Field-Level California Crop Maps 2007-2021

Field-Level California Crop Maps 2007-2021

Abstract

Technological advances in satellite image processing have made crop maps readily available over the last decade. Because of the diversity and complexity of crop production in California, however, reliable crop maps for the state are still scant. To fill this gap, we created field-level crop maps of California (hereinafter, Field-Level California Crop Map (FLCCM)) for 2007-2021. We leverage highly accurate ground-truth labels that exist in 2014, 2016, and 2018 to train our crop classifier using probability random forests. We then feed to our classifier the data for predictors that are available from 2007 to 2021. We release three types of crop predictions, and their corresponding accuracy measures in three formats (.csv, .shp, .rds). Our training algorithm can be applied to other settings in which field-level ground-truth data are scarce but fine-resolution pixel-level data are relatively more abundant. 

Disclamer: The dataset is in the process of being peer-reviewed.