Mapping Eight Crops in Monocropping and Intercropping Farming Systems across Multiple Growing Cycles

Crops in eight farming system classes were identified in monocropping and intercropping systems across multiple growing cycles. It is imperative not to treat the rainy season with multiple growing cycles as a single temporal block. Using the case of the Guinea Savannah of southwest Nigeria (SGS), early maize, late maize, early cassava, late cassava, yam, rice, maize-cassava intercropping, and others (comprising sweet potato, cocoyam, and cowpea) were identified. By incorporating field-based crop inventories, cropping callendars and farmers' farming practices (e.g., crop combinations and crop sequencing) into the workflow, we thus provide a nuanced understanding of the temporal characteristics of cropping practices in complex smallholder farming systems. Sentinel-1 and Sentinel-2 image data were combined into a total of seven experiments. The S1 monthly + S2 bimonthly model's class-wise accuracy exceeded 0.75 for all classes. The analysis of variable importance showed that Sentinel-1 based VV, VH and Sentinel-2 Blue and SWIR1 bands were the most discriminative features. To ensure the scalable approach used for accounting for inter-growing cycle crop dynamics can be adapt for mapping crop mixtures in similar regions where crops grow across multiple growing cycles, we provide scripts in GitHub crop type mapping codes and datasets below.