Modelling the diurnal cycle of evapotranspiration using remote sensing models – are we there yet?

The diurnal cycle of Evapotranspiration (ET) can provide important clues about the functioning of vegetation. However, the Remote Sensing (RS) based ET models, especially the ones that use Land Surface Temperature (LST) as input, often focus on estimating ET at daily and other longer time scales. Modelling the diurnal cycle of ET with RS-based ET models is not fully explored. This study aims (i) to estimate daytime ET as average of the multiple sub-daily ET values (i.e. diurnal ET) and compare this with the one obtained using the traditional temporal scaling approach. This it to evaluate if diurnal ET modelling improves the accuracy of daytime ET estimation and (ii) to estimate diurnal ET using LST with different spatial resolutions and assess if the improvement in the spatial resolution of LST improves the accuracy of the diurnal ET. The study was carried out over five sites on clear sky days and three ET models, Priestley-Taylor Jet Propulsion Laboratory (PT-JPL), Soil Plant Atmosphere and Remote Sensing Evapotranspiration (SPARSE) and Surface Temperature Initiated Closure (STIC). LST from geostationary satellites and Moderate Resolution Imaging Spectroradiometer (MODIS) sensor was used. Two spatial disaggregation approaches were attempted to get fine resolution LST from MODIS. To obtain the diurnal LST from MODIS observations, a semi-empirical Diurnal Temperature Cycle (DTC) model was used. The results indicated that the estimation of daytime ET from the diurnal ET did not improve the accuracy of daytime ET on a consistent basis and is model specific. Further, the accuracy of diurnal ET did not improve when using finer spatial resolution LST values with the ET models only able to capture the general diurnal pattern of ET. The results highlight the limitations of the ET models, and the need to better parameterize the diurnal variation in radiation, aerodynamic and surface characteristics.