Achieving sustainable quantitative water management represents a major environmental challenge for rural watersheds with water scarcity. In these areas, agriculture uses large volumes of water for irrigation compared to available resources. The agro-socio-hydrological MAELIA model (Therond et al., 2014) was developed to test alternative strategies for water management (change in land use, construction of dams, optimization of irrigation strategies, ...). The model simulates hydrology, agricultural lands and management by various stakeholders such as farmers and dam managers. It requires spatialized input data, including soil data, which are not easily available for large scale study. The objective of this study was to quantify the impact of the uncertainty on Available Water Capacity of soil (AWC) on outputs of the MAELIA model at the watershed scale. The study was conducted on the Aveyron Aval - Lère Watershed (~800 km²) in France which is largely covered by agricultural land and where irrigation is largely spread. Soil data were extracted from the Soil Geographical Database of France (Jamagne et al., 1995) with AWC estimations and uncertainty. An experimental design was developed to evaluate the effect of uncertainty associated with the AWC on agronomic variables (yield, irrigation) and hydrological variables (stream flows) predicted by the MAELIA model. Spatial samples of AWC were generated to quantify uncertainty using a Monte-Carlo approach. Results were analyzed on the regional yields, water irrigation consumption and water flows. Yield predictions were more sensitive to uncertainty of AWC for spring crops than for winter crops. The effect on yield was also greater for rainfed crops as compared to irrigated ones. Overall, the predictions were little impacted by AWC uncertainty, provided that an appropriate AWC sampling accounting for correlation was performed, particularly when analyzing at watershed scale output and pluriannual average.