This paper presents an approach to the automated detection of vegetation level change using a Planet mosaic dataset. The approach is applied to the data set covering the state of Queensland, Australia. This dataset consists of a time series of high resolution (3m²) RGB satellite image mosaics each comprised of approximately 95,000 image tiles. The low number (3) of spectral bands combined with the large and diverse geographic area create a variety of challenges for change detection. Image segmentation followed by random forest regression was used to create ranked lists of image tiles containing potential change. The overall approach is conceptually similar to a document retrieval problem in which potentially relevant tiles are displayed to users for inspection. Promising initial results were obtained measured using precision and recall.