Empirical Study of the Accuracy of Remote Sensing Data for Crop Area Estimation in Kano State

Reliable crop area statistics are essential for food security planning in Nigeria, and empirically evaluating the accuracy of satellite remote sensing as a substitute for costly ground surveys fills an important methodological gap for North West Nigerian agricultural monitoring. This study evaluated the accuracy of Sentinel-2 multispectral satellite imagery for estimating cultivated crop area across 12 LGAs in Kano State during the 2022 growing season. Ground truth data were collected through a GPS-based field survey of 1,200 randomly located sample plots across the study area. Supervised classification using Random Forest, Support Vector Machine, and Maximum Likelihood algorithms was applied to cloud-free composite imagery. Random Forest classification achieved an overall accuracy of 87.4 percent and a kappa coefficient of 0.83 for eight land cover classes including five crop types. Maize and sorghum classification achieved the highest producer accuracy at 91.2 percent and 89.7 percent respectively. Groundnut and cowpea, having similar spectral signatures, showed the most classification confusion. Time-series NDVI analysis correctly identified crop growth stages for 84.6 percent of validated plots. The satellite-derived area estimate differed from KNARDA official statistics by 11.3 percent, with the satellite estimate consistently higher, suggesting that official statistics underestimate cultivated area. The study fills a methodological gap in remote sensing validation for northern Nigerian agricultural monitoring and recommends integrating Sentinel-2 based crop area estimation into the KNARDA annual agricultural statistics production framework.

Keywords: remote sensing, crop area estimation, Kano State, Sentinel-2, Random Forest classification