A specific vegetation index for monitoring and tracking the spatiotemporal dynamics of wheat powdery mildew from ground to UAV

Wheat powdery mildew (WPM) is a fungal disease that severely affects the growth of winter wheat, leading to yield losses or failure. A universal method for detecting WPM infection is urgently needed to prevent the occurrence or spread of this disease, but it has not been well addressed so far. To meet the challenge, this work proposes a wheat powdery mildew index (WPMI) to detect WPM infection and track its spatiotemporal dynamics across multiple spatial scales. Leaf and canopy spectral reflectance from ground spectrometers and unmanned aerial vehicle (UAV) imagery were acquired over 3 years (2022 to 2024) under greenhouse and field conditions. Spectral measurements of numerous samples from different scales were explored to observe spectral responses. Two forms of WPMI are proposed, WPMI G = (R760 − R554)/(R661 + R554) and WPMI R = (R760 − R661)/(R661 + R554), based on single-band classification and optimized combinations at the leaf scale. Their superiority in quantifying disease index and presenting differences between healthy and infected samples was evaluated from ground to UAV scales. The combination of superior WPMI G obtained from UAV imagery and hot-spot analysis reveals the spatiotemporal dynamics of WPM infections in smallholder farms. Results showed that the proposed WPMI achieved the highest overall classification accuracy at the leaf scale under the greenhouse (2022: 85% and 86% for WPMI G and WPMI R , respectively) and field conditions (2023: 81% and 80% for WPMI G and WPMI R , respectively; 2024: 80% and 81% for WPMI G and WPMI R , respectively). Compared to traditional vegetation indices (e.g., vegetation indices for monitoring pigment, biomass, or plant stress), the superior WPMI G improved the coefficient of determination ( R 2 ) for quantifying the disease index from ground to UAV scales (ground scale: incremental R 2 values of 0.06 to 0.92, 0.26 to 0.55, and 0.34 to 0.67 for 2022 to 2024; UAV scale: incremental R 2 values of 0.03 to 0.88, 0.24 to 0.51, and 0.11 to 0.65). The proposed WPMI demonstrated consistent and more marked differences at various scales. Using time-series images of WPMI G through hot-spot analysis revealed potential areas of infection or recovery in the field over 3 years and tracked the spatial and temporal dynamics of WPM infection. This work provides a promising strategy for detecting WPM occurrence on smallholder farms and identifying in-field hot spots of the disease spread based on UAV imagery.