Uncertainties of drone-based cropland evapotranspiration estimation

Drone surveys were conducted at a cropland at Kartal, Hungary in 2024 to estimate the evapotranspiration (ET) of the area. There is an eddy covariance tower in the cropland since 2017. Between 27 May and 8 August 9 campaigns were carried out with a DJI M300 drone equipped by a Micasense Altum (MA) multispectral and thermal camera. The leaf area index (LAI) was also measured at 7 points in the sunflower canopy supplemented light interception measurements to estimate the leaf angle distribution of the canopy. Canopy cover, surface temperature, and LAI maps were produced from the MA’s reflectance values and the LAI samples in the 7 points using partial least squares (PLSR) regression to serve as inputs of the pyTSEB model. The spatial average of the ET pixels from the footprint area of the corresponding eddy covariance flux were validated against the eddy covariance ET.

The first results of validation showed very weak relation between the measured and modelled data. The relationship improved considerably when the surface temperature maps taken by the MA were corrected according to the surface temperature measured from the eddy tower by an Apogee infrared radiometer.

Further improvement was reached when the LAI maps were modified based on the leaf angle distribution estimated from the light interception measurements.

While the correlation between the measured and modelled ET is statistically significant, the intercept of the regression is a considerable (~100 W m-2).