High-resolution soil moisture: downscaling of physics-based modeling estimates
Abstract
Accurate distributed estimates of soil moisture are important for several applications in the field of natural hazards, water resources management, and agriculture. They are also key for other scientific advancements, for instance in the understanding of land surface-atmosphere exchanges. For all these, high-resolution is beneficial, if not essential, as it captures small-scale differences.
In this work, we start by demonstrating the importance of spatial resolution for landslide prediction. We first compare the estimates provided by two operational hydrological forecasting systems in Switzerland: a physics-based pan-European modeling framework (TerrSysMP) and a Swiss conceptual hydrological model. We find the latter to be more informative and reconduct the cause into the spatial resolution (12.5x12.5 km2 for the former, 250x250 m2 for the latter). We study this further by comparing tilted-v domains with different sub-grid heterogeneities with the physics-based hydrological model ParFlow-CLM. Results show that, even within 400 x 400 m2, averaging out heterogeneity can impact soil moisture and consequently lead to underestimation of local instabilities. These results confirm the importance of the accuracy and resolution of soil moisture estimates. With the objective of compensating for the lower resolution required for a physics-based model to be run over large domains without prohibitive computational expenses, we explore different methodologies for the efficient downscaling of a coarser resolution soil moisture estimate. We set up the ParFlow-CLM at different resolutions over a catchment in the Upper Colorado River Basin. This allows us a) to have a direct comparison for our downscaled product (rather than just comparing to sparse in-situ measurements), and b) to study how coarse the lower resolution estimates to be downscaled can be without too much information loss. For the downscaling, we explore a wide range of methods, from simple topography-driven techniques to more complex ones, accounting for other covariates, e.g., in soil/climate properties. The key to this exploration is not only to get the most accurate estimates of soil moisture but also to keep the computational expenses low.- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2022
- Bibcode:
- 2022AGUFM.H42G1383L