ISPRS-Archives ISPRS - International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences ISPRS-Archives Int. Arch. Photogramm. Remote Sens. Spatial Inf. Sci. 2194-9034 Copernicus Publications Göttingen, Germany 10.5194/isprs-archives-XLII-3-W6-3-2019 SENSITIVITY OF FOUR CONTEXTUAL REMOTE SENSING BASED SURFACE ENERGY BALANCE MODELS TO SPATIAL DOMAIN Bhattarai N. 1 https://orcid.org/0000-0003-2749-3549 Mallick K. 2 https://orcid.org/0000-0002-2735-930X Jain M. 1 School for Environment and Sustainability, University of Michigan, Ann Arbor, MI 48109, USA Remote Sensing and Natural Resources Modeling, Environmental Research and Innovation Department, Luxembourg Institute of Science and Technology (LIST), L4422, Belvaux, Luxembourg 26 07 2019 XLII-3/W6 3 7 Copyright: © 2019 N. Bhattarai et al. 2019 This work is licensed under the Creative Commons Attribution 4.0 International License. To view a copy of this licence, visit https://creativecommons.org/licenses/by/4.0/ This article is available from https://isprs-archives.copernicus.org/articles/isprs-archives-XLII-3-W6-3-2019.html The full text article is available as a PDF file from https://isprs-archives.copernicus.org/articles/isprs-archives-XLII-3-W6-3-2019.pdf

It remains unclear how the selection of a spatial domain affects the accuracy of evapotranspiration (ET) estimates from contextual remote sensing based surface energy balance (SEB) models, particularly at large spatial scales. We thus tested the effect of spatial domain on four widely implemented contextual remote sensing based SEB models: Surface Energy Balance Algorithm for Land (SEBAL), Mapping ET with Internalized Calibration (METRIC), Simplified Surface Energy Balance Index (S-SEBI), and Triangular ET models. We applied these models on 44 near cloud-free Moderate Resolution Imaging Spectroradiometer (MODIS) thermal images across all of India from 2004 to 2006. Four spatial domains were considered: all of India, agro-ecological regions, 300 km × 300 km grids, and 600 km × 600 km grids and we compared hourly ET estimates from the models against observed ET data at four Bowen Ratio sites in India. Model performance varied across all models and spatial domains. ET values of neighboring pixels across spatial domains formed sharp edges along the boundaries of agro-ecological regions, 300 km × 300 km grids, and 600 km × 600 km grids suggesting that all ET models are highly sensitive to the selection of spatial domain. No single spatial domain was found to be optimal for all models and hence potential uncertainties associated with the selection of spatial domain should be taken into consideration when implementing these models at regional scales. The results from this study provide guidance for future regional-scale implementation of ET models and potential approaches to overcome these challenges.