Journal of Advances in Biology & Biotechnology

Accurate crop yield estimation is essential for food-security planning, supply-chain logistics and crop-insurance indexing. However, field-level forecasting in rainfed cotton systems can be affected by human bias, spatial error and seasonal variability. This study evaluated the comparative accuracy, statistical efficiency and predictive reliability of three field sampling protocols for cotton yield estimation under actual farming conditions. Primary data were collected from 31 commercial cotton plots in Sanglud, Morgaon Sadijan and Ural villages in Akola district, Maharashtra. The evaluated methods were the 5 x 5 m crop-cut experiment, plant-based estimation at the open-boll stage and area-based quadrat sampling, assessed across the first and second picking phases. Predictive performance was compared with ground-truth harvested yields using mean actual yield, percentage error, standard deviation and root mean square error (RMSE). The 5 x 5 m crop-cut method showed the closest agreement with actual yield, with low dispersion (s = 2.37) and RMSE (1.24), producing a +5.97% error in the first picking and a -0.29% error in the second picking. The plant-based method overestimated yield by +73.60% in the first picking and underestimated by -31.20% in the second picking, reflecting selection bias and late-season boll attrition. Quadrat sampling showed persistent overestimation (+18.30% and +30.40%; s = 3.73). The findings support the 5 x 5 m crop-cut protocol as the most reliable approach among the methods evaluated.