Using Satellite Land Surface Temperature to Parameterize Sub-grid Tiling Schemes and Enable Tile-level Calibration of Land Surface Model

Abstract

To better simulate terrestrial energy and water processes, great efforts have been made to improve the representation of spatial heterogeneity in land surface models. HydroBlocks, a field-scale resolving land surface model, was developed to address this challenge while minimizing increases in computational demands by employs a new tiling framework – the hierarchical multivariate clustering (HMC) approach. The HMC is used to cluster grid cells with similar surface characteristics into hydrologic response units (HRUs, i.e., sub-grid tiles) using high-resolution data. HRUs model land surface processes separately and HRUs are connected through subsurface flow. Hence, the HRU configuration plays a critical role in the model performance. Furthermore, since the outputs for each HRU can be mapped in space, it is possible to explicitly calibrate the model at the tile level based on satellite-derived data. To investigate the advantages that satellite observations can have to tune HydroBlocks, the model was set up to run from 2013 to 2019 at the 30-meter spatial resolution and hourly temporal resolution over the Atmospheric Radiation Measurement (ARM) Southern Great Plains (SGP), U.S. This study demonstrated that HydroBlocks benefits from considering the temporal mean and standard deviation map of Landsat LST when generating HRUs. The effects of inputting the albedo, emissivity, and leaf area index derived directly from MODIS into the model were also explored. Finally, six soil and vegetation parameters in HydroBlocks were calibrated by maximizing the linear correlation coefficient of the time-series of simulated LST and observed MODIS LST at each HRU. The regional-level and site-level evaluation indicate the effectiveness of the calibration approach.