Key points are not available for this paper at this time.
Cotton (Gossypium hirsutum L.) is a vital fiber crop that thrives with proper management. Accurate decision-making also improves cotton productivity. However, many field experiments assume uniform conditions and ignore within-field variability caused by nitrogen doses and cultivar differences. Therefore, this field experiment was conducted to develop UAV-based yield-prediction models for cotton under varying nitrogen doses. Five nitrogen treatments were applied as urea fertilizer rates: T0 (0 kg/ha), T1 (100 kg/ha), T2 (120 kg/ha), T3 (140 kg/ha), and T4 (160 kg/ha), with other fertilizer doses kept constant across two cotton varieties, CB Hybrid-1 and CB HYV-15. The findings revealed that increasing nitrogen fertilizer doses boosted vegetative growth and height, but yield decreased at the highest fertilizer rate for both varieties. The most efficient urea fertilizer treatments were T2 (120 kg/ha) for the high-yielding variety and T3 (140 kg/ha) for the hybrid variety. Stepwise Multiple Linear Regression (MLR) and Least Absolute Shrinkage and Selection Operator (LASSO) models were used to predict cotton yield during early growth stages using UAV-based vegetation indices. The Normalized Difference Red Edge (NDRE) was identified as the most effective predictor in the Stepwise MLR model, which performed well at 72 days after sowing (DAS) (R2 = 0.89). The LASSO model also showed strong performance at 79 DAS (R² = 0.75), with the Normalized Difference Vegetation Index (NDVI) as a key predictor. Overall, this study demonstrates that UAV-derived vegetation indices can effectively predict cotton yield and highlight the impact of nitrogen fertilizer on yield.
Islam et al. (Fri,) studied this question.