Progressive Scale Convolutional Network for Spatio-Temporal Downscaling of Soil Moisture: A Case Study Over the Tibetan Plateau

TL;DR

This study tackles high-resolution spatio-temporal downscaling of soil moisture over the Tibetan Plateau by fusing coarse SMAP data with validated high-temporal ERA5-Land variables. It introduces PSCNet, a separable spatio-temporal CNN featuring a multi-frequency temporal fusion (MFTF) module and a bespoke SE block to preserve fine-scale spatial details while modeling complex temporal dynamics. The approach achieves 10-km spatial and 3-hour temporal resolution SM products for 2016–2018, performing strongly against SMAP, in-situ networks, and across temporal generalization tests, with robust spatial coherence and detail preservation. These results demonstrate PSCNet’s potential for reliable, high-resolution SM downscaling in data-sparse, topographically complex regions, enabling improved hydrological and climatic analyses.

Abstract

Soil moisture (SM) plays a critical role in hydrological and meteorological processes. High-resolution SM can be obtained by combining coarse passive microwave data with fine-scale auxiliary variables. However, the inversion of SM at the temporal scale is hindered by the incompleteness of surface auxiliary factors. To address this issue, first, we introduce validated high temporal resolution ERA5-Land variables into the downscaling process of the low-resolution SMAP SM product. Subsequently, we design a progressive scale convolutional network (PSCNet), at the core of which are two innovative components: a multi-frequency temporal fusion module (MFTF) for capturing temporal dynamics, and a bespoke squeeze-and-excitation (SE) block designed to preserve fine-grained spatial details. Using this approach, we obtained seamless SM products for the Tibetan Plateau (TP) from 2016 to 2018 at 10-km spatial and 3-hour temporal resolution. The experimental results on the TP demonstrated the following: 1) In the satellite product validation, the PSCNet exhibited comparable accuracy and lower error, with a mean R value of 0.881, outperforming other methods. 2) In the in-situ site validation, PSCNet consistently ranked among the top three models for the R metric across all sites, while also showing superior performance in overall error reduction. 3) In the temporal generalization validation, the feasibility of using high-temporal resolution ERA5-Land variables for downscaling was confirmed, as all methods maintained an average relative error within 6\% for the R metric and 2\% for the ubRMSE metric. 4) In the temporal dynamics and visualization validation, PSCNet demonstrated excellent temporal sensitivity and vivid spatial details. Overall, PSCNet provides a promising solution for spatio-temporal downscaling by effectively modeling the intricate spatio-temporal relationships in SM data.

Figures (11)

• Figure 1: Elevation distribution and observation networks over the TP. (a) Overview showing elevation and four networks. (b-e) Site distribution with land cover types for Maqu, Ngari, Naqu, and CTP-SMTMN (abbreviated as CTP) networks, respectively. Symbol colors denote data missing rates. Background combines land cover types (30% transparency) with DEM topography. Legend describes FAO land cover classification system 1 (LCCS1) types within each network.
• Figure 2: Overall framework and methodological workflow of SM spatial downscaling.
• Figure 3: The architecture of the proposed PSCNet. The main figure (a) illustrates the overall structure, while (b-f) provide detailed views of its key components: the MFTF Module, the Stage block, the SE block, the feed-forward network (FFN), and the SE Convolutional layer.
• Figure 4: Scatter density plots showing R between downscaled products from ten downscaling methods and original SMAP microwave satellite observations. Overall statistical metrics are displayed in the lower right corner of each subplot. (a) PSCNet; (b) RF; (c) LSTM; (d) ResNet; (e) UNet; (f) ViT.
• Figure 5: Spatial distribution maps of validation metrics (R, bias, and ubRMSE from left to right) for downscaled products from different models, with histograms showing pixel distribution within each metric range.