This document discusses using high-resolution satellite imagery time series to detect land use changes and classify land cover. It presents a project that aims to produce real-time land cover maps using Formosat-2 imagery of southwest France. Supervised classification of imagery from 2008 achieved overall accuracy of 60-72% for differentiating crops, fallows, and various soil work classes like stubble disking and plowing. Change detection between soil states was also explored, with some transitions detected accurately around 90% but many others more difficult to identify reliably. Overall, the results indicate soil state information can help improve crop classification, but direct radiometric classification of soil classes and their changes remains challenging.