Mass and energy flux estimates at different spatial resolutions in a heterogeneous area through a distributed energy–water balance model and remote-sensing data

Abstract

A physically based, distributed energy–water balance model, Flash-Flood Event-Based Spatially Distributed Rainfall–Runoff Transformation-Energy Water Balance (FEST-EWB) model, and remote-sensing data were analysed to study the representativeness of mass and energy fluxes at different spatial resolutions. The analyses were performed in an agricultural area of Barrax (Spain) in the framework of the Sentinel-2 and Fluorescence Experiment (SEN2FLEX). In particular, there were two main objectives: (1) to evaluate the ability of the distributed hydrological model to compute energy and mass fluxes for a heterogeneous area compared to remote-sensing and ground data and (2) to define the length scales of different processes (evapotranspiration (ET) and land surface temperature (LST)) above which the variance of the different variables becomes insignificant for the process, so that bare soil and vegetation behaviours are no longer distinguishable. Mass and energy fluxes were collected from ground data and from an Airborne Hyperspectral Scanner (AHS), with a spatial resolution between 2 and 4 m, and from the Moderate Resolution Imaging Spectroradiometer (MODIS), with a spatial resolution of 1000 m, and then compared with hydrological model outputs. ET and LST spatial variability was analysed at different spatial resolutions using histograms, statistical parameters, and spatial autocorrelation functions. Computed ET with the FEST-EWB model at high spatial resolution (10 m) showed for the three days of analysis a mean relative error of 9.4% compared to AHS data, whereas for land surface temperature comparison a relative error of 1.6% was found. Then, LSTs from AHS and FEST-EWB were aggregated at decreasing spatial resolutions (50, 150, 300, 400, 500, 600, 750, and 1000 m), showing that the thermodynamic variability tends to disappear with a lower number of classes in the histograms and with a decrease of the coefficient of variation (CV) and of standard deviation values. At each scale, a similar behaviour was reported between each pair of images, with the values of standard deviation starting, respectively, from 8.9°C and 9.6°C at 10 m of spatial resolution to 7.7°C and 7.9°C at 1000 m of spatial resolution. Similar results were obtained from the spatial aggregation of ET images. The decrease in standard deviation values of LST and ET with a decrease in the scale became substantial around pixel dimensions equal to 400 m.