MULTI-YEAR ASSESSMENT OF SOIL VEGETATION ATMOSPHERE TRANSFER (SVAT) MODELING UNCERTAINTIES OVER A MEDITERRANEAN AGRICULTURAL SITE Garrigues S (1), Olioso A (1), Calvet J-C (2), Lafont S (2), Martin E (2), Chanzy A (1), Courault D (1), Marloie, O (1), Desfonds V (1) , Bruguier N (1), Renard D (1). (1) UMR 1114, EMMAH, INRA, domaine St Paul-Agroparc, 84000 Avignon, France (2) CNRM-GAME (Météo-France, CNRS), av Coriolis, 31100, Toulouse, France sebastien.garrigues@avignon.inra.fr Mediterranean regions will be particularly sensitive to the impact of climate and land-use changes on water resources and vegetation productivity. These changes will affect the functioning and the dynamic of vegetation as well as the regional water and energy balances, with possible feedbacks on the regional climate. In order to analyze and predict the impact of climate and land-use changes on vegetation productivity and water balance as well as their interactions at the regional scale, a critical step is to quantify the uncertainties associated with land surface models (parametrization, processes) and the data used in these models (atmospheric forcing, vegetation and soil characteristics, crop management practises...). This paper addresses this issue and aims at assessing uncertainties in water (evapotranspiration) and energy flux estimates from Soil Vegetation Atmosphere Transfer (SVAT) models. It focuses on the monitoring of a Mediterranean agricultural site over a large time period (10 years), encompassing several development cycles of distinct crops (wheat, sorghum, maize, peas). The investigated sources of uncertainties include:  The uncertainties in capturing spatial and seasonal variability of atmospheric variables (mainly incoming shortwave radiation, precipitation and air specific humidity) are analyzed comparing simulations forced with local meteorological measurements and simulations forced with re-analysis atmospheric dataset (SAFRAN database).  The representation in the SVAT model of crop rotation, and particularly the influence of the succession of vegetation cover and bare soil on soil moisture and evapotranspiration estimates, are assessed.  The impact on the SVAT simulations of the accuracy of key surface characteristics (soil, vegetation, crop management practises) is then tested comparing simulations feeded with standard values from global reference database (e.g. ECOCLIMAP) and simulations based on in situ or site calibrated values. The ISBA model [1] developed at CNRM/Météo France in the SURFEX surface modeling platform is used in this paper in its A-gs version. ISBA-A-gs [2] simulates the photosynthesis and explicitly describes its coupling with the stomata conductance. This allows representing the impact of [CO2] changes and water stress on both the photosynthesis and the canopy conductance. Besides, ISBA-A- gs can interactively simulate plant growth and mortality to estimate the time course of the plant biomass and the Leaf Area Index (LAI). The experiment was conducted at the INRA, Avignon (France) crop test site (CarboEurope/ICOS site), for which almost 10 years of turbulent fluxes, soil moisture profile and vegetation measurements are available for distinct crop types. The uncertainties in evapotranspiration and energy flux estimates are quantified from both multi- year trend analysis and selected daily cycles spanning a range of atmospheric conditions and phenological stages. Figure 1 and 2 illustrate some preliminary results. While the net radiation flux is correctly simulated, the latent heat flux is globally under-estimated. However, daily plots indicate i) an overestimation of evapotranspiration over bare soil probably due to an overestimation of the soil water reservoir available for evaporation and ii) an under-estimation of transpiration for developed canopy. This paper will further analyse these simulations investigating: - the impact on evapotranspiration estimate of uncertainties in vegetation variables (LAI,