Empirical Upscaling of Point-scale Soil Moisture Measurements for Spatial Evaluation of Model Simulations and Satellite Retrievals

TL;DR

This work tackles the scale mismatch between point-scale soil moisture measurements and coarse model/satellite grids by upscaling point data to 100 m over a 100 km × 100 km agricultural area. It combines spatiotemporal fusion (ESTARFM/ubESTARFM) to generate daily 100 m predictors with an XGBoost-based ML upscaling using 28 OzNet sites as ground truth. Cross-validation shows strong and transferable performance (R ≈ 0.6–0.9 in four-fold CV and 0.6–0.8 in cross-cluster CV) for mapping SM at 100 m, supporting fairer evaluations of model simulations and satellite retrievals. The approach provides a practical path to align in-situ measurements with footprint scales of climatic models and remote sensing products, with future work focusing on independent-data validation.

Abstract

The evaluation of modelled or satellite-derived soil moisture (SM) estimates is usually dependent on comparisons against in-situ SM measurements. However, the inherent mismatch in spatial support (i.e., scale) necessitates a cautious interpretation of point-to-pixel comparisons. The upscaling of the in-situ measurements to a commensurate resolution to that of the modelled or retrieved SM will lead to a fairer comparison and statistically more defensible evaluation. In this study, we presented an upscaling approach that combines spatiotemporal fusion with machine learning to extrapolate point-scale SM measurements from 28 in-situ sites to a 100 m resolution for an agricultural area of 100 km by 100 km. We conducted a four-fold cross-validation, which consistently demonstrated comparable correlation performance across folds, ranging from 0.6 to 0.9. The proposed approach was further validated based on a cross-cluster strategy by using two spatial subsets within the study area, denoted as cluster A and B, each of which equally comprised of 12 in-situ sites. The cross-cluster validation underscored the capability of the upscaling approach to map the spatial variability of SM within areas that were not covered by in-situ sites, with correlation performance ranging between 0.6 and 0.8. In general, our proposed upscaling approach offers an avenue to extrapolate point measurements of SM to a spatial scale more akin to climatic model grids or remotely sensed observations. Future investigations should delve into a further evaluation of the upscaling approach using independent data, such as model simulations, satellite retrievals or field campaign data.

Figures (7)

• Figure 1: The location (highlighted in red rectangle within the NSW State) and the land cover information in 2020 of Yanco agricultural region. The purple circle, red triangle and black dots represent the locations of in-situ sties from CosmOz, OzFlux and OzNet, respectively. The cluster A spans the region of 146.06-146.16 °E and 34.62-34.77 °S (10 × 15 km2) and the cluster B spans the region of 146.25-146.35 °E and 34.92-35.02 °S (10 × 10 km2).
• Figure 2: The experimental design herein.
• Figure 3: Density scatterplots of (a) albedo, (b) NDVI and (c) LST between 01/Jan/2016 and 31/Dec/2019 (excluding training dates of spatiotemporal fusion). The density of scatters was calculated using a binning approach.
• Figure 4: Spatial comparison between MODIS and downscaled predictors on 02/Apr/2017. (a-c) are MODIS albedo, NDVI and LST for the study area, respectively; (d-f) are zoomed windows of MODIS predictors spanning 146.25 to 146.35 °E and -34.95 to -35.05 °N; (g-i) are the downscaled albedo, NDVI and LST for the study area, respectively; and (j-l) are zoomed windows of downscaled predictors covering the same area with (d-f).
• Figure 5: Normalised Taylor diagrams of upscaled SM using XGBoost during 2016-2019 for (a) four-fold cross-validation and (b) cross-cluster validation.