0 100 200 300 400 500 600 700 MURRAY-DARLING BASIN COMMISSION Risks to Shared Water Resources

Impact of the 2003 Alpine Bushfires on Streamflow: Estimated changes in stream exports of sediment, phosphorus and nitrogen following the 2003 bushfires in Eastern Victoria

December 2007

Prepared by the University of Melbourne for the Victorian Department of Sustainability and Environment, and the Murray-Darling Basin Commission. Impact of the 2003 Alpine Bushfires on Streamflow - Estimated changes in stream exports of sediment, phosphorus and nitrogen following the 2003 bushfires in Eastern Victoria

Gary Sheridan, Patrick Lane, Philip Noske, Paul Feikema and Chris Sherwin, School of Forest and Ecosystem Science, The University of Melbourne.

Rodger Grayson, Civil and Environmental Engineering, The University of Melbourne.

Published by Murray-Darling Basin Commission Postal Address GPO Box 409, Canberra ACT 2601 Office location Level 4, 51 Allara Street, Canberra City Australian Capital Territory

Telephone (02) 6279 0100 international + 61 2 6279 0100 Facsimile (02) 6248 8053 international + 61 2 6248 8053 E-Mail [email protected] Internet http://www.mdbc.gov.au

For further information contact the Murray-Darling Basin Commission office on (02) 6279 0100

This report may be cited as: Impact of the 2003 Alpine Bushfires on Streamflow - Estimated changes in stream exports of sediment, phosphorus and nitrogen following the 2003 bushfires in Eastern Victoria

MDBC Publication No. 22/08

ISBN 978 1 921257 62 9

This project was funded by the Department of Sustainability and Environment, the Murray-Darling Basin Commission and the Gippsland Lakes Taskforce.

This work is copyright. Graphical and textual information in the work (with the exception of photographs and logos) may be stored, retrieved and reproduced in whole or in part, provided the information is not sold or used for commercial benefit and its source (Impact of the 2003 Alpine Bushfires on Streamflow - Estimated changes in stream exports of sediment, phosphorus and nitrogen following the 2003 bushfires in Eastern Victoria) is acknowledged. Such reproduction includes fair dealing for the purpose of private study, research, criticism or review as permitted under the Copyright Act 1968. Reproduction for other purposes is prohibited without prior permission of the copyright owners through the Murray-Darling Basin Commission or the individual photographers and artists with whom copyright applies.

To the extent permitted by law, the copyright holders (including their employees and consultants) exclude all liability to any person for any consequences, including but not limited to all losses, damages, costs, expenses and any other compensation, arising directly or indirectly from using this report (in part or in whole) and any information or material contained in it.

The contents of this publication do not purport to represent the position of the Murray-Darling Basin Commission or the copyright holders. They are presented to inform discussion for improvement of the Basin's natural resources.

Front cover images copyright Department of Sustainability and Environment, Victoria. Used with permission Impact of the 2003 Alpine Bushfires on Streamflow - Estimated changes in stream exports of sediment, phosphorus and nitrogen following the 2003 bushfires in Eastern Victoria

Executive Summary

Following the 2003 bushfires a program of water quality monitoring was initiated by the Department of Sustainability and Environment (DSE) in eight Victorian catchments affected by the fires: Dargo River at Lower Dargo Rd., Tambo River at Bindi and Swifts Ck., Mitta Mitta River at Hinnomunjie, Snowy River at McKillops Bridge, Ovens River at Bright, Kiewa River at Bandiana and the Wonnangatta River at Waterford. The objective of the monitoring program was to enable post-fire changes in loads of total suspended solids (TSS), total phosphorus (TP) and total nitrogen (TN) to be estimated.

The spatial extent of the water quality sampling represents the most ambitious attempt to measure fire effects in Australia, and possibly elsewhere. Large impacts were recorded for all constituents, however levels of change were not consistent across the different catchments. Similarly, recovery was found to be highly variable between catchments. Analysis of the water quality datasets revealed very high levels of uncertainty in load estimates. Sources of uncertainty were classified and where possible, attempts were made to quantify, either quantitatively or qualitatively, the magnitude of uncertainty in load estimates. It is critical to the correct interpretation of this report that load estimates are considered in conjunction with the associated uncertainty. The uncertainty analysis was used to group the results from the eight catchments into three “confidence categories” according to the level of confidence in the estimated loads.

Category One: Most confident

The three catchments with the highest level of confidence in the estimated loads are;

• Dargo River at Lower Dargo Road (Station No. 224213)

• Tambo River at Bindi (Station No. 223208) and at Swifts Creek (Station No. 223202)

• Kiewa River at Bandiana, based on TP and TN data only (Station No. 402205)

The confidence in the load estimates from these catchments is due largely to the representativeness of the post-fire storm-event water quality sampling, the number of samples collected both before and after the fire, and the level of agreement between the load estimates using alternative load estimation methods. Estimated load increases for TSS for the year following the fire (2003) vary from no change (Kiewa River) through to a 1400 times increase for the Tambo River at Bindi. Stochastic uncertainty is still very high, with coefficients of variation between 100% and 300% on these factor increases (ie. the standard deviation of the factor increases can be as much as 3 times the value of the factor increase). This high stochastic uncertainty is largely due to the presence of a few extremely high concentration values recorded during major erosion events following the fire. Impact of the 2003 Alpine Bushfires on Streamflow - Estimated changes in stream exports of sediment, phosphorus and nitrogen following the 2003 bushfires in Eastern Victoria

Factor increases in TP loads vary from no change to 400 times, and for TN from no change to 94 times. Both the factor increases and the stochastic uncertainty is generally highest for the TSS loads, and lower for the TP and TN loads respectively.

Recovery after the fire is also highly variable between these catchments, with the Tambo at Swifts Creek showing large annual improvements from 2003 to 2005, while analysis of the data for the Dargo River shows no recovery in estimated loads since the fire. There was no change in estimated export loads from the Kiewa following the fire.

Category Two: Less confident

The two catchments where a moderate level of confidence can be placed in the load estimates are;

• Mitta Mitta River at Hinnomunjie (Station No. 401203)

• Ovens River at Bright (Station No. 403205)

The storm events in the Mitta Mitta and Ovens catchments were sampled erratically during the period 2003 to 2005, reducing the amount of confidence that can be placed in the load estimates. The large increases in TSS loads (ranging from 168 times to 359 times, depending on the load estimation method) and high stochastic uncertainty in the Mitta Mitta are attributed to a small number of extremely high sediment concentrations (>43, 000 mg/L) recorded in late 2003.

Category Three: No confidence

The two catchments where no conclusions can be drawn from the available data are;

• Snowy River at McKillops Bridge (Station No. 222209)

• Wonnangatta River at Waterford (Station No. 224201)

The lack of confidence in the estimates for these sites is mostly due to the non-representative nature of the water quality sampling after the fire. While the data for these sites has been analysed alongside the other six catchments, the results are only presented to allow inspection of the limitations of the data, as the load estimates are considered to be of such high uncertainty that their use is not recommended.

The results above are presented in terms of the expected change in loads relative to the unburnt state for the year of measurement. The relative change in discharge from year to year was small compared to the relative change in concentrations, and as a result conclusions expressed in terms of changes in loads reflect an approximately equivalent proportional change in concentrations of the constituents TSS, TP and TN.

iii Impact of the 2003 Alpine Bushfires on Streamflow - Estimated changes in stream exports of sediment, phosphorus and nitrogen following the 2003 bushfires in Eastern Victoria

Load estimation methods

A range of load estimation methods were applied to the dataset, including;

• Interpolation (where sampling frequency made this method possible)

• Regression (against discharge, and change in discharge)

• Averaging (using arithmetic and geometric means)

• Monte Carlo simulation (using log-normal distributions as inputs to the averaging method)

In addition a range of data stratifications were applied including separating baseflow and event flow via baseflow separation, and separation of the hydrograph into rising limb and falling limbs. Each estimation method is subject to a range of assumptions, and in many cases these did not hold true for the water quality data in this analysis. The most successful and broadly applicable method was the averaging method using the arithmetic mean of the concentration data, stratified into baseflow and event flow concentrations. Monte Carlo simulations using the averaging method also produced similar load estimates.

End of catchment impacts

When interpreting the water quality estimations within this report it is important to note that the sediment and nutrient loads calculated for rivers in the upper catchment are likely to be substantially higher than the loads actually delivered to water impoundments in the lower catchment eg. the Gippsland Lakes. This is because sediment and adsorbed nutrients are stored within stream channels in the lower reaches of the stream network as the slope of the stream channel is reduced. Some of this material may however be re-mobilised during subsequent periods of high flow. It should also be recognised that the burnt areas represent relatively small proportions of the total catchment areas for some receiving waters such as the Gippsland Lakes. Hence the large increases identified in this study, where monitoring was relatively high up in the catchment, will represent much smaller percentage increases at a whole-of-catchment scale. Estimation of end-of-catchment loads of TSS, TP and TN are given in companion reports by Feikema et al. (2005; 2006).

iv Impact of the 2003 Alpine Bushfires on Streamflow - Estimated changes in stream exports of sediment, phosphorus and nitrogen following the 2003 bushfires in Eastern Victoria Table of Contents

1 PROJECT BACKGROUND ...... 7

2 OBJECTIVE ...... 7

3 DETAILS OF WATER QUALITY MONITORING SITES ...... 7

4 REVIEW OF LOAD ESTIMATION METHODS ...... 10

4.1 Averaging ...... 11

4.2 Ratio estimators...... 11

4.3 Monte Carlo Simulation...... 12

4.4 Regression methods ...... 12

4.5 Error estimation...... 13

5 METHODS...... 14

5.1 Water quality sampling ...... 14

5.2 Laboratory analysis...... 14

5.3 Regression ...... 14

5.4 Averaging ...... 15

5.4.1 Baseflow separation...... 15 5.4.2 Uncertainty estimation ...... 15 5.5 Linear interpolation ...... 17

5.6 Monte Carlo Simulation...... 17

5.7 Comparison of methods ...... 18

6 RESULTS...... 18

6.1 Exploratory data analysis ...... 18

6.1.1 Dargo River at Lower Dargo Road ...... 28 6.1.2 Kiewa River at Bandiana...... 28 6.1.3 Mitta Mitta River at Hinomunjie...... 29 v Impact of the 2003 Alpine Bushfires on Streamflow - Estimated changes in stream exports of sediment, phosphorus and nitrogen following the 2003 bushfires in Eastern Victoria

6.1.4 Ovens River at Bright...... 29 6.1.5 Snowy River at McKillops Bridge...... 29 6.1.6 Tambo River at Bindi ...... 30 6.1.7 Tambo River at Swifts Creek ...... 30 6.1.8 Wonnangatta River at Waterford ...... 30 6.2 Evaluation of the interpolation method ...... 31

6.3 Evaluation of the regression method...... 31

6.4 Estimated loads and uncertainty...... 33

6.4.1 Uncertainty estimates using Monte Carlo simulation...... 37 6.4.2 Dargo River at Lower Dargo Road ...... 42 6.4.3 Kiewa River at Bandiana...... 43 6.4.4 Mitta Mitta River at Hinomunjie...... 44 6.4.5 Ovens River at Bright...... 46 6.4.6 Snowy River at Mc Killops Bridge...... 46 6.4.7 Tambo River at Bindi ...... 47 6.4.8 Tambo River at Swifts Creek ...... 48 6.4.9 Wonnangatta River at Waterford ...... 48 6.5 Evaluation of load estimation methodology – a case study ...... 49

6.5.1 Tambo at Bindi dataset...... 49 6.5.2 Interpolation method ...... 49 6.5.3 Regression method...... 51 6.5.4 Average method...... 52 6.5.5 Comparison of methods...... 52 6.6 Recovery ...... 53

7 DISCUSSION AND SUMMARY ...... 53

8 ACKNOWLEDGEMENTS ...... 56

9 REFERENCES ...... 57

APPENDIX 1 ...... 59

APPENDIX 2 ...... 64

APPENDIX 3 ...... 67

vi Impact of the 2003 Alpine Bushfires on Streamflow - Estimated changes in stream exports of sediment, phosphorus and nitrogen following the 2003 bushfires in Eastern Victoria

1 Project Background

The scale and severity of the January – February 2003 North-eastern Victorian bushfires suggested that there may be substantial alteration in the quality and quantity of water available for downstream users. Approximately 1.3 million ha were burnt, including the Upper-Murray, Kiewa, Ovens, Snowy, Tambo and Mitchell river basins. Although elevation of sediment and nutrient loads are widely recognised as consequences of intense wildfire, relatively few studies have been carried out in Australia (eg. Brown, 1972; Prosser and Williams, 1998, Leitch et al. 1984; Chessman 1986; Burgess et al., 1981). Overseas studies provide only general indication of expected impacts. The Bushfire Recovery Program and the MDBC supported a research program through the CRC for Catchment Hydrology involving the School of Forest and Ecosystem Science, The University of Melbourne (formerly the Forest Science Centre) and Sinclair Knight Merz that will allow the measurement and prediction of the magnitude and duration of these changes and an assessment of the implications for water resource management.

One of the projects within this program (Project 5) involves the analysis of water quality data collected from numerous hydrologic stations downstream of the fire affected area in Victoria. These data have been collected from automated pumping samplers installed following the fire, and from routine periodic samples collected before and after the fire.

2 Objective

The objective was to estimate the changes in the in-stream export of total suspended solids (TSS), total phosphorus (TP) and total nitrogen (TN) from Victorian catchments affected by the 2003 bushfires.

3 Details of water quality monitoring sites

The hydrologic stations where water quality data has been evaluated lie to the north and south of the Great Dividing Range, and include the Ovens, Kiewa, Upper Murray, Snowy, Tambo, and Mitchell River basins. Major water impoundments potentially affected by the fire include the Gippsland Lakes, and the Hume and Dartmouth dams. DSE contracted Thiess Environmental Services to install and operate water auto-samplers at existing stream gauging stations in February 2003. The locations of each of the hydrologic stations are shown in Figure 1. The burnt area is shaded in red and the forested area is in green.

7 Impact of the 2003 Alpine Bushfires on Streamflow - Estimated changes in stream exports of sediment, phosphorus and nitrogen following the 2003 bushfires in Eastern Victoria

Water quality data was also collected by DSE from the Mitchell River at Glendale (Station No. 224203) and by the Murray Darling Basin Commission from Snowy Creek at Granite Flat (Station No. 401210). Water quality data from these two sites was not analysed in this study because of very limited event sampling in 2003 and 2004 for the Mitchell River at Glendale site, and because of a lack of pre-fire data for the Snowy Creek at Granite Flat site. A summary of the limited data collected from these sites is given in Table 8 in Appendix 3. A summary of the eight sites investigated in this report is given in Table 1.

Kiewa River @ Bandiana

#S

Mitta Mitta River @ Hinnomunjie #S

Ovens River @ Bright #S Tambo River @ Bindi

#S

#S #S Snowy River @ McKillop Bridge #S

#S #S

#S

Tambo River @ Swifts

Dargo River @ Lower Dargo Rd

Wonnangatta River @ Waterford

Figure 1. The location of the hydrologic stations referred to in this report.

8 Impact of the 2003 Alpine Bushfires on Streamflow - Estimated changes in stream exports of sediment, phosphorus and nitrogen following the 2003 bushfires in Eastern Victoria

Table 1. Summary of water quality and hydrologic data sets investigated in this report. Dates pertain to periods of active water quality sampling. A number of sites have discharge data beyond these dates.

Period Pre Post Post Station Site Area % burnt* Location of fire fire fire Comments 2 # code km (approx) record grab grab auto

Dargo R @ Lower Dargo Rd 224213 DRL 676 62 1990 - ò ò ò Kiewa R @ Bandiana 402205 KRB 1655 24 1990- ò ò ò TSS for period 1990-1992 only Mitta Mitta R @ Hinomunjie 401203 MMR 1533 80 1978- ò ò ò Ovens R @ Bright 403205 ORB 495 55 1978- ò ò ò TN 1985-1998 Snowy R @ McKillops Bridge 222209 SRM 10619 90 1990- ò ò ò Tambo R @ Bindi 223208 TRB 523 90 1978-87 ò x ò Tambo site shifted to Swifts Ck as of Jul 2003 Tambo @ Swifts Creek 223202 TRS 943 70 1978- ò ò ò Autosampler emplaced July 2003 Wonnangatta River @ Waterford 224201 WRW 1979 30 1978-90 ò x ò *% burnt area is the approximate percentage of the catchment burnt above the monitoring station

9 Impact of the 2003 Alpine Bushfires on Streamflow - Estimated changes in stream exports of sediment, phosphorus and nitrogen following the 2003 bushfires in Eastern Victoria

4 Review of load estimation methods

The net load of in-stream TSS exported from the catchment St can be represented by the equation;

T = CQdtS (1) t ∫ 0 where C and Q are the instantaneous sediment concentration and discharge volume respectively, t is time, and t = T is the duration over which the load is calculated. An expression

describing the product C·Q as a function of time is generally not available so St is approximated by;

T i= t t = ∑ iQCS i (2) i=1 where i is the interval number, and when the sampling interval t is short compared to the period of time over which the discharge and concentration vary. While discharge is often frequently recorded, concentration is not (except in the case of in-situ turbidity measurement) and linear interpolation is therefore used between points of measurement given by;

n+1 −+− ttCttC )()( i it +1 ti +1 i ∑∑ qt (3) i=1 ≤< ttt ii +1 +1 − tt ii

where concentrations are denoted Cti , ti i=1,…,n are the times at which concentration is measured, t0 and tn+1 are the times at the start and end of each sub-interval, and qt is the discharge at each timestep.

The problems associated with accurate estimation of sediment and nutrient loads are numerous. They include both the physical constraints of implementing appropriate sampling regimes, and the related statistical issues associated with load estimation (eg. Cohn, 1995; Schwartz and Naiman; 1999, Robinson and Roerish, 1999; Letcher et al., 1999, Fox 2005). Equation 3 is often used to estimate the “true” load from catchments. If the sampling interval is large compared to the period of time over which the discharge and concentration vary then use of Eq 3 will result in large errors. In this case alternative load estimation methods are required, which can be categorised into three broad classes;

1. Averaging

2. Ratio estimators

3. Regression methods (rating curves)

10 Impact of the 2003 Alpine Bushfires on Streamflow - Estimated changes in stream exports of sediment, phosphorus and nitrogen following the 2003 bushfires in Eastern Victoria

The properties of the water quality sampling regime, (eg. frequency, length of period of sampling, fixed interval vs storm event sampling), will to some extent dictate the optimal estimation method. For example, fixed interval sampling over short study durations will tend to under-represent the larger events, resulting in underestimates of the true load. Exploratory analysis of the water quality dataset is required to establish wether or not the assumptions associated with each load estimation method are violated. Ideally, an estimator should be precise (ie. have low variance) and accurate (low bias). An unbiased estimator has an expected value that is equal to the population parameter. Stratification, (eg. by flow category, season, rising or falling limb of the hydrograph), of the data may also improve load estimates. Estimation methods are described in more detail below.

4.1 Averaging

A typical example of an averaging method for the estimation of the load L of a constituent is given by the model;

n c = QL ∑ i (4) i=1 n

th where Q is the total discharge, ci is the concentration of the i sample, and n is the number of samples.

The EPA (1999) study (described earlier) reported that loads of TSS, TP and TN estimated using the averaging method (total discharge multiplied by the arithmetic mean of the sample concentrations) were in error by a factor of 0.91, 0.87, and 1.02 respectively.

4.2 Ratio estimators

A typical example of a ratio estimator for the estimation of the load L of a constituent is given by a model where the average ratio of load to discharge is multiplied by the total discharge;

R = )/( XxyY (5) where y and x are the sample means of yi and xi respectively, YR is the ratio estimate of load and X is the discharge. The ratio estimator is considered a best linear unbiased estimate under two conditions:

1. The relation between Qi and Ci is a straight line through the origin.

2. The variance of Ci is proportional to Qi.

11 Impact of the 2003 Alpine Bushfires on Streamflow - Estimated changes in stream exports of sediment, phosphorus and nitrogen following the 2003 bushfires in Eastern Victoria

These conditions are generally not met so the ratio estimator is biased, (ie. the mean of the estimator is not equal to the population parameter). Preston et al. (1989) tested a number of different estimation techniques and found that stratification under event sampling virtually eliminated bias using the ratio estimator, however this may only indicate that the underlying distributions of the test data were appropriate for ratio estimation (EPA 1999).

The EPA (1999) collected 70 samples routinely on a monthly basis and up to six times per day during flood discharge from the Richmond River (subcatchment area 1790 km2) 3 km upstream of Casino over a one year period from July 1995. Discharge was measured hourly, and sediment and nutrient loads were estimated using linear interpolation on a one hour timestep. The ratio estimator method using flow-weighted mean concentrations overestimated loads of TSS, TP and TN by a factor of 1.58, 1.23, and 1.18 respectively.

4.3 Monte Carlo Simulation

Monte Carlo simulation can be used with either the ratio or the averaging methods. The method involves generating a random population of input values with similar distribution parameters as the observed data. Load estimates are made for each instance of the input data, generating a distribution of estimated loads.

4.4 Regression methods

Regression methods exploit the often strong relationship between discharge and concentration. The advantage here is that discharge is often measured with high frequency (eg. sub-hourly), while concentrations data are more difficult to collect and are therefore less frequent. Generally, log-log regressions are applied because flow and concentration are assumed to be described by a bi-variate log normal distribution. A common log-log model is given by;

Ln(C) = a + bLn(Q) (6) where Q is the discharge, C is the concentration, and a and b are regression parameters. Several studies have found that load estimates based on the regression method are biased and systematically underestimate the true load. Bias correction methods are available. Discharge- concentration relationships can also display hysterisis loops with time at both short and longer time scales, requiring stratification of the data to develop multiple rating curves. Regression methods should not be used when the relationship between discharge and concentration is not log-log. The model includes calculable error within the range of the regression and unknown error due to extrapolation of the regression outside the range over which the regression was developed.

12 Impact of the 2003 Alpine Bushfires on Streamflow - Estimated changes in stream exports of sediment, phosphorus and nitrogen following the 2003 bushfires in Eastern Victoria

In the EPA (1999) study conducted in the Richmond River catchment described above, the regression method provided good estimates of loads of TSS, TP and TN, with estimates in error by small factors of 0.98, 1.01, and 1.09 respectively. The accuracy of these estimates reflects the strength of the relation between discharge and constituent loads in this catchment, yielding r2 values of 0.98, 0.99, and 0.97 for TSS, TP and TN respectively.

Preston et al. (1989) also found that regression methods could provide the lowest estimate error when relationships between flow and concentration were strong and consistent. However, in many of the test cases these conditions were not present and the regression estimators could be imprecise or biased or both.

4.5 Error estimation

The estimation of catchment exports of TSS and other constituents often include considerable uncertainty from a range of sources. There are a number of error taxonomies, however a simple classification listed by Etchells et al. (2005) includes 3 broad error groupings:

1. Knowledge uncertainty, involving our uncertainty in the process / model.

2. Stochastic uncertainty, due to variation in the data. Can be represented by the variance.

3. Measurement uncertainty, associated with errors in instrumentation and measurement.

Knowledge uncertainty can be minimised by better understanding the processes that drive variation in loads, and by ensuring that the effect of these processes is represented within the model. Knowledge uncertainty can be evaluated using ensemble methods, (eg. Etchells et al. 2005), whereby a range of equally valid models are applied, for example, by using ratio and averaging methods, and including a range of stratification approaches within each method. The results from all the approaches are combined, giving a distribution of results from which both the average load, and the error due to knowledge uncertainty can be quantified to some extent.

Measurement uncertainty is associated with instrumentation error, sampling uncertainty (due to non-representative sampling), and inappropriate scaling assumptions. While instrumentation error may be able to be quantified, uncertainty due to inappropriate and non-representative sampling is usually not quantifiable.

Finally, stochastic uncertainty reflects the underlying variability in the data, in this case sediment and nutrient concentration data, and can be readily quantified using estimates of the variance. The analytical propagation of stochastic uncertainty through estimation models can become complex even with relatively simple models. One approach is to use Monte Carlo simulation to produce a population of input data with the same parameters as the observed population. An

13 Impact of the 2003 Alpine Bushfires on Streamflow - Estimated changes in stream exports of sediment, phosphorus and nitrogen following the 2003 bushfires in Eastern Victoria output population of estimated loads allows the uncertainty in the input data to be reflected in the distribution of estimated output load values.

5 Methods

5.1 Water quality sampling

The water quality data analysed in this report results from two different sampling strategies. All of the pre-fire water quality data is from fixed-interval sampling (usually monthly or greater). The post-fire data is from storm event triggered automatic pumping samplers which respond to a change in river stage. They take samples either at fixed time intervals over the event or in response to further pre-programmed changes in river stage. Some of the hydrologic stations also have fixed interval manual samples for the post-fire period. Table 1 gives the period of pre- fire fixed-interval sampling. A more detailed description of the frequency and duration of the pre- and post-fire datasets is given in the results section.

5.2 Laboratory analysis

All post-fire samples have been analysed using standard laboratory techniques (Australian Water Technologies, 1999), for TSS, TN and TP. It is assumed that similar sample management and laboratory protocols have been used in the collection, handling and analysis of the pre-fire water quality samples.

5.3 Regression

Eq 6 was fitted using least squares regression to the pre and post-fire water quality data separately using the logarithm of the 15 min interval discharge data as the independent variable. A range of stratification methods were applied including:

• No stratification,

• Stratification of the water quality data into rising and falling limb (by visual inspection),

• Stratification into individual storm events,

• Stratification by rising and falling limb and by storm event,

• All of the above stratifications with discharge Q replaced with rate of change of discharge ∆Q.

14 Impact of the 2003 Alpine Bushfires on Streamflow - Estimated changes in stream exports of sediment, phosphorus and nitrogen following the 2003 bushfires in Eastern Victoria

The suitability of the regression method for estimating loads was assessed based on the values of the regression parameters.

5.4 Averaging

Constituent loads were estimated using a stratified version of the averaging method by calculating an event mean concentration (EMC) representative of storm event concentrations, and a dry weather mean concentration (DWC), represented by baseflow conditions:

2 = ∑ iQCL i )( (7) i=1

where C i=1 is the event mean concentration and C i=2 is the dry weather mean concentration and Q is the corresponding volume within the flow category. DWC’s and EMC’s were estimated following stratification of the concentration data into event-flow and base-flow categories based on the outflow hydrograph.

5.4.1 Baseflow separation

Baseflow separation algorithms were run on daily flows for the whole period of record at each catchment. The Lyne and Hollick (Nathan and McMahon, 1990) and 2 parameter Chapman and Maxwell (Grayson et al., 1996) algorithms were used. The Lyne and Hollick alpha parameter was set at 0.925, as recommended for forested catchments (Grayson et al., 1996). For the Chapman and Maxwell algorithm the recession coefficient, k, was set at 0.95 and the C parameter at 0.15. These parameters were varied but did not produce significant differences in baseflow. The Lyne and Hollick separation was used in the analyses as it is the most commonly used method. Each day of record was categorised as either baseflow or event flow based on a threshold value of the daily baseflow index (the ratio of baseflow to total flow). (BFI>0.85 was categorised as baseflow). A cut-off value < 1.0 was used after visual inspection to accommodate flows that exceeded the automatically separated baseflow but would be unlikely to carry event-level concentration of sediment and nutrients. The flow records were then used to assign water quality samples to event-flow or base flow categories. The resultant concentrations were then averaged to obtain DWCs and EMCs.

5.4.2 Uncertainty estimation

The error in load estimates using the averaging method were determined by applying the propagation of error formula for Y =f(X,Z, …):

15 Impact of the 2003 Alpine Bushfires on Streamflow - Estimated changes in stream exports of sediment, phosphorus and nitrogen following the 2003 bushfires in Eastern Victoria

2 2 ⎛∂Y ⎞ 2 ⎛∂Y⎞ 2 ⎛∂Y ⎞⎛∂Y⎞ 2 sy = ⎜ ⎟ sx +⎜ ⎟ sz ...++ ⎜ ⎟⎜ ⎟sxz (8) ⎝∂X ⎠ ⎝∂Z⎠ ⎝∂X ⎠⎝∂Z⎠ where, Sx , Sz and Sy are the standard deviations of the X Z & Y measurements respectively, and

∂Y is the partial derivative of the function Y with respect to X. Szx is the estimated covariance ∂X between the X,Z measurements. The covariance term is zero if X and Z measurements are independent, and should be included only if estimated from sufficient data. Applying the propagation of error formula to the load estimation equation gives:

2 2 ⎛ ∂L ⎞ 2 ⎛ ∂L ⎞ 2 ⎛ ∂L ⎞⎛ ∂L ⎞ 2 sL = ⎜ ⎟ s +⎜ ⎟ sv ...++ ⎜ ⎟⎜ ⎟s ⎝∂C⎠ C ⎝∂V ⎠ ⎝∂C⎠⎝∂V ⎠ VC (9)

where sC sv and sL are the standard deviation of the concentration volume and load ∂L ∂L measurements respectively. The partial derivatives simplify, giving = V and = C . ∂C ∂V

Assuming the covariance between V and C is zero (this is unknown due to the short time period of available data, but unlikely), Eq 9 simplifies to:

2 22 += sCsVs 2 L C v (10)

which is in a form used by Hart et al. (1987) for estimating uncertainty in load estimates. In this study the volume measurement V is a single annual value at each site for the year of the load estimate (allocated proportionally to the baseflow and event flow categories based on the 2 baseflow separation), and therefore the variance sv cannot be calculated. A coefficient of

variation (Cv) of 10% for the volume estimate was assumed and the standard deviation sv was subsequently approximated by transposing the coefficient of variation formula;

Cv = sv/V * 100% (11)

The standard deviation SL of the load estimate is calculated for each of the two flow categories

(event flow L1, and dry weather flow, L2) and combined by;

+= sss 22 (12) L 1 LL 2 to give the standard deviation of the estimated total load. The uncertainty in the ratio of the absolute values of predicted post-fire (Bnt) and pre-fire (UnBnt) load (ie. the factor change) is approximated from (Hepworth pers comm.): 16 Impact of the 2003 Alpine Bushfires on Streamflow - Estimated changes in stream exports of sediment, phosphorus and nitrogen following the 2003 bushfires in Eastern Victoria

UnBntBnt = [ UnBntEBnt ]2 + [ ]2 [ UnBntEBntEUnBnt )(/)()var()(/)var()/var( ]4 (13)

This approximation is calculated using the delta method, by expanding the function f(x,y) = x/y in a Taylor series around the mean value (E(X), E(Y)) and taking the variance of the linear terms. There are more terms in the series, though they don't contribute significantly to the estimated variance in the ratio. Note that the covariance is assumed to be zero in this approximation.

5.5 Linear interpolation

Discharge and water quality datasets were plotted and inspected for suitability with the application of the linear interpolation load estimation method. Suitability was assessed based on the extent to which storm events were sampled.

5.6 Monte Carlo Simulation

Concentration data were stratified by flow into event mean concentrations (EMC) and dry weather concentrations (DWC) using the baseflow separation methodology described in Section 5.4.1. For each flow category at each site, for each annual period, and for each constituent, the observed population of concentration values were assumed to be log-normal. The expected value E(X) and variance var(X) of the distribution were determined, respectively, from:

σ 2 XE exp()( μ += ) (14) 2

X = σ 2 − + σμ 2 )2exp()1)(exp()var( (15)

where μ and σ are the mean and standard deviation of the logarithms of the observed data. The observed data for all the sites were assumed to be log-normal because:

a) for the sites with a large number of water quality samples the sample distribution was observed to be approximately log-normal,

b) concentration data cannot be negative, and are often characterised by infrequent, very high concentration values, properties that are consistent with the log-normal distribution, and,

c), researchers commonly observe that water quality data is log-normally distributed (Parkhurst 1998).

The software package PopTools (Hood 2005) was used to generate between 1000 and 60,000 random values with a lognormal distribution with the distribution parameters of the observed data. The number of random values generated was determined by visually assessing a plot of sample size vs estimated mean, so as to determine the sensitivity of the sample parameters to 17 Impact of the 2003 Alpine Bushfires on Streamflow - Estimated changes in stream exports of sediment, phosphorus and nitrogen following the 2003 bushfires in Eastern Victoria the sample size. In general, observed samples with larger variances required a larger sample size to generate a stable mean value.

These concentration distributions were then multiplied by the appropriate discharge volume for each year in each flow strata at each site. The loads based on the DWC and the EMC were added at each iteration to get a total load for that iteration. This population of loads was the output from the simulation.

5.7 Comparison of methods

Where possible, a range of estimation methods were applied to the same concentration dataset. In particular, an attempt was made to apply the interpolation method over the same period for which other estimation methods were applied, to provide an insight into the possible levels of measurement or knowledge uncertainty associated with the estimation methods.

Both the averaging method, (using the arithmetic mean,) and the Monte Carlo simulation were applied to all the data where possible, which also provided a measure of knowledge uncertainty.

6 Results

6.1 Exploratory data analysis

Following the 2003 fires DSE developed a water quality sampling strategy to allow an estimation of the magnitude of the impacts of the 2003 fires on loads of sediment and nutrients exported from a selection of fire affected catchments. The sampling strategy focused on the installation of stage-triggered auto samplers at hydrologic monitoring stations. This type of sampling is particularly suited to load estimation using the linear interpolation method (Eq 3), whereby the concentration corresponding to each discharge value (which is frequently measured) is interpolated from the nearest sediment concentration sample points.

This sampling approach, combined with the interpolation load estimation method, is dependent on the adequate sampling across the storm events for the period of interest. More specifically, if estimation errors are to be avoided, the sampling interval must be short compared to the period of time over which the discharge and concentration vary. For example, if significant storm events are not sampled the resulting estimated load by the interpolation method is likely to be negatively biased because sediment concentrations during events are usually elevated relative to baseflow concentrations.

18 Impact of the 2003 Alpine Bushfires on Streamflow - Estimated changes in stream exports of sediment, phosphorus and nitrogen following the 2003 bushfires in Eastern Victoria

The temporal nature of the storm-based sampling for each site, for each year, is illustrated in Figure 2 to Figure 9. The summary table (Table 2) lists the approximate number of storms captured by sampling, and the percentage of storms captured for the year.

Table 2. Summary of the number of storm events in each catchment, and an approximation of the number and proportion of storms adequately sampled for water quality parameters.

2003 2004 2005

Location events events events Total # Total # Total # Station # # sampled # sampled # sampled % sampled % sampled % sampled

22421 16 6 38 16 10 63 Dargo River 15 7 47 3 40220 21 5 24 26 0 0 Kiewa River 21 0 0 5 40120 19 10 53 14 4 29 Mitta Mitta River 16 5 31 3 40320 13 4 31 12 3 25 Ovens River 13 1 8 5 22220 16 2 13 12 6 50 Snowy River 12 1 8 9 22320 8 6 75 ------Tambo River Bindi ------8 Tambo @ Swifts 22320 17 5 29 9 5 56 9 6 67 Ck 2 22420 16 2 13 12 0 0 Wonnangatta River 12 0 0 1 -- hydrologic station de-commissioned in June 2003

19 Impact of the 2003 Alpine Bushfires on Streamflow - Estimated changes in stream exports of sediment, phosphorus and nitrogen following the 2003 bushfires in Eastern Victoria

Station # 224213 Dargo River @ Lower Dargo Road (2003)

5000 10000 4500 4000 1000 3500 3000

2500 100 2000 1500 10 1000 500

0 1

3 3 3 3 0 03 03 03 03 03 0 0 0 00 0 0 00 00 /2 2 /2 /2 /2 /2 /2 /2 1/20 5/ 7 1 /0 /0 /0 /1 1 2/03/2003 1/04 1 9/08 8/09 7/12 31/01/2003 31/05 30/06/2003 30 2 2 28/10/2003 27 2

Station # 224213 Dargo River @ Lower Dargo Road (2004) 5000 10000

4500 4000 1000 3500 3000 2500 100

2000

Discharge (ML/d) Discharge 1500 10 1000 TSS concentration (mg/L) TSS concentration 500

0 1 4 4 4 4 4 04 04 04 04 04 00 0 00 0 0 00 0 00 00 0 /2 /2 /2 /2 /2 /2 /2 /2 /2 /2 3/2004 3 7 0 /0 /0 /0 /1 1/01 1/01 1 0/04 8/08 6/11 6/12 3 31 3 30/05 29/06/2004 29 2 27/09/2004 27 2 2

Station # 224213 Dargo River @ Lower Dargo Road (2005) 5000 10000 4500

4000 1000 3500 3000 2500 100 2000

1500 10 1000 500 0 1

5 5 5 5 5 5 5 05 0 0 0 00 00 0 00 00 /2 /2 /2 /2005 /2 /2 1 1 5 5/20 6 8/20 9/20 0 2/2005 /0 /0 /0 /0 /0 /0 /0 /1 /1 1 2/03 1/04/2005 1 9 8 7 31 31 30 30/07/2005 2 2 28 27/11/2005 2 Date Figure 2: Daily discharge (ML/d) and the temporal distribution of sediment concentration . (mg/L) samples for the Dargo River at Lower Dargo Road. 20 Impact of the 2003 Alpine Bushfires on Streamflow - Estimated changes in stream exports of sediment, phosphorus and nitrogen following the 2003 bushfires in Eastern Victoria

Station # 402205 Kiewa River @ Bandiana (2003)

14000 1000

12000

10000 100 8000

6000 10 4000

2000

0 1

3 3 3 03 0 03 03 03 03 0 003 003 0 /20 /2 2 /20 /2 /20 /20 /20 /20 4/ 5/2003 8 9 /03 0 0 /07/2003 0 0 1/01/200 2 1/ 1/ 9/ 8/ 31/01 31/05/2003 30/06 30 2 2 28/10 27/11 27/12

Station # 402205 Kiewa River @ Bandiana (2004) 14000 1000

12000

10000 100

8000

6000 10

(ML/d) Discharge 4000

2000 (mg/L) TSS Concentration

0 1 4 4 4 4 4 4 4 4 4 4 04 04 04 00 2 /20 1/ 3/200 0 0 1/ 1/01/200 1/ 1/03/200 0/04/200 0/05/200 9/06/200 9/07/200 8/08/20 7/09/200 7/10/20 6/11/200 6/12 3 3 3 3 2 2 2 2 2 2 2

Station # 402205 Kiewa River @ Bandiana (2005)

14000 1000

12000

10000 100 8000

6000

10 4000

2000

0 1

5 05 05 05 05 05 05 0 0 0 0 0 0 /2 /2 2 /2 7 9 0/ 2 /06/2 /0 /0 1 /11/2 1 1/01/2005 2/03/2005 1/04/2005 1/05/200 0 0 8 8/ 7 7/ 31/01/2005 31/05/2005 3 3 29/08/2005 2 2 2 2 Date Figure 3. Daily discharge (ML/d) and the temporal distribution of sediment concentration

(mg/L) samples for the Kiewa River at Bandiana. Note that samples have been collected 21 for TP and TN (but not for TSS) at this site since 1990. Impact of the 2003 Alpine Bushfires on Streamflow - Estimated changes in stream exports of sediment, phosphorus and nitrogen following the 2003 bushfires in Eastern Victoria

Station # 401203 Mitta Mitta River @ Hinnomunjie (2003)

100000 14000

12000 10000

10000 1000 8000

6000 100

4000 10 2000 0 1

3 3 3 3 0 03 0 03 03 003 003 /20 /200 /200 /2 /2003 /2 /2003 /20 /2003 /20 9 1 /05/2003 /05 07 1/01/20 2/03/20 1/04 1 7/12 31/01 31 30/06 30/ 29/08 28/0 28/10 27/1 2

Station # 401203 Mitta Mitta River @ Hinnomunjie (2004) 100000 14000

12000 10000

10000 1000 8000

6000 100

Discharge (ML/d) Discharge 4000 10 TSS Concentration (mg/L) TSS Concentration 2000

0 1

4 04 004 004 004 004 004 /2 2004 2 /2004 2004 4/2004 6/2004 01/ 03/2 0 05/2 0 07/20 08 /09/200 10/ /11/2004 12/2 1/01 1/03/ 9/ 31/ 31/ 30/ 30/ 29/ 2 28/ 27 27/ 26 26/

Station # 401203 Mitta Mitta River @ Hinnomunjie (2005) 100000 14000

12000 10000

10000 1000 8000

6000 100

4000 10 2000

0 1 5 5 5 5 0 05 0 05 05 005 005 /20 /200 /200 /2 /2005 /2 /2005 /20 /2005 /20 9 1 /01/20 /03/20 /04 /05 07 1 2 1 1/05/2005 7/12 31/01 31 30/06 30/ 29/08 28/0 28/10 27/1 2 Date Figure 5. Daily discharge (ML/d) and the temporal distribution of sediment concentration 22 (mg/L) samples for the Mitta Mitta River at Hinnomunjie. Impact of the 2003 Alpine Bushfires on Streamflow - Estimated changes in stream exports of sediment, phosphorus and nitrogen following the 2003 bushfires in Eastern Victoria

Station # 403205 Ovens River @ Bright (2003)

5000 10000 4500 4000 1000 3500

3000 2500 100 2000 1500 10 1000 500 0 1

3 3 03 003 /2003 2 20 200 1/ 01 /03/2003 05/2003 07/2003 /08/2003 09/ 10/2003 12/2003 1/ 2 1/04/ 1/05/2003 1/ 0/ 9 8/ 8/ 7/1 7/ 31/01/2003 3 30/06/200 3 2 2 2 2 2

Station # 403205 Ovens River @ Bright (2004) 5000 10000 4500

4000 1000 3500 3000 2500 100 2000

Discharge (ML/d) 1500 10 1000 TSS Concentration (mg/L)TSS Concentration 500

0 1

4 4 4 4 04 04 04 00 20 20 200 20 200 1/2 6/ 1/ 0 03/200 0 /08/2004 1 1/ 1/03/2004 1/ 31/01/ 3 30/04/ 30/05/2004 29/ 29/07/2004 28 27/09/ 27/10/2004 26/ 26/12/2004

Station # 403205 Ovens River @ Bright (2005) 5000 10000 4500 4000 1000 3500 3000

2500 100 2000

1500 10 1000 500

0 1

5 5 05 005 /2005 /2 /2005 20 200 1/ 01/2005 /03 04 05 06/200 /08/2005 1 1/01/2005 1/ 2 1/ 1/ 0/ 3 31/05/2005 3 30/07/2005 29 28/09/ 28/10/2005 27/ 27/12/2005 Date Figure 6. Daily discharge (ML/d) and the temporal distribution of sediment concentration 23 (mg/L) samples for the Ovens River at Bright. Impact of the 2003 Alpine Bushfires on Streamflow - Estimated changes in stream exports of sediment, phosphorus and nitrogen following the 2003 bushfires in Eastern Victoria

Station # 222209 Snowy River @ McKillops Bridge (2003)

4000 100000

3500 10000 3000

2500 1000

2000

1500 100

1000 10 500

0 1

3 3 3 03 03 03 03 03 03 03 03 03 03

/01/200 /01/20 /03/200 /04/20 /05/200 /05/20 /06/20 /07/20 /08/20 /09/20 /10/20 /11/20 /12/20 1 2 1 1 31 31 30 30 29 28 28 27 27

Station # 222209 Snowy River @ McKillops Bridge (2004) 4000 100000 3500 10000 3000

2500 1000 2000

1500 100

Discharge (ML/d) Discharge 1000 10 (mg/L) TSS Concentration 500

0 1

4 4 4 04 004 00 /2004 3/2004 /04/2004 /05/2004 1/01/2 1/01 1/03/2 1/0 0 0 8/08/2004 7/09/2004 7/10/200 6/11/200 6/12/20 3 3 3 3 29/06/2004 29/07/2004 2 2 2 2 2

Station # 222209 Snowy River @ McKillops Bridge (2005) 4000 100000 peak at 14,000 ML/d 3500 10000 3000

2500 1000

2000 1500 100

1000 10 500

0 1

5 5 5 0 0 0

1/01/20 2/03/20 1/04/2005 1/05/20 31/01/2005 31/05/2005 30/06/2005 30/07/2005 29/08/2005 28/09/2005 28/10/2005 27/11/2005 27/12/2005 Date Figure 7. Daily discharge (ML/d) and the temporal distribution of sediment concentration 24 (mg/L) samples for the Snowy River at McKillops Bridge. Impact of the 2003 Alpine Bushfires on Streamflow - Estimated changes in stream exports of sediment, phosphorus and nitrogen following the 2003 bushfires in Eastern Victoria

Station # 223208 Tambo River @ Bindi (2003)

200 100000 180 160 10000 140 120 1000

100 80 100 60 40 10 20

0 1

3 3 3 3 0 0 0 0 03 03 03 0 0 0 0 003 0 0 0 /2 /2 /2 /2 /2 /2 1 1 3 4 5 5 6/2003 7/2 8/2003 9/2003 0/2 1/2003 2/2003 /0 /0 /0 /0 /0 /0 /0 /0 /0 /0 /1 /1 /1 1 1 2 1 1 1 0 0 9 8 8 7 7 3 3 3 3 2 2 2 2 2

Station # 223208 Tambo River @ Bindi (2004) 5000 100000

4500 4000 10000 3500 3000 1000 2500

2000 100

Discharge (ML/d) Discharge 1500 1000 10 TSS Concentration (mg/L) TSS Concentration 500 0 1 4 4 4 4 4 4 4 4 4 4 0 0 0 0 0 0 0 0 0 0 /2004 /2004 /01/20 /03/20 /04/20 /06/20 /07/20 /08/20 /09/20 /10/20 /11/20 /12/20 1/01 1/03 31 31 30 30/05/2004 29 29 28 27 27 26 26

Station # 223208 Tambo River @ Bindi (2005) 5000 100000 4500

4000 10000 3500 3000 1000

2500 2000 100 1500 1000 10 500

0 1

5 5 5 5 5 5 5 5 5 0 0 0 0 0 0 0 0 0 0 0 /2 /200 /2 /2 /200 /2 /2

1/01/2005 2/03/2005 1/04/2005 1/05/20 31/01/2005 31/05/200 30/06 30/07 29/08 28/09 28/10 27/11 27/12 Date Figure 8. Daily discharge (ML/d) and the temporal distribution of sediment concentration (mg/L)

samples for the Tambo River at Bindi. Note that this site was decommissioned in July 2003. 25 Impact of the 2003 Alpine Bushfires on Streamflow - Estimated changes in stream exports of sediment, phosphorus and nitrogen following the 2003 bushfires in Eastern Victoria

Station # 223202 Tambo River @ Swifts Creek (2003) 3000 100000

2500 10000

2000 1000 1500 100 1000

10 500

0 1

3 3 3 3 3 3 3 3 0 0 03 0 0 0 0 0 0 0 0 0 0 /2 /2 /200 /2 /2 /2 /20 /2003 /2 1 4 5 5 6 8 2 /0 /03/200 /0 /0 /0 /0 /09/2003 /10 /11 /1 1/01/2003 1 2 1 1 0 9 8 7 3 31/0 3 30/07/2003 2 28 2 27 2

Station # 223202 Tambo River @ Swifts Creek (2004) 3000 100000

2500 10000

2000 1000 1500 100 1000 Discharge (ML/d) Discharge 10 500 (mg/L) TSS Concentration

0 1

4 4 4 4 4 4 4 0 0 0 0 0 0 0 0 /2004 /2 /20 /20 /2004 /2 /20 /20 /2004 3 7 /01/200 /0 /04/2004 0 /09/2004 /12/2004 1 1 0/05 9/ 7/10 31/01 31/03 30 3 29/06 2 28/08 27 2 26/11 26

Station # 223202 Tambo River @ Swifts Creek (2005) 3000 100000

2500 10000

2000 1000 1500

100 1000 10 500

0 1 5 5 5 5 5 5 0 0 0 0 0 05 0 005 0 0 /2 /2005 2 2 /2005 2 /2005 1 3/ 7/20 9/ 1/ 0 0 0 0 /10 /1 1/ 2/ 1/04/200 1/05/2005 1/05/20 0/ 8/ 31/01 3 30/06/2005 3 29/08/20 2 28 27 27/12 Date Figure 9. Daily discharge (ML/d) and the temporal distribution of sediment concentration (mg/L) samples for the Tambo River at Swifts Creek. Note that the auto sampler was not commissioned at this site until July 2003. 26 Impact of the 2003 Alpine Bushfires on Streamflow - Estimated changes in stream exports of sediment, phosphorus and nitrogen following the 2003 bushfires in Eastern Victoria

Station # 224201 Wonnangatta River @ Waterford (2003) 1000 14000

12000

10000 100 8000

6000 10 4000

2000 0 1

3 3 03 03 03 0 0 00 0 00 /2 /2003 /2 /2 3/2003 4/2 5/2003 5/2003 1/2 /01 0 /0 0 0 /10 /1 /12 1/01/2003 2/ 1 1/ 8 7 31 31/ 30/06/2003 30/07/2003 29/08 28/09/2003 2 27 2

Station # 224201 Wonnangatta River @ Waterford (2004) 1000 14000

12000

10000 100

8000

6000 10 Discharge (ML/d) Discharge 4000

TSS Concentration (mg/L) TSS Concentration 2000

0 1

4 4 4 4 4 4 4 04 0 00 00 00 00 00 00 20 /2 /20 /2 /2 /2 /2 /2 3 5 8 /01/ /01 /0 /06 /09 /11 1 1 1/03/2004 1/0 0 9 8/0 7 6 3 3 30/04/2004 3 2 29/07/2004 2 2 27/10/2004 2 26/12/2004

Station # 224201 Wonnangatta River @ Waterford (2005) 1000 14000

12000

10000 100

8000

6000 10 4000

2000

0 1

5 5 5 5 5 5 5 0 05 0 0 00 00 00 00 00 /2 /2 /2 2 /2 /20 1/2005 5 1 0 /04/2 /0 /06 /07 /08/ /09 1/ 1/01/20 2/03/2005 1 1/05/2005 1 0 0 9 8 7/1 3 3 3 3 2 2 28/10/2005 2 27/12/2005 Date Figure 10. Daily discharge (ML/d) and the temporal distribution of sediment concentration (mg/L) samples for the Wonnangatta River at Waterford.

27 Impact of the 2003 Alpine Bushfires on Streamflow - Estimated changes in stream exports of sediment, phosphorus and nitrogen following the 2003 bushfires in Eastern Victoria

Inspection of Figure 2 to Figure 9 shows that in the case of the post-fire datasets many storm events were not sampled. This may have been due to failure of the sampling equipment, or the setting of inappropriate trigger levels for auto samplers. Stream baseflow levels vary markedly throughout the year, and maintaining optimal trigger levels at remote sampling locations is difficult, especially when the samplers are triggered from failure-prone in-stream manual height switches. These difficulties may have led to the observed inconsistent sampling regime in the years following the 2003 fire. Re-evaluation of the sampling regime in late 2004 revealed some of these limitations, however by this point a change in the sampling strategy would have affected the assessment of the temporal recovery. As a result, the sampling approach used earlier in the study was retained across all sites, despite its limitations.

The nature of the water quality datasets available for each of the eight hydrologic stations is discussed in detail in the following sections. A summary listing of the complete water quality dataset is provided in the Appendix.

6.1.1 Dargo River at Lower Dargo Road

Pre-fire: A total of 150 pre-fire sediment concentration values collected monthly over a period of 12 years between 1990 and 2002 were available for analysis. Sixty-four of the pre-fire samples were categorised as base-flow and 86 as event-flow, providing a good dataset for the application of the averaging estimation method.

Post-fire: Four storm events were well sampled for the first 5 months of 2003, followed by three missed events. The two largest events for the year were well sampled in July and August. The remaining seven events for 2003 were not sampled. The 5 largest events were sampled in both 2004 and 2005. However, a large number of smaller events were not sampled, resulting in approximately 63% and 47% of storms being sampled for 2004 and 2005 respectively.

6.1.2 Kiewa River at Bandiana

Pre-fire: A total of only 19 (11 baseflow and 8 event flow) pre-fire sediment concentration values were available for analysis collected between 1990 and 1992. However, at this site a large number of samples have been collected for nutrient analysis, with no accompanying measurement of suspended solids. A total of 605 values for TP and TN collected weekly over a period of 12 years from 1990 to 2002 were available. Therefore, for this site estimates of TP and TN loads are likely to include considerably less uncertainty than estimates of TSS loads.

Post-fire: Five sets of samples were collected in the first six months of 2003, from a range of small to medium storm events. None of the large storm events were sampled at this site. There

28 Impact of the 2003 Alpine Bushfires on Streamflow - Estimated changes in stream exports of sediment, phosphorus and nitrogen following the 2003 bushfires in Eastern Victoria were no storm event samples collected after mid-2003. Weekly fixed interval samples were collected over the post-fire period, however these have not been analysed for TSS.

6.1.3 Mitta Mitta River at Hinomunjie

Pre-fire: A total of 189 pre-fire sediment concentration values (83 baseflow and 106 event flow) collected over a period of 24 years between 1978 and 2002 were available for analysis. Samples were initially collected quarterly, and then monthly from 1990 onwards. The relatively large pre-fire dataset, collected at fixed intervals over a long period, provides a good basis for application of the averaging estimation method.

Post-fire: Storm events were sampled to a variable degree throughout 2003, with initial small events being comprehensively sampled early in the year, and some larger events remaining unsampled later in the year. Overall, 53% of the events were sampled in 2003. In 2004, the four largest events were sampled. However, only approximately 30% of events were sampled overall. Similarly in 2005, four of the largest storms were sampled, though these accounted for only approximately 30% of the storms for the year.

6.1.4 Ovens River at Bright

Pre-fire: A total of 189 pre-fire sediment concentration values (88 baseflow and 101 event flow) were available for analysis collected over a period of 24 years between 1978 and 2002. Samples were initially collected quarterly, and then monthly from 1990 onwards. The relatively large pre-fire dataset, collected at fixed intervals over a long period, provides a good basis for application of the averaging estimation method.

Post-fire: Three small events were sampled early in 2003, while subsequent events of similar or larger magnitude were not sampled during the autumn-winter of 2003. Two of the largest events were captured in winter 2003, while the three other major events were not sampled in the latter part of 2003. Possibly the largest erosion event of the year in the Ovens catchment was not captured at this water quality station, as it occurred in the Buckland Valley, a tributary that enters the Ovens River downstream of the gauging station at Bright. Three events were sampled in 2004, including the largest event for the three year period. Only a single event was sampled in 2005. Overall, approximately 31%, 25%, and 8 % of storm events were sampled in 2003, 2004, and 2005 respectively.

6.1.5 Snowy River at McKillops Bridge

Pre-fire: A total of 149 pre-fire sediment concentration values (67 baseflow and 82 event flow) were available for analysis collected monthly over a period of 12 years between 1990 and 2002.

29 Impact of the 2003 Alpine Bushfires on Streamflow - Estimated changes in stream exports of sediment, phosphorus and nitrogen following the 2003 bushfires in Eastern Victoria

The relatively large pre-fire dataset, collected at fixed intervals over a long period, provides a good basis for application of the averaging estimation method.

Post-fire: Storm events were sampled intensively in late February and early March 2003. The remainder of the year was not sampled, resulting in only about 13% of events being sampled. Six events were sampled in 2004, and one was sampled in 2005.

6.1.6 Tambo River at Bindi

Pre-fire: A total of 35 pre-fire sediment concentration values were available for analysis, however due to missing hydrologic records between 1983 to 1987, only 18 of these could be stratified into baseflow (8) and event flow (10) categories. Samples were collected quarterly over a period of 9 years between 1978 and 1987. The relatively small pre-fire dataset is likely to result in bias in the load estimation, particularly in the event load estimations.

Post-fire: The Tambo river station at Bindi was decommissioned in June 2003, and as a result data is only available for the first six months of the year. Storm events were sampled comprehensively at this site over this period, with six of the eight storm events sampled. No data was collected in 2004 and 2005.

6.1.7 Tambo River at Swifts Creek

Pre-fire: Pre-fire: A total of 196 pre-fire sediment concentration values (99 baseflow and 97 event flow) collected monthly over a period of 24 years between 1978 and 2002 were available for analysis. Samples were initially collected quarterly, and then monthly from 1990 onwards. The relatively large pre-fire dataset, collected at fixed intervals over a long period, provides a good basis for application of the averaging estimation method.

Post-fire: Event sampling did not begin at this site until July 2003, following the decommissioning of the upstream Bindi site. The four major storm events for the second half of the year were sampled. The five major storms in 2004 were also sampled, and the six major storms in 2005 (as of November 2005) were sampled. The storm events for this site are well characterised by sampling over the three year period, apart from the initial gap prior to the installation of sampling equipment.

6.1.8 Wonnangatta River at Waterford

Pre-fire: A total of 47 pre-fire sediment concentration values (25 baseflow and 22 event flow) collected quarterly over a period of 24 years between 1978 and 1990 were available for analysis. The pre-fire sampling provides a limited dataset for load estimation using the

30 Impact of the 2003 Alpine Bushfires on Streamflow - Estimated changes in stream exports of sediment, phosphorus and nitrogen following the 2003 bushfires in Eastern Victoria

averaging method, and considerable uncertainty will be associated with pre-fire load estimates at this site.

Post-fire: Only the first two storm events of 2003 were sampled at this site, with no subsequent sampling over the 2003 to 2005 period.

6.2 Evaluation of the interpolation method

Inspection of Figure 2 to Figure 9 and summary Table 2 shows that except in a limited number of cases the interpolation method cannot be reasonably applied to the data in this study. This is largely due to the fact that:

• the pre-fire sampling regime for water quality is fixed-interval and the sample interval (2- 4 weeks) is large compared to the interval over which the concentration and discharge can be expected to vary (eg. hours to days)

• the post-fire storm event water quality data is incomplete (i.e. each storm event is not captured by the auto-samplers) therefore in many cases (though not all) the sample interval is large compared to the interval over which the concentration and discharge can be expected to vary (eg. hours to days)

Details of the pre-and post-fire datasets are presented in Sections 6.1.1 to 6.1.8.

The true (i.e. interpolated) loads generated from these catchments pre and post-fire are therefore not known, and more general estimation methods are required. Section 6.5 describes a case study of an application of the interpolation method, providing a comparison for the other load estimation methods in this study.

6.3 Evaluation of the regression method

The log of the TSS data were plotted against the log of 15 min discharge data (where available), and split into pre-fire data, post-fire (2003) data. The resulting TSS-discharge relationships are shown in Figure 11 to Figure 12. Except perhaps in one or two cases, these plots indicate that the water quality data are not suited to the regression method of load estimation.

31 Impact of the 2003 Alpine Bushfires on Streamflow - Estimated changes in stream exports of sediment, phosphorus and nitrogen following the 2003 bushfires in Eastern Victoria

100000 100000 100000 10000 10000 10000 1000 1000 1000 100 100 100 10 10 10 1 1 1 0.1 0.1 0.1 1 1 1 10 10 10 100 100 100 1000 1000 1000 10000 10000 10000 100000 100000 100000

Dargo River at Lower Dargo Rd Kiewa River at Bandiana Mitta Mitta River at Hinomunjie 100000 100000 100000 10000 10000 10000 1000 1000 1000 100 100 100 10 10 10 1 1 1 0.1 0.1 0.1 SS (mg/L) 1 1 1 10 10 10 100 100 100 1000 1000 1000 10000 10000 10000 100000 100000 100000

Mitchell River at Glendale Ovens River at Bright Snowy River at McKillop Bridge 100000 100000 100000 10000 10000 10000 1000 1000 1000 100 100 100 10 10 10 1 1 1 0.1 0.1 0.1 1 1 1 10 10 10 100 100 100 1000 1000 1000 10000 10000 10000 100000 100000 100000

Tambo River at Bindi Tambo River at Swifts Creek Wonnangatta River at Waterford Instantaneous Discharge (ML/Day)

Figure 11.Sediment rating curves based on pre-fire grab samples only.

100000 100000 100000 10000 10000 10000 1000 1000 1000 100 100 100 10 10 10 1 1 1 0.1 0.1 0.1 1 1 1 10 10 10 100 100 100 1000 1000 1000 10000 10000 10000 100000 100000 100000

Dargo River at Lower Dargo Rd Kiewa River at Bandiana Mitta Mitta River at Hinomunjie 100000 100000 100000 10000 10000 10000 1000 1000 1000 100 100 100 10 10 10 1 1 1 0.1 0.1 0.1 SS (mg/L) 1 1 1 10 10 10 100 100 100 1000 1000 1000 10000 10000 10000 100000 100000 100000

Mitchell River at Glendale Ovens River at Bright Snowy River at McKillop Bridge 100000 100000 100000 10000 10000 10000 1000 1000 1000 100 100 100 10 10 10 1 1 1 0.1 0.1 0.1 1 1 1 10 10 10 100 100 100 1000 1000 1000 10000 10000 10000 100000 100000 100000

Tambo River at Bindi Tambo River at Swifts Creek Wonnangatta River at Waterford Instantaneous Discharge (ML/Day)

Figure 12. Sediment rating curves based on all post-fire samples.

32 Impact of the 2003 Alpine Bushfires on Streamflow - Estimated changes in stream exports of sediment, phosphorus and nitrogen following the 2003 bushfires in Eastern Victoria

Results for the application of the regression method using stratification by rising and falling limb of the hydrograph, and by storm event are given in the appendix. The same stratifications were repeated replacing the 15 min discharge Q with the change in discharge ∆Q, with results also given in the appendix.

6.4 Estimated loads and uncertainty

The averaging method (Eq. 7) was selected for load estimation across all the sites because the interpolation and regression methods could not be used, and because the properties of the pre- fire and post-fire datasets made this the next-best available option. The limitations of this estimation method are detailed on a site by site basis in Sections 6.4.2 to 6.4.9.

Load estimates of TSS, TP, and TN using the averaging method (Eq. 7) are listed in the first section of Table 3, Table 4, and Table 5 respectively. Estimated loads using Monte Carlo simulations are given in the second half of these tables. The predicted loads for an unburnt condition are compared with each post-fire year, based on the pre-fire concentration data, and the observed discharge for 2003, 2004, or 2005. The proportion of total flow volume in each year categorised as baseflow (B) and event flow (E) is also listed. An example of the baseflow separation method used to determine these proportions is shown in Figure 13. The estimated relative change in load following the fire, (and the standard deviation of the relative change), is also listed. Further descriptive statistics for the input data, including standard errors and deviations and data ranges are given in the Appendix.

60

50

40

30

Flow (ML/day) 20

10

0

5 5 5 5 05 05 05 00 0 00 00 0 0 /2 /2 /2 /2 /2005 /2 /2 3 4/200 4 /0 /04 /0 /0 8/03 2 7 1 23/03 28 12/04/2005 17 22/04 27/04

Figure 13. An example of the application of the baseflow separation algorithm and definition to the outflow hydrograph. The upper blue line is the hydrograph, while the lower pink line is the baseflow component only. The arrows show periods defined as event flow. 33 Impact of the 2003 Alpine Bushfires on Streamflow - Estimated changes in stream exports of sediment, phosphorus and nitrogen following the 2003 bushfires in Eastern Victoria

Table 3. Summary of data used for the calculation of total suspended sediment loads using the averaging method and Monte Carlo simulation. Discharge (Q) is the annual total, and B and E are the proportions of total annual flow apportioned to baseflow and event flow, respectively. Note that the values for the Tambo at Bindi are for a six month period only.

Monte Carlo Method Baseflow Values Eventflow Values Arithmetic Method Log Normal Log Normal Logs of Baseflow Distribution Logs of Eventflow Distribution Total Suspended Solids Conc. (mg/L) Parameters Conc. (mg/L) Parameters Unburnt Baseflow Eventflow Unburnt Burnt Total Unburnt Burnt SD of Total Burnt Total SD of Conc. Conc. Total Q Fraction Fraction Total Load Unburnt Load Burnt Area (tonnes/k (tonnes/k Relative Relative Load Unburnt Load Relative Relative SiteCode Load Calcs. Year (mg/L) (mg/L) (ML) B E (tonnes) SD (tonnes) SD (km^2) m^2/yr) m^2/yr) Change Change Mean SD E(X) SD(X) Mean SD E(X) SD(X) (tonnes) SD (tonnes) Burnt SD Change Change 222209 Snowy Unburnt 2.09 12.02 0.64 0.44 2.08 0.97 1.14 1.06 6 8 222209 Snowy Burnt03 14.20 1498.00 238114 0.38 0.62 1962 480 222084 382074 10619 0.1848 20.91 113 197 2.61 0.35 14.40 5.16 5.40 2.46 4541 92644 1032 1415 589480 5019152 571 4927 222209 Snowy Burnt04 3.50 31.44 204809 0.33 0.67 1784 311 4530 4321 10619 0.1680 0.43 3 2 0.90 0.93 3.76 4.39 2.83 1.31 40 86 924 1315 5551 10728 6 14 222209 Snowy Burnt05 2.60 36.14 282957 0.34 0.66 2454 535 7026 4639 10619 0.2311 0.66 3 2 0.72 0.76 2.74 2.44 3.36 0.78 39 36 1267 1790 7403 6531 6 10 223202 Tambo @ Swifts Unburnt 4.10 5.75 0.76 0.71 2.74 2.20 1.05 0.93 4 5 223202 Tambo @ Swifts Burnt03 26.50 1139.07 35452 0.19 0.81 193 502 32967 77586 895 0.2156 36.83 171 600 2.79 1.44 46.24 122.24 5.31 1.90 1240 7468 145 150 33443 146511 230 1036 223202 Tambo @ Swifts Burnt04 6.40 1049.67 38076 0.22 0.78 205 527 31386 64410 895 0.2296 35.07 153 502 1.64 0.74 6.78 5.83 5.39 1.78 1070 5136 154 156 30805 113108 200 761 223202 Tambo @ Swifts Burnt05 2.25 170.31 44789 0.20 0.80 243 627 6101 10898 895 0.2711 6.80 25 79 0.45 0.90 2.34 2.59 4.32 1.31 176 374 183 188 6474 13797 35 84 223208 Tambo @ Bindi Unburnt 3.13 4.50 0.98 0.60 3.21 2.13 1.28 0.77 5 4 223208 Tambo @ Bindi Burnt03 2662.00 7069.84 3930 0.21 0.79 17 6 24147 35811 523 0.0316 46.20 1459 2165 7.65 0.87 3092.24 3313.37 7.98 1.36 7368 17149 18 14 26103 40333 1462 2534 223208 Tambo @ Bindi Burnt04 na na na 223208 Tambo @ Bindi Burnt05 na na na 224201 Wonnangatta Unburnt 2.84 8.27 0.88 0.56 2.84 1.73 1.58 0.91 7 8 224201 Wonnangatta Burnt03 na 28.27 492095 0.25 0.75 3413 1566 10490 11095 1979 1.7248 5.30 3 4 na na na na 2.67 1.38 38 91 3153 3356 13483 25210 4 9 224201 Wonnangatta Burnt04 na na 380926 0.20 0.80 2741 827 na na 1979 1.3848 na na na 224201 Wonnangatta Burnt05 na na 392306 0.12 0.88 2999 692 na na 1979 1.5154 na na na 224213 Dargo Unburnt 3.30 4.62 0.83 0.76 3.08 2.75 1.12 0.81 4 4 224213 Dargo Burnt03 6.67 119.66 164024 0.26 0.74 701 600 14854 33714 676 1.0376 21.97 21 51 1.85 0.40 6.88 2.87 3.02 1.89 123 725 646 509 13692 69116 21 108 224213 Dargo Burnt04 3.40 202.78 129901 0.24 0.76 559 451 20131 22810 676 0.8262 29.78 36 50 0.94 0.81 3.54 3.39 4.45 1.56 289 927 515 412 27553 121986 54 241 224213 Dargo Burnt05 3.25 220.89 156409 0.22 0.78 677 548 27188 29157 676 1.0020 40.22 40 54 1.07 0.52 3.33 1.85 4.81 1.17 244 421 623 507 28127 42020 45 77 401203 Mitta Mitta Unburnt 2.86 9.35 0.87 0.59 2.83 1.84 1.40 1.03 7 10 401203 Mitta Mitta Burnt03 101.90 1721.10 402824 0.28 0.72 3037 1409 511559 1716019 1533 1.9812 333.70 168 570 3.20 1.81 124.75 623.98 5.08 2.44 3167 62259 2279 3125 817192 6087376 359 2716 401203 Mitta Mitta Burnt04 12.17 305.93 367709 0.27 0.73 2794 1163 83374 128563 1533 1.8226 54.39 30 48 2.00 1.01 12.32 16.46 4.72 1.52 357 1085 2095 2892 96358 190020 46 111 401203 Mitta Mitta Burnt05 3.80 333.33 428286 0.27 0.73 3244 1520 104170 126137 1533 2.1158 67.95 32 42 0.98 0.98 4.30 5.42 4.99 1.50 448 1294 2440 3368 139424 343920 57 162 402205 Kiewa Unburnt 16.64 25.13 2.71 0.50 17.12 9.19 3.07 0.57 25 16 402205 Kiewa Burnt03 na 34.96 568351 0.19 0.81 13340 34837 17880 10862 1655 8.0603 10.80 1.3 4 2.71 0.50 17.12 9.19 3.33 0.72 36 30 13424 7469 18677 14110 1 1 402205 Kiewa Burnt04 na na 575975 0.20 0.80 13484 38358 na na na na na na na 402205 Kiewa Burnt05 na na 564721 0.15 0.85 13478 20228 na na na na na na na 403205 Ovens Unburnt 4.76 9.93 0.88 0.76 3.22 2.85 1.57 0.99 8 10 403205 Ovens Burnt03 2.86 260.99 193478 0.18 0.82 1736 3145 41260 49503 495 3.5080 83.35 24 52 0.87 0.68 3.01 2.29 4.93 1.26 304 598 1337 1432 50991 109088 38 91 403205 Ovens Burnt04 1.25 89.61 172052 0.18 0.82 1548 2795 12678 12341 495 3.1280 25.61 8 17 0.17 0.35 1.26 0.45 3.94 1.21 106 192 1189 1273 16273 29102 14 29 403205 Ovens Burnt05 2.00 22.00 136907 0.16 0.84 1250 2252 2588 2177 495 2.5251 5.23 2 4 0.55 0.63 2.12 1.49 2.78 0.91 24 28 957 1037 2934 2844 3 4

34 Impact of the 2003 Alpine Bushfires on Streamflow - Estimated changes in stream exports of sediment, phosphorus and nitrogen following the 2003 bushfires in Eastern Victoria

Table 4. Summary of data used for the calculation of total phosphorus loads using the averaging method and Monte Carlo simulation. Discharge (Q) is the annual total, and B and E are the proportions of total annual flow apportioned to baseflow and event flow, respectively. Note that the values for the Tambo at Bindi are for a six month period only.

Monte Carlo Method Baseflow Values Eventflow Values Arithmetic Method Log Normal Log Normal Logs of Baseflow Distribution Logs of Eventflow Distribution Total Phosphorous Conc. (mg/L) Parameters Conc. (mg/L) Parameters Unburnt Baseflow Eventflow Unburnt Burnt Total Unburnt Burnt SD of Total Burnt Total SD of Conc. Conc. Total Q Fraction Fraction Total Load Unburnt Load Burnt Area (tonnes/k (tonnes/k Relative Relative Load Unburnt Load Relative Relative SiteCode Load Calcs. Year (mg/L) (mg/L) (ML) B E (tonnes) SD (tonnes) SD (km^2) m^2/yr) m^2/yr) Change Change Mean SD E(X) SD(X) Mean SD E(X) SD(X) (tonnes) SD (tonnes) Burnt SD Change Change 222209 Snowy Unburnt 0.01 0.02 -4.49 0.61 0.01 0.01 -4.19 0.77 0.02 0.02 222209 Snowy Burnt03 0.05 3.35 238114 0.38 0.62 4.56 2.02 498.23 766.46 10619 0.0004 0.0469 109 175 -3.10 0.59 0.05 0.03 -0.44 2.28 8.76 118.77 4.22 2.80 1178.81 7743.93 279 1844 222209 Snowy Burnt04 0.03 0.05 204809 0.33 0.67 4.00 1.78 8.27 4.03 10619 0.0004 0.0008 2 1 -3.80 0.59 0.03 0.02 -3.22 0.63 0.05 0.03 3.69 2.57 8.64 4.78 2 2 222209 Snowy Burnt05 0.02 0.06 282957 0.34 0.66 5.52 2.45 13.09 5.24 10619 0.0005 0.0012 2 1 -4.02 0.38 0.02 0.01 -2.90 0.46 0.06 0.03 5.08 3.50 13.18 5.72 32 223202 Tambo @ Swifts Unburnt 0.02 0.02 -4.36 0.60 0.02 0.01 -4.05 0.70 0.02 0.02 223202 Tambo @ Swifts Burnt03 0.08 0.79 35452 0.19 0.81 0.77 0.47 23.17 35.55 895 0.0009 0.0259 30 50 -2.85 0.78 0.08 0.07 -1.38 1.53 0.82 2.53 0.72 0.54 29.66 244.01 41 342 223202 Tambo @ Swifts Burnt04 0.02 0.74 38076 0.22 0.78 0.82 0.49 22.32 42.40 895 0.0009 0.0249 27 54 -4.02 0.61 0.02 0.01 -1.71 1.68 0.74 2.97 0.76 0.56 21.59 49.15 28 68 223202 Tambo @ Swifts Burnt05 0.06 0.14 44789 0.20 0.80 0.97 0.58 5.34 6.51 895 0.0011 0.0060 6 7 -3.32 1.00 0.06 0.08 -2.43 0.85 0.13 0.13 0.90 0.67 5.07 4.92 6 7 223208 Tambo @ Bindi Unburnt 0.02 0.02 -4.25 0.50 0.02 0.01 -4.03 0.30 0.02 0.01 223208 Tambo @ Bindi Burnt03 3.94 8.77 3930 0.21 0.79 0.07 0.03 30.47 57.18 523 0.0001 0.0580 431 835 1.25 0.59 4.15 2.67 1.24 1.28 7.84 15.90 0.07 0.02 26.35 45.34 366 638 223208 Tambo @ Bindi Burnt04 na na na 223208 Tambo @ Bindi Burnt05 na na na 224201 Wonnangatta Unburnt 0.04 0.04 -3.60 0.83 0.04 0.04 -3.61 0.85 0.04 0.04 224201 Wonnangatta Burnt03 na 0.06 492095 0.25 0.75 19.15 12.47 22.49 23.24 1979 0.0097 0.0114 1 1 na na na na -3.16 0.82 0.06 0.06 19.26 14.98 22.14 23.92 12 224201 Wonnangatta Burnt04 na na 380926 0.20 0.80 14.85 10.02 na na 1979 0.0075 na na na 224201 Wonnangatta Burnt05 na na 392306 0.12 0.88 15.34 11.13 na na 1979 0.0078 na na na 224213 Dargo Unburnt 0.02 0.02 -4.27 0.53 0.02 0.01 -4.16 0.52 0.02 0.01 224213 Dargo Burnt03 0.03 0.22 164024 0.26 0.74 2.84 1.06 27.46 54.74 676 0.0042 0.0406 10 20 -3.64 0.27 0.03 0.01 -2.98 1.69 0.21 0.87 2.87 1.26 24.89 81.77 9 29 224213 Dargo Burnt04 0.01 0.25 129901 0.24 0.76 2.25 0.86 24.98 30.08 676 0.0033 0.0370 11 14 -4.86 0.50 0.01 0.00 -2.14 1.32 0.28 0.61 2.28 1.01 28.34 58.98 12 27 224213 Dargo Burnt05 0.02 0.23 156409 0.22 0.78 2.72 1.05 28.12 32.35 676 0.0040 0.0400 10 13 -4.25 0.43 0.02 0.01 -2.04 1.04 0.22 0.32 2.75 1.24 27.26 43.25 10 16 401203 Mitta Mitta Unburnt 0.02 0.04 -4.16 0.65 0.02 0.01 -3.86 0.76 0.03 0.02 401203 Mitta Mitta Burnt03 0.72 1.42 402824 0.28 0.72 13.49 7.74 492.63 538.79 1533 0.0088 0.3214 37 45 -2.10 1.86 0.69 3.84 -0.77 1.75 2.14 9.66 10.31 7.04 671.13 2114.96 65 210 401203 Mitta Mitta Burnt04 0.02 0.27 367709 0.27 0.73 12.38 7.11 73.67 104.27 1533 0.0081 0.0481 6 9 -3.93 0.66 0.02 0.02 -2.09 1.25 0.27 0.52 9.44 6.50 81.04 149.84 9 17 401203 Mitta Mitta Burnt05 0.02 0.33 428286 0.27 0.73 14.39 8.26 104.61 110.43 1533 0.0094 0.0682 7 9 -3.94 0.26 0.02 0.01 -1.74 1.22 0.37 0.68 11.00 7.57 113.30 195.93 10 19 402205 Kiewa Unburnt 0.03 0.05 -3.64 0.43 0.03 0.01 -3.12 0.61 0.05 0.04 402205 Kiewa Burnt03 0.02 0.04 568351 0.19 0.81 28.08 7.25 21.33 12.82 1655 0.0170 0.0129 1 0.5 -3.93 0.62 0.02 0.02 -3.36 0.56 0.04 0.02 27.97 16.55 21.23 10.57 1 0.6 402205 Kiewa Burnt04 0.02 0.03 575975 0.20 0.80 28.35 7.31 18.02 9.69 1655 0.0171 0.0109 1 0.4 -4.00 0.54 0.02 0.01 -3.56 0.63 0.03 0.02 28.20 16.57 18.19 10.84 1 0.5 402205 Kiewa Burnt05 0.03 0.04 564721 0.15 0.85 28.57 7.51 23.74 11.48 1655 0.0173 0.0143 1 0.5 -3.58 0.44 0.03 0.01 -3.25 0.54 0.04 0.03 28.33 17.22 24.16 12.69 1 0.7 403205 Ovens Unburnt 0.03 0.03 -4.14 0.65 0.02 0.01 -3.79 0.62 0.03 0.02 403205 Ovens Burnt03 0.02 0.33 193478 0.18 0.82 5.85 13.76 53.12 82.07 495 0.0118 0.1073 9 26 -4.03 0.31 0.02 0.01 -1.76 1.10 0.32 0.48 5.13 3.10 48.16 71.09 9 15 403205 Ovens Burnt04 0.01 0.12 172052 0.18 0.82 5.20 12.29 17.89 14.77 495 0.0105 0.0361 3 9 -4.49 0.26 0.01 0.00 -2.43 0.87 0.13 0.14 4.56 2.75 18.90 18.694 5 403205 Ovens Burnt05 0.01 0.04 136907 0.16 0.84 4.15 10.01 5.38 2.82 495 0.0084 0.0109 1 3 -4.35 0.35 0.01 0.00 -3.26 0.55 0.04 0.03 3.65 2.24 5.46 3.13 1 7

35 Impact of the 2003 Alpine Bushfires on Streamflow - Estimated changes in stream exports of sediment, phosphorus and nitrogen following the 2003 bushfires in Eastern Victoria

Table 5. Summary of data used for the calculation of total nitrogen loads using the averaging method and Monte Carlo simulation. Discharge (Q) is the annual total, and B and E are the proportions of total annual flow apportioned to baseflow and event flow, respectively. Note that the values for the Tambo at Bindi are for a six month period only.

Monte Carlo Method Baseflow Values Eventflow Values Arithmetic Method Log Normal Log Normal Logs of Baseflow Distribution Logs of Eventflow Distribution Total Nitrogen Conc. (mg/L) Parameters Conc. (mg/L) Parameters Unburnt Baseflow Eventflow Unburnt Burnt Total Unburnt Burnt SD of Total Burnt Total SD of Conc. Conc. Total Q Fraction Fraction Total Load Unburnt Load Burnt Area (tonnes/k (tonnes/k Relative Relative Load Unburnt Load Relative Relative SiteCode Load Calcs. Year (mg/L) (mg/L) (ML) B E (tonnes) SD (tonnes) SD (km^2) m^2/yr) m^2/yr) Change Change Mean SD E(X) SD(X) Mean SD E(X) SD(X) (tonnes) SD (tonnes) Burnt SD Change Change 222209 Snowy Unburnt 0.27 0.34 -1.43 0.49 0.27 0.14 -1.22 0.51 0.34 0.19 222209 Snowy Burnt03 4.23 10.12 238114 0.38 0.62 74.76 32.92 1875.29 1809.39 10619 0.0070 0.1766 25 27 0.35 1.78 6.92 32.90 1.57 1.38 12.42 29.46 73.20 30.33 2656.94 12035.98 36 165 222209 Snowy Burnt04 0.47 1.57 204809 0.33 0.67 64.95 27.61 246.93 328.31 10619 0.0061 0.0233 4 5 -0.87 0.55 0.49 0.29 -0.15 0.97 1.38 1.74 63.56 27.29 227.03 218.40 4 4 222209 Snowy Burnt05 0.47 0.76 282957 0.34 0.66 89.67 39.20 187.69 72.10 10619 0.0084 0.0177 2 1 -0.93 0.65 0.49 0.35 -0.33 0.33 0.76 0.25 87.65 37.35 186.12 59.41 2 1 223202 Tambo @ Swifts Unburnt 0.25 0.31 -1.53 0.49 0.25 0.13 -1.33 0.53 0.30 0.17 223202 Tambo @ Swifts Burnt03 0.84 4.11 35452 0.19 0.81 10.57 4.81 123.93 133.60 895 0.0118 0.1385 12 14 -0.35 0.70 0.90 0.71 0.89 1.02 4.10 5.54 10.33 5.17 126.29 163.42 12 17 223202 Tambo @ Swifts Burnt04 0.24 3.55 38076 0.22 0.78 11.29 5.05 107.82 164.09 895 0.0126 0.1205 10 15 -1.46 0.29 0.24 0.07 0.34 1.33 3.41 7.49 11.02 5.37 99.88 176.65 9 17 223202 Tambo @ Swifts Burnt05 0.18 1.04 44789 0.20 0.80 13.32 6.00 38.69 41.47 895 0.0149 0.0430 3 3 -1.79 0.55 0.19 0.12 -0.26 0.69 0.98 0.76 13.02 6.46 36.33 31.94 3 3 223208 Tambo @ Bindi Unburnt 0.29 0.41 -1.30 0.40 0.29 0.12 -0.96 0.38 0.41 0.16 223208 Tambo @ Bindi Burnt03 22.60 39.85 3930 0.21 0.79 1.51 0.35 142.38 271.54 523 0.0029 0.2720 94 181 3.05 0.42 22.98 9.99 2.82 1.16 32.70 54.88 1.54 0.54 122.90 136.19 80 93 223208 Tambo @ Bindi Burnt04 na na na 223208 Tambo @ Bindi Burnt05 na na na 224201 Wonnangatta Unburnt 0.23 0.38 -1.56 0.46 0.23 0.11 -1.11 0.53 0.38 0.22 224201 Wonnangatta Burnt03 na 0.57 492095 0.25 0.75 167.43 41.90 210.13 134.71 1979 0.0846 0.1062 1 1 na na na na -0.76 0.64 0.57 0.41 171.91 81.74 208.44 141.23 1 1 224201 Wonnangatta Burnt04 na na 380926 0.20 0.80 132.24 32.22 na na 1979 0.0668 na na na 224201 Wonnangatta Burnt05 na na 392306 0.12 0.88 140.92 35.09 na na 1979 0.0712 na na na 224213 Dargo Unburnt 0.17 0.18 -1.92 0.58 0.17 0.11 -1.88 0.52 0.17 0.10 224213 Dargo Burnt03 0.31 1.45 164024 0.26 0.74 29.00 13.09 189.51 299.57 676 0.0429 0.2803 7 11 -1.22 0.35 0.31 0.11 -0.41 1.13 1.25 1.99 28.45 12.77 169.61 207.22 6 8 224213 Dargo Burnt04 0.13 1.70 129901 0.24 0.76 22.98 10.46 172.03 191.43 676 0.0340 0.2545 7 9 -2.06 0.21 0.13 0.03 -0.01 1.04 1.70 2.38 22.53 10.26 175.29 397.52 8 18 224213 Dargo Burnt05 0.16 1.14 156409 0.22 0.78 27.71 12.83 145.64 156.48 676 0.0410 0.2200 5 6 -1.83 0.06 0.16 0.01 -0.36 0.98 1.13 1.44 27.13 12.54 139.26 167.01 5 7 401203 Mitta Mitta Unburnt 0.19 0.31 -1.85 0.56 0.18 0.11 -1.58 0.72 0.27 0.22 401203 Mitta Mitta Burnt03 2.25 6.56 402824 0.28 0.72 110.68 41.77 2159.16 2301.86 1533 0.0722 1.4085 20 22 -0.02 1.32 2.34 5.06 1.08 1.39 7.74 18.75 95.67 62.75 2323.67 4050.39 24 45 401203 Mitta Mitta Burnt04 0.32 2.08 367709 0.27 0.73 101.44 38.08 590.19 664.43 1533 0.0662 0.3850 6 7 -1.30 0.60 0.33 0.22 0.25 0.96 2.03 2.49 87.61 57.94 588.70 687.40 7 9 401203 Mitta Mitta Burnt05 0.18 1.61 428286 0.27 0.73 117.95 44.59 521.73 446.83 1533 0.0769 0.3403 4 4 -1.76 0.29 0.18 0.05 0.09 0.91 1.66 1.87 102.04 67.49 552.11 541.09 5 6 402205 Kiewa Unburnt 0.28 0.48 -1.38 0.46 0.28 0.14 -0.92 0.59 0.47 0.31 402205 Kiewa Burnt03 0.37 0.70 568351 0.19 0.81 249.43 72.99 361.17 622.74 1655 0.1507 0.2182 1 3 -1.15 0.58 0.38 0.24 -0.73 0.63 0.59 0.42 252.83 146.56 320.75 206.98 1 1 402205 Kiewa Burnt04 0.34 0.57 575975 0.20 0.80 251.98 73.94 302.17 141.57 1655 0.1523 0.1826 1 1 -1.20 0.48 0.34 0.17 -0.70 0.55 0.58 0.35 255.03 146.79 313.67 157.91 1 1 402205 Kiewa Burnt05 0.27 0.68 564721 0.15 0.85 253.02 73.91 349.89 33.19 1655 0.1529 0.2114 1 0.4 -1.32 0.22 0.27 0.06 -0.39 0.03 0.68 0.02 255.86 152.49 350.05 11.10 1 1 403205 Ovens Unburnt 0.20 0.23 -1.73 0.46 0.20 0.10 -1.60 0.47 0.23 0.11 403205 Ovens Burnt03 0.23 2.53 193478 0.18 0.82 43.27 18.62 407.38 433.74 495 0.0874 0.8230 9 11 -1.50 0.21 0.23 0.05 0.49 0.92 2.50 2.86 42.53 17.45 390.25 436.16 9 11 403205 Ovens Burnt04 0.17 1.16 172052 0.18 0.82 38.50 16.61 168.74 191.12 495 0.0778 0.3409 4 5 -1.85 0.40 0.17 0.07 -0.10 0.66 1.12 0.83 37.82 15.52 162.44 121.87 4 4 403205 Ovens Burnt05 0.28 0.50 136907 0.16 0.84 30.74 13.49 63.31 44.10 495 0.0621 0.1279 2 2 -1.27 0.07 0.28 0.02 -0.84 0.46 0.48 0.23 30.17 12.60 61.17 26.99 2 66

36 Impact of the 2003 Alpine Bushfires on Streamflow - Estimated changes in stream exports of sediment, phosphorus and nitrogen

6.4.1 Uncertainty estimates using Monte Carlo simulation

An example of the estimation of the log normal distribution parameters is shown in Figure 14. An example of how sensitivity analysis was used to determine the size of the sample used for the Monte Carlo simulation is shown in Figure 15. Figure 16 (TSS), Figure 17 (TP) and Figure 18 (TN) show the annual load estimates for predicted unburnt, 2003, 2004, and 2005 using Monte Carlo simulation methods. Parameters for the simulation are shown in Table 3, Table 4, and Table 5. Box and whiskers are the 10, 25, 75, and 90 percentiles. The dashed line is the mean and the solid line is the median.

100

80

60 Number of obs Number 40

20

0 0 200 400 600 800 1000 1200 1400 1600 1800 Ovens (Site 403205) 2003 Eventflow

Figure 14. An example of the log-normal distribution fitted (line) to the TSS (mg/L) data (bars) from the Ovens River in 2003. Impact of the 2003 Alpine Bushfires on Streamflow - Estimated changes in stream exports of sediment, phosphorus and nitrogen following the 2003 bushfires in Eastern Victoria

0.35

0.30

0.25

0.20

0.15 Sample Mean 0.10

0.05

0.00

0 10000 20000 30000 40000 50000 60000 Sample Size

Figure 15. An example of a sensitivity analysis for sample size using Monte Carlo simulation with mean zero and standard deviation 1.

38 Impact of the 2003 Alpine Bushfires on Streamflow - Estimated changes in stream exports of sediment, phosphorus and nitrogen following the 2003 bushfires in Eastern Victoria

1x106 Snowy Tambo @ Swifts 800x103 (222209) (223202) 60x103 600x103

400x103

3 200x10 40x103

20x103

20x103

0 0 UnBT03 BT03 UnBT04 BT04 UnBT05 BT05 UnBT03 BT03 UnBT04 BT04 UnBT05 BT05

60x103 Tambo @ Bindi 30x103 Wonnangatta (223208) (224201) 50x103 25x103

40x103 20x103

30x103 15x103

20x103 10x103

10x103 5x103

0 0 UnBT03 BT03 UnBT03 BT03

Dargo 6 Mitta Mitta 60x103 1x10 (224213) (401203)

40x103

3 Total Suspended Solids (tonnes) Solids Suspended Total 500x10

20x103

0 0 UnBT03 BT03 UnBT04 BT04 UnBT05 BT05 UnBT03 BT03 UnBT04 BT04 UnBT05 BT05

40x103 120x103 Kiewa Ovens

(402205) 100x103 (403205) 30x103 80x103

20x103 60x103

40x103 10x103 20x103

0 0 UnBT03 BT03 UnBT03 BT03 UnBT04 BT04 UnBT05 BT05 Sample Period (year)

Figure 16. Total suspended solids annual load estimates for predicted unburnt (UnBT), 2003, 2004, and 2005 (BT) using Monte Carlo simulation methods. Box and whiskers are the 10, 25, 75, and 90 percentiles. The dashed line is the mean and the solid line is the median. Note that the values for the Tambo at Bindi are for a six month period only. 39 Impact of the 2003 Alpine Bushfires on Streamflow - Estimated changes in stream exports of sediment, phosphorus and nitrogen following the 2003 bushfires in Eastern Victoria

2000 60 Snowy Tambo @ Swifts (222209) 50 (223202) 1500 40

1000 30

400 20

10

0 0 UnBT03 BT03 UnBT04 BT04 UnBT05 BT05 UnBT03 BT03 UnBT04 BT04 UnBT05 BT05

60 50 Tambo @ Bindi Wonnangatta

50 (223208) (224201) 40

40 30 30

20 20

10 10

0 0 UnBT03 BT03 UnBT03 BT03

60 Dargo 1500 Mitta Mitta (224213) (401203) 50

40 1000 Total Phosphorous (tonnes) Phosphorous Total 30

20 500

10

0 0 UnBT03 BT03 UnBT04 BT04 UnBT05 BT05 UnBT03 BT03 UnBT04 BT04 UnBT05 BT05

60 120 Kiewa Ovens

50 (402205) 100 (403205)

40 80

30 60

20 40

10 20

0 0 UnBT03 BT03 UnBT04 BT04 UnBT05 BT05 UnBT03 BT03 UnBT04 BT04 UnBT05 BT05 Sample Period (year)

Figure 17. Total phosphorus annual load estimates for predicted unburnt (UnBT), 2003, 2004, and 2005 (BT) using Monte Carlo simulation methods. Box and whiskers are the 10, 25, 75, and 90 percentiles. The dashed line is the mean and the solid line is the median. Note that the values for the Tambo at Bindi are for a six month period only. 40 Impact of the 2003 Alpine Bushfires on Streamflow - Estimated changes in stream exports of sediment, phosphorus and nitrogen following the 2003 bushfires in Eastern Victoria

300 5000 Snowy Tambo @ Swifts 4500 (222209) 250 (223202) 4000 3500 200 3000 150 2500 2000 100 1500

1000 50 500 0 0 UnBT03 BT03 UnBT04 BT04 UnBT05 BT05 UnBT03 BT03 UnBT04 BT04 UnBT05 BT05

Tambo @ Bindi 400 Wonnangatta 250 (223208) (224201)

200 300

150 200

100

100 50

0 0 UnBT03 BT03 UnBT03 BT03

5500 400 Dargo 5000 Mitta Mitta (224213) (401203) 4500 300 4000 Total Nitrogen (tonnes) Total 3500 3000 200 2500 2000 1500 100 1000 500 0 0 UnBT03 BT03 UnBT04 BT04 UnBT05 BT05 UnBT03 BT03 UnBT04 BT04 UnBT05 BT05

700 Kiewa 800 Ovens 600 (402205) (403205)

500 600

400 400 300

200 200 100

0 0 UnBT03 BT03 UnBT04 BT04 UnBT05 BT05 UnBT03 BT03 UnBT04 BT04 UnBT05 BT05 Sample Period (year)

Figure 18. Total nitrogen annual load estimates for predicted unburnt (UnBT), 2003, 2004, and 2005 (BT) using Monte Carlo simulation methods. Box and whiskers are the 10, 25, 75, and 90 percentiles. The dashed line is the mean and the solid line is the median. Note that the values for the Tambo at Bindi are for a six month period only.

41 Impact of the 2003 Alpine Bushfires on Streamflow - Estimated changes in stream exports of sediment, phosphorus and nitrogen following the 2003 bushfires in Eastern Victoria

There are several key points to note with respect to the load estimates presented in Figure 16, Figure 17 and Figure 18, and in Table 3, Table 4, and Table 5.

1. There is very high variability between catchments, with relative increases in loads varying from about a 1500 factor increase in TSS from the Tambo at Bindi, to a 1-3 factor increase from the Kiewa and Wonnangatta Rivers.

2. There is a very high stochastic uncertainty on the estimates of loads, and in many cases the standard deviation of the estimated load is greater than the mean value. Measurement uncertainty due to sampling error is additional to the listed error values.

3. There is variation in estimated loads depending on the load estimation method used, (e.g. arithmetic mean vs Monte Carlo simulation using the averaging method), particularly when the data are strongly skewed and affected by extreme values.

4. The relative post-fire change in TSS loads are generally greater than the relative change in TP loads, which in turn are greater than the relative increase in TN loads.

5. Recovery is highly variable between the different catchments.

The very high variability in load estimates between the catchments, combined with variable levels of measurement uncertainty in the input datasets described in Sections 6.1.1 to 6.1.8, and variable levels of stochastic uncertainty associated with the predicted loads, indicates that only limited broad generalisations regarding the effects of the fires on water quality are possible. Each catchment needs to be considered on an individual basis, taking into account the factors listed above. The following sections provide a detailed analysis of the estimated loads on a catchment by catchment basis.

6.4.2 Dargo River at Lower Dargo Road

A 20 fold increase in TSS load from the Dargo River in 2003 is largely attributable to the 26 fold increase in the EMC in the first year after the fire. The DWC only increased by a factor of 2 and given that approximately 75% of discharge is event flow at this station, it is clear that increases in event flow concentrations are the dominant driver for increased loads in the first year. While the standard deviations on the load estimates are high, (being roughly equivalent to the mean for the pre-fire load, and approximately double the mean for the 2003 load), this is not due to a lack of data, with 150 pre-fire samples and 108 samples in 2003. Figure 2 indicates that while a substantial number of storm events were not sampled, the events that were captured were spread over a range of storm event sizes. This supports the assertion that the EMC samples were representative of storm event conditions. This is an important consideration given the strong influence of the EMC on the total load estimates. 42 Impact of the 2003 Alpine Bushfires on Streamflow - Estimated changes in stream exports of sediment, phosphorus and nitrogen following the 2003 bushfires in Eastern Victoria

Increases in TP and TN were 10 and 7 fold respectively in 2003, with similar levels of uncertainty, and equivalent sample numbers. The relative increase in TSS, TP and TN were of similar magnitude regardless of wether the arithmetic mean or the Monte Carlo simulation was used. This suggests that the knowledge uncertainty associated with the choice of model for this site is low.

The lower proportional increase in TP, compared to TSS, suggests that post-fire sediments include a lower proportion of TP than pre-fire sediments. Given that TP transport in these streams is likely to be predominately in the clay adsorbed form, it is likely that these results indicate an increase in the proportion of coarse sediments, which include less adsorption sites per unit mass than clay sediments.

There is no evidence of recovery in water quality over the 3 post-fire years. Although the mean baseflow concentrations of TSS declined after 2003, the EMCs approximately doubled leading to higher loads and changes relative to unburnt conditions. TP mean concentrations, loads and relative changes remained fairly constant between 2003 and 2005, and although there is more variation in the TN data, the temporal trends are the same. It may be that the poor storm event sampling from August-December in 2003, during which high concentrations were measured from the Tambo at Swifts Creek and the Mitta Mitta, has resulted in an underestimate of the 2003 impacts. The standard deviations of loads are approximately equal to the means for 2004 and 2005, which is low compared to the other sites, but is still high in an absolute sense.

In summary, the number of pre- and post-fire samples, combined with the representativeness of the captured post-fire storm events, provides a reasonable level of confidence in the magnitude of the load estimates for the Dargo River.

6.4.3 Kiewa River at Bandiana

The TSS data for the Kiewa River at Bandiana are poor both before and after the fire. A total of 19 pre-fire fixed interval samples suggests that pre-fire estimated loads would be strongly negatively biased. While a considerable number (93) of samples were collected in 2003, none of these were from large storm events, suggesting that the post-fire EMC is likely to be substantially underestimated.

The Kiewa River at Bandiana dataset differs from the other sites in this study in that there are few TSS data, but a large dataset for TP and TN. 605 pre-fire values are available for TP and TN from the pre-fire period. Weekly fixed-interval sampling for TP and TN continued after the fire, providing approximately 163 samples in addition to the 93 event triggered samples

43 Impact of the 2003 Alpine Bushfires on Streamflow - Estimated changes in stream exports of sediment, phosphorus and nitrogen following the 2003 bushfires in Eastern Victoria

described above. This total number of samples collected before and after the fire provides some confidence in the load estimates. However, the failure to capture the largest post-fire storm events will probably result in a negative bias in the post-fire load calculations.

The load estimates for the Kiewa River show no increase in loads of TP and TN in the post-fire period. The Bandiana gauging site on the Kiewa River is located further down the catchment than the other monitoring sites, and numerous water impoundments exist on the major branches of the Kiewa prior to the monitoring site. While the uncertainty associated with the TP load estimates is smaller than found for the other sites, the uncertainty in the TN load estimate after the fire is large relative to the mean, possibly reflecting the dissolved transport phase of TN.

The large TP dataset can be used as an indicator of change in TSS loads in the absence of comprehensive TSS data for this site. Given the lack of increase in TP loads, it can be reasonably assumed that the TSS loads were also relatively unchanged following the fire. There is no sign of any pattern of change in water quality in the years 2003 to 2005.

In summary, the large and continuous fixed-interval sample dataset, both before and after the fire, indicates that the loads of TSS, TP and TN from the Kiewa River at Bandiana were relatively unchanged by the fire, and have remained this way over the recovery period.

6.4.4 Mitta Mitta River at Hinomunjie

The Mitta Mitta recorded a 168 fold increase in estimated TSS load using the averaging approach and a 359 fold increase using the Monte Carlo simulation. The variation in the estimated load due to the different estimation methods indicates a high knowledge uncertainty. Inspection of the Monte Carlo simulation results for this site (Figure 16) show the mean TSS load estimate is greater than the 75th percentile of the load estimates, suggesting that the concentration data is strongly skewed by a few very large concentration values. The maximum TSS concentration at this site was extremely high, at 43,000 mg/L, measured in December 2003. The concentration ½ an hour later was still 32,000 mg/L, indicating a major erosion event, and perhaps explaining the very high standard deviation at this site. The presence of these few, extreme TSS concentration values, dominates the load estimates by substantially increasing the EMC values, and highlights the important leverage that a few, extreme events can have on the total load estimates. As a result the stochastic uncertainty is very high and the standard deviation of the relative change in the TSS load is 570.

The pre-fire dataset of 189 fixed-interval samples for the Mitta Mitta provides a high level of confidence in the pre-fire load estimates, which is reflected in the relatively low standard deviation (compared to the other sites) on these estimates of about ½ the mean value using the

44 Impact of the 2003 Alpine Bushfires on Streamflow - Estimated changes in stream exports of sediment, phosphorus and nitrogen following the 2003 bushfires in Eastern Victoria averaging method. The 36 fold increase in the DWC after the fire, from 3 mg/L to 102 mg/L is based on only 10 samples, and may be higher than the true baseflow values due to the inherent inaccuracies in separating event and baseflow conditions at the beginning and end of storm events. However this error will have little effect on the load estimates as the baseflow only comprises 27% of the total flow, and the magnitude of the increase in the EMC (185 fold) is the dominant driver of the increased TSS loads. The 91 samples used to calculate the EMC in 2003 were collected from about ½ the storms for that year. However the events cover the spectrum of event magnitudes, which may increase the representativeness of the event samples. Interestingly, the events with the highest TSS concentrations did not involve particularly high peak discharges. This illustrates the difficulties in trying to predict sediment concentrations from discharge data, as is the case with the regression estimation method. This observation also highlights the need to capture even relatively small hydrologic events in order to best estimate TSS loads.

Sampling for TP and TN was the same as for TSS at this site, so conclusions regarding the validity and uncertainty in the TSS load estimates apply equally to the TP and TN load estimates. Using the averaging method, 37 fold and 20 fold increases in TP and TN loads were estimated, respectively. Uncertainties in the pre-fire loads were relatively low, while for post-fire loads the standard deviation was approximately double, and equal to, the mean values for TP and TN respectively.

In summary, the pre-fire dataset provides a high degree of confidence in the pre-fire TSS load estimates for the Mitta Mitta. The post-fire TSS loads are clearly strongly increased following the fire, with estimates ranging from 168 to 359 fold increases, and with extremely high maximum values across several storm events. However the uncertainty in the post-fire load estimate is very high, largely due to the presence of these few extreme values.

Both the averaging and Monte Carlo methods indicate that the heavy impacts of 2003 have subsequently diminished. However the data still indicate a 30 fold increase from estimated TSS loads under unburnt conditions in 2004 and 2005. TP and TN loads decreased by similar magnitudes (4-6 fold) from 2003, and are estimated to be 7 and 4 times the unburnt condition in 2005. There is no significant downward trend in concentrations and loads from 2004 to 2005. The high concentrations still measured during the event in December 2005 (Fig. 4) are notable.

45 Impact of the 2003 Alpine Bushfires on Streamflow - Estimated changes in stream exports of sediment, phosphorus and nitrogen following the 2003 bushfires in Eastern Victoria

6.4.5 Ovens River at Bright

A large pre-fire dataset of 189 samples provides a high level of confidence in the pre-fire estimates of TSS, TP and TN loads. The estimated 24, 9 and 9 fold increase in loads of TSS, TP and TN respectively in 2003 is based on 105 samples from 30% of the storm events over the year, including the two largest storms and three of the smallest storm events. Data from an ongoing manual fixed interval sampling program provides confidence in the DWC values after the fire, though once again most of the leverage on the loads is associated with the 26 fold increase in the TSS EMC values. The SD for the 2003 TSS & TN loads are approximately equivalent to the mean, while for the TP load estimate the SD is approximately twice the mean indicating considerable uncertainty in the 2003 load estimate. Results from the different estimation methods indicates a low level of knowledge uncertainty, particularly for the TP and TN estimates.

Both the averaging and Monte Carlo methods indicate a diminution of load changes and concentrations from 2003-2005. The SDs are approximately equal to the mean for 2004 and 2005, suggesting greater statistical certainty in the results. However, sampling frequency is very low for 2005, resulting in a large measurement uncertainty. The estimated changes for 2005 are around double for all parameters, but the data are sparse.

6.4.6 Snowy River at Mc Killops Bridge

The pre-fire dataset for the Snowy River is good with 149 samples for TSS, TP and TN, illustrated by the SD of the TSS load estimate which is only ¼ of the mean value. The 7 fold increase in DWC for TSS is reasonable given that the samples were collected routinely throughout the year. The 2003 post-fire storm event dataset is very poor with only 13% of the storm events sampled, and none of the larger events sampled. As a result, the sampling uncertainty (ie a form of measurement uncertainty) associated with the post-fire event load estimate is likely to be very high (though is of course unknown) at this site. The load estimates for TSS, TP and TN are therefore highly uncertain and should be considered and used with extreme caution.

The sampling regime from this site has been very variable, with the highest proportion of events measured in 2004. However the distribution of these measured events (Fig. 6) is far from satisfactory for characterisation of impacts. There were no winter events captured in any year, and the 2003 data are dominated by the immediate post-fire event. Although the SD of the relative change in loads is not as high as for some of the other sites, this may simply reflect the poor sampling at this site. It is very difficult to draw conclusions or even summarise the data

46 Impact of the 2003 Alpine Bushfires on Streamflow - Estimated changes in stream exports of sediment, phosphorus and nitrogen following the 2003 bushfires in Eastern Victoria

from this catchment. Anecdotal evidence has suggested that there are still high concentrations of TSS during storm events.

6.4.7 Tambo River at Bindi

The pre-fire dataset for the Tambo River at Bindi is poor with only 8 DWC and 10 EMC values. Despite this small dataset, the pre-fire TSS DWC and EMC values (3 and 4 mg/L respectively) are surprisingly similar to the downstream values from Tambo at Swifts Creek TSS DWC and EMC values (4 and 6 mg/L respectively), which are based on a much larger dataset of 196 samples.

The Tambo at Bindi site is different from the other sites in this study in that it was only operational for the first 6 months of 2003. Load estimates are therefore not annual estimates at this site. However, during the six months of sampling, the storm events were captured better than any of the other sites in the study. As a consequence the measurement uncertainty due to non-representative sampling at this site is probably low. The well sampled storm events, (with a total of 83 post-fire samples collected) also enabled the loads from this site to be estimated using the interpolation method (Eq. 4), and compared with other estimation techniques. The results from this comparison are discussed in Section 6.5 and summarised in Table 7.

Table 7 shows that the interpolation method results in TSS loads from the 01/01/03 to the 20/05/03 of approximately 5000 t, using the averaging method and the geometric mean yields about 6500 t, while with the arithmetic mean yields 16,000 t, and the Monte Carlo simulation yields about 15,000 t. These load calculations are based on discharge of 2314 ML over this period. Note that these estimates are different to those presented in other graphs and tables in this report, which are based on a longer period of discharge (01/01/2002 to 14/07/2003), giving a total of 3930 ML.

Pre-fire TSS loads estimated using these averaging estimation methods are approximately 10t. Estimates of the relative increase due to the fire are approximately 1500 fold. The very high factor increases result from some extremely high EMC values, with the highest concentration for all the sites recorded here at 59,000 mg/L. Estimates of loads using the different estimation methods are similar using the different estimation methods, indicating a low level of knowledge uncertainty.

The same sample datasets were available for TP and TN load estimation as were used for TSS load estimation. Relative increases in TP and TN were approximately 400 and 90 fold respectively based on both the arithmetic method and the Monte Carlo simulation.

47 Impact of the 2003 Alpine Bushfires on Streamflow - Estimated changes in stream exports of sediment, phosphorus and nitrogen following the 2003 bushfires in Eastern Victoria

This monitoring site was decommissioned in July 2003, so no information is available regarding recovery in water quality at this site.

6.4.8 Tambo River at Swifts Creek

The large pre-fire dataset for the Tambo at Swifts Creek (196 samples for TSS, TP and TN) provides confidence that the measurement uncertainty is low for the pre-fire load estimates. However stochastic variability is still high for the pre-fire load estimates with SD values 1 to 2 times the mean value for TSS. Storm event sampling captured about 30% of the events for the year, though these were the largest events, which may lead to over-estimated EMC values. Fixed interval samples collected monthly throughout the year provide confidence in the DWC values for this site. However this will have little effect on the estimated loads, as the 200 fold increase in the EMC provides the dominant leverage on the TSS load estimate. The SD of the relative increase is more than three times the mean TSS load value, indicating that stochastic uncertainty is very high.

The Tambo River at Swifts creek site provides one of the best opportunities for monitoring the recovery in water quality following the fire. Most major storms have been well sampled during 2004 and 2005. It appears that the load increases in TSS, TP and TN estimated in 2003 persisted through 2004 and then declined in 2005. However, the load changes relative to unburnt conditions are still significant; particularly for TSS (> 25 fold). TP and TN changes for 2005 are 6 and 3 fold, respectively. The Monte Carlo simulations show that the mean has moved closer to the median in 2005 as less extreme concentrations were measured. Although the stochastic uncertainties persisted, with SDs double or greater than the estimated means, the representativeness of the sampling regime provides confidence in the temporal variation in the load estimates.

6.4.9 Wonnangatta River at Waterford

Only the first two relatively minor storms 2003 were sampled at this site, so the load estimates will include an extremely high level of measurement uncertainty. The load estimates for 2003 should be considered unreliable. No samples were collected in subsequent years so no evaluation of recovery can be made.

48 Impact of the 2003 Alpine Bushfires on Streamflow - Estimated changes in stream exports of sediment, phosphorus and nitrogen following the 2003 bushfires in Eastern Victoria

6.5 Evaluation of load estimation methodology – a case study

6.5.1 Tambo at Bindi dataset

Most of the storm events for the Tambo at Bindi were reasonably well sampled, which enabled a comparison of a range of load estimation methods, including interpolation, regression and averaging. Depending on the estimation method used, data were stratified by event flow and base flow, rising and falling limb, and/or by storm event.

The Tambo at Bindi hydrological recording site was decommissioned in July 2003 so there is not a full year of discharge record for this site. Discharge at Tambo at Bindi (223208) for the period 01/01/2003 to 14/07/2003 (when hydrologic monitoring ceased at this site) was 3930 ML, about a quarter of the long term average for this period of 16,389. The monitoring site in the lower reaches of the Tambo at Swifts Creek (223202) recorded 4746 ML for the period 01/01/2003 to 31/07/2003, which was also about a quarter of the long term average for this period of 20,832. For the year, the Swifts Creek site recorded 32,361 ML, about half the long term average for this site of 59,825 ML. This comparison with the downstream site suggests that the low discharge value recorded for the first half of the year at the Bindi site is correct. The discharge for the period 01/01/2003 to 20/05/03 when the last water quality sample was collected from this site was 2314 ML.

6.5.2 Interpolation method

The TSS load from the Tambo River at Bindi (223208) was estimated using linter interpolation of instantaneous sediment concentrations between the sampled values. The complete 2003 post- fire dataset is show in Figure 19, along with plots showing the eight individual storm events, and the interpolated TSS concentration values.

Six of the eight major storm events in the period 01/01/2003 to 14/07/2003 were sampled, providing the best available dataset for comparison of the interpolation estimation method with other estimation methods. The load calculated using linear interpolation between the sample points was 4803 t. The sampling across the six sampled storm is variable in quality. Sampling during the largest storm (number 1) missed the peak of the storm, and would have resulted in underestimates of loads. Despite this, the first storm still accounted for about 30% of the total load for the monitoring period. Storm 2 and 8 were reasonably well sampled, while storms 3, 4 and 5 show poor levels of sampling, and storms 6 and 7 were not sampled. A feature of this site was the ongoing high values for baseflow sediment concentration.

49 Impact of the 2003 Alpine Bushfires on Streamflow - Estimated changes in stream exports of sediment, phosphorus and nitrogen following the 2003 bushfires in Eastern Victoria

Tambo @ Bindi Storm # 1 1200 70000 1200 70000 1 1000 60000 1000 60000

50000 50000 800 800 40000 40000 600 600 30000 2 30000 400 8 20000 400 20000 200 10000 45 200 3 7 10000 0 6 0 0 0

Storm # 2 Storm # 3 140 30000 120 6000

120 25000 100 5000

100 20000 80 4000 80 15000 60 3000 60 10000 40 2000 40

5000 20 1000 20

0 0 0 0

Storm # 4 Storm # 5 & 6 10000 140 10000 140 9000 9000 120 120 8000 8000 7000 100 100 7000 6000 6000 80 80 5000 5000 60 60 4000 4000 3000 3000 40 40 2000 2000 20 20 1000 1000 0 0 0 0 TSS (mg/L) TSS (mg/L) Discharge (ML/d) (ML/d) Discharge

Storm # 8 Storm # 7 400 10000 120 5000 9000 4500 350 100 4000 8000 300 3500 7000 80 250 3000 6000

60 2500 200 5000 4000 2000 150 40 1500 3000 100 1000 2000 20 50 500 1000

0 0 0 0

Figure 19. 15 min discharge (ML/d) (blue line), sample sediment concentration values (mg/L) (red points) and interpolated instantaneous sediment concentration values for Tambo River at Bindi (223208). The first graph shows the complete post-fire dataset. Subsequent graphs show the individual storm events labeled in the first graph.

50 Impact of the 2003 Alpine Bushfires on Streamflow - Estimated changes in stream exports of sediment, phosphorus and nitrogen following the 2003 bushfires in Eastern Victoria

6.5.3 Regression method

For this site, the regression method produced poor fitting models in the absence of stratification, regardless of whether the instantaneous flow or instantaneous rate of change of flow was used as the predictor variable. Stratifying by rising or falling limb did not improve model fits. Stratifying by storm event produced variable results. As a result, the regression method could not be applied to the dataset at this site.

Table 6. Summary of the application of the regression method for load estimation to the Tambo at Bindi post-fire water quality data.

Method Comment Regression of concentration values Poor model fit (R2 value of 0.0005) against instantaneous discharge using alldata. Regression of concentration values Poor model fit (R2 value of 0.044 & 0.016 against instantaneous respectively) discharge, stratifying by rising and falling limb. Regression of concentration values against instantaneous rate Poor model fit (R2 value of 0.19) of change of discharge using all data. Regression of concentration values against instantaneous rate Poor model fit (R2 value of 0.038 & 0.288 of change of discharge, respectively) stratifying by rising and falling limb. Regression of concentration values Variable model fit (R2 values between 0.014 and against instantaneous 0.82) discharge stratified by storm event. Regression of concentration values against instantaneous rate Variable model fit (see appendix) of change of discharge stratified by storm event All stratifications. Variable model fit (see appendix)

51 Impact of the 2003 Alpine Bushfires on Streamflow - Estimated changes in stream exports of sediment, phosphorus and nitrogen following the 2003 bushfires in Eastern Victoria

6.5.4 Average method

The averaging method (the product of the arithmetic average of the 81 post-fire concentration values of 6798 mg/L and the total discharge of 2314 ML for the period) was applied to the data without stratification, giving an estimated load of 15,732 t. Stratifying the data into baseflow concentrations (2662 mg/L) and event flow concentrations (6978 mg/L), given 21% baseflow and 79% event-flow, results in a load of 14,050 t, 91% of which is generated during event flow. Using the geometric mean (the exponential of the mean of the logs of the values) of 2852 mg/L rather than the arithmetic mean results in an estimated load of 6600 t, while stratification using the geometric means of baseflow concentrations (2110 mg/L) and event flow concentrations (2909mg/L) results in an estimated load of 6,343 t .

For an equivalent discharge pre-fire using the arithmetic mean of 4.97 mg/L (n=35) and geometric mean of 4.13 mg/L gives a load of 11.5 and 9.6 t respectively. Stratifying into baseflow/event flow gives loads of 9.7 t and 7.9 t for arithmetic and geometric loads respectively.

6.5.5 Comparison of methods

Table 7 provides a comparison of the TSS load estimates using a range of different estimation methods. Estimates vary about three-fold from around 5000 t to around 15,000. Methods using the geometric mean produce results most similar to the interpolation method, while methods using the arithmetic mean, or the Monte Carlo simulation produce estimates at the high end of the range. Parkhurst (1998) suggests that for concentration data where mass balances are being calculated that the arithmetic mean is in most cases superior, except perhaps where the number of samples is low (eg. N < 5), in which case the geometric mean may be superior.

Unfortunately, it is difficult to conclude which method is “best” from this analysis, because the interpolation method could not be applied with a high degree of certainty. It should also be noted that the “best” method for this site may not be the best method for a different catchment with different data properties. However, the analysis does give an indication of the possible range in estimated loads when the input data is relatively good.

52 Impact of the 2003 Alpine Bushfires on Streamflow - Estimated changes in stream exports of sediment, phosphorus and nitrogen following the 2003 bushfires in Eastern Victoria

Table 7. A comparison of load estimates using a range of estimation methods.

Concentration Load (t) Method Discharge (ML) (mg/L) % % % All Base Event Total % Total bas base event data only only event e

Regression << method could not be applied >> Linear interpolatio 2,31 ------4,803 -- -- n 4

Arithmetic 2,31 mean -- -- 6,798 -- -- 15,732 -- -- 4

Geometric 2,31 mean -- -- 2,852 -- -- 6,600 -- -- 4

Arithmetic 2,31 2,66 mean 21 79 -- 6,978 14,050 9 91 4 2 stratified Geometric 2,31 2,11 mean 21 79 -- 2,909 6,343 16 84 4 0 stratified Monte Approx Carlo 21 79 ------10 90 15,000 stratified

6.6 Recovery

Recovery in water quality following the fire is discussed in detail on a site by site basis in Section 6.1.

7 Discussion and Summary

The spatial extent of the sampling represents the most ambitious attempt to measure fire effects in Australia, and possibly elsewhere. Despite the uncertainty surrounding many of the load estimations, it is clear that the 2003 fires have had a significant impact on water quality in most of the sampled catchments. It is also clear that recovery has been variable, and that in the Dargo River catchment at least there has been no discernable recovery. In the case of the Mitta Mitta and Tambo rivers, there has been a decline in loads since the very high 2003 levels, but estimates still indicate they are significantly elevated from pre-fire levels. A parallel, more intensive, study in small (100-200ha) catchments in the East Kiewa found suspended concentrations had declined to pre-fire levels two years after the fires in one catchment, 53 Impact of the 2003 Alpine Bushfires on Streamflow - Estimated changes in stream exports of sediment, phosphorus and nitrogen following the 2003 bushfires in Eastern Victoria

although loads were still elevated due to a 70% increase in streamflow (Lane et al., in press). Although it is difficult to compare fire driven impacts between different environments / climates / fire severity, studies from Australia and overseas generally suggest recovery after 3-4 years. One notable exception is given by Helvey (1980) who found loads elevated for 7 years in the USA.

The changes in loads estimated for the Mitta Mitta and Tambo catchments in 2003 compare with the upper values reported in the literature (see Shakesby and Doerr, 2006) under the average or below post-fire rainfall conditions. They are considerably greater than the estimates for suspended sediment made by Chessman (1986) following the 1983 Ash Wednesday fires. The data were sparse and the calculated loads likely to be subject to substantial uncertainty. Conversely, Brown (1972) suggested that one catchment experienced a thousand-fold increase in sediment exports in the first 18 months after burning. Peak measured concentrations were 143,000 mg/L. However the data on rainfall conditions are incomplete, and the load estimates are from sediment ratings curves with substantial scatter, making comparisons difficult. Prosser and Williams (1998) measured increases in hillslope erosion but found sediment was not transported to the stream network. Leitch et al. (1984) measured a single event that mobilised an estimated 800 t of sediment and ash, 2.9 t of nitrogen and 0.22 t of phosphorous from a 35 ha catchment. The East Kiewa catchments (Lane et al., in press) yielded 9-10 fold increases in total sediment loads, including bedload which exhibited a similar factor increase as in suspended sediment.

The variability of impacts is most likely to be a function of the severity of the burn and speed of subsequent vegetation recovery, the intensity and volume of rainfall following the fires, and the interaction of these factors. Additionally, differences in response may reflect the proximity of the monitoring stations to the burnt area, the total percentage of catchment burnt, and the degree to which riparian areas were impacted. For example, the Kiewa River, which displayed minimal increases in loads following the fire, is buffered by upstream impoundments that act to trap sediment and bound nutrients, and the catchment contains significant unburnt areas.

Within this analysis an attempt has been made to highlight the various components of uncertainty surrounding the load estimates. Although these are very high in some cases, the estimated values represent the most likely impacts based on the data set. It should also be noted that uncertainty analyses are rarely, if ever, reported in the fire-impact literature and that it is most probable that similar or greater levels of uncertainty occur in many studies.

Although the use of auto-samplers is common worldwide, this study has demonstrated some limitations in the methodology. The numerous gaps in the record show these are not “set and

54 Impact of the 2003 Alpine Bushfires on Streamflow - Estimated changes in stream exports of sediment, phosphorus and nitrogen following the 2003 bushfires in Eastern Victoria forget” instruments. Careful attention to trigger levels is crucial, and even where the triggers are correctly set there are frequent unexplained instances of instrument failure. In this case the contracting out of data collection has not been as successful as could be desired. A limitation in the project design was a decoupling of the data acquisition and the data analysis, whereby the scientists who analysed the data were not involved in the data collection. In addition, data collection and the associated analysis was undertaken in series, rather than in parallel, and as a result, a frequent lack of compatibility between the collected data and the intended load estimation technique did not become apparent until later in the study. It must be noted that when the sampling was commissioned there was an expectation that the resultant data set would have been more complete.

The experience of the parallel East Kiewa project (Lane et al. in press) has demonstrated high quality data for the accurate estimation of loads can be captured subject to continual rigorous attention to instrumentation. The resources available for the instrumentation, maintenance and service of the two monitoring stations in the East Kiewa study were considerably greater than for each of the sites reported in this study. In particular, the East Kiewa study relied on continuous measurement of turbidity, using frequently calibrated instrumentation. Uncertainty in the East Kiewa load estimates was in the order of 10% of the estimated load, while the standard deviation of the load estimates in this report are commonly of the same magnitude as the estimated load (ie. 100%). Comparison of the levels of uncertainty associated with these two studies could provide some insight into the level of investment and the experimental methods required to achieve load estimates with low levels of uncertainty in remote forested environments.

Both documentary and anecdotal evidence is strong regarding the poor vegetation recovery in some areas of the upper Mitta Mitta and Tambo Rivers where the fire was particularly severe. There is also evidence of substantial mobilisation and temporary storage of coarse material in these headwater areas, suggesting that there are significant sediment sources still to be transported through the stream system (e.g. Harman and Stewardson, 2004).

When interpreting the water quality estimations within this report it is important to note that the sediment and nutrient loads calculated for rivers in the upper catchment are likely to be substantially higher than the loads actually delivered to water impoundments in the lower catchment eg. the Gippsland Lakes. This is because sediment and adsorbed nutrients are stored within stream channels in the lower reaches of the stream network as the slope of the stream channel is reduced. Some of this material may however be re-mobilised during subsequent periods of high flow. It should also be recognised that the burnt areas represent

55 Impact of the 2003 Alpine Bushfires on Streamflow - Estimated changes in stream exports of sediment, phosphorus and nitrogen following the 2003 bushfires in Eastern Victoria

relatively small proportions of the total catchment areas for some receiving waters such as the Gippsland Lakes. Therefore, the large increases identified in this study, where monitoring was relatively high up in the catchment, will represent much smaller percentage increases at a whole-of-catchment scale.

Estimation of end-of-catchment loads of TSS, TP and TN are given in companion reports by Feikema et al. (2005; 2006). For example, estimates based on catchment modelling using the E2 spatial modelling framework suggest post-fire factor increases in loads from the inland flowing rivers (ie. within the Ovens, Kiewa and Hume catchments) of TSS, TN, and TP would be in the order of 15, 2.7 and 2.8 times respectively (Feikema et al. 2006). These factor increases estimated lower down in the catchment are considerably less than the factor increases reported in this study for the upper catchments.

8 Acknowledgements

The authors would like to acknowledge the financial support for this research from the Victorian Department of Sustainability and Environment, as part of the Bushfire Recovery Program, and from both the North East CMA and the East Gippsland CMA. Guidance and feedback from the project committee is acknowledged, particularly Greg Day, Rae Moran, Robert Argent, Yvette Armstrong, Stuart Minchin, Tracy Walker. Many thanks to Assoc. Prof David Fox, Dr Terri Etchells, Dr K.S. Tan (Department of Civil and Environmental Engineering) and Dr Graham Hepworth (Department of Mathematical Sciences) from The University of Melbourne for statistical and mathematical advise on uncertainty estimation methods.

56 Impact of the 2003 Alpine Bushfires on Streamflow - Estimated changes in stream exports of sediment, phosphorus and nitrogen following the 2003 bushfires in Eastern Victoria

9 References

Australian Water Technologies (1999) Victorian water quality monitoring network and state biological monitoring programme. Manual of procedures. WES Report No. 182/97. Brown JAH (1972) Hydrologic effects of a bushfire in a catchment in South - Eastern New South Wales. Journal of Hydrology 15, 77-96. Burgess JS, Reiger, WA and Olive LJ (1981) Sediment change following logging and fire effects in dry schlerophyll forest in southern New South Wales. IHAS publ. 132: 372-385. Chessman, BC (1986) Impact of the 1983 wildfires on river water quality in East Gippsland, Victoria. Australian Journal of Marine and. Freshwater Research., 37: 399-420 Cohn TA (1995). Recent advances in statistical methods for the estimation of sediment and nutrient transport in rivers. Reviews of Geophysics 33: 1117-1123.

Department of Sustainability and Environment, 2007. Impact of the 2003 Alpine Bushfires on Streamflow - Modelling the impacts of the 2003 bushfires on water quality in the Gippsland Lakes catchments. Feikema P.M, Sheridan G.J, Argent R.M, Lane P.N.J and Grayson R.B, School of Forest and Ecosystem Science, University of Melbourne. Etchells T, Tan KS, Fox DR (2005) Quantifying the uncertainty of nutrient load estimates in the Shepparton Irrigation Region. In Zerger, A. and Argent, R.M. (eds) MODSIM 2005 International Congress on Modelling and Simulation. Modelling and Simulation Society of Australia and New Zealand, December 2005, pp. 2755-2761. ISBN: 0-9758400-2-9. Fox DR (2005) Protocols for the Optimal Measurement and Estimation of Nutrient Loads : Error Approximations. Technical Report 03/05. Australian Centre for Environmetrics, University of Melbourne. Grayson RB, Argent RM, Nathan RJ, McMahon TA, Mein RG (1996). Hydrological recipes: estimation techniques in Australian hydrology. Cooperative Research Centre for Catchment Hydrology, 125 p. Harmon C and Stewardson M (2004) Post-fire coarse sediment yield in the upper Tambo River Basin: results and analysis pf a preliminary field survey and literature review. Unpublished report for East Gippsland Catchment Management Authority. Hart BT, Ottaway EM, And Noller BN, (1987) Magela Creek system, northern Australia, II. Material budget for the floodplain. Australian Journal of Marine and Freshwater Research. 38: 861-876. Helvey JD (1980) Effects of a north central Washington wildfire on runoff and sediment production. Water Resources Bulletin 16 (4), 627-634 Hepworth G. (pers comm.) Statistical Consulting Centre. The University of Melbourne.

57 Impact of the 2003 Alpine Bushfires on Streamflow - Estimated changes in stream exports of sediment, phosphorus and nitrogen following the 2003 bushfires in Eastern Victoria

Hood, G. M. (2005) (PopTools version 2.6.9. Available on the internet. URL http://www.cse.csiro.au/poptools) Lane PNJ, Sheridan GJ, Noske PJ (in press) Changes in sediment loads and discharge from small mountain catchments following wildfire in south eastern Australia Journal of Hydrology Leitch CJ, Flinn DW, van de Graaf, RHM (1984) Erosion and nutrient loss resulting from Ash Wednesday (February 1983) wildfires: a case study. Australian Forestry 46, 173-180 Letcher RA, Jakeman AJ, Merrit WS, McKee LJ, Eyre BD, Baginska B (1999) A review of techniques to estimate catchment exports. EPA NSW Technical Report, 110p. Murray-Darling Basin Commission, 2007. Impact of the 2003 Alpine Bushfires on Streamflow - Modelling the impacts of the 2003 bushfires on water quality in catchments in Victoria and New South Wales. Feikema P.M, Sheridan G.J, Argent R.M, Lane P.N.J and Grayson R.B., School of Forest and Ecosystem Science, University of Melbourne. Nathan, R.J., McMahon, T.A., 1990. Evaluation of automated techniques for baseflow and

recession analysis. Water Resource Research 26(7), 1465-1473.

Parkhurst D (1998) Arithmetic versus geometric means for environmental concentration data. Environmental Science and Technology 32 (3) pp92-98. Preston SD, Bierman Jr VJ, Silliman SE (1989). An evaluation of methods for the estimation of tributary mass loads. Water Re sources Research 25 (6): 1379-1389. Prosser IP and Williams L (1998) Effect of wildfire on runoff and erosion in native Eucalyptus forest. Hydrological Processes 12: 251-264 Robinson DM, Roerish ED. (1999) Influence of various water quality sampling strategies on load estimates for small streams. Water Resources Research 35 (12): 3747-3759. Shakesby RA., Doerr SH. (2006). Wildfire as a hydrological and geomorphological agent. Earth-Science Reviews, 74: 269-307 Schwarts SS, Naiman DQ. (1999) Bias and variance of planning level estimates of pollutant loads. Water Resources Research 35 (11): 3475-3487.

58 Impact of the 2003 Alpine Bushfires on Streamflow - Estimated changes in stream exports of sediment, phosphorus and nitrogen following the 2003 bushfires in Eastern Victoria

APPENDIX 1

Summary of regression parameters estimated using the regression estimation method. Regression parameters are summarised by site, by year, and by storm event. Data are stratified by rising and falling limb. Both discharge, and change in discharge, are used as the predictive variable.

59 Impact of the 2003 Alpine Bushfires on Streamflow - Estimated changes in stream exports of sediment, phosphorus and nitrogen following the 2003 bushfires in Eastern Victoria

2003 Data Catchment Dargo Tambo @ Bindi Tambo @ Swifts Ovens @ Bright Mitchell @ Glenadale Snowy @ McKillop Mitta Mitta @ Hinnomunjie Wonnangatta @ Waterford Time Period (Up to 16 Aug 2003) (Up to 20 May 2003) (Aug to Dec 2003) (Up to August 2003) (Aug Only 2003) (Up to 12 March 2003) (Up to 30 Dec 2003) (Up to 21 May 2003) Polynomial Polynomial Polynomial Polynomial Polynomial Polynomial Polynomial (2nd Polynomial (2nd Regression Type Linear (2nd Order) Linear (2nd Order) Linear (2nd Order) Linear (2nd Order) Linear (2nd Order) Linear (2nd Order) Linear Order) Linear Order) All Data SS v Ins Q 0.5599 0.6298 0.0005 0.2229 0.3265 0.0007 0.0778 0.5418 0.5436 0.0294 0.087 0.008 0.0218 SS v Delta Q 0.0654 0.19 0.25 0.0264 0.1366 0.1919 0.0974 0.173 0.0149 0.0356 0.0706 0.1093

All Data - Rising Limb SS v Ins Q 0.64 0.76 0.044 0.117 0.1799 0.3422 0.0153 0.0687 (Only 1 data point) 0.0304 0.0304 0.0109 0.08 SS v Delta Q 0.23 0.038 0.11 0.0126 0.1195 0.2848 0.3247 0.8193 0.2921 0.5179

All Data - Falling Limb SS v Ins Q 0.61 0.91 0.016 0.016 0.2989 0.388 0.006 0.0907 0.6996 0.8723 0.1013 0.2012 0.0212 0.1156 SS v Delta Q 0.37 0.86 0.288 0.49 0.625 0.7101 0.1918 0.1996 0.247 0.0378 0.1926

Individual Storms Storm 1 (All Data) (12/04/03) (27 Feb - 1 Mar/03) (27/2/03) (15-17 Apr 03) SS v Ins Q 0.05 0.078 0.2 0.1604 0.1485 0.2395 0.2964 0.3238 0.1345 0.4639 SS v Delta Q 0.05

(Rising Limb) SS v Ins Q 0.7817 0.8123 0.276 0.0403 0.3954 SS v Delta Q 0.46

(Falling Limb) 0.6772 (poly), SS v Ins Q 0.4081 0.7594 0.7237 0.8008 0.3618 0.5715 (power) 0.0591 0.6708 SS v Delta Q 0.84

Storm 2 (All Data) (19,20/05/03) (8/3/03) (14/4/03) (20,21 May 03) SS v Ins Q 0.34 0.45 0.474 0.5061 0.5067 0.7741 0.094 0.8005 0.3998 0.4461 SS v Delta Q 0.01 0.4696 0.7204

(Rising Limb) SS v Ins Q 0.3246 0.93 0.99 0.4461 0.7705 0.8598 0.8175 0.9507 SS v Delta Q 0.1107

(Falling Limb) SS v Ins Q 0.796 0.694 0.8924 0.635 0.9238 0.6601 0.0358 0.2331 SS v Delta Q 0.4591 0.6913

Storm 3 (All Data) (23,24/07/03) (12/03/03) (23,24/7/03) SS v Ins Q 0.9 0.82 0.018 0.054 0.64 0.0101 0.0109 0.4338 SS v Delta Q

(Rising Limb) SS v Ins Q 0.967 0.8 0.81 0.4656 0.9582 0.9082 SS v Delta Q

(Falling Limb) SS v Ins Q (Only two points) 0.978 0.977 0.998 0.9627 SS v Delta Q

60 Impact of the 2003 Alpine Bushfires on Streamflow - Estimated changes in stream exports of sediment, phosphorus and nitrogen following the 2003 bushfires in Eastern Victoria

2003 Data Catchment Dargo Tambo @ Bindi Tambo @ Swifts Ovens @ Bright Mitchell @ Glenadale Snowy @ McKillop Mitta Mitta @ Hinnomunjie Wonnangatta @ Waterford Cont. (Up to 16 Aug 2003) (Up to 20 May 2003) (Aug to Dec 2003) (Up to August 2003) (Aug Only 2003) (Up to 12 March 2003) (Up to 30 Dec 2003) (Up to 21 May 2003)

Storm 4 (All Data) (24/7/03) (24/8/03) SS v Ins Q 0.64 0.7 0.079 0.73 0.56 0.662 0.003 0.476 0.556 SS v Delta Q

(Rising Limb) SS v Ins Q 0.94 0.54 0.765 0.589 0.657 0.7853 SS v Delta Q

(Falling Limb) SS v Ins Q 0.896 0.88 0.967 0.445 0.9459 SS v Delta Q

Storm 5 (All Data) (24,25/8/03) SS v Ins Q 0.0017 0.0142 0.73 0.73 0.35 0.52 SS v Delta Q 0.6177 0.8455

(Rising Limb) SS v Ins Q 0.5997 0.9215 0.969 0.99 0.7417 0.793 SS v Delta Q

(Falling Limb) SS v Ins Q 0.928 0.969 0.8617 0.9118 SS v Delta Q

Storm 6 (All Data) SS v Ins Q 0.7412 0.9319 0.263 0.268 0.725 0.802 SS v Delta Q 0.42 0.64 0.219

(Rising Limb) SS v Ins Q 0.727 0.865 0.687 0.847 SS v Delta Q

(Falling Limb) SS v Ins Q 0.88 0.998 0.866 0.981 SS v Delta Q

Storm 7 (All Data) SS v Ins Q 0.0463 0.1033 SS v Delta Q

(Rising Limb) SS v Ins Q SS v Delta Q

(Falling Limb) SS v Ins Q SS v Delta Q

61 Impact of the 2003 Alpine Bushfires on Streamflow - Estimated changes in stream exports of sediment, phosphorus and nitrogen following the 2003 bushfires in Eastern Victoria

2004 Data Catchment Dargo Tambo @ Bindi Tambo @ Swifts Ovens @ Bright Mitchell @ Glenadale Snowy @ McKillop Mitta Mitta @ Hinnomunjie Wonnangatta @ Waterford Polynomial Polynomial Polynomial Polynomial Polynomial Polynomial Polynomial (2nd Polynomial (2nd Regression Type Linear (2nd Order) Linear (2nd Order) Linear (2nd Order) Linear (2nd Order) Linear (2nd Order) Linear (2nd Order) Linear Order) Linear Order) All Data NA NA NA NA SS v Ins Q 0.06 NA NA 0.6 0.7 0.57 0.59 0.53 0.53 0.2682 0.3478 0.23 0.25 NA NA SS v Delta Q NA NA 0.33 0.44 NA NA NA NA NA NA All Data - Rising Limb NA NA NA NA SS v Ins Q 0.16 NA NA 0.69 0.83 0.64 0.67 NA NA 0.236 0.31 0.48 0.49 NA NA SS v Delta Q NA NA 0.07 0.07 0.28 0.32 NA NA NA NA NA NA All Data - Falling Limb NA NA NA NA SS v Ins Q 0.002 0.02 NA NA 0.54 0.65 0.55 0.64 0.54 0.54 0.28 0.29 0.25 0.5 NA NA SS v Delta Q NA NA 0.79 0.79 (Note: All Data was one 0.46 0.95 0.31 0.31 NA NA NA NA Storm on 14-15/12/2004) NA NA Individual Storms 24-26/04 NA NA 24 Apr #1 18-20 Jun 6-10/09/2004 30/08-1/09 NA NA Storm 1 (All Data) NA NA NA NA SS v Ins Q 0.76 0.81 NA NA 0.4 0.52 0.32 0.35 0.11 0.44 0.78 0.86 NA NA SS v Delta Q NA NA 0.73 0.85 NA NA NA NA NA NA (Rising Limb) NA NA NA NA SS v Ins Q 0.72 0.8 NA NA 0.94 0.94 0.95 NA NA SS v Delta Q NA NA 0.66 NA 0.62 0.62 NA NA NA NA NA NA (Falling Limb) NA NA NA NA SS v Ins Q 0.64 0.92 NA NA 0.06 0.57 0.75 0.96 0.71 0.88 NA NA SS v Delta Q NA NA 0.7 NA 0.89 0.89 NA NA 15-18/06 NA NA 24 Apr #2 4-5 Aug 3-16/11/04 8/9-16/9 NA NA Storm 2 (All Data) NA NA NA NA SS v Ins Q 0.7 0.73 NA NA 0.83 0.86 0.34 0.53 No Good No Good 0.18 0.19 NA NA SS v Delta Q NA NA No Good No Good 0.44 0.54 NA NA NA NA NA NA (Rising Limb) NA NA NA NA SS v Ins Q 0.98 0.99 NA NA 0.91 0.91 0.14 0.28 0.66 0.5 0.54 NA NA SS v Delta Q NA NA NA NA NA NA NA NA (Falling Limb) NA NA NA NA SS v Ins Q 0.97 NA NA 0.92 0.99 0.12 0.42 0.96 0.85 0.96 NA NA SS v Delta Q NA NA 0.61 0.62 NA NA 30/08-2/09 NA NA 8-11 Sept 8-10 Aug 29/09-1/10/04 NA NA Storm 3 (All Data) NA NA NA NA SS v Ins Q 0.76 0.78 NA NA 0.71 0.73 0.01 0.46 0.66 0.73 NA NA SS v Delta Q NA NA 0.01 0.29 NA NA NA NA NA NA (Rising Limb) NA NA NA NA SS v Ins Q 0.96 NA NA 0.98 0.99 0.008 0.76 0.69 0.9 NA NA SS v Delta Q NA NA 0.96 0.97 NA NA NA NA NA NA (Falling Limb) NA NA NA NA SS v Ins Q 0.95 0.99 NA NA 0.94 0.94 0.99 0.73 0.84 NA NA SS v Delta Q NA NA 0.0075 NA NA 8/09/2009 NA NA 12-16 sept NA NA Storm 4 (All Data) NA NA NA NA SS v Ins Q 0.58 0.7 NA NA 0.97 0.98 NA NA SS v Delta Q NA NA NA NA NA NA NA NA (Rising Limb) NA NA NA NA SS v Ins Q 0.83 0.9 NA NA 0.93 NA NA SS v Delta Q NA NA NA NA NA NA NA NA (Falling Limb) Only two recordings NA NA NA NA SS v Ins Q NA NA 0.97 NA NA SS v Delta Q NA NA NA NA

62 Impact of the 2003 Alpine Bushfires on Streamflow - Estimated changes in stream exports of sediment, phosphorus and nitrogen following the 2003 bushfires in Eastern Victoria

2004 Data Catchment CONT. Dargo Tambo @ Bindi Tambo @ Swifts Ovens @ Bright Mitchell @ Glenadale Snowy @ McKillop Mitta Mitta @ Hinnomunjie Wonnangatta @ Waterford Polynomial Polynomial Polynomial Polynomial Polynomial Polynomial Polynomial (2nd Polynomial (2nd Regression Type Linear (2nd Order) Linear (2nd Order) Linear (2nd Order) Linear (2nd Order) Linear (2nd Order) Linear (2nd Order) Linear Order) Linear Order) 14-Oct NA NA 6-10 Nov NA NA Storm 5 (All Data) NA NA NA NA SS v Ins Q 0.47 0.48 NA NA 0.48 0.48 NA NA SS v Delta Q NA NA NA NA NA NA NA NA (Rising Limb) NA NA NA NA SS v Ins Q NA NA NA 0.82 0.9 NA NA SS v Delta Q NA NA NA NA NA NA NA NA (Falling Limb) NA NA NA NA SS v Ins Q 0.97 NA NA 0.86 0.96 NA NA SS v Delta Q NA NA NA NA 6 - 7 of Nov NA NA 12-13 Nov NA NA Storm 6 (All Data) NA NA NA NA SS v Ins Q 0.06 NA NA 0.72 0.75 NA NA SS v Delta Q NA NA NA NA NA NA NA NA (Rising Limb) NA NA NA NA SS v Ins Q NA NA NA 0.9 0.9 NA NA SS v Delta Q NA NA NA NA NA NA NA NA (Falling Limb) NA NA NA NA SS v Ins Q 0.95 NA NA 0.996 NA NA SS v Delta Q NA NA NA NA 13-14/12 NA NA 13-16 Nov NA NA Storm 7 (All Data) NA NA NA NA SS v Ins Q no good 0.1 NA NA 0.6 0.65 NA NA SS v Delta Q NA NA NA NA NA NA NA NA (Rising Limb) NA NA NA NA SS v Ins Q 0.99 NA NA 0.85 NA NA SS v Delta Q NA NA NA NA NA NA NA NA (Falling Limb) NA NA NA NA SS v Ins Q NA NA NA 0.8 0.95 NA NA SS v Delta Q NA NA NA NA NA NA 13-15 Dec NA NA Storm 8 (All Data) NA NA NA NA SS v Ins Q NA NA 0.86 0.87 NA NA SS v Delta Q NA NA NA NA NA NA NA NA (Rising Limb) NA NA NA NA SS v Ins Q NA NA 0.95 0.96 NA NA SS v Delta Q NA NA NA NA NA NA NA NA (Falling Limb) NA NA NA NA SS v Ins Q NA NA 0.88 0.89 NA NA SS v Delta Q NA NA NA NA

2003 and 2004 Data Catchment Dargo Tambo @ Bindi Tambo @ Swifts Ovens @ Bright Mitchell @ Glenadale Snowy @ McKillop Mitta Mitta @ Hinnomunjie Wonnangatta @ Waterford Polynomial Polynomial Polynomial Polynomial Polynomial Polynomial Polynomial (2nd Polynomial (2nd Regression Type Linear (2nd Order) Linear (2nd Order) Linear (2nd Order) Linear (2nd Order) Linear (2nd Order) Linear (2nd Order) Linear Order) Linear Order) All Data NA NA 0.12 0.123 NA NA SS v Ins Q 0.32 0.35 NA NA NA NA SS v Delta Q NA NA NA NA NA NA NA NA All Data - Rising Limb NA NA NA NA SS v Ins Q NA NA NA NA SS v Delta Q NA NA NA NA NA NA NA NA All Data - Falling Limb NA NA NA NA SS v Ins Q NA NA NA NA SS v Delta Q NA NA NA NA

63 Impact of the 2003 Alpine Bushfires on Streamflow - Estimated changes in stream exports of sediment, phosphorus and nitrogen following the 2003 bushfires in Eastern Victoria

APPENDIX 2

Summary of pre- and post-fire water quality data.

Site Time BaseFlow=1 Confidence Confidence Eventflow=2 Means -0.95 0.95 N Std.Dev. Std.Err. Minimum Maximum TSS 222209 PRE 1 2.09 1.84 2.34 67 1.04 0.13 1 6 TSS 222209 PRE 2 12.02 -0.88 24.93 82 58.73 6.49 1 520 TSS 222209 POST05 1 2.60 0.03 5.18 5 2.07 0.93 1 6 TSS 222209 POST05 2 36.14 13.53 58.76 7 24.45 9.24 7 82 TSS 222209 POST04 1 3.50 -2.38 9.38 4 3.70 1.85 1 9 TSS 222209 POST04 2 31.44 20.81 42.08 36 31.45 5.24 1 120 TSS 222209 POST03 1 14.20 8.42 19.98 5 4.66 2.08 8 21 TSS 222209 POST03 2 1498.00 835.22 2160.78 61 2587.86 331.34 2 13000 TSS 223202 PRE 1 4.10 0.73 7.48 99 16.93 1.70 1 170 TSS 223202 PRE 2 5.75 3.15 8.36 97 12.92 1.31 1 90 TSS 223202 POST05 1 2.25 -1.73 6.23 4 2.50 1.25 1 6 TSS 223202 POST05 2 170.31 111.34 229.29 105 304.76 29.74 2 2000 TSS 223202 POST04 1 6.40 0.47 12.33 5 4.77 2.14 2 14 TSS 223202 POST04 2 1049.67 685.55 1413.79 137 2155.15 184.13 5 9100 TSS 223202 POST03 1 26.50 6.70 46.30 6 18.87 7.70 1 57 TSS 223202 POST03 2 1139.07 575.07 1703.06 90 2692.78 283.84 2 15000 TSS 223208 PRE 1 3.13 1.55 4.70 8 1.89 0.67 1 6 TSS 223208 PRE 2 4.50 2.50 6.50 10 2.80 0.89 1 9 TSS 223208 POST05 1 0 TSS 223208 POST05 2 0 TSS 223208 POST04 1 0 TSS 223208 POST04 2 0 TSS 223208 POST03 1 2662.00 697.89 4626.12 5 1581.84 707.42 510 4200 TSS 223208 POST03 2 6978.04 4377.55 9578.53 77 11457.32 1305.68 1 59000 TSS 224201 PRE 1 2.84 2.08 3.60 25 1.84 0.37 1 9 TSS 224201 PRE 2 8.27 2.37 14.18 22 13.32 2.84 1 64 TSS 224201 POST05 1 0 TSS 224201 POST05 2 0 TSS 224201 POST04 1 0 TSS 224201 POST04 2 0 TSS 224201 POST03 1 0 TSS 224201 POST03 2 28.27 20.96 35.59 66 29.77 3.66 1 150 TSS 224213 PRE 1 3.30 2.31 4.28 64 3.93 0.49 1 25 TSS 224213 PRE 2 4.62 3.32 5.92 86 6.06 0.65 1 34 TSS 224213 POST05 1 3.25 0.24 6.26 4 1.89 0.95 2 6 TSS 224213 POST05 2 220.89 171.00 270.77 89 236.82 25.10 3 930 TSS 224213 POST04 1 3.40 -0.59 7.39 5 3.21 1.44 1 9 TSS 224213 POST04 2 202.78 159.10 246.46 109 230.09 22.04 1 990 TSS 224213 POST03 1 6.67 0.93 12.40 3 2.31 1.33 4 8 TSS 224213 POST03 2 119.66 66.13 173.18 105 276.58 26.99 1 1600 TSS 401203 PRE 1 2.86 2.39 3.32 83 2.12 0.23 1 17 TSS 401203 PRE 2 9.35 4.06 14.64 106 27.48 2.67 1 260 TSS 401203 POST05 1 3.80 -0.36 7.96 5 3.35 1.50 1 9 TSS 401203 POST05 2 333.33 251.91 414.75 97 403.98 41.02 3 2000 TSS 401203 POST04 1 12.17 -4.07 28.40 6 15.47 6.32 3 43 TSS 401203 POST04 2 305.93 183.60 428.27 61 477.66 61.16 3 2200 TSS 401203 POST03 1 101.90 -12.94 216.74 10 160.54 50.77 4 420 TSS 401203 POST03 2 1721.10 491.82 2950.38 91 5902.60 618.76 1 43000 TSS 402205 PRE 1 16.64 12.05 21.23 11 6.83 2.06 5 30 TSS 402205 PRE 2 25.13 11.18 39.08 8 16.69 5.90 9 63 TSS 402205 POST05 1 0 TSS 402205 POST05 2 0 TSS 402205 POST04 1 0 TSS 402205 POST04 2 0 TSS 402205 POST03 1 0 TSS 402205 POST03 2 34.96 30.13 39.78 93 23.42 2.43 1 140 TSS 403205 PRE 1 4.76 0.74 8.78 88 18.97 2.02 1 180 TSS 403205 PRE 2 9.93 4.29 15.57 101 28.56 2.84 1 280 TSS 403205 POST05 1 2.00 0.16 3.84 4 1.15 0.58 1 3 TSS 403205 POST05 2 22.00 12.99 31.01 19 18.69 4.29 1 83 TSS 403205 POST04 1 1.25 0.45 2.05 4 0.50 0.25 1 2 TSS 403205 POST04 2 89.61 70.36 108.85 81 87.03 9.67 1 330 TSS 403205 POST03 1 2.86 1.31 4.41 7 1.68 0.63 1 5 TSS 403205 POST03 2 260.99 198.27 323.71 98 312.82 31.60 5 1700

64 Impact of the 2003 Alpine Bushfires on Streamflow - Estimated changes in stream exports of sediment, phosphorus and nitrogen following the 2003 bushfires in Eastern Victoria

Site Time BaseFlow=1 Confidence Confidence Eventflow=2 Means -0.95 0.95 N Std.Dev. Std.Err. Minimum Maximum TP 222209 PRE 1 0.014 0.011 0.017 67 0.012 0.001 0.005 0.067 TP 222209 PRE 2 0.022 0.014 0.030 82 0.036 0.004 0.005 0.320 TP 222209 POST05 1 0.019 0.011 0.027 5 0.007 0.003 0.011 0.025 TP 222209 POST05 2 0.060 0.035 0.085 7 0.027 0.010 0.025 0.110 TP 222209 POST04 1 0.026 0.003 0.048 4 0.014 0.007 0.013 0.040 TP 222209 POST04 2 0.048 0.038 0.057 36 0.028 0.005 0.011 0.140 TP 222209 POST03 1 0.052 0.014 0.089 5 0.030 0.014 0.021 0.098 TP 222209 POST03 2 3.349 2.019 4.678 61 5.189 0.664 0.005 21.000 TP 223202 PRE 1 0.016 0.013 0.019 99 0.016 0.002 0.005 0.140 TP 223202 PRE 2 0.023 0.019 0.028 97 0.023 0.002 0.005 0.120 TP 223202 POST05 1 0.057 -0.053 0.166 4 0.069 0.034 0.019 0.160 TP 223202 POST05 2 0.135 0.100 0.170 105 0.181 0.018 0.013 1.200 TP 223202 POST04 1 0.021 0.008 0.034 5 0.010 0.005 0.008 0.032 TP 223202 POST04 2 0.742 0.502 0.982 137 1.419 0.121 0.005 7.100 TP 223202 POST03 1 0.078 -0.003 0.158 6 0.076 0.031 0.025 0.230 TP 223202 POST03 2 0.787 0.529 1.045 90 1.232 0.130 0.014 4.700 TP 223208 PRE 1 0.016 0.009 0.023 8 0.009 0.003 0.007 0.034 TP 223208 PRE 2 0.019 0.015 0.022 10 0.006 0.002 0.012 0.030 TP 223208 POST05 1 0 TP 223208 POST05 2 0 TP 223208 POST04 1 0 TP 223208 POST04 2 0 TP 223208 POST03 1 3.940 1.503 6.377 5 1.963 0.878 1.400 6.500 TP 223208 POST03 2 8.768 4.595 12.942 77 18.389 2.096 0.058 120.000 TP 224201 PRE 1 0.038 0.024 0.051 23 0.032 0.007 0.007 0.110 TP 224201 PRE 2 0.039 0.022 0.057 22 0.040 0.009 0.011 0.180 TP 224201 POST05 1 0 TP 224201 POST05 2 0 TP 224201 POST04 1 0 TP 224201 POST04 2 0 TP 224201 POST03 1 0 TP 224201 POST03 2 0.061 0.045 0.076 66 0.062 0.008 0.008 0.330 TP 224213 PRE 1 0.016 0.014 0.018 64 0.008 0.001 0.005 0.042 TP 224213 PRE 2 0.018 0.016 0.020 86 0.010 0.001 0.005 0.063 TP 224213 POST05 1 0.015 0.007 0.023 5 0.006 0.003 0.008 0.025 TP 224213 POST05 2 0.225 0.170 0.281 89 0.263 0.028 0.018 1.200 TP 224213 POST04 1 0.009 0.003 0.014 5 0.005 0.002 0.005 0.016 TP 224213 POST04 2 0.250 0.193 0.308 109 0.304 0.029 0.009 1.400 TP 224213 POST03 1 0.027 0.010 0.044 3 0.007 0.004 0.020 0.034 TP 224213 POST03 2 0.216 0.129 0.303 105 0.449 0.044 0.005 2.300 TP 401203 PRE 1 0.021 0.015 0.026 83 0.025 0.003 0.005 0.197 TP 401203 PRE 2 0.039 0.012 0.065 106 0.135 0.013 0.005 1.400 TP 401203 POST05 1 0.020 0.013 0.027 5 0.006 0.003 0.016 0.030 TP 401203 POST05 2 0.329 0.257 0.400 97 0.353 0.036 0.015 1.600 TP 401203 POST04 1 0.024 0.007 0.040 6 0.016 0.007 0.009 0.044 TP 401203 POST04 2 0.266 0.166 0.365 61 0.387 0.050 0.007 2.100 TP 401203 POST03 1 0.717 -0.264 1.697 10 1.370 0.433 0.024 3.600 TP 401203 POST03 2 1.419 1.050 1.787 91 1.771 0.186 0.012 6.400 TP 402205 PRE 1 0.029 0.027 0.031 359 0.015 0.001 0.006 0.190 TP 402205 PRE 2 0.054 0.049 0.060 246 0.044 0.003 0.012 0.300 TP 402205 POST05 1 0.030 0.021 0.039 10 0.012 0.004 0.013 0.054 TP 402205 POST05 2 0.044 0.028 0.060 11 0.023 0.007 0.017 0.092 TP 402205 POST04 1 0.021 0.017 0.025 26 0.011 0.002 0.005 0.050 TP 402205 POST04 2 0.034 0.026 0.042 26 0.021 0.004 0.008 0.098 TP 402205 POST03 1 0.023 0.016 0.031 17 0.014 0.003 0.006 0.060 TP 402205 POST03 2 0.041 0.036 0.046 128 0.028 0.002 0.007 0.210 TP 403205 PRE 1 0.027 0.009 0.045 88 0.085 0.009 0.005 0.790 TP 403205 PRE 2 0.031 0.021 0.041 101 0.049 0.005 0.006 0.430 TP 403205 POST05 1 0.014 0.005 0.022 4 0.005 0.003 0.010 0.021 TP 403205 POST05 2 0.044 0.032 0.056 19 0.024 0.005 0.010 0.110 TP 403205 POST04 1 0.012 0.007 0.016 4 0.003 0.001 0.008 0.015 TP 403205 POST04 2 0.124 0.101 0.147 81 0.104 0.012 0.008 0.480 TP 403205 POST03 1 0.018 0.013 0.023 7 0.005 0.002 0.011 0.025 TP 403205 POST03 2 0.333 0.229 0.437 98 0.519 0.052 0.019 2.900

65 Impact of the 2003 Alpine Bushfires on Streamflow - Estimated changes in stream exports of sediment, phosphorus and nitrogen following the 2003 bushfires in Eastern Victoria

Site Time BaseFlow=1 Confidence Confidence Eventflow=2 Means -0.95 0.95 N Std.Dev. Std.Err. Minimum Maximum TN 222209 PRE 1 0.272 0.231 0.313 67 0.168 0.021 0.048 1.060 TN 222209 PRE 2 0.340 0.294 0.385 82 0.209 0.023 0.112 1.300 TN 222209 POST05 1 0.470 0.081 0.859 5 0.313 0.140 0.230 0.900 TN 222209 POST05 2 0.762 0.545 0.978 12 0.341 0.098 0.460 1.800 TN 222209 POST04 1 0.470 0.165 0.775 5 0.246 0.110 0.230 0.780 TN 222209 POST04 2 1.574 0.893 2.255 50 2.397 0.339 0.220 13.000 TN 222209 POST03 1 4.229 -2.755 11.213 5 5.625 2.516 0.267 13.000 TN 222209 POST03 2 10.120 7.116 13.125 61 11.731 1.502 0.307 45.000 TN 223202 PRE 1 0.249 0.216 0.281 98 0.161 0.016 0.081 1.135 TN 223202 PRE 2 0.310 0.268 0.351 96 0.204 0.021 0.103 1.143 TN 223202 POST05 1 0.183 -0.031 0.398 3 0.086 0.050 0.090 0.260 TN 223202 POST05 2 1.037 0.811 1.263 103 1.157 0.114 0.210 7.500 TN 223202 POST04 1 0.241 0.160 0.321 5 0.065 0.029 0.160 0.293 TN 223202 POST04 2 3.546 2.619 4.472 137 5.485 0.469 0.170 26.000 TN 223202 POST03 1 0.840 0.358 1.321 6 0.459 0.187 0.249 1.400 TN 223202 POST03 2 4.111 3.143 5.079 90 4.621 0.487 0.236 21.000 TN 223208 PRE 1 0.290 0.179 0.401 6 0.105 0.043 0.163 0.403 TN 223208 PRE 2 0.411 0.281 0.540 9 0.168 0.056 0.226 0.782 TN 223208 POST05 1 0 TN 223208 POST05 2 0 TN 223208 POST04 1 0 TN 223208 POST04 2 0 TN 223208 POST03 1 22.600 10.736 34.464 5 9.555 4.273 13.000 36.000 TN 223208 POST03 2 39.851 20.029 59.673 77 87.332 9.952 0.600 600.000 TN 224201 PRE 1 0.231 0.187 0.274 22 0.097 0.021 0.103 0.414 TN 224201 PRE 2 0.376 0.291 0.461 22 0.191 0.041 0.103 0.700 TN 224201 POST05 1 0 TN 224201 POST05 2 0 TN 224201 POST04 1 0 TN 224201 POST04 2 0 TN 224201 POST03 1 0 TN 224201 POST03 2 0.566 0.478 0.655 66 0.359 0.044 0.100 2.000 TN 224213 PRE 1 0.171 0.146 0.195 64 0.099 0.012 0.029 0.568 TN 224213 PRE 2 0.179 0.149 0.209 86 0.141 0.015 0.041 1.116 TN 224213 POST05 1 0.160 1 0.000 0.160 0.160 TN 224213 POST05 2 1.144 0.878 1.410 90 1.271 0.134 0.110 5.700 TN 224213 POST04 1 0.130 0.096 0.164 5 0.027 0.012 0.100 0.170 TN 224213 POST04 2 1.701 1.334 2.068 109 1.931 0.185 0.100 9.000 TN 224213 POST03 1 0.308 0.021 0.595 3 0.116 0.067 0.237 0.441 TN 224213 POST03 2 1.449 0.974 1.925 105 2.455 0.240 0.140 13.000 TN 401203 PRE 1 0.186 0.158 0.214 83 0.127 0.014 0.039 0.661 TN 401203 PRE 2 0.309 0.183 0.435 106 0.655 0.064 0.048 6.699 TN 401203 POST05 1 0.175 -0.270 0.620 2 0.050 0.035 0.140 0.210 TN 401203 POST05 2 1.611 1.320 1.901 95 1.427 0.146 0.120 7.300 TN 401203 POST04 1 0.322 0.089 0.554 6 0.222 0.091 0.150 0.740 TN 401203 POST04 2 2.079 1.448 2.710 61 2.464 0.315 0.140 13.000 TN 401203 POST03 1 2.254 0.014 4.493 10 3.131 0.990 0.305 8.900 TN 401203 POST03 2 6.560 4.935 8.185 91 7.801 0.818 0.260 32.000 TN 402205 PRE 1 0.281 0.265 0.296 359 0.146 0.008 0.084 0.970 TN 402205 PRE 2 0.477 0.438 0.516 246 0.310 0.020 0.105 1.830 TN 402205 POST05 1 0.272 0.225 0.319 9 0.061 0.020 0.190 0.390 TN 402205 POST05 2 0.680 1 0.000 0.680 0.680 TN 402205 POST04 1 0.338 0.269 0.407 26 0.171 0.033 0.160 0.760 TN 402205 POST04 2 0.572 0.451 0.693 26 0.299 0.059 0.150 1.390 TN 402205 POST03 1 0.370 0.267 0.472 17 0.200 0.049 0.141 0.770 TN 402205 POST03 2 0.700 0.462 0.937 128 1.358 0.120 0.128 13.000 TN 403205 PRE 1 0.199 0.175 0.223 88 0.113 0.012 0.051 0.692 TN 403205 PRE 2 0.229 0.203 0.256 100 0.135 0.013 0.080 0.900 TN 403205 POST05 1 0.280 1 0.000 0.280 0.280 TN 403205 POST05 2 0.496 0.302 0.690 17 0.378 0.092 0.230 1.900 TN 403205 POST04 1 0.168 0.052 0.284 4 0.073 0.036 0.112 0.270 TN 403205 POST04 2 1.159 0.861 1.458 81 1.350 0.150 0.140 12.000 TN 403205 POST03 1 0.228 0.187 0.269 7 0.045 0.017 0.155 0.271 TN 403205 POST03 2 2.532 1.982 3.081 98 2.739 0.277 0.280 14.000

No samples available for WRW post-fire baseflow. Calculations assume this is equal to the pre fire baseflow value. Note that that TRB and TRS values are for a six month period only in 2003.

66 Impact of the 2003 Alpine Bushfires on Streamflow - Estimated changes in stream exports of sediment, phosphorus and nitrogen following the 2003 bushfires in Eastern Victoria

APPENDIX 3

Summary of water quality data collected by DSE from the Mitchell River at Glendale (Station No. 224203) and by the Murray Darling Basin Commission from Snowy Creek at Granite Flat (Station No. 401210). Water quality data from these two sites was not analysed in this study because of very limited event sampling in 2003 and 2004 for the Mitchell River at Glendale site, and because of a lack of pre-fire data for the Snowy Creek at Granite Flat site. The Snowy Creek at Granite Flat was instrumented to measure turbidity continuously from December 2004. A summary of the limited data collected from these sites is given in Table 8.

Table 8. Summary of water quality data from the Mitchell River at Glendale (Station No. 224203) and from Snowy Creek at Granite Flat (Station No. 401210).

Date of collection Number Location Station TSS TP TN of storm event of No. (mg/L) (mg/L) (mg/L) samples samples Monthly Grab pre-fire 197 4 0.0226 0.216 15-16/08/2003 Mitchell 24 na na na 14-15/12/2004 River at 224203 19 180 0.151 1.28 05-06/02/2005 Glendale 23 33 0.049 0.456 Monthly grab post- 6 7 0.020 0.279 fire

Monthly Grab pre-fire 0 ------09-20/06/2004 42 69 0.090 0.889 Snowy 04/08/2004 14 71 0.106 1.091 Creek at 401210 08-10/09/2004 25 180 0.215 1.82 Granite Flat 04-05/02/2005 18 82 0.102 0.954 Monthly grab post- 0 ------fire na: samples or data corrupted and unsuitable for analysis

67