Yield and plant height predictions of irrigated maize through unmanned aerial vehicle in North Florida

Monitoring crops and producing correct yield estimates are both crucial components of decision-making in agricultural systems. This is particularly important for crops like maize, with a national gross production value above US $82 billion in 2021. Several crop parameters, e.g., plant height (PH), leaf nutrient status, etc., are in use for improving maize grain yield (MGY) prediction models. Traditional PH measurements can be time-consuming and labor-intensive; thus, unmanned aerial vehicles (UAVs) might be efficient and practical tools to predict in-season MGY. Two small-plot trials were conducted in North Florida to estimate in-field PH (IFPH) using UAVs (UAV-PH) and compare different regression approaches to estimate MGY using different time points (from 17 to 133 days after planting [DAP]). A multispectral camera was mounted onto an Aurelia X6 Standard hexacopter to capture crop canopy height. Overall, the IFPH values were higher than UAV-PH but strongly correlated (r = 0.95). Moderate to strong correlations were observed between MGY and PH at all time points except 17 DAP. The prediction of MGY using IFPH or UAV-PH was better through machine learning algorithms (Random forest regressor [RFR] and Support vector regression [SVR]) compared to simple/multiple linear regression (S/MLR), with R2, MAE, RMSE, and nRMSE ranges of 0.91–0.94, 553.86–1574.62 kg ha−1, 708.94–195.04 kg ha−1, and 14.2–39.1 % for best-performing models of RFR and SVR, respectively. Furthermore, S/MLR usually overestimated low-yielding plots (MGY < 12000 kg ha−1) and underestimated high-yielding plots (MGY > 12000 kg ha−1). The results showed that UAVs are a potential tool for quicker decision-making in maize cropping systems in North Florida since the crop-height model accurately estimated ground PH measurements (IFPH) at different time points.