Random Forest change classification and automatic feature selection for mapping changes from grasslands to arable lands in the Czech republic

This dataset was created in order to demonstrate mapping changes from grasslands to arable lands with help of new implemented algorithm based on random forest variable importance matrics (MDA and MDG). There are two datasets. Imagery data and training data in order to train random forest classifier. Imagery data contain time series Leaf Area Index dataset composed of 23 images from Sentinel 2 satellite. Training data contain point shapefile that serves as reference dataset for training and validation purpose. Study area is located in the Czech republic with high occurence of changes from grasslands to arable lands.