The use of crop models in sloping areas is questionable when relief is not taken into account, as relief affects infiltration, radiation, and aerodynamic processes. The objective of this work is to evaluate the performance of FAO-AquaCrop model in simulating crop evapotranspiration, water balance, and biomass production in hilly areas using in situ measurements. The experiment was conducted in the Cap-Bon region, north-eastern Tunisia, on two wheat fields located on opposite sloping rims (A and B) and one control field (C) on a flat terrain: field A is SE-oriented with 5% slope and B is NW-oriented with 6% slope. Three flux stations were used to monitor automatically actual evapotranspiration (ET) and climatic factors whereas soil moisture and biomass production were measured manually. Model’s outputs were compared to actual measurements using statistical indicators: slope of the regression line, root mean squared error (RMSE), and the coefficient of determination (R2). Actual ET varied between 1 and 2 mm during crop initial stage and 3–4 mm during mid-season stage. The ET/ETo ratio during mid-season was 0.81, 0.74, and 1.03, respectively for fields A, B, and C, well below the commonly used value of 1.15. Comparison between measured and simulated ET shows a substantial overestimation of the model in sloping fields with 6–20% higher averages and a RMSE of 0.47–0.77 mm/day. AquaCrop seems to simulate reasonably well water balance, particularly in flat conditions. RMSE of water content in the top 100 cm soil-layer was in the range 41–67 mm/m, representing a relative error of 11–21%. Simulated and measured biomass values presented similar trends (R2 = 0.86–0.94) with a systematic difference, indicating that AquaCrop outputs could be improved by a correction factor.