This article develops models for environmental data recorded by meteorological satellites. In general, such data are continuously available for suitable space and time units and are intrinsically nonstationary. Space-time auto-regression (STAR) is a class of models that can be used in monitoring and forecasting, but it must be adapted to nonstationary processes. A set of adaptive recursive estimators is then proposed to estimate STAR parameters that change both over space and time. An extensive application to the normalized difference vegetation index (NDVI), for a region of sub-Saharan Africa, illustrates and checks the approach.