A Spatial Downscaling Framework for SMAP Soil Moisture Based on Stacking Strategy

Abstract

Soil moisture (SM) data can provide guidance for decision-makers in fields such as drought monitoring and irrigation management. Soil Moisture Active Passive (SMAP) satellite offers sufficient spatial resolution for global-scale applications, but its utility is limited in regional areas due to its lower spatial resolution. To address this issue, this study proposed a downscaling framework based on the Stacking strategy. The framework integrated extreme gradient boosting (XGBoost), light gradient boosting machine (LightGBM), and categorical boosting (CatBoost) to generate 1 km resolution SM data using 15 high-resolution factors derived from multi-source datasets. In particular, to test the influence of terrain partitioning on downscaling results, Anhui Province, which has diverse terrain features, was selected as the study area. The results indicated that the performance of the three base models varied, and the developed Stacking strategy maximized the potential of each model with encouraging downscaling results. Specifically, we found that: (1) The Stacking model achieved the highest accuracy in all regions, and the performance order of the base models was: XGBoost > CatBoost > LightGBM. (2) Compared with the measured SM at 87 sites, the downscaled SM outperformed other 1 km SM products as well as the downscaled SM without partitioning, with an average ubRMSE of 0.040 m3/m3. (3) The downscaled SM responded positively to rainfall events and mitigated the systematic bias of SMAP. It also preserved the spatial trend of the original SMAP, with higher levels in the humid region and relatively lower levels in the semi-humid region. Overall, this study provided a new strategy for soil moisture downscaling and revealed some interesting findings related to the effectiveness of the Stacking model and the impact of terrain partitioning on downscaling accuracy.

Watershed scale soil moisture estimation model using machine learning and remote sensing in a data-scarce context

“…Step 3: Calculate the Prediction Error of the downscaling approach Prior to applying the models for high-resolution soil moisture prediction, the model performance was evaluated within the spatial support of the θSMAP pixels (approximately 9 km). Given that the study area is relatively small compared to other studies (Bai et al, 2019;Rao et al, 2022;Xu et al, 2024), model generalization error was assessed through repeated 10-fold cross-validation with grid search, as implemented according to (Krstajic et al, 2014(Krstajic et al, ) (2014 3) in the mlr package (Schratz et al, 2021).…”

“…This makes them less suitable for agricultural, hydrological, and environmental applications requiring daily and high spatial detail information (Vergopolan et al, 2021). Several methods have been proposed to enhance the spatial resolution of remote soil moisture estimates through a process called "downscaling" (Abbaszadeh et al, 2019;Bai et al, 2019;Cui et al, 2019;Fang et al, 2019;Guevara & Vargas, 2019;Hernandez-Sanchez et al, 2020;Liu et al, 2020;Mao et al, 2019;Montzka et al, 2020;Peng et al, 2017;Shangguan et al, 2024;Sishah et al, 2023;Xu et al, 2024;Zhu et al, 2023). Recently, machine learning techniques such as random forest (Hengl et al, 2018) have achieved advancements in the downscaling of remote soil moisture estimates, either spatially (Bai et al, 2019;Chen et al, 2019;Zappa et al, 2019;Zhao et al, 2018) or temporally (Lu et al, 2015;Mao et al, 2019;Xing et al, 2017).…”