Do UAV Datasets Correlate with the Yield of Durum and Bread Wheat Cultivars? Implications for different Nitrogen Application rates

Quantifying yields provides essential information for adjusting planting dates, enhancing crop breeding productivity, and regulating agricultural inputs. Wheat is a major cereal crop that remains vulnerable to abiotic and biotic stresses, all affecting yields. To identify high-yielding wheat cultivars, a randomized block (RB) field experiment was conducted with ten durum and ten bread wheat cultivars planted in triplicate under low (24 kg/ha) and high (48 kg/ha) nitrogen application rates. Unmanned aerial vehicle (UAV) imagery was collected at the grain-filling wheat stage. These UAV images were processed using Pix4Dmapper following a Pix4D manual guide, including the following steps: (1) initial processing; (2) point cloud and mesh; and (3) figital surface model (DSM), Orthomosaic, and index. The correlation coefficient (r) was used to assess the relationship between crop height, spectral bands, and vegetation indices with durum and bread wheat cultivars. A total of 120 crop cut yield samples were collected in all RB subplots, consisting of ten durum and ten bread wheat cultivars collected in triplicate. The harvested triplicates for each wheat cultivar were averaged to yield 30 representative samples (3 samples per cultivar) for durum and bread cultivars to identify high-yielding varieties. A simple linear regression (SLR) model was constructed using highly correlated spectral bands and vegetation indices, along with three samples of mean values per ten cultivars of durum and bread wheat to estimate yield. The findings showed that all evaluated spectral bands and indices were highly correlated (r ≥ 0.8 ±) with measured yield values. The SLR model produced varied yield estimation results, with the coefficient of determination (R²) ranging from 0.6 to 0.99 and root mean square error (RMSE) from 0.001 to 0.301 t/ha for durum and bread wheat cultivars, respectively. The findings from this study showed that all evaluated UAV datasets have a strong relationship with different wheat cultivars and can be used to estimate yield in real time.