Vegetation indexes are widely used as a proxy of  vegetation status, they are often used to monitor and assess  qualitatively and quantitatively the growing season. The  Normalized Vegetation Index (NDVI) is the most widely used in  agriculture, frequently as a proxy for different physiological and  agronomical aspects, such as drought stress and crop yield losses  evaluation. NDVI forecast is usually correlated to precipitation  however, in Mediterranean and arid climates, it is not well  correlated due to prolonged dry periods and sparse precipitation  events. In this study, we forecast Mediterranean permanent  grassland NDVI at 7 and 30 days ahead using machine learning  and two soil moisture products as predictors, simulated soil  moisture values and satellite-based Soil Water Index (SWI)  values. Results show that both products can be used as reliable  predictors of permanent grassland in Mediterranean areas.  Predictions at 7 days are more accurate and better forecast the  negative effect of drought on vegetation dynamics than 30 days.  This study shows the potential of using a simple methodology and  readily available data to predict the grassland growth dynamic in  the Mediterranean area .