were filtered and analysed for MRP within 24h. DRP and TRP were measured on filtered and unfiltered samples respectively using the manual method H2a of Rayment and Higginson (1992). In 2000, natural runoff samples collected from Vasey were filtered and analysed for MRP within 24h. Filtered (n=95) and unfiltered (n=100) samples were analysed for DRP and TRP respectively using the manual method H2a of Rayment and Higginson (1992). All samples were then stored at –15oC for approximately 4 months, thawed and analysed for DRP and TRP using a Lachat flow injection system (Huberty and Diamond 1996). Because any effects of storage time may have been confounded by the differences in MRP analysis methods, 20 unfiltered and 20 filtered samples were analysed on the same day using both the manual and automated methods described above.

vi Differences in the precision and bias of the means between sample groups were tested using a statistical model that was originally developed to compare paired analysis results from two different laboratories (Jorgensen 1985).

Results and Discussion

There were no significant differences in the mean DRP concentrations between samples analysed within 24 h and those stored frozen for 4 months prior to analysis (Table C-1). Mean TRP concentrations, however, decreased by 39 and 15 percent after frozen storage, reflecting a significant bias (P<0.001). The variance in DRP concentrations was less for immediately analysed sample than for frozen samples in 1999 but was greater in 2000. The variance in TRP concentrations decreased by up to 32% in samples analysed by automated method after frozen storage. There was only a 5% difference in the mean MRP concentrations measured using either the manual or automated colorimetric methods.

Table C-1: Statistics for TRP and DRP analysed within 24 h and after being stored frozen for 4 months

TRP DRP mg/L mg/L <24h Stored % change <24h Stored % change frozen over time frozen over time 1999 Mean 0.31 0.19 -39 *** 0.19 0.19 0 Variance 0.050 0.034 -32 * 0.032 0.036 11 * N2526 Outliers 21 removed 2000 Mean 0.18 0.16 -15 *** 0.15 0.14 -2 Variance 0.043 0.031 -27 *** 0.032 0.027 -15 *** N9693 Outliers 42 removed *, ** and *** indicate significant differences at P<0.05, 0.01 and 0.001

vii Table C-2: Statistics for MRP concentrations measured using manual and automated methods

Manual Automated % method method difference MRP (mg/L) Mean 0.45 0.47 5 *** Variance 0.050 0.034 11 *** N40

Discussion

The mean MRP concentration measured during automated analysis was only 5% higher than the when using the manual method, so the method of colorimetric MRP analysis was unlikely to have caused the decrease in TRP concentrations measured after frozen storage. Instead, the decrease in TRP concentrations may have been due to complexation with organic material >0.45µm in size. Uptake by micro-organisms was unlikely due to the low storage temperatures reducing microbial activity, and adsorption to container walls was also unlikely considering there was no significant change in DRP concentrations. The larger decrease in 1999 samples may have been due to microbial uptake post-thawing as not all samples were analysed within 24hrs of thawing due to large sample throughput in that year.

Conclusion

The comparison of mean DRP concentrations measured within 24 hrs of sample collection and after storage at –15oC for 4 months suggested there was no need to adjust DRP values measured after frozen storage. The results do suggest, however, that TRP concentrations measured within 24hrs of frozen samples being thawed will underestimate the value measured within 24hrs of sample collection by approximately 15%.

viii

Minerva Access is the Institutional Repository of The University of Melbourne

Author/s: Melland, Alice Rowena

Title: Pathways and processes of phosphorus loss from pastures grazed by sheep

Date: 2003-12

Citation: Melland, A. R. (2003). Pathways and processes of phosphorus loss from pastures grazed by sheep. PhD thesis, School of Agriculture and Food Systems, The University of Melbourne.

Publication Status: Published

Persistent Link: http://hdl.handle.net/11343/39116

File Description: Pathways and processes of phosphorus loss from sheep pastures

Terms and Conditions: Terms and Conditions: Copyright in works deposited in Minerva Access is retained by the copyright owner. The work may not be altered without permission from the copyright owner. Readers may only download, print and save electronic copies of whole works for their own personal non-commercial use. Any use that exceeds these limits requires permission from the copyright owner. Attribution is essential when quoting or paraphrasing from these works.