INFORMATION TO US£RS

This manuscript has been reproduced from the microfilm master. UMI films the text directly from the original or copy submitted. Thus, some thesis and dissertation copies are in typewriter face, while others may be from any type of computer printer.

The quality of this reproduction is dependent upon the quality of the copy submitted. Broken or indistinct print, colored or poor quality illustrations and photographs, print bleedthrough, substandard margins, and improper alignment can adversely afreet reproduction.

In the unlikely event that the author did not send UMI a complete manuscript and there are missing pages, these will be noted. Also, if unauthorized copyright material had to be removed, a note will indicate the deletion.

Oversize materials (e.g., maps, drawings, charts) are reproduced by sectioning the original, beginning at the upper left-hand comer and continuing from left to right in equal sections with small overlaps. Each original is also photographed in one exposure and is included in reduced form at the back of the book.

Photographs included in the original manuscript have been reproduced xerographically in this copy. Higher quality 6* x 9" black and white photographic prints are available for any photographs or illustrations appearing in this copy for an additional charge. Contact UMI directly to order.

University Microfilms Internationa! A Beil & Howell information Company 300 North 2eeO Roaa. Ann Arbor Ml 48106-1346 USA 313 761-4 700 800 521-06C0 Order Number 9236155

Soil phosphorus dynamics in a humid tropical silvopastoral system

Cooperband, Leslie Rose, Ph.D.

The Ohio State University, 1992

Copyright ©1993 by Cooperband, Leslie Rose. All rights reserved.

UMI 3(H) N. Zeeh Rd. Ann Arbor. MI 4X106 SOIL PHOSPHORUS DYNAMICS IN A HUMID TROPICAL SILVOPASTORAL SYSTEM

DISSERTATION

Presented in Partial Fulfillment of the Requirements for the Degree Doctor of Philosophy in the Graduate School of The Ohio State University

By

Leslie Rose Cooperband, B.A., M.S.

* 4 * * * #

The Ohio State University

1992

Dissertation Committee: Approved by T.J. Logan R. Lai R.E.J. Boerner S. Traina

^^A dvteor m DepartmenrsQf Agronomy a Carlos Fernandez Gutierrez, mi mano derecha

V a los agricultores del asentamiento Neguev

ii ACKNOWLEDGEMENTS

The breadth and depth of the dissertation herein would have been impossible to achieve without the numerous people and institutions who assisted me. From the project's inception to the final writing stage, I have received unquantifiable support spanning two hemispheres including family, friends, colleagues, faculty, funding agencies and technical personnel. I would like to thank them all deeply and mention several specifically. First, I would like to thank the various funding agencies whose financial support made this research possible. These include (in chronological order): the Tinker Foundation; The Midwest Universities Consortium for International Activities; The Ohio State University College of Agriculture's Office of International Affairs; the Fullbright Scholarship Program; and the Lindbergh Fund. Their financial committment allowed me to conduct field research outside the U.S.; an expensive proposition beyond the means of most graduate students. My advisory committee, Terry Logan, Ralph Boerner, Rattan Lai and Sam Traina, provided their valuable time and extremely constructive criticism from project design to data analysis and dissertation writing. My major advisor, Terry Logan has been a true mentor as well as an intellectual sparring partner and emotional confidant. He solidified my confidence in my own abilities by letting me loose in the wilds of Costa Rica. He spent countless hours with me discussing, arguing and philosophizing. I truly feel

1 have grown both personally and professionally from our relationship. Among the Ohio contingency, I thank my wonderful closest friends and partners-in-crime in impartial alphabetical order: Jeff Herrick, Peggy Logan, Nancy Lust, Margaret Reeves and Michelle Wander. In the process of trying to solve the world's and each other's problems, they have enriched my life immeasurably. I would also like to thank Billie Harrison, Jane Ross, Doug Beak, Brad Fuller and the rest of the Soil Chemistry lab folks for their assistance in lab analyses and their general good nature. Billie, thanks for your moral support and comeraderie all these years. I'd also like to express great appreciation to the Soil Characterization Lab people, especially Lee Burras and Sandy Jones who never said no when I asked for help. My sincere appreciation is also extended to Dr. Robin Taylor, my statistics guru in Wooster. Without Robin's creative energy and infinite patience, my dissertation would lack the scientific and intellectual rigor it now has. I also express deep gratitude to Peggy Richards; she kept the "little engine that could" on track, especially when it felt like derailing. Within the course of the research's evolution there were several key people whose suggestions proved extremely fruitful. Among them, I thank Dr. Dave Janos who helped me clarify many of my diffuse early ideas. I am also indebted to Dr. Wes Jarrell for turning me on to the anion exchange membranes. There were numerous people at the Cdntro Agrondmico Tropical de Investigacidn y Ensenanza (CATIE) who provided invaluable logistic, technical, intellectual and moral support. Foremost was my lab technician, Carlos Fernandez. His dedication to quality and infinite patience made my lab and field work bearable and even fun. Gracias, Carlos, por tu ayuda tan valiosa. I would also like to thank Drs. Francisco Romero and Rolain Borel for accepting me as a member of the Silvopastoral Systems Project and for providing me with critical logistical and labor support, especially in the experiment's implementation. Within the Silvopastoral Systems Project team, I extend sincere appreciation to Ebal Oviedo who carried me around on the back of a motor bike for two months in Neguev trying to convince farmers to work with the "gringa loca." Gracias, Ebal, por tu companerismo y tu confianza en el proyecto. I also thank Erick Lopez, Luis Carlos Saborfo, Leonel Solano, Luis Angel Sanchez, Carmelo Ghana and the other field workers for help in the field work. I thank Dr. Donald Kass for his comeraderie and for general assistance throughout my stay at CATIE. I also thank Isabel Rojas of the Universidad Nacional for her help in the mycorrhizae research. Other individuals in the CATIE soils lab and the Nitrogen Fixing Trees Project who provided assistance in lab work include Mario Jimenez, Patricia Leandro and Roberto Diaz-Romeu. Lastly but certainly not least, I thank Dr. Maria Julia Mazzarino for her unbounded friendship and for her unceasing enthusiasm for nutrient cycling. She truly inspired me. I cannot forget to thank the farmers who participated in the research project. Without them, there would be no dissertation. Their willingness to forfeit one quarter hectare of their small farms for more than three years

v impressed me deeply. I am grateful for their humility and their hopefulness. A Nelson Esquivel y su esposa Mima, les agradezco con todo mi corazon. At the end of my long acknowledgements road, I owe deep thanks to my family: my mother, sister, brother and Chirin£, for their acceptance and understanding. I often wondered if they thought I was totally crazy for the things I did, but in the final analysis I know they love me. Finally, to my dad who is here in spirit; after all, he was the guy who sparked the curiosity flame in me. VITA

May 8 , 1 9 6 0 ...... Born - Boston, Massachusetts

1 9 8 2 ...... B.A., Barnard College, Columbi University, New York, New Yo

1982-198 3 ...... Horticulture Specialist, New York City Parks Department, Brooklyn, New York

1983-1984 ...... Ecologist/Environment Educato The Nature Conservancy, Long Island, New York

198 4 ...... Research Assistant, Dr. Daniel Janzen, Costa Rica

198 5 ...... Project Consultant, U.S. Agenc for International Development, Washington, D.C.

1986 ...... M.S., Zoology Department, Th Ohio State University, Columbus, Ohio

1986-1988 ...... Consultant, Midwest Universiti Consortium for International Activities, Columbus, Ohio

vii PUBLICATIONS

Cooperband, L.R. and T.J. Logan. 1991. Measuring in situ changes in labile soil P in a humid tropical silvopastoral system with anion exchange membranes. ASA Abstr. p 241.

Cooperband, L.R. and T.J. Logan. 1991. Phosphorus cycling in a humid tropical silvopastoral system and effects on pasture biomass availability and species composition. ASA Abstr. p 347.

Cooperband, L. R. and T.J. Logan. 1989. Phosphorus adsorption and diffusion in a silvopastoral volcanic ash soil of Costa Rica. ASA Abstr. p 198.

Logan T.J. and L. R. Cooperband. 1987. Soil erosion on steeplands of the humid tropics and subtropics, pp 21-38 in Southgate, D. and J. Disinger (eds). Proceedings of international symposium on sustainable development of natural resources in the third world. Westview Press. Boulder, CO.

Cooperband, L.R. 1986. Soil conservation in hillside agriculture: a case study of highland Venezuela, Peru and Guatemala, in Proc. First International Seminar on Soil and Water Conservation. Santo Domingo, Dominican Republic.

Cooperband, L.R. 1985. Ornithology in the neotropics: a directory. American Ornithologist's Union. Washington, D.C. 95 pp.

FIELDS OF STUDY

Major Field: Agronomy Studies in: tropical agriculture, agroforestry, nutrient cycling LIST OF TABLES

TABLE PAGE 1.1. Monthly and annual precipitation totals (mm) over an 18 year period for the El Carmen weather station, 6 km. east of Neguev settlement ...... 15 1 .2. Analysis of variance model for soil chemical properties ... 31 1 .3. Probability (P) values for both main effects and interactions from the analysis of variance model for soil chemical properties ...... 39 1 .4. Cation exchange properties of study soil at two depths from the beginning to end of study ...... 41 1.5. Study soil pH in water, 0.1 M KCI and their difference for two depths at the beginning and end of study ...... 43 1.6. Sodium bicarbonate-EDTA extractsble K and micronutrients at two depths from the beginning and end of stu d y ...... 45 1.7. Organic carbon, total Kjeldahl N (%) and C:N ratio for two depths atthe beginning and end of study ...... 46 1.8 . Changes in C:N ratio among treatments from beginning to end of study ...... 48 1.9. Phosphorus forms in study soil at two depths from the beginning and end of study ...... 49 1.10. Analysis of variance table for anion exchange resin P 50 1.11. Soil particle density and particle size analysis ...... 55 1.12 Analyis of variance tables for soil bulk density ...... 57 1.13 Soil bulk density via core method presented as means by farm and by treatment for the beginning and end of study ...... 58 1.14 Mineral oxide characteristics of study soil's surface (0-15 cm) and subsurface (15-30 cm) horizons ...... 63 2.1 General soil chemical characteristics of Neguev series surface (0-15 cm) and subsurface (15-30 cm) horizons... 95 2.2 Major phosphorus forms for the Neguev soil series 96

ix 2.3 Effect of agitation time and number of 1 M NaCI extractions on AEF-P extractability ...... 118 2.4 Analysis of variance table of treatment effects ...... 118 2.5 Effect of P addition, incubation time and soil moisture content on anion exchange filter paper P extraction as mg P/kg filter paper (top); and analysis of variance table of main effects and 2 -way interactions (bottom) ...... 1 2 0 2.6 Effect of agitation time and number of extractions on AEM-P extractability ...... 128 2.7 Limiting equivalent ionic conductances for selected inorganic anions in aqueous solutions at 25 °C ...... 133 3.1 Soil chemical characteristics of bulk soil taken from study farms after methyl bromide fumigation ...... 160 3.2 Percent hyphal and vesicular infection of cacao (Theobroma cacao) feeder roots used as VAM inoculum . 160 3.3 Inherent pasture VAM infection levels measured from farm field sam p les ...... 160 3.4 One-way ANOVAs by species for effect of inoculation on both percent hyphal infection andpercent vesicles 169 3.5 Mean hyphal and vesicle infection levels for the two grass species and for Erythrina berteroana from seedlings and cu ttin g s ...... 169 3.6 Analysis of variance for inoculation and regression analysis for hyphal and vesicle infection against growth and nutrient parameters measured for the two grass sp ecies ...... 171 3.7 Analysis of variance for inoculation and regression analysis for hyphal and vesicle infection against growth and nutrient parameters for Erythrina berteroana seedlings and vegetative cuttings ...... 177 4.1 Comparison of mass loss (and P loss) rate constants (k) among residue types on sod or soil for the greenhouse study ...... 2 1 2 4.2 Analysis of variance table for the effects of residue type, placement surface and time on AEM-P under greenhouse conditions (with the no residue controlremoved) ...... 214 4.3 Univariate repeated measures analysis of variance table for effects of residue type and placement surface on AEM-P by residue application rate ...... 214 4.4 Multivariate repeated measures analysis table for effects: AEM-P, residue type, placement surface and their interactions by application rate over time (greenhouse study) ...... 215 x 4.5 Greenhouse anion exchange membrane (AEM-P) (mg P/L extract solution) by residue type, placement surface and application rate (means pooled over time) ...... 221 4.6 Test for improvement of fit of the 3-parameter exponential model over the 2 -parameter negative exponential model (field decomposition rates) ...... 228 4.7 Comparison of shape (c) parameter values between and within (mass and nutrients) residue types from the 3- parameter exponential decomposition model fitted to the field decomposition d ata ...... 228 4.8 F-statistics of stepwise multiple regression between the field decomposition model shape (c) parameter for mass and residue C:N, C:P, Lignin:N and % Lignin ...... 230 4.9 F-statistics of stepwise multiple regression between the field decomposition model shape (c) parameter for residue P and residue C:N, C:P, Lignin:N and % Lignin ...... 230 4.10 F-statistics of stepwise multiple regression between the AEM-P-2nd derivative residue P regression rate (b) parameter and residue C:N, C:P, Lignin:N and % Lignin .. 230 5.1 General soil chemical characteristics of Neguev series surface (0-15 cm) and subsurface (15-30 cm) horizons... 258 5.2 Major phosphorus forms for the Neguev soil series ...... 259 5.3 Corresponding sampling dates and results from preliminary ANOVA of treatment effects on non­ transformed AEM-P (farm means pooled) ...... 269 5.4 Univariate repeated measures F-tests for constant (grand mean), farm and treatment main effects and their interaction. Output from repeated measures ANOVA on log-transformed, index-normalized AEM-P ...... 287 5.5 Univariate repeated measures F-tests by farm for constant and treatment effects on log-transformed, index- normalized AEM-P ...... 287 5.6 Single degree of freedom polynomial contrasts for farm 3. Significant P values (,0.050) indicate that treatments differ at the turning points of the given order polynomial. 294 5.7 Univariate repeated measures F-tests for constant (grand means), farm and treatment main effects and their interaction. Output from ANOVA on log soil moisture or pF ...... 294 5.8 Major P forms in the surface (0-15) horizon from the beginning (Sept. 1987) and end (Dec. 1990) of study ...... 3 0 9

xi 6.1 Botanical inventory of most common pasture species found in native grass pastures in the Neguev settlement. Included are latin names, common local name and general classification ...... 320 6.2 Changes in Erythrina berteroana biomass production over tim e ...... 323 6.3 Grazed (GT) and non-grazed (T) treatment differences in Erythrina biomass production over tim e ...... 323 6.4 Treatment differences in biomass P produced for Erythrina biomass components (leaves, green and woody stems) pooled over tim e ...... 326 6.5 Changes in Erythrina biomass P content over time ...... 326 6 . 6 Mean pasture biomass and biomass P production (grazing cycles pooled) among farms and treatments ...... 333 6.7 Mean dung biomass and biomass P by farm and treatment (G and GT); n = 2 0 ...... 333 6 . 8 General soil chemical characteristics of Neguev series surface horizon (0-15 cm) by farm. Included are exchangeable bases (sum of Ca, Mg, K), exchangeable Al, bicarbonate-extractable P, organic and total soil P, organic C and total kjeldhal N ...... 341 A.1 Mean monthly temperature minima and maxima (oC) and total monthly rainfall (mm) recorded at Barquero farm, Neguev settlement ...... 355 A.2 Neguev series chemical and physical characterization (source: Wielemaker, 1990a) ...... 356 B.1 Concentrations of additional nutrients including K, Ca, Mg, Cu, Zn, Mn in both inoculated and non-inoculated treatments by plant species ...... 360 B.2 Mean biomass (g) and nutrient content (mg) parameters for the two grass species and for E. berteroana grown from seed and vegetative cuttings ...... 361 C.1 Models for greenhouse decomposition data (mass and nutrient loss) including residual mean squares (RES MS) and model parameters (A = y-intercept, B = slope parameter, C = shape parameter and D = 4th parameter). Only significant model fits are included ...... 363 C.2 Mass and nutrient loss (both concentration and total element m ass per litter bag), % mass remaining (MAREM), lignin content (LIGN), carbon:nutrient and lignin:N ratios for decomposing greenhouse residues 367

xii C.3 Models for field decomposition data (mass and nutrient loss) including residual mean squares (RES MS) and model parameters (A *=y-intercept, B = slope parameter, C= shape parameter, D=4th parameter). Only significant model fits are included ...... 369 C.4 Mass and nutrient loss (both concentration and total element mass per litter bag), lignin content (LIGN), carbon:nutrient and ligniniN ratios for decomposing field residues...... 371

xiii LIST OF FIGURES

FIGURE PAGE 1.1 Cross-sectional view of the Atlantic coast landscape showing the location of the Neguev settlement with respect to both altitude and proximity to the Turrialba volcano and the Caribbean Sea (from Wielemaker, 1990b) ...... 7 1.2 Parent materials of the Atlantic coastal plain of Costa Rica (from Wielemaker, 1990a) ...... 8 1.3 Soil chronology of the Atlantic coastal plain of Costa Rica (from Wielemaker, 1990a) ...... 10 1.4 Cross-sectional schematic of the Neguev settlement landscape surface spanning an 8 km radius between the Destierro and Peje rivers. The three layers show the stratification of different-aged volcanic materials. The dark surface represents the most recent mudflows underlain by the polka dotted layer of older mud/lahar flows. The finely dotted areas represent highly weathered alluvial strata (from Wielemaker, 1990b) ...... 11 1.5 Atlantic coastal plain Atlas map showing location of the Neguev settlement within the Limon Province, Canton of Siquirres (a) (from Chinchilla, 1987). Neguev settlement showing location of major transecting rivers and dominant soil series (b) (from Wielemaker, 1990b) ...... 12 1 . 6 Monthly precipitation from the El Carmen weather station (18-yr. record, 6 km east of Neguev settlement) compared with monthly totals (17-mo. record) measured in Neguev, Barquero farm ...... 16 1.7 Location of the five study farms. Participant farmers are as follows: 1 = Eudoro Barquero; 2 = Rodrigo Castillo; 3 = Nelson Esquivel; 4 = Rodrigo Guerrero; 5 = Roberto Espinoza ...... 23

xiv Mean monthly ambient temperature maxima and minima over the 17-month study ...... 26 P adsorption isotherm for the surface (015 cm) and subsurface (15-30 cm) horizons ...... 52 Calculated ion activity products (IAP) in adsorption isotherm equilibrium solutions compared to solubilities of variscite and amorphous Al phosphate (AIOHHPO 4 ). Activities estimated with the Geochem speciation model (Spositc and Mattigod, 1980) ...... 53 Moisture retention characteristics of the Neguev soil surface horizon ...... 60 Percent total pores drained over the range of pore size diameters from macropores to micropores. Pore size diameters correspond to the difference between successively increasing suction pressures ...... 62 Differential thermal analysis (DTA) for the farm- composited clay (< 2*/m) fraction of the Neguev series surface (0-15 cm) and subsurface (15-30 cm) horizons... 65 X-ray diffractogram of the Neguev surface (0-15 cm) horizon clay (< 2pm) fraction ...... 67 Differential x-ray diffractogram for untreated and CBD- treated < 2 pm fraction, surface horizon ...... 68 Scanning electron microscope (SEM) photograph of the < 2 fjm fraction showing an apparent tubular structure ...... 69 X-ray diffractogram of kaolinite standard untreated (a) and formamide treated (b) ...... 70 X-ray diffractogram of halloysite standard (API#12) untreated (a) and formamide treated (b) ...... 71 X-ray diffractogram for the surface (0-15 cm) horizon (A) and subsurface (15-30 cm) horizon (B), < 2 pm fraction untreated (a) and formamide treated (b) ...... 72 Conceptual diagram of the interactions between the soil matrix, soil solution and anion exchange membrane for H2 PO4 '. Soil solution P is buffered by labile soil P adsorbed to the soil matrix as illustrated by the idealized P adsorption isotherm. P adsorbed to the anion exchange membrane is, in turn, in equilbrium with soil solution P as described by the idealized correlation curve ...... 91 Conceptual model relating changes over time in soil solution P, dynamically exchanged P and infinite sink sorbed P ...... 92 P adsorption isotherms for surface (0-15 cm) and subsurface (15-30 cm) horizons of the Neguev soil series 97 xv 2.4 Soil moisture retention characteristics of surface (0*6 cm) horizon of the Neguev soil serie s ...... 98 2.5 Experimental design of the silvopastoral field experiment. The tree species used was Erythrina berteroana planted from 2.5 m vegetative cuttings. The 2 X 2 factorial design was replicated on five neighboring farm s ...... 100 2.6 Tree row and transect grid sampling plan for placing anion exchange filter paper d is c s ...... 105 2.7 Schematic diagram of the procedure for anion exchange membrane preparation and in situ use in the field...... 114 2.8 Equation describing the statistical normalizing index used to analyze anion exchange membrane(AEM) P fluxes as a function of tree pruning and grazing. The index normalizes AEM-P with respect to both the experimental control treatment (no trees, no grazing) and AEM-P levels prior to tree pruning and pasture grazing ...... 117 2.9 Correlation between anion exchange filter paper (AEF)- extracted P (referred to as resin P on the y-axis) and equilibrium soil solution P ...... 122 2.10 Anion exchange filter paper (resin) extractable P as a function of distance from selected trees; the two tree treatments, with and without grazing (cattle) are com pared ...... 124 2.11 Soil solution P estimated from anion exchange filter paper (AEF)-P at three sampling dates for the four experimental treatm ents...... 125 2.12 P sorbed by various resin materials as a function of soil solution P. Materials were incubated in soil for six days. ± 1 S D ...... 126 2.13 The effect of repeated P soaking solution-NaCI extraction on anion exchange membrane P sorption (referred to as resin P) over a range of P soaking solution concentrations 129 2.14 The effect of increasing N 0 3 " and SO4 .2 - concentrations on AEM P sorption (expressed as % interference) from three P solution concentrations (0.25, 0.50, 1.5 mg P/L) 131 2.15 Change in soil solution P as a function of added P after 5 and 19 days equilibration time. ± 1 SD ...... 135 2.16 Changes in AEM-P (expressed as resin P) as a function of incubation time in soil over the range of P added. ± 1 SD 135

xvi 2.17 The relationship between P addition and soil pM over the range of AEM incubation periods (top). Changes in soil pH over time for the various P levels added (bottom) ...... 136 2.18 The correlation between day 2 incubation AEM-P (expressed as resin-P) and day 5 (equilibration period) soil solution P. ± 1 SD ...... 137 2.19 The correlation between day 8 incubation AEM-P (expressed as resin-P) and day 19 equilibration soil solution P. ± 1 SD. Note that soil solution P at 19 d equilibration is approximately an order of magnitude less than soil solution P measured after 5 d equilibration ...... 137 2.20 AEM-P (expressed as resin-P) from the control soil (0 mg P/kg added) after 2, 4, 6 and 8 d incubation in the soil. ± 1 SD ...... 139 2.21 The relationship between percent P remaining in dung (top), Erythrina berteroana leaves (middle) and pasture grass residues (bottom) over the duration of a field decomposition study and concurrent soil AEM-P under each respective residue type. Note that AEM-P under dung residues is about four times greater than both Erythrina leaves and pasture grass residue AEM-P ...... 140 2.22 Fluctuations in AEM-P (index-normalized and log- transformed) before and after tree pruning and cattle grazing for the three non-control field treatments. The y- axis is in log scale. Any AEM-P increases reflect P fluxes greater than the control treatment (non-grazed, no trees) and the time prior to tree pruning ...... 141 2.23 The relationship between AEM-P (index-nomalized, log- transformed) and soil moisture suction. High soil moisture suction values indicate relatively dry conditions while low values indicate wet to near soil moisture saturation...... 143 2.24 Log index AEM-P as related to the log of lag~^ soil moisture suction. There is a suggested inverse linear relationship between the two, albeit not very strong ...... 144 3.1 Total plant mass as a function of either percent hyphal or percent vesicle infection for the two grass species. Figs. a & b are for Homo/eps/s aturensis and Figs. c & d are for Paspalum conjugatum...... 173 3.2 Paspafum conjugatum total plant P content as a function of either percent hyphal or percent vesicle infection ...... 174

xvii 3.3 Total plant mass for E. berteroana grown from seedlings and new growth (stem and leaf) mass for E. berteroana grown from vegetative cuttings as a function of either percent hyphal or vesicle infection. Figs. a & b are for seedlings and figs. c & d are for cuttings ...... 178 3.4 Total plant or new growth P as a function of either percent hyphal or vesicle infection for E. berteroana seedlings (a & B) and cuttings (c & d) ...... 179 3.5 Frequency distribution (% total by treatment) of nodules for inoculated and non-inoculated E. berteroana seedlings. The nodulation index is a qualitative assessment of nodule abundance where 1 - no nodules, 2 = few and small,3 = many and small and 4 = abundant and large ...... 182 4.1 General schematic of phosphorus cycle (source: Walbridge, 1991) ...... 192 4.2 Mass loss over time (expressed as a log function) for Erythrina leaves (a); Erythrina leaves and stems (b); pasture grass (c) and cattle dung (d) in the greenhouse study ...... 209 4.3 Residue P loss over time (no specific fundtions fitted) for Erythrina leaves (a); Erythrina leaves and stems (b); pasture grass (c) and cattle dung (d) in the greenhouse study ...... 210 4 .4 Labile soil P fluxes (expressed as AEM-P) for residues decomposing on sod; residue types compared to no residue control; for Erythrina leaves (a); Erythrina leaves and stems (b); pasture grass (c) and cattle dung (d) in the greenhouse study ...... 216 4.5 Labile P fluxes (as AEM-P) for all residues decomposing on bare soil ...... 217 4.6 Comparisons of P fluxes (as AEM-P) for no residue control, sod versus bare soil ...... 219 4.7 Comparisons of P fluxes between residues on sod versus bare soil (as AEM-P) for Erythrina leaves (a); Erythrina leaves and stems (b); pasture grass (c) and cattle dung (d) 220 4.8 Fluctuations in soil moisture (pF) over the course of litter bag decomposition in the field (n = 5 for each sampling d a te )...... 222 4.9 Fluctuations in soil temperature over the course of litter bag decomposition in the field (n = 6 for each sampling d a te )...... 223

xviii Mass, P, N and C loss from decomposing Erythrina leaves fitted to the 3-parameter exponential model. (Note: mass and nutrient values are scaled to fit on the same graph).. 225 Mass, P, N and C loss from decomposing pasture grass fitted to the 3-parameter exponential model. (Note: mass and nutrient values are scaled to fit on the same graph).. 226 Mass, P, N and C loss from decomposing cattle dung fitted to the 3-parameter exponential model. (Note: mass and nutrient values are scaled to fit on the same graph).. 227 Regression between shape (c) parameter in the 3- parameter model and residue C:P ratio ...... 232 Relationship between dung P loss over time and AEM-P beneath litter bags. Note: AEM-P scale is approximately 5X those for Erythrina and pasture grass ...... 233 Relationship between Erythrina leaf P loss over time and AEM-P underneath litter bags ...... 234 Relationship between pasture residue P loss over time and AEM-P underneath litter bags ...... 235 Linear relationship between AEM-P and the second derivative of dung P ...... 238 Linear relationship between AEM-P and the second derivative of Erythrina leaf P ...... 239 Linear relationship between AEM-P and the second derivative of pasture grass P ...... 240 Regression between the rate constant (from the AEM-P- 2nd derivative residue P relationship) and the residue C:P ratio ...... 241 Flow diagram of sources, sinks and processes regulating soil P dynamics (from: Sanyal and De Datta, 1 9 9 1 ) ...... 250 Hypothesized pattern of P fluxes in the silvopastoral system. Erythrina trees are expected to scavenge labile soil P more effciently than pasture grasses and sequester P in their leaf biomass. Upon pruning, P in leaf biomass is released via decomposition processes into the upper soil stratum to which pasture roots have access. Cattle also recycle P by consuming both pasture and Erythrina leaves and defecating. They alter P dynamics by concentrating and redistributing easily soluble P sources ...... 255 Moisture retention curve for the Neguev soil series, surface horizon ...... 260 Field experimental design. A 2 X 2 factorial including trees and grazing. Grazed treatments are approximately twice as large as non-grazed (clipped) treatments ...... 261 xix 5.5 Transect sampling design used to monitor labile P dynamics as a function of tree pruning and grazing during the June 1989 tree pruning. Tree spacing was 3 m within rows and 6 m between rows ...... 263 5.6 Non-transformed AEM-P fluxes following November 1990 tree pruning and grazing ...... 268 5.7 AEM-P fluxes following tree pruning and grazing expressed as the ratio of treatmentx relative to the non­ grazed, no tree experimental control ...... 271 5.8 Index-transformed AEM-P fluxes over time. November 1990 pruning. Farm 1 (a) and Farm 3 (b| ...... 273 5.9 Labile soil P fluxes (as AEF-P) for all treatments following tree pruning and grazing, June 1989. Farm 1 (a) and Farm 2 (b)...... 276 5.10 Anion exchange filter paper P (expressed as resin P) as a function of distance from tree for the two tree treatments (sampling date means pooled). June 1989 tree pruning. Farm 1...... 278 5.11 Soil solution P (estimated from AEF-P) close to (25-75 cm) and far away (1 50-300 cm) from trees (both tree treatment means pooled) as a function of sampling time. The first sampling point represents one week prior to pruning; the two subsequent points represent one and three weeks after pruning, respectively ...... 279 5.12 Gravimetric soil moisture fluxes following tree pruning and grazing, June 1989. Farm 1 (a) and Farm 2 (b) ...... 281 5.13 Labile soil P fluxes (as AEM-P) following tree pruning and grazing, May 1990. Farm 2. Surface (0-2.5 cm) horizon (a) subsurface (5-8 cm) horizon (b). Note the difference in scale for the two horizons ...... 283 5.14 Labile soil P fluxes (as AEM-P) by treatm ent for surface (0- 2.5 cm) and subsurface (5-8 cm) horizons following tree pruning and grazing, May 1990. Trees alone (a); Grazed, no trees (b); Trees and Grazing (c); No trees, No grazing control (d) ...... 285 5.15 Gravimetric soil moisture changes following tree pruning and grazing, May 1990. Farm 2 ...... 286 5.16 Labile soil P fluxes (as AEM-P treatm entxontrol ratio) following tree pruning and grazing, November 1990. Farm 1 ...... 289 5.17 Labile soil P fluxes (as AEM-P treatm entxontrol ratio) following tree pruning and grazing, November 1990. Farm 1 with last sampling point removed ...... 290 xx Labile soil P fluxes las AEM-P treatm entxontrol ratio) following tree pruning and grazing, November 1990. Farm 3 ...... 291 Labile P fluxes (as AEM-P treatm entxontrol ratio) by treatment for farms 1 and 3. November 1990 pruning .... 292 Soil moisture fluxes (as pF) for both farms (1 and 3) during November 1990 tree pruning and grazing (treatment means pooled for each farm) ...... 296 Mean square error (MSE) of residuals from regressions of log AEM-P (non-normalized values) against several time lags of log soil moisture tension (pF). The lowest residual MSE corresponds to the best-fitting regression ...... 297 Time lag (-1 sampling date) relationship between soil moisture (pF) and labile soil P (as log non-normalized AEM- P), all treatm ents plotted together...... 298 Individual treatment AEM-P - pF regression relationships. No trees, no grazing control (a); Grazed, no trees (b); Trees alone (c); Trees and grazing (d) ...... 299 Field experimental design based on a 2 X 2 factorial including trees and grazing. The two grazed treatments were 900 m2, the non-grazed treatment was 400 m^, and the non-grazed no trees control was 300 m2 ...... 316 Differences in Erythrina leaf biomass production among farms at each pruning event (treatment means pooled). Different letters signify differences at each pruning (p < 0 .0 5 ) ...... 325 Erythrina leaf biomass P production over successive prunings in non-grazed treatment. Standard error of 0.07 differentiates farms for each pruning event ...... 328 Erythrina leaf biomass P production over successive prunings in grazed treatment. Standard error of 0.06 signifies significant differences among farms at each pruning event ...... 329 Fluctuations in pasture biomass on offer among treatments relative to an estimate from the regression between biomass and rainfall (overall SE —372). Arrows indicate tree prunings ...... 330 Monthly rainfall corresponding to grazing cycles averaged from Neguev (Farm 1) and the El Carmen weather station ( 6 km east of Neguev ...... 331 Farm differences (SE = 180) in mean pasture biomass on offer between successive pruning events ...... 334

xxi 6 . 8 Treatment differences (SE —168) in mean pasture biomass on offer between successive pruning events ...... 335 6.9 Fluctuations in pasture biomass P among treatments over time (overall SE = 1.27). Arrows indicate tree prunings.... 337 6.10 Pasture biomass P differences among treatments between successive pruning events (SE = 0.44 for treatment comparisons within a given time period). Letters denote significant differences over tim e ...... 338 A. 1 Soil map and legend for the Neguev settlement (from de Bruin, 1 9 9 1 ) ...... 357

xxii TABLE OF CONTENTS

DEDICATION...... ii ACKNOWLEDGEMENTS...... iii VITA...... vii LIST OF TABLES...... ix LIST OF FIGURES...... xiv INTRODUCTION...... 1

CHAPTER

I. BASELINE SOIL CHARACTERISTICS AND CHANGES IN SELECTED PROPERTIES OVER TIME...... 5 INTRODUCTION...... 5 Site Description ...... 6 FIELD SITE SELECTION...... 20 Experimental Design and Criteria for Site Selection 20 Farm Selection Process ...... 21 Tree Species Selection ...... 22 Climatogical Monitoring ...... 25 Baseline Soil Characterization ...... 25 Soil Chemical Methods ...... 27 Soil Physical Analyses ...... 34 Soil M ineralogy ...... 36 RESULTS...... 38 Soil Chemical Parameters ...... 38 Soil Physical Parameters ...... 54 Soil Mineralogy ...... 61 DISCUSSION...... 73 Baseline Characterization: Soil Chemical Param eters...... 73 Baseline Characterization: Soil Physical Parameters 78 Baseline Characterization:Soil Mineralogy ...... 79 Changes in Selected Soil Chemical and Physical Properties Over the Study Period ...... 81

xxiii II. MEASURING IN SITU CHANGES IN LABILE SOIL P WITH ANION EXCHANGE RESIN-IMPREGNATED MEMBRANES ...... 84 INTRODUCTION...... 84 THEORY ...... 89 METHODS AND MATERIALS...... 94 Site Description ...... 94 Calibration of Anion Exchange Filter Paper Methodology ...... 99 RESULTS...... 116 Anion Exchange Filter Paper M ethodology ...... 116 Anion Exchange Membrane Methodology ...... 123 DISCUSSION...... 142 Dynamic Exchanger vs. Infinite Sink ...... 142 Competition From Other Anions ...... 146 Relationship Between Resin Material-P and Soil Solution P ...... 146 Extrapolation of AEM Method to Field Use ...... 148 CONCLUSIONS...... 149

III. RESPONSIVENESS OF THE SILVOPASTORAL SYSTEM'S TWO DOMINANT GRASS SPECIES, PASPALUM CONJUGATUM AND HOMOLEPStS ATURENSIS, AND THE LEGUMINOUS TREE COMPONENT, ERYTHRfNA BERTEROANATO VESICULAR- ARBUSCULAR

IV. DECOMPOSITION OF LEGUMINOUS TREE {ERYTHRfNA BERTEROANA), PASTURE GRASS AND CATTLE DUNG RESIDUES AND EFFECTS ON LABILE SOIL P DYNAMICS...... 187 INTRODUCTION...... 187 MATERIALS AND METHODS...... 193 Site Description ...... 194

xxiv Greenhouse Experiment; Decomposition and Dynamics of P Release from Erythrina residues, pasture residues and Cattle Dung ...... 195 Field Decomposition Study and Labile Soil P Dynamics Beneath Decomposing Residues ...... 204 RESULTS...... 207 Greenhouse Decomposition ...... 207 Concurrent Labile Soil P Dynamics ...... 210 Field Decomposition ...... 217 Residue P loss and Labile Soil P Fluxes in the Field 230 DISCUSSION...... 236 Comparisons between the Greenhouse and the Field...... 236 Placement Surface Effects...... 241 Best-fitting Decomposition Models ...... 242 Relationship between Residue P Loss and Labile Soil P Fluctuations ...... 243 CONCLUSIONS...... 245

V. LABILE SOIL PHOSPHORUS FLUXES AS A FUNCTION OF TREE PRUNING AND GRAZING IN A HUMID TROPICAL SILVOPASTORAL SYSTEM...... 246 INTRODUCTION...... 246 METHODS AND MATERIALS...... 255 Site Description ...... 255 Monitoring Labile Soil P Fluxes from the First Tree Pruning {June, 19891 ...... 261 Monitoring Labile Soil P Fluxes Using Anion Exchange Membranes ...... 263 RESULTS...... 274 First Tree Pruning June 1 9 8 9 ...... 274 Third Tree Pruning May 1990 ...... 281 Fourth Tree Pruning November 1990 ...... 283 DISCUSSION...... 299 LeguminousTree Effects ...... 299 Grazing Effects and the Relationship Between Trees and Grazing Over the 18-Month Study ...... 303 Relationship Between Soil Moisture Fluctuations and Labile Soil P D ynam ics ...... 306 Farm Differences in Labile P Dynamics ...... 307 CONCLUSIONS...... 309

XXV VI. SILVOPASTORAL SYSTEM-LEVEL CHANGES IN LEGUMINOUS TREE (ERYTHRINA BERTEROANA) AND PASTURE BIOMASS PRODUCTION...... 311 INTRODUCTION...... 311 MATERIALS AND METHODS...... 314 Site Description ...... 314 General Management of Field Experiment ...... 316 Climatogical Monitoring ...... 318 Pasture Biomass Determination ...... 318 Tree Biomass Measurements ...... 320 Cattle Dung Biomass Estimates ...... 321 RESULTS...... 321 Erythrina Biomass Production ...... 321 Changes in Pasture Biomass Production ...... 327 Cattle Dung Biomass Production ...... 339 DISCUSSION...... 340 Farm Differences in System Component Biomass Production...... 340 Phenology of Silvopastoral System Production ...... 342 Predicted P Dynamics Among System Components ....343 CONCLUSIONS...... 346

SUMMARY AND CONCLUSIONS...... 347 Changes in Soil Chemical Properties Over Time ...... 347 Importance of Biological Mediation of P Cycling...... 348 Erythrina berteroana*s Ability to Enhance P Cycling ....350 Effects of Tree Pruning and Grazing on Pasture Biomass Production and Management Implications 351

APPENDIX A ...... 354 APPENDIX B ...... 359 APPENDIX C ...... 362

LIST OF REFERENCES...... 372

xxvi INTRODUCTION

In developing countries of the humid tropics, timber exploitation and agricultural expansion are-the primary means to earn foreign exchange and reduce national debt. Lowland primary forests are cleared for removal of precious hardwoods often with little regard for sustainable production or environmental impacts. In Central America, for example, intact forests have been cleared at rates as high as three percent annually (Buschbacher, 1986). Once rudimentary infrastructure is in place, a cascade of secondary exploitation ensues. The pioneer mentality often governs decisions in natural resource use. Many countries with fairly large expanses of tropical forest have laws which promote deforestation by increasing land value for cleared lands and giving automatic ownership to those who replace forests with "improved" land uses. As a result, landless squatters, who cut and burn timber-extracted lands for subsistence cultivation, invade to take advantage of short-lived boosts in soil fertility. In addition, land speculators lay claim to vast tracts by burning and establishing pastures for cattle ranching. Cattle ranching is viewed as a relatively inexpensive way to maintain large areas free of secondary forest vegetation, thereby increasing land value with minimal inputs. Resultant pasture management is extremely laissez faire. Practices including continuous grazing, overstocking, no fertilizer inputs and lack of high quality forage eventually lead to overgrazing and declines in both soil 1 2 fertility and pasture productivity. The end product is degraded pasture; i.e., an increase in undesirable forage species (weeds) to the point where it is no longer economically or ecologically viable to maintain livestock (Serrao et al., 1979). Soils with inherentlly low fertility (low nutrient levels, acid, aluminum toxicity, high phosphorus fixation) exacerbate the problem and accelerate the process of degradation; such soils constitute large areas of the humid tropics (Sanchez, 1987; Sanchez and Benites, 1987; Sollins et al., 1988). Degraded pasture lands are abandoned, more forests are cleared and the cycle of deforestation continues. In an attempt to either break the deforestation cycle or at least slow its pace, researchers and development practicioners have proposed the use of alternative farming systems, particularly those which minimize dependency on expensive external inputs (synthetic fertilizers, pesticides, machinery, introduced species). Agroforestry systems, which incorporate fast-growing, multipurpose trees with crops or pasture, have been included among these alternative systems. They have been hypothesized to maintain agricultural productivity through enhanced nutrient cycling, improved soil structure and increased organic matter levels (Huxley, 1987, Nair, 1984), Research on agroforestry systems in both the new and old world tropics has shown that ameliorative tree effects are often long term and dependent on inherent soil fertility status (Sanchez, 1987; Lai 1989a; Palm, 1988). Within the context of agroforestry, systems which combine leguminous trees (particularly those that fix atmospheric nitrogen) with pasture, or silvopastoral systems, are hypothesized to improve both livestock and pasture productivity. Such trees can be used as supplemental livestock forage due to their relatively high nutrient content (Torres, 1983; 3 Perdock et al., 1982). Several studies have demonstrated increased live weight gain and milk production in animals fed diets including leguminous tree fodder (Pezo et al. 1990). Experiments in which leguminous trees were planted with pasture grasses showed signficant increases in pasture production and crude protein content (Torres, 1983; Daccarett and Blydenstein, 1968; Bronstein, 1984). These studies were limited to above­ ground effects (shading, forage quality tree biomass production), however, and did not include livestock grazing as an experimental factor. Few studies have measured soil nutrient dynamics in silvopastoral systems, specifically those nutrients limiting pasture productivity. Phosphorus is considered the major nutrient limiting plant growth in highly weathered soils or those derived from volcanic materials due to their high P retention capacities (Parfitt et al., 1989; Vitousek and Denslow, 1987). Contradictory to its importance, the dynamics of phosphorus availability in high P-retainng soils is relatively unknown, particularly the role of leguminous trees in phosphorus cycling. Hypothetically, leguminous trees have a high P demand because symbiotic rhizobium need P to fix nitrogen (Munns and Franco, 1981; Sa and Israel, 1991). Accordingly, leguminous trees should be effective scavengers for scarce bioavailable soil P via deep and extensive rooting and by symbiotic association with vesicular-arbuscular mycorrhizae (VAM). If left to grow unmanaged, these trees would sequester P in their biomass to a greater extent than associated pasture species. Under such conditions, they potentially could out-compete pasture species for P and create P deficiencies in pasture biomass. Alternatively, if managed to maximize tree leaf biomass production where most of the P is concentrated and then 4 pruned, the primings could be used to return nutrients to the soil surface from where pasture roots would gain access. Tree prunings would release P via decomposition and subsequent mineralization. Spatial redistribution into the upper soil stratum would facilitate pasture P uptake, particularly if the overall input was great enough to overwhelm the soil's high P-retention capacity. Within the context of cattle grazing, P cycling dynamics would be altered because cattle tend to concentrate P into localized high-P sources; i.e., dung, and create P flushes from compensatory root dieback following grazing (Buschbacher, 1987). In addition, there would be some P export from the silvopastoral system in the form of cattle biomass. The general objectives of my dissertation research were to: 1) study the dynamics of labile (bioavailable) soil P in a low fertility, high P-retaining soil as a function of leguminous tree pruning and cattle grazing; and 2 ) determine the effects of leguminous trees, tree pruning and cattle grazing on pasture biomass production in a silvopastoral system. I established

Erythrina berteroana (an arboreous legume) in native grass pastures in the Atlantic coastal plain of Costa Rica. I measured soil P fluxes as well as changes in pasture biomass over an 18-month period using a 2 X 2 factorial experimental design (trees and grazing were the two main factors). The resultant treatments included: 1 ) a grazed pasture with trees; 2 ) a grazed pasture without trees; 3) a non-grazed (clipped) pasture with trees; and 4) a non-grazed (clipped) no trees pasture (the experimental control). The system was managed using a five-week grazing cycle coupled to a five- month tree pruning regime. The block of four treatments was replicated on five farms in the Neguev settlement (a government-established settlement area for small farmers). I conducted a complete baseline characterization of the study site including: geologic and socio-economic history, vegetation, climate and soil chemical, physical and mineralogical properties (Chapter I). To detect small-scale spatial and temporal changes in labile soil P, I modified a technique using anion exchange membranes (AEM) in situ (Chapter II). I performed a greenhouse experiment to assess the importance of vesicular-arbuscular (VA) mycorrhizae in E. berteroana and pasture grass species' P uptake and overall growth (Chapter III). In separate greenhouse and field experiments, I determined decomposition rates, P release characteristics and subsequent labile soil P fluxes associated with the principal silvopastoral system components, Erythrina leaves and stems, pasture grass clippings and cattle dung (Chapter IV). Moving from the micro- to the macro-scale, I monitored labile soil P fluxes intensively as a function of tree pruning and cattle grazing using the in situ AEM technique in the main field experiment (Chapter V). Finally, in an attem pt to relate treatment effects to above-ground silvopastoral system productivity, I measured changes in pasture biomass production and pasture biomass P at each grazing cycle (Chapter VI). CHAPTER I

BASELINE SOIL CHARACTERISTICS AND CHANGES IN SELECTED PROPERTIES OVER TIME

INTRODUCTION

Site Description

Geology and Geomorphology of Costa Rica's Atlantic Coast

The Atlantic zone of Costa Rica is part of the coastal lowland overlying the Depression of Nicaragua (a narrow canal-like geological formation east of the Central Cordillera) which extends from Nicaragua to Costa Rica along the Caribbean coast (Nobbe and Hazeu, 1987). During the Quaternary, volcanic activity was at its peak, forming two principal volcanic mountain chains: the Guanacaste and the Central (de Bruin, 1991). The study area is located in the Santa Clara coastal plain at the foothills of the Turrialba volcano, a prominent volcano of the Central Cordillera (Figure 1.1). The Santa Clara plain was formed from prequaternary-age alluvial sediments interlayered with quaternary-age laharic volcanic deposits (Figure 1.2). The older alluvial sediments are actually fluvially redeposited volcanic material underlying laharic deposits composed of andesitic-basaltic lava, tuffs and other pyroclastic materials. The lahars, or volcanic debris flows (debris

6 7

40 too ISO Kl

Figure 1.1. Cross-sectional view of the Atlantic coast landscape showing the location of the Neguev settlement with respect to both altitude and proximity to the Turrialba volcano and the Caribbean Sea (from Wielemaker, 1990b). ZONA ATLANTICA NORTE

LtTHO-SI

lie otmm Et unem. KPMin IM LKEKt i m i n IMLKOCI

4 .UWML K N I i n tfU H T K Z K II

n v u n - m c M K o c m tn (H .ctifoa«‘t>

M M iin c loan v iw CIMILLCM CCXIML IflMKMMtVI

M C tlT IC MKM OF TMC (W ILLCM K TtLMKII (CMLY H.CI9t«CCHtl

MMLT1C n o t tKII-PlCISTKCtCt

COMLKCMTK MCXI ifiia-iuinccK. fuacn* r.t

q w . i w i w m in .n c o c , u m m r . i

M M M tw o w m niactiK fHOCOC. t»W D M r .t iM r titn uirttts (oiimcdc n um i mocfnc, noMi r.t

Figure 1.2. Parent materials of the Atlantic coastal plain of Costa Rica (from Wielemaker, 1990a).

oo 9 carried from volcanic eruptions including mud and volcanic materials), covered and cut into the older sediments, in addition to the volcanic lahars, there are more recent alluvial sediments deposited from rivers and streams which crossed over the volcanic-formed landscape (Figure 1.3). The resulting topography, highly dissected and undulating, consists of river terraces, natural levees, flood basins, deep (up to 1 0 m) incisions and steep (~30%) slopes (de Bruin, 1987; 1991). Many of the valleys associated with the steep slopes are poorly drained swamps because they are not connected to any major rivers (Wielemaker, 1990b). It is evident from cross-sectional view of the landscape surface (Figure 1.4) that there are at least three principal strata: pre-quaternary alluvium covered by two quaternary-aged lahar flows (Lansu, 1988; Wielemaker, 1990b). The older lahar remnants, located on the more stable slope ridge tops, are highly weathered, clayey, very deep and homogeneous (C horizon found at depth greater than 120 cm). The more recent lahars have developed into shallower, less weathered and coarser textured soils. Interspersed among these strata and associated with current river systems are deposits of even younger materials from the foothills of the Turrialba volcano. These deposits contain a wealth of easily mineralizable pyroclastic materials (Wielemaker, 1990b). The Neguev settlement area, in which my field sites were located, lies within the Cantons of Siquirres and Guacimo in the northeast part of Limon Province between 10° 10* and 10°15'N latitude and 83°30'and 83°35* E longitude (Figures 1.5 a & b). The altitude ranges from 10 to 50 m above sea level. The Parismina river bounds the settlement area to the west and ZONA ATLANTICA NORTE

VOUHGCST SOILS

I 10 20 30

Figure 1.3. Soil chronology of the Atlantic coastal plain of Costa Rica (from Wielemaker, 1990a). 11

2 • km —t Iftio DMtterro R» PlM

H CorrUiita d* lode m6 % n c l w t i

Corrlooto do lode o d i ooWjoo

Cotfofo* olw loln may meieerfaodee

Figure 1.4. Cross-sectional schematic of the Neguev settlement landscape surface spanning an 8 km radius between the Destierro and Peje rivers. The three layers show the stratification of different-aged volcanic materials. The dark surface represents the most recent mudflows underlain by the polka- dotted layer of older mud/lahar flows. The finely dotted areas represent highly weathered alluvial strata (from Wielemaker, 1990b). Figure 1.5. Atlantic coastal plain Atlas map showing location of the Neguev settlement within the Limon Province, Canton of Siquirres (a) (from Chinchilla, 1987). Neguev settlement showing location of major transecting rivers and dominant soil series (b) (from Wielemaker, 1990b) 13 north, while the Destierro, Williamsburg, Germania, Dos Vueltas and Peje rivers dissect the settlement in northeasterly directions. The following landscape types are distinguishable within the

settlement's boundaries: 1 ) flat, low-lying areas of young alluvial origin; 2 ) relatively high and flat fluvio-laharic ridges ; 3) highly dissected areas of fluvio-laharic origin; and 4) undulating hills of volcanic alluvial origin (Wielemaker, 1990a). The age of these landscapes increases from type 1 to 4; due to such age and position differences, development of corresponding soils has resulted in a soil chronosequence. The younger alluvial soils are shallow, low in fine-sized particles and rich in weatherable minerals. The intermediate age soils contain amorphous clay minerals and possess andic properties such as low bulk density, high soil porosity and thixotropy (a reversible gel-sol transformation in which the soil becomes wet as it is squeezed). The older soils have formed crystallized minerals with increased weathering, first as halloysites and later as kaolinites. Soil chemical properties such as pH and cation exchange capacity (CEC) decrease with increased soil weathering along this soil chronosequence.

Characteristics of the Study Site Soil

The Neguev series is the penultimate in this soil chronosequence. It is deep to very deep (>150 cm), well-drained, clayey, acid, dark brown to dark yellowish brown (hue 1 0 VR) with slightly weathered gravels at depths greater than 120 cm (de Bruin, 1987; 1991). The soil is classified as an andic humitropept, family very fine, kaolinitic, isohyperthermic using US Soil Taxonomy (Soil Survey Staff, 1990). The soil formed from fluvio-laharic 14 materials on the high flat ridges and slopes of the dissected landscape I Wielemaker, 1990b). Soil texture is classified as clay with approximately

60% < 2 pm diameter particles. The series is acid (pH ranges from 4.2- 5,5) and low in CEC. The A horizon extends to a depth between 10-20 cm and is distinguished from the B horizon by a slightly darker chroma (10YR 3/3) than the B horizon. The structure is strong, fine subangular blocky to crumb and this horizon is not thixotropic. The B horizon is very deep (30-

> 1 0 0 cm), porous and has a moderate to weak subangular blocky structure. There is no evidence of clay illuviation in this horizon. In the lower depths, there are few to many weathered gravels around which the soil is thixotropic (Nobbe and Hazeu, 1987). The C horizon begins between 120-130 cm depth and contains weathered gravel and stones of laharic origin in a thixotropic, porous massive structure (see profile description, Appendix A).

Climate

The Atlantic coastal plain is hot and wet. The annual rainfall averaged over an 18 year period (1972-90) from the El Carmen weather station ( 6 km east of Neguev settlement, 15 m above sea level; it is the closest complete weather station) is 3666 mm; this amount is distributed fairly evenly throughout the year (Table 1.1, Figure 1.6). Nonetheless, February through April are relatively dry while June, August and October- December are quite wet. Particularly heavy rainfall is usually associated with cold front storms or exhausted hurricanes known as "temporales" (Nobbe and Hazeu, 1987). The mean annual daily temperature recorded at Table 1.1. Monthly and annual precipitation totals

Year Jan Feb Mar Apr May Jim Jul Aug Sep Oct Nov Dec Avg/yr. Sum 197i 714.5 350.0 7>.4 331.4 1 5 li 426.5 781.4 466.4 401.7 638.8 186.6 487.1 393.6 4725.1 1973 305.2 229.2 31.8 105.8 497.1 237.8 338.5 114.6 98.1 179.2 649.8 597.1 282.0 3384.2 1974 361.3 214.2 128.0 424.7 147.4 217.0 449.2 418.9 164.4 174.8 298.1 432.9 285.9 3430.9 1975 210.8 77.3 187.1 195.8 208.7 358.9 354.8 440.7 199.7 262.1 717.7 611.0 318.7 3824.6 1976 357.8 182.1 87.9 167.7 394.6 204.7 1001.1 342.3 608.2 1488 519.0 296.4 359.2 4310.6 1977 171.6 100.0 213.4 108.5 92.0 577.4 769.4 365.6 319.6 361.0 334.5 169.7 298.6 3582.7 1978 124.3 356.6 222.2 131.9 244.1 383.4 365.0 319.9 282.9 167.7 478.7 287.7 280.4 3364.4 1979 118.1 172.4 89.4 512.6 227.3 392.9 157.2 537.2 243.5 123.6 353.5 513.3 286.8 3441.0 1980 257.7 270.4 35.0 263.1 117.8 506.7 289.4 231.3 160.8 258.4 423.8 916.5 309.2 3710.9 1981 242.0 312.2 194.2 376.6 204.5 230.0 305.0 320.9 139.6 144.4 1062.8 393.1 327.1 3925.3 1982 127.8 102.9 79.1 89.7 156.5 222.8 1180.7 630.6 231.1 527.5 357.3 296.3 333.5 4002.3 1983 266.1 242.8 307.1 41.9 523.4 197.6 393.0 433.3 198.0 330.8 159.7 222.7 276.4 3316.4 1984 507.8 331.6 45.7 96.1 260.5 196.0 138.6 480.7 159.8 353.3 347.3 346.3 272.0 3263.7 1985 72.5 184.5 115.9 108.3 174.0 574.6 210.3 399.7 147.1 177.9 330.1 263.4 229.9 2758.3 1986 502.6 23.3 309.3 232.8 257.6 358.6 420.8 439.2 190.0 282.3 273.0 228.4 293.0 3515.9 1987 395.3 119.2 88.3 397.4 204.3 179.8 349.7 377.9 169.3 714.6 169.3 347.7 292.7 3512.8 1988 557.8 308.1 265.0 114.8 274.6 142.8 252.9 302.5 249.9 372.8 245.4 492.1 298.2 3578.7 1989 198.8 243.6 143.4 301.8 245.5 418.2 466.4 181.5 104.2 375.5 198.9 298.2 264.5 3174.0 1990 471.6 56.8 353.0 174.1 631.9 3630 384.5 697.4 204.9 163.6 360.3 430.7 357.7 4291.8

Avg/mo. 313.9 204.1 156.5 219.7 263.8 325.6 452.0 396.5 224.9 287.2 392.9 400.5 303.1 3637.6 Precipitation (mm) yr. record, 6 km east of Neguev settlement! compared with monthly totals totals monthly with compared settlement! Neguev of east km 6 record, yr. Figure 1.6. Monthly precipitation from the El Carmen weather station (18- (18- station weather ElCarmen the from precipitation Monthly 1.6. Figure 1-ot eod esrdi euvstlmn, aqeo farm. Barquero settlement, Neguev in measured record! (17-month J F M A M J J A S O N D El Carmen s Month Neguev 17 El Carmen is 25.1° C (1976-88); the minimum average is 23.5° and maximum average is 25.9° C %de Bruin, 1991; (Lansu, 1988; Nobbe and Hazeu, 1987). The difference between the average daily temperature of the hottest and coldest months is 2.4° C. Relative humidity is > 80% throughout the year. Accordingly, the soil moisture regime is classified as perudic and the soil temperature as isohyperthermic. Rainfall exceeds evapotranspiration for more than 10 months of the year (Wielemaker, 1990b). Such conditions create the potential for rapid soil weathering and soil loss via erosion. In addition, if saturated conditions persist, localized anaerobiosis can become a problem for plant root oxygen consumption.

Vegetation and Land Use

According to the Holdridge life-zone classification system, most of the Atlantic zone was once entirely premontane forest, basal belt transition. Within this classification type, there are two major forest types: tropical lowland and alluvial rain forest (Nobbe and Hazeu, 1987). The lowland forest is usually associated with well-drained, highly leached soils while the alluvial forest is situated on poorly-drained, slightly developed alluvial sediments. With the construction of the railroad from the country's Central Valley to the Port of Limon and the initiation of large banana (Musa spp.) and cacao (Theobroma cacao ) plantations (1871-1938), the region underwent significant development and subsequent deforestation (Rojas and Waaijenberg, 1990; Rojas and van Sluys, 1990). Timber exploitation was 18 sparse, initially because of sufficient reserves in more accessible parts of the country. However, in recent years, deforestation has been occurring at rates as high as 3% annually (Buschbacher, 1986). Cattle farming has taken hold only within the past 15-20 years as infrastructure from logging has improved land accessibility (van der Kamp, 1990). The current landscape is a mosaic of remnant primary forest, timber-exploited secondary forests, charral (exploited forestland with mostly woody perennial shrubs and grasses), pastures, subsistence cropland and large expanses of export

crops including banana (Musa spp.), pineapple (Ananas comosus), macadamia and tropical house plants. The Neguev settlement area is a microcosm of the Atlantic zone in that it contains the same continuum of vegetation types and land uses; the remnant primary forest patches are quite small, however, and almost non-existent.

Social History of the Neguev Settlement Area

Prior to 1979, Neguev was a large, single owner hacienda. Most of the land was logged for timber, although a large percentage still contained relatively undisturbed tropical forest. The cleared sections were used for livestock grazing and banana plantations. Around 1980, the hacienda was invaded by landless squatters organized by the Small Farmers Union of the Atlantic (UPAGRA). Many of the people were former banana plantation workers, small farmers from different regions of the country (many came from Guanacaste, the northwest province of Costa Rica), and unemployed workers from San Jose (Lansu, 1988). The parastatal agency responsible for agrarian reform, the Institute for Agrarian Development or IDA, 19 intervened and subdivided the hacienda into parcels of 10-17 ha, thus creating the Neguev IDA Settlement. The total area is 5340 ha (494 km2) and includes 311 land holdings. Shortly after the settlement was established, squatters received titles to their parcels and IDA implemented various agricultural development programs. Over 70% of the total land area within Neguev is pasture because of the soil's low fertility. Farmers grow maize ( Zea mays), beans {Phaseoius vulgaris), cassava (Manihot escu/entis) and plantain {Musa spp.) on the more fertile alluvial soils for farm and local consumption. In recent years, IDA and international donor agencies have introduced agricultural development schemes to produce non-traditiona! export crops including pineapple ( Ananas comosus), passion fruit ( Passifiora eduiis) and heart of palm (Bactris gasipaes). These export cash crops are rapidly replacing subsistence crops and pasture, creating a more dynamic and diverse agriculture in Neguev settlement. Of the five farms selected for my field experiment, all farmers were part of the original group of squatters. When they invaded the Neguev hacienda, the parts of their land currently in pasture were either abandoned pastures or scrubby charral (Participant farmers, pers. comm., 1991). In all cases, they removed the woody scrub vegetation to re-establish native grass pastures. The pastures on which experimental plots were established, then, had been cleared of primary forest for more than 2 0 years and maintained as grazed pasture for at least 1 0 years. 20

FIELD SITE SELECTION

Experimental Design and Criteria for Site Selection

The objectives of this study were: 1) to better understand how

leguminous trees affect pasture ecosystems; and 2 } to determine the effect of leguminous trees on soil P cycling and pasture primary production in a grazed silvopastoral system. The field experiment was a 2 X 2 factorial design with presence and absence of trees and grazing as the two factors. The resultant treatments included: 1) a grazed pasture with trees (GT), 2) a grazed pasture without trees (G), 3) a non-grazed plot with trees (T), and 4) a non-grazed plot without trees (the experimental control, P). Tree spacing

was set at 6 X 3 m, pruning frequency at five-month intervals and cattle stocking rate at 2 animal units per hectare per year. The stocking rate represents the average stocking rate (n = 39 farms) for farms in Neguev settlement (unknown author, 1986; CATIE-AUW-MAG Survey of Neguev settlement). These design parameters were determined from previous studies (at CATIE) and are discussed in greater detail along with experiment establishment and management in Chapter VI. The research was conducted in the context of a larger silvopastoral systems project for the Atlantic zone of Costa Rica. This project sought to improve pasture and livestock production for the region's resource-limited farmers through enhanced incorporation of fast-growing, multipurpose trees, particularly legumes. As such, I was looking for soils representative of the 21 region's soil fertility constraints; i.e., acid, high P retention capacity, low exchangeable bases. I was also looking for native grass pastures which exhibited some form of degradation such as weed infestation and had not been altered recently either chemically or by introduction of improved pasture species; i.e., areas with pastures at least 10 years old. From both soils and vegetation points of view, I was looking for "worst case scenarios" to test potential leguminous tree effects on both soil and pasture parameters. Another important component was resource-limited farmer participation. If the system was designed for eventual farmer adoption, I felt that on-farm field experiments were a critical part of this process. I wanted to select farmers who: 1 ) had similar cattle numbers and hectares of native grass pastures and whose livelihood did not depend solely on livestock production (cattle used as the farm's "piggy bank"); 2 ) w ere interested in improving their present level of pasture productivity; but 3 ) did not have access to conventional or large-scale operation methods of improvement including herbicides, fertilizers and improved pasture varieties; and 4) would be willing to contribute on-farm resources (mainly animals and some labor) to maintain the field experiment for a minimum of two years. Given these criteria, an IDA settlement seemed like the most appropriate location because it would provide greatest farmer and farm homogeneity.

Farm Selection Process

I located farms and established the experimental plots during June- September 1987. To locate soils with the desired characteristics, I used soil 22 maps {scale 1:50,000) prepared by the University of Wageningen, The Netherlands-CATIE Atlantic Zone Program (Figure A. 1). Initially, I visited sites throughout the northeast Atlantic region, testing soil pH in water and NaF in the field with pH paper to verify mapping units. I was looking for low pH soils (pH in water < 5.0) and soils with andic properties (pH in NaF tests for presence of amorphic minerals; an NaF pH > 10 suggests presence of poorly crystallized minerals whose hydroxyl groups are displaced by F~ when NaF is added). After much searching, I focused on farms in Neguev settlement because there seemed to be greater abundance of highly weathered soils and native grass pastures. I decided to limit my search to a single soil series and to pastures with similar botanical composition to minimize among-site variation (soil type was not a factor in the experimental design). When a particular farm met these criteria, I interviewed the farmer to determine his willingness to participate. Within

six weeks, I had selected five farms within a 6 km radius in the Neguev settlement (Figure 1.7). On each farm, we established treatment plots in a randomized complete block design.

Tree Species Selection

I focused my selection on fast-growing legumes, mainly because of their nutrient supply potential. I wanted to use a species already native to the Costa Rican pasture landscape and familiar to farmers. Since the experiment's design was essentially an extention/intensification of the living fence concept, I narrowed my selection to species most frequently used for live fences: Erythrina spp. and Gliricidia sepium. In addition to farmer Figure 1.7. Location of the five study farms within the Neguev settlement. Participant farmers are as follows: 1 = Eudoro Barquero; 2 = Rodrigo Castillo; 3 - Nelson Esquivel; 4 = Rodrigo Guerrero; 5 = Roberto Espinoza. w co 24 familiarity, other important selection criteria included multiple use potential (forage, fuel wood, medicinals), forage quality (livestock digestability and palatability), ease of obtaining homogeneous sources of numerous vegetative cuttings, stake establishment and survival, leaf biomass production, ability to withstand repeated and frequent prunings, and worldwide applicability potential.

I chose Gtiricidia septum initially because it is the most abundant live fence species in the Atlantic coastal region. Previous agroforesty and animal nutrition trials also suggested that Gtiricidia was suited to a range of soil types and had slightly higher digestability than Erythrina poeppigiana (65% versus 50%), a leguminous tree used mostly for light shade in coffee plantations (F. Romero and R. Borel, CATIE, pers. comm., 1987). It also had the advantage over Erythrina species in that it could be used for fuelwood and medicinals as well as forage and mulch. The major disadvantage is that it produced less leaf biomass than Erythrina species and had mixed success in palatability trials.

I planted over 400 Gtiricidia stakes in September 1987. Abnormally heavy rains in September and October compounded with a lepidopteran feeding frenzy on Gtiricidia live fences throughout the Atlantic coastal region resulted in high stake mortality. By February of the following year, mortality was over 50% on all five farms, so I decided to replace Gtiricidia with

Erythrina berteroana, another leguminous species common to live fences. I planted Erythrina stakes in early May 1988 in the same manner as Gtiricidia.

Erythrina establishment was successful (near zero mortalility). 25 Climatogical Monitoring

Beginning in May, 1989 and ending in December, 1990 I monitored daily precipitation and maximum and minimum air temperatures on one farm (Farm 1: Sr. Eudoro Barquero) in Neguev (Figure 1.8). I measured daily rainfall (mm) using a "Tru-check" wedge-shaped bucket rain gauge and max/min ambient temperatures in the shade using an Hg max-min thermometer.

Baseline Soil Characterization

I took composite soil samples weighing 1 - 2 kg to characterize baseline soil chemical, physical and mineralogical properties on all five farms. After treatments were demarcated, I collected at least 10 random subsamples per composite sample. I included treatment plot and soil depth as the two main factors in the sampling design: four treatments (grazing/trees, grazing/no trees, trees/no grazing, no trees/no grazing) and two soil depths (0-15 cm and 15-30 cm). For the two larger grazed plots (900 m^). | took two composite samples per depth, and for the smaller, non-grazed plots (400 m^), | took one composite sample per depth. For each farm, then, I took a total of 1 2 composite samples, or six per depth. These field moist soil samples were mixed well and transported to The Ohio State University in Columbus, Ohio where they were air-dried and sieved through a 2 -mm mesh sieve for subsequent chemical, physical and mineralogical analyses. I undertook composite soil sampling as well as other baseline measurements in August 1987 before trees were planted. I Temperature (oC) the 17*month study. 17*month the Figure 1.8. Mean monthly ambient temperature maxima and minima over minima and maxima temperature ambient monthly Mean 1.8. Figure 20 24 22 26 28 16 30 34 32 J A S O N D J F M A M J J A S O N -® - Maxima Maxima - -® ot (uy 99Nvme 1990) 1989-November (July Month - m - Minima 26 27 conducted final characterization sampling in the same manner in November- December 1990.

Soil Chemical Methods

Because I performed some of the baseline and final characterization analyses in two different laboratories (Ohio State University, Agronomy Department, Soil Chemistry Lab and Soils Lab at CATIE, Costa Rica), I repeated certain analyses for both datasets to eliminate possible error from performing similar analyses in different locations. I selected samples from

both initial and final characterization datasets, and reran total P, NaHC 0 3 ~ EDTA extractable P, TKN and exchangeable bases. There were no significant differences between values obtained from the Soil Chemistry Lab at OSU and those from the Soils Lab at CATIE.

Soil pH in Water

I used a 1:1 soihsolution ratio of 5 g soil to 5 mL deionized water; pH was measured in the soil water suspension after mixing (McLean, 1982).

Soil pH in KCI

I used a similar procedure as pH in water, replacing deionized water with 1 M KCI (method used to determine net AEC or CEC). 28 Exchangeable Cations

I used the pH-buffered ammonium acetate procedure (Thomas, 1982) except the solution was buffered at pH 4.8 and not pH 7 (pH 4.8 better reflects the soil's true pH and therefore wouldn't overestimate exchangeable bases). I measured Ca, Mg, K, Na (exchangeable Na was so insignificant, it wasn't measured for final soil characterization and is not reported in results).

Exchangeable Acidity

I used the unbuffered KCI method (Thomas, 1982).

Total Kjeldahl Nitrogen (TKN)

For initial soil characterization, I used a modification of Lachat's QuickChem method No. 10-107-06-2-G (Lachat, 1990b) in which 0.5 g of air-dried, sieved soil is digested for 3 h at 350 °C with 2.5 mL concentrated

H2 SO4 and a mercuric oxide catalyst. An aliquot of the digest is then run on the Lachat QuickChem Automatic flow Injection Ion Analyzer using the salicylate colorometric test for NH 3 . For TKN characterization at the end of the study period, I followed the method according to Bremner and Mulvany

(1982) except I used 5 mL H 2 SO4 without adding water. I also left the soil and acid to react overnight prior to heating instead of only 30 minutes. 29 Organic Carbon

I used the dry combustion method, in which a 2 g soil sam ple in combination with Mn02 is combusted in an oxygenated furnace at 900-950

°C (Soil Survey Staff, 1972). Evolved CO 2 is trapped in an ascarite bulb and the bulb's weight difference is used to determine % organic C. CaC 0 3 is used as a standard.

Total P

For baseline characterization of total P I followed the Lachat

QuickChem method No. 10-115-01-1-H for PO 4 -P in Kjeldahl digests (Lachat, 1990a). I determined P using the Lachat QuickChem Autoanalyzer reduced ascorbic acid phosphomolybdate colorometric procedure. For determination of total P at the end of the study period, I used the conc.

HNO3 /HCIO4 digestion method (Olsen and Sommers, 1982). I tested standard curves and several samples at three wavelengths (440, 460, 470 nm) with and without NaHS 0 3 (NaHSC> 3 is used to reduce As sometimes present as ASO 3 at concentrations which also produce the blue color in complexation with molybdate). I found no significant difference in absorbance values between samples containing and free of NaHS 0 3 ; therefore, I chose not to add it. The most stable wavelength was found to be 470 nm and was used for the sample absorbance readings. To test for method comparability, I ran 10 samples from both 1987 and 1990 samplings using the Lachat TKP method and compared results with the acid digest method. 30

Organic P

I used the ignition method (Olsen and Sommers, 1982). To ensure adequate pH range, I added known amounts of P to several samples and determined percent P recovered.

NaHC 03~EDTA Extractable P, K, Cu, Fe, Mn and Zn (modified Olsen)

I used the CATIE soils lab method (Diaz-Romeu and Hunter, 1978) in

which soil is extracted with a solution of 0.5 M NaHC 0 3 and 0.01 M EDTA adjusted to pH 8.5; soil: solution ratio of 1:10 (2.5 g soil in 25 mL extracting solution).

Statistical Analysis of Soil Chemical Data

Because the sampling design was the same for all of the above soil chemical parameters, I devised a multivariate ANOVA model using the SYSTAT MGLH procedure. The model included farm, treatment, soil depth and sampling time as main effects and all corresponding 2-, 3- and 4-way interactions (Table 1.2; Wilkinson, 1990). F statistics were calculated using error terms corresponding to each main effect and interactions associated with those main effects. Error terms consisted of all interactions with farm since farm variation represents variation due to plot (random) effects, t ran the model for each dependent variable (soil chemical parameters) and list only the F statistic probability (P) values as the results of these ANOVAs. I 31

Table 1.2. Analysis of variance modal for soil chemical properties.

Model statement:

dependent variable *■ constant + farm + treatment ■+■ farm*treatment + depth + depth‘treatment + depth‘farm + depth* treatment "farm + time + time‘treatment + time‘farm + time‘ treatment* farm + time‘ depth + time ‘ depth ‘treatment •+- time‘depth‘farm + time*dapth‘treatment‘farm.

ANOVA Table

Source i Degrees of Freedom

Farm 4 Treatm ent 3 Error a * farm‘treatment 12

Depth 1 Error b ■ depth‘farm 4

Depth *T reatment 3 Error c • depth *trt‘farm 12

Time 1 Error d ■ time‘farm 4

Time‘Treatment 3 Error e ■ time *trt‘farm 12

D epth‘ Time 1 Error f - time * depth * farm 4

Time * Depth*Treatment 3 Error g - time ‘ depth‘tn ‘farm 12 i Separate error terms calculated for specific main effects and their corresponding interaction terms. 32 performed post tests (either least signigicant differences, LSD's or Tukey's HSD multiple comparisons test) for those effects with a significant F- statistic (P < 0.05).

Anion Exchange Resin P

I used the buried bag technique (Gibson et al., 1985) with Dowex 1 X

8 anion exchange resin, 20-50 pm diameter. Approximately 1 g was sewn into very fine mesh (< 2 0 pm) permeable polyethylene bags, 3 X 3 cm 2 square. I buried bags randomly in each treatment plot in the surface horizon only (0-7.5 cm depth): 18 bags each in G and GT treatments, 8 in the T treatment and 6 in the P treatment, for a total of 50 bags per farm. I left these bags in the soil at all five farms for one week in August 1987. I analysed P using the following procedure: 1) removed resin from bags and reweighed it (I had to reweigh the resin because many of the bags had been disfigured by an unknown burrowing soil animal which had widened the mesh openings thus spilling some of the resin beads); 2 ) placed resin from each bag in 50 mL centrifuge tubes and added 25.00 mL of 0.5 M HCI; 3) shook for 1 h on a low-speed horizontal shaker and decanted the acid extract; 4) reweighed resin in centrifuge tubes to determine carry-over from first extraction and then added another 25 mL of 0.5 M HCI for a second extraction. P was determined in both extracts according to ascorbic acid-reduced phosphomolybdate method (Olsen and Sommers, 1982) using a final volume of 25 mL, 10 mL aliquot and 4 mL of reagent. I conducted MGLH analyses of variance on the data for both farm and treatment effects using the SYSTAT MGLH ANOVA procedure (Wilkinson, 1990). 33

Phosphorus Sorption Isotherm

I determined P sorption isotherms for the 0-15 (surface) and 15-30 cm (subsurface) horizons. I took composite samples for each depth from the five farms and ran the experiment in duplicate, t shook air-dried, sieved soil at a soiksolution ratio of 1:100 with six P levels (0, 5, 10, 15, 25, 50 mg/L) in 0.05 M NaCI for 72 h. I used a wide soiksolution ratio to both minimize precipitation of phosphate minerals and to add enough P to saturate the soil's high P retention capacity. After equilibration with the P solutions, I measured the pH of the suspensions. I centrifuged the suspension and measured P concentration in the equilibrium solutions using the ascorbic acid-reduced phosphomolybdate method (Olsen and Sommers, 1982). Using a subset of the samples, I measured total dissolved Al via the ferron-phenanthroline method (Barnhisel and Bertch, 1982). I also measured dissolved organic C (DOC) directly in the equilibrium solution with a

Dohrmann-Xertex Carbon Analyzer. Using pH, Al, DOC and PO 4 and the Geochem program (Sposito and Mattigod, 1980), I calculated Al-P ion activity products (IAP) for all P levels. These were then compared to lAPs for variscite and amorphous aluminum phosphate to determine whether the major mechanism for soil P retention was sorption and/or mineral precipitation. 34 Soil Physical Analyses

Bulk Density

I used the core method (Blake and Hartge, 1986a) for oven-dry soil bulk density (105 °C). After field treatment plots were established and before trees were planted, I took 10 cores from the 900 plots and five cores from the 400 m P plots from the surface soil horizon. For final characterization, I stratified sampling in tree treatments. I took six cores from each of the four treatments but, from the tree treatments, I took three next to randomly selected trees and three between rows. The core volumes for beginning (to) and end (te) samplings were slightly different; the former had a volume of 137.41 cm3 and the latter 99.73 cm3, The statistical analysis included the following: 1) an MGLH ANOVA on the te data set including farm, treatment (only analyzed the two tree treatments) and distance from tree; 2 ) because I found no significant tree distance effect, I performed a second ANOVA on the to and te data sets including farm, treatment and sampling time as main effects. I used an incomplete factorial design (no interaction terms with farm were included in the model statement) because farm 4 was not sampled at the end of the study.

Particle Size Analysis

I used the pipette method (Gee and Bauder, 1986a) and pretreating with H2 O 2 to remove organic matter (Zelazny and Quereshi, 1972). I conducted pre-analysis comparisons of organic matter removal using both 35 h 2 ^ 2 and NaOCI. Results for percent clay were similar between the two

pretreatment methods within a 2 % margin of error, so I used H 2 O 2 pretreatment. Statistical analysis involved an ANOVA model with farm and depth main effects and their interaction.

Particle density

I used the pycnometer method (Blake and Hartge, 1986b) except that I used 50 mL volumetric flasks in place of pycnometers. I used bulk soil samples composited by farm and performed an ANOVA for farm effects.

Moisture Retention Curve

For the low tensions, I used undisturbed soil cores from the surface horizon (99.73 cm3 volume) placed on a sand-filled tension table (points determined include: saturated, pF 1.7, 2.0, 2.3; Klute, 1986). For the intermediate tensions (pF 2.7 and 3.0), I used cores equilibrated on ceramic pressure plates. I took two cores per treatment plot per farm for a total of 60 cores. For high tensions, I used disturbed samples equilibrated on ceramic pressure plates (3 and 15 bar or pF 3.48 and 4.18). The soil samples used were the bulk composite samples collected in 1987. I converted gravimetric moisture content to volumetric (%wt. x BD) and plotted pF (-log soil H 2 O suction in cm) against volumetric moisture content (% vol). I also calculated the percentage of total pores drained at each suction pressure (%voli-%vo| 2 ^total porosity). Soil moisture retention was determined for baseline characterization only. 36

Soil Mineralogy

I performed most of the mineralogical analyses on the clay fraction only. For the chemical analyses, I used samples composited from all five farms from both the 0-15 and 15-30 cm horizons. The 0-15 cm sample was used for x-ray diffraction, thermal analysis and scanning electron microscope analyses. I separated, Mg-saturated, and freeze-dried the clay fraction (< 2 pm) according to standard procedures (Jackson, 1975).

Sodium Citrate-Bicarbonate-Dithionite-Fe (CBD-Fe)

f determined the reductant-soluble Fe content of the clay fraction

(%Fe and %Fe2 C>3 ) from citrate-buffered dithionite extraction (Jackson, 1975). I also analyzed x-ray diffraction patterns of both untreated clay and CBO-treated residues and calculated a differential x-ray diffractogram to determine the Fe mineralogical characteristics.

Acid Ammonium Oxalate Extractable Fe and Al

I determined %Fe and %AI in whole soil oxalate extracts according to McKeague et al.(1971) and Searle and Daly (1977). Using soil samples composited for farm and depth (five samples per depth or a total of 1 0 samples run in duplicate), I extracted 2.5 g of whole soil with a total of 130 mL acid ammonium oxalate at pH 3.0 in the dark. I extracted three times because of high iron content (manifested as intense yellow color of the 37 extract). The extract solution was brought to 250 mL final volume. I measured Al and Fe concentrations in five-fold diluted extracts by atomic absorption spectrometry.

X-Ray Diffraction (XRD)Analysis

I prepared parallel-oriented aggregate specimens by pipetting 2 mL of Mg-saturated dispersed clay suspensions (containing approximately 30 mg clay) onto glass slides and allowing to air-dry (Brady et al., 1986). I used a Philips XRG-3100 generator, PW 1316/90 wide-range goniometer (using CuKa radiation) fitted with a theta-compensating slit diffractometer and DMS-41 control panel to generate the diffraction patterns with the following instrument parameters: initial 20= 10°, final 20= 65°, 820=0.1°, 5T = 40 s and number of data points = 551.

Differential Thermal Analysis (DTA-DSC)

I obtained thermograms using composite samples for both the 0-15 and 15-30 cm horizons with a Dupont 990 Thermal Analyzer, (Dixon,

1966). Using a differential scanning calorimeter cell, N 2 atmosphere and 10 mg samples, scans were made from 40 to 580 °C at a rate of 25 °C/min.

Test for Halloysite

Because the peaks obtained from x-ray diffraction suggested the presence of either kaolinite and/or halloysite, I performed a rapid test for 38 presence of halloysite (Churchman et al., 1983). This test involves first running an XRD spectrum on an untreated clay specimen as described previously. The same slide is then sprayed with formamide, allowed to dry for 20-30 min and the XRD spectrum rerun. If halloysite is present, there should be a peak at 1.04 nm d-spacing. Kaolinite does not expand beyond 0.7 nm d-spacing.

Scanning Electron Microscopy

1 prepared 0-15 cm depth, < 2 pm samples for electron microscopy according to McKee and Brown (1977). After adhering clay particles to an Al sample stub, the specimen was placed in a Denton high vacuum evaporator and coated evenly first with C and then with Au. I used a Cambridge S-4 Stereoscan Electron Microscope and Polaroid film to identify and photograph features of interest.

RESULTS

Soil Chemical Parameters

General Trends

When viewed as a group, the main variables which affect significantly almost all soil chemical parameters were farm, soil depth and sampling time (Table 1.3). Except for Olsen-extracted Cu and Mn and soil C:N ratio, treatment effects were not statistically significant. This trend suggested I =1 o =1 o &! aglflSiirSfglU pooooooooooooo §si§§§i§s§§§§§

1 * * 2

2 $ c 3 3 'ISP 3 ooooooooooooooooooopopppp i»§i§§5ii=§§8sSii§§5i§is§ ooooooooo §mtiis§s ooooooooooooooooooopopppp 28i§3S28385SS8§S§§|§gB$S2 ooooopppooooooooooooooooo ©oopoppoppo oooopopoooooooooooooooooo §§?2|S§f§§S§§8§§0S§81§§1§ m t m q i au»iratuaaNUv^naM ^atS'idt

AI(%ECECI - parcant of Effaetiva cation exchanga capacity h axchangaaWa Al Total P-Inc * total P via incineration mathod; Total P-Dtg •* total P via acid dioaation. Olaan P and micromitrianta art aquivaiant to aodhim biearbonate-EDTA axtractod. Org P It porcem total P aa organic P via incineration mathod. TKN - total kjaldaht nitrogen axprawad at percent. S t 4 0 that initial differences among farms overshadowed any possible treatment effects. There was a sharp difference between the two soil horizons for all parameters except exchangeable Al and acidity (exchangeable Al controls exchangeable acidity) and C:N ratio. Significant sampling time effects suggested that changes had occurred over the three-year field experiment for many of the soil chemical properties. Exchangeable divalent cations (Mg and Ca) and sum of bases as well as organic C and the C:N ratio did not change significantly over time. Percent organic P and Olsen-extracted P and Fe were considered borderline. There was a significant interaction between depth and time for pH in H 2 O, and delta pH indicating a significant decrease in pH over time for the surface soil horizon but not the subsurface. There was a similar significant depth-time interaction for all CEC variables suggesting that farms behaved differently with respect to each parameter. For Ca, for example, farm 1 exhibited a significant increase over time in both soil depths whereas farm 3 Ca increased significantly only in the subsurface horizon. Mg either decreased slightly (as in farms 2 and 5) in the surface horizon only or did not change over time in the subsurface. K decreased significantly in the surface horizon but didn't change in the subsurface except for farm 4 where it decreased over time.

Cation Exchange Capacity

Consistent with the general characteristics of the study soil, bases extracted in NH4 -acetate (pH adjusted to 4.8) were low (mean sum of bases for five farms was 2.29 cmolc/kg) and unbuffered KCI-exchangeable acidity was high (mean for the five farms was 1.45 cmolc/kg; Table 1.4). For the Table 1.4. Cation exchange properties of study soil at two depths from the beginning and end of study.

Ca______Mfl______K______Sum o Bases Al______H______Exch. Acidity ECEC _____ Al t% of ECEC beg end beg end beg end beg end beg end beg end beg end beg end beg end

arm

0 -15 cm

1 1.33 1.63 1.01 0.93 0.40 0.32 2.73 2.88 0.92 1.63 0.56 0.26 1.48 1.88 4.21 4.76 21.6 34.0 2 0.66 0.67 0.84 0.65 0.33 0.23 1.83 1.55 0.69 1.19 0.31 0.22 1.00 1.41 2.83 2.96 25.1 40.1 3 0.76 0.83 0.81 0.71 0.51 032 2.08 1.85 1.40 2.12 0.45 0.26 1.85 2.40 3.94 4.25 36.0 50.1 4 0.92 0.83 1.17 1.00 0.63 0.46 2.73 2.30 0.89 1.52 0.31 0.31 1.21 1.83 3.93 4.13 22.8 36.9 5 0.69 062 0.91 0.85 0.49 0.32 2.09 1.79 1.31 2.36 0.57 0.38 1.70 2.74 3.79 4.53 30.0 51.9

15 -30 cm

1 0.48 1.13 0.40 0.42 0.24 0.19 1.13 1.74 0.90 1.99 0.41 0.26 1.32 2.25 2.44 3.99 36.9 49.7 2 0.21 0.36 0.29 0.28 0.18 0.14 0.69 0.79 0.59 1.24 0.24 022 0.83 1.46 1.51 2.25 39.2 55.5 3 0.30 0.76 027 0.37 0.27 0.21 0.84 1.33 1.45 2.20 0.37 0.27 1.81 247 2.65 3.80 54.3 59.9 4 0.45 0.45 0.44 0.48 0.42 0.28 1.31 1.20 1.07 1.79 0.24 0.26 1.31 2.05 2.63 3.26 41.0 54.9 5 0.27 0.28 0.27 0.34 0.20 0 16 0.74 0.79 1.63 3.33 0.56 0.39 2.19 3.73 2.93 4.51 55.9 73.8 LSD- 0.22 LSD- 0.08 LSD- 0.07 LSD-0.27 LSD-0.15 LSD= 0.14 LSD=0.15 LSD-0.35 LSD- 5.51

LSD'S are differences among all independent variables for each dependent variable. ECEC * Effective cation exchange capacity = sum of bases + exchangeable acidity 42 most part, farm 2 had the lowest values for all variables, while farms 1, 3 and 4 had slightly higher CEC's. Overall, exchangeable bases were significantly lower in the subsoil than the surface soil (mean is 0.94 cmolc/kg) while exchangeable Al and acidity did not differ appreciably with depth. Exchangeable Al dominated both total acidity and ECEC. Both exchangeable Al and percent ECEC present as Al increased substantially from beginning to end of the study. In the subsurface horizon of farm 5, it approached Al toxicity levels. (According to Sanchez and Cochrane (1980), Al saturation > 60% is considered toxic to plants.) Farm 1 exhibited a significant increase in exchangeable Ca for both soil depths over time (from 1.33 to 1.63 cmolc/kg); this increased ECEC. However, there was no significant change in the other four farms. Mg, K and H either decreased with time or didn't change at all. Because exchangeable bases were so low, their contribution to changes over time was minimal.

Soil pH

Soil pH in H 2 O and KCI were low (means of 5.17 and 4.00 respectively for the surface horizon), and there was a significant decrease in

H2 O pH with depth (5.17 to 4.88). Both pH in water and in 1 M KCI appeared to decrease significantly over time for all farms (from 5.17 to 4.80

H2 O pH; from 4.00 to 3.84 KCI pH; Table 1.5). This decrease w as probably regulated by increases in exchangeable Al. H 2 O pH did not change significantly over time in the 15-30 cm horizon, but it did decrease significantly for KCI pH (from 4.03 to 3.85). The difference between the Table 1.5. Study soil pH in water, 0.1 M KCI and their difference for two depths at the beginning and end of study.

Farm pH H20 pH KCI Delta pH beg end beg end beg end

0 -1 5 cm

1 5.25 4.80 4.01 3.73 1.23 1.08 2 5.32 4.83 4.10 3.94 1.22 0.89 3 5.06 4.79 3.89 3.83 1.17 0.96 4 5.08 4.86 3.97 3.94 1.11 0.93 6 5.12 4.74 4.04 3.74 1.08 1.00

15 - 30 cm

1 4.94 4.81 4.03 3.76 0.91 1.05 2 4.96 4.86 4.16 3.94 0.80 0.93 3 4.93 4.80 3.91 3.88 1.02 0.93 4 4.79 4.89 4.01 3.96 0.78 0.93 5 4.78 4.86 4.06 3.73 0.72 1.14 LSD - 0.08 LSD - 0.05 LSD - 0.10

LSD’S are differences among all independent variables for each dependent variable. 4 4

H2 O pH and KCI pH {delta pH) w as always positive (ranges from 0.72-1.23 across both soil depths), indicating a net CEC for both soil horizons. While pH decreased over time, the delta pH did not change.

NaHCC> 3-EDTA Extractable K and Micronutrients

Like exchangeable K, NaHC 0 3 ~EDTA extractable (or modified Olsen) K decreased over time across farms and soil depth (Table 1.6). Modified Olsen-extractable K ranged from 104 to 137 mg/kg in the surface horizon and 81-116 mg/kg in the subsurface horizon. Both Cu and Mn increased significantly, while Zn decreased over time. Except for farm 5, Fe increased substantially over time for the surface horizon (from 248.9 to 333.7 mg/kg) while the subsurface horizon exhibited no real change (except for farm 3). Again, Fe increases may have been influenced by soil pH declines which increase the solubility of Fe.

Organic C and N

As expected, percent organic C and TKN were higher in the surface than the subsurface horizon (Table 1.7). Organic C ranged from 3.17-

3.85% in the 0-15 cm horizon and 1.92-2.73% in the 1 5-30 cm horizon. Organic C did not change over the 3-year study period, but TKN did decrease among farms and both soil depths (from 0.38 to 0.29% and 0.23 to 0.19% in the surface and subsurface horizons, respectively). Differences among farms in the C:N ratio were more strongly influenced by N content (farms 3 and 4 with the two lowest C:N ratios had significantly higher TKN Table 1.6. Sodium bicarbonate-EDTA extractable K and micronutrients at two depths from the beginning and end of study

Farm K Cu Zn Mn Fe beg end beg end beg end beg end beg end

0 -15 cm

1 132.61 102.86 12.06 30.08 113.76 49.59 2.20 4.16 187.06 365.88 2 103.58 73.13 13.66 23.70 136.21 65.31 1.76 2.75 249.09 352.75 3 107.90 92.14 16.14 25.39 59.57 23.23 2.06 3.73 327.88 407.50 4 136.98 137.48 16.62 27.03 167.04 74.95 3.36 5.67 238.62 273.00 5 111.28 97.99 12.97 23.87 231.05 121.63 2.49 4.30 242.00 254.50

15 • 30 cm

1 115.63 64.84 10.36 31.91 100.84 44.08 1.38 3.51 167.65 187.25 2 91.30 46.80 9.65 23.75 132.30 47.56 1,03 2.12 169.56 166.38 3 80.83 64.35 11,14 27.68 49.45 25.56 1.43 3.60 160.71 309.50 4 100.21 86.78 13.51 27.87 132.42 54.94 2.56 4.99 206.79 202.00 5 104.40 51.68 9.86 26.53 186.78 74.88 1.95 4.65 172.26 160.38 LSD = 16.58 LSD= 1.80 LSD ='9.87 LSD= 0.37 LSD == 61.42

LSD'S are differences among all independent variables for each dependent variable.

■h* CJl Table 1.7. Organic carbon, total Kjeldahl N (%) and C:N ratio for two depths ______at the beginning and end of study. ______

Farm______Organic C (%> ______TKN 1%)______C:N Ratio beg end beg end beg end

0 -16 cm

1 3.17 3.12 0.31 0.27 1 0 4 9 11.51 2 3.17 3.20 0.33 0.27 10.15 11.78 3 3.19 3.68 0.50 0.31 6.45 12.04 4 3.85 3.26 0.42 0.31 9.25 10.45 5 3.50 3.18 0.33 0.29 10.65 10.85

15 - 30 cm

1 2.18 2.02 0.17 0.17 12.69 12.25 2 2.06 1.99 0.26 0.17 8.44 11.88 3 2.25 2.49 0.28 0.21 8.33 11.91 4 2.73 2.14 0.25 0.20 11.00 10.81 5 1.92 2.01 0.19 0.19 10.61 10.85 LSD = 0.42 LSD - 0.05 LSD = 2.54

LSD'S are differences among all independent variables for each dependent variable. 4 7 values) than C level. The C:N ratio exhibited a significant treatment effect and treatment X time interaction (Table 1.8). The data suggested that there were initial differences among treatments in the subsurface horizon only. The two grazed treatments, G and GT, exhibited higher C:N ratios (11.52, 12.35) than the non-grazed treatments (T, P) in the subsurface horizon (8.14, 8.84). In contrast, there were no significant differences by the end of the study, suggesting that either organic C increased for all treatments at both depths or TKN decreased (except in the two grazed treatments' subsurface horizons).

Phosphorus Forms

Because P was the focal point of this dissertation, I attempted to characterize P over a wide availability continuum (Table 1.9). Anion exchange resin P (AER-P), considered to correlate with the most labile soil P fraction, was determined in situ only at the beginning of the study. As with other soil chemical parameters, there was a significant farm effect (Table 1.10). Farm 2 had significantly less AER-P than the other farms (0.18 mgP/g resin); farms 4 and 5 exhibited the greatest AER-P (0.50 and 0.58 mgP/g resin). In general, modified Olsen or NaHC 0 3 -EDTA extractable P, another measure of labile P, was low (mean for all farms in the surface horizon was 2.79 mgP/kg soil) while total P was very high (1385 mg/kg). The percent of total P as organic was also quite high, ranging between 60- 75 %. Modified Olsen P increased for all farms over time in both soil depths but dramatically on farms 3 and 4 (3.28 to 11.7 and 4.05 to 15.12 Table 1.8. Changes in C:N ratio among treatments from beginning to end of study.

Treatment Beginning End

0-15 cm Trees alone (T) 8.51 11.36 Grazed, no trees |G) 9.33 11.18 Grazed, trees (TGI 9.51 11.00 Non-grazed, no trees IP) 10.25 11.77

15-30 cm Trees alone (T) 8.14 11.60 Grazed, no trees (G) 11.52 11.56 Grazed, trees (TG) 12.35 11.45 Non-grazed, no trees (P) 8.64 11.55 Overall LSO - 2.54. Table 1.9. Phosphorus forms in study soil at two depths from the beginning and end of study.

Farm AE# NaHC03 extr. P Organic P Incin. Total P Acid dig Total P Org P (% Inc. Tot. Resin P beg end beg end beg end beg end beg end mgP/kg soil

0 -15 cm

0.325 1.95 3.63 900.21 615.47 1310.55 821.25 1263.68 1318.83 68.70 74.93 2 0.179 2.20 3.34 716.43 480.55 1091.41 662.11 1158.66 1237.58 65.35 72.54 3 0.378 3.28 11.69 893.06 705.73 1507.06 1196.67 1697.51 1813.63 59.26 59.21 4 0.503 4.05 15.12 1055.29 853.44 1731.84 1452.19 2153.99 2199.49 60.79 58.96 5 0.578 2.47 6.12 847.28 702.58 1284.07 967.27 1104.24 1472,34 65.99 72.67

15 - 30 cm

1 nd 2.09 3.27 779.69 503.75 1199.59 701.25 1087.67 1211.21 64.93 71.81 2 nd 2.00 2.90 764.98 383.67 1104.81 567.74 1055.11 1088.63 69.09 67.64 3 nd 3.01 10.16 730.14 677.62 1359.76 1161.99 1411.27 1735.70 53.62 58.65 4 nd 3.63 14.68 940.71 790.00 1606.79 1384.22 1393.25 2141.06 58.44 57.11 5 nd 2.28 5.75 788.39 612.50 1258.21 872.34 1242.62 1402.38 62.63 70.33 LSD = 0.087 LSD= 0.52 LSD = 167.4 LSD = 95.21 LSD = 181.94 LSD= 9.68

LSD'S are differences among all independent variables for each dependent variable. 0 Anion exchange resin extractable P in mg P/g resin measured at beginning of study only.

ID 50

Table 1.10. Analysis of variance table for anion exchange resin P.

Source Degrees of Freedom F-ratio P

Farm 4 5.17 0.001 Treatment 3 1.25 0.292 Farm*Trmnt 12 1.2 0.285 Error 227 mgP/kg). Organic P decreased significantly over time (from 882.5 to 671.6 mg/kg in the surface horizon), while the percent of total P in organic form marginally increased for some farms. Both, however, remained high relative to other, non-volcanic soils (Fassbender and Bornemisza, 1987). Total P, m easured via HNO3 -HCIO4 digestion was similar to total P via incineration (values ranged from approximately 1055 to 1700 mg/kg with the exception of farm 4 with extremely high values of 2154 mg/kg). Over time, total P increased significantly for farms 1, 2 and 5 in the surface horizon and farms 3, 4 and 5 in the subsurface.

P Sorption Isotherm

The P retention capacity for the study soil was high; approximately 2000 mg P/kg soil and higher for the surface and subsurface horizons. (Figure 1.9). The disparity between modified Olsen P and total P values corroborated the adsorption isotherm. Changes in Al-P ion activity products (IAP) with increasing P addition showed that all P levels were undersaturated with respect to variscite (Figure 1.10). However, the Geochem model suggested that, at the four highest P additions, the soil was supersaturated with respect to amorphous Al-phosphate. As such, I could not conclude that the P retention characteristics of this soil series resulted solely from adsorption; rather, precipitation of amorphous AJ-phosphate minerals may contribute substantially to the soil's high P retention capacity. P P Adsorbed (m g/kg) 2500 Figure 1.9. P adsorption isotherm for the surface (015 cm) and subsurface subsurface and cm) (015 surface the for isotherm adsorption P 1.9. Figure 1-0 m horizons. cm) (15-30 qiiru Pcn. (mg/liter) conc. P Equilibrium usrae Soil Subsurface tn N> p p IAP . - 0 - 5 - 0 - 5 - 0 - 5 - .0 -3 .5 -3 .0 -4 .5 -4 .0 -5 .5 -5 .0 -6 6.5 qiiru sltoscmae t ouiiiso aict n aopos Al amorphous and variscite of solubilities to compared solutions equilibrium Figure 1.10. Calculated ion activity products (IAP) in adsorption isotherm isotherm adsorption in (IAP) products activity ion Calculated 1.10. Figure phosphate {AtOHHPO^. Activities estimated with the Geochem speciation speciation Geochem the with estimated Activities {AtOHHPO^. phosphate oe (pstad atgd 1980). Mattigod, (Spositoand model Variscite Amorphous Al-phosphate o (20“) (H2P04“ log CO U 1 54 Soil Physical Parameters

Soil Particle Density and Particle Size Analysis

Soil particle density for the soil series Neguev was 2.62 g/cm^ (mean of 5 farms; Table 1.11). This value fell within the documented range of 2.6- 2.7 g/cm^ for most mineral soils (Hillel, 1982). Particle density ranged from 2.32-2.70 for soils with andic properties, the lower values reflecting higher soil organic matter content (Maeda et al., 1977). Soil texture or particle size analysis was dominated by the < 2 pm or clay fraction (Table 1.11). All farms had total clay close to 60% but farms 1 and 5 had significantly higher (65%) clay fractions. As a result, sand and silt fractions were slightly lower for those two farms. Among farm variation may have been due to differences in aggregate stability or inconsistent organic matter

removal pretreatment with H 2 O 2 . If pretreatment was incomplete and organic matter not completely removed, stable aggregates of clay-size particles may have contributed to either sand or silt fractions, thus underestimating the clay fraction. El Swaify (1980), in contrast, stated that clay recovery was not dependent on organic matter removal; rather, clay recovery was enhanced by either ultrasound treatment or peptizing agents like Na-hexametaphosphate and NaOH.

Soil Bulk Density

As stated in the methods, I measured soil bulk density at the beginning and end of the field study. The end-of-study sampling design 55

Table 1.11. Soil particle density end particle size analysis.

Farm Particle density Clay Sand Silt gcm-3 -----Total (%) ------1 2.63 65.11 a 7.54 a 27.53 a 2 2.63 58.51 b 11.36 b 30.27 b 3 2.58 56.59 b 13.73 be 29.81 ac 4 2.62 58.02 b 11.72 b 3 0.50 b 5 2.62 65.12 a 9.78 b 25.36 a No significant differences among farms for particle density. Farms with different letters differ at the p -0,05 level (Tukey's HSD multiple comparisons post test) for each particle size fraction. 56 differed from the beginning in that I sampled close to or away from the tree trunk base in those treatments with trees. Therefore, I conducted two analyses of variance; one for the tree treatment data set which included tree distance (Table 1.12a) and one for the complete data set which included sampling time (Table 1.12b). In the first ANOVA, there were significant farm and treatment effects (p < 0.05 level) but no significant tree distance effect. Although there were significant treatment*distance and treatment*farm*distance interactions, inspection of the means revealed no consistent pattern; i.e., for certain farm and treatment combinations, bulk density may have been lower close to the tree base, while in others it may have been lower away from the tree base. Because tree distance was not considered an important distinguishing variable, tree distance values were pooled as overall treatment means for the second ANOVA. This ANOVA was an incomplete factorial because I was able to sample only four out of five farms at the end of the study. Results suggested a strong farm effect, a borderline treatment effect (p = 0.06) and strong significant interaction between sampling time and treatment. This interaction can be explained easily when the treatment means are reviewed (Table 1.13). Treatment P was the only treatment with a significant decrease in bulk density over the three-year study (from 0.89 to 0.83 g/cm^l. This probably resulted from the sustained lack of animals along with no trees (the presence of trees may have slowed the rebound effect from lack of grazing in the trees, no grazing treatment). When treatment means were pooled over farm, there was no change in bulk density, except for a borderline decline in farm 5. In all cases, the soil bulk density range of 0.84-0.91 was subtantially less than 1.0 g/cm3, a range Table 1.12. Analysis of variance tables for soil bulk density.

Table 1.12a. ANOVA for end of study data, tree treatments only.

Source Degrees of Freedom F-ratio P

Farm 3 4.65 0.008 Treatment 1 10.11 0.003 Tree distance 12 0.80 0.377 Farm*Trmnt 3 1.20 0.326 Farm*Dist 3 0.46 0.710 Trmnt*Dist 1 6.39 0.017 Trmnt#Farm*Dist 3 2.99 0.046 Error 32 This ANOVA includes only 4 farms and 2 tree treatments

Table 1.12b. ANOVA on complete data including sampling time.

Source Degrees of Freedom F-ratio P

Farm 4 16.18 0.000 Treatment 3 2.48 0.062 Farm*Trmnt 12 1.27 0.239 Sampling time 1 2.40 0.123 Time*Trmnt 3 6.45 0.000 Error 222 This is an incomplete design for reasons discussed in text. Table 1.13. Soil bulk density via core method presented as means by farm and by treatment for the beginning and end of study. ______

Farm beg. end ------g cm-3— 1 0.85 0.85 2 0.86 0.86 3 0.84 0.83 4 0.90 nd 5 0.91 0.88 LSD - 0.03

Treatment beg. end ------g ,cm-3— T 0.87 0.85 G 0.87 0.87 TG 0.87 0.88 P 0.89 0.83 LSD = 0.03 Bulk density not measured for farm 4 at end of study. Treatment T = trees alone; G = grazed, no trees; TG = grazed with trees; P = non-grazed, no trees. 59 characteristic of soils with andic properties and high clay content (El-Swaify, 1980; Shoji and Ono, 1978; Maeda et al., 1977; Wielemaker and Lansu,

1991).

Moisture Retention Characteristics

Volumetric moisture content plotted against pF or the inverse log of water suction {in cm) indicated that soil moisture changed little from saturated conditions to tensions close to 1 bar {Figure 1.11). The curve, calculated from farm means (ANOVA showed no significant difference among farms for all points on the curve except for 3 bar or pF 3.5), showed that saturated moisture content was about 63% by volume and decreased gradually to approximately 48% at 1 bar {pF 3). Field capacity, which several researchers suggest is more appropriately characterized at pF 1.7 (Lai, 1980; Maeda et al., 1977), was 53% by volume. The greatest decrease in soil moisture content occurred between 1 and 3 bar (pF 3 and 3 .5) and coincided with the break between measurements made on intact soil cores and those made with disturbed soil samples. Lansu (1988), found a similar abrupt change in volumetric water content between pFs 2.7 and 3.0 and suggested that the drop may due to a change in methodology. Even at 1 5 bar tension {pF 4.2), however, volumetric moisture content still hovered around 20%. This relatively high moisture retention at such high tensions was most likely a function of the soil's high clay content. Plant available water, calculated as the difference in water content between pFs 1.7 and 3.5 was 29.2% volumetric or 30.7% gravimetric. horizon. MoistureNeguevretention soilthe of characteristicsFiguresurface 1.11. pF (-log cm water suction) 0 2 1 3 4 5 10 lmerc itr c e (%) t ten n co oisture m etric olum V 20 uk soil bulk 30 40 50 ol cores •oil 070 60 0 6 6 1 The total porosity, averaged over all farms, was 64%. There were two important pore size ranges which accounted for the greatest percentage of total pores drained (Figure 1.12): 1) those drained between saturation and pF 1.7 (pore diameter > 0.0025 cm), and 2) those drained between pFs 3.0 and 3.5 (pore diameter between 0.00015-.00005 cm). The first

range, which constituted about 2 0 % of the total porosity, corresponded to the macropores while the second range, accounting for almost 52%, represented the proportion of micropores present. Again, the high percentage of micropores was most likely a function of the soil's high clay content (El Swaify, 1980; Gavande, 1968). The low percentage of macropores was evidence for soil compaction (Lai, 1989b).

Soil Mineralogy

Mineral Oxide Characteristics

Values obtained for oxalate extractable Fe and Al were high (1.64% Fe, 0.94% Al) but consistent with other soil chemical and mineralogical analyses which characterized the soil series as a highly weathered andisol (Table 1.14). There was no significant difference between the surface and subsurface horizons. Although oxalate is believed to extract poorly crystallized Fe forms, the high values obtained may be an overestimate of this fraction due to the presence of magnetite. Magnetite was separated mechanically from whole soil using a magnet, and X-ray diffraction confirmed its presence. If magnetite is present in substantial amounts, it catalyzes the oxalation reaction, thereby dissolving a crystallized mineral 62 55

50

45

y 4 0

~ 35 w O o 30 CL 2 25 o I- * 20 © o © 15 Q_ 10

2500 1500 750 300 150 50 9.5

Pore Size Diameter (microns)

Figure 112. Percent of total pores drained over the range of pore size diameters from macropores to micropores. Pore size diameters correspond to the difference between successively increasing suction pressures. 63

Table 1.14. Mineral oxide characteristics of study soil's surface <0-15 cm! and subsurface (15-30 cm) horizons.

Oxalate-Fe Oxalate-AI CBD-Fe OxahCBD Fe ------%------Ratio

0-15 cm

1.64 0.94 9.79 0.17

15-30 cm

1.48 1.04 9.76 0.15 CBD-Fe is citrate bicarbonate dithionite extractable Fe. 64 along with the amorphous Fe and overestimating the poorly-crystallized Fe fraction (J. Bigham, OSU, pers. comm., 1991). The percent Fe extracted with citrate bicarbonate dithionite (CBD) in the < 2 pm fraction was also high 9.8%) and most closely correlated with the soil's crystallized Fe oxide content (geothite, magnetite, maghemite). The ratio of oxalate- Fe:CBD-Fe is a measure of percent poorly crystallized Fe. The values obtained (15-17%) were considered high relative to oxisols but were probably representative of volcanic material-derived soils (Bigham et al., 1978). The ratio should be viewed with some caution since the oxalate-Fe was probably not an accurate estimation of the amorphous fraction for this soil.

Differential Thermal Analysis

DTA produced two prominent endotherms for both surface (0-15 cm) and subsurface (15-30 cm) horizons: the first at 250 °C and the second at 440°C (Figure 1.13). There was a broad and weak exotherm between 250- 440°C which reflected organic matter oxidation (Calhoun and Carlisle, 1971). The 250 °C endotherm represented gibbsite while the 440°C asymmetric endotherm was characteristic of either poorly crystallized kaolinite or a mixture of kaolinite and halloysite. The endotherm for kaolinite usually occurs between 500-550°C (Bigham, pers. comm., 1991), while halloysite normally exhibits an asymmetric 575°C peak (Calhoun and Carlisle, 1971). By calculating the area associated with each endotherm, I determined the percent kaolinite/halloysite (30.8% for the 0-15 cm horizon; 38.0% for the 15-30 cm horizon), gibbsite (16% for 0-15 cm depth; 19.2% 65

0-15 cm

15-30cm

Figure 1.13. Differential thermal analysis (DTA) for the farm-composited clay (< 2//m) fraction of the Neguev series surface (0-15 cm) and subsurface (15-30 cm) horizons. for 15-30 cm depth) and Fe-oxides (15.6% for 0-15 cm depth; 15.5 % for 15-30 cm depth). Although these were only estimates, there appeared to be slightly higher kaolinite/halloysite and gibbsite in the subsurface horizon.

X-ray Diffraction, Scanning Electron Microscopy and the Formamide Test for Halloysite

The x-ray diffractogram for the < 2 pm fraction corroborated results from DTA; the dominant minerals were poorly crystallized kaolinite or halloysite, gibbsite and iron oxides (Figure 1.14). The differential pattern of untreated and CBD-treated samples showed sharp peaks for gibbsite, goethite, magnetite and maghemite; these peaks substantiated the high percentages obtained from both CDB extraction and DTA (Figure 1.15). The asymmetry and breadth of the two kaolinite/halloysite peaks (0.74 and 0.36 nm) created doubts about the exact mineralogy of the 1:1 aluminosilicate present. Pure kaolinite usually gives a sharper, more symmetrical peak (J.Bigham, pers. comm., 1991). In addition, SEM photographs suggested the presence of tubular crystals, formations more characteristic of halloysite than kaolinite (Figure 1.16). X-ray diffractograms of kaolinite standards exhibited no change in the d-spacing when treated with formamide (Figure 1.17). Halloysite standards, on the other hand, produced the characteristic shift from 0.7 nm to 1.0 nm when treated with formamide (Figure 1.18). Such a shift reflected hydration of the halloysite structure when treated with formamide. X-ray diffractograms for both formamide-treated surface and subsurface < 2 pm study soil samples did not show the shift associated with halloysite (Figure 1.19). Therefore, there Gl 0,487nm 1.2 -

1.1 - 0.439nm 1.0 GO Gl Gibbsite 0.9 H 0.417m ri MG GO GoethHe GO+G) 0.253nm 0.8 0.173nm MG K KaoNnlte 0.148nm 0.7 H Hatayslte 0.336nm 0.6 MG Magnetite H,K SMaghemKe 0.5 0.739nm

0.4 - MG 0.297nm 0.3 -

0.2 -

0.1 -

0.0 70 50 30 10 Degrees 29 CuKa 35kV 20mA

Figure 1.14. X-ray diffractogram of the Neguev surface (0-15 cm) horizon clay (< 2//m) fraction. 68

Qi OO QO 0 0 + 0 1 a«17n*i K 0.173nm MO H 0.i4anm

azranm

MG \ ICL297 MX Untreated O.TMnmI

* l A -

CBD-treated ILA

DXRD

70 50 30 10 Degrees 26 CuKa35kV20mA

Figure 1.15. Differential x-ray diffractogram for untreated and CBD-treated < 2 um fraction, surface horizon. Figure 1.16. Scanning electron microscope (SEM) photograph of the < 2 fjm fraction showing an apparent tubular structure. 70

0 .7 l4 n m

A ' A ' A ' A 1 A 1 J 1 1 1 1 ■ j ■

Degrees 26 CuKa 35KV 20mA

Figure 1.17. X-ray diffractogram of kaolinite standard untreated (a) and formamide treated (b). 1.01

e

Dogmas 20 CuKa 35kV 20mA

Figure 1.18. X-ray diffractogram of halloysite standard {API#12) untreated (a) and formamide treated (b|. B

i

I f b •.Ml

-A,A, 4 ‘4 , fcliitA,A, l ,4 l l 1 I 1 DagrawM CtfsmvtflmA bfMN MDWIM

Figure 1.19. X-ray diffractogram for the surface (0-15 cm) horizon (A) and subsurface (15-30 cm) horizon (B), < 2//m fraction untreated (a) and formamide treated (b)

vi to was no conclusive evidence for halloysite in the study soil. The most likely scenario is that kaolinite was present in a very poorly crystalized form.

DISCUSSION

Baseline Characterization: Soil Chemical Parameters

The volcanic-material derived soils of Central and South America have been well characterized; their soil chemical properties exhibit striking homogeneity (Muller et at., 1968; Fassbender and Bornemisza, 1987; Alvarado, 1984; Sanchez and Cochrane, 1980). Even volcanic-ash derived soils from Japan possess similar ranges of exchangeable bases: Ca 0.2-3.2; Mg 0.3-1.5; K 0.3-0.7 (Shoji and Ono, 1978; Egawa, 1984). Ca is generally the most abundant exchangeable base followed by Mg and then K. Exchangeable Na is present in very low amounts (Egawa, 1984). The extremely low values obtained for exchangeable Ca, Mg and K for the study soil are typical of highly leached volcanic-material derived soils, and they corroborate results obtained by de Bruin (1991) and Lansu (1988) for the Neguev soil series. In a survey of soils from Central America, Muller et al.(1968) found a notable decrease in exchangeable bases in Costa Rican soils under acid conditions. Sanchez and Cochrane (1980) mention the low CEC of many soils of the tropics and consider an ECEC < 4 cmolc/kg soil to be a major constraint to crop production. They estimate that 80% of oxisols and ultisols in the tropics have saturated Al > 60%, based on soil pH values less than 5.0. In general, Al dominates both the exchangeable acidity and effective CEC of the Neguev soil series. The importance of free 74 Al hydroxide is further evidence for the soil's advanced stage of weathering (van Dooremolen et al., 1990a). Just as ECEC is dominated by Al, the soil's pH is strongly regulated by free Al. There is a highly significant correlation (albeit not very linear as r^<0.42 values suggest) between both pH in H 2 O and pH in KCI and exchangeable Al and % ECEC as Al (F statistics and P values for the regression H 2 O pH against exchangeable Al and % ECEC/AI are as follows: for exchangeable Al, F= 16.7, p<0.0001; for % ECEC/AI, F= 57.1, p <0.0001; for KCI pH regression: exchangeable Al, F = 13.7, p<0.0001; % ECEC/AI, F = 26.2, p <0.0001). As with CEC, the Neguev series pH range is typical of many acid, volcanic-ash derived soils (Alvarado, 1984; de Bruin, 1991; Shoji and Ono, 1978). In a study relating volcanic deposit age with soil properties along a soil chronosequence, van Dooremolen et al. (1990b) found that KCI pH decreases with increase in soil age. Furthermore, the ratio of pyrophosphate-extracted Al to oxalate-extracted Al increases asymptotically as KCI pH decreases. This ratio, a measure of both crystallinity (1 being highly crystallized, 0 being highly amorphous) and Al associated with organic compounds (pyrophosphate supposedly extracts Al associated with the soil's organic components), suggests that older, more weathered volcanic soils have more crystallized, C-complexed Al than more recently deposited ash soils. Very few studies present complete soil chemical characterization of volcanic-ash derived soils; therefore, there are almost no reference values for parameters like modified Olsen-extractable K and micronutrients. Values obtained for K fall in a range similar to NaOAc-extracted K. Kass et al. (1985), report Olsen-extractable Mn ranging from 52-168 mg/kg for soils in 75 the San Carlos region of Costa Rica. These values are more than an order of magnitude higher than those obtained for the Neguev soil series (highest value is 5.7 mg/kg) and, in fact, the San Carlos soils were chosen to study common bean response under high Mn and Al levels. Double dilute acid- extractable Mn exceeding 100 mg/kg is considered plant toxic (Sanchez and Cochrane, 1980). There are considerable citations for organic C, N and C:N ratios, no doubt reflecting the concern for key soil constraints to plant production (Alvarado, 1984; Bornemisza, 1966; Fassbender and Bornemisza, 1987; Lansu, 1988; Muller et al., 1968). For many acid soils with inherently low CECs, organic matter contributes a targe percentage of the CEC; hence the critical link to soil fertility (Lai, 1989a). In a survey of 167 Central American soils, Fassbender and Bornemisza (1987) found that organic C ranged widely (0.4-12.2% with a mean of 2.96%). The Neguev soil has a slightly higher than average % OC when compared with this survey, yet its organic C content is much lower than soils of more recent volcanic deposition. (Alvarado (1984) cites 9.2% OM or 5.4% OC for volcanic ash-derived soils in Guatemala and Costa Rica). Both Lansu (1988) and de Bruin (1991) obtained values nearly identical to my results for both C (3-5%) and N (~ 0.3%). Over 70% of the soils analyzed from Central America have TKN values between 0.1-0.4% (Fassbender and Bornemisza, 1987). The literature on volcanic-ash soils also reports C:N ratios ranging from 9-12; again, the Neguev soil falls within this range. Relative to other non-volcanic origin soils of the tropics, the C:N ratio is low, probably due to the relatively higher TKN values (Lai, 1989a; Sanchez and Cochrane, 1980). 76 Given the mineralogical characteristics of soils derived from volcanic materials, anion nutrients, particularly P, present the greatest limitations to plant availability (Bertsch and Cordero, 1984; Shoji and Ono, 1978; Parfitt, 1980). As such, the literature on volcanic soils P forms abounds (Bornemisza, 1966; Fassbender, 1968; Fassbender et al., 1968; van Dooremolen et al., 1990b; Condron et al., 1985). For soils of the tropics, much of the early literature on P forms considered pH buffered, salt solution extractions like Olsen (NaHC 0 3 -EDTA, pH 8.5), Egner-Riehm (0.02 N Ca-

(actate, pH 3.8) and NH 4 CI best correlated with the plant-available P fraction (Balerdi et al., 1968; Fassbender, 1968). Many of the studies used a succession of increasingly harsh chemical extractants to characterize the increasingly recalcitrant P fractions. The organic P fraction is believed to follow salt extracts in terms of plant P supply, whereas the Fe-P, Al-P and Ca-P fractions do not correlate well with plant uptake (Fassbender et al., 1968; Condron et al., 1985), In general, most of the literature on volcanic-ash derived soils cites high total and organic P fractions and low plant-available P. For Costa Rican soils with andic properties, organic P ranges from 60-80% of total P (Kass, 1991) and for pasture soils in general, the range is between 50-80% of total P (Condron et al., 1985). In a detailed study of Costa Rican soils, Palencia O. and Martini (1970) cite plant available P ranges from 0.1 -6.3 mg/kg. P retention fluctuates between 80-95% (Bertsch and Cordero, 1984). Fassbender et al. (1968) classified 110 Central American soils as either Ca-phosphate or Al-phosphate dominated. The Neguev soil would fall into the Al-phosphate category. The Al-P group has a high total P content (1242 mg/kg average) and organic phosphates dominate (52% of total P is 77 organic). Most soils in the Al-P group are extremely P deficient ( 6 6 %) and only 15% have an adequate P supply for plants. Shoji and Ono (1978) and Daage (1987) concur that the sesquioxide constituents of andosols (e.g. gibbsite) contribute as much to the soils' high P retention capacities as the poorly-crystallized minerals like allophane and imogolite. Indeed, several studies have found a strong correlation between total Al and oxalate- extractable Al with P retention capacity (Kass, 1991; Russell et al., 1988; Le Mare, 1982; Saunders, 1965). The Neguev soil’s P adsorption isotherm depicts more graphically the incongruity between high total P and low available P. As the Geochem model predicts, the high concentrations of free Al coupled with high soluble P at high P additions leads to precipitation of amorphous Al-phosphates. Shongwe’s(1989) study of acid Swaziland soils showed that without P additions adsorption was the dominant mechanism of P retention. With inorganic P additions, the soils became supersaturated with respect to variscite, suggesting that precipitation of Al-P minerals is an important component of acid soil P retention. Parfitt (1989) also confirms that AJ- phosphates are likely to precipitate at high levels of P addition in soils containing large amounts of reactive Al. In summary, the Neguev soil does not behave like a classic andosol or a classic oxisol; it’s high P retention capacity is more characteristic of volcanic ash-deposited soils, whereas its low pH, low CEC and high exchangeable Al are more reminicent of highly weathered soils. Organic C and N levels are intermediate when placed along an andosol-oxisol continuum. As van Dooremolen et al. (1990a) point out, the Neguev series falls along the older end of the spectrum of volcanic- derived soils. 78

Baseline Characterization: Soil Physical Parameters

The Neguev soil's high clay content is consistent with PSA characterization of highly weathered soils (Ei Swaify, 1980; Egawa, 1984). Oxisol clay content ranges from 56-90%. The Neguev soil is not classified as an oxisol, yet its clay content lies within the lower end of this range. Lansu (1988) reports percent clay values for the Neguev soil under pasture as 57-58% in the 0-30 cm horizon. The variation in percent clay can be a function of the methodology used for particle size analysis (PSA). For soils with poorly crystallized mineralogy, PSA is not a good index property since soils with allophane and other poorly crystallized minerals do not disperse well {Maeda et al., 1977). Consistent with the soil's low bulk density and high clay content, the Neguev soil has a high moisture rentention capacity, even at high suction pressures. According to Warkentin and Maeda (1980), andosols have a characteristically high 15 bar (pF 4.2) or wilting point moisture content. As compared with the 0.33 bar range for oxisols with high clay contents (28- 42%), the Neguev soil's 0.33 bar (pF 2.3) volumetric moisture content is much higher, nearly 50%. Total porosity for andic soils, calculated from saturated moisture content, ranges from 65-77 (Shoji and Ono, 1978); the Neguev soil total porosity is approximately 64%. The shape of the moisture release curve is similar to those for soils with high allophane content; i.e. an S-shaped curve with a linear portion between 0.01-1 bar or pF 1.5-3.0 (Warkentin and Maeda, 1980; Maeda et al., 1977; Gavande, 1968; Egawa, 1984). Although the Neguev soil has no allophane, the high clay and 79 moderate organic matter contents coupled with high total porositv contribute to the soil's andosol-like moisture rentention characteristics. For non-compacted andic soils, the greatest percentage of pores are drained between pF 1-1.7; these pores reflect the soil's permeability or drainage capacity (Egawa, 1984). Sollins and Radulovich (1988) demonstrate that the high percentage of macropores in well-aggregated oxic dystropepts of Costa Rica result in preferential or channelized flow between aggregates, bypassing micropores even when they aren't fully saturated. They further propose that this preferential flow serves to prevent nutrient leaching from the soil matrix, and that when tropical forests are cleared and converted to pasture, much of the soils' macropores are lost, thereby creating greater susceptiblity to nutrient loss from micropores. The Neguev soil's low percentage of pores drained under low suction pressures suggests that permeability is not great and consequently preferential flow has been reduced. Anecdotal observation of water ponding following prolonged rainfall is further evidence for the soil's compaction and low permeability despite its well-drained status. The Neguev soil's pore distribution curve is best described as two peak; the first around pF 1.3 and the second, larger peak around pF 3.0. The low tension peak, corresponding to non-capillary porosity, is usually 8.4-11.9% for andosols; while the high tension peak, or capillary porosity, averages around 23% (Egawa, 1980).

Baseline Characterization: Soil Mineralogy

The aforementioned chemical and physical properties of the Neguev soil are, largely, manifestations of the soil's mineralogy. As such, 8 0 characterization of the dominant minerals helps explain the somewhat oxic, somewhat andic nature of the Neguev series. For example, CBD-extractable Fe for oxisols ranges between 7-14% (Bigham et al., 1978); the Neguev soil's CBD-Fe (9.8%) falls squarely within this range. However, oxalate extractable Al for oxisols is usually between 0.24-0.55%; the Neguev soil has almost twice that amount (0.94-1.04%), suggesting that there is still a significant amorphous Al component. As Blume and Schwertmann (1969) suggest, the soil's low pH coupled with some downward movement of organic matter may actually retard the crystallization process resulting in high relative and absolute amounts of poorly crystallized Fe and Al oxides. DTA and XRD analyses provide further evidence for the dominance of Fe and Al oxide minerals, but do not resolve the exact nature of the kaolinite/halloysite contribution to the soil's mineralogy. The tack of a 1.0 nm peak in X-ray diffractograms in conjunction with the formamide test rule out the presence of hydrated halloysite (Calhoun et al., 1972). However, the broad 0.74 nm XRD peak along with the 440°C DTA endotherm create ambiguity as to the actual kaolinite/halloysite constituents. Perhaps Herbillon (1980) best resolves this ambiguity by suggesting that kaolinites in oxic materials are largely poorly crystallized. Such poor crystallinity is often exhibited in the XRD pattern as broad peaks which can be confused further if oxide coatings are present. Indeed, some soil kaolinites display such broad diffraction bands as to preclude an accurate determination of the mineral. There have been several studies of Central and South American soils relating soil chronology with mineralogy (Bornemisza, 1969; Calhoun and Carlisle, 1971; Calhoun et al., 1972; Gomez et al., 1981; van Dooremolen 81 et al., 1990b). Most have used a minimum of XRD and DTA techniques to characterize dominant minerals along volcanic deposition chronosequences. According to Herbillon (1980), most variable charge soil clays can be considered as a "continuum from completely disordered, through poorly ordered to well crystallized material." The typical weathering sequence for soils derived from volcanic materials begins with allophane succeeded by metahalloysite and imogolite followed by hydrated halloysite and ending with kaolinite. The weathering process is accelerated at lower elevations where soil temperatures are constantly high and rainfall is abundant and intense (Calhoun and Carlisle, 1971). Within Costa Rica, soils of the Atlantic coastal region are considered relatively young, in general, due to the presence of amorphous minerals and halloysite (Bornemisza, 1969). van Dooremolen et al.'s (1990b) thorough survey of the Atlantic coastal region's soils provides conclusive XRD evidence for the dominance of kaolinite, gibbsite and goethite in the older volcanic deposits like the Neguev series.

Changes in Selected Soil Chemical and Physical Properties Over the Study Period

The most important result for all soil chemical parameters studied is that the experimental treatments had no significant effect over the three- year period. This is not uncommon for experiments involving trees, in general, and agroforestry experiments, in particular (Sanchez, 1987; Kass et a., 1991; Vilas B., 1990; Lai, 1989a; Rao and Coe, 1992). In most cases, the effects from residue inputs and subsequent changes in organic matter fractions do not manifest themselves until five years after tree establishment. Rao and Coe (1992) stress the importance of evaluating 82 agroforestry systems over a minimum of five years to: 1 } better identify true long-term trends and 2 ) account for the interaction between climatic variation and treatment effects. Lai (1989a), for example, found increases in organic C, N and exchangeable Ca and Mg in agroforestry treatments

(Leucaena feucocephafa with maize) after five years. Despite this general trend, there are some specific changes over time among the five farms. Exchangeable Ca, for example, increases on farm 1; relative to the other farms, the increase is significant, but the absolute change in terms of soil fertility is probably minor. It is also unusual since both exchangeable Al increased and soil pH decreased markedly.

In fact, the increases in exchangeable Al, NaHC 0 3 -EDTA extractable

Fe and decrease in both H 2 O and KCI pH are dramatic for all farms at both soil depths. The two are most likely related synergistically. In general, divalent and trivalent cations (A|3 + , Ca2 + , Fe3 + ) increase at the expense of monovalent cations (K + , H + ). This phenomenon is a common by­ product of the weathering process; i.e. polyvalent cations are held selectively by the soil matrix whereas the monovalent cations are more easily leached (Egawa, 1984; Haynes and Swift, 1986). Although organic C and N do not change over time, there are treatment differences in the C:N ratio. In the surface horizon, the C:N ratio increases significantly in the trees alone (T) treatment; in the subsurface, it increases for both non-grazed treatments. The non-grazed treatments are storing carbon. Alternatively, the T treatment is adding inorganic N as N mineralized from decomposing Erythrina leaves. This inorganic N input tends, in turn, to increase mineralization of organic soil N reserves, resulting in decreases in TKN (Bornemisza, 1966). Lower TKN translates to higher 83 C:N ratios. Most likely, however, these treatments, because they have no export of residues, are behaving as fallowed agro-ecosystems. The changes in soil P forms can be explained in a similar manner as

TKN. Organic P, for instance, decreases among all farms while NaHC 0 3 ~ EDTA extracted P and total P increase. In a Costa Rican study monitoring changes in organic P following deforestation, Bornemisza (1966) found that increased soil temperatures tended to increase organic P mineralization. The mineralization led to increases in the inorganic P pools which, in turn, stimulated further mineralization of the organic pool. Bertsch and Cordero (1984) also found that when inorganic P was added to P-deficient soils, mineralization of organic P and N reserves increased. Because these and the other soil chemical changes have occurred, for the most part, at the farm level and not at the treatment level, they may be reflecting the more global, gradual changes from deforestation and consequent changes in land use; i.e. nutrient leaching and increased acidification (Wielemaker and Lansu, 1991; Sollins and Radulovich, 1988). Soil bulk density w as the only soil physical param eter m easured over time. When treatment means are averaged by farm, there is no change. However, when farm means are averaged by treatment, there is a significant decrease in the control treatment after three years, as might be expected. When compared with the bulk density for the forest soil on the sam e soil series (0.63 g/cm3), the control treatm ent bulk density is still relatively high. Nonetheless, it appears to be rebounding from the compaction incurred through continuous grazing (Humphreys, 1991). The non-grazed tree treatment has not undergone such a significant decline in soil bulk density which suggests that the tree roots are slowing the rate of 84 decline. Lai (1989b), in contrast, found that alleycrop agro-ecosystems (leguminous trees with maize and cowpea) had lower bulk densities relative to non-agroforestry treatments. This may not be the most appropriate comparison since annual crops have such different rooting patterns than pasture grasses. Although the grazed treatments have been converted from continuous to rotational grazing, there probably has not been sufficient time for bulk density to reflect the change in livestock management. In addition, the possible ameliorative tree effect in a grazed system has not developed yet at this early stage in the silvopastoral system's establishment. CHAPTER II

MEASURING IN SITU CHANGES IN LABILE SOIL P WITH ANION EXCHANGE RESIN-IMPREGNATED MEMBRANES

INTRODUCTION

Historically, the quantitative assessment of nutrient bioavailability, particularly P, has lagged behind its conceptual characterization. We accept, for example, that diffusion supplies almost all plant P and that transport to plant roots depends on the concentration gradient between the soil matrix and the root surface along with the diffusion coefficient (Bhadoria et al., 1991; Kovar and Barber, 1988; Abrams and Jarrell, 1992). We also acknowledge that the general model for available P includes an intensity factor, a quantity factor, a buffer capacity factor as well as rate and diffusion factors (Dalai and Hallsworth, 1976). The intensity factor, or soil solution P, is determined by the amount of solid phase P (quantity factor) which will be released into solution. The buffer capacity, or labile P, reflects the soil's ability to replenish the soil solution as plants withdraw P (Fox, 1981). The kinetics of the equilibrium reactions between soil matrix P

solution P <-*■ labile P are, in turn, affected by soil mineralogy (Al and Fe sesquioxide content, poorly crystallized mineral component), soil structure, soil pH and organic P pools (Aharoni et al., 1991; Mattingly, 1975; Parfitt et al., 1989; Tiessen et al., 1984).

85 8 6 Conventional estimates of plant available P employed chemical extractants. The basic protocol was to extract an air-dried, sieved (<2 mm) soil sample with a mild salt or acid solution using short extraction times and a small extractant volume to soil weight ratio (Wild, 1988). Despite their widespread use, chemical extractants are not well understood in terms of their mode of action and selectivity (Curtin et al., 1987). Furthermore, they are limited in their ability to measure accurately bioavailable P because they are operationally defined by the chemical extractant used (will react differently with different soil types) and they only measure static P pools (Abrams and Jarrell 1992). Another major disadvantage is that they may mobilize P forms other than those that are truly plant available (Logan, 1982; Menon et al., 1989). More than three decades ago, Amer et al. (1955) proposed the use of anion exchange resins in an attempt to better quantify labile soil P. They postulated that the resin would remove anions like P from the soil solution without returning another phosphate anion in exchange. In this way, they hypothesized that the resin would behave as an ion sink whose sorption rates would depend on the same factors and processes which regulate nutrient supply to plant roots. This was a major methodological breakthrough in that the rate of resin sorption depended solely on the rate of desorption and/or dissolution from the soil matrix and not on the properties of the resin itself (Palma and Fassbender, 1970; Krause and Ramlal, 1987). Early studies with ion exchange resins (AER) involved equilibration of disturbed soil samples with granular resin via shaking followed by physical 87 separation of soil and resin and analysis of resin-sorbed nutrients (Amer et al., 1955; Igue and Fuentes, 1971; Sibbesen, 1977; 1978). Igue and Fuentes (1971) found resin P correlated most closely with either Ca- or Al-P forms, thus making inferences about the sources of resin-extractable P. The AER procedure became useful as a batch technique for routine analysis of plant available P. Although it was a major improvement over chemical extraction methods, it proved inadequate for precise characterization of labile P because it did not account for rate phenomena like diffusion (Abrams and Jarrell, 1992; Vang et al., 1991b). Vaidyanathan and Nye (1966) introduced the use of strong base anion exchange papers placed on the soil surface to measure diffusive soil P fluxes in a less disturbed setting. However, the filter paper introduced several sources of error including: mass flow from the wet soil to the relatively dry paper; imperfect contact between the paper and the soil surface caused by large interaggregate cavities; and microflora colonization of the filter paper after several days incubation with moist soil. Barraclough and Tinker (1981) attempted to resolve these problems using hydrophyllic polymeric membranes to measure P diffusion in both field and disturbed soil cores. Skogley et al. (1990; 1991) and Vang et al. (1991a: 1991b) promoted mixed resin, rigid porous capsules as a phytoavailability index under saturated soil paste conditions. They claimed that the resin functions as an infinite sink allowing continuous ion adsorption over long periods because the volume of soil from which nutrients diffuse extends only a few millimeters from the surface of the resin capsule. They also asserted that the mixed-bed resins (cation and anion exchange together) better simulates plant P uptake than anion exchange 88 resins alone. In all of these cases, nutrient flux rate was diffusion- controlled, and the principal mechanism of sorption was equilibrium (non- selective or non-ligand bond) anion exchange with a large capacity exchanger. Several researchers have expanded upon these techniques for field use (Sibbesen, 1977; 1978; Binkley and Matson, 1983; Gibson et al., 1985; Lajtha and Schleshinger, 1988; Skogley et al., 1990). Most have experimented with the buried resin bag technique which incorporated granular ion exchange resins in polyethylene mesh bags to measure nutrient changes for both undisturbed and agricultural ecosystems. Such studies took the ion exchange resin methodology one step further by developing in situ techniques which perturbed the ecosystem less. Nonetheless, they have made no attempt to understand the mechanism of resin sorption under field conditions, nor have they explored resin kinetics as related to either soil properties or to bag/capsule geometry. For example, they have not provided a theoretical model which includes anion diffusion inside the resin- filled bag. In addition, they have not investigated the effect of varying resin placement time on nutrient sorption characteristics. In general, most studies have assumed that the resin behaves as an infinite sink; i.e., once nutrients like P are adsorbed to the resin, they are not desorbed. These factors may not be critical for use as a bioavailability index, but they are important for measuring nutrient fluxes under field conditions. As such, these studies have limited interpretive capability with respect to nutrient fluxes. Recently, interest has developed in the use of rigid ion exchange materials such as oxide-impregnated papers and synthetic resin membranes 89 to measure P bioavailability and ion nutrient suppiy (Huettl et al., 1979; Van der Zee et al., 1987; Menon et al., 1989; Sharpley, 1991; Abrams and Jarrell, 1992). In the case of oxide-coated resins and filter papers, they take advantage of the strong affinity of Al and Fe oxides for P and the formation of essentially non-reversible ligand bonds between the metal and phosphate anions. Such materials are true sinks for P. In fact, P can only be desorbed from the material by strong acid extraction. These materials have not been tested under field conditions, however. The resin- impregnated membranes, although like resin beads are dynamic exchangers for P, have been shown to correlate strongly with bioavailable P and don't present the three dimensional limitations of resin bags (Abrams and Jarrell, 1992). To date, all of these techniques have been evaluated as soil test methodologies. None have been characterized for their potential to measure nutrient dynamics under field conditions. Both filter papers and membranes hold promise for in situ measurements of ion fluxes because they are essentially two-dimensional structures which don't possess the diffusion problems of resin-filled nylon bags. In this paper, I evaluate several anion exchange materials for use in a field experiment of P dynamics in a Costa Rican silvopastoral system. I refine and calibrate methodology for anion exchange filter papers (AEF) and resin-impregnated membranes (AEM), relate AEF-P and AEM-P to soil solution P for a high P-fixing soil, discuss the theory of AEM sorption under field conditions and present two examples of field use. Resin-impregnated membranes were used in situ to estimate changes in labile soil P related to different management practices in the silvopastoral system. 90

THEORY

Ion exchange materials (including all resins, ion exchange filter papers and resin-impregnated membranes) can be viewed as competitive exchangers with those soil solids that are in dynamic equilibrium with the soil solution dissolved species. In the case of P at a relatively acid pH range

(4.3-5.0), H2 PO4 ' is transferred via the soil solution from the soil solid phase to the ion exchange material (Figure 2.1). The equilibrium reaction which describes the interaction between the resin and soil solution P:

R-NH4 + CI- + H2 PO4 - ++ R-NH4 + H2 PO4 " + Cl- (1)

where R-NH4 + represents the resin matrix with embedded, positively- charged, ammonium functional groups. The reaction is simple ionic exchange of adsorbed Cl' for other anions in solution. In contrast, the equilibrium reaction of H 2 PO4 - with metal oxide coated materials (either Fe- oxide coated filter papers or Al-oxide coated resin) can be characterized mainly as surface precipitation (there also may be some adsorption via ligand exchange):

Fe-OH + H2 PO4 - -» Fe-H2P04 + OH- (2)

This reaction is essentially irreversible. The resultant functional model for exchange resins responds to soil solution P dynamics whereas the metal oxide coated material acts as a P accumulator (Figure 2.2). This Labile Soil P (mg/kg) Soil Solution p (mg/L)p Solution Soil Soil Figure2. bufferedbylabile soilsoil Padsorbedthe matrixto illustrated as the by soilsolutionanion and exchangemembrane for H idealized Panion adsorbedthe exchangeto P adsorption isotherm. idealizedcorrelation curve. membrane ininis,turn,equilbrium with sou solution the describedby Pas H2PC4- 1 Conceptualdiagram soilinteractionsofthe matrix,. the between Soil Solution Soil 2P04- H 2 < Q. s til PO 4 H2P04* . Soil*.solution Pis Soil Solution P (mg/L)P Solution Soil Anion Exchange Exchange Anion Membrane (0 Solution P (mg/L) iue .. ocpul oe rltn si slto P solution soil relating model Conceptual 2.2. Figure dynamics to P sorption of[ a) a dynamic exchange exchange dynamic coated oxide (metal a sink infinite a) of[ an materials). b) sorption and P reslnc to dynamics 0.0 2.0 3.0 4.0 5.0 1.0 5 0 5 20 15 10 5 0 na c r e g n a h c x e ic am yn D niie i > k sin Infinite l ol i P n tio lu so il o S Time

92 93 mechanistic distinction is important in terms of interaction with other anions in solution. Because the mechanism for resin materials is ion exchange, there will be competition between H 2PO 4' and other anions at the resin sorption surface, particularly if other anion activities are high. Conversely,

anion competition is not expected to be important when H 2 PC>4 ~ precipitates on metal oxide surfaces since it converts to a stable solid phase. Another noteworthy outcome of these mechanistic differences is how the two material types react over time. Metal oxide-coated materials will always accumulate P over time. Anion exchange materials will behave as either sinks or exchangers for P depending on: 1) the intrinsic anion

exchange capacity of the resin material; 2 ) the amount of time in contact with the soil; and 3) the soil’s P retention capacity. Throughout the literature, resin materials are described as infinite sinks, likely because their exchange capacities remain large over the study period and/or the soils' P fixation capacities are low enough to minimize competition for P between the resin and soil solid phase. In general, then, most anion exchange resins react rapidly with

H2 PO4 ', and the rate of sorption is limited by the rate of desorption or dissolution in the case of agitated systems, and by pore and film diffusion in

the case of in situ resin placement. Resins can be used to estimate

instantaneous soil solution H 2 PO4 " concentrations by regression analysis. Resin behavior (either as a sink or an exchanger) can be calibrated with soils of differing P retention capacities. Once this relationship is established, resin materials can be used in the field over time to estimate changes in net 94 labile soil P. If sufficient information is available on soil moisture content, porosity and tortuosity, diffusion coefficients can also be calculated.

METHODS AND MATERIALS

Site Description

The bulk of this study was conducted in the laboratory. However, the field experiment for which the methodology was developed was located in the Atlantic coastal plain of Costa Rica (10° N 83° W). The area receives 3630 mm rainfall annually and the ecological lifezone is lowland humid tropical rainforest. The soil is an andic humitropept and is located on the ridgetops of a slightly undulating landscape. The geologic origin is volcanic ash and lahars from the late Pleistocene; due to its stable position on the landscape and high rainfall, however, it has mineralogical characteristics of more highly weathered soils (kaolinite, gibbsite and iron oxides). In general, the Neguev series is deep, well-drained, acid, low in exchangeable bases, high in exchangeable Al, high in total P but low in available P, clayey and low in bulk density (Tables 2.1 and 2.2). In addition, the Neguev soil has an extremely high P fixation capacity (> 2000 mg/kg) and a high moisture retention capacity over a wide range of soil moisture tensions (Figures 2.3 and 2.4).

The field experiment, a silvopastoral system, was a 2 X 2 factorial design with cattle (grazing) and trees as the two independent variables Table 2.1. General soil chemical characteristics of Neguev series surface <0-15 cm) and subsurface (15-30 cm) horizons.!

Soil pH Exchanoeable Bases KCI-exch. ECEC# Organic C TKN® Clay Bulk

Depth Ca Mg K Acidity Content Density

...... cm 1:1 ciiiuiq /Ay sun* Mg/m^

soil:H?0

0-15 5.1 -5.3 0.87 0.95 0 .4 7 1.45 3.74 3.38 0.38 60.7 0.87

(0.28) (0.15) (0.11) (0.35) (0.53) (0.30 (0.081 (4.1) (0.03)

15-30 4.8- 5.0 0.34 0.33 0.26 1.49 2.43 2.23 0.23 ------

(0.12) (0.08) (0.09) (0.521 (0.54) (0.31) (0.05) { Values averaged over five study farms from 1987 soil sampling. Numbers in parentheses are standard deviations. # ECEC is effective cation exchange capacity = sum of NH 4 OAC (pH 4.8) extr. bases + KCI-exch. acidity. @ TKN is total kjeldahl nitrogen.

u> 01 Table 2.2. Maior phosphorus forms for the Neauev soil series. S

Soil Depth (cm I NaHC03-EDTA Organic P Acid Digest

extractable P Total P

______— mg P/kfj soil—

0 - 15 2.79 (0.861 882 (122) 14-76 (445)

1 5 - 3 0 2.60 (0.70) 801 (81) 1238 (166) § Values are averaged over five study farms from 1987 samples. Numbers in parentheses are standard deviations. P P Adsorbed (m g /k g ; 3000 2000 0 5 0 5 0 5 30 25 20 15 10 5 0 Figure 2.3. P adsorption isotherms for surface (0-15 cm) and subsurface subsurface and cm) (0-15 surface for isotherms adsorption P 2.3. Figure (15-30 cm) horizons of the Neguev soil series. soil Neguev the of horizons cm) (15-30 qiiru Pcn. (mg/liter) conc. P Equilibrium usrae Soil Subsurface horizonNeguevsoil the of series. iue24 Soilmoistureretention (0-6cm) characteristics surfaceof Figure 2.4. pF (-log cm water suction) 0 2 3 4 1 5 10 ouerc osue otn (%) content moisture Volumetric 20 uk coil bulk 0 3 0 4 60 0 5 ol cores ■oil 0 7 98 99

(Figure 2.5). I planted Erythrina berteroana (a tropical leguminous tree) from

2 . 6 m cuttings in native grass pastures on five farms, all within a 6 km radius. Each farm contained the randomized block of four treatments and served as a replicate. Grazed treatment plots were 900 m2 and non-grazed treatments 400 m2. The experiment was managed with a five-week grazing cycle coupled to a five-month tree pruning regime. Tree prunings were left on the ground as supplemental forage.

Calibration of Anion Exchange Filter Paper Methodology

I had conducted baseline characterization of labile soil P in the field plots using the buried bag anion exchange resin method (Binkley and

Matson, 1983). I decided not to pursue this procedure for more in-depth in situ measurements because of logistical problems (the bag mesh was disrupted by soil fauna resulting in resin bead loss and cumbersome reweighing of resin for P determination). Therefore, when I began investigating possible techniques for measuring in situ fluxes in labile soil P, I experimented first with anion exchange filter (AEF) papers. Prior to testing under field conditions, however, I performed a series of laboratory experiments to calibrate the method for the high P fixing Neguev soil series. I assessed the relationships between soil moisture, time of AEF incubation in soil and AEF-P extractability. I also established an empirical correlation between AEF-P and soil solution P. Once these relationships had been established, I applied the technique to the field experiment. Pasture with trees & grazing (900 m2) ^

^ ^

Trees/No Grazing 3 m (400 m2)

Nongrazed pasture Grazed pasture (300 m2) (900 m2)

Figure 2.5. Experimental design of the silvopastoral field experiment. The tree species used was Erythrina berteroana planted from 2.5 m vegetative cuttings. The 2 X 2 factorial design was replicated on five neighboring farms. 101 Effect of Agitation Time and Number of Extractions on 1 M NaCI P Extra eta bility

The objectives of this experiment were to: 1) to determine the optimum agitation time for extracting P off the anion exchange filter paper (AEF) with 1 M NaCI as the extracting solution (a previous study comparing

1 M NaCI with 0.5 M HCI showed no significant difference in their extracting ability; NaCI was chosen because it was less harsh; i.e. it didn't extract labile P from soil particles themselves, only from the filter paper); and 2) to determine the number of extractions needed to remove all P adsorbed to the filter paper. The experimental design included: two P levels (5, 10 mg P/L), three agitation times (1,2,4 hours) and two sequential extractions with 1 M NaCI. The resultant 12 treatments were performed in duplicate. I weighed dry, clean Whatman DE81 anion exchange filter paper discs (0.2 mm thick, 2.5 cm diam.) to four decimal places and soaked them separately in 10 mL of each P concentration solution overnight. The filter discs are weakly basic anion exchangers with diethylaminoethyl functional groups and possess an ion exchange capacity of 0.17 m m o l c / m 2 . The following day, I removed the filter papers from P solutions, rinsed with double deionized water (DDW) and placed them in 50 mL centrifuge tubes with 25 mL 1 M NaCI. I shook samples on a reciprocating shaker at low speed for the specified agitation times. Following each extraction time, I decanted extract solutions without removing the filter paper and weighed and recorded the entrained solution volume in each centrifuge. I then added a second dose of 25 mL 1 M NaCI to the centrifuge tubes with their 1 0 2 respective filter papers and placed them on the shaker again for the specified agitation times. The second extract solution was then decanted and saved separately. I determined P concentration in both the initial P soaking solutions after AEF removal and in the 2 NaCI extract solutions using the ascorbic acid-reduced phosphomolybdate method (Olsen and

Sommers, 1982) on a Beckman DU - 6 spectrophotometer (882 nm).

Effect of Time of AEF placem ent and Soil Moisture Content on AEF P Adsorption

Because P sorption is largely a diffusion-driven process, it was important to establish an empirical relationship between AEF placement time, soil moisture content and AEF-P for the Neguev soil. Accordingly, I performed a laboratory experiment in which I incubated AEFs in P-amended soil over a period of days at different moisture contents (near saturation and pFs 1.8 and 3.5). The experimental design included five P additions (0, 5,

10, 15, 20 mg P/L added), three gravimetric moisture contents ( 6 8 , 31,

27%), and four placement times (2, 4, 6 , 8 d).

I w etted five 1 . 8 kg subsamples of air-dried, sieved ( 2 mm), surface horizon (0-15 cm) soil composited from the five study farms to pF 1.8 moisture tension (determined on disturbed samples to be 31% by weight; the non-disturbed moisture content at pF 1.8 is much higher.) I then incubated each wetted subsample with successively higher P levels (O, 315, 630, 945, 1260 mg P/kg soil) for 28 d. Following incubation, I divided each P level soil subsample into three groups and wet them to the three aforementioned moisture levels. For each 103 moisture X P level, I further divided the soil into four incubation or AEF placement time intervals. There were three replicates for each P level X moisture X placement time treatment. I placed one filter paper disc per replicate in plastic vials (25.8 cm 3 volume). After removing AEFs at the appropriate time intervals, I cleaned off excess soil, extracted with 15 mL 1 M NaCI (shook for 1 h), centrifuged and filtered the extract through 0.2 /jm pore diam. Nucleopore filters. I measured P concentration in the extracts as described above. I performed statistical analyses using the PC SAS General Linear Model (GLM) for analysis of variance and treatment comparisons.

Correlation between AEF-P and Soil Solution P

Based on the results of the experiment described above, I tested AEF P extraction capability over a wider P range and related AEF-P to direct measures of soil solution P. I equilibrated composited surface horizon

Neguev series soil w etted to 6 8 % gravimetric moisture content at nine P levels (0, 250, 500, 750, 1000, 1250, 1500, 1750, 2000 mg P/kg soil) for three days. Following equilibration, I divided each P level sample into two groups, one for AEF-P extraction and the other for soil solution P determination. For the AEF-P extraction, I held constant both AEF placement (incubation) time and soil moisture content (six days and 67% gravimetric moisture content) and determined AEF-P in the manner described previously. I determined soil solution P using vacuum extraction by placing approximately 100 g of soil from each P level on a vacuum funnel 104 lined with Whatman filter paper #1 and wetting with deionized water until just saturated. I let the saturated soils equilibrate overnight covered with parafilm, extracted via vacuum suction the following day, filtered the extracted solution and analysed P concentration using the same procedure as for AEF-P. I developed a polynomial regression between AEF-P and soil solution P using PC SAS.

Field Placement of AE Paper Discs

To test AEF-P extractability under field conditions, I used AEF discs to monitor soil P fluxes as a function of tree pruning and cattle grazing in the above-described field experiment. I monitored all four treatments on two farms for a period of 33 days. I began sampling 7 d before tree pruning and repeated sampling 7 and 21 d after pruning. For those treatments with trees, I included distance from trees as an independent variable. I selected, at random, two groups of four trees and superimposed a grid pattern of 36 sampling points (Figure 2.6). Distances between tree rows included 25, 75, 150 and 300 cm while those within rows were 75 and 1 50 cm from each tree. For treatments without trees, I used an artificial grid pattern with a 150 cm sampling interval to select 16 randomized sampling points. At each sampling point, regardless of treatment, I buried three AEF discs to a depth of approximately 10 cm. During each of the three sampling periods, I left AEF discs in the field for six days. I collected corresponding gravimetric soil moisture samples on the same day that discs were removed from the field. I determined AEF-P in the <4------300 cm

«•—* f V » « » f . . • T i • 25 cm J

I *75

*150 cm

• 300 Transact 000 cm • 150 cm

• 75

• 25 cm 1r

T * • * t . . . f

Figure 2.6. Tree row and transect grid sampling plan for placing anion exchange filter paper discs. 106 manner described above and used the AEF-P - soil solution P regression equation to report changes in soil solution P as a function of time for the various treatments. I analyzed treatment and dependent variable effects using PC SAS GLM procedure for analysis of variance and post tested significant differences using least square means procedures (p < 0.050).

Development of the Anion Exchange Resin Impregated Resin Methodology

Although the anion exchange filter paper technique proved feasible for both laboratory and field use, it possessed three major limitations: 1 ) the filter paper's anion exchange capacity was not as great as most ion exchange resins, minimizing the filter paper's P extracting capabilities especially in a high P fixing soil; 2 ) soil could not be removed completely from the filter paper, creating a source of unquantifiable error in AEF-P estimates; and 3) filter papers did not stand up well to field conditions. In many instances, I could only recover torn pieces with soil particles adhered to them. As a result, I had to weigh all recovered filter papers and report AEF-P as mg P/g filter paper. This limitation proved to be extremely time-consuming. I, therefore, decided to investigate other anion exchange materials for in situ use. I screened several different materials including the anion exchange filter papers and, from these, selected the Ionics® anion exchange resin-impregnated membrane (AEM), Type 204-U-386. These are thin (0.57 mm thickness), anion-selective membranes comprised of a homogeneous cross-linked vinyl copolymer reinforcing fabric and embedded with quaternary ammonium anion exchange groups (Ionics, 1987). They have an exchange capacity of 2.80 107 molc/kg dry resin, a strong ability to exclude cations, and a high resistance to contamination by organic materials. They are also quite durable physically and chemically. I performed a series of laboratory experiments to calibrate AEM-P extractability including: NaCI extraction time, AEM recyclability, AEM selectivity for anions other than P, AEM sorption kinetics under conditions of high soil P retention and correlation with soil solution P. I then tested the method in the field under two kinds of experimental conditions: 1 ) a decomposition study in which the membranes were placed in the soil surface underneath residue-filled litter bags and collected successively over a 1 2 0 d period; and 2 ) a repeated measures study of soil P fluxes (as a function of tree pruning and grazing) in which permanently marked quadrats within field treatments were monitored repeatedly every 4 or 7 d for 80 d. The decomposition study represents a single placement and successive collection approach, whereas the P flux study represents a repetitive, short-duration placement (membranes never left longer than 7 d in the field).

Screening of Anion Exchange Materials

I evaluated several rigid, essentially two-dimensional anion exchange materials for their P extractability and correlation with vacuum extracted soil solution P: 1) Whatman DE81 anion exchange filter paper disks (2.5 cm diameter); 2) Silica-PVC membrane sheets, and 3) Ionics® anion-type (204- U-386) sheets. I cut sheets of the silica-PVC and Ionics membranes into squares with areas equal to that of the filter paper discs. The filter paper 108

discs were evaluated here in unsaturated (only soaked in deionized H 2 O) and NaCI saturated forms. I incubated a composite of surface (0-15 cm)

soil samples from the five experimental farm sites with P as NaH 2 P0 4 -H2 0 at rates of 0, 250, 500, 750 and 1000 mg/kg for 72 h. I then placed discs or squares of each anion exchange material in small vials (32.2 cm^J half filled with soil at a moisture content of 67% by weight and then covered by

an equal amount of soil. I removed the anion exchangers after 6 d, washed them free of soil with deionized water in a squirt bottle (I could not remove

all soil from the silica-PVC material), extracted by shaking with 15 mL 1 M NaCI for 1 h, and filtered the extract through Whatman #42 filter paper. During the exchange materials' incubation period, 1 water saturated samples of each P-equilibrated soil for 24 h and vacuum extracted soil solution as described above (100 g samples using 3 h extraction time). I analyzed both the extracts and soil solution for P as ascorbic acid-reduced phosphomolybdate (Olsen and Sommers, 1982) using a Milton Roy Spectronic 20D spectrophotometer (882 nm in a 1 cm cell).

Effect of NaCI Extraction Time on Ionics® AEM P Extractability

I soaked membranes for 2 h in three P concentrations (0, 2, and 4 mg/L P as NaH 2 P0 4 ‘H2 0 ). I then rinsed them with deionized water to remove excess P and placed them in 50 mL centrifuge tubes with 15 mL 1 M NaCI to shake at low speed on a reciprocating shaker for either 1, 2 or 4 h. When removed from the shaker after each period, I decanted the 109 extracting solution and re-extracted the membranes with another 15 mL of 1 M NaCI. I determined P in the extracts as described above.

Recyclability of the Ionics 0 AEM

My objectives were two-fold: 1) to determine whether AEMs could be saturated with P and extracted with NaCI repeatedly without loosing sorption ability, and 2) to assess AEM-P extractability after multiple use in the field. I determined AEM recyclability by soaking clean, NaCI-saturated membranes in a range of P solutions and successively extracting with NaCI and resoaking in P. I placed three membranes in each 50 mL P solution (0,

2, 4, 6 , 8 , and 10 mg P/Lj for 2 h, stirring occasionally. I then removed the membranes, rinsed them three times with deionized water and extracted them with 15 mL 1 M NaCI for 1 h. After extracting, I removed the membranes from the extract solutions, cleaned them with deionized water and placed them again in fresh P solutions for soaking. The soaking and extracting process was repeated for a total of three cycles. I analyzed AEM-P extracts as described above.

I undertook the second objective by placing 1 0 clean, unused membranes and 1 0 used (had been placed in the field repeatedly at least five times) in a 1 M NaCI saturating solution overnight. I then rinsed the membranes with deionized water and placed each in a beaker containing 1 0 mg P/L solution to soak overnight. The following day, I removed and rinsed the membranes and extracted AEM-P with 15 mL 1 M NaCI for 1 h. I analyzed extracts as above. 110 Ionics® AEM Anion Selectivity

I studied the relative selectivity of the Ionics 9 anion-type (204-U-386)

AEM for H 2 PO4 *, NO 3 * and SO 4 .2 - in non-soil, aqueous solutions to assess possible interference from other anions common in the soil solution with AEM P extraction. I selected a range of P solution concentrations representative of field conditions and combined each P level with increasing concentrations of either NC> 3 ' or SO 4 .2 -. l chose N 0 3 " and SO 4 .2 - ranges to reflect "worst case scenarios." I placed clean AEM squares (6.25 cm^) in 50 mL polypropylene centrifuge tubes containing P solutions at concentrations of 0.25, 0.50 and 1.50 mg P/L with either NO 3 at 0, 5, 50 or 100 mg N/L or SO 4 at 0, 50, 500 or 1000 mg S/L; all solutions were prepared as Na salts using a total volume of 40.00 mL. All solutions were shaken on a reciprocal shaker at low speed overnight. I then removed the membranes, rinsed them in distilled water, extracted them in 15 mL 1 M NaCI and analyzed extracts for P as described previously. Results are presented as percent anion interference of AEM-P sorption {[Px XPxNyl / Px X 100 where Px is P in pure solution at level x and PxNy is P at level x in combination with N at level y).

AEM Kinetics and Correlation with Soil Solution P

I equilibrated composited soil from the field study with P for 72 h at rates of 0, 250, 500, 1000 and 1500 mg/kg as NaH 2 P0 4 .H2 0 at a gravimetric soil moisture content of 55%. I divided each P level soil into 111 aliquots which were then wetted to existing (55%), 65 and 75% gravimetric moisture and left to incubuate for an additional 96 h in parafilm-covered beakers, mixing at least once daily to prevent anaerobiosis. Despite this effort, the two higher moisture levels became anaerobic, and I eliminated soil moisture as a factor in the experimental design, but included the high P (1500 mg/kg)-high moisture (75%) treatment for comparison. I presaturated

squares (6.25 cm2) of Ionics 9 anion-type (204-U-386) AEM in 1 M NaCI 24 h prior to use, rinsed with deionized water to remove excess salt and then placed them in vials containing approximately 23 g (oven dry wt. equivalent)

of the P-treated soils for periods of 2, 4, 6 and 8 d. Upon removal from the vials, I cleaned and NaCI-extracted the membranes as above and measured soil pH in 1:1 (soil: water) subsamples taken from each vial. I determined soil solution P by wetting aliquots of each treated soil to 75% moisture by weight and extracting the soil solution by suction after 48 h equilibration. This was done twice: immediately following the initial 72 h P incubation (while the membranes were incubating), and approximately two weeks later. I analyzed P in extracts and soil solution as described previously. I determined P addition effects on AEM-P extractability and soil pH using SYSTAT’s MGLH analysis of variance and correlated AEM-P with soil solution P by polynomial regression (Wilkinson, 1991). Soil solution measured after the initial 72 h incubation was regressed with AEM-P extracted at days 2 and 4, while the second soil solution determination was

regressed with AEM-P extracted at days 6 and 8 . 1 1 2 in Situ Application of the AEM Method

Basic procedure for fieid use The basic pre-field preparation process involved cutting the membranes, presaturating with Cl' using 1 M NaCI, attaching som e kind of durable thread for easy recovery and cleaning with deionized water before taking to the field. Ionics, Inc. (65 Grove St., Watertown, MA 02172) supplies the anion exchange membrane in 30 cm x 30 cm sheets packed in ethylene glycol to prevent dessication. I cut sheets into 2.5 cm x 2.5 cm squares. Once membranes were cut into squares, I presaturated membranes in 1M NaCI at least 24 h before use. I chose Cl' as the counterion because it was not expected to be a confounding anion in the soil solution system. I sewed dental floss (one could use nylon fishing line or any durable thread) to the membrane and attached a brightly colored flag (red or orange works best in pasture grass) at the other end so it could be found in the field easily. I stored m em branes in NaCI in the refrigerator until use to prevent bacterial and fungal growth. When ready to take to the field, I washed membranes with deionized water and transported them in water-containing wide-mouth bottles (Ionics, Inc. recommends storage and transport in an aqueous medium to prevent cracking and disfiguration upon drying; i.e., they can loose both physical and chemical properties if they crack or split). I placed membranes in the surface soil by opening a vertical slit in soil with a hand trowel and gently sliding the membrane vertically into the slit. I then closed the slit by pressing the soil firmly together (Figure 2.7). I collected membranes from the soil by tugging gently on the dental floss; in most cases, the membrane 113 slid out easily- I then removed large aggregates adhered to the membrane and carried the membranes back to the lab in water-filled plastic bottles. In the lab, I used a squirt bottle with distilled water and my fingertips to massage off any remaining soil particles. I placed individual membranes (or replicates) in beakers with distilled water until they were extracted and extracted AEM-P in 50-mL centrifuge tubes with 15 mL 1M NaCI by shaking on a reciprocating shaker for 1 h. I then removed membranes and analyzed P in the NaCI extract as ascorbic acid-reduced phosphomolybdate. I recycled used membranes by rinsing twice with distilled water and resoaking in 1M NaCI until they were reused. 1 refrigerated membranes for long-term storage.

Field Decomposition Study Details of this experiment are presented elsewhere in the dissertation. Basically, I placed three types of residue from the field experiment: cattle dung, leaves of the tropical leguminous tree Erythrina berteroana and native pasture, in plastic mesh litter bags and laid them on the soil surface (in the pasture) in one of the field plots. Concurrently, I placed three Ionics® AEM squares (6.25 cm^l vertically beneath each residue bag 0-2.5 cm below the soil surface. I later collected litter bags and their corresponding membranes from the field after 1, 2, 4, 7, 14, 21, 28, 35, 42, 49, 56, 63, 70, 77 and P extracted from AEM in 15 ml 1M Ionics AEM NaCI chloride saturated (~ 6 cm2 area) AEM removed from ■ extract solution

Analyze P colorimetrically

AEM placed in slit in soil surface

Figure 2.7. Schematic diagram of the procedure for anion exchange membrane preparation and in situ use in the field. 115 120 d. I analyzed residues for total P in acid digests and calculated P remaining as percent of initial total P. I determined AEM-P as described previously.

‘y

Labile Soil P Fluxes as a Function of Tree Pruning and Grazing In general, I monitored labile soil P with AEMs before and after one tree pruning and three cattle grazing cycles. I took measurements on the block of four treatments on two of the five study farms. I randomly selected and permanently marked 1 m2 quadrats within each treatment plot; five quadrats in the grazed and three in the non-grazed treatments. Quadrats within the two tree treatments were located close to tree trunks to maximize tree effects. I began measurements 12 d before pruning by placing 10 AEMs in the surface (0-2.5 cm) soil of each quadrat. I removed them 4 d later and put in a new group for another 4 d (5 d before pruning). I continued this 4 d sampling interval for 4, 8, 13, 17 and 22 d after pruning. I made subsequent measurements every 7 d until 70 d after pruning. At each sampling date, then, I removed AEMs and placed a new group in the same quadrat such that the quadrat was considered the repeated measures unit. I processed membranes and determined AEM-P as described above. I also monitored, concurrently, soil moisture suction at each quadrat using ceramic cup tensiometers. When I plotted AEM-P as a function of time, the data were not normally distributed. In addition, I wanted to relate observed variation to either treatment effects and/or soil moisture fluxes. I began with the assumption that there may have been block differences present prior to 116 sifvopastoral system perterbation (in this case pruning followed by grazing); i.e., the treatments, because of differences in management practices, possessed different system equilibria at time zero. Given this condition, I applied a data-normalizing index used in insecticide effica&y trials known as the Henderson and Tilton Percent Mortality Index (Taylor, 1987). This index normalizes treatments relative to the experimental control (in this case, the non-grazed, no trees treatment) and to silvopastoral system AEM-P status at the time prior to system perterbation; i.e. pruning followed by grazing (Figure 2.8). I then log transformed the index-normalized AEM-P data and performed repeated measures analysis of variance using SYSTAT (Wilkinson, 1991). Ultimately, I plotted treatment AEM-P fluxes as a function of time and compared AEM-P with changes in soil moisture status.

RESULTS

Anion Exchange Filter Paper Methodology

Effect of Agitation Time and Number of Extractions on AEF-P Extractability

The majority of AEF-P was extracted in the first 1 M NaCI extraction regardless of P level and shaking duration (Table 2.3). Less than 5% was detected in the second extraction and this was probably P in the entrained solution from the first extraction. Agitation time had no significant effect on NaCI extraction of P from the AEF at the lower P level, but there was a slight increase from 2 to 4 h at the higher P level (Table 2.4). It is reasonable to assume that the AEF will rarely be exposed to soil solution 117

(aemp £71 )(aemp rctij AEMPINDEX = 100 {aempcu){aemp

awnpet1 - control treatment membrane P at time 1 aempratx * treatment a membrane P at time x aem p^ * control treatment membrane P at time x aeaemp,at1 « treatment a membrane P at time 1

Figure 2.8. Equation describing the statistical normalizing index used to analyze anion exchange membrane(AEM) P fluxes as a function of tree pruning and grazing. The index normalizes AEM-P with respect to both the experimental control treatment (no trees, no grazing) and AEM-P levels prior to tree pruning and pasture grazing. Table 2.3 Effect of agitation time and number of 1 M NaCI extractions on AEF P extractability.

Agitation Time (hi ...... i ...... __7...... __ A______P (mg/LI 1st 2nd 2nd 1st 2nd 2nd 1st 2nd 2nd soaking extract extract (%sum) extract extract (%sum) extract extract (%sum) solution ...... 'W rp —__ — 5 2.085 0.104 4.75 2.000 0.125 5.88 1.940 0.104 5.09

10 4.275 0.188 4.20 4.335 0.209 4.60 4.695 0.188 3.84

% sum is equivalent to: (//g R 2nd extr///g P 1st + 2nd) * 100. No significant differences between extraction times within a given P soaking solution level; significant differences between two P levels (p<0.0001).

Table 2.4. Analysis of variance table of treatment effects.

Source df F-Statistic Pr > F P concentration in 1 st extraction P level (P) 1 1138.33 0.0001 Time (T) 2 1.71 0.2580 PX T 2 5.51 00440 Error 6 P concentration in 2nd extraction P level 1 7.20 0.036 Time 2 0.20 0.824 P XT 2 0.00 1.000 Error 6 119 concentrations as high as 10 mg/L, particularly for the high P fixing Neguev soil. Therefore, a single extraction with 1 M NaCI shaken for 1 h is adequate to extract AEF-P.

Estimation of Soil Solution P with Anion Exchange Filter Paper

The most significant effect on AEF-P extractability was soil moisture content (Table 2.5 ). In addition, the highly significant moisture content- incubation time interaction verified that incubation duration is only important when the soil is close to saturated moisture content. The highest AEF-P values were obtained at the highest moisture level. There was a large difference between AEF-P at 67% and 31% moisture contents, but no significant difference between 31% and 27%. These results confirm the importance of diffusion in AEF-P extractability; when moisture is limiting, P diffusion to the filter paper is impeded spatially and only soil solution in direct contact with the AEF contributes to AEF P sorption. In addition, the amount (mass) of P in solution is moisture dependent such that dissolution of P compounds will be enhanced with increasing moisture content. Under field conditions of my experiment, ambient gravimetric soil moisture ranges between 50-68%. Such conditions ensure that field soil moisture levels will almost always be within the range at which P diffusion to the filter papers will not be spatially hindered and dissolution of soluble P compounds will be favored. When moisture is not limiting, the distance over which P can diffuse as well as the total P amount to the filter paper increases; therefore Table 2.5. Effect of P addition, incubation time and soil moisture content on anion exchanQe filter paper P extraction as mg P/kg filter paper (top); and analysis of variance table of main effects and 2-way interactions (bottom).

P Addition Incubation Time (Days) (mg P/kg soil) 2 4 6 8 Soil Moisture - 67% bv wt.

0 0.717 0.713 1.440 1.140 315 0.703 0.941 1.610 1.380 630 0.654 0.982 2.830 1.100 945 0.760 1.100 1.140 1.120 1260 0.725 0.686 0.677 0.683 Soil Moisture * 31 % bv wi,

0 0.692 0.716 0.673 0.700 315 0.664 706.000 0.720 0.735 630 0.668 0.632 0.696 0.776 945 0.640 0.633 0.664 0.681 1260 0.639 0.655 0.664 0.710 Soil Moisture = 27% bv wt.

0 0.692 0.925 0.812 0.672 315 0.653 0.661 0.648 0.734 630 0.685 0.677 0.754 0.716 945 0.652 0.676 0.652 0.695 1260 0.678 0.713 0.660 0.658

Model df F Value Pr > F P Level 4 2.15 0.0772 Moisture 2 24.95 0.0001 Time 3 5.27 0.0018 P Level * Moisture 8 1.83 0.0758 P Level * Time 12 1.10 0.3660 Moisture * Time 6 4.28 0.0005 121 incubation time becomes an important determinant of AEF P extractability. At the 67% moisture level, AEF P extraction increased significantly from days 2 to 4 to 6 but leveled off between days 6 and 8 . This suggests that

AEF P extractability reached equilibrium within 6 - 8 d. The incubation time had no significant effect on AEF P extractability at the highest P addition, however. There may be a sufficiently high soil solution P concentration at this level to override rate-limiting diffusion effects. The relationship between AEF-P and soil solution P was curvilinear and the derived regression equation explained 98% of the variation (Figure 2.9). The filter paper's P extractability improves with higher soil solution equilibrium P concentrations. This concentration-dependent efficiency has been confirmed numerous times in resin P extraction studies (Amer et al., 1955; Barraclough and Tinker, 1981; Vaidyanathan and Nye, 1966). The correlation, however, was weak at low solution P concentrations; i.e., the AEF was not sensitive to low-level changes in soil solution P. This suggests that the AEF is a weaker competitor than the soil matrix for soluble P. Under field conditions, then, P additions to the soil solution (P mineralized from decomposing residues or roots) would have to be relatively large to be within the AEF's range of detection. In spite of this potential drawback, the filter papers should be able to detect and magnify micro-site changes as well as non-equilibrium soil solution fluxes.

Field Test of AEF-P Method In general, field use affirmed that, despite aforementioned limitations, the method detected significant treatment effects (at the 0.05 level) despite Resin P (mg/kg resin) 30 0 1 0 iue29 Correlationanionfilter between exchange Figure(AEF)-extractedpaper 2.9. P (referredresiny-axis)P theP on as to and equilibrium soil solution P. e i P 50 + .9 on - 03 SolnP 0.35 - SolnP 5.39 + 5.02 = PResin E q u i l i b r i u m P c o n c . ( m g / l i t e r ) 2345678 0.98 I 123 the targe inherent system variability (Figures 2.10-2.11). For example, AEF- P was significantly higher close to the trees than 150-300 cm away from the trees in the grazed treatment (Figure 2.10). The difference in distance effect between grazed and non-grazed treatments probably reflects management differences; in the grazed treatment, cattle altered the distribution of decomposing branches whereas in the non-grazed treatment branch distribution remained relatively uniform. With distance means pooled, there was a significant treatment by time interaction for all treatments (Figure 2.11). The emergent pattern suggested that the two non-grazed treatments exhibited increases in soil solution P (as estimated by AEF-P) as early as one week after pruning and grazing,whereas the two grazed treatments did not respond until three weeks after pruning. The agro-ecological significance of these results will be discussed elsewhere; suffice it to say that the AEF-P methodology was sensitive enough to detect biologically meaningful changes in labile P.

Anion Exchange Membrane Methodology

Screening of Anion Exchange Materials

All materials adsorbed P as a function of increasing soil solution P, in the order: Silica-PVC > > Ionics® AEM > Cl-saturated filter paper > unsaturated filter paper (Figure 2.12). The high apparent P sorption by the Silica-PVC material is misleading in that it was due almost entirely to binding of soil particles to the membrane. It was impossible to remove soil from this Resin P (mg/kg resin paper) k > 2 3- 4 - 0 5 6 1 - ■ | 5 10 5 20 5 300 250 200 150 100 50 0 ------i i - i - — -Trees ofdistance from selected trees;twotree thetreatments, withwithoutand AnionFigureexchange filter 2.10. paper (resin) extractable functionaP as grazing(cattle) are compared. Alone i I ■i Distance ------rmte (cm) tree from , ■ J j . I i , , 1 T r e e s A l o n e re ad Cattle and Trees re ad Cattle and Trees ______iI __ 350 Soil Solution P (m g/L ) 2.5 2.5 2.0 0.5 0.5 1.5 1.5 0.0 1.0 —▲ ▲— Figure 2.11. Soil solution P estimated from anion exchange filter paper paper filter exchange anion from estimated P solution Soil 2.11. Figure AF- ttresmln ae o h fu xeietltet ents. treatm experimental four the for dates sampling three at (AEF)-P 0-6 re alone Trees ate n pasture and Cattle re ad cattle and Trees Pasture ie (days) Time 20 0 -2 4 1 33 -3 7 2 ouin . aeil wr icbtd n ol o i dy. 1 SD. 1 ± soil of days. six for soil function in a as incubated materials were resin Materials various by P. sorbed P solution 2.12. Figure

Resln-P (mq P/cm2 resin) 0.0 0.2 0.4- 0 0.8 . 6 0.0 - “■— Fitter Fitter “■— A '" Fitter Fitter '" A 0.5 ol ouin (mg/L) P Solution Sotl Paper* Paper* Paper* Paper* 1.0 nauae ^ lonlca ^ unaaturated Cl Cl at at d te ra tu sa 1.5 2.0 SMca-PVC — # — 2.5 3.0 127 material. On the other hand, the material with the second best sorption characteristics, the Ionics'* AEM, was very easy to clean and stood up the best to handling in the laboratory and field. P sorption by the paper filters was much Jess than that of the Ionics® AEM. In addition, the paper filters tended to tear when wet and were difficult to separate from the soil interface. Based on these results, the Ionics® AEM was selected for further study.

Effect of Agitation Time and Number of NaCI Extractions on AEM P Extractability

All of the AEM-sorbed P was extracted after shaking for 1 h and with a single 15 mL, 1 M NaCI aliquot {Table 2.6). There were no significant differences at either P soaking solution level among the three extraction durations. In addition, there was no P detected in the second extraction with NaCI. These results confirm the expected: 1 M Cl" is sufficient to replace any anions adsorbed to the membrane, thereby serving as an effective extracting solution.

Recyclability of the Ionics 9 AEM

It appears that the Ionics® AEM can be reused one or more times without a change in P sorption (Figure 2.13}. There were no significant differences in AEM-P for the 2, 6 and 8 mg P/L solutions over the three extraction times. There was a slight increase in AEM-P between the first and third use at the 4 mg P/L solution, and a marked increase after one use 1 2 8

Table 2.6. Effect of agitation time and number of extractions on AEM P extracts bility.

Agitation Time

2------4------

P (mg/Li 1st 2nd 1st 2nd 1st 2nd

soaking solution extract extract extract extract extract extract

------r « D ____

0 0.0 0.0 0.0 0.0 0.0 0.0

(0.0) (0.0) (0.0) (0.0) (0.0) (0.0)

2 0.71 0.0 0.71 0.0 0.72 0.0

(0.06) (0.0) (0.04) (0.0) (0.04) (0.0)

4 1.36 0.0 1.40 0.01 1.40 0.01

(0.13) (0.0) (0.04) (0.01) (0.06) (0.01) n = 6 for 1st extractration; n = 3 for 2nd extraction. Numbers in ( ) are standard deviations. anion exchange membrane P sorption (referred to as resin P) over a range of of range a over P) resin as to (referred sorption P membrane exchange anion Figure 2.13. The effect of repeated P soaking solution-NaCl extraction on on extraction concentrations. solution-NaCl sofution soaking P soaking P repeated of effect The 2.13. Figure Resin P (mg P/L in extract) ouin Cnetain m P/L) (mg Concentration P Solution 0

Cycle 3 Cycle2 Cycle 1 2

4

6

8

10 129 130 with the 10 mg P/L solution. Most likely, P was not completely desorbed by the NaCI after the first extraction and was carried over to the subsequent two extractions. The highest P solution represents a somewhat unrealistic condition in that such high concentrations are rare in the field. In addition, placing the membranes in aqueous P solution rather than in soil increases the amount of P that would be sorbed under normal field use. The comparison between new and used membranes showed no significant difference in their ability to extract P from a 10 mg P/L solution (AEM-P for new membranes was 0.412 mg P/L extract and 0.421 mg P/L for used membranes); further evidence for their ability to withstand repeated use without altering sorption capability. Field experience also suggested that the anion exchange membranes can be used for a minimum of three times.

Anion Selectivity on the Ionics 0 AEM

The Ionics® AEM had a high selectivity for P versus NC> 3 ' or S 0 4 2 - when these ions were present in solution at relatively low concentrations

(Figure 2.14). Regardless of the initial solution P concentration, NO 3 ' at 5 mg N/L did not reduce AEM P sorption significantly, yet reduced it by 50% and 75% at 50 and 100 mg N/L respectively. Sulfate at 50 mg S/L also decreased AEM P sorption by 50%, while the two highest SO 4 .2 - solution concentrations reduced AEM P sorption by approximately 95%. While the lowest N 0 3 " or SO 4 .2 - levels are representative of seasonal field fluxes for this agro-ecosystem, extremely high levels are rarely expected to be observed. Nonetheless, NO 3 ' soil solution concentrations could reach 50 ocnrtos 02, ,0 15 g P/L}. mg 1.5 0,50, (0.25, concentrations AEM P sorption (expressed as % interference} from three P solution solution P three from interference} % as (expressed sorption P AEM Figure 2.14. The effect of increasing N increasing of effect The 2.14. Figure

Percent Interference 1.0 Nltrale-N: 5,50,100 mg/L ■ Sulfate-S: 50,500,1000 mg/L mg/L 50,500,1000 Sulfate-S: ■ mg/L 5,50,100 Nltrale-N: 0 .2 5 0 .2 5 0 .2 5 0.5 0.5 0.5 1.5 1.5 1.5 1.5 1.5 1.5 0.5 0.5 0.5 5 .2 0 5 .2 0 5 .2 0 nta Slto P oe (mg/L) Cone. P Solution Initial 3 0 n SO and ' 2 4 - concentrations on on concentrations - mg N/ L as N is mineralized from decomposing legume residues {Vilas, 1990). The results confirm that Ionics® AEM anion selectivity, as for all exchange materials, is governed by ion size (and hydration) and valence. Within the range representative of soil solution concentrations for the tested anions, then, the AEM is very selective for P. Ionics, Inc. (1987) reported low limiting equivalent ionic conductance for H 2 P0 4 ‘ relative to NO 3 ' and

HS0 4 ~ in aqueous solution (Table 2.7). This is a measure of the anions' conductance through the membrane's aqueous phase (when hydrated, membranes are approximately 46% water; R. MacDonald, Ionics, Inc., pers. comm., 1992). Anions with low conductance are less likely to pass through the membrane, and, therefore, more likely to be adsorbed on surface exchange sites than smaller, more mobile anions. Ideally, if AEM-P is to be correlated directly with soil solution P, concentrations of competing anions should be measured as well. This kind of anion selectivity experiment should be repeated in a soil medium, with the ultimate objective of developing a predictive model for each anion in relation to its competitors. 133

Table 2.7. Limiting equivalent ionic conductances for selected inorganic anions in aqueous solutions at 25°C.

Limiting Equivalent Ionic Ion Conductance (mho-cm^/equiv.) §

ci- 76.4

H2 PO4 - 33.0

HSO4 - 50.0

NO3 - 71.4 OH- 198.6 S Values were obtained in the presence of the anion exchange membrane; the lower the conductance, the stronger the adsorption to the membrane (from Ionics, 1987). 134 AEM Kinetics and Correlation with Soil Solution P

Soil solution P increased with added P but was almost an order of magnitude higher five days after the initial P addition than after 19 days equilibration (Figure 2.15). This indicates that sorption or precipitation of the added P was continuing during the study; i.e., the soil solution at higher P additions had not reached steady state equilibrium after five days of incubation. AEM-P also decreased significantly with time at the higher P additions (p< 0.0001} (Figure 2.16). This dem onstrates the ability of the membrane to reflect rapid changes in soil solution P and confirms that this material is a dynamic exchanger rather than a P sink under these soil conditions. AEM-P was correlated significantly with soil pH (p< 0.0001) but the relationship was not linear (r^ = 0.39). Soil pH, in turn, was highly positively correlated with P addition (r2 = 0.90) but not with incubation time (Figure 2.17). These results suggest that the H 2 P0 4 - ions displaced OH- ions almost immediately and that the increase in pH probably decreased soil (Al-oxide) P sorption to a certain extent (Violante et al., 1991). There w as a strong correlation between AEM-P and soil solution P when day 5 (equilibration time) soil solution P was regressed against day 2 (AEM incubation time) membrane extraction (Figure 2.18). Sorption was nearly linear up to 2 mg P/L in soil solution. When day 19 (equilibration time) soil solution P was regressed against day 8 (AEM incubation period) AEM-P, there was a similar relationship despite much lower soil solution P concentrations (Figure 2.19). 5 Day Incubation 135 19 Day Incubation _i Q. D> E CL oC 3 O CO

o O r o 1000 2000 P Added (mg P/kg soil)

Figure 2.15. Change in soil solution P as a function of added P after 5 and 19 days equilibration time. ± 1 SD.

Time of Resin in Soil 2 d a y s 4 d a y s co & 6 d a y s CM * — 8 d a y s E o CL

O. £ *5 0 > DC

1000 2000 P Added (mg P/kg soil)

Figure 2.16. Changes in AEM-P (expressed as resin P) as a function of incubation time in soil over the range of P added. ± 1 SD. - m - o ~ m - aao — »oo

— 1000 -D- 1SOOA ~ m ~ 1 S 0 0 B 136 e.20

5 .SO

6 . 4 0

*

3 5 . 0 0

4 . 0 0

o 2 4 6 a Number of Inoubotton day*

5 .4

xo . o CO

4 .5

0 6 0 0 1200 P Added (mg P/kg ooil) Figure 2.17. The relationship between P addition and soil pH over the range of AEM incubation periods (top). Changes in soil pH over time for the various P levels added (bottom). s <0.015 + 1.14X - 0.094X2 R2 s 0.998

c «o £ CM E u o.

a . i co a DC

2 4 6 8 Soil solution P (mg/L) Figure 2.18. The correlation between day 2 incubation AEM-P (expressed as resin-P) and day 5 (equilibration period) soil solution P. ± 1 SD.

1.0 -0.006 4- 2.45X - 1.69x2 R2 ■ 0.903

m 0 -® “ 0 CM | 0.6 - CL o> a. 0.4 H CL< C <0 v 02- r r M *

0.0 0.0 0 .4 0.6 0.8 Soil Solution P (mg P/L)

Figure 2.19. The correlation between day 8 incubation AEM-P (expressed as resin-P) and day 19 equilibration soil solution P. ± 1 SD. Note that soil solution P at 19 d equilibration is approximately an order of magnitude less than soil solution P measured after 5 d equilibration. 138 There were no significant changes in AEM-P in the zero P addition

control soil over the incubation period of 2 to 8 d (Figure 2.20). Again, this confirms that AEM-P mirrors soil solution P dynamics rather than accumulating P over time. It also demonstrates the membrane's ability to detect low labile P levels in this high P-sorbing soil.

Application of the AEM Method to P Release in the Field from Residue Decomposition

Residue P release characteristics fit first order negative exponential functions for the most part (Figure 2.21). Rates of P release appeared to differ among the residue types, however. AEM-P increased in time lag fashion which appeared to track P released from the decomposing residues. The membranes detected two major P pulses over the study period; the magnitude and timing of the dung P pulses differed significantly from those

of the Erythrina leaves and pasture grass. Most importantly, the membranes were able to detect biologically meaningful changes in labile soil P under field conditions.

Application of AEM Methodology to Soil P Fluxes as Related to Tree Pruning and Cattle Grazing

Log-transformed, index-normalized AEM-P exhibited an underlying periodicity over time among the four field treatments (Figure 2.22). Despite large within-treatment variation, the membranes were able to detect P fluxes which could be explained in terms of both micro-climatological (soil moisture) and treatment-induced (tree pruning and grazing) changes. When added) after 2, 4. 2, after added) iue22. AEM-Presin-P)(expressed as fromcontrolFigure soil the 2.20. <0 P/kgmg

Resin-P (tig P/cm2 resin) 04)00 04)10 0.014 0 6 and and i o Rei i Si (days) Soil in esin R of s Tim oto si: m Pk soil P/kg mg 0 soil: Control 2 8 dicbto ntesi. 1 SD. ± incubationd insoil. the 4 1

0 9 3 1 ■ Oun«P 140 — *w# 1 H SO r I .. « M ft. I

as M t s IN tu D «ya H) Um FMM

4 — e M s •> M m I

M ri Dnyn In tftn FtoW

W i ••

as

o as M IN 1M Deye WUieFleM

Figure 2.21. The relationship between percent P remaining in dung (top). Erythrina berteroana leaves (middle) and pasture grass residues (bottom) over the duration of a field decomposition study and concurrent soil AEM-P under each respective residue type. Note that AEM-P under dung residues is about four times greater than both Erythrina leaves and pasture grass residue AEM-P. 141

Trees Grazed — •*" Trees & grazing

< 3

i C 1 >

1.0 0 20 6040 80 Days

Figure 2.22. Fluctuations in AEM-P (index-normalized and log-transformed) before and after tree pruning and cattle grazing for the three non-control field treatments. The y-axis is in log scale. Any AEM-P increases reflect P fluxes greater than the control treatment (non-grazed, no trees) and the time prior to tree pruning. 142 AEM-P data were plotted with soil moisture data as a function of time, a time lag pattern emerged (Figure 2.23). There was a suggested inverse

linear relationship between the log of lag ~1 soil moisture suction and log AEM-P (Figure 2.24). For example, if soil moisture was high at sampling date 1 (shown as low soil moisture suction), AEM-P at sampling date 2 would increase relative to the previous sampling date. It is not clear, however, how long the actual lag period was; it was probably less than the sampling interval of 4 or 7 d. As in the decomposition study, the AEM in situ method produced ecologically significant results and was sensitive enough to detect small-scale temporal and spatial changes.

DISCUSSION

Dynamic Exchanger vs. Infinite Sink

The concept of anion exchange resin as competitor with the soil matrix for labile P is well established (van der Zee et al., 1987; Curtin et al., 1987). Since the mechanism of resin P sorption is ion exchange, the widely accepted analogy between plant roots and resin materials is somewhat dubious. Indeed, it seems simplistic to assume that anion exchange resin materials (including resin beads, filter paper and resin-impregnated membranes) would behave solely as infinite sinks for P. Implicit in the notion of competition is a two-way interaction between the soil matrix and the resin material. Under differing soil mineralogical and chemical conditions, the soil will either compete strongly or weakly with the resin AEM-P Soil moisture 2.2 3 0

2.0 o x 3 o (0 ■o -20 k c a A Z 1*8* o, s ui £ < 1.6- 3 O) -10 « o o —J E 1.4* o CO

1.2 40 50 60 70 80 90 Day

Figure 2.23. The relationship between AEM-P (index-nomalized, log- transformed) and soil moisture suction. High soil moisture suction values indicate relatively dry conditions while low values indicate wet to near soil moisture saturation. 144

■ trees * grazed • trees ft grazing 1.6-

T 1.4- l " - % ot a e § . 1.2- * 4 .* £ 2 1.0- . a “ *■ • ; # n “ • A • § 0.8‘ A " • B O0.6 o» 3 0.4-

0.2-£ —1—' TI * ' 1 I * '---- *1 ■ 1 1 * 1l'1"^ ’ i I J ■ 1.2 1.4 1.6 1.8 2.0 2.2 2.4 2.6 Log Index AEM-P

Figure 2.24. Log index AEM-P as related to the log of lag-1 soil moisture suction. There is a suggested inverse linear relationship between the two, albeit not very strong. 145 material for labile P, and the kinetics of P sorption on both surfaces will be altered (Aharoni et at., 1991}. Moreover, the interaction is expected to be dynamic as it reflects microsite fluctuations in soil pH and organic acid activities (Violante et al., 1991; Kafkafi et al., 1988; Lopez-Hernandez et al., 1986; Traina et al., 1986; Fox and Comerford, 1992}. It is not surprising, then, that the Neguev series, an acid soil dominated by Al and Fe oxides and poorly crystallized kaolinite, would compete strongly with both the anion exchange filter paper and resin- impregnated membrane for P, and actually desorb P from the membrane under non-equilibrium, high soil solution P conditions. Krause and Ramlal (1987) also found that anion flow to the resin could be reversed either by changes in soil pH or by plant root and microbial uptake. Although the bulk of the resin literature supports the concept of resin as infinite sink (Sibbesen, 1978; Skogley,1991; Yang et al., 1991a & b; Gibson et al., 1985}, most of these studies were conducted on neutral to high pH, calcareous soils or under near-saturated soil moisture conditions which favor unidirectional flow of dissolved P to the resin. Yang et al. (1991a & b) used mixed bed (cation and anion) with H+ as the counter-cation; this acidified the soil microsite, enhancing P solubility in their Ca-dominated soils. The dynamic exchange nature of resin membrane P sorption (AEM-P} evokes a more complex interpretation of P fluxes over time. Since sorption and desorption at the membrane surface probably occur relatively rapidly, it is difficult to assess quantitatively the labile P flux pattern which results during the membrane incubation period (the time during which the membrane is in contact with the soil}. At best, AEM-P represents an 146 integrated value of all P fluxes (both to and from the membrane) over the sampling interval. Perhaps a more precise, quantitative description of this integral can be achieved by measuring AEM-P in minimally-disturbed soil cores over very short time intervals.

Competition From Other Anions

As stated previously, the Ionics® anion exchange membrane has a

strong affinity for H 2 PO4 ' relative to other anions (Ionics, Inc., 1987). However, the mechanism for P sorption is ion exchange; and, consequently, other anions can interfere with AEM P sorption, particularly when their soil solution activites are high. Vaidyanathan and Nye (1966), comparing P flux

to anion exchange resin paper using Cl', NO 3 ', HCO3 " and SO4 .2 - as resin saturating ions, found P diffusion to the resin paper was not affected by the monovalent anions but was slowed slightly by the divalent anion (S O ^ -). They suggest that P diffusion to the resin material would be impeded by other anions only if the other anions affect the equilibrium between soil solution and soil surface P. In contrast, Skogley (1991) and Yang et al. (1991b) assert that anions and cations are adsorbed to resin capsules independently and selectively based on their soil solution activities and rates of diffusion to the resin sink. While this may be valid for their study conditions, it does not address the possibility of anions either outcompeting or displacing resin-adsorbed P if the other anions* solution concentrations overwhelm soil solution P concentration. In addition, it ignores possible synergistic interactions among counter-ions both negatively and positively charged; i.e. ions which alter soil pH or P solubility. 147

Relationship Between Resin Material-P and Soil Solution P

The function which best described the relationship between soil solution P and both AEF- and AEM-P was L-shaped (Langmuir equation), regardless of whether the systems were at equilibrium. The relationship between AEF-P and soil solution P at low solution P concentrations was weak, however; suggesting low AEF sensitivity to small-scale changes in soil solution P status. AEM-P, in contrast, exhibited a fairly strong linear relationship at the lower end of the soil solution P spectrum. These results confirmed the membrane's significant affinity for P at low solution P levels and, in turn, its ability to track micro-scale P fluxes. The AEM method, when used to measure labile soil P fluxes, is, in fact, an improvement over direct measurements of soil solution P because: 1) soil solution methods are difficult to perform and are estimates at best (Sposito, 1989); and 2) soil solution measurements do not account for the soil's P replacement ability (Abrams and Jarrell, 1991; Dalai and Hallsworth, 1976; Gibson et al., 1985). Additional research has taken the relationship between AEM and soil solution one step further using bioassays (Abrams and Jarrell, 1992). They found a significant exponential correlation between AEM-P (in terms of their bioavailability index which includes a diffusion coefficient) and plant P uptake, and suggested that the AEM-P vs. plant uptake curve mirrors the exponential function which describes the saturating nature of plant response to available P. Other studies using anion exchange resins have obtained strong correlations between resin P and plant P uptake (Palma and 148 Fassbender, 1970; Dalai and Hallsworth, 1976; Yang et al., 1991b). Dalai and Hallsworth (1976) inferred from the highly significant correlation between resin P and young plant P uptake that the resin P reflected the importance of both intensity and kinetics of labile soil P in early plant development.

Extrapolation of AEM Method to Field Use

The impetus for laboratory AEM calibration was, as stated, in situ field use. Despite major limitations associated with soil disturbance and manipulation in the lab, the membrane incubation study and correlation with soil solution P elucidated important mechanisms of membrane function over both time and a wide range of soluble P concentrations. In general terms, we know that AEM P sorption is rapid, and under low ambient soil solution concentrations, incubation time has little effect on AEM P sorption. When large P pulses are introduced and the system's equilibrium (soil matrix ++ soil solution) is disrupted, AEM-P reflects those equilibrium shifts. Most importantly, we know that P flow to the membrane is not unidirectional; AEM-P can be adsorbed and desorbed, depending on its competitive ability relative to the soil matrix. The effects of soil moisture on AEM P sorption have yet to be evaluated properly in a semi-controlled, minimally disturbed context (perhaps using membranes placed in undisturbed soil cores equilibrated to differing soil moisture contents). Indeed, unsuccessful lab experiments as well as field use suggest a strong link between soil moisture and AEM-P fluxes. 149

The AEM in situ use in two distinct field settings corroborates laboratory findings; i.e., AEM-P is extremely dynamic over relatively short sampling periods. Another revelation from the field studies is that AEM sensitivity enhances (perhaps magnifies) both spatial and temporal variation. This is simultaneously beneficial and problematic. The benefits relate to their ability to detect the more subtle agro-ecosystem microsite changes which may be important to soil-plant nutrient cycling (Krause and Ramlal, 1987; Binkley and Matson, 1983; Lajtha and Schlesinger, 1988). The difficulty arises in interpretation of highly variable, non-normal (according to parametric statistics) datasets. Although complex, such datasets are characteristic of field measurements. The large inherent variation predicates the need for large sample sizes and frequent sampling intervals. Instead of being viewed as a disadvantage, however, we should utilize the inherent complexity to encourage more non-conventional means of data analysis and interpretation.

CONCLUSIONS

1. The Ionics® anion exchange membrane effectively measures changes in soil solution P and labile P in a high P-retention andic humitropept from Costa Rica.

2. The membrane sorbs P over a wide range of soil solution P concentrations, but is able to detect and sorb P at very low soluble P concentrations. The relationship between AEM-P and soil solution P is 150 L-shaped regardless of whether the soil matrix P was in equilibrium with soil solution P or not, suggesting that the anion exchange membranes are most sensitive to P fluxes at low solution concentrations.

3. The Ionics® anion exchange membrane can be used to estimate accurately soil solution P and, in turn, labile soil P if interference from other anions and soil moisture effects are quantified.

4. The anion exchange membrane behaves as a dynamic exchanger rather than an infinite sink, particularly in the context of a low pH, high P-fixing soil. Because the mechanism of P sorption is simple ion exchange, it is reasonable to assume that P is adsorbed and desorbed from the membrane surface as the soil chemical micro-environment changes. The membrane competes with the soil matrix for labile P, and this competition occurs in both directions. As such, the membrane measures net change and not total accumulation in labile P status over the period during which it is in contact with the soil. This distinction is critical for field data interpretation.

5. The Ionics® anion exchange membrane can be used in situ to measure P fluxes in a variety of field settings. The method is extremely sensitive to micro-scale spatial and temporal changes in labile soil P. CHAPTER III

RESPONSIVENESS OF THE SILVOPASTORAL SYSTEM'S TWO DOMINANT GRASS SPECIES, PASPALUM CONJUGATUM AND HOMOLEPSIS A TURENSIS, AND THE LEGUMINOUS TREE COMPONENT, ERYTHR/NA BERTEROANA. TO VESICULAR-ARBUSCULAR (VA> MYCORRHIZAL INFECTION

INTRODUCTION

Few would dispute that low phosphorus availability is a major constraint to pasture productivity on soils of the neotropics (Jehne, 1980; Janos, 1983; Sieverding and Saif, 1984; Ae et al., 1990). Although volcanic soils and those with Fe and Al oxides contain large amounts of total P, the majority exists as stable organic and inorganic mineral-bound complexes, leaving little in plant available form. As such, many plant species native to soils with high P retention capacities have evolved mechanisms to access a greater range of soil P pools. As Clarkson <1985) states, "growth is the pacemaker for nutrient inflow to the plant and this inflow can be regulated by varying the relative size of the absorbing system or the capacity of its uptake mechanisms." The majority of these mechanisms are biologically mediated and include root extension, root exudation of organic compounds, P-solubilizing bacteria (Borie, 1985} and, most notable, the symbiotic relationship between plant roots and fungi

151 152 known as mycorrhizae (Salinas and Sanchez, 1976; Jehne, 1980; Janos, 1988). Most tropical rainforest species form associations with endophytic vesicular-arbuscular mycorrhizae (VAM); very few form ectomycorrhizal associations (Janos, 1983). Several studies investigating mechanisms of VAM fungi P acquisition have concluded that VAM utilize both the same P pools as non-VAM roots (i.e. the most labile P fractions; Barea and Azcon- Aguilar, 1983; Ae et al., 1990; Ojala et al., 1983) as well as metal oxide- complexed P pools not available to non-VAM roots (Bolan et al., 1984; 1987; Parfitt, 1979; Sanyal and De Datta, 1991; Boerner, 1992a). Mycorrhizae confer advantages over non-VAM roots both spatially and kinetically. They increase total root surface area by external hyphal extension, thereby gaining greater access to positionally unavailable P pools. In addition, they appear to have a lower Michaelis constant (Km) for P uptake than non-mycorrhizal roots, which reduces the effective threshold value for P absorption from soil and leads to greater plant P uptake at lower soil solution concentrations (Habte and Manjunath, 1987; Barea and Azcon- Aguilar, 1983). By improving plant P status, VAM stimulate increased plant growth and confer greater tolerance to extreme soil physical and chemical conditions relative to non-mycorrhizal plants (Jehne, 1980; Bethlenfalvay et al., 1989). Numerous studies have shown that the extent of VAM effectiveness (plant benefits) is plant species specific and is inversely related to available soil P (Stribley et al., 1980; Abbott and Robson, 1987; Siqueira, 1987;

Huang and Yost, 1987; Yost and Fox, 1979). Cool season or C 3 grasses 153 have been shown to be less mycorrhizal dependent than warm season C 4 grasses, for example; and nitrogen-fixing legumes are highly mycorrhizal dependent relative to grasses (Hetrick et al., 1988; Wilson et al., 1991; Sieverding and Saif, 1984). Janos (1980; 1983) proposed a continuum of VAM dependency for tropical plants ranging from facultative (can attain reproductive maturity without mycorrhizae when soil fertility is relatively high) to obligate mycotrophy (cannot survive to reproductive maturity without mycorrhizal infection). Dependency is functionally defined as either the level of P availability below which plants do not grow without mycorrhizae or relative plant growth with and without mycorrhizae. He further hypothesized that mycotrophic status is a function of soil fertility, and suggested that changes in soil fertility or in the probability of mycorrhizal infection can influence vegetation succession. During early succession, when soil fertility is expected to be relatively high, facultatively mycotrophic plant species would dominate; whereas in late succession, soil fertility is low and almost all species are obligately mycotrophic (Boerner, 1992b, c).

Legumes associated with nitrogen-fixing Rhizobium bacteria, in all cases studied, form a tripartate symbiosis with VA mycorrhizae (Barea and Azcon-Aguilar, 1983; Mosse, 1977; van Kessel et al, 1985). According to Beck and Munns (1984), nodule initiation is plant growth rate dependent, and growth reduction by low soil P could be critical to nodule formation. As such, VAM have a dramatic effect on legume nodule activity, which is strongly linked to plant productivity (Bethlenfalvay et al., 1987). Their effect on legumes is not simply to increase plant P concentration which 154 leads to increased plant growth and nodulation; more subtlely, the symbiosis produces higher photosynthetic rates at lower P and N concentrations. The C, N and P supply/demand relationship among host plant, VAM fungus and

Rhizobium bacteria, then, is a fundamental expression of the three symbionts' activities as both sources and sinks for each others' products. The outcome of this association is a positive feedback loop in which VAM infection increases root P concentration followed by increased shoot P concentration leading to more leaf and root growth which stimulates further root exploration for P (Huang and Yost, 1987). From an agricultural perspective, VAM-infected legumes can be managed either to reduce fertilizer inputs or as nutrient suppliers for use in multiple cropping systems. Sieverding and Saif (1984) showed that herbaceous legumes which were VAM-inoculated in the field with only 20 kg P/ha fertilizer added produced 6 8 % more biomass than non-inoculated plants after three months of growth. Likewise, Siqueira (1987) reduced the

P requirement for Styiosanthes (an herbaceous forage legume) by 97% with VAM inoculation. Stated in terms of fertilizer equivalent, VAM inoculation produced yields equivalent to those obtained with 400 mg P/kg soil fertilizer application. Huang et al. (1985) provide further evidence for the potential benefits of field VAM inoculation in a leguminous forage tree, Leucaena ieucocephaia. They obtained four times taller and 80 times greater dry stem weight for mycorrhizal plants compared with non-mycorrhizal plants, van Kessel et al. (1985) established VAM-mediated nutrient transfer from a legume {Glycine max) to a non-legume (Zea mays) which resulted in significantly higher leaf and root N in VAM-inoculated relative to non- 155

inoculated maize. Therefore, field VAM inoculation has the potential to increase plant productivity on soils with inherently low available P. Assuming that legumes are more mycotrophic than grasses {legumes are almost exclusively obligate mycotrophs whereas grasses are usually facultative), one could argue that under low available soil P conditions, VAM-infected legumes would be better scavengers for P than grasses and, therefore, stronger P bioaccumulators. Accordingly, VAM-infected legumes should enhance P cycling when planted in association with less mycotrophic pasture grasses {Ae et al., 1990). Given my tenet dissertation hypothesis that leguminous trees enhance P cycling on a high P-fixing soil, it w as important to evaluate the role of VA mycorrhizae in both leguminous tree and pasture grass P uptake and biomass production. The objectives of this study, then, were to: 1) assess inherent levels of VA-mycorrhizal infection in

native grass pastures of the field study farms, and 2 ) determine plant responsiveness to VAM inoculation using biomass and nutrient uptake

parameters for the two dominant grass species (Paspatum conjugation Berg and Homo/apsis aturensis Chase) and the leguminous tree, Erythrina berteroana Urban. 156

MATERIALS AND METHODS

Characterization of Inherent VAM Infection in Native Pastures

Site Description of Study Farms

The study farms were located in the Atlantic coastal plain of Costa Rica in the Limon Province <10° N 83° W; 30-50 m above sea level). The area receives 3630 mm rainfall annually, most of which is distributed fairly evenly throughout the year. The Holdridge ecological lifezone is lowland humid tropical rainforest although the settlement in which the farms are located was deforested, for the most part, over 20 years ago. Presently, the landscape is a mosaic of subsistence farms, pastures, export crop plantations (banana, pineapple, macadamia, foliage houseplants) and secondary forest patches. The study soil is classified according to US Soil Taxonomy as an andic humitropept, series Neguev and is derived from volcanic debris and mudflows from the late Pleistocene. Due to its stable position on the landscape and high rainfall, the Neguev soil has undergone intense weathering and, accordingly, possesses a low pH, low CEC and very high P retention capacity (see Chapter I for detailed description of soil chemical, physical and mineralogical properties). The field experiment, a silvopastoral system, included cattle grazing and presence or absence of

Erythrina berteroana as the two main effects (a 2 X 2 factorial design). Blocks of the four treatments were established on five farms in 1987. 157

Root Collection and Percent Rootlength Infected with VAM

I measured VAM infection levels in pasture species roots prior to field experiment establishment. Field levels of VAM infection would be used as baseline characterization of the study farms and for comparison with levels obtained under greenhouse conditions. Pasture roots were excavated by cutting 30 cm X 30 cm pieces of turf and separating out fine feeder roots

(roots were a composite of Homo/epsis aturensis, Paspa/um conjugatum, broad-leaved weeds and Cyperaceae). Roots were placed directly into vials containing formalin-acetic acid-alcohol (FAA) solution (90 mL 95% ethanol:5 mL glacial acetic acid:5 mL 37% formaldehyde). I collected 18 root samples per farm, using a random sampling pattern. I washed excess soil from the roots using deionized water and placed a subsample from each vial in Omnisette® tissue microcassettes for clearing and staining (modified procedure of Phillips and Hayman, 1970; Kormanick and McGraw, 1982). I determined total root length and percent infection using the grid-intersection method of Giovanetti and Mosse (1980).

Greenhouse Experiment: Plant Responsiveness to VAM Inoculation

The experimental design included plant species (3) to which either inoculum (cacao feeder roots) or autoclaved inoculum (non-inoculated) were added. I grew the two grass species from seed for 15 weeks after which time I measured above and below ground biomass, root volume and levels of VAM infection. Concurrently, I grew E. berteroana from both seed and 158 vegetative cuttings, seedlings were grown for 30 weeks, cuttings for 18 weeks. Upon harvest, I measured biomass, nodulation and VAM infection levels.

Seed collection

I collected seeds from Erythrina berteroana from trees planted at the field study farms. I also collected seeds from the two principal grass species, Homo/epsis aturensis and Paspatum conjugatum, and composited seeds from all treatments within all farms. Following harvesting, seeds were allowed to dry for several days before planting.

Seed germination

Erythrina seeds had to be scarified before planting. I scarified by rubbing with sand paper to remove a small patch of the seed coat. For each species, I prepared a bed of non-sterilized vermiculite and sowed several rows. Erythrina seeds germinated in 2*3 days, while grass seeds, which were not scarified, took at least two weeks to germinate. During this period, I watered the trays (24 cm X 50 cm) with tap water.

Soil sterilization

During the period that seeds were germinating, I took a composite soil sample from all five farms (approx. 40 kg of unamended field-moist soil) and 159 sterilized it with methyl bromide gas. I placed soil in a large plastic bag three layers thick with a cannister of methyl bromide inside. The bag was sealed tightly and the cannister perforated to release the gas. I let it fumigate 3 d before punching holes in the bag to let it aerate. Once the excess gas escaped, the soil remained inside the closed bag to avoid contamination. I conducted a routine chemical analysis of the soil to serve as baseline information (Table 3.1).

Seedling Transplantation and Inoculation with VAM

Approximately one month after seeds were planted, I transplanted seedlings to individual pots (one plant per pot) containing approximately 700-800 g field moist, sterilized soil. For both grass species, I used pots 12 cm diam. and 11 cm deep. The pots used for Erythrina were 14 cm diam. and 14 deep. For each species there were seven replicates per treatment, except for Homotepsis, where n = 6 . The same day seedlings were transplanted, I added the VAM inoculum. VAM inoculum was prepared according to Janos (1984). I collected cacao feeder roots {Theobroma cacao) from the cacao plantation at CATIE, washed them thoroughly to remove all soil and discarded the large, woody fraction. I then cut fine roots into small pieces and divided two groups into equal amounts: one for inoculum and the other for autoclave sterilization to serve as the control. In addition, I soaked some additional feeder roots overnight in tap water to collect other, non mycorrhizal micro-organisms. The following day, this liquid was filtered Table 3.1. Soil chemical characteristics of bulk soil taken from the study farms after methyl bromide fumigation. ______

______Exchangeable ______Extractable__ pH 8 ______K______Ca______Mg ______Acidity P ______Cu______Zn______Mn ------cmolc/kg soil ------VQ/Q soil ------

4.4-4.5 0.27 1.14 1.24 1.87 17.0 24.1 4.1 142.6 ______10.021 (0.091 (0.081 (0.23) 10.7) (2.8) (0.81 <8.11 n — 3 for all parameters measured; numbers in parentheses are standard deviations. § pH range (1:1 soil:H20).

Table 3.2. Percent hyphal and vesicular infection of cacao (Theobroma cacao) feeder roots used as VAM inoculum.

Sample f hyphal infection (%) vesicles (%) Homolepsis, non-inoc 0 0 Homolepsis, inoc 33 43 Paspalum, non-inoc 0 O Paspalum, inoc 20 65 Paspalum, inoc 15 41 Paspalum, inoc 63 14

i Cacao feeder roots were recovered from pots of respective treatments upon harvest; therefore, samples are identified according to plant species from which cacao roots were taken and whether roots had been sterilized via autoclave (non-inoc) or used to inoculate with VAM (inoc).

Table 3.3. Inherent pasture VAM infection levels measured from farm field samples. ______

Farm VAM infection (%) 5 1 72.57 ( 9.841 2 69.14 ( 7.58) 3 69.05 ( 9.38) 4 75.51 (10.61) 5 66.93 (10.94) i n per farm - 18; numbers in parentheses are standard deviations. 161 through Whatman # 1 paper (qualitative grade) and added to all pots (both VAM inoculated and non-inoculated). All pots received an N,K nutrient solution four days after planting (25:20 N:K; 50 mL total volume of which 25 mL was 25 //g/mL N and 25 mL was 20 //g/mL K) to prevent nutrient limitation from affecting plant growth. The same volume of nutrient solution was added once more two weeks later. Seedlings were watered with tap water as needed to maintain relatively constant soil moisture.

Growth Parameters Measured

At the end of 15 weeks, I harvested the two grass species for leaf, stem, and root dry weight biomass and root volume determinations. Biomass samples were dried in a 70°C forced air oven and then ground in a Wiley mill through a 1 mm mesh stainless steel screen. I determined foliar nutrient content (N,P, macro (Mg, Ca, K) and micro nutrients (Cu, Mn) using

HNO3 -HCIO4 acid digestion (Briceno and Pacheco, 1984). P content is the

only nutrient reported for Paspalum because there was not sufficient plant material for complete analysis of other nutrients. Means and standard deviations for those samples analyzed are presented to discuss the high Mn

levels obtained (Table B.1). For Homo/epsis, only those samples from the VAM-inoculated treatment had sufficient material for chemical analysis; the non-inoculated plants did not. As such, no nutrients are reported for

Homo/epsis. 162

Erythrina Grown from Vegetative Cuttings

Due to Erythrina's woody phenology, I let the seedlings grow longer than the grasses before harvesting. In addition, I reran the experiment using

Erythrina grown from vegetative cuttings, because it is rarely planted from seed in the field. I cut 40 cm long stakes from a single Erythrina berteroana tree to minimize inherent variation. Stakes were planted in sterilized soil along with cacao feeder root VAM inoculum and autoclaved roots for the control pots {pot size: 24 cm diam. 22 cm tall) About one week after planting, the control pots developed a fungal infection, Pironema, a nonpathogenic saprophytic fungus common to sterilized environments. No intervention was taken and the fungus disappeared in 1 - 2 weeks with no adverse effects. During the course of the experiment numerous stake leaves senesced (turned yellow and browned along the edges) and abscised. I applied different remedies including: 1) a non-VAM micro-organism slurry prepared from cacao roots and deionized water and filtered through a 1 0 /ym mesh cloth and 2) inoculation Rhizobium strain CR751 10 weeks after stake planting. I harvested after 18 weeks from planting, selecting 1 0 stakes at random (5 VAM-inoculated and 5 non-inoculated) for analysis.

Parameters measured for Erythrina stakes included leaf and regrowth stem biomass, and foliar chemical analysis. Roots were collected to evaluate VAM infection, but there was not enough material for biomass and nutrient analyses. I measured leaf, stem and root biomass, presence of nodules, and foliar chemical analysis (determined as for grasses) for 163

Erythrina seedlings. Nodules were evaluated using a qualitative index which included presence/absence, abundance, size and color (red indicates nodules are actively fixing N): 1 = absent; 2 = present and few; 3 = present, numerous and red; and 4 = present, abundant, large and red. Biomass samples were ground through a 1 mm mesh screen for complete chemical analysis (N as TKN according to Diaz-Romeu and Hunter, 1978; and P, macro and micro elements in HN 0 3 -HCI0 4 -digests according to Briceho and Pachecho, 19841. For the non-inoculated seedlings, there was not enough material to do chemical analysis for each sample so I composited the material available and ran it as a single sample. I had also collected young and old leaves from the source tree to conduct chemical analyses as a comparison. As with the pasture grasses, N and P are the only nutrients discussed, but means and standard deviations of selected samples are presented for additional nutrients (Table B.1),

VAM Infection Determination

I cleared and stained roots for percent VAM infection according to Koske and Gemma (1989). Percent infection was determined using the slide intersection method (Sverding, 1983; Biermann and Linderman, 1981). This involved mounting between 15-20 root segments per sample, approximately 1 cm long, on a microscsope slide and passing at least 4 times over the slide to note the presence or absence of infection at each intersection (a modification of the gridline intersect method). Percent infection is determined as the number of infected intersections divided by the total 164 number of intersections observed X 100. While scoring for percent intersections infected with VAM hyphae, I also noted the percent of infected intersections with vesicles. Vesicles, believed to be involved in carbohydrate storage, may serve as more definitive evidence of active plant- VAM symbiosis (Sieverding, 1989; Barea and Azcon-Aguilar, 1983). Both percent hyphal and vesicle infection were evaluated for the two grass

species, Erythrina seedlings and Erythrina cuttings. The cacao feeder root inoculum was also collected from several inoculated and control treatments upon plant harvesting and roots were evaluated for percent infection (Table 3.2).

Statistical Analyses

I performed one-way ANOVAs by species on the effect of inoculation on both percent hyphal and vesicle infection using SYSTAT's MGLH procedure for analysis of variance (Wilkinson, 1990). In addition, I performed ANOVAs by species for the effect of inoculation on each biomass (total plant mass, root mass, shoot mass, rootishoot ratio, root volume) and nutrient (total N, P) parameter. I also conducted linear regression analysis for percent hyphal infection and percent vesicle infection against each biomass and nutrient parameter. 165

RESULTS AND DISCUSSION

Inherent Levels of VAM Infection in Field Study Pastures

VAM infection levels of the native grass pastures ranged from 67- 76% of total rootlength and among farm differences in infection level were not significant (Table 3.3). These were consistent with levels reported for tropical pasture species (Sieverding and Saif, 1984). When compared with VAM infection levels from a survey of tropical rainforest roots and adjacent sugarcane field roots (Rose and Paranka, 1987), percent infection in pastures was high (forest roots in the litter/humus layer averaged 46% infection and only 23% in the mineral soil layer; sugarcane roots were infected with VAM in only 20% of rootlength). Wallace (1981) reported infection frequency for a variety of grazing lands ranging from 36-88%; the majority of termperate grasslands hovered around 30-40% infection, however (Wallace, 1987). Berbara et al. (1985) surveyed eight species of tropical herbaceous legumes for levels of VAM infection and found a range between 27-51%. Wallace suggests that mycorrhizae confer grazing tolerance by maximizing below-ground plant biomass, promoting tillering and prostrate growth and maintaining constant photosynthetic rates even under severe clipping regimes. His results for grazing effects on VAM infection levels are contradictory, however. In the earlier study using different clipping regimes (1981), there was a positive correlation between grazing (clipping) intensity and percent VAM infection, whereas in an actual grazing study with 1 6 6 ungulates and cattle (19871, grazing did not influence levels of VAM colonization.

Plant Responsiveness to VAM Inoculation under Greenhouse Conditions

Soil Chemical Characterisitics after Fumigation with Methyl Bromide and Ramifications for Plant Nutrient Uptake

Soil chemical properties were altered by fumigation with methyl bromide (Table 3.11. When compared with values obtained for non­ fumigated field soil (see Chapter I), fumigated soil pH was lower (4.4-4.5 vs.

4.7-5.3), NaHC0 3 -EDTA extractable P was higher (17.0 vs. 5.4 mg/kg),

NaHCC>3 -EDTA extractable Zn was subtantially lower (4.1 vs. 67 mg/kg) and NaHC0 3 -EDTA extractable Mn was 36 times higher (142.6 mg/kg vs.

4.1) than non-fumigated soil. Exchangeable bases,acidity and NaHC 0 3 - EDTA extractable Cu were not affected by fumigation. Other studies have reported side effects of sterilization on soil chemistry including enhanced P availability and dramatic increases in extractable Mn (Stribley, 1987; Williams-Linera and Ewel, 1984; Borie, 1985). Williams-Linera and Ewel (1984) obtained a six-fold increase in extractable Mn with steam sterilization, but claim that this did not affect appreciably seedling germination and mortality. Although their findings with respect to plant effects are somewhat reassuring, the increase in Mn I obtained with methyl bromide fumigation was extreme. This increase in extractable Mn was also reflected in plant Mn levels, which ranged from 972-2860 mg/kg dry weight plant material (Table B.1). In comparison, a study of forages from the Siquirres canton of Costa Rica (the region from 167 which seed was collected for this experiment), averaged 257 mg Mn/kg plant dry matter, with only 11 % of samples having values greater than 500 mg Mn /kg (Vargas and Fonseca, 1989). Although Siquirres canton forage Mn contents are relatively high when compared with other regions of Costa Rica, the greenhouse values I obtained are two to three times greater than Siquirres values and are well above the cited livestock toxicity level of 1000 mg Mn/kg (Vargas and Fonseca, 1989).

Erythrina Mn contents were also quite high: 2860 and 1423 mg

Mn/kg for VAM-inoculated and non-inoculated Erythrina stakes, respectively; and 1521 and 972 mg Mn/kg for VAM-inoculated and non- inoculated Erythrina seedlings, respectively (Table B.1). In contrast, the source tree leaves from which Erythrina cuttings were taken had only 100 mg Mn/kg. Humphreys (1981) cites Mn toxicity threshold values (Mn plant tissue concentration at which yield is reduced by 5% below maximum yield levels) for several tropical legumes as follows: Centrosema pubescens 1600 mg M/kg; Leucaena feucocephala 550 mg Mn/kg; Desmodium unicatum

1160; and Styiosanthes humilis 1140. The only woody legume, Leucaena feucocephala, has the lowest Mn toxicity threshold, suggesting that woody legumes are less tolerant to high Mn levels than herbaceous species. It is difficult to assess the effect of such high soil/plant Mn levels on the overall experimental results. Perhaps the grasses were able to tolerate these levels as there was no mortality. The high Mn may have adversely affected Erythrina cuttings and may have caused the leaf yellowing and abscission I observed. It also may have affected nodulation in Erythrina cuttings. 168

Effect of Inoculation on Hyphal and Vesicle Infection Levels

For the most part, inoculation significantly increased both percent hyphal and hyphal/vesicle infection for nearly all species (Table 3.4).

Inoculation did not affect hyphal infection levels for Homo/epsis aturensis probably because of high within-treatment variation, yet the effect of inoculation on the level of vesicle infection was highly significant (Table

3.5). The effect of inoculation on Erythrina seedlings was greater than for

Erythrina cuttings. Though some VAM hyphae were present in non- inoculated Erythrina seedlings and cuttings, neither ever produced vesicles.

Infection levels for both grasses and Erythrina were lower than the inherent field levels, although this is a common byproduct of greenhouse inoculation experiments (Koslowsky and Boerner, 1989; Mosse, 1977; Borie, 1985).

Erythrina berteroana field infection levels, for example, were almost 50% higher than greenhouse levels (I had excavated roots from two stakes in the field and found hyphal infection levels at 63% and vesicle infection at 11,5- 12.5%). The significant effect of inoculation on percent VAM infection for all species indicated that, although some non-inoculated plants were infected, the treatment distinction was maintained. Analysis of the cacao feeder root inoculum substantiated this claim; the inoculum infection levels paralleled those obtained for the treatment plants, ranging from 15-63% for hyphae alone and 14-65% for hyphae and vesicles (Tables 3.2 and 3.5). Table 3.4. One*way ANOVAs by spades for effect of inoculation on both percent hyphal Infection

Species ANOVA: effect of ANOVA: effect of inoculation inoculation on % hyphal on % vesicles Infection Homohpsis ttvrtnsis ns *** # Ptspa/um conjvqstvm *• • * * £ryfftnoa berferoane seedlings *** • fryrhrms berferoane cuttings • I % hyphal infection Indicates than only intraradical hyphae were observed, whereas % vesicles indicates that both intraradical vesicles and hyphae were scored. 0 * significant at p< 0.05; ** significant at p< 0.01; *** significant at p< 0.001.

Table 3.5. Mean hyphal end vesicle infection levels for the two grsss species and for Erythrina berferoane from ssedHnoa and cuttings. Numbers in parentheses are standard deviations.

Hyphal Infection 1%) Vesicle Infection (%) Species ______n I non-inoculated ______Inoculated non-inoculated ______Inoculated Ptsptktm contugttum6 n 18.13 19.891 48.44 (19.48) 5.05 (6.00) 28.63(18.61)

Homoitptis Bturtnsi,s 6 61.97(14.571 47.52 (28 28) 0.00 (0.00) 51.08(22.66)

£. berferoane 4/7 4.17 (7.22) 41.14 (8.06) 0.00 (0.00) 10.20 (4.17) seedlings £. bertarcane cuttings S 10.64 (10.82) 32.79 (6.86) 0.00 (0.00) 8.92 (5.05) | Where two numbers appear for a given plant species, the top number is n for non-inoculated treatment and the bottom number fa n for inoculated treatment. 170 The source of infection in non-inoculated plants is problematical. I am certain that it did not come from the autoctaved inoculum control roots since autoclaving effectively destroyed any intraradical VAM present in the cacao roots (0% infection). It is possible that watering and soil splashing could have contributed to contamination, or that the microorganism wash contained some spores of Gigasporaceous fungi. Another likely scenario is that macrofauna present in the greenhouse transferred spores among pots (S. Rabatin, pers. comm., 1992). Alternatively, soil fumigation with methyl bromide did not eliminate completely the viable VAM spore population (Fitter, 1977; Wallace, 1987).

Effect of Inoculation and Infection on Plant Biomass and Nutrient Content

Grasses

As may have been expected, Homo/epsis aturensis and Paspalum con/ugatum, responded differently to both inoculation and infection (Table

3.6). For Homo/epsis, inoculation resulted in significantly higher total plant, root and shoot biomass as well as greater root volume. Total plant mass of inoculated plants, for example, was approximately eight times greater and root volume was almost three times greater than non-inoculated plants (Table B.2), However, the relationship between these growth parameters and infection (either hyphal or vesicle) was not linear (Table 3.6).

Paspalum biomass was also enhanced more than twofold by inoculation, but in contrast to Homo/epsis, Paspalum biomass (above and below ground as well as total) was correlated linearly and positively with vesicle infection (Table 3.6). Root volume for inoculated Paspalum was Table 3.6. Analysis of variance for inoculation and regression analysis for hyphal and vesicle infection against growth and nutrient parameters measured for the two grass species. ______

Parameter Regression on % Regression on % ANOVA on ______hyphal infection ______vesicles______inoculation I Homo/eps/s aturensis Total m ass ns ns * Root mass ns ns * Shoot m ass ns ns # R:S ratio ns ns ns Root volume ns ns * Paspalum con/ugatum Total mass ns ** r* - 0.42 * * Root mass ns •* r2 - 0.50 • * Shoot mass ns 0 * r2 - 0.38 * * Total P * r2 - 0.31 ** r2 - 0.46 • ♦ ♦ P concentration ns ns ns R:S ratio ns ns ns Root volume ** r2 « 0.61 ns • ♦ 5 * significant at p< 0.05; •• significant at p< 0.01; *** significant at p< 0.001. # regression significant at p< 0.090, r2 - 0.24. 172 approximately 30% greater than non-inoculated plants and was correlated linearly with the level of hyphal, not vesicle infection. Although there was not enough plant material to quantify nutrients in Homo/epsis, there was a strong mycorrhizal infection effect on Paspaium's total P content (6.63 mg/plant in inoculated vs. 2.53 mg/plant in non-inoculated) manifested as positive linear correlations for both hyphal and vesicle infection. When viewed graphically, these interspecific distinctions were more obvious (Figures 3.1 and 3.2). In fact, the regressions with vesicle infection, for both species, reflected less variation and, therefore, more linear relationships (except for Homo/epsis). There was also a weak negative (and not statistically significant) trend between between

Homo/epsis biomass production and percent hyphal infection. Such interspecific differences in response to VAM infection evoke speculation that the two grasses are located at different points along the facultative-obligate mycotrophic continuum (Janos, 1980; 1983).

Paspaium, because it produced a clear-cut linear response to infection, is perhaps less facultative and closer to obligate mycotrophy than Homo/epsis, whose response produced no clear pattern. Alternatively, interspecific differences in plant response reflect differences in VAM effectiveness (Boerner, 1990; Collins Johnson et al., 1992). Anecdotally, most cattle farmers prefer Paspalum ("pasto natural") over Homo/epsis as forage, claiming that Paspalum is both better quality and more palatable than

Homo/epsis (Neguev settlement farmers pers. comm., 1990). It is tempting to make inferences about responsiveness to VAM infection and forage quality. Although there are few studies comparing two 173

M

4X

r o

u

u>

OjO

0 to 40 M SO 100 0 so 40 00 • 0 100 NrMM Moetod RoottMigtb bwow t Mid*d Rootlangtii with VololM

11 11 tl 10 *

3 1

1

0

0 10 10 00 40 CO H 70 0 W 10 SO 40 SO §0 70 Pw w Mn i i I H h Ui iqi Ii with Vn Um

Figure 3.1. Total plant mass as a function of either percent hyphal or percent vesicle infection for the two grass species. Figs. a & b are for Homofepsis aturensts and Figs. c & d are for Paspatum conjugatum. Tatal Plant p (nj) « 0 either percent hyphal or percent vesicle infection. vesicle percent or hyphal percent either Figure 3.2. 3.2. Figure " S 0 4 S 0 70 00 SO 40 00 SO "■ 0 * < n * i a l * a Pmspa/umconjugatum total plant P content as a function of function a as content P plant total 10 0 P iriw t t iriw P 0 O O 0 O O 70 SO SO 40 SO SO t t *t*gh ih m m VmW with R*atl*ngth M m N

4 7 1 175 warm season, or C 4 , grasses, those comparing forage legumes with grasses have demonstrated that legumes produce larger responses to VAM infection in terms of both biomass production and nutrient uptake (Bolan et al., 1987; Siqueira, 1987). If certain host-VAM associations result in enhanced nutrient uptake, particularly P (e.g. Boerner, 1990), the relationship could be viewed as a means to improve forage quality. In addition, improved nutrient status could augment resilience to grazing and reduce resistance to disease (Jehne, 1980). Wallace (1981) found that VAM infection intensity was positively correlated with grass root production, tillering and grazing intensity; suggesting that mycorrhizae promote growth forms better adapted to grazing stress. If differences in forage quality are legitimate, it may be possible to promote growth of more desirable forage species by managing for the most beneficial host-VAM fungi relationships (Collins Johnson et al., 1992; Rosendahl et al., 1990; Sieverding, 1990). Inoculation did not affect the root:shoot ratio in either grass; i.e., VAM infection did not stimulate root production any more than shoot production. Usually, the root:shoot ratio is inversely related to VAM- influenced plant growth (Bethlenfalvay et al., 1989). This relationsip is often the result of plant stress. Fitter (1977) found Lofium perenne's root:shoot ratio decreased as a function of VAM infection and competition from Ho/cus fanatus. Wallace (1981) also found that mycorrhizal infection stimulated root growth relative to shoot growth in intensively grazed

Panicum coforatum . In the first case, the VAM association was detrimental

(resulted in decreased competitive ability for L. perenne relative to H. 176

lanatus); in the second case, the VAM effect was beneficial as a grazing survival strategy.

Leguminous tree

Inoculation and mycorrhizal infection had a strong effect on Erythrina

berteroana seedling growth and nutrient uptake, but the effect was less

clear for E. berteroana grown from vegetative cuttings (Table 3.7). Total seedling mass and total P content for inoculated seedlings were nine times greater and total N content was five times greater than non-inoculated plants (Table B.2). The root:shoot ratio was not affected by inoculation and showed no relationship with either hyphal or vesicle infection levels.

All biomass and nutrient parameters measured for E. berteroana seedlings (including total N and P, N:P ratio and N concentration) exhibited significant linear relationships with hyphal infection (Figures 3.3 and 3.4). Both plant N concentration and the N:P ratio exhibited negative linear correlations with percent vesicle infection; i.e. N concentration was higher in non-inoculated plants than inoculated (104 vs. 59 mg N/g plant material) even though total plant N in inoculated plants was five times higher than non-inoculated plants. The higher N concentration in non-inoculated plants was probably an artifact of relatively high total N in very little total plant mass.

In contrast to E. berteroana grown from seed, biomass production

(new growth) and P uptake for E. berteroana grown from cuttings (stakes) were negatively affected by mycorrhizal infection (Tables 3.7, B.2). New growth mass for non-inoculated stakes was slightly higher than inoculated 177

■

4 i I *

i 1

o 0 0020 SO 40 SO •010 0 ■ 10 IS to N n M M n M S m Mm o Ui Pw m M Mooted RMllinglh with Vh Mm

S I

1 1

1 1 o 0 0 I 4 ■ 10 12 M 0 1 0 so so 40 so so N w m M m M RoMtenfth wMh Vn M n

Figure 3.3. Total plant mass for E. berteroana grown from seedlings and new growth (stem and leaf) mass for E. berteroana grown from vegetative cuttings as a function of either percent hyphal or vesicle infection. Figs. a & b are for seedlings and figs. c & d are for cuttings. T«M Hwm QttwWi P 1*1) Tfttal Plant P (* * ) « « U utns c & (cd). cuttings Figure 3.4. Total plant or new growth P as a function of either percent percent either of function fora infection P as vesicle or growth hyphal new or plant Total 3.4. Figure H 1 > 9 4 SO 40 90 >0 10 0 ■ 0 voit O HailliigO f o W Pvroairt 10 to 0•0 00 40 00 . berteroana E. w n PfOaiW seedlings (a & (aB> seedlings and ' I • t ' I I ' t • I ' M m M WaoOanj with h t 1 ' I I ' I ' I 10 voWVN V m 1 I o I u •0 178 1 7 9

Table 3-7 Analysis of variance for inoculation and regression analysis for hyphal and vesicle infection against growth and nutrient parameters for Erythrina bartaroana seedlings and vegetative cuttings.

Parameter Regression on % Regression on % ANOVA on hyphal infection vesicles inoculation I Erythrina bartaroana grown from seed Total m ass * r* - 0.54 ns * Root m ass • * r2 - 0.69 ns ♦ ♦ Shoot mass # * r2 - 0.48 ns • Leaf m ass * r2 - 0.40 ns * Stem mass * r2 - 0.52 ns «- N concentration ** r2 - 0.67 •• r2 - 0.59 * ♦♦ Total N * r2 - 0.42 ns • P concentration ns ns ns Total P ** r2 - 0.69 ns 9 * ♦ N:P ratio * r2 - 0.46 ** r2 - 0.65 * * R:S ratio ns IS ns ns Erythrina bartaroana grown from cuttings New shoot mass ns ns Leaf m ass ns ns ns Stem m ass ns ns ns ## Total N ns ns ns Total P ns ns i * significant at p< 0.05; ** significant at p< 0.01; *** significant at p< 0.001 # shoot mass includes both leaves and stems. @ regression significant at p< 0.087, r2 - 0.32. IS regression significant at p< 0.073. r2 - 0.35. ## difference significant at p< 0.054. 9 9 regression significant at p< 0.085, r2 - 0.35. 180 stakes {7.0 vs 5.9 g) and total new growth P difference was of a similar magnitude (13.9 vs. 10.3 mg P/plant). There were no significant linear effects, however, between either hyphal or vesicle infection levels and any of the growth or nutrient parameters (Figures 3.3 and 3.4). In addition, both inoculated and non-inoculate stakes lacked nodules upon harvest, suggesting that either the Rhizobium strain added was not compatible with

E, berteroana or that Mn toxicity may have impeded nodule development. The slight negative effect incurred from inoculation suggests that, in the initial phase of stake establishment, the cost of carbon drain to support the mycorrhizal symbiont outweighed the potential benefit of enhanced nutrient uptake from the plant's association with the fungus (Buwalda and Goh, 1982; Clarkson, 1985; Bethlenfalvay et al., 1987). During this early stage, the stake's ability to generate both new above-ground growth and roots is a function of stake size (nutrient and energy storage capacity). Therefore, it relies entirely on internal reserves and not nutrient uptake from the soil for initial establishment and, accordingly, any association with a carbon-demanding symbiont would diminish those internal reserves. It is probably not coincidental that there were no nodules on stake roots upon harvest. Perhaps the limited carbon and nutrient reserves present were inadequate to support either VAM fungi or Rhizobium bacteria symbionts. Seedling establishment operates on a totally different strategy. Initial root growth is critical for nutrient uptake because seed and cotyledon reserves are extremely limited. If association with VAM fungi at an early stage promotes root growth and subsequent enhanced nutrient uptake, the ensuing positive feedback mechanism confers advantage (increased growth) 181 over seedlings with no VAM association (Habte and Manjunath, 1987; Huang and Yost, 1987; Huang et al., 1985). There are numerous studies which demonstrate positive response to mycorrhizal inoculation and infection in both herbaceous and arboreous legumes, and corroborate the results I obtained for Erythrina berteroana seedlings. Janos (1975) evaluated numerous tropical rainforest tree species, legumes among them, and found that inoculation significantly increased seedling height, leaf number and leaf size in all species. Huang et al. (1985) note five-fold increases in leaf area for Leucaena teucocephaia seedlings inoculated with VAM fungi. Sieverding (1989) also surveyed several leguminous tree nursery species for seedling growth with and without VAM inoculation and obtained growth increases in inoculated plants from 1.3-5 times those of non-inoculated plants (species included Acacia saiigna, Leucaena feucocephaia, inga cerstediana). Russo and Mora (1986) showed that E. berteroana's congener, E, poeppigiana (used as a shade tree in coffee plantations in Costa Rica), when inoculated with either non­ sterilized soil or VAM inoculum (added to sterilized soil), produced increases in both seedling height at harvest and above-ground biomass. VAM and

Rhizobium together had a positive effect on leaf area as well. Indeed, mycorrhizal infection had a significant effect on nodulation in E. berteroana seedlings (Figure 3.5). The non-inoculated seedlings produced no nodules after 30 weeks while the inoculated seedlings had numerous, moderate-large red nodules. Several studies allude to the synergistic relationship between Rhizobium and VA- mycorrhizae (Mosse,1977; Barea and Azcon-Aguilar, 1983; Sieverding and Saif, 1984). Most confirm that 1 8 2

non-inoculated (n-4) B inoculated (n-7) 1 0 0

80 -

*

o 60 2323234853485353484848484890535353535348484823232323235353

0 ■o > s 40 cO o 3

u- 2 0 -

1 4 absent abundant Nodulation Index

Figure 3.5. Frequency distribution (H total par treatment) of nodules tor inoculatod and non-inoculated Erythrina bartaroana seedling*. Tho nodulation index is a qualitative assessment where t-no nodulos. 2-few, 3-numsrous and 4~abundant. 183 VAM infection stimulates nodulation. van Kessel et al. {1985) found that leguminous plants infected with both Rhizobium and VAM showed increases in nodulation and N fixation compared to non-mycorrhizal plants. They attribute the increased N fixation to increased P uptake via VAM hyphae.

Although it is beyond the scope of this study to assess the level of E. berteroana's mycorrhizal dependence, one can infer from both growth and nodulation evidence that it is probably an obligate mycotroph.

Differences between hyphal and vesicle infection

The presence of intraradical hyphae with and without vesicles along with differences in infection effects from the two have led to speculation about their significance. One possible explanation is that they represent infection by different VAM families. Areas infected without vesicles may represent those VAM species which produce only extraradical vesicles or auxiliary cells; eg. those from the family Gigasporaceae, whereas areas with both hyphae and vesicles represent intraradical species of the family Glomaceae. It is impossible to verify this hypothesis since spore identification was not within the scope of this study. Nonetheless, given the considerable growth period prior to harvest, it is likely that plants without vesicles after 15 weeks would not have produced vesicles at some later point in their growth (R. Boerner, pers. comm., 1992). It is difficult to extrapolate from other studies* findings, but it is quite possible that the mixed species inoculum contained both VAM families.

Janos (1975}has isolated Sc/erocystis and Acauiospora genera from cacao 184 feeder root inoculum, both from the interadical vesicle family Glomaceae. He notes, however, that Gigasporaceous fungi are present in tropical plant species; they are simply not as common as Glomaceous species (Janos, pers. comm., 1992). Hyphal regrowth from auxiliary or extra-radicle vesicles of Gigasporaceae can initiate intraradical mycorrhizal infection (Pons and Gianinaz2 i-Pearson, 1985). It is likely, then, that the non-vesicle, intraradical infection I observed was as functional as the vesicle/hyphal infection. Several studies support the hypothesis that VAM species-host plant associations differ in their effectiveness; i.e. certain mycorrhizae species produce large, positive growth effects with certain plants and less significant or even deleterious effects with other plant species (Sieverding and Toro, 1987; Fitter, 1977; Carling and Brown, 1980; Boerner, 1990, 1992c). Furthermore, others suggest that effectiveness can vary as a function of host-fungus ontogeny and changes in external soil chemical and p hysical conditions (Bethlenfalvay et al., 1989; Collins Johnson et al. 1992; Rosendahl et al., 1990; Sieverding, 1990). This observation could explain why most of the growth and nutrient effects for Paspaium conjugatum grass were correlated with levels of vesicle infection (presumably Glomaceae) while most of the growth parameters for Erythrina seedlings were correlated with levels of hyphal infection (presumably Gigasporaceae). The effectivity hypothesis could also be evoked to explain differences in VAM-related growth response between the two grasses. Another possible explanation for infection differences is that areas without vesicles represent non-functional infections or that the infection 185 function is not simple nutrient acquisition. Certain mycorrhizal associations are not true mutualisms, and the host-fungus symbiosis can be modified during the plant’s phenology and as a function of soil P conditions (Abbott and Robson, 1987; Aziz and Habte, 1987; Habte and Manjunath, 1987;

Janos, 1983). This is probably not valid since Erythrina seedlings exhibited significant growth enhancement related to increased levels of hyphal infection (S.Rabatin, pers. comm., 1992). It is possible, as Allen et al. (1980; 1982) suggest, that growth enhancement was not a function of VAM-mediated nutrient uptake (which would promote the presence of VAM structures involved in nutrient transfer, arbuscules, and the end product of nutrient transfer, vesicles); rather, the VAM association altered the hormone balance in the host plant. Allen et al. have found that VAM-infected plants had increased activity of growth hormones, such as cytokinins and gibberellins, relative to non-VAM plants. They further suggest that increased cytokinin levels have a positive feedback effect on VAM root infection (trigger increased root infection by reducing resistance to fungal invasion) and subsequent P utilization. Alternatively, mycorrhizal infection could have promoted enhanced levels of rhizosphere phosphatase which could bypass the need for VAM-mediated P uptake (Dodd et al., 1987).

CONCLUSIONS

1. Growth response to mycorrhizal infection differed between the the two grass species, Paspaium conjugatum and Homolepsis aturensis. Such differences may reflect differences in VAM-host effectiveness, or represent 186 plant differences along the facultative-obligate mycotrophy continuum. As the two grasses also differ in their perceived forage value, it may be possible to manipulate mycorrhizal infection in pastures to promote VAM species which promote proliferation of more desirable forage species.

2. Erythrina berteroana seedlings were highly responsive to inoculation with VAM fungi. In addition, nodulation appeared to be linked strongly to mycorrhizal infection. The increased biomass and foliar P levels support the hypothesis that mycorrhizae are important for enhanced P cycling in the

silvopastoral system. Since VAM infection results in greater Erythrina leaf P contents, more P will be released into the available P pool as leaves decompose and P is mineralized.

3. Erythrina grown from vegetative cuttings responds either weakly or negatively to inoculation with VAM, at least in its initial growth phase. In the initial phase of stake establishment, the cost of carbon drain to support the mycorrhizal symbiont may outweigh the benefits derived from the host- fungus association. This relationship should be investigated in the context of establishment in agroforestry systems since Erythrina (as well as many other leguminous trees) is planted almost exclusively from vegetative stakes. At best, the negative VAM effect may be an artifact of greenhouse establishment or Mn toxicity, and is probably ephemeral in the field. It would be important to evaluate infection effects over a longer time period, both in the greenhouse and in the field, to assess the dynamic nature of the cost/benefit relationship. 187 4, Differences between grass and leguminous tree response to vesicle and hyphal infection may be related to infection by different VAM families and/or differences in host plant-VAM species effectivity. CHAPTER IV

DECOMPOSITION OF LEGUMINOUS TREE {ERYTHRINA BERTEROANA), PASTURE GRASS AND CATTLE DUNG RESIDUES AND EFFECTS ON LABILE SOIL P DYNAMICS

INTRODUCTION

Soil nutrient dynamics in many agro-ecosystems are closely tied to decomposition and nutrient release patterns from ecosystem residues. Residue inputs are considered analogous to litter layers in undisturbed ecosystems; organic material gradually decomposes and becomes incorporated into the soil organic matter pool (Budelman, 1988). In the process, nutrients critical to plant growth are mineralized and added to the bioavailable pool. The analogy is somewhat tenuous in that naturally produced litter is often nutrient-poor; i.e., it constitutes senescent and nutrient-retranslocated plant material. In contrast, agro-ecosystem residues can be manipulated to enhance nutrient content and thereby maximize their contribution to bioavailable nutrient pools. Prunings from leguminous trees, for example, are often used in agroforestry and alley-cropping systems as crop mulches, mainly because they offer a relatively concentrated, readily available nitrogen source (Palm, 1988).

1 8 8 189 Several researchers have postulated synchronicity between residue decomposition and crop nutrient demands as a means to improve crop production, particularly in agroforestry systems (Swift, 1985; Szott et al., 1990; van der Kruijs et al., 1989). They suggest that the efficiency of nutrient availability can be manipulated by varying the quantity, quality and timing of residue inputs. Contingent to our ability to achieve such precise manipulations, however, is an in-depth understanding of the factors affecting both the decomposition process and subsequent nutrient release characteristics. In addition, we need to understand how the soil matrix (both abiotic and biotic constituents) and live plant roots (rhizosphere) interact with residue nutrient release patterns. In general, litter decomposition is regulated by; 1) ambient and edaphic conditions ; 2) litter quality; and 3) the community of decomposer organisms (Swift, 1986; Blair et al., 1990). The three are related integrally and operate in a feedback loop. Soil pH, soil moisture, C>2 supply, temperature (soil and air) and rainfall are considered the major environmental factors affecting organic matter decomposition because they, in turn, affect decomposer activity (Hopkins et al., 1988). Litter quality is often defined in terms of relative nutrient concentrations (C:N and C:P ratios as well as actual concentrations) and amounts of labile (easily degradable) versus recalcitrant compounds. Most studies concur that C:N and C:P ratios in excess of 20:1 and 100:1 respectively result in microbial immobilization of N and P and slower decomposition rates (Huffman et al., 1991; Parmelee et al., 1989; Smeck, 1985). In addition, relatively high concentrations of certain plant structural and secondary compounds like lignin and polyphenolics are known to modify decomposition rates and nutrient release 190 (Palm, 1988; Szott et al., 1990; Vilas Boas, 1990). Residue manipulation strategies, then, can be based on differences in residue quality. Residues with high nutrient contents and low recalcitrant compound contents are hypothesized to decompose and mineralize nutrients quickly, which could result in a large, brief nutrient pulse subject to plant uptake or leaching. Residues low in nutrients and high in recalcitrant compounds decompose more slowly, which could result in build-up of soil organic matter pools. Either or both strategies could be employed to achieve either immediate plant supply and/or organic matter accumulation. Nutrients are released from decomposing residues via either simple leaching of easily soluble compounds or microbially-mediated mineralization of organically-bound compounds (Blair, 1988). The release characteristics (rate dynamics) depend on the nutrient's initial concentration in the litter and its structural configuration in the litter matrix ,as well as microbial demand and labile soil reserves. If nutrients are scarce and soil replenishment slow, microbes will immobilize those nutrients, resulting in release rates slower than actual mass loss rates. In addition, as decomposition proceeds, there is often an interaction between release rate and microbial production of new, more recalcitrant compounds or physical isolation in stable organo-mineral complexes (Coleman et al., 1984; Hopkins et al., 1988; Duxbury et al., 1989). Conversely, nutrient release from rapidly decaying litter could prime decomposition of adjacent recalcitrant material leading to a mineralization pulse (Blair et al., 1990). Phosphorus is usually the most limiting nutrient for plant growth in tropical and subtropical regions and yet the dynamics of residue contribution to P fertility are not welt understood. Phosphorus cycling at the ecosystem 191 leveJ differs from carbon and nitrogen cycling in that biotic and abiotic components interact strongly (Duxbury et al., 1989; Smeck, 1985). The biological subcycle, whose pathways include inputs from living and dead biomass and microbe-driven transformations to either labile or stable organic P forms, is much more dynamic than the pedological pathway whose endpoint is an occluded P sink; i.e., stable forms including mineral P precipates, P sorbed via ligand exchange and organic P physically isolated in microaggregates (Figure 4.1). Moreover, the biologically-mediated pathways are in constant battle with the system's tendency to degrade to more stable occluded P forms. Dynamic equilibria do exist between labile soil P pools and solution P, however, and labile organic P pools contribute significantly to solution P replenishment. It is critical, then, to enhance the short-term dynamic equilibria to keep P in more plant-available forms. The interactions between biotic and abiotic components often prescribe P mineralization and immobilization patterns. High P-retaining soils, for example, shift equilibrium reactions toward abiotic immobilization, which, in turn, decreases microbial P supply, leading to microbial P immobilization. This scenario would translate to reduced decomposition rates and reduced P mineralization from decaying residue. Several studies, suggest, however, that the plant rhizosphere (root exudates, less mineral soil, higher organic matter content, more biologically active) can moderate adsorptive effects from high P-fixing soils and keep more P in soil solution for both plant and microbial use (Dalai, 1979; Comerford and Dyck, 1988). In an attempt to elucidate P dynamics as a function of leguminous tree pruning and cattle grazing in a humid tropical silvopastoral system, it was critical to isolate P fluxes associated with individual system inputs; i.e.. VEGETATION

•nd

PUUIT RESIDUES PRIMARY MINERALS

STABLE SOLUTION P SECONDARY MINERALS SOIL MICROBES

OCCLUDED P LABILE OBGANC P

WffM*. wbturti ind Mraimww

OBILE INORGANIC P

LEACHING LOSSES

Figure 4.1. General schematic of phosphorus cycle (source: Walbridge, 1991). 193

leguminous tree (Erythrina berteroana) leaves and stems, pasture grass clippings and cattle dung. In addition, it was important to quantify the various interactions between soil P retention, residue type, pasture sod roots, timing of P release, and the resultant soil solution P concentration. I generated the following set of hypotheses to address specific aspects of these relationships: 1. Those residues with relatively high P contents and subsequently rapid decomposition rates are expected to release P into the soil solution more rapidly than low P content, slow decomposing residues.

2. Cattle dung is expected to release P more slowly than Erythrina leaves even though it has higher P content. This may be related to a greater content of recalcitrant compounds in dung relative to Erythrina leaves.

3. If Erythrina leaf and stem residues are combined, P released from rapidly decomposing leaf material may be immobilized by microbes decomposing woody material thereby slowing overall P release from Erythrina residues. 4. The relationship between residue application rate and P release (as detected ultimately in the soil solution) is expected to be curvilinear; i.e., for low amounts added, the amount detected in the soil solution is expected to be negiible. It may, in fact, be a threshold phenomenon rather than a continuous function such that some critical amount of P must be present before significant P pulses are detected. 5. Soil P retention (immobilization) is expected to be a major sink for P released from decomposing residues. For the purposes of this experiment, P retention will include adsorption, precipitation of amorphous Al-phosphate minerals and microbial immobilization. Microbial immobilization is included because it would be extremely difficult to separate biotic effects from 194 mineralogical effects; i.e., the experiment could not include a treatment with sterilized soil. Therefore, overall P retention is expected to supress P concentrations in the soil solution which, in turn, should decrease residue decomposition and P mineralization. 6. Pasture sod roots are expected to be the other major sink for P released from decomposing residues, and are expected to have a similar effect on soil solution P as soil P retention. Alternatively, pasture roots may reduce soil P sorption via production or organic ligand compounds, resulting in enhanced soil solution P levels and more rapid P mineralization. The objectives of this study were to determine decomposition rates and subsequent soil P dynamics for the principal silvopastoral system residues under both controlled (greenhouse) and field conditions. In the greenhouse, I proposed to evaluate: 1) the effect of placement surface and rhizosphere (bare soil versus sod) on residue decomposition, nutrient release and labile soil P dynamics; and 2) the effect of residue application rate on labile soil P dynamics. In the field, I sought to determine: 1) residue decomposition rates and nutrient loss rates within a variable environmental context; and 2) the relationship between residue P loss and labile soil P fluxes directly beneath decomposing residues.

MATERIALS AND METHODS

To meet the dual objectives of determining both specific (reductionist) and multiple (integrated) effects on decomposition rates, P release characteristics and subsequent changes in labile soil P, I conducted 195 decomposition experiments in both controlled (greenhouse) and field environments. In both cases, I used the field experiment soil, the Neguev series. I also obtained silvopastoral system component residues, including

Erythrina berteroana leaf and stem biomass, pasture grass clippings and cattle dung, as composite samples from the field experiment treatments. I include, therefore, a brief description of the field site along with methods for the greenhouse and field decomposition experiments separately.

Site Description

The field experiment was located in the Atlantic coastal plain of Costa Rica (10° N 83° W). The area receives 3630 mm rainfall annually and the ecological lifezone is lowland humid tropical rainforest. The soil is an andic humitropept (series Neguev) and is located on the ridgetops of a slightly undulating landscape. The geologic origin is volcanic ash and lahars from the late Pleistocene; due to its stable position on the landscape and high rainfall, however, it has mineralogical characteristics of more highly weathered soils (kaolinite, gibbsite and iron oxides). In general, the Neguev series is deep, well-drained, acid, low in exchangeable bases, high in exchangeable Al, high in total P but low in available P, clayey and low in bulk density. In addition, the Neguev soil has an extremely high P retention capacity (> 2000 mg P/kg soil) and a high moisture retention capacity over a wide range of soil moisture tensions. The field experiment, a silvopastoral system, was a 2 X 2 factorial design with cattle (grazing) and trees as the two independent variables. I planted Erythrina berteroana (a tropical leguminous tree) from 2.6 m 196 cuttings (stakes) in native grass pastures on five farms, all within a 6 km radius (all inside the Neguev settlement). All farms were located on the Neguev soil series upon which I established the randomized block of four treatments (farms were considered as experimental replicates). Grazed treatment plots were 900 m2 and non-grazed treatments 400 m2.

Greenhouse Experiment: Decomposition and Dynamics of P Release from Erythrina residues, pasture residues and cattle dung

I assessed effects of residue type, residue application rate and placement surface (i.e., sod or bare soil) on both decomposition rates and labile soil P using two concurrent experiments. I measured: 1) mass and nutrient loss via litter bag decomposition (one application rate only) in one set of containers (flats) on both sod and bare soil; and 2) labile soil P fluxes in another set of fiats (aluminum pans) using a 4 X 3 X 2 experimental design (4 residue types, 3 application rates, 2 decomposition surfaces).

Material Collection from the Field

Soil collection To determine amount of soil needed, I calculated the surface area of experimental containers (aluminum pans: 1000 c m 2 ) ancj assum ed a depth of 4 cm soil; thus giving a soil volume of 4000 cm^. The soil field-moist bulk density was approximately equal to 1 g / c m ^ ; therefore, 4 kg per container were needed or a total of about 300 kg soil. The 300 kg were collected as a composite sample (0-20 cm depth) from three of the five study farms. I mixed the soil thoroughly by hand and broke up the large clods. I also removed most roots, biological debris and stones while 197 attempting to maintain soil macrofauna alive (earthworms eventually vanished from the pans, however). I then transferred homogenized soil into the experimental containers for both the labile soil P and decomposition components of the experiment.

Pasture sod I collected pasture sod by cutting intact pieces of sod from one farm site (farm 5) to minimize site-related variation. Although sod was collected from outside the field treatment plots so as not to damage treatments, it was completely representative of the botanical composition within treatment plots (dominated by Paspa/um conjugatum, Homo/epsis aturensis, broad- leaved weeds, cyperaceous weeds). I was careful to transport the sod with sufficient soil to minimize transplant shock. Once in the greenhouse, I gently washed soil from sod roots and immediately transplanted into homogenized soil and watered. I maintained both sod and non-sod treatments moist (watered daily to prevent drying out) for 4 weeks prior to residue placement to give the microcosms time to equilibrate.

Labile Soil P Dynamics Component

Experiments/ Design The experimental design of the P dynamics component was a 4 X 3 X

2 factorial; the factors being: 1) residue type (Erythrina leaves, Erythrina leaves and woody stems mixed, clipped pasture, cattle dung); 2) residue application rate; 400, 600, 800 kg/ha DM; and 3) placement surface (bare soil or sod). The control was placement surface with no residue applied. The 24 treatments were replicated in duplicate. The dependent variable 198 w as labile soil P measured as anion exchange membrane-extractable P (AEM-P using the Ionics** anion exchange membrane; see Chapter II).

Residue Application Rate Determination

I calculated a range of dry-weight biomass values (kg/ha) representative of those found in the field using field biomass data for each residue type. Although the ranges differed slightly for different residue types, I decided to use the same three application rates for all residue types to minimize variation among treatments: 400, 600, 800 kg/ha dry matter. The lowest rate was calculated as the minimum amount needed ion a fresh weight basis per 1000 cm^, the size of the experimental containers) to produce detectable levels of labile soil P measured with anion exchange membranes (AEM). Initial fresh weights were calculated for each residue type using means of dry matter content (%) according the following relationship: FW = DW/% DM or DW/(1-% moisture); where FW is fresh weight, DW is dry weight (oven-dried at 70°C) and DM is dry matter. I calculated means for DM content for Erythrina biomass (25.2%) from the second pruning event (November 1989), for pasture clippings (21.1%) from numerous grazing cycle biomass data (1989), and for cattle dung (11.8%) from dung collected during the first pruning event (June/July, 1989).

Field Collection of Residues and Placement on Treatment Pans

I collected all material from one farm (farm 5) to minimize variability associated with each residue type. I composited dung samples using dung 1 9 9 from numerous animals which was either freshly deposited or had been deposited no more than 2 h prior to collection. Using a machete, I then cut and collected random pasture patches in each of the four treatment plots. For Erythrma leaf and stem biomass, I selected branches at random from at least 20 different trees in both grazed and non-grazed treatment plots with trees. The trees had been pruned approximately three months prior to collection time. Leaf biomass was composited from numerous branches and from resprouts at tree bases. I transported residues in ice coolers from the study site in the Neguev settlement to the greenhouse in Turrialba (approximately 1.5 h drive) and placed residues on treatment pans in the same day. Prior to placement on treatment pans, all residue types were mixed well and plant residues were chopped into 5-10 cm long pieces. For

Erythrina stems and leaves, stems of various diameters were chopped and mixed with chopped leaves such that all pans received at least some stems with leaves. I spread loose residues evenly over treatment pans and covered treatments with 8 mm X 8 mm mesh black plastic screening (material used for shading greenhouses) to simulate conditions of residues in decomposition bags.

Anion Exchange Membranes, Membrane Placement, and Extractable P Determination

As mentioned, I used Ionics® anion exchange membrane type 204-U- 386 to measure changes in labile soil P over the eight-week decomposition study. The membrane consists of cross-linked copolymers of vinyl monomers containing ammonium anion exchange groups (see Chapter II for 2 0 0 more thorough desciption of membrane properties). Membranes used in the experiment were cut to 2.5 cm X 2.5 cm or 6.25 cm2 surface area and presaturated with Cl’ anions using a 1 M NaCI solution (minimum 24 h prior to soil placement). To facilitate membrane recovery and removal, I sewed pieces of unwaxed dental floss into the corner of each membrane square. I placed three membranes per treatment pan prior to residue placement (day 0 of experiment). Membranes were always buried just below the surface (0-2.5 cm) as the pans were fairly shallow. Spacing between membranes was approximately 12 cm. I left membranes in the pans for 7 d at which time they were removed, washed with double deionized water (DDW) to remove soil and debris and stored in DDW until NaCI extraction (never longer than overnight). On the same day that membranes were removed, I buried a new group for another 7 d. The entire procedure was repeated for a total of eight weeks. I used 15 mL of 1 M NaCI per AEM to extract orthophosphate-P adsorbed to the membranes. Membranes were extracted for 1 h at low speed and, because extract solutions were debris-free, I took a 10 mL aliquot directly for analysis. (For the first two weeks' membranes, I had included a filtering step in the process but it turned out that the filter papers were contaminated with P as deduced from a separate experiment discussed below. As a result, the filtering step was eliminated). Samples were analyzed according to the ascorbic acid-reduced phosphomolybdate method (Olsen and Sommers, 1982) and P concentration determined using a Spectronic 20D spectrophotometer. Once membranes had been extracted with NaCI and the extract saved for analysis, AEMs were washed with DDW 201 and resaturated with 1 M NaCI to be recycled for subsequent labile soil P measurements. After running "blanks" (clean membranes extracted with NaCI, filtered, etc.) along with greenhouse membranes for the first two weeks and finding relatively high P concentrations in those blanks, I decided to run a separate experiment to determine the cause of contamination. I tried running blanks without filtration in both DDW and NaCI and discovered that the source of P contamination was the filter paper (Baxter qualitative paper). In an attempt to recuperate the first two weeks' greenhouse AEM-P data, I calculated a contamination factor (mean of n = 50) from the contamination experiment data and subtracted this factor from the first two week's AEM-P extract concentrations. In cases where the adjusted value was negative (<

0), they were corrected to 2 ero.

Soil Moisture Prior to placing the first group of membranes, I determined gravimetric soil moisture for each pan. I decided to maintain soil moisture in the range of 55-70 % (by weight) so that moisture would not limit diffusion. I determined gravimetric moisture content for each pan at time prior to residue placement, reweighed each pan once per week and adjusted water content accordingly. I also watered pans daily (estimating amount water needed) because greenhouse conditions were often very hot, and surface soil tended to dry out by late afternoon (especially pans with bare soil). 2 0 2 Expenmentai maintainence I placed pans at random on the benches in the east-facing section of the greenhouse and rotated the pans weekly. I adjusted gravimetric soil moisture to minimize microclimate effects on a daily basis. I removed emerging seedlings from bare soil treatments to eliminate potential root effects in non-sod treatments. Sod treatments were left unclipped during the eight-week study.

Statistical Analysis

I performed univariate analysis of variance using SYSTAT to determine residue type, placement surface and application rate effects on AEM-P fluxes (Wilkinson, 1990). Because the sampling unit (each treatment combination) had been monitored repeatedly over time, I also performed repeated measures analysis of variance for both univariate and multivariate effects on AEM-P over time. I conducted analyses using both unadjusted AEM-P (without the no residue control AEM-P values removed) and adjusted (with control AEM-P removed) AEM-P values.

Litter Bag Decomposition

I measured mass and nutrient loss for the four residue types concurrently with the AEM-P measurements; however, this part of the experiment was conducted in separate containers. Initially I had hoped to measure both simultaneously in time and space, but the amount of DW biomass needed for residue chemical analysis (C, N and P) throughout the 203 eight weeks of the experiment greatly exceeded the range of DW biomass values encountered in the field. I, therefore, sacrificed the ability to make direct comparisons between residue P release and labile soil P. The experimental design consisted of two factors: the same four residue types as used in the labile P portion of the experiment and the two placement surfaces, sod versus soil. Again, the eight treatments were duplicated for a total of 16 experimental units. I calculated the minimum amount of dry matter needed for chemical analysis as 2.5 g DW per week and doubled this amount to ensure sufficient material to determine total DW biomass (per litter bag) needed. I then converted this to fresh weight per bag per residue type using the same DM content and equation used in the AEM-P part of the study. I constructed litter bags from black plastic (polyethylene) screening used for shading greenhouses, mesh size 4 mm. Litter bags for plant residues were 10 cm X 10 cm while those for dung were 5 cm X 10 cm. The dung bags also contained a piece of fiberglass fine mesh screen (2 mm mesh size) placed inside the bag on its underside to prevent fine particle loss. The same day that I initiated the AEM-P component of the experiment (i.e., put membranes in the soil and placed fresh residues on sod or soil), I filled litter bags with fresh residues. I placed bags randomly on their respective decomposition surfaces (bare soil or sod). I also took a separate sample of each residue type for DM content at time zero. I arranged decomposition containers (wooden flats, 32 cm X 40 cm) randomly among the AEM containers to minimize micro-climate differences between the two studies. I took soil samples from the flats several days before bag placement to determine soil moisture content (% by weight). I 2 0 4 watered daily to maintain soil moisture constant but did not weigh these flats as they exceeded the balance's capacity. I removed one bag from each flat on the same day that AEMs were removed, weighed, oven-dried (70° C), reweighed to determine mass on dry-weight basis, and removed material from bags for grinding and subsequent chemical analyses.

Methods for Chemical Analyses

I determined residue P concentration on oven-dried (70° C) ground (Wiley mill mesh size 40) following the nitric-perchloric acid digestion procedure (Briceno and Pacheco, 1984). I determined C and N concentrations via dry combustion-reduction, using the Carlo-Erba Automated C-N Analyzer (model 1500, series 2). Lignin content was measured as acid-digestable lignin (ADL) according to Goering and van Soest (1970).

Statistical Analysis

I determined the best-fitting models for both mass and nutrient loss over time using a curve-fitting program (Taylor, 1978). The program fitted each dataset (mass, P, N and C loss) to a set of functions or models, iteratively, and calculated minimal residual MSE for each. The function with the lowest residual MSE was considered the best-fitting model. The models used included: 1) Y = exp (a + b/x) or x_1 2) Y = exp(a + blogx) or x^ 205 3) Y = exp (a + bx or x ^ 4) Y = exp (a + bx) or x** 5) Y = exp (a +bx^) or x^ 6) Y = e(a + bxc) or 3 parameter model Model 2 corresponds to a logarithmic relationship, model 3 a square root, model 4 the negative exponential (the model most widely used to describe litter decomposition processes which assumes a constant decomposition rate), model 5 the physical diffusion model (density dependent rate constant) and model 6, the three parameter model, which assumes that the rate of decomposition decreases with the remaining fraction in a non-linear fashion. The program also generated estimates of parameter values (a, b and c) which were used to assess decomposition rates and compare treatments. I also performed a separate regression analysis using the conventional natural log (% mass or % nutrient remaining) against time to determine rate constants (k) and to compare with the curve-fitting output.

Field Decomposition Study and Labile Soil P Dynamics Beneath Decomposing Residues

Although the greenhouse decomposition study provided an environment in which treatment effects could be evaluated with reduced variation from edaphic factors (soil moisture, ambient temperature), it was extremely limited in its ability to assess biomass and nutrient loss dynamics similar to field conditions. In addition, the experimental design limitations which resulted in separating the litter bag and AEM-P components precluded any direct correlations between residue P loss and labile soil P fluxes. Accordingly, I conducted a decomposition study in the field to: 1) determine 2 0 6 decomposition rates of Erythrina leaves, pasture grass clippings and cattle dung in the context of variable field conditions; and 2) relate residue P loss characteristics to micro-site changes in labile soil P (soil directly under decomposing residues). I chose the highest application rate from the greenhouse experiment, 800 kg DM/ha, as the sole amount applied for all three residues. I then calculated the initial dry weight per cm2 of litter bag needed based on the amount needed for chemical analyses and an estimated 60% loss after three months (determined to be 8 g dry weight per bag). I constructed litter bags as descibed by Vilas Boas (1990). The bag dimensions were 12 cm X 20 cm with an internal surface area of 225 cm2. The upper side of the bag was fabricated from black polyethylene mesh with 8 mm X 8 mm openings. The bag underside (in contact with soil) was made from fiber glass mesh with 1 mm X 1 mm openings. The two mesh types were joined together with staples. The large upper mesh was chosen to permit free entry of most decomposer organisms, while the fine mesh on the underside was used to prevent particle loss (Anderson and Ingram, 1989). I collected litter material from farm 1 (the same farm where the bags were placed) in the same manner as described for the greenhouse study and weighed residues into bags in the field. I then placed three bags per residue type in contact with the soil surface (in the pasture) in the no trees no grazing control treatment of farm 1 (total of 135 bags: 9 bags X 15 collection dates, in 300 m2 area). Concurrently, I placed three Ionics® AEM squares (6.25 cm2) vertically beneath each litter bag, 0-2.5 cm below the soil surface. 20 7 I later collected litter bags and their corresponding anion exchange membranes from the field after 1, 2f 4, 7, 14, 21, 28, 35, 42, 49, 56, 63, 70, 77 and 120 d. I also monitored soil moisture (using ceramic cup tensiometers) and soil temperature in the 300 m2 pasture over the duration of the experiment (mean of five permantly placed tensiometers and six soil thermometers). Upon harvest, I removed soil and visible insects from the litter bags and dried residues for mass loss determination. I then ground and analyzed residues for total P, N and C as described previously. I determined AEM-P as described in Chapter (I.

Statistical Analysis

I used the same curve-fitting program and models as the greenhouse decomposition data. Following a preliminary analysis of the six models listed above, I performed a second models comparison using the negative exponential model, the three parameter model, and the double exponential model (two distinct rate constants) as well as a four parameter model included for exploratory analysis only. This generated minimal MSE values, and model parameters (a-d, depending on the number of parameters in each model). Using model parameters from the three parameter model (since it gave the lowest minimal residual MSEs), I calculated expected values and plotted (on log-log scale) scaled observed versus model-estimated values against time for each residue type. Mass and nutrient values for each residue type were scaled independently using a common multiplier so that they would all fit on the same graph. 208

Since the c, or shape, parameter is a measurement of the data's

deviation from the negative exponential model (c =1 ), I quantified the

deviation between the two and three parameter models using an F-test for

improvement of fit (Fimprovement Df = n-| MSEmocjell "n2 ^'®^model2 /f

(n-|-n2 )* MSEmocjeli; F1 4 /1 3 at a 0.05). I also compared c (shape)

parameter values using 95% confidence intervals (1,96*standard error) for

mass and nutrient loss both between and within residue types. I related residue P release and labile soil P (AEM-P) both qualitatively and quantitatively. I deduced a mathematical relationship between the two by accounting for biotic and abiotic factors controlling residue P mineralization and immobilization. I tested this mathematical relationship using SYSTAT's linear regression analysis (Wilkinson, 1990). I then performed stepwise multiple linear regression using the slope (b) parameter from the residue P-AEMP regression against litter quality parameters including: C:N ratio, C:P ratio, lignin content and lignin:N ratio.

RESULTS

Greenhouse Decomposition

In general, mass loss best fit a log function regardless of residue type and placement surface (Table C.1 and Figure 4.2). Carbon loss was also logarithmic and followed the same pattern as mass loss (as expected), whereas N and P loss exhibited no consistent patterns (Table C.2).

Phosphorus loss in dung, pasture grass and Erythrina leaf and stem residues w as minimal over the eight weeks (Figure 4.3). However, Erythrina leaf 0.7 2 0 9

I f

W«M WOO*

0.74

0.7 0.7*

.70

WNt

Figure 4.2. Mass loss over time (expressed as a log function) for Erythrina leaves (a); Erythrina leaves and stems (b); pasture grass (c) and cattle dung (d) in the greenhouse study. Trttl Faiiwa |riii p («f) ^ *•** ^ W J1 J w - « is ate ug d i te rehue study. greenhouse the in (d) dung cattle Erythrina Figure 4 .3 . Residue P loss over tim e (no specific functions fitted) for for fitted) functions specific (no e tim over loss P Residue . .3 4 Figure leaves (a); (a); leaves WMk W**h Erythrina * leaves and stem s (b); pasture g rass

of % mass and % P remaining to determine rate constants, k (Table 4.1). For all residues except pasture grass, there was no significant difference in mass loss rate between the two placement surfaces, sod and bare soil. Pasture grass residues on sod decomposed significantly faster than on bare soil. Coincident with pasture grass mass loss, pasture grass P loss rate on sod was also significantly faster than on soil. In addition, it was the only residue type with a significant negative exponential regression for P loss, confirming the curve-fitting results lack of consistency among models. Residue decomposition rates within a given placement surface displayed significant differences as well. On sod, dung decomposed significantly slower than all plant residues. On bare soil, the trend was similar except that Erythrina leaves decomposed the most rapidly followed by leaves and stems and pasture grass (not different from each other) and dung the slowest.

Concurrent Labile Soil P Dynamics

Univariate analysis of variance (including time as an independent variable) revealed significant effects on labile soil P (measured as AEM-P) from residue type and placement surface as well as a significant interaction 2 1 2

Table 4.1. Comparison of mass loss (and P loss S) rate constants (k as

Residue Sod Bare Soil Type k r2 k X2

Dung -0.025 a 0.824 -0.017 a 0.755 b c Erythrina leaves -0.055 a 0.720 -0 .0 6 4 a 0 .9 1 0 a a Erythrina leaves & stems -0.043 a 0.865 -0 .0 4 4 a 0 .7 9 3 a b Pasture grass (mass) -0.067 a 0.774 -0.039 b 0.807 a b Pasture grass (P) -0.034 a 0.580 -0.017 b 0 ,5 1 4 ab c against time. $ Other residues' P regressions not included because they were not significant. Letters beside k constant are for comparing between sod and soil placement; letters below are for comparing among residue types within each placement surface. Significant differences as 95% confidence intervals (± 1.96*SE). 213 between the two (Table 4.2). Moreover, the trend was consistent among the three residue application rates. These effects were corroborated in the repeated measures univariate ANOVA which, in reality, differs little from the straightforward factorial ANOVA exept for the temporal component (Table 4.3). Labile soil P (as AEM-P) w as not affected by time, regardless of application rate, suggesting that P fluxes overall did not vary over time (Table 4.4). In addition, there were no signficant effects from any of the experiment's main factors on AEM-P fluxes over time. When soil P fluxes were compared qualitatively to residue P loss from the litter bag component of the study, it was not surprising that there was no signficant fluctuation in labile soil P since residue P loss w as minimal over the eight w eek period. Despite the lack of significant fluctuations in labile soil P overall, there were noteworthy distinctions in P flux patterns among residue types and between placement surfaces. When compared to the sod no residue control, both Erythrina leaves and pasture grass residues on sod exhibited distinct peaks in AEM-P at six weeks after residue placement (Figure 4.4).

In contrast to Erythrina leaves alone, leaves and stems combined resulted in an initial decline of AEM-P relative to the control, inferring immobilization. Dung residues produced two peaks in AEM-P: the first at one week after residue placement and a second, broader peak between weeks 4-5. Comparisons between AEM-P fluxes on sod and bare soil showed greater P dynamics on sod versus bare soil. In fact, AEM-P levels for all residues on bare soil regardless of application rate were barely detectable (Figure 4.5). The difference was apparent in the no residue control as well, intimating the presence of sod-induced soil solution P enhancement 214

Table 4.2. Analysis of variance table for the effects of residue type, placement surface and time on AEM-P under greenhouse conditions (with the no residue control removed). ______

Source df F-ratio P value Residue type (R) 3 3.77 0 .0110 Placement (P) 1 18.65 0.0001 Time (T) 8 1.72 0 .0 9 2 0 R X P 3 4.19 0 .0060 R X T 24 1.39 0 .1050 P X T 8 1.72 0 .0920 R X P X T 24 1.37 0 .1190 Error 360

Table 4.3. Univariate repeated measures analysis of variance table for effects of residue type and placement surface on AEM-P by residue application rate. ______

Source df F-ratio P value 400 kg /ha Residue (R) 3 3.15 0.035 Placement (P| 1 11.09 0.002 R X P 3 2.96 0.0 4 4 Error 40 600 kg/ha Residue (R) 3 4.51 0.008 Placement (P) 1 28.50 0.0001 R X P 3 2.32 0.090 Error 40 800 kg/ha Residue (R) 3 3.03 0.040 Placement (P> 1 14.02 0.001 R X P 3 3.41 0.026 Error 40 Table 4.4. Multivariate repeated measures analysis table for effects: AEM- P, residue type, placement surface and their interactions by application rate over time (greenhouse study). ______

Source df F-ratio P value 400 kg/ha Constant (AEM) 7 0.376 0.916 AEM X Resid 21 0.739 0.791 AEM X Place 7 0.381 0.913 A X R X P 21 0.848 0.659 Error 280 600 kg/ha Constant (AEM) 7 0.820 0.571 AEM X Resid 21 0.790 0.731 AEM X Place 7 1.187 0.310 A X R X P 21 0.980 0.488 Error 280 800 kg/ha Constant (AEM) 7 1.536 0.155 AEM X Resid 21 1.399 0.117 AEM X Place 7 1.609 0.133 A X R X P 21 1.348 0.144 Error 280 £ l r AEII-P (m| P/l tatrtrt tofn) 0 - # 0 . 0 . 0 0 # . 1 # Erythrina cattle dung (d) in the greenhouse study. greenhouse the in (d) dung cattle eopsn o sd rsdetpscmae t o eiu oto; for control; residue no to compared types residue sod; on decomposing Figure 4.4. Lt bile soil P fluxes (expressed as AEM-P) for residues for AEM-P) as (expressed fluxes P soil Lt bile 4.4. Figure r l tra n s - O l c r m a O - evs (a); leaves fc M W WMk M i r | n d l u # • nn (Ml M ( n rn T • Erythrina evsad tms b; atr gas c and (c) grass pasture (b); s stem and leaves £ I «.#• ■m 'Control Wnnk W mk * M # HI H # M t* r - T

- DtM# 216

ae soil. bare Figure 4 .5 . Labile P fluxes (as AEM-P) for all residues d eco m p o sin g on on g sin o p m eco d residues all for AEM-P) (as fluxes P Labile . .5 4 Figure AEM-P (mg P/L extract sol'n) 0.00 0.02 0.05 0.03 0.04 0.01 0123456789 ~o- ~o- Tr Control Control "O, Tro#— Week leaf leaf Pasture — Dung 17 21 21 8 (Figure 4.6). When residue types were viewed individually, sod versus soil comparisons further illuminated sod-induced effects on labile soil P (Figure

4.7). In all cases except the Erythrina leaves and stems where P fluctuations on sod were lower than on soil, P fluctuations and, in general, labile soil P levels, were significantly higher on sod than on bare soil. With the time factor removed, the previously noted interaction between residue type and placement surface became extremely apparent (Table 4.5). As the P flux comparisons suggested, there were no significant differences in AEM-P among residue types on bare soil, regardless of residue application rate. On sod, however, there were significant differences among residues at all application rates. In all cases, dung AEM-P was greatest, followed by Erythrina leaves alone and pasture grass; Erythrina leaves and stems AEM-P was consistently the lowest. As shown previously, Erythrina leaf and stem residue AEM-P on sod was indistinguishable from that on bare soil. The overall effect from application rate was minimal; Erythrina leaves and pasture grass AEM-Ps at 400 kg/ha application rate were significantly lower than both 600 and 800 kg/ha, while dung AEM-P did not increase with higher application rate.

Field Decomposition

As is characteristic of the Costa Rican Atlantic coastal plain, both soil temperature and soil moisture fluctuated little over the 100+ d course of field decomposition (Figures 4.8 and 4.9). Soil moisture vacillated within its usual field capacity to near saturation range, and soil surface temperature (except for one day where it dropped to a chilly 20 °C) hovered around 25- ess ae soil. bare versus Figure 4.6. Comparisons of P fluxes (as AEM-P) for no residue control, sod sod control, residue no for AEM-P) (as fluxes P of Comparisons 4.6. Figure

AEM’P (mg P/L extract sol'n) 0.00 0.02 0.04 0.03 0.06 0.05 0.01 0123456789 o soil o s ro o B d o S Week 19 21 AEM-P (mi P/L extract torn) 0.07 1 .0 0 O.Ol atr rs () n ctl ug (d). dung cattle and (c) grass pasture for AEM-P) (as soil Figure 4.7. Comparisons of P fluxes betw een residues on sod versus bare bare versus sod on residues een betw fluxes P of Comparisons 4.7. Figure l00 flj0 c f # f " J - OJM e.M 1 2 1 0

i a > i o y —r |— — k M W W m I Erythrina -••I*Ml Ml lae (a); leaves Erythrina | 5 - | IU a < CL . o £ M o 0.00 P M a - «ai 0.00 o.os o.w 0.t2 o.-m 0

lae n se (b); s stem and leaves 1

2 -••4 Waa* k n W Ml M ■ ■ Bars m I =4 0 2 2 221

Table 4.5. Greenhouse anion exchange membrane (AEM) P (mg P/L extract solution) by residue type, placement surface and application rate (means pooled over time). ______

Residue Type Sod Bare Soil § No residue control 0.019 ab 0.0 0 4 400 kg/ha Erythrina leaves 0.014 b 0.0 0 4 Leaves & stems 0.013 b 0.001 Pasture grass 0.014 b 0.004 Dung 0.046 a 0.003 600 kg/ha Erythrina leaves 0.024 b 0.002 Leaves & stems 0.010 c 0.003 Pasture grass 0.027 ab 0.005 Dung 0.048 a 0.010 800 kg/ha Erythrina leaves 0.031 b 0.004 Leaves & stems 0.006 c 0.007 Pasture grass 0.025 b 0.003 Dung 0.061 a 0.004 S No significant differences among residues for all application rates for bare soil. Values with different letters within each application rate are significantly different (95% confidence interval or ± 1.96*SE). All residues on sod are significantly different from those on bare soil, except for Erythrina leaves & stems. In addition, AEM-Ps at lowest application rate for Erythrina leaves and pasture grass are significantly different from the two highest application rates. eopsto i tefed n=5 o ahsmln date). sampling each for (n = 5 field the in decomposition Figure 4.8. Fluctuations in soil moisture (pF) over the course of litter bag bag litter of course the over (pF) moisture soil in Fluctuations 4.8. Figure pF (log soil moisture suction) 2.0 2.2 1.0 1.2 2.4 1.4 2.6 3.0 1.6 1.8 0 20 as n h field the in Days 40 SD 60 80 100 2 2 2 Figure 4.9. Fluctuations in soil temperature over the course of litter bag bag litter of date). course sampling the each over for In = 6 field the intemperature soil in Fluctuations decomposition 4.9. Figure Soil Temperature (oC) 223 22 4 35 °C. As such, the contribution to decomposition dynamics from environmental factors was probably minor. The iterative curve fitting of mass and nutrient loss showed that the 3-parameter model was the best-fitting model in most cases (Table C.3). The 4-parameter model did not improve the fit to any appreciable extent. The double exponential model was performed separately and did not improve the fit as well; therefore it was not included in the table. The 3- parameter model assumes a non-constant, non-linear decomposition rate; i.e., the rate decreases with the remaining fraction in a non-linear fashion. Accordingly, it also describes the structure of the decomposition pattern on a temporal scale. For all residue types (Erythrina leaves, pasture grass clippings and dung), observed values for mass, N and C loss paralled closely those predicted from the 3-parameter model parameters (Figures 4.10-4.12; Table C.4). Observed P loss, however, deviated slightly from the model curve. Because the model's shape (c) parameter was not equal to 1 (would revert to the 2-parameter , negative exponental model if c = 1), it could be used to compare data structure either among residue types or relative to a fixed set of values. The more different the shape parameter from 1, the less likely the constant decomposition rate (negative exponential) was adequate to describe observed mass and nutrient loss dynamics. I compared c parameter values relative to the case where c = 1 or the negative exponential model (Table 4.6) and among residue types (Table 4.7).

Erythrina mass, N and C loss as well as dung P and C loss and pasture C loss exhibited significant improvements over the negative exponential model when fit to the 3-parameter model, implying that those loss rates were Mass Loss (g or mg) 10 .1 values are scaled to fit on the same graph). same the on fit to scaled are values Figure 4.10, Mass, P, N and C loss from decomposing decomposing from loss C and N P, Mass, 4.10, Figure itd ote -aaee xoeta oe. Nt: asad nutrient and mass (Note: model. exponential 3-parameter the to fitted * Moss ■ Phosphorus i Nitrogen ♦ Corbon ♦ Nitrogen i Phosphorus ■ *Moss ie (Days) Time Erythrins leaves

1000 225 Mass Loss (g or mg) to the 3-parameter exponential model. (Note: mass and nutrient values are values nutrient fitted graph). and grass same mass (Note: the pasture on fit model. to decomposing exponential scaled from loss C and 3-parameter N P, the to Mass, 4,11. Figure o s hshrs Ntoe ♦ Carbon ♦ Nitrogen i Phosphorus ■ Moss —» (— ie [Days] Time 10

100 226 Mass Loss (g or mg) h 3prmtrepnnil oe. Nt: as n ntin aus are values nutrient and mass (Note: model. exponential 3-parameter the iue .2 Ms, , n C osfo eopsn atedn fte to graph). fitted same dung the cattle on fit to decomposing scaled from loss C and N P, Mass, 4.12. Figure Ms ■ hshrs Ntoe ♦ Carbon ♦ Nitrogen • Phosphorus ■ Moss* ie (Days] Time a i

1000 227 Figure 4.6. Test (F-parameter) for improvement of fit of the 3-parameter exponential model over the 2-parameter negative exponential model (field decomposition rates). ______

Residue Type Mass PN C Dung 1.68 5.07* 1.27 2.92* Erythrina 3.69* -0.97 4.62* 2.70* Pasture 1.93 -0.74 1.89 3.09* * significant improvement over the 2-parameter negative exponential model; a = 0.05.

Table 4.7. Comparison of shape (c) parameter values between and within (mass and nutrients) residue types from the 3-parameter exponential decomposition model fitted to the field decomposition data. ______

Residue Shape (c) Parameter From 3-Parameter Type ______Model Mass P N C Dung 1.283 a 0.441 b 1.446 a 1.289 a a b a a Erythrina 0.754 b 0.975 a 0.648 b 0.779 b a a a a Pasture 1.400 a 1.070 a 1.847 a 1.468 a a b a a Significant differences represent 95% confidence intervals. Letters beside numbers are for comparing c parameters between residue types. Letters below numbers are for within residue type comparisons among mass and nutrients. 229

density dependent (Table 4.6). However, Eyrthrina and pasture P loss characteristics were not signficantly different from the negative exponential model, inferring a constant release of P from those decomposing residues. The difference between dung and plant residues was probably the initial large drop in dung P (likely soluble inorganic P) followed by a much slower release pattern. According to Ezcurra and Becerra (1987), the shape (c) parameter can also be used as an indirect measure of labile:recalcitrant content in decomposing residues. If residues are rich in relatively labile compounds, then c < 1 and the decomposition rate will not decrease much until the more recalcitrant fractions dominate. If residues have relatively more recalcitrant compounds, then c > 1 and the initial rate b (the rate constant also known as k) will decrease more sharply as the first fractions decompose. Dung P decay, for example, was significantly less than 1 confirming its high soluble P content relative to both Erythrina leaves and pasture grass (Table 4.7). Erythrina mass, C and N shape (c) parameters were also less than 1; while those for dung and pasture grass were greater than 1; suggesting that Erythrina leaves have less recalcitrant compounds than pasture grass and dung. I performed stepwise multiple regression of c parameters against litter quality indices (initial C:N, C:P, lignin:N ratios and lignin content) to investigate the relationship between the shape parameter and initial litter quality (Tables 4.8 and 4.9; Table C.4). There were no significant relationships between mass loss dynamics and initial litter quality (N and C produced similar non significant regressions; therefore they weren't included). The same held for residue P loss, except there was a borderline Table 4.8. F-statistics of stepwise multiple regression between the field decomposition model shape (c) parameter for mass and residue C:N, C:P, Lignin:N and % Lignin. ______

Independent Variable F-statistic P value C:N ratio 5.615 0.254 C:P ratio 0.110 0.796 Lignin:N ratio 0.911 0.515 Lignin content (%) 0.230 0.715

Table 4.9. F-statistics of stepwise multiple regression between the field decomposition model shape (c) parameter for residue P and residue C:N, C:P, Lignin:N and % Lignin.

Independent Variable F-statistic P value C:N ratio 0.040 0.8 7 4 C:P ratio 70.144 0.0 7 6 LigninrN ratio 2.532 0.357 Lignin content (%) 1.744 0.413

Table 4.10. F-statistics of stepwise multiple regression between the AEM- P-2nd derivative residue P regression rate (b) parameter and residue C:N, C:P, Lignin:N and % Lignin. ______

Independent Variable F-statistic P value regression r^ C:N ratio 0.011 0.933 0.011 C:P ratio 1546.43 0 .016 0.999 Lignin:N ratio 3.921 0.298 0.797 Lignin content (%) 1.192 0.472 0.544 231 significant regression between the shape parameter and the C:P ratio (Figure 4.13). When the c parameter was close to 1 (decomposition rate constant), the C:P ratio was in excess of 200. The paucity of points used to generate the regression preclude any definitive statements about the relationship. Speculatively, since C:P ratios greater than 200:1 usually result in microbial immobilization, the negative exponential model may adequately describe decomposition when immobilization is the dominant pathway; whereas the 3-parameter model best describes decomposition when mineralization is dominant, at least in the initial stages.

Residue P loss and Labile Soil P Fluxes in the Field

Upon qualitative inspection, labile soil P (AEM-P) pulses appeared to be related to patterns of residue P release; however the exact relationship was not readily discernible (Figures 4.14-4.16). Dung P, for example, exhibited an initial rapid release within the first 4-7 d after residue placement (Figure 4.14), Remaining P continued to decline at a fairly rapid rate until approximately 25 d after residue placement and remained relative unchanged for another 50 days. Concurrent measurements of labile soil P (as AEM-P) exhibited a small peak initially followed by a large, broad peak coincident with the sharpest decrease in residue P. Erythrina leaf residue P loss, in contrast, exhibited an initial increase in residue P (immobilization) followed by an increase in the release rate at approximately 20 d from initial residue placement (Figure 4.15). Subsequent P release followed an acceleration-deceleration pattern with AEM-P peaks occurring immediately after the sharpest increases in residue P release. Pasture grass residue P Figure 4.13. Regression between shape (cl param eter in the 3-param eter eter 3-param the in eter param (cl shape residue between and Regression model 4.13. Figure

C C Parameter (Decomposition model) 0.2 0.3 0.4 0.5 0.6 0.7 0.6 0.9 1.0 50 C:P - 014 004 r2 0.986 - r‘2 0.004x + 0.174 - y 100 ratio. : Rto n Residue RatioC:P in 150 0 20 300 250 200 232 Erythrina beneath litter bags. Note: AEM-P scale is approximately 5X th o se for for se o th 5X AEM-P and approximately time is over loss P scale AEM-P dung Note: een betw bags. litter Relationship beneath . 4 .1 4 Figure

Residue P (mg) remaining 20 30 10 40 0 5 0 a atr grass. pasture d an 25 0 as n h field the in Days — ■—mg P dung dung P —■—mg -— AMP dung --O— AEM-P 50 5 7 100 125 5 - - 15 - - 10 20

AEM-P (mg P/L extract soPn) 3 3 2 E- udret lte bags. litter between underneath Relationship AEM-P 4.15. Figure Residue P (mg) remaining 10 12 14 16 18 2 4 6 8 0 Q E- Erythrina Erythrina P AEM-P ”Q- g -m -■ 25 as n h field the in Days 50

Erythrina 75 leaf P loss over time and and time over loss P leaf 100 125 - 3 - - r 4 - 5 - 2 6

AEM-P (mg P/L extract sol'n) 4 3 2 AEM-Plitterbags. underneath Figure 4.16. Relationship between pasture residue P loss over time and time over Ploss residue pasture Relationshipbetween Figure4.16. Residue P (mg) remaining 5 2 20 10 15 0 5 0 g pasture P mg E- pasture AEM-P 25 as n h field the in Days

75 100 125 0.0 1.0 2.0 3.0 .0 4

AEM-P (mg P/L extract sol'n) 235 236

and AEM-P followed similar patterns to those for Erythrina leaves, but the magnitude of the labile soil P pulses was lowest among the three residue types (Figure 4.16). Considering the biotic and abiotic factors controlling immobilization and mineralization rates of residue P, I deduced the following scenario. The easily solubilized P compounds release P at fairly rapid rates. If the total amount of P in this pool is sufficiently large, it should satisfy microbial demands and result in net P release. This pulse should be detected in the soil solution temporarily (as in the case of the rapid initial peak from dung). Excess P would also result in increased microbial growth. As P is released from the residue and more microbes incorporate P into their biomass, the microbial population comes into equilibrium with its P supply and the C:P ratio widens. Residue P becomes scarce, and microbial activity results in net P immobilization; the overall result is that the rate of P released decelerates and eventually flattens out. The microbes continue to assimilate C at a faster rate than P, and population growth will bottom out (will also get microbial death); the coupling of these phenomena should result in a narrowing of the C:P ratio and an acceleration in the rate at which P is released from the residue. Eventually, P mineralization will dominate and the microbial population will have lagged behind this acceleration in P release: result is another P pulse in the soil solution. Therefore, it was the acceleration and deceleration of the residue P loss rate or the second derivative which appeared to regulate soil solution P pulses. The regression between the second derivative of residue P loss and labile soil P (as AEM-P) proved significant for all three residue types; however, the degree of linearity of the relationship differed among the three 237 {Figures 4.17-4.19). In general, as the second derivative became smaller, the rate of P released decelerated and less P was mineralized into the soil solution. As the second derivative increased, P release accelerated and more P w as detected in the soil solution. To test the hypothesis that litter quality was regulating changes in the relationship between residue P loss and labile soil P, I regressed the

regression coefficient (the slope parameter from the 2 nd derivative of residue P - AEM-P regression) against the various litter quality indices (C:N,

C:P, LignimN. % Lignin: Table 4.10). There was, in fact, a significant inverse linear relationship between the C:P ratio and the regression coefficient b, although there were too few data points to make the relationship conclusive (Figure 4.20). When the C:P ratio was high, the slope was close to zero, indicating either deceleration or no change in the rate of P loss. When the C:P ratio was low, in the case of dung, the slope was steep, insinuating an acceleration in the residue release rate.

DISCUSSION

Comparisons between the Greenhouse and the Field

Differences in decomposition rates and best-fitting models between the greenhouse and the field are not unusual given the differences in microclimate and soil faunal activity (Hopkins et al, 1988; Gonzalez and Sauerbeck, 1982). It is very likely that the suite of larger decomposers responsible for initial break-down of large residue pieces was absent or greatly reduced in the greenhouse. The smaller bag size used in the fdn P. dung of Figure 4.17. Linear relationship between AEM-P and the second derivative derivative second the and AEM-P between relationship Linear 4.17. Figure

AEM*P (mg P/L extract sol'n) 0.0 2.0 3.0 4.0 1.0 5.0 6.0 7.0 •0.06 sde (g 2d derivative 2nd (mg) P esidue R - 212 607 r2 0.454 - r~2 6.067* + 2.162 - y 0.340.14 0.54 238 of of Figure 4.18. Linear relationship between AEM-P and the second derivative derivative second the and AEM-P between relationship Linear 4.18. Figure Erythrina

AEM-P (mg P/L extract sol'n) -0.5 0.0 0.5 2.0 1.0 2.5 3.0 1.5 la P. leaf - 0.10 sde (g 2d rvai e ativ eriv d 2nd (mg) P esidue R - 036 1.272 + 0.336 - y 0.30 k ‘ - 0.313 - r‘2 0.50 0.70 9 3 2 fpsue rs P. grass pasture of Figure 4.19. Linear relationship between AEM-P and the second derivative derivative second the and AEM-P between relationship Linear 4.19. Figure AEM-P (mg P/L extract sol'n) -0.5 0.0 0.5 10 02 00 02 04 0.60 0.40 0.20 0.00 -0.20 eiu P m) n derivative 2nd (mg) P Residue - 003 097 r' 0.600 - '2 r 0.957X + 0.043 - y 40 24 Figure 4.20. Regression between the rate constant (from the AEM-P-2nd AEM-P-2nd ratio. C;P the (from residue the constant and rate relationship) P the residue between derivative Regression 4.20. Figure Rate constant 10 0 2 4 3 7 9 5 6 8 1 25 - 8. 002 ‘ - 0. 9 9 .9 0 - r‘2 0.032X • 3 5 .3 8 - y : Rto n Residue in Ratio C:P 175125 225

275 241 242 greenhouse study may also have impeded decomposer access to internal residue reserves and may have created an O 2 -limited environment. It is not uncommon, therefore, to have lower rates of mass and nutrient losses in the greenhouse as compared to the field; and as such decomposition rates obtained in the greenhouse study should not be used to predict processes in the field (Gonzalez and Sauerbeck, 1982). Nonetheless, this was not the objective of the greenhouse study; it was designed to compare relative rates among residue types and to evaluate the effect of placement surface on decomposition.

Placement Surface Effects

Enhanced labile soil P dynamics (and to a lesser extent decomposition rates) under sod is, on the one hand, contrary to the hypothesized increase in plant P demand-uptake resulting in decreased levels of soil solution P in the rhizosphere (Blair and Boland, 1978). On the other hand, numerous studies confirm the existence of decreased soil P sorption in the presence of live roots or their exudates, particularly in high P-fixing soils (Fox and Comerford, 1992; 1990; Lopez-Hernandez et al., 1986; 1984; Violante et al., 1991; Kafkafi et al., 1988; Traina et al., 1986). Low molecular weight organic acids are known to complex with free Al and Fe that would otherwise form poorly crystallized precipitates with orthophosphate anions. Roots are also known to acidify the rhizosphere which, within a certain range, decreases P sorption (Gillespie and Pope, 1990). Roots also release phosphohydrolase enzymes which break phosphate ester bonds, thereby releasing organically-bound P forms independent of C mineralization (Smeck, 1985). 2 4 3 Given the Neguev soil's high P retention capacity, the rhizosphere provides a more favorable micro-environment for maintaining P in soil solution. An alternative scenario is that, on a volume basis, there was less mineral soil present in the surface soil under sod. Accordingly, there would be less competition from the mineral soil constituents for scarce P. Nonetheless, it does not exclude the possibility that organic compounds (either biotically generated or evolved from the soil organic matter) reduce soil P sorption thereby enhancing P bioavailability. Decomposition rates would, in turn, relate to enhanced P bioavailability because microbes could draw on soil P pools for their nutrient needs thus reducing the potential for immobilization of residue P into microbial biomass (van Veen et al., 1987; Singh and Jones, 1976).

Best-fitting Decomposition Models

Most decomposition studies evoke the constant-rate negative exponential model to describe mass and nutrient loss dynamics (Olson, 1963; Weider and Lang, 1982; Palm, 1988; Hueveldop et al., 1985; Vilas Boas, 1990). Alternatively, they have proposed the double exponential model which promotes the existence of two distinct pools, each with different but constant decay rates (Hargrove et al., 1991). It seems rather simplistic to assume that organic matter can be divided into two unique fractions each with its own stable decomposition rate (Ezcurra and Becerra, 1987; Andren and Paustian, 1987). Moreover, there are no chemically operative guidelines along which such fractions could be separated. Most likely, organic matter can be characterized as a continuum of degradable 24 4 compounds; decomposition proceeds in a cascade-like fashion such that new, more recalcitrant (either spatially or chemically) compounds are formed over time (Andren and Paustian, 1987). The 3-parameter model best describes this concept because the shape parameter confers a non-linear change reflective of an evolving resource quality. For extremely labile constituents in litter, the 3-parameter model is also a better predictor of mass and nutrient loss because it captures the behavior of the instantaneous fractional loss rate (Schlesinger and Hasey, 1981). Usually, the initial loss rate of soluble compounds is so rapid that a constant decay rate does not describe it adequately. This was certainly true for the dung P loss pattern and the Erythrina mass loss.

Relationship between Residue P Loss and Labile Soil P Fluctuations

The integral relationship between residue quality, microbial population dynamics and nutrient release from decomposing residues has been demonstrated both directly and indirectly (Parmelee et al., 1989; Polglase et al., 1992; Beare et al., 1989; Budelman, 1988). Residues with higher initial nutrient concentrations and lower carbon:element ratios have resulted in greater decomposition rates and greater net mineralization. Such residues were also found to support larger and more diverse decomposer organisms, including bacteria, fungi and nematodes (Parmelee et al., 1989). Polglase et al. (1992) suggest that many plants have evolved strategies for accumulating soluble P compounds in above ground tissues as means to minimize microbial immobilization when plant residues decompose. In this 245 way, mineralization returns nutrients relatively rapidly to the bioavailable pool for subsequent plant reutilization.

Studies have confirmed the lability of Erythrina residues relative to other leguminous tree mulches (Palm, 1988; Hueveldop et al., 1985; Vilas Boas, 1990). It has low lignin and poly phenolic contents coupled with high total N; both favorable quality indices in terms of degradability. In addition, dung decomposition has been shown to increase available soil P in the surface soil layer (0-2.5 cm) after the first week of deposition followed by decreases due to mineral soil P retention (Omaliko, 1984; Buschbacher, 1987). These findings corroborate the ephemeral nature of P fluxes. Microbially-mediated P release appears to follow a time-lagged feedback loop in which the microbial population dynamics (the equlibrium between population growth and P substrate availability) are always one step behind the changes in the residue P loss rate. The faster the P loss rate changes, the less likely the microbes are to be in equilibrium with their P supply which, in turn, lessens the P demand from biotic constituents. Singh and Jones (1976) and van Veen et al. (1987) have confirmed that the quality of residues which satisfy microbial P demand often dictates labile soil P fluctuations. Singh and Jones (1976) for example, cite a 30 d peak in available soil P following residue placement and attributed it to microbes using only limited amounts of total P reserves to fuel their own growth. Subsequent declines in available P were attributed to microbial immobilization of an increasingly scarce P supply (from both residues and soil organic P pools). In summary, then, nutrient turnover patterns and subsequent pulses in available soil forms are strongly influenced by the 2 4 6 degree of nutrient immobilization and how it changes over time (Chauhan et al., 1981; Blair and Boland, 1978).

CONCLUSIONS

1. In this high P-retaining soil, there is a strong rhizosphere-induced enhancement of soil solution P levels. Without the presence of live roots, any P mineralized from decomposing residues is removed quickly from the soil solution (or bioavailable pool) by soil retention (both biotic and abiotic). Plant roots maintain P in bioavailable form; this may be due to either exudation of organic substances (acids or enzymes) or volumetric displacement of mineral component by the organic component (thereby minimizing the effect from mineral soil).

2. Overall, Erythrina leaf residues decompose faster than both pasture grass clippings and cattle dung. The P release characteristics for both

Erythrina and pasture grass follow negative exponential decay functions, whereas dung P is released as an initial large pulse (soluble inorganic P) followed by more gradual mineralization of organically-bound P or dissolution of inorganic P solids. The overall magnitude of dung P released is about 4-5 times greater than that of both Erythrina and pasture grass.

3. The model which best describes residue mass loss under field conditions is a three-parameter exponential decay function. This model assumes that 247 the decomposition rate is not constant and, in fact, varies with the remaining residue quality (i.e., elemental and lignin ratios).

4. The relationship between residue P release patterns and subsequent changes in labile soil P is a function of the acceleration-deceleration of the rate of P release. As the rate of P release accelerates, there is probably an ephemeral shift in the equilibrium between microbes and their substrate, resulting in a temporary excess of mineralized P. This solubilized P remains in soil solution until plant uptake or subsequent microbial reutilization remove it. The time-lagged feedback loop between microbial P immobilization and mineralization of decomposing residues creates a brief window of opportunity for plant P uptake. CHAPTER V

LABILE SOIL PHOSPHORUS FLUXES AS A FUNCTION OF TREE PRUNING AND GRAZING IN A HUMID TROPICAL SILVOPASTORAL SYSTEM

INTRODUCTION

Agroforestry systems, which incorporate fast-growing trees with annual or perennial crops and/or pasture, are considered possible alternatives to current, non-sustainable land use practices in the humid tropics (Sanchez, 1987; Benites, 1990). Such systems, particularly those utilizing nitrogen-fixing tree species, have the potential to improve soil chemical, physical and biological properties through enhanced nutrient cycling, improved soil structure and increased soil organic matter levels (Huxley, 1987, Nair, 1984; Young, 1986).

Many of the hypotheses generated about nutrient dynamics in agroforestry systems are based on analogs from comparatively undisturbed ecosystems and fallows (Jordan, 1985; Kang and Wilson, 1987: Anderson, 1986; Ewel, 1986). Usually, soils of the humid tropics have inherently low nutrient exchange and buffering capacities. Tropical forests, as such, have evolved biological nutrient-conserving mechanisms which, in essence, bypass the soil phase. They include: large root biomass, concentration of roots in the soil surface, associations with symbiotic micro-organisms

2 4 8 2 4 9 (mycorrhizae and N-fixing microorganisms) and low nutrient concentrations in litter (Jordan, 1985; Vitousek, 1984; Vitousek and Sanford, 1986). Sollins and Radulovich (1988) also report that well-aggregated highly weathered soils under forests exhibit preferential water flow along decayed root channels, animal burrows and other biologically-created macropores. Rapid flow through these macropores bypasses fine pore spaces where most nutrients are held, thus preventing nutrient loss via leaching. Resultant biologically-mediated nutrient cycles are highly efficient, with almost no loss from either above- or below-ground ecosystem compartments. Fallow system nutrient cycles, although somewhat less "closed" than forest cycles, do conserve nutrients by storage in above­ ground biomass coupled with high litter production and consequent build-up of soil organic matter (Kang and Wilson, 1987; Swamy and Ramakrishnan, 1987). Soil phosphorus cycling, unlike other major nutrients like carbon, nitrogen and sulfur, involves equilibrium reactions among both organic and inorganic constituents (Duxbury et al,, 1989). The relative strengths of interacting P sources and sinks depend upon soil mineralogy, degree of weathering, P content of organic inputs and activity of soil organic fractions (Figure 5.1). Labile P, which regulates soil solution P and plant P uptake, fluctuates as a function of complex physio-chemical and biochemical mechanisms including: plant root depletion, leaching of water-soluble forms from standing dead biomass and litter, immobilization and mineralization from organic pools and adsorption and desorption from inorganic pools (Cole et al., 1977; Tate, 1984; Sanyal and De Datta, 1991). 2 5 0

Structural P (C/P-SOO) Actlvo Slow Posshre •Oil P COlIP »- coll P (C/P"S0to200l Milobdle P (c/p-sotoao: (C/P«20 to200) IC/P-SOIolSO) Ul) K*Mi Plant Lablt* Secondary P Occluded P residue P P K*Ui

I • InunobiTtxollon M - Mineralization • m K i, K* ,K j ,tC4 • Con at ant* Mi ■ Combined moisture Prim ary P temperature factor

Figure 5.1. Flow diagram of sources, sinks and processes regulating soil P dynamics (from: Sanyal and De Datta, 1991). 251 The contribution to P cycling from organic P pools becomes substantial when a large percentage of the total soil P is organic (eg., grasslands) and when soils are dominated by high P-retaining minerals like allophane and poorly crystallized sesquioxides (Bowman and Cole, 1978; Vitousek and Denslow, 1986; Stewart and Tiessen, 1987). The availability of inorganic P and organic substrates regulates microbial growth which, in turn, determines immobilization and mineralization rates. The ultimate limit on organic P availability is mineralization rate rather than the amount of organic P present (Tate, 1984). Mineralization rates are controlled by the interactions among soil microbes, plant P demands and P complexation with the soil matrix. Formation of stable P-mineral complexes (both organic or

inorganic) can prevent enzymatic hydrolysis of R-OPO 3 functional groups and retard mineralization (Harrison, 1982). Ecosystem conversion to agriculture disrupts internal nutrient cycling and alters the pools and processes involved; in most cases, land clearing results in increased mineralization and nutrient loss via leaching, runoff and/or volatilization (Adepetu and Corey, 1977; Duxbury et al., 1989). Phosphorus loss occurs via relatively rapid mineralization of organic P reserves (Mueller-Harvey et al., 1985) and subsequent occlusion by the soil matrix (Parfitt et al., 1989; Vitousek and Denslow, 1986; Stewart and Tiessen, 1987). In either case, the P flush associated with land clearing is brief and available P declines sharply unless P is converted to new organic forms (Serrao et al. 1978). When tropical forests are converted to pastures, P cycling is strongly affected by livestock grazing (Buschbacher, 1987; Mott, 1975; Tate, 1984). The residue pool generated from dead plant material (thatch) along with 252 excreted material which grazing animals deposit become important components in the P transport system. Cattle are particularly important in pasture nutrient cycling because they convert nutrients from relatively dispersed and unavailable forms (grass) to concentrated and available forms (dung and urine). The overall grazing effect, then, is localized enhancement of nutrient turnover. Buschbacher (1987), for example, found that dung had very high P content relative to Brachiaria grass, yet cattle concentrated over 50% of their dung on only 30% of the pasture area. The high P solubility and patchy distribution have the potential to benefit pasture plant P uptake. Jackson et al. (1990) demonstrated that plants adapted to low P soils could increase P uptake kinetics by as much as 80% when roots encountered nutrient-rich patches. In fact, the degree to which roots responded to P-rich patches (plasticity in P uptake) was correlated with low soil P status, suggesting that plants have evolved mechanisms to take advantage of highly ephemeral available P pools. In a similar study, Caldwell et al. (1987) (evoking interspecific plant competition) showed that plants can shift their nutrient acquisition abilities following system perturbations, such as grazing or herbivory, to exploit localized P flushes. Just as plant roots can exploit animal residue-deposited P pools, they can also utilize plant litter-generated P pulses. A key hypothesis related to nutrient cycling in agroforestry systems is that the tree component in the system accumulates limiting nutrients in its own biomass and that these nutrients are redistributed to companion plants as litter decomposes (Young, 1989; van Noordwijk and Dommergues, 1990). Specifically, trees are hypothesized to: 1 ) mine nutrients from weathering minerals in subsurface horizons; and 2 ) trap nutrients from the soil solution that would 25 3 otherwise be lost by leaching and recycle them through litter decomposition to the soil surface. Although P loss via leaching is minimal, trees can convert inorganic P to organic P forms which have a lower potential for soil matrix retention. Cavigelli and Thien (1989) demonstrated that lupine

[Luptnus spp.), an herbaceous legume, accumulated more P in its roots than grasses, and depleted available P pools (Bray-1 extractable soil P) less than crops like sorghum and wheat. Their findings suggest that legumes access P pools other than those most readily available and, in turn, serve as P bioaccumulators. Stewart and Tiessen (1987) and Virginia (1986) also provide evidence that deep-rooted plants (particularly leguminous, N-fixing trees) facilitate P movement from the subsoil to the surface soil. They concentrate P in the most active root zone through P-enhanced litter deposition and decomposition. Furthermore, Gillespie (1989) and Gillespie and Pope (1990) showed that, for highly weathered soils, leguminous trees (black locust) enhanced P uptake of companion plants via acidification of the rhizosphere coupled with changes in P diffusion parameters. They proposed that high rooting densities of tree-crop interplantings minimized diffusion distances, thereby allowing greater diffusion of solubiliized P to roots of non N-fixing species. Such studies confirm the potential nutrient-accumulating and nutrient redistribution capabilities of many N-fixing leguminous trees. However, many of these studies did not investigate P cycling holistically or in undistiburbed field settings;, most characterized P dynamics from either controlled pot experiments or static measurements of ecosystem components. Moreover, agroforestry and silvopastoral nutrient dynamics 254 hypotheses have not been tested rigorously under field or on-farm conditions. The benefits derived from leguminous tree-pasture associations are expected to be both long- and short-term. The tree-pasture association is anticipated to increase nutrient accumulation in the agro-ecosystem as a whole over time; i.e., create a greater nutrient reservoir than would grazed pasture alone. Equally important, leguminous trees are predicted to stimulate short-term P pulses; such pulses are believed to be crucial to pasture P uptake, particularly when soil-plant competition for labile P is intense and P availability to plants so ephemeral. Overall, nutrient dynamics are predicted to reflect a balance between short-duration high turnover from system perturbations (pruning and grazing) and long-term nutrient enrichment. The general hypothesis for this study relates to the specific case of acid, low fertility and high P-retaining soils (Figure 5.2). Within this context of limited available P, leguminous trees (specifically Erythrina berteroana) are expected to sequester P from sources not readily available (either chemically or spatially) to companion pasture species. As the trees grow, they will concentrate P in their leaf biomass. If left unpruned, the trees would solely compete with pasture grasses for scarce available P. However, if managed to maximize leaf biomass production and then pruned, accumulated

Erythrina P should be mineralized from decomposing leaf residues. This increase in mineralized P should, in turn, result in labile P (soil solution P) pulses in the upper soil stratum to which more shallow-rooted pasture grasses have access. These pulses should persist and be detected via in situ measurement because P will be converted to organic forms less likely to Mineralized P a HtPO

Figure 5.2. Hypothesized pettern of P fluxes In the sitvopsstoral system. Erythrina trees are expected to scavenge labile soil P more effclently than pasture grasses and sequester P in their leaf biomass. Upon pruning, P In leaf blomess Is released via decomposition processes Into the upper soil stratum to which pasture roots have access. Catlle also recycle P by consuming both pasture and Erythrina leaves and defecating. They alter P dynamics by concentrating and redistributing easily soluble P sources. 255 25 6 be adsorbed by the high P-fixing soil matrix. If grazing is superimposed on this scenario, labile P dynamics may be altered (temporarily dampened or heightened) due to: 1) some P export from the silvopastoral system via cattle biomass: 2 ) conversion of plant residues to animal dung which has more concentrated and localized P release characteristics; and 3) nutrient flush following grazing related to compensatory root dieback. The objectives of this study, then, are to: 1 ) monitor intensively the dynamics of labile soil P in situ as a function of tree pruning and grazing in a leguminous tree (Erythrina berteroana) - native grass pasture silvopastoral system; 2 ) determine specific effects of grazing, distance from trees and soil depth on labile soil P fluxes; and 3) assess the relationship between soil moisture fluctuations and soil solution P dynamics.

METHODS AND MATERIALS

Site Description

The field experiment was located in the Atlantic coastal plain of Costa Rica (10° N 83° W). The area receives 3630 mm rainfall annually and the ecological lifezone is lowland humid tropical rainforest. The soil is an andic humitropept (series Neguev) and is located on the ridgetops of a slightly undulating landscape. The geologic origin is volcanic ash and lahars from the late Pleistocene; due to its stable position on the landscape and high rainfall, however, it has mineralogica! characteristics of more highly weathered soils (kaolinite, gibbsite and iron oxides). In general, the Neguev 257 series is deep, well-drained, acid, low in exchangeable bases, high in exchangeable Al, high in total P but low in available P, clayey and low in bulk density (Tables 5.1 and 5.2). In addition, the Neguev soil has an extremely high P retention capacity (> 2000 mg P/kg soil) and a high moisture retention capacity over a wide range of soil moisture tensions (Figure 5.3). The field experiment, a silvopastoral system, was a 2 X 2 factorial design with cattle (grazing) and trees as the two independent variables

(Figure 5.4). I planted Erythrina berteroana (a tropical leguminous tree) from

2 . 6 m cuttings (stakes) in native grass pastures on five farms, all within a 6 km radius (all inside the Neguev settlement). All farms were located on the Neguev soil series upon which I established the randomized block of four treatments (farms were considered as experimental replicates). Grazed treatment plots were 900 and non-grazed treatments 400 m2. | planted trees one year prior to the experiment's initiation to allow sufficient time for establishment (one year is considered adequate time for establishment since trees are planted as large stakes and they root fairly rapidly). I planted trees using a 6 m X 3 m spacing. The total number of trees in the grazed and non-grazed plots was 40 and 25 respectively. In both treatments, there were sufficient rows to sample non-edge trees. The experiment was managed as a five-week grazing cycle coupled to a five-month tree pruning regime. I used an animal stocking rate of 2 . 0 animal units/ha/yr, the estimated average for the Neguev settlement. Cattle were let into the grazed plots for 4 d; the pasture was then left to recuperate for the remaining 31 d. The same day that the animals were let out of their plots, the pasture in the non-grazed plots was cut to the same Table 5.1, General soil chemical characteristics of Neouev series surface (0-15 cm! and subsurface 115-30 cm) horizons.I

Soil pH Exchangeable Bases KCI-exch. ECEC# Organic C TKN$ Clay Bulk Depth Ca Mg K Acidity Content Density cm 1:1

0 - 1 5 5.1 -5 .3 0.87 0.95 0.47 1.45 3.74 3.38 0.38 60.7 0.87

(0.281 (0.151 (0.11) (0.35) (0.53) (0.30) (0.08) (4.1) (0.03)

15 -30 4.8 -5.0 0.34 0.33 0 2 6 1.49 2.43 2.23 0.23 ------

(0.121 (0.08) (0.09) (0.52) (0.54} (0.31) (0.05) I Values averaged over five study farms from 1987 soil sampling. Numbers in parentheses are standard deviations. # ECEC is effective cation exchange capacity - sum of NH 4OAC (pH 4.81 extr. bases + KCI-exch. acidity. @ TKN is total kjeldahl nitrogen. 258 TaMaS.2. Maiornhotnhonis forma for tha Naoutv soi) aariaa. i

Soil Dapth (cm) NaHC03-EDTA Organic P Total P axtractabla P

- ...... —...... mo PAa tail ......

0 - IS 2 .7 9 (0.86) 682 <1221 1476 (446)

15-30 2 .6 0 (0.701 801 181) 1238 (1661 i Valms art avaragad ovar fivt study fanns from 1967 samplas. Numbara in paranthasas art standard daviationa. pF (-log cm water suction) F igure 5.3. M oisture rete n tio n cu rv e tor N eguev so il il so eguev N tor horizon. e ) rv cm cu n (0-15 tio n e rete c rfa u s oisture M , s ie r 5.3. e s igure F 0 2 3 4 1 5 10 ouerc osue otn (%) content moisture Volumetric 20 u k soilbulk 30 40 50 60 70 260 Pasture with trees & grazing (900 m2) # m M u w U rn # # 1 ^ ^ 1 Trees/No Grazing 3m (400 m2)

Nongrazed pasture Grazed pasture (300 m2) (900 m2)

Figure 5.4. Field experimental design. A 2 X 2 factorial including trees and grazing. Grazed treatments are approximately twice as large as non-grazed (clipped) treatments. 262

height as the grazed pasture using machetes. When Brythrina trees were pruned, tree prunings were left on the ground as either supplemental forage

or mulch. No residues {either Brythrina or pasture clippings) were removed from the non-grazed plots, whereas the grazed plots incurred some export in the form of cattle biomass.

Monitoring labile soil P fluxes from the first tree pruning (June, 1989)

I used anion exchange filter paper (AEF) discs to monitor soil P fluxes as a function of tree pruning and cattle grazing. I monitored all four treatments on two farms (farms 1 and 2) for a period of 33 days. I began sampling 7 d before tree pruning and repeated sampling 7 and 21 d after pruning. For those treatments with trees, I included distance from trees as an independent variable. I selected, at random, two groups of four trees and superimposed a grid pattern of 36 sampling points (Figure 5.5). Distances between tree rows included 25, 75, 150 and 300 cm while those within rows were 75 and 150 cm from each tree. For treatments without trees, I used an artificial grid pattern with a 150 cm sampling interval to select 16 randomized sampling points. At each sampling point, regardless of treatment, I buried three AEF discs to a depth of approximately 10 cm. During each of the three sampling periods, I left AEF discs in the field for six days. I collected corresponding gravimetric soil moisture samples on the same day that discs were removed from the field. I determined AEF-P in 1 M NaCI extract solutions (method described in Chapter II; results reported as mg P/g filter paper disc) and used the AEF-P - soil solution P regression 300 cm ^ ------

T • • • T

• 28 cm 4i k i 7S

• too cm

• 300 Transect 800 cm • 180 cm

• 78

• 28 cm 1r

T«*■ • * • T . . . T

Figure 5.5. Transect sampling design used to monitor labile P dynamics as a function of tree pruning and grazing during the June 1989 tree pruning. Tree 263 spacing was 3 m within rows and 6 m between rows. 2 6 4 equation (also in Chapter II) to estimate changes in soil solution P as a function of time for the various treatments. I analyzed treatment and dependent variable effects using both PC SAS GLM and SYSTAT procedures for analysis of variance and post tested significant differences using least square means procedures (p < 0.05).

Monitoring Labile Soil P Fluxes Using Anion Exchange Membranes

Basic Procedure for Anion Exchange Membrane Field Use

I used Ionics® anion exchange transfer membranes (type 204-U-386) for all subsequent in situ determinations of labile P (see Chapter II for more complete discussion of methodology development). The basic pre-field preparation process involved cutting the membranes, presaturating with Cl" using 1 M NaCI, attaching some kind of durable thread for easy recovery and cleaning with deionized water before taking to the field. Ionics, Inc. supplies the anion exchange membrane in 30 cm x 30 cm sheets packed in ethylene glycol to prevent dessication. I cut sheets into 2.5 cm x 2.5 cm squares, rinsed off packing liquid with deionized water and presaturated membranes in 1M NaCI at least 24 h before use. I chose Cl" as the counterion because it was not expected to be a confounding anion in the soil solution system. I sewed dental floss (could use nylon fishing line or any durable thread) to the membrane and attached a brightly colored flag (red or orange works best in pasture grass) at the other end so it could be found in the field easily. I stored membranes in NaCI in the refrigerator until use to prevent bacterial and fungal growth. 265 When ready to take to the field, I washed membranes with deionized water and transported them in water-containing wide-mouth bottles (Ionics, Inc. recommends storage and transport in an aqueous medium to prevent cracking and disfiguration upon drying; i.e. both physical and chemical properties can change if they crack or split). I placed membranes in the surface soil by opening a vertical slit in soil with a hand trowel and gently sliding the membrane vertically into the slit (see Figure 2.7 in Chapter II}. I then closed the slit by pressing the soil firmly together, taking care to minimize disturbance of soil structure and to maximize membrane-soil contact. I collected membranes from the soil by tugging gently on the dental floss; in most cases, the membrane slid out easily (if the dental floss disengaged from the membrane, I had to excavate with a hand trowel to recover the membrane). I then removed large aggregates adhered to the membrane and carried the membranes back to the lab in water-filled plastic bottles. In the lab, I used a squirt bottle with distilled water and my fingertips to massage off any remaining soil particles. In general, soil did not cling to the membrane and subsequent filtering of the extract solution was not necessary. I placed individual membranes (or replicates) in beakers with distilled water until they were extracted and extracted AEM-P in 50- mL centrifuge tubes with 15 mL 1M NaCI by shaking on a reciprocating shaker for 1 h. I then removed membranes and analyzed P in the NaCI extract as ascorbic acid-reduced phosphomolybdate. I recycled used membranes by rinsing twice with distilled water and resoaking in 1 M NaCI until they were reused. I refrigerated membranes for long-term storage. 266

May 1990 Tree Pruning

I monitored one farm only (farm 2 from previous experiment) as a preliminary field test of the anion exchange membranes (AEM) and to determine the effect of AEM placement depth on AEM P extractability. The sampling frequency was the same as in the AE filter paper experiment: 7 d prior to pruning and 14 and 21 d after pruning. Within each treatment, I established 1 m^ quadrats; four in the two grazed treatments and three in the non-grazed treatments. In both treatments with trees, I located quadrats in close proximity to the trees to maximize detection of tree effects. I staked and tagged the four corners of each quadrat to facilitate reuse and subsequent identification. When trees were pruned, I specifically placed prunings within quadrats (leaves and branches) to ensure decomposition of Erythrina biomass within the quadrats. Within each quadrat, I placed 8 membranes at the 0-2.5 cm soil depth and 5 membranes at 5-7.5 cm soil depth. At each sampling date, I removed the resident AEM group and placed a new group such that the quadrat was considered the repeated sampling unit. I also took soil samples for gravimetric moisture determination on the same day that AEMs were removed. I took two composite samples per quadrat using a tube soil sampler to a depth of 0-5 cm (3 subsamples per composite sample).

November 1990 Tree Pruning

This was a much more intensive sampling frequency than that used for the two previous prunings. I monitored labile soil P with AEMs before 2 6 7 and after one tree pruning and three cattle grazing cycles. I took measurements on the block of four treatments on two of the five study farms (farms 1 and 3). I randomly selected and permanently marked 1m2 quadrats within each treatment plot; five quadrats in the grazed and three in the non-grazed treatments. Quadrats within the two tree treatments were located close to tree trunks to maximize tree effects. I began measurements 12 d before pruning by placing 10 AEMs in the surface (0-2.5 cm) soil of each quadrat. I removed them 4 d later and put in a new group for another 4 d (5 d before pruning!. I continued this 4 d sampling interval for 4, 8, 13, 17 and 22 d after pruning. I made subsequent measurements every 7 d until 70 d after pruning. At each sampling date, then, I removed AEMs and placed a new group in the same quadrat such that the quadrat was considered the repeated measures unit. I processed membranes and determined AEM-P as described above. I also monitored, concurrently, soil moisture suction at each quadrat using ceramic cup tensiometers.

Statistical analysis of AEM-P as a function of tree pruning and grazing

When I plotted non-transformed AEM-P by treatment (means and SEs from each quadrat per treatment were pooled for the two farms) against sampling date, there was no discernible pattern (Figure 5.6). There were fluctuations through time, but it was difficult to determine whether they reflected background noise or ecologically meaningful changes. I performed ANOVAs for treatment effects (farm means pooled) at each sampling date (Table 5,3). Although there were significant treatment differences at most dates, they were probably due to the large separation of the experimental AEM-P (mg P/L extract solution) oebr 90 re rnn ad grazing. following and fluxes pruning AEM-P tree Non-transformed 1990 November 5.6. Figure 4.00 0.50 2.00 4.50 0.00 2.50 3.00 3.50 1.00 1.50 1 2 3 4 5 6 7 8 90 80 70 60 50 40 30 20 10 0 o- ae ln -- ae n trooa no , razed -o-G alone ea ra -T -o rn & Graz* & Prune -•-C o n tr o l l o tr n o -•-C a (rm is sampling) (from firstDay Grazed— Trooa & Graz* . j 68 26 Tabia S.3. Corresponding sampling datas and results from preliminary - ANOVA of traatmant offsets on non-transformad AEM-P (farm maans pooladl.

Sampling Coinciding avant F-statistic P r> F Davs 0-4 12 day* bsfore pruning 3 .7 6 0.0 1 1 3

5-8 5 days bsfore pruning 3 3.16 0.0001

9-12 pruning and grazing + 4 days 6.91 0.0 0 0 2

12-16 8 days aftar pruning 3.59 0.0142

16-20 14 days aftar pruning 2.22 0.0861

20-24 18 days aftar pruning 2 .8 3 0 .0 3 8 9

24-28 22 days aftar pruning 4 .4 6 0 .0 0 4 5

28-36 29 days aftar pruning 13.56 0.0001

35-42 35 days aftar pruning 1.97 0 .1 1 8 6

42-49 42 days aftar pruning + grazing 5 .82 0 .0 0 0 7

49-66 49 days aftar pruning 0 .5 4 0.6563

70-77 63 days aftar pruning + grazing 0.89 0.4471

77-84 70 days aftar pruning 17.29 0.0001 270 control treatment (no trees, no grazing) relative to the other three treatments. As a first attempt at data disentanglement, I decided to normalize treatments relative to the experimental control treatment fno trees, no grazing). I calculated ratios for each treatment relative to the control for each sampling date and plotted the ratios against time (Figure 5.7), The resultant pattern was still difficult to interpret. I then decided to use a data normalizing index which would acount for treatment differences not only relative to the control, but also relative to starting conditions. I began with the assumption that there may have been treatment differences present prior to silvopastoral system perturbation (in this case pruning followed by grazing). Given this condition, I applied a data-normalizing index used in insecticide efficacy trials known as the Henderson and Tilton Percent Mortality Index (Taylor, 1987). This index normalizes treatments relative to the experimental control (non-grazed, no trees) and to silvopastoral system AEM-P status at the time prior to system perturbation; i.e., pruning followed by grazing.

(contrlto)(trmnttsc) AEM- P INDEX = 100 * ( 1 ) (contrltxMtrmttO)

The numerator of the ratio consists of contrlt0 as the control treatment AEM-P (no trees, no grazing) at time zero, or starting conditions, and trmnttx as any non-control treatment AEM-P at any time subsequent to time zero. In the denominator, contrltx is the control treatment AEM-P at any time subsequent to starting conditions and trmntto is any non-control treatment AEM-P at time zero. Trees alone -o- Grazed, no trees 271

Control —e-Graze & Trees 3.0 Prune a Graze

G raze

c 2.0

o 1.0

0.0

0 10 20 30 40 50 60 70 80 90 Day (from first sampling)

Figure 5.7. AEM-P fluxes following tree pruning and grazing expressed as the ratio of treatments relative to the non-grazed, no trees experimental control. 272 When I plotted the index-transformed data against time, the separation among treatments was even more defined than before (Figure 5.8). However, the range of corrected AEM-P values was large (from +98 to -2409); and, as such, the data were not normally distributed (variance- dependent means). To eliminate the negative number range from the scale and to stabilize the variance, I readjusted the data further. I created a new matrix (matrix 2) by subtracting index-AEM-P values from the largest positive number [98.2 -fx)J, thus eliminating all negative numbers and then taking the log-|o of (1 + matrix 2) values. I then performed repeated measures analysis of variance using SYSTAT (Wilkinson, 1991) for farm means combined and for each farm separately. I chose a repeated measures approach over individual ANOVAs at each point in time because it better accounted for the temporal component of the dataset. When variables are measured at different times, the changes induced by time on the sampling unit may be much greater than the changes attributed to applied treatments (Mead, 1988). The repeated measures analysis assumes that the variance among experimental units over time is heterogeneous (the same variable measured at different times is not sufficiently similar) and, therefore, single quantitative analyses of variance are not meaningful. In the model:

Dep. variable (1-5) = Const. + Factor-] + Factor + Factor-] *Factor 2 (2) where the dependent variable is measured over five sampling periods and the treatment includes two factors, the repeated measures ANOVA A T* h i m 00 00 TO 00 00 4 0 f t r m O ral ■ r ■ 1' i i 1 'T ' 1 i ,,"i 1 i *0 M Day Day ■ ■ i Day (M o Oral ■awp*»0) ■ ■ 0 0 tttOOOAOOOOOTOOO » 100 ■100 ■MO rfUJV HIIMMN mm Figure 5.8. Index-transformed fluxes AEM-P pruning. over time. Farm November 1 and (a) Farm 19903 1b). tree 2 7 4 computes effects from: 1) the underlying variation inherent in the system {tests the null hypothesis that the constant is not signficantly different from zero); 2) the two treatment factors separately; and 3) the interaction between the two treatment factors. It calculates effects for each component in the model as: 1) univariate repeated measures F-tests which assess the overall effect from each component over time; and 2) single degree of freedom polynomial contrasts which fit increasing order polynomials to the model. Significant polynomials {p<0.05) describe: 1) the mathematical shape of the dependent variable's underlying periodicity {if, for example, there is a significant third order polynomial for effects = constant, the overall shape of the curve has two significant turning or inflection points); and 2) the points in time at which treatments differ significantly; i.e., treatment divergences from both the system's underlying periodicity and from the other treatments. Using the fitted or adjusted residuals from the repeated measures ANOVA, I detransformed AEM-P back to the ratio of treatment at time x relative to both the control treatment and the starting conditions: [(controltQ) (trmnttx)] / [{trmnttp) (controltx)] . (3) where to is time prior to pruning and tx is any time subsequent to time zero. Results are expressed as this AEM-P ratio against time; significant treatment fluxes, therefore, represent changes relative to both space (control treatment) and time (time before pruning). Related to the process of discerning treatment effects, I hoped to remove, or at least, describe the variation associated with external environmental conditions. Given the importance of soil moisture for soluble P movement to the anion exchange membrane, the most logical parameter 275 for regression analysis was soil moisture tension. I performed regression analysis using log non-transformed AEM-P and log soil moisture suction or pF over a range of time lags i + 2. + 1, 0, -1, -2). I also conducted a repeated measures ANOVA for soil moisture (pF) alone, using farm and treatment as the two independent variables.

RESULTS

First Tree Pruning June 1989

In general, there were significant treatment effects (at the 0.05 level) despite the large inherent system variability (Figure 5.9), P fluxes, measured either as anion exchange filter paper P (AEF-P) or estimated soil solution P, differed for the two farms, however. Farm 1, because it exhibited more significant treatment differences than farm 2, was used for a more complete statisitcal analysis including tree distance effects. Several analyses of variance using the same main effects resulted in different F- statistics for AEF-P and estimated soil solution P. The discrepancy between the two dependent variables probably reflected the different ways they were obtained. The soil solution P values were estimated from the regression equation developed in vitro (see Chapter II), whereas AEF-P was measured directly in the field. In addition, because the correlation equation relating the two had a negative y-intercept, many of the very low AEF-P values AEF-P (mg P/g M e ) 14 10 12 - 14-20 0-0 a- f*, # r i* (r* a# If#**, * -* -a tM ttoM « n T - o - rnn ad rzn, ue 99 Fr 1 a ad am (b). 2 Farm and (a) 1 Farm 1989. June grazing, and pruning Figure 5.9, Labile soil P fluxes (as AEF-P) for all treatm ents following tree tree following ents treatm all for AEF-P) (as fluxes P soil Labile 5.9, Figure a (rm is sampling) first (from Day - I n n l * lf**« lf**« * l n n I - , l n i f • m tr*** 27-32 ft. ttl « f 14 10 12 0 2 4 0 •N* f*, ff*l* H If***, * -•-N O- ** t * # it* l*** -T -O a (rm is sampling) first (from Day 14-20 nl if*** I l in c f - > -

* < • > » * . m tm * 276 2 7 7 produced zero soil solution P values. Numerous zero values diminished the probability that means would differ significantly. Results are presented as either AEF-P or estimated soil solution P, and the two are discussed interchangeably. With sampling dates pooled, AEF-P close to trees (25-75 cm) was significantly higher than AEF-P away from trees (150-300 cm) for the trees and grazing treatment on farm 1 (Figure 5.10). In the trees alone treatment, however, there was no significant distance effect. The tree distance by treatment interaction may be a cattle-induced artifact or a legitimate interaction effect. When I simplified the distance factor to near-tree and away-from-tree and pooled grazed and non-grazed (tree) treatment means, there was a significant interaction between sampling time and tree distance (Figure 5.11). Soil solution P near trees did not differ significantly with time, although there was a suggestive increasing trend at the third sampling. Soil solution P far from trees peaked at 7-14 d after tree pruning and grazing and declined at 21-27 d after pruning. Again, differences may be related to tree biomass distribution and relative amounts. With distance means pooled, there was a significant treatment by time interaction for all treatments on farm 1 (Figure 5.9a). One week prior to tree pruning and grazing, there were no significant differences among the four treatments. The trees and cattle treatment showed no significant change in soil solution P (as estimated by AEF-P) over the 33-day measurement period. However, there was an increasing trend by the last sampling time (21-27 days after pruning). The trees alone treatment showed a significant peak at 14-20 days (7-14 days after pruning) followed Resin P (mg/kg resin paper) 5 4 3 6 2 0 5 10 5 20 5 30 350 300 250 200 150 100 50 0 Figure 5.10. Anton exchange filter paper P (expressed as resin P) as a a as P) resin as (expressed P paper filter exchange Anton 5.10. Figure en old. ue 99te rnn. am 1. Farm pruning. tree 1989 June pooled). means function of distance from tree for the two tree treatments (sampling date date (sampling treatments tree two the for tree from distance of function Distance from tree (cm) tree Distancefrom

TreesAlone re ad Cattle and Trees 8 7 2 1.5

1.3 • \ o> E 1.0

c 0.8 • '■M.2 D O 0.5 • i/i 9^o (/> 0.3

0.0 0 -6 14-20 27-33 Time (days)

Figure 5.11. Soil solution P (estimated from AEF-P! close to (25-75 cm) end far away (150-300 cm) from trees (both tree treatment means pooled) as a function of sampling time. The first sampling point represents one week prior to pruning; the two subsequent points represent one and three weeks 279 after pruning, respectively. 2 8 0 by a sharp decline at the third sampling. Soil solution P in the grazed pasture without trees increased from the time prior to pruning and grazing to three weeks after grazing. P fluxes in the no trees, no grazing control mirrored those of the trees alone treatment. The emergent pattern revealed a sharp distinction between grazed and non-grazed treatments; the two non­ grazed treatments exhibited increases in soil solution P as early as one week after pruning and grazing followed by a decline two weeks later, whereas the two grazed treatments did not begin to respond until three weeks after pruning and grazing. During the initial phase of silvopastoral system response to tree pruning and grazing, then, grazing effects overshadowed or dampened possible tree effects. The P flux pattern for farm 2 was quite different than farm 1 (Figure 5.9b). There were significant treatment differences prior to tree pruning and grazing; the control AEF-P was significantly higher than the other three treatments. One week after pruning and grazing, however, control AEF-P decreased and remained unchanged for the remainder of the sampling period. The two tree treatments' AEF-P decreased significantly after the first sampling time and remained relatively constant at weeks 1 and 3 after pruning. AEF-P in the grazed, no trees treatment did not change significantly over the entire sampling duration. In summary, no clear pattern emerged which separated treatments either by tree pruning or grazing effects. Concurrent measurements of gravimetric soil moisture showed little fluctuation over time in either farm (Figure 5.12). Throughout the 33 d sampling period, soil moisture fluctuated between 55 and 70% by weight or 47-60% by volume (gravimetric soil moisture X soil bulk density of 0.85 ■Tra ■ •nw i M IfMa ►- C aalral -tru*4 t tr*M 70 2 8 1 iprm * Otaaa

f J f 82 * 1 2

80

0-8 14-21 27-88 Day (from flrat s a m p lin g )

■ T i»a tlta t i a i a 4 , m it -a - C » a r * l -•itia l a tra

7 8

7 4

S 7 0

! m 6 2 1 o to 88

S4

50

0-8 14-21 27-83 Oay (from flrat sampling)

Figure 5.12. Gravimetric soil moisture fluxes following tree pruning and grazing, June 1989. Farm 1 (a), and Farm 2 (b). 2 8 2 kg/m^/bulk density of water). This range corresponded to the near­ saturated field capacity range of the soil's moisture retention curve (Figure 5.3). For farm 1, the grazed, no trees treatment showed an increase in soil moisture content from samplings dates 1 to 2, with no significant change thereafter (Figure 5.12a). The other three treatments exhibited a gradual increasing trend over time. There were no significant differences among treatments for farm 2 (Figure 5.12b), but all exhibited an increase in soil moisture content from sampling time 1 to 2 (55 to 71%). Nonetheless, there was no apparent relationship between soil moisture content and AEF-P fluctuations for either farm or treatment. Laboratory incubation studies showed a significant positve effect of soil moisture on AEF-P extractability (Chapter II). Perhaps, under field conditions however, P diffusion to the AE filter paper surface was not affected by such a narrow range of soil moisture contents.

Third Tree Pruning May 1990

Only one farm (farm 2) was monitored for labile soil P fluxes during the May 1990 tree pruning. As stated in the methods, I measured labile soil P using anion exchange membranes (AEM) and not anion exchange filter papers. In general, labile soil P fluxes exhibited the same pattern as measured one year earlier (Figure 5.13). There were marginally significant treatment differences in AEM-P during the week prior to system perturbation (pruning and grazing); the grazed and trees treatment AEM-P was significantly higher than the other three treatments. Following pruning and T ra * * * im m 1-1 ■ r u a * . m lr* * « Tra** ft |iu a 1.0 **■ ir * * * , ft* ■*■!

0.0

0.7

ui

0.1

0.0

0-7 14-21 20.38 Day (from flrai sampling)

0.80 -o - Tra** *i*«* Cut**. •• ira —ftu* ft It***

0.45

0.40 mo *» -j ST r CL 0.20 0.15

0.05

0.00

0-7 28-35 Oay {from flrai ••mpllng)

Figure 5.13. Labile soil P fluxes (as AEM-P) following tree pruning and grazing. May 1990. Farm 2. Surface (0-2.5 cm) horizon (a) and subsurface (5-8 cm) horizon (b). Note the difference in scale for the two horizons. 2 8 4 grazing, AEM-P decreased and treatments were indistinguishable at both 7 and 21 d after pruning. There was a very strong soil depth effect on AEM-P dynamics (Figure 5.13). The overall range of AEM-P for the 0-2.5 cm surface horizon was almost twice as great as that for the 5-8 cm subsurface: 0.15-1.1 versus 0.025-0.55 mg P/L extract solution. Whereas AEM-P in surface horizon showed significant changes over time, AEM-P in the subsurface remained low among all treatments throughout the sampling period (Figure 5.14). The two tree treatments exhibited consistently higher AEM-P in the soil surface relative to the subsurface, while the two non-tree treatments diverged only at the last sampling date. These results suggested that, for this high P fixing soil, labile P dynamics occurred only in the very thin, top soil layer; and, as such, were largely biologically mediated. Gravimetric soil moisture content fluctuated in a similar pattern as AEM-P fluxes; i.e., relatively high at the first sampling date followed by a statistically significant drop and then a slight increase (Figure 5.15). Despite this superficial similarity, there was no significant correlation between soil moisture content and AEM-P, Overall, there were no differences among treatments, and the soil moisture range was similar to levels encountered previously (56-80%).

Fourth Tree Pruning November 1990

Repeated measures analysis of variance for farms 1 and 3 together revealed a significant underlying periodicity in AEM-P among the four field treatments (Table 5.4). The univariate analysis F-test for constant was 285

f „ ! ~ S“ i “ i “

I I t i i u

Figura 8.14. LabQa soil F ftuxas (as AEM-P} by treatment lor surface (0*2.5 cm) «nd subsurface (5-8 cm) horizons Mowing m pruning and grazing, May 1990. Traas alone (a); Grazed, no traoa (b); Traas and Grazing (c); No traos. No grazing control (d). olwn ta puig n gaig My 90 Fr 2. Farm 1990. May grazing, pruning and changes foTlowing moisture trao soil Gravimetric 5.15. Figure Soil Moisture (% by wt.) 50 55 60 65 70 75 80 85 - 1-1 28-35 14-21 0-7 □Tos olono -□-Troos a (rm is sampling) first (from Day -■-Control

Gae k troos k ■Grazed r«, o troos no Gru«d,

6 8 2 2 8 7

Table 5.4. Univariate repeated measures F-tests for constant (grand mean), farm and treatment main effects and their interaction. Output from repeated measures ANOVA on log-transformed, index-normalized AEM-P. ______

Hypothesis test for effect ______df______F-statistic______P value Constant 11 13.27 0.0001 Farm 11 11.49 0.0001 Treatment 22 0.75 0.7820 Farm X Treatment 22 1.60 0.0490

Table 5.5. Univariate repeated measures F-tests by farm for constant and

Hypothesis test for effect df F-statistic P value Constant - Farm 1 11 12.36 0.0001 Treatment - Farm 1 22 0.71 0.8210 Constant - Farm 3 11 12.42 0.0001 Treatment - Farm 3 22 2.49 0.0010 2 8 8 highly significant, confirming the strong underlying signal inherent in the dependent variable evolving over time. In addition, there were several significant multiple-order polynomials which fitted the overall dataset (output not shown) which further corroborated the complexity of the temporal- dependent pattern. When the farm effect was added to the model, it too was highly significant; i.e. either the two farms' periodicities were different or the treatments were responding differently on each farm. The significant farm by treatment interaction confirmed the latter. For example, when the repeated measures ANOVA was run for each farm separately, there was no significant treatment effect on farm 1 (Table 5.5). However, treatment strongly affected the AEM-P pattern for farm 3. In fact, 3rd, 5th, 9th and 10th order polynomials significantly fit the AEM-P pattern with treatment included in the model, implying treatment divergences (differences) at the 2nd, 4th, 8th and 9th turning points in the overall pattern (Table 5.6). The interfarm treatment differences became apparent when viewed graphically (Figures 5.16-5.18). As stated in the methods, AEM-P ratio peaks represent increases relative to both the control treatment (no trees, no grazing) and AEM-P starting levels. Both farms exhibited similar AEM-P fluxes early in the cycle with small peaks 8 and 17 d after pruning (corresponding to days 16 and 24 on the x-axis) and a large peak just prior to the second grazing cycle (35 d after pruning). The only point at which treatments differed for farm 1 was the large peak prior to the second grazing, where the trees alone treatment showed a significant increase in AEM-P relative to the other two treatments (Figure 5.16). When the last sampling point AEM-P was removed from the farm 1 cycle and the scale reduced, both the underlying periodicity and suggested treatment -□-Trees alon* Grazed, no troo* 5.5 Grazed A tr**« 5.0

4.5

4.0

o 3*5 Prun* & Graz* m 3.0 o. z 111 < 2.0

1.5

1.0

0.5

0.0

0 10 20 30 40 50 60 70 80 90 Day (from first sampling)

Figure 5.16. Labile soil P fluxes (as AEM-P treatment: control ratio) following tree pruning and grazing, November, 1990. Farm 1. am wt ls smln pit removed. point sampling last with 1 Farm iue .7 Lbl si P lxs a AMP ratio) AEM-P (as fluxes P soil Labile 5.17. Figure olwn te puig n gaig Nvme, 1990. November, grazing, and pruning tree following AEM-P ratio 0.0 0.5 2.0 1.0 2.5 1.5 1 2 3 4 5 6 7 8 90 80 70 60 50 40 30 20 10 0 rn & Qraz* &Prun* a (rm is sampling) first (from Day OTaa alone -O-Traaa Gae, o treat no -Grazed, - + rzd trees A Qrizad

AEM-P ratio Figure 5.18. Labile soil P fluxes (as AEM-P ratio) ratio) November, AEM-P grazing, (as 3. and fluxes Farm pruning P 1990. soil tree Labile following 5.18. Figure 0.0 0.5 2.0 1.0 1.5 1 2 3 4 5 6 7 8 90 80 70 60 50 40 30 20 10 0 rn 4 Graz* 4 Pruna -o-Traea alona *+G razad. no traas traas no razad. *+G alona -o-Traea a (rm is sampling) first (from Day o Gaa & traas & —o-Grazad 291 u IJ IJ -•-fm I -o-fani 1 -*-ram a M ■ 4 J

4J - IJ 4J -

IJ IJ IJ - ! « U i J - s*1 « u - * KJ II -

M •J M u N M Day (Iran Hral lawfSat) Day (ftwa M w a m

npaa iMk. Tnaa 4 m Qaato—>■ r%m 1.11a. Traaa I tm ha Vaamaau

Figure 5.19. Labile P fluxes (as AEM-P treatmentxontrol ratio) by treatment for farms 1 and 3. November 1990 pruning. 2 9 2 293 differences became pronounced (Figure 5.17). The trees and grazing treatment AEM-P, for example, appeared to differ from the other two treatments 17 d after pruning. In comparison, the trees and grazing AEM-P on farm 3 exhibited a clear and significant divergence from the other treatments beginning approximately 28 d after pruning and continuing for the remainder of the sampling period (Figure 5.18). The other two treatments’ AEM-P fluctuated very little; trees alone peaked at 8 d after pruning and remained unchanged thereafter, while grazed, no trees peaked approximately 14 d after the second grazing cycle. When farm 1 and farm 3 were compared by treatment separately, periodicity differences were more striking (Figure 5.19). Grazed, no trees AEM-P for the two farms exhibited similar patterns until 22 d after pruning after which time farm 1 peaked just prior to the second grazing and farm 3 about 14 d later. The trees alone cycles also diverged 22 d after pruning, with farm 1 exhibiting a second, larger peak at the onset of the second grazing cycle. The two grazed and trees periodicities remained similar throughout most of the sampling period. However, farm 3 AEM-P peaked a third time 14 d after the second grazing, while farm 1 remained low. These interfarm treatment differences may have resulted from differences in grazing pressure, Erythrina biomass inputs or soil fertility status. The relationship between AEM-P fluxes and soil moisture was not easily discernable. A separate repeated measures ANOVA for soil moisture fluctuations (measured as pF or log soil moisture suction in cm H 2 O) resulted in no significant differences among treatments, farms or their interaction (Table 5.7) There was a significant underlying periodicity without main effects; all polynomials except 10th, 19th, 22nd and 25th 2 9 4 Table 5.6 Single degree of freedom polynomial contrasts for farm 3. Significant P values (< 0.050) indicate that treatments differ at the turning points of the given order polynomial. ______

Polynomial Order F-statistic ______P value 1 2.50 0.131 2 1.70 0.231 3 15.65 0.001 4 1.67 0 .3 7 0 5 6.23 0 .0 1 7 6 0.09 0 .9 1 7 7 2.68 0 .1 1 7 8 0 .5 9 0 .5 7 5 9 5.07 0 .0 3 0 10 5.67 0 .0 2 3 11 1.18 0 .3 4 7

Table 5.7. Univariate repeated measures F-tests for constant (grand mean), farm and treatment main effects and their interaction. Output from repeated

Hypothesis test for effect df F-statistic P value C onstant 29 70.92 0.0001 Farm 29 14.67 0.0001 Treatment 58 0.91 0 .6 7 3 0 Farm X T reatm ent 58 1.07 0 .3 4 4 0 295 orders were significant, and the overall F-statistic for the constant was highly significant. The repeated measures analysis corroborated the visually obvious: soil moisture fluctuated markedly over the sampling period (Figure 5.20). Although fluxes were apparent, in fact, they fell wiithin the range encountered during previous pruning measurements; i.e., the near saturated to field capacity range of the soil's moisture retention curve. The lack of significant difference in soil moisture content between farms eliminated the likelihood that AEM-P treatment differences were due to differences in external environmental conditions. Regression analysis between log-transformed (non-index normalized) AEM-P and a range of time lagged pF produced significant but non-linear relationships between AEM-P and pF no time lag, time lag-1 and time lag~2 (Figure 5.21). The regression with the highest r^ (0.210) and lowest residual mean-square error was time lag-1. When log AEM-P was plotted against pF lag-1, the suggestive linear relationship was apparent (Figure 5.22). However, when viewed by treatment, the relationships were inconsistent (Figure 5.23). The no trees, no grazing control AEM-P had the strongest correlation with pF (r^ = 0.42), followed by the grazed, no trees and trees alone treatments (r^ = 0.21). There was no relationship whatsoever between grazed and trees AEM-P and pF ( r2 = 0.098). Interestingly, the log (non-index normalized) AEM-P-pF regression goodness- of-fit was inversely related to index-normalized AEM-P treatment divergence from other treatments over time: the grazed and trees treatment AEM-P flux, which diverged from the other treatments over time, was least correlated with soil moisture fluxes. This further confirms the relatively minor influence of soil moisture on AEM-P fluxes, particularly when treatment 296 Farm 1 —o- Farm 3

3.0 Pruna & Graza

2.5 c o M C o 2.0 w9 3 'Ci CO o E 1.5 o <0 oOft 1.0 u. a.

0.5

0.0

0 10 20 30 40 50 60 70 90 Day (from first sampling)

Figure 5.20. Soil moisture fluxes (as pF) for both farms (1 and 3) during November 1990 tree pruning (treatment means pooled for each farm). 2 9 7

0.14

0 .1 3 a.

0.12

0.11

0.10

2 0.09

0.06

-2 1 0 1 2 Time Lag

Figure 5.21. Mean square error fMSE) of residuals from regressions of log AEM-P(non-index normalized} against several pF time lags. The lowest residual MSE corresponds to the best fitting regression. ffi Trees alone ■ Grazed, no traaa

* Control • Grazad & traaa

1 .0 - i y - -1.428 + 0.616X r*2 - 0.210 ♦ 0.5 -

SI Q. a B 2 ui 0.0 - <

wE 0 -0.5 - 0 -B * * ♦ » of SI c IS 0 CO 1 9 c 9 o ■1.0 -

■1.5 n

- 2.0 — i | r - T T T 1.4 1.7 2.0 2.3 2.6 Log soil moisture tension (pF) lag-1

Figure 5.22. Time lag relationship (-1 sampling date) between labile soil F (as log AEM-P) and son moisture (as pF), all treatments plotted together. Lap NwMrMlMiM4 AEM-P r 04 -0 M re, o rzn cnrl a; rzd n re () Tes ln () Tre and rees T (c); alone Trees (b); trees no Grazed, la); control grazing no trees, 3 0 rzn (d). grazing iue 23. niiultet n AMP- Frgeso rltosis No relationships. regression pF - AEM-P ent treatm Individual . 3 .2 5 Figure 10 -) 10 to S 4 4 4 4 4 4 S4 14 84 14 14 14 14 IS Mlt, TrM l Ml lMH MuM a a M u M . M PlpMTH M L M lM M t M ll r»M T , t l M t w p F |»l Mta. Ol ll (M WM. p k u p m . M W M ( l ll lM O . a t . M M M ttL l V » W| 1.7 pF T Im * L ap ap L * Im T pF a. * i * v F p TK m p a L 2.1 -1 fa Lap a L Tfcaa F p • i r i i j s _i i I HI A a. < r “ i 10 7 4 4 . 24 1 24 2 14 4 2 1.1 4 1 14 14 17 a p n u M M . . M M U p a L m I T F p m m utlwiM. tu Mm m , M M T m m -1 *1 *l r* 299

3 0 0 effects are great. Incidentally, because the control treatment AEM-P was most strongly correlated with soil moisture and the other treatments were normalized relative to the control, any variation associated with soil moisture was removed with the index normalization.

DISCUSSION

LeguminousTree Effects

Erythrina trees affected labile soil P dynamics both spatially and temporally. The spatial effect was more pronounced when trees were combined with cattle grazing; i.e., labile soil P was higher underneath the tree canopies of grazed plots than under the canopies of non-grazed, clipped plots. It is possible that cattle grazing altered the distribution of decomposing Erythrina biomass such that remaining leaves and branches were concentrated around tree bases. This unequal distribution could have created an artificial tree distance effect. Alternatively, there could have been legitimate tree effects underneath tree canopies which were magnified by grazing. Singh et al. (1989), for example, found enhanced available P under Poplar and Eucalyptus tree canopies by as much as 33%. Virginia (1986) also reported greater N and P accumulation under mesquite iProsopis) canopies relative to non-tree interspaces. Both attribute available P enrichment under trees mainly to deposition from mineralized litter P with some input from root decay and/or root P leakage. 301 If pasture growing within the tree canopy radius benefits from the soil's enhanced available P status, it should lead to greater new biomass production than outside the tree's radius; which, in turn, may result in greater livestock consumption. If cattle selectively graze pasture close to trees, there should be greater pasture root dieback close to trees resulting in higher soil solution P (AEF-P). While pasture biomass production differences within and outside the tree canopy radius were not determined and cattle foraging behavior not quantified, it is also possible that cattle simply grazed pasture close to trees more intensively than in inter-tree row spaces because there were more Erythrina leaves close to trees upon which cattle preferred to browse. Anecdotal evidence suggests that cattle, upon entering the experimental plots, browsed selectively on Erythrina leaves over pasture grass. It seems logical that they would then eat pasture in close proximity to the greatest abundance of Erythrina. In either case, cattle created greater pasture biomass spatial heterogeneity than could ever be achieved by simply cutting pasture with a machete, which, subsequently, accentuated soil solution P {AEF-P) spatial heterogeneity. Over the four week period following tree pruning and grazing, tree canopy effects manifested differences in soil solution P temporal patterns, regardless of grazing. The stability of soil solution P near trees contrasted with the more dynamic P flux away from trees and could reflect both tree microsite effects as well as differences in decomposing residue distribution. A more constant source of decomposing biomass close to the tree base may have resulted in a more gradual but persistent release of mineralized P into the soil solution. In addition, the tree-pasture rhizosphere may have greater levels of biochemical activity than pasture alone rhizosphere. These 3 0 2 biochemical processes may be mediated by: 1) differences in microbial composition and activity (Swift, 1986); 2) different composition and quantities of root exudates (Singh et al., 1988; Fox and Comerford, 1990; Gillespie and Pope, 1990); and 3) greater mineralization of labile organic P pools which replenish solution P (Sharpley, 1985; Tiessen et al., 1984). Root exudation of low molecular weight organic acids, for example, can either chelate free Al or acidify the rhizosphere thereby maintaining more P in the soil solution. The tree-induced zone of heightened biochemical activity coupled with decomposing Erythrina residue, could have led to P immobilization shortly following residue placement. Three weeks later, however, solution P appeared to be increasing and was significantly higher than solution P away from trees. Although tree prunings were distributed fairly evenly between and within tree rows, it is likely that Erythrina residues were more sparse at 150-300 cm from the trees. Because the tree rhizosphere effects were probably minimal at this distance, both microbial P immobilization as well as solubility enhancement reactions would not influence soil solution P dynamics as strongly. The significant but short-lived soil solution P peak detected one week after pruning may reflect easily mineralizable P released from localized and concentrated sources (either Erythrina leaves or dung). The pulse was highly ephemeral because there were no tree rhizosphere effects present to counteract the soil's strong adsorption capacity (Vitousek and Denslow, 1986; Adepetu and Corey, 1977). It is also not surprising that there was such a strong soil depth effect on labile soil P dynamics (May 1990 pruning), given the soil's high P retention capacity and P’s inherently low mobility. The comparatively high 303 AEM-P in the upper 2.5 cm and the lack of change in the lower 5-8 cm during the entire 33 d sampling suggest that labile P dynamics were almost exclusively controlled by biologically-mediated, upper rhizosphere processes (Till and May, 1971). Alternatively, it is also possible that there was greater root biomass in the 0-2.5 cm soil layer than 5-8 cm, and, therefore, there was less mineral soil present to fix mineralized P. In either case, results confirm the importance of organic processes in regulating P availability (Ewel, 1986).

AEM-P fluxes were most pronounced and more immediate in the two tree treatments; this too supports the hypothesis that leguminous tree rhizosphere effects are more pronounced than those of the pasture rhizosphere (Virginia, 1986). In the case of surface residue decomposition, P fluxes resulted initially from immobilization followed by mineralization three w eeks after pruning. P immobilization in the inital phase of decomposition is common when residues have C:P ratios greater than 300 and available soil P reserves are low (Singh et al., 1988; Chauhan et al., 1979; 1981). Pasture removal, either via grazing or clipping, did not induce significant change either spatially (depth) or temporally until three weeks after grazing. These results suggest that mineralization from pasture root dieback did not affect soil solution P until three weeks after grazing and was evident only in the most active, upper root zone. 304 Grazing Effects and the Relationship Between Trees and Grazing Over the 18-Month Study

Grazing effects overshadowed tree effects on labile P dynamics in the first pruning event (June 1989). This seems reasonable since trees were young (had been planted from stakes one year prior to pruning) and

Erythrina leaf biomass production was not maximized (because trees had been left unpruned for 12 months, pruned biomass was mostly woody material). The AEF-P peak detected in the two non-grazed (clipped) treatments one week after pruning and grazing could reflect the P flush associated with a fairly uniform pasture clipping (decreased plant uptake) coupled with microbial mineralization of easily degradable sources like pasture roots (Saunders and Metson, 1971). The absence of similar AEM-P peaks in the two grazed treatments suggests that cattle grazing did not produce the same intensity of pasture root dieback as pasture clipping. In addition, some of the pasture biomass was converted to either cattle biomass or dung. Mineralization of the most soluble dung P could have produced fairly rapid, pronounced AEM-P peaks (as seen in the controlled field decomposition study, Chapter IV) in a very localized pattern, but its overall treatment effect was probably diluted (Buschbacher, 1987). In the grazed treatments, then, pasture P uptake and microbial mineralization from easily degradable sources acted in opposition; in the non-grazed clipped treatments, those same processes occurred in synergistic consort. One year later, tree treatments were beginning to exhibit different labile P dynamics than non-tree treatments, albeit not very striking. One week subsequent to tree pruning, both tree treatments exhibited significant 305 decreases in AEM-P followed by increases two weeks later. These fluxes probably reflect immobilization and mineralization of Erythrina residues. In contrast, the two non-tree treatments exhibited no significant change in AEM-P over the four week sampling. Although tree effects were more pronounced than those from the 1989 pruning, tree and grazing effects were still acting independently on labile P dynamics; at least within the 30 + day sampling duration. Perhaps only the fourth pruning (November 1990) P dynamics provide sufficient data to evaluate individual tree and grazing effects and their interaction. Regardless of farm and treatment, AEM-P exhibited a strong underlying periodicity. This periodicity was likely a manifestation of the living pasture biomass biochemical rhythm; i.e., the dynamic equilibria between plant P uptake (changing with pasture growth and death), rhizosphere effects, microbial biomass fluctuations and soil P adsorption and desorption (Saunders and Metson, 1971; Blair and Boland, 1978; Yavitt and Wieder, 1988). Underlying P fluctuations were altered to a certain extent by system perturbations (pruning and grazing). In fact, many of the peaks observed in the P flux pattern correspond temporally to those observed in the field decomposition study (Chapter IV). Both pasture grass clippings and

Erythrina leaf residues, for example, exhibited AEM-P peaks 10-30 d after residue placement. These correspond to the first small peaks observed in all treatments beginning approximately 7 after pruning and continuing for another 14-18 d. The trees alone treatment AEM-P peak (farm 1 only) which occurred just prior to the second grazing corresponds to the decomposition study’s Erythrina AEM-P peak (approx. 2.5 mg P/L extract 306 solution) around 35 d after residue placement. Although these correlations are purely qualitative, they support the hypothesis that AEM-P ebbs and peaks measured in the silvopastoral system as a whole resulted from immobilization and mineralization processes associated with residue decomposition (Dalai, 1979). Yavitt and Wieder (1988) interpreted sharp decreases and increases in labile P as either pH-derived P solubility enhancement (decomposing residues releasing organic acids which decreased pH thereby increasing P solubility) or immobilization-induced increase in phosphatase activity leading to increased P mineralization (low substrate-induced enzyme production). The grazed tree treatment AEM-P exhibited significant divergence from the other treatments over time (on farm 3). This suggests a positive synergistic relationship between trees and grazing. Decomposition of

Erythrina biomass initially undergoes the aforementioned immobilization-

mineralization "tug-o-war." The Erythrina biomass P release characteristics (Chapter IV) do exhibit a bimodal AEM-P flush; the first peak corresponds to flush from the easily mineralizable material (25-40 d after residue placement), and the second, larger peak corresponds roughly to the same period that grazed tree treatment AEM-P peaks (approximately 70 d after residue placement). This second peak is not present in the trees alone treatment, however; intimating a priming effect from grazing. Both Mott (1975) and Tate (1984) proposed that grazing accelerates P turnover, ultimately resulting in greater system retention of nutrients. The second and third grazing cycles appear to induce additional waves of mineralization of the more recalcitrant organic P. This could be due, in part, to plant biomass P conversion to more soluble dung P. More 307 likely, however, the pasture biomass loss and root dieback associated with grazing has stimulated the suite of biochemical processes resulting in enhanced available soil P: decreased plant uptake coupled with increased microbial activity leading to increased mineralization of previously deposited

Erythrina residues and other sources of organic P (Tate, 1984; Mott, 1975).

Relationship Between Soil Moisture Fluctuations and Labile Soil P Dynamics

Overall, the relationship between soil moisture and labile soil P (either AEF- or AEM-P) fluxes was not strong. The relatively constant wet conditions (field soil moisture levels rarely dropped below 55% by weight) ensured that microbial processes were never water-limited (Harrison, 1982), nor was P diffusion to anion exchange materials impeded. As such, there w as no significant correlation between gravimetric soil moisture and labile soil P during the first and third pruning events, regardless of treatment. There was a significant and weakly linear time lag relationship between soil moisture ( as pF) and AEM-P during the fourth pruning event, however. The variation explained by this relationship was inversely related to treatment AEM-P divergence over time. Fluctuations in the grazed tree treatment AEM-P were not related to soil moisture at all, while the non- grazed, no trees control treatment AEM-P was strongly related to preceding soil moisture conditions. When treatment effects are large, then, they appear to overshadow silvopastoral system variation associated with external environmental conditions like soil moisture. 308 Farm Differences in Labile P Dynamics

The disparity in labile P fluctuation patterns among farms suggests that there are inherent farm differences not necessarily reflective of treatment effects. It is possible, although stocking rates were uniform, that grazing pressures differed between farms compared during both the first and fourth pruning events. It is also possible that there were farm differences in

tree biomass production which could have resulted in Erythrina leaf biomass production insufficient to produce significant effects. A more plausible explanation, however, is that there are inherent farm differences in soil P fertility status (Table 5.8). When farms 1 and 2 were sampled in June 1989, farm 1 had signficantly higher anion exchange resin-extractable P and organic P (comparing measurements from beginning of study). Modified

Olsen-extractable (NaHC 0 3 -EDTA) P did not differ between farms however. When farms 1 and 3 were monitored in November 1990, both Olsen P and organic P were signficantly higher in farm 3 than farm 1 (comparing measurements from the end of the study). It appears, then, that sufficient levels of organic P are needed to fuel the biologically mediated processes regulating both tree (residue decomposition) and grazing (decreased plant P uptake, dung decomposition, root dieback) effects on labile soil P dynamics. Several studies concur that, for soils low in available P (highly weathered ultisols, high P-fixing andosols), bioavailable P dynamics are controlled largely by the mineralization of organic P (Tiessen et al., 1984; Sharpley, 1985; Parfitt et al., 1989; Stewart and Tiessen, 1987). Unless the inorganic P input is sufficient to minimize immobilization, microbes will utilize inorganic labile P pools. If inorganic labile P pools are depleted, 309

Tabia 5-8- Major P form* in tha surfaco (0*15 cm) horizon from tha baginning (Sept. 1987) and and (Dae. 1990) of study. ______

Farm AE Rosin P N*HCC >3 axtr. P Organic P Total P

bag, only ______bag. ______and______bag. ______and______bag. ______and - ...... - mg P/ kg soil------

1 0.328 1.95 3.63 900.21 615.47 1310.55 821.25 2 0.179 2.20 3.34 716.43 480.55 1091.41 662.11 3 0 .3 7 8 ______3.28 11.69 893.06 705.73 1507.06 1196.67 LSD-0.087 LSD-0.52 LSD-167.4 LSD-95.21 LSD* ara valid for both farm and tima diffarancas. 310 microbes will depend on the rate of mineralization of organic P pools to supply them with enough P to decompose the added residue (Chauhan et al., 1979; Blair and Boland, 1978). Those native organic P pools can be critical to maintaining labile P dynamics when residue inputs are P pauperate. If residues contain sufficient P (C:P ratio < 200 or P content > 0.2%), microbes will be less likely to draw P from native organic P pools, resulting in greater net mineralization. Given the significant difference in P dynamics on two farms with inherently different organic P reserves, it is probable that the 18-month old silvopastoral system w as utilizing the soil's organic P reserves to fuel system P dynamics. In fact, the significant decrease in organic P on all farms from 1987 to 1990 corroborates this conjecture. As the system matures and above and below ground biomass accumulates, however, organic P is expected to recycle mainly as a function of system management practices (with minimal input from soil organic matter priming); and, therefore, soil organic P reserves should be replenished.

CONCLUSIONS

1. Erythrina berteroana, an N-fixing leguminous tree, significantly increased labile soil P both spatially and temporally. Tree effects over time were more pronounced within the influence of the tree canopy than outside the tree canopy radius. 311 2. Grazing effects overshadowed tree effects early in the silvopastoral system’s establishment. After 18 months, however, trees and grazing produced positive synergistic effects on labile P dynamics, maintaining higher, sustained soil solution P than either trees or grazing alone.

3. The pasture ecosystem P dynamics reflect an underlying periodicity which only occurs in the near (0-2.5 cm) soil surface. The relative lack of labile P fluctuations even 5 cm below the soil surface is strong supportive evidence for complete biological (plants and rhizosphere microbes) mediation of P dynamics in this high P fixing soil.

4. Soil solution P variation associated with soil moisture fluctuations can be masked when tree and grazing signals are significantly strong. In general, the soil moisture range is so narrow in this system (between field capacity and near saturation) that both biological and physiochemical processes are not greatly affected. The lack of significant soil moisture effects in a non­ disturbed field settting contrasts with significant effects for a similar soil moisture range on disturbed soil samples in the lab (Chapter I)).

5. Inherent levels of soil organic P are needed to augment silvopastoral system P mineralization, at least in the system's early phase of development and nutrient accumulation. CHAPTER VI

SILVOPASTORAL SYSTEM-LEVEL CHANGES IN LEGUMINOUS TREE {ERYTHRINA BERTEROANA) AND PASTURE BIOMASS PRODUCTION

INTRODUCTION

In much of the Latin American humid tropics, cattle rearing is the predominant production system on lands recently cleared of primary rain forests {Toledo and Torres, 1990; Buschbacher et al., 1988). Laissez-faire management practices including continuous grazing, overstocking, no fertilizer inputs and lack of high quality forage eventually lead to overgrazing and declines in both soil fertility and pasture productivity. The end product is degraded pasture; i.e., an increase in undesirable forage species {weeds) to the point where it is no longer economically or ecologically viable to maintain livestock {Serrao et al., 1979). Soils with inherentlly low fertility (low nutrient levels, acid, aluminum toxicity, high phosphorus retention) exacerbate the problem and accelerate the process of degradation. In fact, on extremely infertile soils where nutrient reserves {particularly phosphorus) have been depleted by mismanagement, pasture species are often unresponsive to fertilizer additions (Crowder and Chheda, 1982; Buschbacher et al., 1988). In an attempt to improve existing grazing land productivity and ameliorate land degradation associated with deforestation, researchers and 312 3 1 3 development practicioners have proposed farming systems which maximize utilization of on-farm resources and minimize dependency on expensive external inputs (synthetic fertilizers, pesticides, machinery, introduced species). The pasture ecosystem, unlike annual cropping systems, has a propensity for nutrient conservation; i.e., it does not experience the large and inevitable nutrient export associated with crop harvest (Humphreys, 1991). Silvopastoral systems which integrate woody perennials and herbaceous pasture species have the potential to improve the pasture's inherent nutrient-conserving capabilities through enhanced nutrient cycling (Chiyenda and Materechera, 1989; Singh, 1990). One potential role for trees in the pasture landscape is the live fence. Many farmers in Central and South America plant leguminous trees to delineate farm and pasture borders (Budowski, 1985; Russo, 1990; Pezo et al., 1990; Toledo and Torres, 1990). Trees are pruned on an annual basis mainly to produce new posts. The live fence reduces maintainance costs and provides limited shade for grazing animals. It is an under-utilized resource, however, because it could be managed for supplemental livestock forage production. Due to their relatively high nutrient content, leguminous trees possess considerable forage value (Torres, 1983; Perdock et al., 1982). Several studies have demonstrated increased live weight gain and milk production in animals fed diets including leguminous tree fodder (Pezo et al., 1990). If planted in close association with pasture grasses (i.e., plantation style), trees can affect the pasture beneficially as well. Experiments in which leguminous trees (Erythrina poeppigiana) were planted with pasture grasses showed signficant increases in pasture production and crude protein 314 content relative to both a non-legume tree treatment (Cordia a/iodora) and a no tree control (Torres, 1983; Daccarett and Blydenstein, 1968; Bronstein,

1984). Singh (1990) found that pasture associated with Acacia and Aibizia species was seven times more productive than within traditional land use systems. Christie (1975) obtained threefold increases in buffel grass yield under Eucalyptus (a non-leguminous tree). Other studies have demonstrated the nutrient accumulating potential of tree-pasture associations over relatively short time periods (Singh, 1990; Montagnini et al., 1989; Toledo and Torres, 1990). Reciprocally, livestock grazing can have beneficial effects on tree and pasture productivity. Dung pats are known to concentrate considerable amounts of N and P, for example. Nutrients returned in livestock dung can lead to increased pasture and tree growth when grazing is sustained and intense (Peterson et al., 1956; Humphreys, 1991). Livestock grazing also reduces nutrient competition from grasses which can also result in increased tree growth (Jusoff, 1988). If livestock forage on high (nutrient) quality tree prunings, they may indirectly benefit the pasture ecosystem as a whole by reducing grass intake and increasing nutrient cycling through richer excreta (dung enriched from tree prunings) (Pezo et al., 1990). Few studies have evaluated integrated effects from both trees and grazing on pasture production. Nor have they measured nutrient fluxes, specifically limiting nutrients like phosphorus, among silvopastoral system components. I hypothesize that, if leguminous trees and livestock grazing together promote enhanced nutrient cycling (specifically P cycling, the limiting nutrient in the system), their combined effects should manifest themselves in the system's above-ground components; i.e., pasture biomass 315 and nutrient production should increase. The objectives of this study, then,

were to determine the effects of leguminous trees (Erythrina berteroana), tree pruning and cattle grazing on pasture biomass production and biomass P.

MATERIALS AND METHODS

Site Description

The field experiment was located in the Atlantic coastal plain of Costa Rica (10° N 8 3 ° W). The area receives 3630 mm rainfall annually and the ecological lifezone is lowland humid tropical rainforest. The soil is an andic humitropept (series Neguev) and is located on the ridgetops of a slightly undulating landscape. The geologic origin is volcanic ash and lahars from the late Pleistocene; due to its stable position on the landscape and high rainfall, however, it has mineralogical characteristics of more highly weathered soils (kaolinite, gibbsite and iron oxides). In general, the Neguev series is deep, well-drained, acid, low in exchangeable bases, high in exchangeable Al, high in total P but low in available P, clayey and low in bulk density. In addition, the Neguev soil has an extremely high P fixation capacity (> 2000 mg/kg) and a high moisture retention capacity (> 60 % by volume) over a wide range of soil moisture tensions.

The field experiment, a silvopastoral system, was a 2 X 2 factorial design with cattle (grazing) and trees as the two independent variables (Figure 6.1). The resultant treatments were: 1) a grazed pasture with trees Pasture with trees & grazing (900 m2) # # #

$ j j $ ^ ”

Trees/No Grazing 3m (400 m2)

Nongrazed pasture Grazed pasture (300 m2) (900 m2)

Figure 6.1. Field experimental design based on a 2 X 2 factorial including trees and grazing. The two grazed treatments were 900 m2, the non- grazed, tree treatment was 400 m2 and the non-grazed, no trees control was 300 m2. 317

(GT); 2) a grazed pasture without trees (G); 3} a non-grazed pasture with trees (T); and 4) a non-grazed pasture without trees, the experimental

control (P). I planted Erythrina berteroana (a tropical leguminous tree) from vegetative cuttings (large stakes) in native grass pastures on five farms, all within a 6 km radius (inside the Neguev settlement). All farms were located on the Neguev soil series upon which I established the randomized block of four treatments (farms were considered as experimental replicates). Grazed treatment plots were 900 m2 and non-grazed treatments 400 m2.

General Management of Field Experiment

The experiment was managed as a five-week grazing cycle coupled to a five-month tree pruning regime. I used an animal stocking rate of 2.0 animal units/ha/yr, the estimated average for the Neguev settlement (see Chapter I). Two to three animals weighing 300-450 kg each were let into the grazed plots for 4 d; the pasture was then left to recuperate for the remaining 31 d. The same day that the animals were let out of their respective treatment plots (either with or without trees), the pasture in the non-grazed plots was cut to the same height as the grazed pasture using machetes. No residues (either Erythrina or pasture clippings) were removed from the non-grazed plots, whereas the grazed plots incurred some export in the form of cattle biomass. I planted trees one year prior to the experiment's initiation to allow sufficient time for establishment (one year is considered adequate time for establishment since trees are planted as large stakes and they root fairly rapidly). I cut 2.6 m-long stakes (diameter ranging from 6-10 cm) from a 3 1 8 single live fence approximately 10 km from the study farms. This length (height) was chosen to ensure that, when planted, cattle would not have access to canopy regrowth. I planted trees using a 6 m X 3 m spacing based on results from previous agroforestry studies in Costa Rica (Bronstein, 1984; Russo and Budowski, 1986). The total number of trees in the grazed (GT) and non-grazed plots (T) was 40 and 25, respectively. Although the tree spacing was identical for the two treatments, the tree density on a per hectare basis was 444 trees/ha for GT and 625 trees/ha for T. During the first year of tree establishment, I protected trees from livestock damage (cattle would rub against trees, and if they weren't rooted sufficiently, they could be pushed over) by attaching them to barbed wire grounded to firmly planted wooden posts (Silvopastoral Systems Project (SSP), 1989). Previous trials had shown that this was the most successful way to protect stakes from cattle uprooting or debarking. In this way, I was able to implement the grazing regime shortly after tree planting. Trees were left unpruned for approximately 1 2 months, after which time, I began pruning every five months. The five-month pruning frequency was chosen because previous studies had shown that 4-6 month pruning frequencies maximized leaf biomass production and nutrient content without premature tree mortality (Joachim and Kandiah, 1934; Budowski et al., 1985; SSP,

1989). When Erythrina trees were pruned, tree prunings were left on the ground as either supplemental forage or mulch. 313 Climatogical Monitoring

Beginning in May, 1989 and ending in December, 1990 I monitored daily precipitation and maximum and minimum air temperatures on one farm (Farm 1) in Neguev. These measurements were taken to provide more accurate climatological data for the Neguev settlement because the Atlantic coastal region is characterized by high microclimatic variability. I measured daily rainfall (mm) using a "Tru-check" wedge-shaped bucket rain gauge and max/min ambient temperatures in the shade using a mercury max-min thermometer (the ambient temperature data were presented in Chapter I). In addition, I obtained monthly rainfall measurements from the El Carmen weather station for comparison; 6 km east of Neguev settlement and 15 m above sea level; it was the closest complete weather station.

Pasture Biomass Determination

I conducted an informal survey of pasture floristic composition for baseline characterization (Table 6.1). I measured pasture biomass before and after each grazing cycle using the comparative yield method (Haydock and Shaw, 1975). I took measurements for each treatment plot on all five study farms for 17 consecutive grazing cycles (May 1989- November 1990). Results are presented as pasture biomass on offer; i.e., the standing biomass just prior to grazing. The comparative yield method is based on the belief that it is less biased to estimate yield from visual criteria rather than attempt to estimate actual weight (Shaw et al., 1985; Crowder and Chheda, 1982). The method included: 1) setting a fixed biomass scale based on pasture height 320 Table 6.1. Botanical inventory of most common pasture species found in native grass pastures in the Neguev settlement. Included are latin names, common local name and general classification.

Species ______Local Name Category Phaseoius peduncuiaris unknown Pyifantus niruri unknown Lascicsis sorghoidea unknown S/da rhomifoiia unknown Miconia spp. unknown Winterigia spp. unknown Stei/aria ovata unknown Ischaemun Ind/cum Ratana grass Paspalum fascicuiatum Gamalote grass Digitaria horizontal/s grass Brachiaria spp. grass Setaria sphaceiata San Juan grass Cynodon n/emfuens/s Estrella grass Centrosema spp. legume Desmodium spp. legume Panicum iaxum Pasto Natural natural grass Paspafum conjugatum Pasto Natural natural grass Panicum frondescens Pasto Natural natural grass Homolepsis aturens/s Arocillo natural grass Panicum piiosum Pasto Natural natural grass ipomoea spp. Churistate weed Pseudolephanthopus spicatus Lechuguilla weed Kyliinga brevifofia Cyperaceae weed Ageratum conyzoides Santa Lucia weed Cyperus ferax weed Cyperus iuzuiae weed Scieria pterota Cyperaceae weed Borreria iacris weed Cyperus diffusus weed Hyptis capitata weed Drymaria cordata Mielsilla weed Soianum torvum weed Paspaium virgatum Zacaton weed Psidium guajava Guayaba weed Mimosa pudica weed Borreria iatifofia weed 321 and density; the lowest rank <11 corresponded to the lowest, thinnest pasture and the highest rank (5) corresponded to the tallest, most dense pasture; and then 2). ranking visual estimates of biomass in each treatment (60 in grazed plots, 40 in non-grazed plots). I ranked visuals by throwing a 25 cm X 25 cm quadrat randomly within each treatment plot. In conjunction with the visual estimates, I cut quadrats representative of each rank, dried them at 70 °C to determine their dry weight (DW) biomass and subsampled for nutrient (P content) analysis (Briceno and Pacheco, 1984). Overall pasture biomass was estimated using a standard formula which relates the relative frequency of each rank to its corresponding DW biomass. I performed univariate analysis of variance for farm and treatments effects on biomass production over time (Wilkinson, 1990).

Tree Biomass Measurements

I measured Erythrina biomass production for four pruning events (June 1989, November 1989, May 1990, October 1990). I randomly selected 10 trees within the grazed plots and six within the non-grazed plots to be evaluated repeatedly on each of the five farms. When trees were pruned, I separated all new regrowth from each tree (growth subsequent to last pruning) into leaf, green stem and woody stem (large branches). I weighed each component in the field for fresh weight (FW) biomass and took a subsample for DW and P content determinations. Remaining prunings along with those from non-sampled trees were left in the treatment plots to be eaten and/or to decompose. Biomass production for leaves, green stems and woody materials was calculated as DW kg/ha. I performed univariate analysis of variance for farm, treatment and pruning event effects 322 on tree component biomass production.

Cattle Dung Biomass Estimates

During two separate grazing cycles (both coincident with tree pruning), I counted the number of dung pats per 900 m2 in G and GT treatments on all five farms immediately after animals had been let out of the plots. I also randomly selected 10 pats per treatment to determine total FW in the field. I then collected subsamples from each dung pat for DW and P content determinations. I calculated both dung DW biomass and biomass P on a per hectare basis.

RESULTS

Erythrina Biomass Production

For grazed (GT) and non-grazed (T) treatments combined over the four pruning events, the woody material produced the greatest amount of biomass (307.48 kg/ha), followed by leaves (227.68 kg/ha) and green stems (58.39 kg/ha). There was a significant increase in biomass production over time (i.e., from the first to the fourth pruning event) for tree components combined and for each component separately (Table 6.2; p < 0.0001). The increase in green stem biomass did not appear until the third pruning event and there was a substantial increase (71%) from the third to the fourth pruning. Woody material production increased gradually (35%) from the first to the second pruning and stabilized between the third and 323

Table 6.2. Changes in Erythrina berteroana biomass production over time.

Pruning Event ______Leaf______Green Stem______W oody Stem ------kg/ha ------1 98.77 d 14.07 c 159.90 c 2 219.60 c 18.71 c 246.19 b 3 270.15 b 45.34 o 409.89 a 4 322.21 a 155.46 a 413.93 a Letters correspond to significant differences (p < 0.0001) within tree biomass components over the four pruning events (Least square means ± SE).

Table 6.3. Grazed (GT) and non-grazed (T) treatment differences in Erythrina biomass production over time. ______

Pruning ______Leaf______Green Stem______W oody Stem T GT T GT T GT ------kg/ha 1 112.42 85.11 13.75 14.38 157.68 162.13 2 272.22 166.97* 22.12 15.29 301.64 190.75* 3 326.26 214.04* 55.34 35.34 531.76 288.02* 4 409.30 235,11* 264.72 46.20* 542.91 284.96* * denotes significant differences (p < 0.0001) between the two treatments for a given biomass component (Least square means± SE). 3 2 4 fourth prunings, suggesting that woody biomass production had reached some sort of equilibrium by the fourth pruning. Leaf biomass, however, exhibited a steady linear increase over the four prunings. There was a significant interaction between treatments and pruning events such that biomass production in the non-grazed treatments increased to a greater extent than grazed treatments with successive prunings (Table 6.3). There was no signficant difference between the two treatments at the first pruning for all biomass components (leaves, green and woody stems), yet by the fourth pruning, the non-grazed treatment was producing more than twice the amount of biomass as the grazed treatment. There were also differences in leaf biomass production among farms over the four pruning events; the trends were the same for the two treatments, however the magnitude of production was consistently greater in the non-grazed treatm ent (Figure 6.2).

There were similar trends in Erythrina biomass P. Again, the non- grazed treatment produced approximately two times more biomass P than the grazed treatment over all tree components, farms and pruning events (0.434 versus 0.222 kg P/ha). There were significant increases in biomass P production for both treatments over time, but the extent of increase was greater for the non-grazed treatment compared to the grazed treatment (Table 6.4). Among the three tree components (treatments and pruning events pooled), leaf biomass produced the most P (0.523 kg/ha) followed by woody stems (0. 325 kg/ha) and green stems (0.135 kg/ha). The relatively high leaf biomass P was probably a combination of significantly higher P concentration (0.223% overall) compared to stem material (0.176 and 0.105 %} and high biomass production. In contrast, woody stem biomass P 3 2 5 Farms

□ 1 2 §3 4 0 0

300

CD

C o o ■o 200 o

CO CO E o m 100

1 2 3 4 Pruning Event (June 1989-Nov 1990)

Figure 6.2. Differences in Erythrina leaf biomass production among farms at each pruning event (treatment means pooled). Different letters denote differences at each pruning (p<0.05). 3 2 6 Table 6.4. Treatment differences in biomass P produced for Erythrina biomass components (leaves, green and woody stems) pooled over time. ______

Pruning Event Non-grazed ______Grazed______------k g p / h a ------1 0.153 0.130 ns 2 0.327 0.196 3 0.406 0.237 4 0.849 0.323 * denotes significant differences (p < 0.0001) between the two treatments for a given biomass component (Least square means ± SE).

Table 6.5. Changes in Erythrina biomass P content over time.

Pruning Event ______Leaf______Green Stem______Woody Stem ------P concentration (%) ...... — 1 0.240 b 0.195 a 0.108 b 2 0.211 c 0.181 b 0.115 a 3 0.192 d 0.157 d 0.091 d ______4 ______0.250 a 0.173 c 0.106 c Numbers with different letters signify significant differences (p <0.05) within a given biomass component; i.e., leaf, green stem or woody material (Least square means ± SE). 32 7 was solely a function of high biomass production because P content was extremely low (0.105%). The amount of woody stem P ranged from 0.208 kg P/ha (farm 2) to 0.485 kg P/ha (farm 4); this amount was still inferior to leaf biomass P production despite lower (25% less) leaf biomass production relative to woody biomass production. Focusing on leaf biomass P, there was approximately 60% more P produced from leaves in the non-grazed treatment than the grazed treatment (0.650 kg/ha versus 0.388 kg/ha). However, there was no significant difference in leaf P concentration between the two treatments (p < 0.946), suggesting that differences in the amount of P produced reflected solely differences in total biomass production. Erythrina leaf P concentration decreased from the first to the second and third prunings but was highest in leaves from the fourth pruning (Table 6.5). In contrast, stem biomass P content tended to decrease over time (for both green and woody stems), implying a shift in P allocation from stems to leaves over time. There were also signficant differences in leaf biomass P among farms (4=1 >3 = 5>2) which remained fairly consistent over the four prunings (Figures 6.3-6.4).

Changes in Pasture Biomass Production

There was an underlying periodicity in pasture biomass production regardless of treatment (Figure 6.5). Some of the variation could be explained by fluctuations in rainfall (Figure 6.6). There w as a significant regression between monthly rainfall and biomass on offer (p< 0.027); however, it was not linear (r2 = 0.015). There was a slight improvement of fit when biomass on offer was regressed against a quadratic of rainfall, Biomass P Production (kg/ha) event. ro o 00 dfeetae frs o ec pruning each for farms differentiates 0.07 of error production over P biomass Erythrina leaf 6.3.Figure ucsie rnns n o-rzd ramn. Standard treatment. non-grazed prunings in successive .0 ~ 0.40 0.00 0.20 0.60 0.80 1.00 1.20 - 1.40 - Non-grazed 1 □ m2 rnn Event Pruning ■ 5 ■ 4 @ 3 § 3 2 Farm 328 Biomass P Production (kg/ha) successive prunings in grazed treatment. Standard error error event. pruning Standard each over treatment. production P grazed in prunings biomass leaf Erythrina successive 6.4. Figure of 0.06 signifies significant differences among farms at at farms among differences significant signifies 0.06 of 0.00 0.20 0.40 0.30 0.70 0.50 0.80 .0 - 0.60 0.10 1 rnn Event Pruning 3 2 Farm 554224824449 329 Pasture Biomass on Offer (kg/ha) iue .. lcutos n atr boas n offer on biomass pasture in Fluctuations 6.5. Figure ros niae re prunings. SE-372). (overall tree rainfall indicate and Arrows biomass between regression among treatments relative to an estimate from the the from estimate an to relative treatments among 40 - 4400 2600 2600 - 3200 80 - 3800 60 - 5600 00 - 5000 ----- 2 6 1 1 1 1 18 16 14 12 10 8 6 4 2 0 rzn Cce (May Cycle Grazing 1—|—i—|—i—|—i—|—i—|—i—|—i—|—i—|—i— - G —*—P GT —G -o T 11 Estimate 99Nv 1990) 1989-Nov 0 3 3 Monthly rainfall (mm) Figure 6.6. Monthly rainfall corresponding to grazing grazing to corresponding rainfall Monthly 6.6. Figure yls vrgd rm euv Fr 1 ad h El the and 1) (Farm Neauev from averaged cycles amn ete sain 6 m at f Neguev). of east km (6 station weather Carmen 200 400 300 500 0 0 7 600 100 0 2 4 6 8 10 911 12 8 13 14 7 15 616 17 5 4 3 2 1 rzn Cycle Grazing 331 332 suggesting a curvilinear relationship between biomass production and rainfall (p < 0.010; r2 = 0.028). When an estimate of biomass production (based on the linear regression) was plotted with observed treatment fluctuations, it appeared that both the non-grazed, no tree control (P) and the grazed, no trees (G) biomass followed the estimated biomass fairly closely (except for the last 3-4 cycles where G biomass increased relative to the estimate). In contrast, there was a clear trend of higher biomass production in both tree treatments, relative to both the estimate and to the non-tree treatments. This was confirmed when cycle means for biomass production were pooled by both treatment and farm (Table 6.6). The grazed tree treatment produced signficantly more biomass than all other treatments and the non- grazed tree treatment production was similar to the grazed no trees treatment. There were also significant differences among farms similar to the trend for Erythrina biomass production (4> 5> 3> 2 = 1). When the time effect was consolidated into biomass production between successive tree prunings (three groupings: 1st-2nd, 2nd-3rd and 3rd-4th prunings), there was no significant time effect on biomass production; i.e., there were no significant differences among the three designated periods (p<0.656). However, there were significant interactions between time between successive prunings and both farm and treatment effects. Although overall production levels did not change over time, the relative production among farms fluctuated over time (Figure 6.7). This was likely a reflection of site variation and differences in grazing pressure. Treatment differences also varied with successive prunings (Figure 6.8). Between 1st and 2nd prunings, there was little difference among treatments; between the 2nd and 3rd prunings, non-grazed tree treatment 33 3

Table 6.6. Mean pasture biomass and biomass P production (grazing cycles pooled) among farms and treatments. ______

Farm N Biomass S Biomass P P conc. (%) -kg/ha ------1 64 3848.89 d 9.111 e 0.237 e 2 64 3929.97 d 9.469 d 0.244 d 3 64 4149.35 c 12.536 b 0.302 a 4 56 4852.73 a 14.147 a 0.297 b 5 64 4527.44 b 11.580 c 0 .2 6 0 c

Treatm ent T 78 4212.96 b 11.968 a 0.286 a G 78 4260.09 b 10.347 c 0.243 d GT 78 4502.73 a 11.833 a 0.262 c P 78 4070.93 c 11.326 b 0.281 b S Least square means with different letters are significantly different (± SE; p < 0.05).

Table 6.7. Mean dung biomass and biomass P by farm and treatment (G and GT); n ^ 2 0 .______

Farm Biomass Biomass P G GT G GT — kg/ha ------1 269.40 171.71 * 1.264 0.838 * a b a c 2 97.35 94.61 ns 0.562 0.576 ns d e d c 3 132.86 152.39 * 0.807 0.954 # c c b b 4 142.35 196.89 * 0.685 1.043 * be a c a 5 155.76 136.85 * 0.642 0.647 ns b d c d Letters below means within a given treatment signify significant differances among farms (p < 0.05). Asterisks signify significant differences between the two treaments within a given farm. (Least square means ± SE). Pasture Biomass on Offer (kg/ha) iue .. am ifrne (E10 i mean in (SE-180) differences Farm 6.7. Figure atr bo s o ofr ewe s sve pruning e ssiv e c c su between offer on ass biom events. pasture 40 - 2400 00 - 4000 20 - 3200 80 - 4800 60 - 5600 60 - 1600 0 - - 800 0 - ucsie rnn Events Pruning Successive - 23 3-4 2-3 1-2

Farm □ 5 ■ 3 § 2 E ED ED

4 1

4 3 3 Pasture Biomass on Offer (kg/ha) 2400 4000 3200 - 4000 iue .. ramn dfeecs S-6) n mean in (SE-168) differences Treatment 6.8. Figure rnn events. pruning atr boas n fe bten uccessive su between offer on biomass pasture 60 - 1600 800 - 23 3-4 2-3 1-2 ucsie rnn Events Pruning Successive 01000153025348484848232301022353535348484848484823230100312302535353014848230100002323531653 914^249^75254544959592250120242422829527681655269294^98284

Treatment GT ■ m Q ■ T □ p

5 3 3 3 3 6 biomass was significantly greater than other treatments, and by the last period, the two grazed treatments had higher production relative to the non- grazed treatments. Pasture biomass P production was similar to overall biomass production; however, there was a significant increase in biomass P among all treatments over time (Figure 6.9). With farm and treatment means pooled, pasture biomass P production increased by approximately 13% from the first period (prunings 1-2) to the second period (prunings 2-3) and an additional 6% between the second and third (prunings 3-4) period (10.114 ± 0.22 to 11.638 ± 0.20 to 12.35 ± 0.23 kg P/ha (Figure 6.10). Overall, there was significantly more P produced in the two tree treatments relative to the non-tree treatments (Table 6.6). Not only was there increased biomass P over time (associated with more mass in general), there was also a significant increase in P concentration (p < 0,0001). For farm and treatment means pooled, P concentration increased from 0.238% to 0.277% between the first and second periods and from 0.277% to 0.289% between the second and third periods. There was no significant interaction between successive pruning periods and treatment effects, implying that treatment differences remained constant over time. Overall, the two non-grazed treatments had significantly higher biomass P content (0.286 and 0.281%) than the two grazed treatments; in both cases, however, the tree treatments’ P content was superior to those of the non-tree treatments. There were also significant differences among farms such that farm 3 with the highest biomass P content (0.302%), although it produced less overall biomass than farm 5, resulted in higher biomass P relative to farm 5 (Table 6.6). Pasture Biomass P (kg/ha) ramns vr ie oeal E12) Arrows prunings. SE-1.27). among (overall tree P time indicate over biomass pasture treatments in Fluctuations 6.9. Figure 12 13 14 10 15 6 1 11 7 6 8 9 2 6 1 1 1 1 18 16 14 12 10 8 6 4 2 0 rzn cce My 99Nv 1990) 1989-Nov (May cycle Grazing 37 33 338

Treatment 14 □ T Q 13 GT ^949 P

«T 12

11

CO CO « 10 o ID © 9

CO o - 8

7 -

6

1-2 2-3 3-4 Successive Pruning Events

Figure 6.10. Pasture biomass P differences among treatments between successive pruning events (se-0.44 for treatment comparisons within given time period). Letters denote significant differences over time. 339

Coincidentally. P concentration differences in Erythrina leaves among farms paralled those of pasture biomass P content; i.e., farms 3 and 4 had the highest Erythrina leaf P content followed by farm 5 and then farms 1 and 2.

Cattle Dung Biomass Production

There were no significant differences in dung biomass and biomass P produced between tree and no tree treatments, yet there were differences among farms (Table 6.7). Farm differences were probably due to unavoidable variations in animal size, age class (cows vs. calves or yearlings) and eating patterns. Dung biomass and P production averaged over all farms was 159.55 kg DM/ha and 0.792 kg P/ha for the grazed, no trees treatment and 150.49 kg MD/ha and 0.811 kg P/ha for the grazed tree treatment. The slightly higher (albeit not significant statistically) dung P produced in the tree treatment relative to the non-tree treatment in spite of slightly lower biomass production may be a reflection of the significantly higher P concentration (p < 0.027, n = 100) in tree treatment dung versus non-tree treatment dung (0.548% in GT versus 0.520% in G). The higher P concentration in dung from tree treatments compared to non-tree treatments, may, in turn, reflect consumption of Erythrina leaves and/or enhanced P content in pasture biomass exhibited in tree treatments. 3 4 0

DISCUSSION

Farm Differences in System Component Biomass Production

Farm differences in production levels were fairly consistent among silvopastorai system components (namely Erythrina and pasture biomass). Such consistency, which was also maintained over time, suggests that the farms possess inherently different below-ground properties even though they are located on the same soil series. When general soil fertility characteristics among farms are reviewed (exchangeable bases and Al, bicarbonate extractable, organic and total P, total C and N), there is a strong suggestive correlation between above-ground biomass performance and soil fertility status (Table 6.8). Farm 4, with the highest pasture and tree biomass production over time, had the highest levels of exchangeable bases, available, organic and total P and organic C. Farm 2, which consistently produced the lowest levels of pasture and tree biomass, had the lowest levels of all soil fertility parameters. These results reinforce the importance of the soil as a chemical reservoir, particularly for limiting nutrients, and suggest that farms with relatively high available P levels will fuel greater nutrient fluxes. Many concur that soil fertility status can have a large effect on pasture biomass productivity (Snaydon, 1981; Humphreys, 1991; Crowder and Chheda, 1982). Specifically, labile soil nutrient pools may influence the rate at which nutrients are mineralized from senescent material and, in turn, the rate of new shoot growth. In addition, soil fertility status may influence species responsiveness to nutrient additions (Crowder and Chheda, 1982). In 341

Table 6 .8 . General soil chemical characteristics of Neguev series surface horizon (0-15 cm) by farm. Included are exchangeable bases (sum of Ca, Mg, K), exchangeable Al, bicarbonate-extractable P, organic and total soil P, t Farm Exch. Exch. Al Bicarb- Organic Total P Organic TKN Bases extr. P P C —ciiy/ny—■ ------%- 1 2.73 0.92 1.95 900.21 1310.55 3.17 0.27 2 1.83 0.69 2.20 716.43 1091.41 3.17 0.27 3 2.08 1.40 3.28 893.06 1507.06 3.19 0.31 4 2.73 0.89 4.05 1055.29 1731.84 3.85 0.31 5 2.09 1.31 2.47 847.28 1284.07 3.50 0.29 LSD 0.27 0.15 0.52 167.40 95.21 0.30 0.05 342 extremely unfertile soils, for example, grasses may not respond to nutrient inputs and their overall production levels may be reduced. In summary, then, nutrients in leaf biomass are strongly dependent on soil pools: the more nutrients in the soil, the greater the transfer and, in turn, the more nutrients will be available for primary production purposes (Budelman, 1989).

Phenology of Silvopastoral System Production

Pasture biomass production exhibited temporal periodicity, but the trend over the 18-month study was stable. Although biomass production did not change significantly over time, there was an increase in both the amount and concentration of biomass P. All treatments exhibited this trend, yet it was most pronounced in the two treatments with trees. This suggests that all treatments are in a nutrient agrading phase and that the mere impositition of rotational grazing (a by-product of experimental implementation; previously, systems were grazed continuously) has had a beneficial effect on pasture productivity. However, the magnitude of the effect has been enhanced by the presence of trees. The higher P concentration in the two non-grazed treatments relative to the grazed treatments also confirms the fallow-like status incurred when grazing is removed (Montagnini et al., 1989).

Erythrina biomass production changed dramatically over time, especially leaf production. Woody branch production increased from the first to second prunings, but leveled off between the third and fourth prunings. There was also a contrasting increase in leaf P content coupled to 343 a decrease in stem P content. The stabilizing woody material production along with a decrease in stem P concentration suggests that, as the trees are maturing and the effects of repeated pruning increase, resource allocation is shifted towards leaf biomass production and leaf nutrient accumulation. This is a common phenomenon for aging, repeatedly pollarded trees; i.e., leaves and small stems contain more than half of the total nutrient pools although their actual mass is inferior to the woody components (Maghembe et al., 1985; Salazar and Palm, 1990). The shift toward increased nutrient levels in the leaf biomass also confirms Erythrina's ability to mine nutrients from the soil.

Relative to other leguminous trees' P content, including Leucaena leucocephala (0.20%), Inga edulis (0.22%), Cassia reticulata (0.31%) and

Gliricidia sepium (0.24%) (Salazar and Palm, 1990; Budelman, 1989), E. berteroana P content from the fourth pruning (0.25%) can be considered intermediate to high. Overall tree biomass production (sum of leaf, stem and branch from fourth pruning was 891.6 kg/ha) is low compared to that of seven-year old Acacia and Albizia planted in the same tree density (1420- 4455 kg/ha) (Budelman, 1989). However, this is not surprising given

Erythrina's comparatively young age.

Predicted P Dynamics Among System Components

Above and below ground system components can be classified as either net sources or sinks for P to predict P dynamics and pasture production in each treatment. In the non-grazed no trees control, there is no P contribution from either dung or Erythrina leaves; the only net source is 34 4 from pasture grass clippings. Pasture biomass production should, therefore, be minimal compared to treatments which have those other P sources. In fact, overall pasture biomass production in the control treatment was lowest among the four treatments. Conversely, there is no export of nutrients from this system; there are no cattle. Therefore, one would predict P accumulation in the biomass over time despite low production levels. Indeed, the control treatment biomass had a higher P content compared to the grazed treatments. When cattle are added to the picture (the grazed no trees treatment), the system receives an additional and concentrated P source (dung) as well as the stimulatory (enhanced C-fixation) effects associated with grazing (Dyer et al., 1991). However, it also incurs some export in the form of cattle biomass. Therefore, one would predict higher biomass production relative to the non-grazed control but lower nutrient accumulation. This is exactly what happened.

The inclusion of Erythrina trees adds another net source of P because the greatest P-concentrating organs (tree leaves) are pruned at the height of their nutrient-accumulating phase. In the non-grazed tree treatment, then, one would predict greater pasture biomass production relative to the control treatment. Overall pasture biomass production was greater than the control but was similar to production in the grazed no trees treatment. The similarity between the non-grazed tree and grazed no tree treatments implies that the effects from dung and/or grazing stimulated pasture growth to the same extent as the application of Erythrina residues alone. In terms of P accumulation, the non-grazed tree treatment should result in enhanced pasture P levels because there is no net export. Moreover, the amount of 345 tree biomass and biomass P produced was twice that of the grazed tree treatment. Lower tree biomass production in the grazed treatment may be the outcome of cattle browsing on accessible regrowth between pruning events. The overall result in the non-grazed tree treatment was the highest pasture biomass Pproduction and P concentration of the four treatments. Finally, the combined effects of trees and cattle should result in tw o concentrated P sources: Erythrina leaves and cattle dung. This should result in overall higher pasture biomass production relative to the other treatments. Indeed, this treatment produced the highest levels of pasture biomass among the four treatments. Biomass P was also high relative to the non-tree treatments. However, because there is some P export from the system, P concentration in the pasture biomass was less than both non- grazed treatments. Nonetheless, pasture P content was higher than the grazed no trees treatment. This could have resulted from direct P release from decomposing Erythrina leaf residues or from P-enriched dung. Dung in the tree treatment had a higher P concentration than dung in the no-trees treatment, presumably because cattle fed on relatively high P content

Erythrina leaves. In summary, the grazed tree treatment had the greatest nutrient turnover which resulted in greater pasture biomass production, whereas the non-grazed tree treatment had the greatest nutrient accumulation; greater P storage at the expense of reduced turnover. It appears that Erythrina residues provide sufficient P to result in greater availability for pasture utilization and that grazing stimulates the cycling of P. 3 4 6 CONCLUSIONS

1. There was an increase in Erythrina leaf biomass production over the course of four successive prunings. Non-grazed treatments produced approximately twice as much tree biomass as grazed treatments. For both treatments, P concentrated in the leaves at the expense of woody components over time. It appears that tree leaf productivity has not yet reached equilibrium and is still increasing after more than two years.

2. Pasture biomass production was greatest in the grazed treatment with trees, followed by the grazed no trees treatment. Between the two non- grazed treatments, the treatment with trees produced more biomass than the no tree control.

3. Pasture P production and concentration was greatest in the non-grazed tree treatment, confirming the hypothesis that Erythrina residues would promote increased P availability for pasture uptake. Nutrient turnover and subsequent pasture biomass production was greatest in the presence of trees and grazing, suggesting that grazing stimulates P cycling.

4. Inherent soil fertility, particularly inherent P pools, appears to be an important contributor to P cycling. If the soil base provides sufficient nutrients for above-ground incorporation, nutrient turnover from plant and animal residues will be enhanced, resulting in greater productivity overall. SUMMARY AND CONCLUSIONS

Changes In Soil Chemical Properties Over Time

The Costa Rican silvopastoral system of Erythrina berteroana and native grass pasture has undergone notable changes below ground over the 40-month study (baseline measurements taken in August 1987 and final measurements in December 1990). The significant declines in soil pH and exchangeable monovalent cations with concurrent increases in extractable Al and Fe suggest that the soil system is still undergoing an equilibrium shift from forest to pasture ecosystems, even 15-20 years after rain forest clearing. Although the most dramatic changes in soil chemical properties occur within the first five years following forest clearing, more gradual changes associated with nutrient leaching are possible over the period described (Bushbacher et al., 1988; Werner, 1984). Alternatively, it is also possible that the observed changes in soil chemical properties, when viewed in a longer time-scale context, merely represent normal fluctuations associated with the pasture ecosystem's own dynamic equilibrium. The significant increase in soil C relative to N in only the non-grazed tree treatment in both surface (0-15 cm) and subsurface (15-30 cm) horizons implies that organic C is accumulating to a greater extent than N in this system. In addition, N may not have accumulated as much as C because N mineralized from decomposing Erythrina leaves may have stimulated mineralization of soil organic N reserves, resulting in a temporary

347 348 decrease in TKN (Boernemisza, 1966). The presence of leguminous trees and tree pruning increases the C:N ratio in the surface soil to a greater extent than recycling from pasture grass clippings alone (control treatment). In summary, the increase in the C:N ratio in the non-grazed tree treatment relative to the grazed treatments supports the hypothesis that the lack of grazing is analogous to fallow conditions and that leguminous trees enhance the fallow status of the system. Since there is no nutrient or organic matter export from the non-grazed system, one should expect organic C accumulation over time. Although there was a treatment-induced (non-grazed treatment) increase in soil organic C, organic P reserves decreased among all farms regardless of treatment. Simultaneously, available P (NAHCO 3-EDTA extractable) increased. With treatment means pooled by farm, there was an increase in P mineralization from the experiment's initiation to its end. This increased mineralization was either another manifestation of gradual soil chemical changes associated with post-deforestation equlibrium shifts or an indication of enhanced nutrient turnover associated with changes in pasture management (from continuous to rotational or no grazing).

Importance of Biological Mediation of P Cycling

The Neguev soil series has a high P retention capacity (> 2000 mg P/kg soil as determined from P sorption isotherm). The soil's highly weathered mineralogy (kaolinite, gibbsite and iron oxides), low pH and high Al content contribute to its propensity for P sorption via ligand exchange and precipitation of Ai-P secondary minerals. It is not unusual, then, that the biotic components of the pasture ecosystem rely strongly on 349 biologically-mediated mechanisms to keep P in bioavailable form. These include symbiotic association with endomycorrhizal fungi and rhizosphere- induced enhancement of soil solution P levels. Mycorrhizal fungal (VAM) infection of Paspafum conjugatum, one of the dominant grass species in the "pasto natural" complex, increased P uptake and growth to a greater extent than VAM infection of Homoiepsis aturensis, a less "desirable" (from farmer's perspective) grass species. The difference in growth response to mycorrhizal infection between the two grasses may be due to differences in VAM species-plant host effectiveness or differences in plant dependence on VAM for nutrient acquisition and growth (facultative versus obligate mycotrophy).

The silvopastoral system's leguminous tree, Erythrina berteroana, when grown from seed, was highly responsive to inoculation with VaM fungi. Rhizob/um nodulation, P uptake and biomass production were strongly enhanced by VAM fungal infection. The lack of nodules on non-

VAM inoculated seedlings implies that E. berteroana could not survive in the field without mycorrhizae; i.e., it is probably an obligate mycotroph. In contrast to seedlings, E. berteroana grown from vegetative stakes responded weakly or negatively to VAM inoculation. This negative growth response demonstrates the dynamic cost-benefit relationship between plant host and fungal symbiont. In seedlings, root establishment is critical to ensure adequate nutrient (P) uptake for growth; therefore, the nutrient- acquiring benefits derived from seedling association with VAM over-ride the costs of host fungal maintenance. In contrast, the Erythrina stake, during the early growth phase, has sufficient nutrient reserves to maintain itself while roots get established. As such, the cost of C drain to maintain the 350 fungus outweighs the potential benefits derived from fungus-mediated nutrient (P) uptake. The relationship between costs and benefits could shift, however, once mycorrhizal roots become the major conduit through which limiting nutrients like P are acquired. The other important biological mechanism for enhancing soil P biovailability in the pasture ecosystem appears to be rhizosphere induced; i.e., root exudates of organic acids which complex with Al and release P from A1-P compounds, phosphate-hydrolyzing enzymes which break organic P bonds, and general P leakage associated with root P uptake. An alternative explanation is that live pasture roots which are most dense in the upper soil stratum simply displace and/or minimize the mineral soil fraction which contributes the most to P retention. In either case, the existence of rhizosphere-induced enhancement of soil solution P was deduced indirectly from experiments both in the greenhouse and in the field. In the greenhouse decomposition study, any P mineralized from decomposing residues was barely detectable in the soil solution of bare soil, whereas under sod, there were significant P pulses in the soil solution. In addition, there was an underlying periodicity in soil solution P levels under sod without decomposing residues present; there were no such fluctuations in bare soil with no residues. Periodicity in bioavailable P was observed among all treatments in the field as well and was only detectable in the upper 2.5 cm. Both greenhouse and field measurements provide strong supportive evidence that not only do pasture roots maintain more P in soil solution, but they exhibit some phenological pattern of root exudation and nutrient (P) uptake. 351 Erythrina berteroana’s Ability to Enhance P Cycling

Erythrina berteroana*s ability to enhance P cycling in the silvopastoral system has been demonstrated in several ways. Its association with VAM fungi resulted in greater biomass and biomass P production. Since E. berteroana is probably obligately mycotrophic in the field, mycorrhizae likely play a critical role in enhancing P cycling through greater leaf P production.

Secondly, Erythrina leaves decomposed significantly faster than both pasture grass clippings and cattle dung. The amount of P released from

Erythrina was also significantly higher than that from grass residues. Although the magnitude of dung P released into soil solution was 4-5 times greater than those from both plant residues, the overall impact of dung on P cycling in silvopastoral system is highly localized and, therefore, probably of reduced importance relative to Erythrina residues.

Lastly, Erythrina leaf biomass and biomass P production as well as leaf P concentration increased linearly over the 18-month study period. This suggests that, as the trees and their response to repeated pruning mature, more P is concentrating in the leaf component. Since the leaves are the tree's greatest nutrient source, in terms of both livestock forage and pasture mulching purposes, Erythrina's potential for supplying P (becoming a net source rather than net sink because of pruning management! appears to be increasing over time.

Effects of Tree Pruning and Grazing on Pasture Biomass Production and Management Implications

Although grazing effects overshadowed tree effects on soil P cycling early in the silvopastoral system's establishment, tree pruning and grazing 352 combined increased the magnitude of P fluxes by the fourth tree pruning. This translated to greatest pasture biomass production in the grazed tree treatment relative to the other treatments There appears to be a synergistic relationship between tree pruning (and subsequent decomposition of tree residues) and grazing such that nutrients (P) released from decomposing

Erythrina residues are made more bioavailable through grazing. This could result from several sources: soluble P released from dung, compensatory pasture root dieback and diminished pasture P uptake following grazing. Enhanced soil solution P levels, in turn, probably lead to greater P uptake of pasture rebounding from grazing which results in greater above-ground pasture biomass production. Overall, then, tree pruning and grazing combined promote nutrient turnover which stimulates above-ground pasture biomass production. In contrast, the non-grazed tree treatment fostered greater P accumulation at the expense of biomass production. In summary, the presence or absence of grazing results in a trade-off: greater biomass production with lower P concentration when grazed, versus less biomass produced with greater P accumulation when not grazed. Regardless, leguminous trees and tree pruning promote greater nutrient (P) turnover and subsequent increases in pasture biomass and/or greater nutrient accumulation compared to both no­ tree systems. Whether enhancing P cycling or increasing pasture biomass and biomass P production, inherent levels of soil P pools augmented both tree and grazing effects. Soil organic P levels seemed to be positively related to the magnitude of labile soil P fluxes associated with tree pruning and grazing. Available (NaHC 0 3 -EDTA extractable) and total soil P levels 353 appeared to be important determinants of above-ground pasture biomass production when compared among farms. Such results substantiate the hypothesis that the silvopastoral system, during its initial establishment phase, needs some minimum (threshold perhaps) soil P resource base to fuel P cycling both below and above ground. Perhaps as the silvopastoral system and, in particular, the tree component reach some steady-state equilibrium, organic matter levels will increase and the reliance on native soil P pools to prime P cycling will diminish.

In terms of pasture management, the field experiment provides strong evidence in support of planting leguminous trees in pastures (specifically

Erythrina berteroana), even in the short term. In addition, there is a need to build up the soil’s P fertility so that tree effects can be maximized. This can be achieved by placing certain grazing areas into fallow and maintaining a clipped pasture planted with leguminous trees. Alternatively, farmers could add a single dose of P fertilizer (either via manure or inorganic synthetic additions, but sufficient to overcome the soil’s P retention capacity) to give the biologically-mediated P cycling an initial boost. Managing via fallows is obviously less expensive for the farmer, but the benefits are attained more gradually and some grazing area would have to taken out of production. Fertilizer additions produce more immediate effects but the costs may be prohibitive. If neither strategy is feasible for resource-limited farmers, the short- and long-term benefits derived from simply planting leguminous trees in grazed pastures, in terms of supplemental forage and enhanced pasture biomass production, may outweigh the increased labor costs associated with tree planting, pruning and management. APPENDIX A

ADDITIONAL INFORMATION FROM CHAPTER I

354 355

Table A.1. Mean monthly temperature minima and maxima (°C) and total monthly rainfall (mm) recorded

Month Min Temp Max Temp Precipitation!

Jul-89 22.17 30.52 nd Aug-89 22.58 32.87 nd Sep-89 22.57 33.02 nd Oct-89 22.46 32.40 168 Nov-89 22.70 31.52 314 Dec-89 21.97 31.94 405 Jan-90 21.88 30.40 480 Feb-90 20.91 30.66 59 Mar-90 21.92 30.61 362 Apr-90 22.64 31.78 146 May-90 23.62 31.46 817 Jun-90 23.78 31.41 531 Jul-90 23.23 30.67 331 Aug-90 23.62 30.82 588 Sep-90 23.81 31.71 206 0ct-90 23.10 32.87 184 Nov-90 23.70 31.67 409 Dec-90 nd nd 376 October 1989 includes only 21-31 October, not the entire m onth's rainfall. 356

Table A.2. Neguev soil series chemical and physical characterization (source: Wielemaker, 1990a). ______

Profundidad 0-4 •"14 14-29 29—04 •4-99 pH (11,0) 4 .7 4.2 4.2 2 .5 2 .a pH(XCl) 4 .0 2.0 2 .9 2 .9 2.4 pH(HaP) 0.4 O.C 0 .9 9 .0 9.1 Acidts cxtr (aaq/ioog) 1.0 3.0 2.C 2 .0 2.0 CXC (aaq/lOOg) 11.0 2 7 .C 25.0 24.7 2C.« Znttreuble c«tionM Ca (aaq/ioog) 1.2 1.0 1 .9 0 .5 2.2 Ho (Mq/100g) 2 .5 0.5 0.2 0.2 0.1 K (aaq/lOOg) 1.0 0.7 0 .7 0.2 0.1 Ha (aaq/lOOg) 0.2 0.2 0.2 0.2 0.2 Hatarla orgdnica (%) 0.9 5.0 2 .9 1 .1 0.7 Extr. per deido oxal. Tm (%) 1.1 1.2 0.9 0.75 0.95 A1 (%) 0.4 0.45 0.55 0.45 0.70 Ratancidn da P (%) 00 00 94 95 9C T axtura Arcilia («) 42 •• •0 70 72 Liao («) 20 22 24 22 10 Arana (%) 27 11 15 7 10 Dansidad aparante o.oc 0.02 0.79 Pat. agua a pP4.2 (%) MAPA DE 357 SUELOS

Figure A . 1. Soil map and legend for the Neguev settlem ent (from de Bruin, 1991). 00 It) CO t* '< * t |: | M j:* A * H .♦ I t Hth ;; lit «f v\h\ Ijiii! sfji: H r 5 fii in; f t t|! »«- i ■■1111 5 :1 :*U ill!;!!!-* i!IM iiliij. ij;,,,,H *» p i ! m *h :i :u<{ i JU:!l h 1 Cl !j 0 0 m il *3 !> I iliii s : s : : s 3S8SS a* • -i1! . *1 l!“I K ,i ? !; Shff* w l- < il i 5 i f a a * 4 f a Jill? >ia 5:5 s:i:i l iiij !; !| &;=!*) il ilill!* •«**»i *{.i 1 • m i \ w , fii!! <'I i T 11 ! ii*fl *! ! |ii« ij ifi! •t: m !* ! fii Mi f! iiil IlIiHs iliii! iliii liHul if!

II 2 B -IB B B j IB |B \ I APPENDIX B

ADDITIONAL DATA FROM CHAPTER III

359 Table B. t. Concentatfons ot additional nutrients including K, Ca, Mg, Cu, Zn, Mr in boh Inoculated and non-lnoculated treatments by plant specie*.

Shoot ______Root Treatment n K C* Mg Cu Zn Mn K ca Mg Cu Zn Mn

Paspalum 7 1.1910 46 0.1410.02 0.4310.03 2515 4316 13261126 0.3410.09 0.010.04 0.161 0.02 72114 104146 15291423 Inoculated

Paspakm non­ 7 1.771 0.3* 0.1& 0.01 0.391 0.09 2317 4615 11291 131 0.4010.14 0.051 0.03 0.211 0.04 52120 125142 7631434 inoculated

Homotop&is inoculated 4/1 5 2.1M0.21 0.1310.02 0.331 0.06 2216 6617 12651 57 0.95 0.03 0.16 76 134 1320

Homdepsis non­ 0 * inoculated Erythrina cuttings 2 2.791 0.16 1.0110.12 0.S11 0,04 1011 5614 28801 234 2.351 0.99 0 331 0.05 0.311 0.15 55116 1261 X 10001 26 inoculated 9

Erythrina cuttings non­ 2 2.461 O X 0.761 OX 0.4210.16 711 4615 14231 579 2.6310.10 0.641 0.54 0.361 0.05 5914 1201 17 146011047 inoculated

Erythrina seedlings 2 1.4010 70 0.621 006 0.571 0.02 1616 9613 15211 1063 1.3510.21 0.371 0.05 0.461 0.06 190193 1051 21 560165 inoculated Erythrint seedlings 1 1.91 0.35 0.45 10 79 972 1.35 0.21 0.21 16 74 400 non­ inoculated

Erythrina source bee | 2 3.001 0.71 2.061 0.49 0.591 0.05 B±0 4714 1001 0 360 5 n «4 tor shoot samples and n >1 for root samples. # not enough plant material to perform chemical analyses. 9 values tor boh Erythrin* grown from cuttings and seedlings are weighted averages from leaf and stem concentrations. | Erythrina source tree values represent leaf material collected from the tree used to make vegetative cuttings; no roots were collected. Table 6.2. Mean biomass (g) and nutrient content (mg) parameters for the two grass species and for E. berteroana grown from seed and vegetative cuttings. ______

Paspaium conjugatum Homolepsis aturensis Parameter non- inoculated inoculated non-inoculated inoculated n = 6 n = 7 n = 6 n_2 _fi Shoot mass 0 2.04 (0.84) 5.04 (1.70) 0.09 (0.07) 0.94 (0.83) Root mass 0.61 ( 0 .2 0 ) 1.70 (0.63) 0.07 (0.06) 0.41 (0.33) Total mass 2.65 (1.031 6.73 (2.25) 0.16 (0.13) 1.35 (1.15) R:S ratio 0.31 10.05) 0.34 (0.07) 0.83 (0.39) 0.75 (0.73) Root volume 3.00 10.95) 10.17 <4.22) 1.07 (0.65) 2.70 (1.29) Total P 2.53 10.81) 6.63 (1.79) ...... 5

Erythrina berteroana seedlings Erythrina berteroana vegetative cuttings Parameter non-inoculated inoculated non-inoculated inoculated n == 4 n == 7 a = 5 n - 5 Leaf mass 0.04 (0 .0 2 ) 0.70 (0.54) 3.16 (0.31) 2 .8 8 (0.47) Stem mass 0.16 (0.05) 1.42 (0.81) 3.84 (0.58) 3.03 (0.55) Shoot mass 0 .2 0 10.06) 2 .1 2 (1.34) 7.00 (0.41) 5.91 (0.74) Root mass 0.13 (0.06) 0.84 (0.33) ...— — — R:S ratio 0 .6 6 (0.27) 0.46 (0.14) ...— — — Total mass 0.33 (0 .10) 2.96 11.64) — — ...... 5 Total P 0.73 (0.23) 6.58 (2.81) 13.89 (2.30) 10.27 (1.96) Total N 34.18 (11.751 174.95 (102.49) 768.02 (114.01) 660.43 (190.89) $ not enough material to determine. 0 all biomass parameters in g, P and N in mg, root volume in cm3. Numbers in parentheses are standard deviations. APPENDIX C

ADDITIONAL DATA FROM CHAPTER IV

362 u u m u n n m m iifiiitiiliH nrnf I ii- 7 js |

i l l

I llilliiiiilliilgiiilliiisiiiigiiililliiliiliiiliiil |4 f III r* £ £ t ill a f p 28 28 1581 u 8 SSI 3 31

g 3! il CO m CO Eaotf-P 1/x 7 0.490 2 1 7 6 0.0GB Eaod-P tog(«) 7 0.466 2 2 6 4 -0 3 4 4 Eaod-P • o n to 7 0.420 2 3 2 2 •0.062 Eaod-P x(nag axp) 7 0367 2 2 9 3 -O.OIB Eaod-P x*2 7 0L311 2 2 9 4 •0303 Eaod-P 3 -p a w n 6 0391 2261 0 3 3 2 3 Eaod-P 4-param 6 0 3 6 3341 •1376 0311 0.273 Eae*-P 1/* 7 2311 2 1 0 7 2 1 1 7 Em HP to0t»> 7 1367 2 2 7 9 -2101 EaoH-P •***> 7 1.766 2441 -0.146 EaeH-P x (tmq axp) 7 1326 2 3 4 4 •2044 EaoN-P x*2 7 1341 2 2 0 2 •2006 EaeH-P 3 p r a n 6 1.797 2 2 0 4 •0306 1 3 3 3 EaolFP 4-param 6 2311 2 9 0 0 •1390 0.479 ELSaod-P 1/X 7 0360 1.726 2 1 6 6 EL8eod-P ■oOt*) 7 0392 1 3 2 6 -0.102 ELSaod-P acn(x) 7 0 3 4 3 2 0 6 •0.126 ELSaod-P x(nog axp) 7 0 3 9 3 1 3 6 •0333 ELSaod-P **2 7 0337 1396 -0 3 0 4 ELSaed-P 3-param 6 0316 1.774 0 3 1 2 -4 3 0 4 EL8aod-F 4-param 6 0 3 6 6 3 3 3 0 -1 3 1 3 2301 -0.101 FI fianlFP 10c 7 0 3 6 6 1.72 2 1 7 3 ELSaoHP taSOO 7 1373 1 3 9 3 -211 ELBaoH-P aqrtO*) 7 1.144 2066 •2134 ELSaoO-P x(nagwq>) 7 1310 1 3 6 3 J ELSaeH-P 7 1330 1 3 9 -0 3 0 3 ELSaefrP 3-param 6 1399 1.77 2 0 4 6 -2 3 6 8 ELSaoH-P 4-param 6 1303 2 4 4 0 •1316 2 0 0 3 •0.106 Paod-P 1/x 7 0 3 2 7 2371 2 1 3 6 Paod-P loQO*) 7 0 3 9 7 2 6 6 6 -2102 Paod-P aqrtOt) 7 0.736 2 9 0 6 •0.134 Paod-P x (n a g a v ) 7 0 3 6 6 2 3 0 -0336 Paod-P x"2 7 1391 2 3 2 7 •2004 Paed-P 4*param 6 0372 3 3 9 4 -1.026 0.109 0.01 PaoH-P 1/X 7 4 2 7 9 2 4 6 9 ft m i PaoK-P toO0«) 7 0 2 8 6 2 3 7 3 •2067 PaoH-P aqrtfx) 7 0 3 1 5 2 3 4 3 •2 0 7 PaoH-P x(nag axp) 7 0366 2 3 8 6 •2016 PaeH-P x*2 7 0 3 0 7 2 3 6 2 -2002 PaeM-P 4-param 6 0372 2 4 0 -0317 0 -0.067 D aed - P 1/x 7 2 3 6 4 2361 2 0 6 6 D e o d -P *og°0 7 14047 4661 4X063 x(hag axp) 7 18087 4 6 0 7 •0.016 x*2 7 22087 4 4 7 7 4X002 3-param 6 8.16 4 4 0 * 0073 -1.184 4-param 6 16064 x y q 4X103 0.006 -0.062 1ft 7 18.136 4.461 0041 tofl&O 7 21082 4 0 0 2 -0.018 »qn(x) 7 22.461 4 0 1 6 4X018 x(nag axp) 7 23.068 4 4 8 6 4X004 x*2 7 23068 4 4 8 8 0 3-param 6 14.806 4 4 6 6 0.006 -4.133 4-param 6 28036 4 0 1 7 4X116 0 •0 0 1 8 1ft 7 11047.13 7.014 0241 logfx} 7 7460017 7 0 3 3 -0.177 •VKx) 7 1181606 7071 4X228 xO0(x) 7 3361402 7064 4X228 Eaoll-C aqHW 7 26186.06 7.719 -0.323 Esoil-C x(rtag axp) 7 23521.44 7.406 -0.006 Esoil-C x*2 7 31047-36 7.372 -0.012 Eaoll-C 3-param 6 27273.64 7.524 -0.124 0.885 Esoil-C 4-param 5 32834.66 7.673 -0.269 0.66 0.066 ELSsod-C 1 /X 7 8456.294 7.203 0.174 ELSsod-C •ogW 7 3076.159 7.623 -0.127 ELSsod-C sqrt(x) 7 6606.006 7.694 -0.166 ELSsod-C x(nag axp) 7 9100.183 7.667 -0.046 ELSsod-C xA2 7 16260.16 7.406 -0.006 ELS«od-C 4-param S 6660.801 7.607 •0.084 0.091 -0.119 ELSsoN-C 1/X 7 103066.4 7.216 0.224 ELSaoll-C tog(x) 7 92506.69 7.519 -0.171 ELSsoil-C aqrt(x) 7 97306.38 7.754 -0.225 ELSsoN-C x(nag axp) 7 108311.6 7.576 •0.061 ELSso»-C x*2 7 132976.6 7.466 -0.006 ELSsoil-C 4-param 6 120536.8 7.614 •0.005 0.116 -0.16 Psod-C 1/x 7 46675.62 7.300 0.240 Paod-C tog(x) 7 18246.27 7.662 -0.197 Psod-C sqrt(x) 7 16373.54 7.041 -0.269 Paod-C x(nag axp) 7 24583-34 7.743 -0.077 Psod-C x*2 7 53448.17 7.627 -0.000 Paod-C 3-param 6 18376.4 8.12 -0.451 0.343 Paod-C 4-param 5 22083.12 8.077 •0.406 0.366 -0.003 Psoll-C 1/x 7 8253.814 7.445 0.172 PsoU-C log(x) 7 3319.486 7.673 •0.126 Psoil-C sqrt(x) 7 7214.868 7.841 •0.162 Psoil-C x(nag axp) 7 14229.96 7.711 •0.043 Psoll-C xA2 7 27673.86 7.638 -0.004 Psoil-C 4-param S 4648.161 7.767 -0.084 •0.018 -0.128 Osod-C 1/x 7 1286423 7.493 0.106 Dsod-C tofl(x) 7 4467.947 7.643 -0.066 Dsod-C »qrt(x} 7 3141.313 7.766 -0.115 Osod-C x(nag axp) 7 4350.281 7.678 -0.032 Dsod-C xA2 7 0806.396 7.627 •0.003 Dsod-C 3-param 6 3665.22 7.765 -0.114 0.501 Dsod-C 4-param S 4146.443 6.127 -0.474 0.291 0.067 Dtoil-C 1/x 7 4821.172 7.51 0.080 Dsoil-C tog(x) 7 1329.235 7.63 -0.067 Dsoil-C sqrt

§ RESID/SURF induda* rasidua typs, placement surtao* (tod or bar* toil) and m ast or alannant. E - Erythrina laavas ELS > Erythrina laavas and atoms P > pastura grass O - dung i mhhhmmn **'* I

9909000—0 8M«888Mo« 0000090 — 0 0009000—9 I NWWMMf* WW 0 0 0 0 0 0 0 9 0 w yuuu66*« 90000900 — 900000000 I ********* 288848883 * * * £ * * 3*2 8828=6338 823388823 o ^ o o o o o o o8 « o * e o o o o 0 0 0 0 * 0 0 0 0 o o o a o a o a o 000090009 « l i S 8 «38 S 51*353 * 8*8 3*S 9 tt**£8 688888868 382 688=38

359123158 2 88828235 •«8 * 55SS 9 882228325 322288285 4** 2***** 282838388 * 8 * 8888*8 288888388 822338288 k->*ua««Ha 99 — 0 0 9 0—9 — 00900900 9 0 = 0 9 0 0 0 0 SH 8326888 882223238 868328388 £ 8 * * * 8 * = * 666882228 o o a o 0 « o a « ■ 00990000 090009000 000000999 282232383 8 8 38 28888 828388882 ********* 883823883 o o ■ a □ o a a 990909909 000099000 009000000 2 8888388 38 B9 flg338 888333888 8888 * 8**8 883383388 Muuaaauttw — 009 — — — — 9 moommnhvh 0 0 0 0 0 0 0 0 0 a s a c K H s ^ i 833833323 28=382288 8 * 858 * 8*8 886638633 a o a • a o a * 099009090 009000009 090000000 8 883 S 88 S £38838888 668683282 ** 98 ** 88 * 883888866

888288888 282888888 288228888 888588884 228822228 £883383 X1 868283883 888388888 * 8588***8 868688688 o a o a a o o a 000090009 090000099 « o o o * » « e «000990900 882828883 8 88 X8383 338888886 226888883 *«* = 4** 8 * * 0 a 90 i 8 8 8 a a a a a a a a S 686363338 = 6 0 6 6 = 6 0 6 mimmmm + mmm 0 9 0 0 8 0 0 0 8 ' • aa-^awau- — 0 — 99 0 0 0 0 0—0900000 9 0 0 0 0 0 0 0 — I oa-> -a « a a a — 9000 — 9 — 0 0900 — — 900 • « a a a - a * o 90 — — 900 — 9 009900900 **W»«18>N*0 SSSSSoSSS

2=2888268 883888383 838322286 228222385 868838855 s ^ a a a a a a ^ 999900990 900—000 — 0 0—0090000 9 9 —0—00 —0 a l S a - a ^ a a 90—090099 0 0 0 0 0 6 - 0 — •- 0.- 263. 0 9 0 9 = 6 0 8 0 ^••*-a"*»aa 90000 — 900 00009—900 O0O—09000 0909 — — — 00 s s i i S i s i i m m m I l s i S i i i S i s i i i H i i i l i i i i i U 000090000 a 5S 2 o So3 - S 3o885900 688 - 5X220 258 i 2s - 5o s S S o i Z s i 0 00900—090 000090009 0 0 0 0 0 0 0 0 0

£ i . i i i H a a s s 3s 83 s 53 5583 X2825 531XX3IXX 5238=5558 638863683 3 3888384 628838333 886328368 888888882

— 8 o o = ■>•6 o 0 0 9 0 0 0 —80 02-228889 9 8 8 0 = 0 090 8 3888288 338238882 283 68 888 3 222836833 688883822

S ! a £ Z S I 33 838338283 888888883 882883386 668 626236 822838388 828686682 838888886 888632888 288828822 - 0 0 0 0 —0*0 9 —099099 — 00000900 — — 0- — 00 — 0 8 8388288 382882288 228833868 283688828 62=223832 * 0 9 ■»* 0 m s 8 8 ■9 3 O 0 O 0 0 s 3 S 2 2 L 9 £ RF WEEK HASSta) SO MAREU SO P SO N SO C SO LKW P SO N SO C SO CP SO CM SO UQ.N — % ■ — — tno-

3 0 4 30 000 10000 000 019 000 269 016 4721 013 1302 62 00 1243 67 20000 39 246.47 067 1636 092 4.31 2 1 422 030 9614 691 013 003 174 036 4336 016 54 06 729 109 19222 126.4 36679 7661 2692 379 2 2 333 004 62 44 062 022 001 239 o n 4635 012 76 03 646 46 16300 207 21660 636 19.S3 082 2 3 33* 0 13 63 26 296 016 004 207 042 4646 004 63 11 737 123 1063 6 604 30767 70.43 2296 466 2 4 335 004 77 61 099 019 001 190 016 4666 015 64 04 636 44 15636 249 24637 1912 2403 210 2 5 297 016 6907 426 015 002 166 4667 43 09 264 374 6627 9972 35900 23.06 2 « 3 19 033 74 19 736 020 006 236 036 4713 004 61 09 749 101 1509 9 1522 249 79 6322 2041 479 2 7 30* 022 71 51 310 021 004 211 030 46 49 0 31 63 06 644 47 14299 1115 23034 4124 2233 337 2 • 314 001 7309 033 016 009 200 020 46 63 036 55 07 626 63 14704 54 266.71 34 89 2334 252

2 0 331 000 100 00 000 026 000 206 004 4312 007 331 136 00 1102 19 23322 36 16692 027 2117 039 256 2 1 324 0 01 96 66 0 27 023 001 136 003 42 66 029 126 03 625 24 22346 212 17415 6 21 2709 104 2 2 446 006 6310 120 030 001 169 006 4269 006 131 01 640 23 1901 6 306 144 74 373 2266 099 2 3 426 006 60 23 107 026 001 194 002 4305 041 119 04 762 01 16339 69 13391 631 2346 006 2 4 403 0 14 75 69 266 029 001 193 004 4323 006 115 01 776 13 1742 7 592 15176 339 2247 0.39 2 3 4 11 011 77 40 2 13 030 001 166 009 4314 009 123 02 771 59 17729 526 14396 709 2303 109 2 « 4 16 0 16 76 25 306 026 001 167 006 4306 006 114 0.2 777 07 17692 736 13664 434 2304 074 2 7 393 033 7319 633 032 003 197 030 4263 047 123 00 766 51 16627 1686 134 50 1333 2209 366 2 • 383 006 72 50 107 030 000 161 001 4314 020 116 02 607 05 1660 9 320 14390 066 2364 030

2 0 321 000 100 00 000 040 ooo 167 003 40 72 0 13 206 00 972 26 2121 3 70 10179 034 21.94 065 703 2 1 326 007 <00)6 1 36 040 001 136 001 36 90 0 43 21 0 OS 619 07 20969 516 9962 464 29.66 040 2 2 511 010 96 06 190 046 004 161 006 36 56 004 233 23 925 11 20210 413 6716 666 2166 071 2 3 469 006 6912 122 042 001 160 001 39 76 026 197 09 643 16 16634 137 94 77 376 2210 031 2 4 4 71 006 90 40 163 042 001 164 003 36 30 023 19 6 03 967 29 16606 436 94 11 377 2147 019 2 3 477 0 01 91 46 0 14 044 003 191 006 30 29 016 21 0 14 906 32 16719 106 69.46 336 20.63 060 2 « 4.62 0 11 66 66 217 044 002 167 010 36 26 016 201 IS 663 67 16137 319 6034 403 21.02 1.09 2 7 463 004 6667 061 046 001 193 002 3601 003 222 OS 691 02 16059 166 tt-30 250 2026 025 2 • 463 003 69 23 034 044 003 166 002 3632 040 206 14 672 15 1791 1 75 6776 2033 044 Mew* end dewelow (SO) el iw ww legs ee rwidw Dpi pw d*» RESIDUE 1*ERYTHR1HA LEAVES. 2* ERYWfflNA LEAVE S6STEMS, WASTURE. 4-DUNG SURFACE: 1-SOO 2-flARE SOL 368 369 TABLE C3. Models lor field decomposition data (mass and nutrient loss) Including residual mean squares (RES MS) and model paramei-y-lntercept, B- slope parameter, C- shape parameter and D ■ 4th parameter. Only significant model fits are Included.

RESIDUE MODEL DF RES MS MODEL PARAMETERS B D•Mass 1/x 14 1.788 1.795 0.195 D•Mass iog(x) 14 0353 2.126 -0.102 D Mass •qrt(x) 14 0348 2.192 -0.068 D Mass x(neg axp) 14 0.133 2.093 -0.007 D Mass x*2 14 3302 2319 0.000 D Mass 3-param 13 0.117 2.066 -0.002 1383 D Mass 4-param 12 0338 3.736 -1.652 0.193 0381 D P 1/X 14 0.831 0372 0320 D P log(x) 14 0.196 1372 -0351 D P sqrt(x) 14 0.120 1333 -0.175 D P x(neg exp) 14 0.199 1304 -0.021 D P xA2 14 0.455 1.168 0.000 D P 3-param 13 0.127 1396 -0331 0.441 D P 4-param 12 0.148 3.495 -2.120 0.180 0380 D N 1/X 14 0.046 0.097 0.183 DN tog(x) 14 0327 0367 -0.063 DN aqrt(x) 14 0315 0.420 -0.055 D N x(neg axp) 14 0.009 0341 -0.006 D N x*2 14 0.009 0379 0.000 DN 3-param 13 0.008 0309 -0.001 1.446 D N 4-param 12 0.016 2.923 -2399 0.149 0.452 DC 1/X 14 0395 0.879 0315 DC iog(x) 14 0.184 1353 -0.116 D c sqrt(x) 14 0.069 1333 -0.079 D c x(neg exp) 14 0.021 1322 •0.009 D c x*2 14 0.035 1.143 0.000 D c 3-param 13 0.016 1.192 -0.002 1389

D c 4-param 12 0.037 1370 -0.082 0393• 0.091 E Mass 1/X 14 0310 0376 0320 E Mass log(x) 14 0.161 1328 -0.175 E Mass sqrt(x) 14 0.034 1350 •0.122 E Mass x(neQ exp) 14 0.030 1.183 ■0.014 E Mass x*2 14 0.145 1.075 0.000 E Mass 3-param 13 0.022 1340 -0.039 0.754 E Mass 4-param 12 0.024 3351 -2.144 0302 0.471 E P 1/x 14 0.136 -0.098 0388 E P tog(x) 14 0.056 0.432 -0.170 E P eqrt(x) 14 0.018 0370 •0.122 E P x(negexp) 14 0.009 0.414 -0.014 E P xA2 14 0.026 0314 0.000 E P 3-param 13 0X39 0.416 •0.016 0375 E P 4-param 12 0.006 3.130 -2.734 0304 0.609 E N 1/x 14 0385 0313 0375 E N tofl(x) 14 0.062 0358 -0307 370

E -N sqrt(x) 14 0.021 1.006 -0.146 E -N x(neo exp) 14 0.029 0.811 -0.017 E -N XA2 14 0.099 0.668 0.000 E N 3-param 13 0.019 0.916 •0.075 0.648 E N 4-param 12 0.019 3152 -2301 0301 0.470 E C 1/X 14 0.491 0.628 0322 E C tofl(x) 14 0.159 1.187 -0.177 E c #qrt(x) 14 0.035 1313 -0.124 E c x(neg exp) 14 0.027 1.145 -0.014 E c x*2 14 0.129 1.037 0.000 E c 3-param 13 0.022 1.194 -0.036 0.779 E c 4-param 12 0.025 3.436 -2370 0300 0.496 P Mass 1/x 14 0.687 1.096 0310 P Mass log(x) 14 0.337 1.472 -0.118 P Mass sqrt(x) 14 0.149 1554 •0.081 P Mass x(neg exp) 14 0.068 1.441 •0.009 P Mass xA2 14 0.072 1.359 0.000 P Mass 3-param 13 0.059 1.400 -0.001 1.400 P Mass 4-param 12 0.084 1.445 -0.017 0.890 0.031 P P 1/x 14 0.204 0.126 0354 P P tog(x) 14 0.090 0.619 -0.159 P P sqrt(x) 14 0.030 0.754 -0.115 P P x(neg exp) 14 0.011 0.607 -0.013 P P xA2 14 0.030 0.506 0.000 P P 3-param 13 0.012 0596 -0.010 1.070 P P 4-param 12 0.009 3.159 -2.584 0312 0.682 P N 1/x 14 0.051 -0.128 0.155 P .4 tog(x) 14 0.029 0.179 •0.098 P N sqrt(x) 14 0.017 0351 -0.067 P N x(neg exp) 14 0.011 0.155 -0.007 P N xA2 14 0.009 0.084 0.000 P N 3-param 13 0.010 0.092 0.000 1.847 P N 4-param 12 0.016 3372 -3338 0.152 0.561 P C 1/X 14 0.571 0.945 0318 P C log(x) 14 0.285 1336 -0.123 P c sqrt(x) 14 0.127 1.425 -0.084 P c x(neg exp) 14 0.055 1.308 •0.009 P c xA2 14 0.050 1325 0.000 P c 3-param 13 0.043 1360 -0.001 1.468 P c 4-param 12 0.115 3.596 -2.306 0.188 0.521

E - Erythrina leaves P - pasture grass D - dung OOOOOOOODODOOOOO i*1 l a438**ft88B2-...,-o S 33 88* ft 8 8 5 2 ■* . » - o a 3 3 S V * & 8 V a ; n

1*499833*432288* = 8*43X88*4838948* 9 358*88*3 2*32 2*4

ooooaooeooooooo opepa-^oooosoopee — 00 — 00000000000 8 a a a a * s s a ± s 9 a a * 8 It 8 9 3 S a 8 8 it 3 8 8 U 5 8 *8*8 38*82 3 28 35 8 ila. .w « ^ 9400004000000000 0000000000000400 35553*333a88882* 8 8828 (3SB8!* 28* Bba * * a a * 8 a a 9 * 8 & 3 2 a q If ooooOQOOaeesoop Boooeoflooeooeoeo 000000000000400 1 5 8S222S222222S28 88 B2S88S8B283S29 asss£32 88ssaass/ — — - fDUUUUUUMy^yuUbUW SSiaSSKSttiSlttil 4289=4888=8*8*35 =S84*i3i88*is332

oooeoooopooeoooo 000004000000000 I 3=8388 39888 8838 3838883838233388 32*8*238*838888 888 288 8888888888 3**4*38**4*444*4 388888882 22222A3 K8£S«3 5 2«8a*8883 33 88228 2382888 88 *32*38232883*3*8

— o — oooooooooo N^MOOWOOOOOOOOOO O M N — OMOOOpOOOOO 8 V a 9 B * A 2 2 8 8 3 8 3 S 8 2288988282388238 4*898**882**828

*

o o 8 5 8 8i8<48MU82 8 43 W 'p*o = **33'4^888484 &2338 = 3X8*8::3385 4S9s98**9gi24952 8 9**4***94844881: \

o^-*eof3p^^«^o^oo« 528*83*83848=98 88 8828 22888 88888 2 * * =82*82*2 2£8»

2388888 882833888 A22 388H*£2i53S25 888422=88288889* 88 38823 283 332888 38*988 388832 8=82 3948*8=83888*883

^ W ^ ^ -w H*U*3l$4«p*yH*GIS«y 3— |3482o2.O*0o8p 8 2 2 8 * t 3 8 18 8 8 8 8 8 S « »if is ft » g s 5 * it a ft it a tt 9 3282*2**2 389*8 ^mmnnnnt a i m s m y i ^ B S m m m m n n 882 4882283388988 83138828=3389333 88*39984*3994999 ^88388332830882 = S33I£ s382ai5tBB§ jmSISSIUPISUa 833388888883 383 23888*8282829828 83888484*42=928 33ii3S22SS?5i823 3?gs3?mt'MS23St 988832548*84884= 28383388S&8o83S8 388238222283383* 32828882333 48823 88 * 28 3 *8.••„•.8 .4*48 8 82S-i.Sv,-o 3B2 23t833t8 32 38 283***988**9*338 3 * 8 tt 3 8 9 3 9 a a = 8 = a

*2 82 88 8 2888 22882 *48a*aS8a«8882X8 2 2 8 8 9 8 2 3 8 2 8 3 8 8 3 3 23 2*8 3 9*2 2*98=2* *88838838*988388

oo — o — o — pooooooo OHW^MHH0^OeO0-«O 83238 8 32 3888838 S88 888**88*8X8=3 248 8988 98888828

8 t

9 o 8 3 §

ILZ LIST OF REFERENCES

Abbott, L., and A. Robson. 1987. Managing va mycorrhizal fungi in pastures. Pages 10-12 ]n 0. Sylvia, L. Hung, and J. Graham, (eds). Mycorrhizae in the next decade: practical applications and research priorities. Proc. 7th N.A. Conf. on Mycorrhizae. Institute of Food and Agricultural Sciences, University of Florida, Gainsville, Fla.

Abrams, M., and W. Jarrell. 1991. Effective diffusion coefficients in soils as measured by the planar ion sink sheets method, unpublished paper.

. 1992. Bioavailability index for phosphorus using ion exchange resin- impregnated membranes. Soil Sci. Soc. Am. J. (in press Sept.-Oct.).

Adepetu, J. , and R. Corey. 1977. Changes in N and P availability and P fractions in two soil from Nigeria under intensive cultivation. Plant Soil 46:309-316.

Ae, N., J. Arihara, K. Okada, T. Yoshihara, and C. Johansen. 1990. Phosphorus uptake by pidgeon pea and its role in cropping systems of the Indian subcontinent. Science 248:477-480.

Aharoni, C., D. Sparks, S. Levinson, and I. Ravina. 1991. Kinetics of soil chemical reactions: relationships between empirical equations and diffusion models. Soil Sci. Soc. Am. J. 55:1307-1312.

Allen, M., T. Moore Jr., and M. Christensen. 1980. Phytohormone changes in Boute/ona gracilis infected by vesicular-abuscular mycorrhizae: I. cytokinin increases in the host plant. Canad. J. Bot. 58:371-374.

Allen, M., T. Moore Jr., and M, Christensen. 1982. Phytohormone changes in Boute/oua gracilis infected by vesicular-arbuscular mycorrhizae. II. altered levels of gibberellin-like substances and abscisic acid in the host plant. Canad. J. Bot. 60:468-471.

372 373 Alvarado, A. 1984. Alumfnio activo en suelos derivados de cenizas volc£nicas de Costa Rica y Guatemala. Turrialba 34(31:396-398.

Amer, F., D. Bouldin, C. Black, and F. Duke. 1955. Characterization of soil phosphorus by anion exchange resin adsorption and 32P equilibration. Plant Soil 6(4):391-408.

Anderson, J., and J. Ingram (eds). 1989. Tropical soil biology and fertility: a handbook of methods. CAB International, Wallingford, England. 171 p.

Anderson, J. 1986. Plant nutrient-use efficiencies and soil processes in natural and agricultural ecosystems. Pages 20-29 in R. Prinsley and M. Swift, (eds). Amelioration of soil by trees. Commonwealth Science Council, London, England.

Andren, O., and K. Paustian. 1987. Barley straw decomposition in the field: a comparison of models. Ecology 68(5): 1190-1200.

Aziz, T., and M. Habte. 1987. Soil solution P requirements of mycorrhizal cowpea grown in eroded soil. Page 143 1 q D. Sylvia, L. Hung, and J. Graham, (eds). Mycorrhizae in the next decade: practical applications and research priorities. Proc. 7th N.A. Conf. on Mycorrhizae. Institute of Food and Agricultural Sciences, University of Florida, Gainsville, Fla.

Balerdi, F., L. Muller, and H. Fassbender. 1968. Estudio de fdsforo en suelos de america central. III. Comparacidn de cinco m6todos qufmicos de andlisis de fdsforo disponible. Turrialba 18(4):348-360.

Barea, J., and C. Azcon-Aguilar. 1983. Mycorrhizas and their significance in nodulating nitrogen-fixing plants. Advances in Agronomy 36:1-54.

Barnhisel, R., and P. M. Bertsch. 1982. Aluminum. Pages 275-300 in A. Page, R. Miller, and D. Keeney, (eds). Methods of soil analysis part 2: Chemical and microbiological properties. 2nd ed. ASA, SSSA, Inc., Madison, Wisconsin.

Barraclough, P., and P. Tinker. 1981. The determination of ionic diffusion coefficients in field soils: I. diffusion coefficients in sieved soils in relation to water content and bulk density. J. Soil Science 32:225-236.

Beare, M., J. Blair, and R. Parmelee. 1989. Resource quality and trophic responses to simulated throughfall: effects on decomposition and nutrient flux in a no-tillage agroecosystem. Soil Biol. Biochem. 21 (8): 1027-1036. 374 Beck, D., and D. Munns. 1984. Phosphate nutrition of Rhizobium spp. Appl. Environ. Microbiol. 47{2):278-282.

Benites, J. 1990. Agroforestry systems with potential for acid soils of the humid tropics of Latin America and the Caribbean. For. Ecol. Mgmnt. 36:81- 101.

1990. Agroforestry systems with potential for acid soils of the humid tropics of Latin America and the Caribbean. For. Ecol. Mgmnt. 36:81-101.

Berbara, R., L. Freire, L. Lima, and N. Amaral Sobrinho. 1985. Infeccidn micorrfzica en ocho especies de leguminosas usadas como abono verde. Unpublished report. Universidad Federal Rural do Rio de Janeiro, Rio de Janeiro, Brazil. 12 .

Bertsch, F., and A. Cordero. 1984. Fertilidad de Typic Dystrandepts de Costa Rica. II. Aniones (N,P,B,S,Mo), materia orgdnica y textura. Turrialba 34(2):199-205.

Bethlenfalvay, G., M. Brown, and W. Newton. 1987. Photosynthetic water- and nutrient-use efficiency in a mycorrhizal legume. Pages 231-233 jrt D. Sylvia, L. Hung, and J. Graham, (eds). Mycorrhizae in the next decade: practical applications and research priorities. Proc. 7th N.A. Conf. on Mycorrhizae. Institute of Food and Agricultural Sciences, University of Florida, Gainsville, Fla.

Bethlenfalvay, G., R. Franson, M. Brown, and K. Mihara. 1989. The Glycine- Glomus-Bradyrhizobium symbiosis. IX. Nutritional, morphological and physiological responses of nodulated soybean to geographic isolates of the mycorrhizal fungus Glomus mosseae. Physiol. Plant. 76:226-232.

Bhadoria, P., J. Kaselowsky, N. Claassen, and A. Jungk. 1991. Soil phosphate diffusion coefficients: their dependence on phosphorus concentration and buffer power. Soil Sci. Soc. Am. J. 55:56-60.

Biermann, B., and R. Linderman. 1981. Quantifying vesicular-arbuscular mycorrhizae: a proposed method towards standardization. New Phytol. 87:63-67.

Bigham, J., D. Golden, L. Bowen, S. Buol, and S. Weed. 1978. Iron oxide mineralogy of well-drained ultisols and oxisols: I. characterization of iron oxides in soil clays by Mossbauer spectroscopy. X-ray diffractometry and selected chemical techniques. Soil Sci. Soc. Am. J. 42:816-825. 375 Binkley. D., and P. Matson. 1983. Ion exchange resin bag method for assessing forest soil nitrogen availability. Soil Sci. Soc. Am. J. 47:1050- 1052.

Blair, G., and O. Boland. 1978. The release of phosphorus from plant material added to soil. Aust. J. Soil Res. 16:101-111.

Blair, J., R. Parmelee, and M. Beare. 1990. Decay rates, nitrogen fluxes, and decomposer communities of single- and mixed-species foliar litter. Ecology 71(51:1976-1985.

Blair, J. 1988. Nutrient release from decomposing foliar litter of three tree species with special reference to calcium, magnesium and potassium dynamics. Plant Soil 110:49-55.

1988. Nitrogen, sulfur and phosphorus dynamics in decomposing deciduous leaf litter in the southern Appalachians. Soil Biol. Biochem 20(51:693-701.

Blake, G., and K. Hartge. 1986. Bulk density. Pages 363-376 in A. Klute, (ed). Methods of soil analysis part 1: Physical and mineralogical methods. 2nd ed. ASA, SSSA, Inc., Madison, Wisconsin.

Blume, H., and U. Schwertmann. 1969. Genetic evaluation of profile distribution of aluminum, iron and manganese oxides. Soil Sci. Soc. Amer. Proc. 33:438-444.

Boerner, R. 1992a. Plant lifespan and response to inoculation with vesicular- arbuscular mycorrhizal fungi. III. Responsiveness and residual soil P levels. Mycorrhiza. (in press).

-...... 1992b. Plant lifespan and response to inoculation with vesicular- arbuscular mycorrhizal fungi. I. Annual vs. perennial grasses. Mycorrhiza. (in press).

1992c. Plant lifespan and response to inoculation with vesicular- arbuscular mycorrhizal fungi. II. Species from weakly mycotrophic families. Mycorrhiza. (in press).

Boerner, R. 1990. Role of mycorrhizal fungus origin in growth and nutrient uptake by Geranium robertianum. Am. J. Botany 77:483-489. 376 Bolan, N., A. Robson, N. Barrow, and L. Aylmore. 1984. Specific activity of phosphorus in mycorrhizal and non-mycorrhizal plants in relation to the availability of phosphorus to plants. Soil Biol. Biochem. 16(4):299-304.

Bolan, N., A. Robson, and N. Barrow. 1987. Effects of vesicular-arbuscular mycorrhiza on the availability of iron phosphates to plants. Plant and Soil 99:401-410.

Borie, F. 1985. Fdsforo org£nico y micorrizas "VA" en suelos volcdnicos de Chile. Unpublished report. Universidad de la Frontera, Temuco, Chile. 18 p.

Bornemisza, E. 1966. El fdsforo orgdnico en suelos tropicales. Turrialba 16(1):33-38.

Bornemisza, E. 1969. Minerales de arcilla en suelos centroamericanos y de Panama. Turrialba 19(11:97-102.

Bowman, R., and C. Cole. 1978. An exploratory method for fractionation of organic phosphorus from grassland soils. Soil Sci. 125(21:95-101.

Brady, K., J. Bigham, W. Jaynes, and T. Logan. 1986. Influence of sulfate on Fe-oxide formation: comparisons with a stream receiving acid mine drainage. Clays and Clay Minerals 34(3):266-274.

Bremner, J., and C. Mulvany. 1982. Nitrogen-total. Pages 595-624 in A. Page, R. Miller, and D. Keeney, (eds). Methods of soil analysis part 2: Chemical and microbiological properties. 2nd ed. ASA, SSSA, inc., Madison, Wisconsin.

Briceno, J., and R. Pacheco. 1984. Mdtodos analfticos para el estudio de suelos y plantas. 61. Universidad de Costa Rica, Centro de Investigaciones Agronomicas, San Jose, Costa Rica.

Bronstein, G. 1984. Produccidn de pasto asociado con por6 ( Erythrina poeppigiana) con laurel (Cordia attiodora) y sin arboles. M.Sc. Thesis. CATIE, Turrialba, Costa Rica.

Budelman, A. 1988. The decomposition of leaf mulches of Leucaena feucocephafa, Gfiricidia sepium and Ffemingia macrophyUa under humid tropical conditions. Agroforestry Systems 7:33-45.

-...... 1989. Nutrient composition of leaf biomass of three selected woody leguminous species. Agroforestry Systems 8:39-51. 377 Budowski, G., R. Russo, and E. Mora. 1985. Productividad de una cerca viva de Erythrina berteroana Urban en Turrialba, Costa Rica. Turrialba 35(1):83-86.

Buschbacher, R., C. Uhl, and E. Serrao. 1988. Abandoned pastures in eastern Amazonia. II. nutrient stocks in the soil and vegetation. J. Ecol. 76:682-699.

Buschbacher, R. 1986. Tropical deforestation and pasture development. Bioscience 36(1):22-28.

1987. Cattle productivity and nutrient fluxes on an Amazon pasture. Biotropica 19:200-207.

Buwalda, J., and K. Goh. 1982. Host-fungus competition for carbon as a cause of growth depressions in vesicular-arbuscular mycorrhizal ryegrass. Soil Biol. Biochem. 14:103-106.

Caldwell, M., J. Richards, J. Manwaring, and D. Eissenstat. 1987. Rapid shifts in phosphate acquisition show direct competition between neighbouring plants. Nature 327:615-616.

Calhoun, F., and V. Carlisle. 1971. Infrared spectra of selected Colombian Andosols. Soil Crop Sci. Soc. Florida Proc. 31:157-161.

Calhoun, F. G., V. Carlisle, and Z. C. Luna. 1972. Properties and genesis of selected Colombian andosols. Soil Sci. Soc. Amer. Proc. 36(31:480-485.

Carling, D.r and M. Brown. 1980. Relative effect of vesicular-arbuscular mycorrhizal fungi on the growth and yield of soybeans. Soil Sci. Soc. Am. J. 44:528-532.

Cavigelli, M., and S. Thien. 1989. Soil phosphorus bioavailability in rotations. Agron. Abstr. Pages 235

Chahan, B., J. Stewart, and E. Paul. 1981. Effect of labile inorganic phosphate status and organic carbon additions on the microbial uptake of phosphorus in soils. Canad. J. Soil Sci. 61:373-385.

Chauhan, B., J. Stewart, and E. Paul. 1979. Effect of carbon additions on soil labile inorganic, organic and microbially held phosphate. Canad. J. Soil Sci. 59:387-396. 378 Chinchilla, E. 1987. Atlas cantonal de Costa Rica p. 371*374. Institute de Fomento y Asesoria Municipal (IFAM), San Jose, Costa Rica.

Chiyenda, S., and S. Materechera. 1989. Effect of incorporating prunings of Leucaena feucocephala, Cassia siamea and Cajanus ca/an on yield of maize in alley cropping system. Pages 119-128 in J. van der Heide, (eds). Nutrient management for food crop production in tropical farming systems. Institute for Soil Fertility (IB) and International Institute for Tropical Agriculture (IITA), Haren, The Netherlands.

Christie, E. 1975. A note on the significance of Eucalyptus popu/nea for buffel grass production in infertile semi-arid rangelands. Trop. Grasslands 9(3):243-246.

Churchman, G., J. Whitton, G. Claridge, and B. Theng. 1983. A rapid test for halloysite. Soil Taxonomy News (Washington, D.C.) 5 February: 10. USAID-Soil M anagement Support Service, .

Clarkson, D. 1985. Factors affecting mineral nutrient acquisition by plants. Ann. Rev. Plant Physiol. 36:77-115.

Cole, C., G. Innis, and J. Stewart. 1977. Simulation of phosphorus cycling in semiarid grasslands. Ecology 58(1): 1-15.

Coleman, D., C. Cole, and E. Elliott. 1984. Decomposition, organic matter turnover, and nutrient dynamics in agroecosystems. Pages 83-104 in R. Lowrance, B. Stinner, and G. House, (eds). Agricultural ecosystems: unifying concepts. John Wiley & Sons, Inc., New York, NY.

Collins Johnson, N., P. Copeland, R. Crookston, and F. Pfleger. 1992. Mycorrhizae: possible explanation for yield decline with continous corn and soybean. Agron. J. 84:387-390.

Comerford, N., and W. Dyck. 1988. Interaction of forest floor material and mineral soil on orthophosphate sorption. New Zeal. J. For. Sci. 18(2): 191- 198.

Condron, L., K. Goh, and R. Newman. 1985. Nature and distribution of soil phosphorus as revealed by a sequential extraction method followed by 31P nuclear magnetic resonance analysis. J. Soil Sci. 36:199-207.

Crowder, L., and H. Chheda. 1982. Tropical grassland husbandry. Longman Group Ltd., Essex, UK. 379 Curtin, D., J. Syers, and G. Smillie. 1987. The importance of exchangeable cations and resin-sink characteristics in the release of soil phosphorus. J. Soil Sci. 38:711-716.

Daage, F. C. 1987. Current programs, problems and strategies for land clearing and development on volcanic ash soils. Pages 195-206 in R. Lai, M. Nelson, H. Scharpenseel, and M. Sudjadi, (eds). Tropical land clearing for sustainable agriculture: proceedings of an IBSRAM inaugural workshop. International Board for Soil Research and Management (IBSRAM), Bangkok, Thailand.

Daccarett, M., and J. Blydenstein. 1968. La influencia de arboles leguminosos sob re forraje que crece bajo ellos. Turrialba 18(4):405-408.

Oalal, R. 1979. Mineralization of carbon and phosphorus from carbon-14 and phosphorus-32 labelled plant material added to soil. Soil Sci. Soc. Am. J. 43:913-916.

Dalai, R., and E. Hallsworth, 1976. Evaluation of the parameters of soil phosphorus availability factors in predicting yield response and phosphorus uptake. Soil Sci. Soc. Am. J. 40:541-546.

Daniels Hetrick, B., D. Gerschefske Kitt, and G. Thompson Wilson. 1988. Mycorrhizal dependence and growth habit of warm-season and cool-season tallgrass prairie plants. Canad. J. Bot. 66:1376-1380. de Bruin, S. 1987. Estudio semidetallado de los suelos del asentamiento Neguev y los alrededores. CATIE-AUW-MAG Atlantic Zone Program, Guapiles, Costa Rica. 26 p. de Bruin, S. 1990. La aptitud de los suelos en el Asentamiento Neguev. Pages 27-30 in W. Wielemaker, (eds). Atlantic Zone Programme Workshop on soil information; Working document no. 10. CATIE-AUW-MAG Atlantic Zone Programme, Turrialba, Costa Rica, October 2-4.

1991. Estudio detallado de los suelos del asentamiento Neguev. Unpublished report for Atlantic Zone Programme. CATIE-AUW-MAG Atlantic Zone Programme, Guapiles, Costa Rica. 23 .

Diaz-Romeu, R., and A. Hunter. 1978. Metologla de muestreo de suelos, andlisis qufmico de suelos y tejido vegetal e investigacidn en invernadero. CATIE, Turrialba, Costa Rica. 61. 3 8 0 Dixon, J. 1966. Quantitative analysis of kaolinite and gibbsite in soils by differential thermal and selective dissolution methods. Clays Clay Mineralogy 14:83-89.

Dodd, J., C. Burton, R. Burns, and P. Jeffries. 1987. Phosphatase activity associated with the roots and the rhizosphere of plants infected with vesicular-arbuscular mycorrhizal fungi. New Phytol. 107:163-172.

Dulin, P. 1982. Agroforestry as an appropriate land use system in the American tropics. Agroforestry for the humid tropics, Short Course presentation. USAID/CATIE, Turrialba, Costa Rica. 17 p.

Duxbury, J., S. Smith, J. Doran, C. Jordan, L. Szott, and E. Vance. 1989. Soil organic matter as a source and a sink of plant nutrients. Pages 33-68 jn D. Coleman, J. Oades, and G. Uehara, (eds). Dynamics of soil organic matter in tropical ecosystems. NifTAL Project, Honolulu, Hawaii.

Dyer, M., M. Acra, G. Wang, D. Coleman, D. Freckman, S. McNaughton, and B. Strain. 1991. Source-sink carbon relations in two Panicum coforatum ecotypes in response to herbivory. Ecology 72<4):1472-1483.

Egawa, T. 1984. Properties of soils derived from volcanic ash. Pages 249- 302 in K. Tan, (eds). Andosols: Benchmark papers in soil science series. Van Nostrand Reinhold Co. Inc., New York, New York.

El Swaify, S. Physical and mechanical properties of oxisols. 1980. Soils with variable charge. Ed. B. Theng, 303-324. New Zealand Society of Soil Science, Offset Publ., Palmerston North, New Zealand.

Ewel, J. 1986. Designing agricultural ecosystems for the humid tropics. Ann. Rev. Ecol. Syst. 17:245-271.

Ezcurra, E., and J. Becerra. 1987. Experimental decomposition of litter from the Tamaulipan Cloud Forest: a comparison of four simple models. Biotropica 19(41:290-296.

Fassbender, H. 1968. Phosphate retention and its different chemical forms under laboratory conditions for 14 Costa Rica soils. Agrochimica 12(6):512- 521.

Fassbender, H. , L. Muller, and F. Balerdi. 1968. Estudio de fbsforo en suelos de america central: II. Formas y su relacidn con las plantas. Turrialba 18(4):333-347. 381 Fassbender, H. W., and E. Bornemisza. 1987. Qurmica de suelos con 6nfasis en suelos de America Latina. 2nd ed. rev. IICA, San Jose, Costa Rica.

Fitter, A. 1977. Influence of mycorrhizal infection on competition for phosphorus and potassium by two grasses. New Phytologist 79:119-125.

Fox, R. 1981. External phosphorus requirements of crops. Pages 223-240 in R. Dowdy, J. Ryan, V. Volk, and D. Baker, (eds). Chemistry in the soil environment. ASA Special Publication No. 40. ASA, SSSA, Madison, Wl.

Fox, R., M. Habte, R. Huang, and C. Powell. 1987. Mycorrizal effectiveness as indicated by plant P uptake. Page 248 in D. Sylvia, L. Hung, and J. Graham, (eds). Mycorrhizae in the next decade: practical applications and research priorities. Proc. 7th N.A. Conf. on Mycorrhizae. Institute of Food and Agricultural Sciences, University of Florida, Gainsville, Fla.

Fox, T., and N. Comerford. 1992. Influence of oxalate loading on phosphorus and aluminum solubility in spodosofs. Soil Sci. Soc. Am. J. 56:290-294.

1990. Low molecular-weight organic acids in selected forest soils of the southeastern USA. Soil Sci. Soc. Am. J. 54:1139-1142.

Fuentes, R., K. Igue, and E. Bornemisza. 1972. Mineralizacidn de P orgdnico en suelos volcanicos de Costa Rica. Fitotecnica Latinoamericana 8(1):71-76.

Gavande, S. 1968. Water retention characteristics of some Costa Rican soils. Turrialba 18(1):34-39.

Gebhardt, H., and N. Coleman. 1974. Anion adsorption by allophanic tropical soils: III. phosphate adsorption. Soil Sci. Soc. Am. Proc. 38:263- 266.

Gee, G., and J. Bauder. 1986. Particle size analysis. Pages 383-412 in A. Klute, (eds). Methods of soil analysis part 1: Physical and mineralogical methods. 2nd ed. ASA, SSSA, Inc., Madison, Wisconsin.

Gibson, D., I. Colquhoun, and P. Greig-Smith. 1985. A new method for measuring nutrient supply rates in soils using ion-exchange resins. Pages 73-79 in A. Fitter, D. Atkinson, D. Read, and M. Usher, (eds). Ecological interactions in soil, plants, microbes and animals. Special publication no. 4 British Ecological Society. Blackwell Scientific Publ., Oxford, England. 382 Gillespie, A. 1989. Modelling nutrient flux and interspecies root competition in agroforestry plantings. Agroforestry Systems 8:257-265.

Gillespie, A., and P. Pope. 1990. Rhizosphere acidification increases phosphorus recovery of black locust: I. induced acidification and soil response. Soil Sci. Soc. Am. J. 54:533-537.

Gillespie, A., and P. Pope. 1990. Rhizosphere acidification increases phosphorus recovery of black locust: II. model predictions and measured recovery. Soil Sci. Soc. Am. J. 54:538-541.

Giovannetti, M., and B. Mosse. 1980. An evaluation of techniques for measuring vesicular arbuscular mycorrhizal infection in roots. New Phytol. 84:489-500.

Goering, H. , and P. van Soest. 1970. Forage fiber analysis (apparatus, reagents, procedures, and some applications). Agriculture Handbook No. 379. Agricultural Research Service, USDA.

Gomez, V., A. Chernyakhovskiy, N. Chizhikova, Y. Shchurygina, and B. Gradusov. 1981. Mineral composition of volcanic soils in the tropical region of Costa Rica. Soviet Soil Science 13(4):95-102.

Gonzalez A., M., and D. Sauerbeck. 1982. Decomposition of I^C-labelled plant residues in different soils and climates of Costa Rica. Pages 141-146 In Proceedings of Regional Colloquium on soil organic matter. UNDP-IAEA. CENA, Piracicaba, S.P. Brazil.

Habte, M., and A. Manjunath. 1987. Soil solution phosphorus status and mycorrhizal dependency in Leucaena Jeucocepha/a. Applied and Environ. Microbiol. 53(4):797-801.

Hargrove, W., P. Ford, and Z. Somda. 1991. Crop residue decomposition under controlled and field conditions. Pages 99-108 in Proc. International Soil Tillage Research Organization (ISTRO) 12th International Conference: Soil Tillage and Agricultural Sustainability. 8-12 July 1991. ISTRO, Ibadan, Nigeria.

Harrison, A. 1982. Labile organic phosphorus mineralization in relationship to soil properties. Soil Biol. Biochem. 14:343-351.

Haydock, K., and N. Shaw. 1975. The comparative yield method for estimating dry matter yield of pasture. Aust, J. Exptl. Agric. Anim. Husb. 15:663-670. 383

Hayman, D. 1975. Phosphorus cycling by soil micro-organisms and plant roots. Pages 67*91 in N. Walker, (eds). Soil microbiology. Butterworth & Co., Ltd., London, England.

Haynes, R., and R. Swift. 1986. Effects of soil acidification and subsequent leaching on levels of extractable nutrients in a soil. Plant Soil 95:327*336.

Herbillon, A. 1980. Mineralogy of oxisols and oxic materials. Pages 109-126 in B. Theng, (eds). Soils with variable charge. New Zealand Society of Soil Science, Offset Publ., Palmerston North, New Zealand.

Heuveldop, J., L. Alpizar, H. Fassbender, G. Enriquez, and H. Folster. 1985. Sistemas agroforestales de caf6 ( Coffea arabica) con laurel ( Cordia aiiiodora) y cafe con pord ( Erythrina poeppig/ana) en Turrialba, Costa Rica. II. Produccidn agricola maderable y de residuos vegetales. Turrialba 35(4): 347- 355.

Hillel, D. 1982. Introduction to soil physics. Academic Press, Inc., Orlando, Florida. 364 p.

Hopkins, D., R. Shiel, and A. O'Donnell. 1988. The influence of sward species composition on the rate of organic matter decomposition in grassland soil. J. Soil Sci. 39:385-392.

Huang, R., and R. Yost. 1987. Early physiological events in VA mycorrhizal Leucaena leucocephaJa. Page 253 io D. Sylvia, L. Hung, and J. Graham, (eds). Mycorrhizae in the next decade: practical applications and research priorities. Proc. 7th N.A. Conf. on Mycorrhizae. Institute of Food and Agricultural Sciences, University of Florida, Gainsville, Fla.

Huang, R., W. Smith, and R. Yost. 1985. Influence of vesicular-arbuscular mycorrhiza on growth, water relations and leaf orientation in Leucaena leucocephafa (Lam.) de Wit. New Phytol. 99:229-243.

Huettl, P., R. Wendt, and R. Corey. 1979. Prediction of algal-available phosphorus in runoff suspensions. J. Environ. Gual. 8(1): 130-132.

Huffman, S., C. Cole, and N. Scott. 1991. Phosphorus transformations during crop residue decomposition in laboratory experiments: mathematical simulation. ASA Abstr.

Humphreys, L. 1991. Tropical pasture utilisation, 27-29. Cambridge University Press, Cambridge, U.K. 3 8 4

. 1991. Tropical pasture utilisation. Cambridge University Press, Cambridge, UK.

Hundal, H., C. Biswas, and A. Vig. 1988. Phosphorus sorption characteristics of flooded soil amended with green manures. Trop. Agric. 65(21:185-187.

Huxley, P. 1986. Rationalizing research on hedgerow intercropping: an overview. Working Paper No. 40. International Research Council for Research in Agroforestry (ICRAF), Nairobi, Kenya. 151 p.

Igue, K., and R. Fuentes. 1971. Retencidn y solubilizacibn de 32p en suelos dcidos de regiones tropicales. Turrialba 21 (4>:429-434.

Ionics, I. 1987. Properties and characteristics of anion transfer membranes type 204-U-386. Bulletin No. AR 204.3-A. W atertown, Mass. 2 p.

Jackson, M. 1975. Soil chemical analysis-advanced course. 2nd ed., 10th printing. Department of Soil Science, University of Wisconsin, Madision, Wisconsin.

Jackson, R., J. Manwaring, and M. Caldwell. 1990. Rapid physiological adjustment of roots to localized soil enrichment. Nature 344(6261 }:58-60.

Janos, D. 1975. Effects of vesicular-arbuscular mycorrhizae on lowland tropical rainforest trees. Pages 437-446 in Sanders, Mosse, and Tinker, (eds). Endomycorrhizas. Academic Press, London, England.

. 1984. Methods for vesicular-arbuscular mycorrhiza research in the lowland wet tropics. Pages 173-187 jn E. Medina, H. Mooney, and C. Vazquez-Vanes, (eds). Physiological ecology of plants of the wet tropics (Tasks for vegetation science 12). Junk, The Hague, Netherlands.

1983. Tropical mycorrhizas, nutrient cycles and plant growth. Pages 327-345 jn S. Sutton, T. Whitmore, and A. Chadwick, (eds). Tropical rainforest: ecology and management. Special publication No. 2 of British Ecol. Soc. Blackwell Scientific Publ., Oxford, England.

1980. Mycorrhizae influence tropical succession. Biotropica 1 2(Supplement):56-64.

. 1988. Mycorrhiza applications in tropical forestry: are temperate-zone approaches appropriate? Pages 133-188 jn F. Ng, (eds). Trees and 385 Mycorrhiza. Forest Research Institute of Malaysia, Kuala Lampur, Malaysia. 305.

Jehne, W. 1980. Endomycorrhizas and the productivity of tropical pastures: the potential for improvement and its practical realization. Tropical Grasslands 14(3):202-209.

Joachim, A., and S. Kandiah. 1934. The change in composition and decomposability of typical Ceylon green manures with age. Tropical Agriculturist 82{1):3-19.

Jordan, C. 1985. Nutrient cycling in tropical forest ecosystems: principles and their application in management and conservation. John Wiley & Sons, Ltd., New York, NY. 190 p.

Jusoff, K. 1988. Influence of sheep grazing on soil chemical properties and growth of rubber (Hevea bras/f/ensis) in Malaysia. Agrofor. Syst. 7:115-120.

Kafkafi, U., B. Bar-Yosef, R. Rosenberg, and G. Sposito. 1988. Phosphorus adsorption by kaolinite and montmorillinite: II. organic anion competition. Soil Sci. Soc. Am. J. 52:1585-1589.

Kang, B., and G. Wilson. 1987. The development of alley cropping as a promising agroforestry technology. Pages 227-243 in H. Steppler and P. Nair, (eds). Agroforestry: a decade of development. International Council for Research in Agroforestry (ICRAF), Nairobi, Kenya.

Kass, D. 1991. Development of appropriate technology for overcoming different mechanisms of P retention in Central AmericaFinal Report, AID-SCf Project 936-5442. CATIE, Turrialba, Costa Rica. 78 p.

Kass, D., A. Barrantes, W. Bermudez, W. Campos, M. Jimenez, and J. Sanchez. 1991. Resultados de seis anos de investigacidn de cultivo en callejones (alley cropping) en "La Montana", Turrialba, Costa Rica. El Chasqui (CATIE Publication) 19:5-24.

Kass, D. L., M. Jimenez, W. Bermudez, and L. Cedeno. 1985. Respuesta del frijol comun (Phaseo/us vulgaris L.) var. Talamanca a aplicaciones de caliza y fdsforo en suelos con altos niveles de alumrnio y manganeso en la zona dtlantica de Costa Rica. Pages 165-175 jn XXXI Renunidn Anual del PCCMCA. San Pedro Sula, Honduras, April 16-19. 386 Klute, A. 1986. Water retention: laboratory methods. Pages 635-662 in A. Klute, (ed). Methods of soil analysis part 1: Physical and mineralogical methods. 2nd ed. ASA, SSSA, Inc., Madison, Wisconsin.

Kormanik, P., and A. McGraw. 1982. Quantification of vesicular-arbuscular mycorrhizae in plant roots. Pages 37-45 in N. Schenck, (ed). Methods and principles of mycorrhizal research . Amer. Phytopathological Soc., St. Paul, MN.

Koske, R., and J. Gemma. 1989. A modified procedure for staining roots to detect VA mycorrhizas. Mycol. Res. 92(4):486-505.

Koslowsky, S., and R. Boerner. 1989. Interactive effects of aluminum, phosphorus and mycorrhizae on growth and nutrient uptake of Pan/cum virgatum L. (Poaceae). Environmental Pollution 61:107-125.

Kovar, J., and S. Barber. 1988. Phosphorus supply characteristics of 33 soils as influenced by seven rates of phosphorus addition. Soil Sci. Soc. Am. J. 52:160-165,

Krause, H., and D. Ramlal. 1987. In situ nutrient extraction by resin from forested, clear-cut and site-prepared soil. Can. J. Soil Sci. 67:943-952.

Lachat Instruments Co. 1990. Lachat QuickChem Method No. 10-115-01-1- H:TKP. Lachat methods manual for QuickChem Automated Ion Analyzer. Chapter 20. Lachat Instruments, Mikwaukee, Wisconsin.

. 1990. Lachat QuickChem method No. 10-107-06-2-G:TKN. Lachat methods manual for QuickChem Automated Ion Analyzer. Chapter 8. Lachat Instruments, Milwaukee, Wisconsin.

Lajtha, K., and W. Schlesinger. 1988. The biogeochemistry of phosphorus availability along a desert soil chronosequence. Ecology 69(1):24-39.

Lai, R. 1989. Agroforestry systems and soil surface management of a tropical alfisol: III. Changes in soil chemical properties. Agroforestry Systems 8:113-132.

. 1980 . Physical and mechanical characteristics of alfisols and ultisols, with particular reference to soils in the tropics. Pages 253-280 in B. Theng, (ed). Soils with variable charge. New Zealand Society of Soil Science, Offset Publ., Palmerston North, New Zealand. 38 7 Lai, R. . 1989. Agroforestry systems and soil surface management of a tropical alfisol; IV. Effects on soil physical and mechanical properties. Agroforestry Systems 8:197-215.

Lansu, A. 1988. Soil structure under four land use types in the settlement Neguev, Atlantic zone of Costa Rica. Field report No. 18. Atlantic Zone Programme CATIE-UAW-MAG, Turrialba, Costa Rica. 12 .

Le Mare, P. 1982. Sorption of isotopically exchangeable and non­ exchangeable phosphate by some soils of Colombia and Brazil, and comparisons with soils of southern Nigeria. J. Soil Sci. 33:691-707.

Logan, T. 1982. Mechanisms for release of sediment-bound phosphate to water and the effects of agricultural land management on fluvial transport of particulate and dissolved phosphate. Hydrobiologia 92:519-530.

Lopez-Hernandez, D., G. Siegert, and J. Rodriguez. 1986. Competitive adsorption of phosphate with malate and oxalate by tropical soils. Soil Sci. Soc. Am. J. 50:1460-1462.

Lopez-Hernandez, D., J. Rodriguez, and G. Siegert. 1984. Competencia entre aniones orgdnicos (malato y oxalato) con ortofosfato por los sitios de adsorci6n en suelos tropicales. Turrialba 34(3):291-296.

Maeda, T., H. Takenaka, and B. Warkentin. 1977. Physical properties of allophane soils. Advances in Agronomy 29:229-264.

Maghembe, J., E. Kariuki, and R. Haller. 1983. Biomass and nutrient accumulation in young Prosopis/u/Jflora at Mombasa, Kenya. Agroforestry Systems 1:313-321.

Mattingly, G. 1975. Labile phosphate in soils. Soil Science 119(5):369-375.

McKeague, J., J. Brydon, and N. Miles. 1971. Differentiation of forms of extractable iron and aluminum in soils. Soil Sci. Soc. Proc. 35:33-38.

McKee, T., and J. Brown. 1977. Preparation of specimens for electron microscopic examination. Pages 809-833 in J. Dixon and S. Weed, (eds). Minerals in soil environments. Soil Sci. Soc. Am., Madison, Wisconsin.

McLean, E. 1982. Soil pH and lime requirement. Pages 199-224 in A. Page, R. Miller, and D. Keeney, (eds). Methods of soil analysis part 2: Chemical and microbiological properties. 2nd ed. ASA, SSSA, Inc., Madison, Wisconsin. 388

Mead, R. 1988. The design of experiments: statistical principles for practical applications. Cambridge University Press, Cambridge, England.

Menon, R., L. Hammond, and H. Sissingh. 1989. Determination of plant- available phosphorus by the iron hydroxide-impregnated filter paper (Pi) soil test. Soil Sci. Soc. Am. J. 53:110-115.

Montagnini, P., E. Stijfhoorn, and F. Sancho. 1989. Soil chemical properties and root biomass under plantations of native trees species, grass cover and secondary forest vegetation in the Atlantic lowlands of Costa Rica. Belowground Ecol. Bull. 1(1):6-8.

Mosse, B. 1977. Plant growth responses to vesicular-arbuscular mycorrhiza: X. Responses of Styfosanthes and maize to inoculation in unsterile soil. New Phytol. 78:277-288.

Mott, G. 1975. Nutrient recycling in pastures, unpublished paper (Gainsville, FL,). University of Florida.

Mueller-Harvey, I., A. Juo, and A. Wild. 1985. Soil organic C, N, S and P after forest clearance in Nigeria: mineralization rates and spatial variability. J. Soil Sci. 36:585-591.

Muller, L. 1961. Un apar£to micro-Kjeldahl simple para andlisis rutinarios de mat6rias vegetales. Turrialba 11(1 ):17-25.

Muller, L., F. Balerdi, R. Diaz-Romeu, and H. Fassbender. 1968. Estudio del fdsforo en suelos de america central: I. UbicaciOn, caracterfsticas ffsicas y qufmicas de los suelos estudiados. Turrialba 18(41:319-332.

Muller, M., V. Sundman, O. Soininvaara, and A. Merilainen. 1988. Effect of chemical composition on the release of nitrogen from agricultural plant materials decomposing in soil under field conditions. Biol. Fert. Soils 6:78- 83.

Munns, D., and A. Franco. 1981. Soil constraints to legume production. Pages 133-152 in P. Graham and S. Harris, (eds). Biological nitrogen fixation technology for agriculture. Workshop held at CIAT, March 9-13. Centro Internacional de Agricultura Tropical (CIAT), Cali, Colombia.

Nair, P. 1984. Soil productivity aspects of agroforestry. International Council for Research in Agroforestry (ICRAF), Nairobi, Kenya. 3 8 9 1984. Soil productivity aspects of agroforestry. International Council for Research on Agroforestry (ICRAF), Nairobi, Kenya. 85 p.

Newman, E. 1966. A method of estimating the total length of root in a sample. J. Appl. Ecol. 3:139-145.

Nobbe, H., and G. Hazeu. 1987. Soil survey of the area east of Rio Jimenez and the northern part of Neguev. Unpublished report for Atlantic Zone Programme. CATIE-AUW-MAG Atlantic Zone Programme, Guapiles, Costa Rica.

Ojala, J., W. Jarrell, J. Menge, and E. Johnson. 1983. Comparison of soil phosphorus extractants as predictors of mycorrhizal dependency. Soil Sci. Soc. Am. J. 47:958-962.

Olsen, S., and L. Sommers. 1982. Phosphorus. Pages 403-430 lo A. Page, R. Miller, and D. Keeney, (eds). Methods of soil analysis part 2: Chemical and microbiological properties. 2nd ed. ASA, SSSA, Inc., Madison, Wisconsin.

Olson, J. 1963. Energy storage and the balance of producers and decomposers in ecological systems. Ecology 14(21:322-331.

Omaliko, C. 1984. Dung decomposition and its effects on the soil component of a tropical grassland ecosystem. Trop. Ecol. 25:214-220.

Palencia O., J., and J. Martini. 1970. Caracterfeticas morfoldgicas, ffsicas y qufmicas de algunos suelos derivados de cenizas volcanicas en centroamerica. Turrialba 20(3):325-332.

Palm, C. 1988. Mulch quality and nitrogen dynamics in an alley cropping system in the Peruvian Amazon, Ph.D. Dissertation. North Carolina State University, Raleigh, NC. 112 p.

Palma, G., and H. Fassbender. 1970. Estudio del fdsforo en suelos de America Central. V. Uso de resinas de intercambio para evaiuar la disponibilidad de P. Turrialba 20(3):279-287.

Parfitt, R. 1989. Phosphate reactions with natural allophane, ferrihydrite and goethite. Journal of Soil Science 40:359-369.

1980. Chemical properties of variable charge soils. Pages 167-193 in B. Theng, (eds). Soils with variable charge. New Zealand Society of Soil Science, Offset Publ., Palmerston North, New Zealand. 3 9 0

Parfitt, R., L. Hume, and G. Sparling. 1989. Loss of availability of phosphate in New Zealand soils. J. Soil Sci. 40:371-382.

Parfitt, R. 1979. The availability of P from phosphate-goethite bridging complexes. Desorption and uptake by ryegrass. Plant Soil 53:55-65.

Parmelee, R., M. Beare, and J. Blair. 1989. Decomposition and nitrogen dynamics of surface weed residues in no-tillage agroecosystems under drought conditions: influence of resource quality on the decomposer community. Soil Biol. Biochem. 21(11:97-103.

Perdock, H,, M. Thamotharam, J. Blom, H. van den Born, and C. van Veluw. 1982. Practical experience with urea-ensiled straw in Sri-Lanka. in T. Preston, C. Davis, F. Dolberg, M. Hague, and M. Saadullah, (eds). Maximum livestock production from minimum land. Proc. Third Seminar. Agricultural University, Dept, of Animal Sci., Bangladesh.

Peterson, R., W. Woodhouse Jr., and H. Lucas. 1956. The distribution of excreta by freely grazing cattle and its effect on pasture fertility: II. effect of returned excreta on the residual concentration of some fertilizer elements. Agron. J. 48:440-449.

Pezo, D., M. Kass, J. Benavides, F. Romero, and 0. Chaves. 1990. Potential of legume tree fodders as animal feed in Central America. Pages 163-175 in C. Devendra, (ed). Shrubs and tree fodders for farm animals. Proc. Workshop in Denpasar, Indonesia, 24-29 July. IDRC, Canada.

Phillips, J., and D. Hayman. 1970. Improved procedures for clearing roots and staining parasitic and vesicular-arbuscular mycorrhizal fungi for rapid assessment of infection. Trans. Br. Mycol. Soc. 55(1 ):158-160.

Polglase, P., E. Joketa, and N. Comerford. 1992. Phosphorus, nitrogen and carbon fractions in litter and soil of southern pine plantations. Soil Sci. Soc. Am. J. 56(2):566-572.

Pons, F., and V. Gianinazzi-Pearson. 1985. Observations on extra-matrical vesicles of Gigaspora margarita in vitro. Trans. Br. Mycol. Soc. 84(1): 168- 170.

Rao, M., and R. Coe. 1992. Evaluating the results of agroforestry research. Agroforestry Today 4{1):4-9. 391 Rojas, A., and H. Waaijenberg. 1990. Introduccidn. Pages 1-5 in M. de Onoro, (eds). El asentamiento Neguev: Interaccidn de campesinos y estado en el aprovechamiento de los recursos naturales. Informe t6cnico no. 162; Program paper no. 7. CATIE-AUW-MAG Atlantic Zone Programme, Turrialba, Costa Rica.

Rojas, A., and F. van Sluys. 1990. Historia agraria. Pages 6-17 |jq M. de Onoro, (eds). El asentamiento Neguev: Interaccidn de campesinos y estado en el aprovechamiento de los recursos naturales. Informe t6cnico no. 162; Program paper no. 7. CATIE-AUW-MAG Atlantic Zone Programme, Turrialba, Costa Rica.

Rose, S., and J. Paranka. 1987. Root and VAM distribution in tropical agricultural and forest soils. Pages 56 in D. Sylvia, L. Hung, and J. Graham, (eds). Mycorrhizae in the next decade: practical applications and research priorities. Proc. 7th N.A. Conf. on Mycorrhizae. Institute of Food and Agricultural Sciences, University of Florida, Gainsville, Fla. and not resin- extractable P.

Rosendahl, S., C. Rosendahl, and U. Sochting. 1990. Distribution of VA mycorrhizal endophytes amongst plants from a Danish grassland community. Agric. Ecosys. Environ. 29:329-335.

Roskoski, J. 1981. Nodulation and N2 fixation by Inga jinicuii, a woody legume in coffee plantations. Plant Soil 59:201-206.

Russell, J., E. Kamprath, and C. Andrew. 1988. Phosphorus sorption of subtropical acid soils as influenced by the nature of the cation suite. Soil Sci. Soc. Am. J. 52:1407-1410.

Russo, R., and E. Mora. 1986. Effect of vesicular-arbuscular mycorrhizae on the nodulation off. poeppigiana. Pages 101-104 in Final project technical report: Erythrina spp. (Phase I: 3-P-82-001 5). CATIE, Turrialba, Costa Rica.

Russo, R., and G. Budowski. 1986. Effect of pollarding frequency on biomass of Erythrina poeppigiana as a coffee shade tree. Agrofor. System 4:145-162.

Russo, R. 1990. Erythrina (Leguminosae: Papilionoideae): a versatile genus for agroforestry systems in the tropics. J. Sustain. Agric. 1(2):89-109.

Sa, T., and D. Israel. 1991. Energy status and functioning of phosphorus- deficient soybean nodules. Plant Physiol. 97:928-935. 3 9 2 Salazar, A., and C. Palm. 1990. Effects of P fertility on rice and cowpea yields alley-cropped with leguminous trees. Agron. Soc. Am. Abstr. .

Salinas, and Sanchez P.A. 1976. Soil-plant relationships affecting varietal and species differences in tolerance to low available soil phosphorus. Ciencia e Cultura 28(2): 156-168.

Sanchez, P., and T. Cochrane. 1980. Soil constraints in relation to major farming systems of tropical America. Pages 107-139 in Soil related constraints to food production in the tropics. International Rice Research Institute (IRRI), Los Banos, Phillipines.

Sanchez, P. 1987. Soil productivity and sustainability in agroforestry systems. Pages 117-138 jn H. Steppler and P. Nair, (eds). Agroforestry: A decade of development. ICRAF, Nairobi, Kenya.

Sanchez, P., and J. Benites. 1987. Low-input cropping for acid soils of the humid tropics. Science 242:1521-1527.

Sanyal, S., and S. De Datta. 1991. Chemistry of phosphorus transform ations in soil. Advan. Soil Sci. 16:2-120.

Saunders, W. 1965. Phosphate retention by New Zealand soils and its relationship to free sesquioxides, organic matter and other soil properties. New Zealand J. Agri. Res. 8:30-57.

Saunders, W., and A. Metson. 1971. Seasonal variation of phosphorus in soil and pasture. N.Z. J. Agric. Res. 14:307-328.

Schlesinger, W., and M. Hasey. 1981. Decomposition of chaparral shrub foliage: losses of organic and inorganic constituents from deciduous and evergreen leaves. Ecology 62(3):762-774.

Searle, P., and B. Daly. 1977. The determination of aluminum, iron, manganese and silicon in acid oxalate soil extracts by flame emission and atomic absorption spectrometry. Geoderma 19:1-10.

Serrao, E., I. Falesi, J. Veiga, and J, Neto. 1979. Productivity of cultivated pastures on low fertility soils in the Amazon of Brazil. Pages 195-225 in P. Sanchez and L. Tergas, (eds). Pasture production in acid soils of the tropics. Centro Internacional de Agriculture Tropical (CIAT), Cali, Colombia.

Sharpley, A. 1991. Soil phosphorus extracted by iron-aluminum-oxide- impregnated filter paper. Soil Sci. Soc. Am. J. 55(4):1038-1041. 39 3

1985. Phosphorus cycling in unfertilized and fertilized agricultural soils. Soil Sci. Soc. Am. J. 49:905-911.

Shaw, N., L. ’tMannetje, R. Jones, and R. Jones. 1985. Pasture measurements. Pages 235-250 in N. Shaw and Bryan W.W., (eds). Tropical pasture research: principles and methods. CAB Bulletin 51. Commonwealth Agricultural Bureau, Lavenham Press Ltd., Suffolk, England.

Shoji, S., and T. Ono. 1978. Physical and chemical properties and clay mineralogy of andosols from Kitakami, Japan. Soil Science 126:297-312.

Shongwe, G, N. 1989. Effect of pH on Al-orthophosphate interactions in some acidic Swaziland soils. Pages 102-105 in . Ph.D. dissertation. Ohio State University, Columbus, Ohio. 260.

Sibbesen, E. 1977. A simple ion-exchange resin procedure for extracting plant-available elements from soil. Plant Soil 46:665-669.

. 1978. An investigation of the anion-exchange resin method for soil phosphate extraction. Plant Soil 50:305-321.

Sieverding, E. 1983. Manual de metodos para la investigacidn de la micorrfza vesiculo-arbuscular en el laboratorio. CIAT, Cali, Colombia.

Sieverding, E., and S. Toro T. 1987. Growth of coffee and tea plants in nurseries inoculated with different vam fungal species. Pages 58. in D. Sylvia, L, Hung, and J. Graham, (eds). Mycorrhizae in the next decade: practical applications and research priorities. Proc. 7th N.A. Conf. on Mycorrhizae. Institute of Food and Agricultural Sciences, University of Florida, Gainsville, Fla.

Sieverding, E. 1989. Class handouts for course on VA-mycorrhizae, 366- 374. Universidad de Costa Rica, San Jose, Costa Rica.

Sieverding, E., and S. Saif. 1984. VA mycorrhiza management: a new, low cost, biological technology for crop and pasture production on infertile soils? Unpublished discussion paper for CIAT Annual Review. CIAT, Cali, Colombia.

Sieverding, E. 1990. Ecology of VAM fungi in tropical agrosystems. Agric. Ecosys. Environ. 29:369-390. 3 9 4 Silvopastoral Systems Project. 1989. Srstemas silvopastoriles para el trOpico humedo bajo. Informe final: primera fase. Centro Agrondmico Tropical de Investigacibn y Ensenanza (CATIE), Turrialba, Costa Rica. 184.

Singh, B., and J, Jones. 1976. Phosphorus sorption and desorption characteristics of soil as affected by organic residues. Soil Sci. Soc. Am. J. 40:389-394.

Singh, K., H. Chauhan, D. Rajput, and D. Singh. 1989. Report of a 60 month study on litter production, changes in soil chemical properties and productivity under Poplar (P. deitoides) and Eucalyptus (E. hybrid) interplanted with aromatic grasses. Agroforestry Systems 9:37-45.

Singh, P. 1990. Agrosilvopasture systems in India. Pages 183-195 in C. Devendera, (ed). Shrubs and tree fodder for farm animals. Proc. Workshop in Denpasar Indonesia, 24-29 July. IDRC, Canada.

Singh, Y., B. Singh, M. Maskina, and O. Meelu. 1988. Effect of manures, crop residues and green manure (Sesbania acuJeata) on nitrogen and phosphorus transformations in a sandy loam at field capacity and under waterlogged conditions. Biol. Fertil. Soils 6:183-187.

Siqueira, J. 1987. Mycorrhizal benefits to some crop species in a P-deficient oxisol of southeastern Brazil. Pages 59 iQ 0. Sylvia, L. Hung, and J. Graham, (eds). Mycorrhizae in the next decade: practical applications and research priorities. Proc. 7th N.A. Conf. on Mycorrhizae. Institute of Food and Agricultural Sciences, University of Florida, Gainsville, Fla.

Skogley, E. 1991. The universal bioavailability environment/soil test UNIBEST Unpublished paper given at International Symposium on Soil Testing and Plant Analysis in the Global Community. Orlando, Florida, Aug 22-27.

Skogley, E., S. Georgitis, J. Yang, and B. Schaff. 1990. The phytoavailability soil test - PST. Commun. in Soil Sci. Plant Anal. 21:1229- 1243.

Smeck, N. 1985. Phosphorus dynamics in soils and landscapes. Geoderma 36:185-199.

Snaydon, R. 1981. The ecology of grazed pastures. Pages 13-29 in F. Mortey, (eds). World Animal Science: grazing animals. Elsevier Sci. Publ. 395 Soil survey staff. 1972. Soil survey laboratory methods and procedures for collecting soil samples: Method 6A2b. USDA Soil Conservation Service. Soil survey investigations. Report No. 1. U.S. Govt. Printing Office, Washington, D.C.

1990. Keys to soil taxonomy 1q SMSS Technical monograph no. 19 4th ed. AID, USDA, SMSS, Virginia Polytechnic Institute.

Sollins, P., and R. Radulovich. 1988. Effects of soil physical structure on solute transport in a weathered tropical soil. Soil Sci. Soc. Amer. J. 52{4):1168-1173.

Sollins, P., P. Robertson, and G. Uehara. 1988. Nutrient mobility in variable- and permanent-charge soils. Biogeochem. 6:181-199.

Sonzogni, W., S. Chapra, D. Armstrong, and T. Logan. 1982. Bioavailability of phosphorus inputs to lakes. J. Environ. Qual. 11 (4):555-563.

Sposito, G., and S. Mattigod. 1980. Geochem: A computer program for the calculation of chemical equilibria in soil solutions and other natural water systems. The Kearney Foundation of Soil Science, University of California, Riverside, Riverside, California. 92 p.

Sposito, G. 1989. The chemistry of soils. Oxford University Press, Inc., New York, NY. 277 p.

Stewart, J., and H. Tiessen. 1987. Dynamics of soil organic phosphorus. Biogeochem. 4:41-60.

Stribley, D. 1987. Mineral Nutrition. Pages 59-70 in G. Safir, (ed). Ecophysiology of VA-mycorrhizal plants. CRC Press, Boca Raton, Fla.

Stribley, D., P. Tinker, and R. Snellgrove. 1980. Effect of vesicular- arbuscular mycorrhizal fungi on the relations of plant growth, internal phosphorus concentration and soil phosphate analyses. J. Soil Science 31:655-672.

Suarez, J., A. Senra, and J. Galindo. 1981. Study of the contribution of nutrients to the soil by the faeces of cows grazing in three rotational systems. J. Agric. Sci. 15:95-102.

Swamy, P., and P. Ramakrishnan. 1987. Contribution of Mikania micrantha during secondary succession following slash and burn agriculture (Jhum) in northeast India II. nutrient cycling. Forest Ecol. Mgmnt. 22:239-249. 3 9 6

Swift, M. 1985. Tropical soil biology and fertility: planning for research |q Report of 2nd workshop of TSBF. Special Issue 9. International Union of Biological Sciences, Biology International, Paris, France.

1986. Integration of nutrient cycling in the plant-soil system: the role of biological processes. Pages 30-37 in R. Prinsley and M. Swift, (eds). Amelioration of soils by trees. Commonwealth Science Council, London, England.

Szott, L., C. Palm, P. Sanchez, J. Perez, E. Fernandez, A. Salazar, R. Scholes, B. Panshanasi, and C. Davey. 1990. Agroforestry systems for acid soils in the humid tropics, unpublished report. Tropical Soils Research Program, North Carolina State University and Instituto Nacional de Investigaciones Agropequaria e Agroindustrial, Peru.

Tate, K. 1984. The biological transformation of P in soil. Plant Soil 76:245- 256.

Taylor, R. 1987. On the accuracy of insecticide efficacy reports. Forum: Environ. Entomol. 16(1): 1-8.

. 1978. The relationship between density and distance of dispersing insects. Ecol. Entomol. 3:63-70.

Thomas, G. 1982. Exchangeable cations. Pages 159-166 in A. Page, R. Miller, and D. Keeney, (eds). Methods of soil analysis part 2: Chemical and microbiological properties. 2nd ed. ASA, SSSA, Inc., Madison, Wisconsin.

Tiessen, H., J. Stewart, and C. Cole. 1984. Pathways of phosphorus transformations in soils of differing pedogenesis. Soil Sci. Soc. Am. J. 48:853-858.

Till, A., and P. May. 1971. Nutrient cycling in grazed pastures IV. the fate of sulphur-35 following its application to a small area in a grazed pasture. Austr. J. Agric. Res. 22:391-400.

Toledo, J., and F. Torres. 1990. Potential of silvopastoral systems in the rain forest. Pages 35-52 in Agroforestry land-use systems. Spec. Publ. No. 90-02. Nitrogen Fixing Tree Association (NFTA), Honolulu, Hawaii.

Torres, F. 1983. Role of woody perennials in animal agroforestry. Agrofor. Syst. 1:131-163. 3 9 7 Traina, S., G. Sposito, D. Hesterberg, and U. Kafkafi. 1986. Effects of pH and organic acids on orthophosphate solubility in an acidic, montmorillonitic soil. Soil Sci. Soc. Am. J. 50:45-52.

unauthored. 1986. Unpublished survey data of livestock production in Neguev settlementCATIE-AUW-MAG Atlantic Zone Program, Guapiles, Costa Rica.

Vaidyanathan, L., and P. Nye. 1966. The measurement and mechanism of ion diffusion in soils. II. An exchange resin paper method for measurement of the diffusive flux and diffusion coefficient of nutrient ions in soils. J. Soil Science 17(2): 175-183. van der Kamp, E. 1990. Aspectos econbmicos de la ganaderla en pequena escala y de la ganaderfa de carne en la zona btlantica de Costa Rica. Field report no. 51. CATIE-AUW-MAG Atlantic Zone Programme, Turrialba, Costa Rica, August. van der Kruijs, A., J. van der Heide, and B. Kang. 1989. Observations on decomposition rates of leaves of several shrub and tree species applied as mulch under humid tropical conditions. Pages 361-366 in J. van der Heide, (ed). Nutrient management for food crop production in tropical farming systems . Institute for Soil Fertility, Haren, The Netherlands. van der Zee, S., L. Fokkink, and W. van Riemsdijk. 1987. A new technique for assessment of reversibly adsorbed phosphate. Soil Sci. Soc. Am. J. 51:599-604. van Dooremolen, W., W. Wielemaker, N. van Breemen, E. Meijer, and L. van Reewijk. 1990. Chemistry and mineralogy of andosols of various age in a soil chronosequence on andesitic lahars in Costa Rica. Chem, Geol. 84:139- 141.

. 1990. Chemistry and mineralogy of andisols of various age in a soil chronosequence on andesitic lahars in Costa Rica. Pages 39-42 in W. Wielemaker, (eds).Atlantic Zone Programme Workshop on soil information; Working document no. 10. CATIE-AUW-MAG, Turrialba, Costa Rica. van Kessel, C., P. Singleton, and H. Hoben. 1985. Enhanced N-transfer from a soybean to maize by vesicular-arbuscular mycorrhizal (VAM) fungi. Plant Physiol. 79:562-563. van Noordwijk, M., and Y. Dommergues. 1990 . Root nodulation: the twelfth hypothesis. Agroforestry Today 2(21:9-10. 39 8 van Veen, J., J. Ladd, J. Martin, and M. Amato. 1987. Turnover of carbon, nitrogen and phosphorus through the microbial biomass in soils incubated with 14C-, 1°N-, and 32P-labelled bacterial cells . Soil Biol. Biochem. 19(5):559-565.

Vargas G., E., and H. Fonseca Z. 1989. Contenido mineral y protbico de for raj es para ruminantes en pastoreo en Costa Rica. Universidad de Costa Rica, San Jose, Costa Rica.

Vilas Boas, O. 1990. Descomposicibn de hojarasca y mineralizacibn del nitrbgeno de la matbria orgbnica del suelo bajo cuatro sistemas agroforestales en Turrialba, Costa Rica. Tesis de Maestria, CATIE. Cbntro Agronbmico Tropical de Investigacibn y Ensenanza (CATIE), Turrialba, Costa Rica. 152 p.

Violante, A., C. Colombo, and A. Buondonno. 1991. Competitive adsorption of phosphate and oxalate by aluminum oxides. Soil Sci. Soc. Am. J. 55:65- 70.

Virginia, R. 1986. Soil development under legume tree canopies. For. Ecol. Manag. , vol. 16.

Vitousek, P., and J. Denslow. 1986. Nitrogen and phosphorus availability in treefall gaps of a lowland tropical rainforest. J. Ecol. 74:1167-1178.

Vitousek, P. 1984. Litterfall, nutrient cycling, and nutrient limitation in tropical forests. Ecology 65(1 ):285-298.

Vitousek, P., and R. Sanford. 1986. Nutrient cycling in moist tropical forest. Ann. Rev. Ecol. Syst. 17:137-167.

Vitousek, P., and J. Denslow. 1987. Differences in extractable phosphorus among soils of the La Selva biological station, Costa Rica. Biotropica 19(2): 1 67-170.

Wada, K. 1980. Mineralogical characteristics of andosols. Pages 87-107 in 8. Theng, (ed). Soils with variable charge. New Zealand Society of Soil Science, Palmerston North, New Zealand.

Walbridge, M. 1991. Phosphorus availability in acid organic soils of the lower North Carolina coastal plain. Ecology 72(6):2083-2100. 3 9 9 Wallace, L. 1981. Growth, morphology and gas exchange of mycorrhizal and nonmycorrhizal Panicum co fora turn L., a C4 grass species, under different clipping and fertilization regimes. Oecologia 49:272-278.

1987. Mycorrhizas in grasslands: interactions of ungulates, fungi and drought. New Phytol. 105:619-632.

Warkentin, B., and T. Maeda. 1980. Physical and mechanical characteristics of andosols. Pages 281-302 in B. Theng, (ed). Soils with variable charge. New Zealand Society of Soil Science, Offset Publ., Palmerston North, New Zealand.

Weider, R., and G. Lang. 1982. A critique of the analytical methods used in examining decomposition data obtained from litter bags. Ecology 63(6):1636-1 642.

Werner, P. 1984. Changes in soil properties during tropical wet forest succession in Costa Rica. Biotropica 16(11:43-50.

Wielemaker, W., and A. Lansu. 1991. Land-use changes affecting classification of a Costa Rican soil. Soil Sci. Soc. Am. J. 55:1621-1624.

Wielemaker, W. (ed.). 1990a. Atlantic Zone Programme Workshop on soil information . Working document no. 10. CATIE-AUW-MAG, Turrialba, Costa Rica.

Wielemaker, W. 1990b. Clima, geologia y suelo. Pages 18-30 in M. de Onoro, (eds). El asentamiento Neguev: Interaccibn de campesinos y estado en el aprovechamiento de los recursos naturales. Informe tecnico no. 162; Programme paper no. 7. CATIE-AUW-MAG Atlantic Zone Programme, Turrialba, Costa Rica.

Wild, A. 1988. Plant nutrients in soil: phosphate. Pages 695-742 In A. Wild, (ed). Russell's soil conditions and plant growth. 11th ed. Longman Scientific & Technical and Wiley & Sons, Inc., New York, NY. orr.

Wilkinson, L. 1990. SYSTAT: The system for statistics. SYSTAT, Inc., Evanston, IL.

Williams-Linera, G., and J. Ewel. 1984. Effect of autoclave sterilization of a tropical andept on seed germination and seedling growth. Plant and Soil 82:263-268. 400 Wilson, G., B. Hetrick, and A. Schwab. 1991. Reclamation effects on mycorrhizae and productive capacity of flue gas desulfurization sludge. J. Environ. Qual. 20:777-783.

Yang, J., E. Skogley, and B. Schaff. 1991. Nutrient flux to mixed-bed ion exchange resin: temperature effects. Soil Sci. Soc. Am. J. 55(31:762-767.

Yang, J., E. Skogley, S. Georgitis, B. Schaff, and A. Ferguson. 1991. Phytoavailability soil test: development and verification of theory. Soil Sci. Soc. Am. J. 55(5):1358-1365.

Yavitt, J., and R. Wieder. 1988. Nitrogen, phosphorus and sulfur properties of some forest soils on Barro Colorado Island, Panama. Biotropica 20(1 ):2- 10.

Yost, R., and R. Fox. 1979. Contribution of mycorrhizae to P nutrition of crops growing on an oxisol. Agron. J. 71:903-908.

Young, A. 1986. Effects of trees on soil. Pages 10-19 in R. Prinsley and M, Swift, (eds). Amelioration of soil by trees. Commonwealth Science Council, London, England.

1989. Ten hypotheses for soil-agroforestry research. Agroforestry Today 1(1>:13-16.

Zelazny, L. , and J. Qureshi. 1972. Chemical pretreatment effects on clay separation and mineralogical analyses of selected Florida soils. Soil Crop Sci. Soc. Florida Proc. 32:117-121.

Zou, X. 1992. Phosphorus transformations in soils: a new method and tree species effects. Ph.D. Dissertation. Colorado State University, Fort Collins, CO.