Advancing the Use of Synthetic Aperture Radar for Mapping Agricultural Smallholdings Across Diverse Agroecological Zones in Nigeria

Mapping smallholder agricultural fields across Nigeria's diverse agroecological zones is a fundamental challenge for food security monitoring due to field size heterogeneity, inter-cropping complexity, and cloud cover limitations affecting optical sensors during the growing season. This study advances SAR-based methods for mapping smallholder agricultural fields using Sentinel-1 C-band and ALOS-2 L-band imagery across Sudan Savanna, Guinea Savanna, and Derived Savanna agroecological zones. A phenology-based SAR analysis framework exploiting temporal backscatter signatures is developed to distinguish crop from non-crop land and map major crop types including maize, sorghum, millet, cassava, and cowpea. Machine learning classifiers trained with field survey data from 2,400 plots achieve overall classification accuracies of 86 percent for crop versus non-crop and 74 percent for multi-class crop type mapping. The study evaluates the complementarity of multi-frequency SAR and the marginal benefit of fusing SAR with limited-availability cloud-free optical data. A methodology for scaling field delineation using watershed segmentation adapted to sub-hectare field boundaries is developed and tested. Keywords: SAR, smallholder agriculture, crop mapping, Nigeria, agroecological zones