Predicting In-season Sorghum Yield Potential Using Remote Sensing Approach: A Case Study Of Kano In Sudan Savannah Agro- Ecological Zone, Nigeria

Estimating crop yield prior to harvest using remote sensing techniques has proven to be successful. However, accuracy of estimation still varies across crops and landscapes. This study was conducted to examine the applicability of Sentinel-2B for estimating sorghum yield during the 2018 rainy season in three locations (Bebeji, Dawakin Kudu and Rano) within the Sudan Savannah agro-ecological zone of Nigeria. SAMSORG 45 (an early maturing improved sorghum variety) was established in five (5) randomly selected farmer plots in each of the three LGAs. The relationship among different vegetation indices, Normalized Difference Vegetation Index (NDVI), Green Normalized Difference Vegetation Index (GNDVI), Ratio Vegetation Index (RVI) and grain yield were determined using linear regression analysis. Models at different growth stages were then compared using root mean square error (RMSE), coefficient of variations (CV) and coefficient of determination (R2) respectively. The results from the statistical analysis showed that NDVI was superior to GNDVI and RVI for grain yield estimation, indicating low RMSE, high R2 and low CV values at early vegetative (40 days after sowing, DAS), reproductive stage, and entire crop-life cycle. The estimate at 40DAS, reproductive stage, and entire crop-life cycle showed RMSE of 0.04, 0.03, 0.02, R2 (0.75, 0.77 0.93), CV (13.7%, 27.3%, 39.2%) respectively. In addition, RVI had the best fit for stalk yield estimates, having RMSE (0.06, 0.04, 0.01), R2 (0.5, 0.83, 0.98) and CV (15.7%, 19.9% 38.5%) at 70DAS, reproductive stage, and entire crop-life cycle respectively. This study therefore concludes that sorghum yield could be accurately predicted in-season with NDVI and RVI for grain and stalk yields using Sentinel-2B.