1. Modelling water dynamics in coffee systems: Parameterization of a mechanistic model over two production cycles in Costa Rica. Pablo Siles, Patrice Cannavo, Julie Sansoulet, Jean-Michel Harmand and Philippe Vaast CATIE (Centro Agronómico Tropical de Investigación y Enseñensa), Turrialba, Costa Rica CIRAD (Centre de Coopération Internationale en Recherche Agronomique pour le Développement), Montpellier, France

2. Introduction Water is key issue in the coffee regions of Central America: –at plot level, complementarity /competition between coffee and various shade tree species –at landscape level, coffee located in mountainous areas (erosion) and rainy zones providing water for communities downstream. –Study with Inga (up to 6 species), predominant genus used in CA (70% shaded coffee)

3. Material and Methods Trial established at CICAFE, Central valley of Costa Rica, Date : 1997 Optimal ecological conditions High altitude (~1200 m), temperate (~22°C) high rainfall (>3000 mm), 3 dry Months slope < 5% Monoculture Coffee (MC): Coffea arabica density : 5000 plants ha -1 Agroforestry System (AFS) Inga densiflora density : 277 trees ha -1 Shade 40-55% Intensive fertilization regime:

4. Monitoring of water fluxes during 2 years GR = I + E + T + Rn + ΔS + D Interception I. densiflora Coffee Runoff Transpiration Soil evaporation Gross Rainfall  Soil water stock Drainage Interception I = GR - (Stemflow +Throughfall)

5. Transpiration : Sap flow (coffee & tree)Runoff Soil water content Stemflow: Inga and coffee

6. Influence of shade trees on throughfall Reduction in throughfall in AFS by 14.4% in 2004 and 7.6% in 2005 20042005 MCAFSMCAFS LAI Coffee4.714.644.603.80 LAI I. densiflora-1.32-1.22

7. Influence of shade trees on stemflow Higher stemflow in AFS (41%) could be explained by differences in architecture of coffee plants (40 cm taller, longer branches) in spite of lower coffee LAI Low contribution of tree stemflow to the system (1% of rainfall)

8. Influence of trees on rainfall interception System Total rainfall Throughfall Stemflow Interception (mm) (%) (mm)(%) (mm)(%) 2004 AFS1426103872.8167*11.722115.5 MC1426121485.184*6.01268.9 2005 AFS1725132476.820411.819611.4 MC1725 143483.2 1247.2 1679.6 74% higher in AFS 18% higher in AFS

9. AFSMC Throughfall*77%83% Tree Stemflow1%- Coffee Stemflow10.5%7% Interception11.5%10% Transpiration34%25% Runoff3%8% Drainage (>200 m)50.5%57% Order of magnitude of various components for 2005 I. densiflora Coffee Runoff Transpiration Soil evaporation Gross Rainfall  Soil water stock Drainage Interception

10. Adaptation of Model “HYDRUS” Comparison of simulated (solid line) and observed (circles) soil volumetric water contents in the 0-30 and 60-90 cm soil layers in AFS with allocation of water uptake in the various soil layers according to root density 0-30 cm soil layer in AFS 60-90 cm soil layer in AFS

11. Water drainage (in mm d -1 ) at 200 cm soil depth in AFS Cumulative values in AF system over wet season

12. Conclusion Shade trees modify the coffee architecture resulting in increased coffee stemflow and a lower throughfall Runoff was decreased in AFS due to coffee architecture and litter cover, hence less soil erosion and better water quality Lower runoff offsets higher interception in AFS, hence a higher infiltration in AFS Higher transpiration in AFS slightly lowers drainage in AFS The soil water content simulated adequately by the Hydrus model (one the first time this model is used in agroforestry) Sound basis to estimate the amount of water drainage and hence nutrient leaching (nitrate) No OVERGENERALIZATION, present AFS with only one tree species (Inga), Andosol (fast infiltration) and high rainfall regime (>2500 mm). The challenges are to use this approach 1) in more complex systems and 2) in conditions of lower precipitation and different soils (currently underway in India) and 3) upscaling.