Parameterization and evaluation of mechanistic crop models for estimating Urochloa brizantha cv. BRS Piatã productivity under full sun and in silvopastoral system.

Portuguese. Co-orientador: José Ricardo Macedo Pezzopane, Embrapa Pecuária Sudeste.

English. Silvopastoral systems are a kind of agroforestry system in which trees or shrubs are combined with animals and pastures. Silvopastoral systems are important to intensify pasture production and mitigate climate change effects. However, very few studies have been performed to adapt crop models to simulate these systems. The aim of this study was to parameterize and test the mechanistic crop models APSIM and CROPGRO for estimating Urochloa brizantha cv. BRS Piatã productivity under full sun and in a silvopastoral system, to evaluate the models' performance to simulate tree-crop interactions, and to develop tools to improve these simulations. For this purpose, four field experiments were conducted under full sun to investigate cutting management under irrigated and rainfed conditions and grazing management under rainfed conditions with high and low N supply. Another experiment was carried out in a silvopastoral system with the trees arranged in simple rows, in East-West orientation, with 15 m between rows and 2 m between plants in the rows. This experiment was conducted under grazing management and rainfed conditions with the pasture variables, microclimate and soil water content being assessed at four distances from the North row (0.00 m, 3.75 m, 7.50 m and 11.25 m). The forage mass simulations for the pasture at full sun, performed using the APSIM-Tropical Pasture model, showed good agreement between observed and estimated data (R2 between 0.82 and 0.97, d between 0.92 and 0.98, and NSE ranging from 0.72 to 0.92), while the simulations with the CROPGRO-Perennial Forage model achieved good precision (R2 between 0.65 and 0.93) and good accuracy (d from 0.86 to 0.97, and NSE from 0.60 to 0.90), for the various managements and environmental conditions. Even considering the promising performance of both models for pastures under full sun, they already needs to be tested in other locations, climate conditions, soils, and grazing or cutting intensities, to prove its accuracy and reach enough confidence. The pasture growth simulations at the silvopastoral system indicated that the APSIM-Tropical Pasture was efficient when only competition by solar radiation was considered (R2 from 0.69 to 0.88, d from 0.90 to 0.96, and NSE between 0.51 and 0.85), but inefficient when considering only competition by soil water (R2 between 0.58 and 0.85, d between 0.58 and 0.82, and NSE from -4.07 to -0.14). The CROPGRO-Perennial Forage achieved good performance on pasture growth simulation at the distances 0.00 m, 3.75 m, and 7.50 m from the trees (R2 from 0.75 to 0.90, d from 0.93 to 0.96, NSE between 0.74 and 0.85). Despite the good results, improvements should be performed in both models for simulating all factors that affect forage growth in silvopastoral systems.