Multivariate calibration of an ecohydrological model using spatial patterns of remote sensing-derived land surface temperature

Authors: DUETHMANN, D. - Leibniz Institute; Anderson, Martha; MANETA, M. - University Of Montana; TETZLAFF, D. - Leibniz Institute

Submitted to: Hydrological Processes
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 10/25/2023
Publication Date: 11/14/2023
Citation: Duethmann, D., Anderson, M.C., Maneta, M., Tetzlaff, D. 2023. Multivariate calibration of an ecohydrological model using spatial patterns of remote sensing-derived land surface temperature. Hydrological Processes. 628. Article e130433. https://doi.org/10.1016/j.jhydrol.2023.130433.
DOI: https://doi.org/10.1016/j.jhydrol.2023.130433

Interpretive Summary: The EcH2O model is a spatially distributed modeling system describing water and energy flows through ecosystems and their relationships with vegetation, carbon, and soil moisture dynamics at catchment scales. The model has been widely used to study impacts of climate and land-use on water quality and availability toward building more resilient landscapes. While many experiments have focused on calibrating EcH20 to better represent observed streamflow, less emphasis has been placed on calibrations aimed at accurately capturing heat and evaporation fluxes across landscape, although these fluxes have large impacts on habitat, vegetation growth, and quantification of water use. This study uses both observed streamflow and satellite measurements of land-surface temperature derived from Landsat thermal band imagery at 100-m spatial resolution to improve model parameterization. The study, conducted over a mixed forest-grassland-agricultural landscape in northeast Germany, found that both types of measurements can be effectively used together to constrain model parameters to obtain better representations of water movement and evaporative losses across the catchment. The use of satellite data in the calibration process will be of benefit for large area modeling in regions with limited ground-based measurements.

Technical Abstract: Distributed hydrologic models aim at estimating hydrologic variables for each spatial model unit but their calibration at the basin scale is usually inefficient to ensure an adequate simulation of spatial patterns. Considering remote sensing-derived observations for model calibration in addition to streamflow is a suitable strategy to better constrain model parameters, improve process-consistency and the representation of spatial patterns of hydrologic models. In this regard, land surface temperature (Ts) is an interesting variable as it is at the core of the surface energy and water balance and closely related to evapotranspiration. This study therefore aims at assessing the benefits of integrating spatial patterns of satellite-derived Ts into calibration of the process-based ecohydrologic model EcH2O. We furthermore explore the value of an increasing number of Ts images in the calibration period. The study is performed in a mixed land cover catchment in the lowlands of NE Germany and makes use of Landsat-derived Ts data. Our results demonstrate the value of satellite-derived Ts data for reducing uncertainties of energy-balance related vegetation parameters, which are hardly constrained when the model is calibrated to streamflow only. The trade-off between good simulations for streamflow and Ts was only small, but good model performance with respect to streamflow does not preclude low performance regarding spatial patterns of Ts. Improvements in simulated Ts could already be achieved by including only very few images of satellite-derived Ts in model calibration. Spatial patterns in observed Ts are shown to be strongly related to land cover class and vegetation dynamics, and our results indicate that further model improvements in spatial Ts patterns may be possible by better representing observed variations of leaf area index within the ecohydrologic model.