Fine‐Scale Spatial Variation of Evapotranspiration Across Tropical Land‐Cover Mosaics Revealed by
UAV
Thermography
Medha Bulusu, Florian J. Ellsäßer, Hendrayanto, Dirk Hölscher, Alexander Röll ABSTRACT
A key flux in the hydrological cycle, evapotranspiration (ET) recycles terrestrial precipitation to the atmosphere. Its energetic equivalent, latent heat flux, is central to the surface energy balance. Evaporative fraction (EF) therein quantifies the proportion of available surface energy dissipated as latent rather than sensible heat. Previous studies point to high mean ET from tropical old‐growth forests and mature oil palm plantations, but fine‐scale assessments of ET and EF across heterogeneous land‐cover mosaics are lacking. Here, uncrewed aerial vehicle (UAV) thermography coupled with energy balance modelling is particularly suitable, covering sizable areas at high resolution (~10 cm). We applied it across landscapes with secondary forest, oil palm, and shrubland, and used a one‐source energy balance approach to model ET and EF. The objective was to assess variability among and within land‐cover classes, including fine‐scale correlations and clustering. UAV‐derived ET was compared to sap flux‐derived tree transpiration estimates in 20‐year‐old secondary forest, showing significant linear relationships ( R 2 = 0.56–0.66, p < 0.001). This complements previous method validations and points to UAV thermography as a reliable tool for fine‐scale ET mapping. While mean ET varied with shortwave radiation across flights, EF remained more stable and was used in subsequent analyses. Mean EF was highest in secondary forest (0.82–0.87), intermediate in logged forest and shrubland (0.80–0.82), and lowest in young oil palm (0.78). EF variability was high in shrubland, logged forest, and young oil palm, and much lower in secondary forests. Semivariograms indicated land‐cover‐dependent correlation ranges between 7 m (young oil palm) and 43 m (secondary forest). Global Moran's I indicated moderate spatial clustering, while local indicators of spatial association revealed strong patterns, with high‐value clusters dominating secondary forest and low‐value clusters in oil palm and logged forest. Our findings thus indicate low and variable EF in systems with limited canopy cover, that is, (young) oil palm and shrubland, and high EF at high spatial continuity in structurally more complex and closed secondary forest, thus adding spatially coherent evaporative cooling and microclimate regulation as further arguments for preserving and promoting secondary forests in tropical mosaic landscapes, aside from well‐documented benefits for biodiversity and other ecosystem functions.