1/15 Land use and crop evapotranspiration in Tensift/Marrakech plain: inter-annual analysis based on MODIS satellite data I. Benhadj (1) , Duchemin B. (1) , Er-Raki S. (2) , Hadria R. (1) , Maisongrande P. (1) , V. Simonneaux (1,2) , B. Mougenot (1) , S. Khabba (2) , Kharrou M.H. (3) , A. Chehbouni (1) (1) Centre d’Etudes Spatiales de la BIOsphère (CESBIO) Unité mixte de recherche UPS-CNRS-CNES-IRD 18 avenue. Edouard Belin,bpi 2801– 31401 Toulouse cedex 9, France (2) Faculté des Sciences Semlalia, Université Cadi Ayyad de Marrakech Avenue Prince My Abdellah, BP 2390 – 40000 Marrakech, Maroc (3) ORMVAH, Office Régional de Mise en Valeur Agricole du Haouz Avenue Hassan II, BP 2411 - 40000 Marrakech, Maroc Corresponding author: duchemin@ird.fr Abstract In this study, we aim at monitoring the water balance over the semi-arid plain of Tensift/Marrakech, a 2000 km² intensively cropped area in center of Morocco. This requires firstly to map the land use, secondly to monitor the vegetation dynamics, and thirdly to evaluate evapotranspiration, which is the key-variable of water balance in semi-arid plains. In this context, we investigate the potential offered by Terra-MODIS satellite, which provide a costless daily global coverage of the Earth. We use a six-year archive of 16-day composite NDVI images from 2000-2001 to 2005-2006 agricultural seasons. However, the use of medium (250 m) spatial resolution data makes difficult to directly monitor land surfaces. Indeed, each pixel (mixed pixel) generally includes different types of surface, and consequently its spectral response results from the contribution of each land classes. In a first phase, the land use and the vegetation dynamics are retrieved using linear unmixing model applied on NDVI time series. Identification of end-members, i.e. the specific NDVI time course of individual land classes, is based on the assumption that pure pixels can be identified directly from MODIS NDVI images. The approach is set up to map the land use fractions of the three classes that are the most important for agricultural water management in the study area (non cultivated areas, orchards, annual crops). In a second phase, the information on land use and vegetation dynamics is used to estimate evapotranspiration. The method is adapted from the FAO-56 algorithm, which computes crop water needs from a reference evapotranspiration (ETo) and cultural crops coefficients (Kc). ETo is calculated by applying spatial interpolation of the meterorological data available in the study area. The crop coefficients, which vary according to the crop type, phenological stage and soil water content, are retrieved for each land classes according to their NDVI time courses using various scenarii of irrigation. The spatio-temporal patterns of evapotranspiration maps are analysed to regional driving variables (climate and water availability).