FAO-56 dual crop coefficient (FAO-2Kc) based model are increasingly applied at large scale for agricultural water monitoring, requiring field-scale data over the spatial extent of interest. Given the lack of in-situ measurements, satellite products can be used to estimate indirectly the parameters through calibration. However, a lack of knowledge about model sensitivity can lead to suboptimal use of satellite data. This study aims to analyze the sensitivity of SAMIR, a FAO-2Kc-based model using satellite data. The Sobol method was applied for evapotranspiration (ET) and deep percolation (DP) simulations on 37 contrasted agricultural seasons. Results indicate that SAMIR's sensitivity mainly depends on the modeled water stress. We proposed a proxy for the model sensitivity which can determine 84% (73%) of the ET (DP) among the agricultural seasons. An interaction analysis allowed reducing the calibration problem to the adjustment of only two parameters (a_Kcb and Zr_max), accounting for most of the sensitivity.