Drone maps spotted crop water stress before other plant measures changed

What the drones detected

The researchers used the Priestley–Taylor Two-Source Energy Balance model with drone observations to estimate field-scale evapotranspiration. The model used thermal infrared data for surface temperature, multispectral data for green plant area, and laser-based data for plant structure.

Evapotranspiration estimates agreed well with measurements from an eddy-covariance system, with a root mean square error of 0.14 millimeters per hour and an R² of 0.49. Agreement improved to an R² of 0.81 when senescence periods were excluded. Thermal infrared maps revealed crop water-stress signals before multispectral or laser-based plant measures changed.

The estimates were highly sensitive to surface temperature, making accurate thermal calibration important. Green plant area and total plant area gave similar inputs during peak growth, but green plant area more effectively represented declining canopy function during stress and senescence, especially in winter wheat, whose dense structure could create cooling unrelated to transpiration.

Why early stress matters

Mapping evapotranspiration at fine spatial resolution can show how water use varies within a field. Earlier detection of crop water stress could help identify areas that need closer attention while supporting localized water-use monitoring, although the study did not test whether acting on these maps improved irrigation or yields.

Evidence and limits

This was a field study using drone observations collected across three crop seasons in one 10-hectare field. Researchers checked the model against eddy-covariance measurements, which estimate exchanges of water and energy between the field and the atmosphere. The results depended strongly on accurate thermal calibration, and estimates were less consistent when senescing vegetation was included. The abstract does not establish how well the approach would perform in other fields, climates or crop varieties.