Soil moisture retrospective estimation and interpretability analysis integrating meteorological drivers and environmental factor representations

Retrospective soil moisture estimation is essential for understanding land-surface hydrological processes, vegetation water regulation, and drought-related environmental responses. However, in-situ soil moisture observations are often discrete and irregular, while their temporal evolution is jointly affected by meteorological forcing and soil-depth conditions. To address this problem, this study proposes a Mamba-based retrospective soil moisture estimation framework that integrates Meteorological-Driven and Environmental-Factor Collaborative Gating (MDECG) with Multi-Scale Feature Rearrangement and Progressive Fusion (MFRPFM). The MDECG module jointly incorporates meteorological drivers and soil-depth information through a data-driven collaborative gating mechanism to adaptively enhance informative external forcing, while the MFRPFM module captures short-term fluctuations and long-term trends through multi-scale temporal representation. Experiments were conducted on the 2009-2016 soil moisture dataset of a north subtropical evergreen and deciduous broad-leaved mixed forest, with estimation targets at 10, 20, 30, 40, and 50 cm soil depths. For each target date, the model uses a 30-day window of ERA5-Land meteorological variables, including target-day records, together with soil-depth information. The estimation is performed after the complete target-day reanalysis inputs become available, representing retrospective same-day estimation rather than advance forecasting. The proposed method achieved the best average performance across the five soil depths, with an MAE of 0.00916 ± 0.00032, MSE of 0.000133 ± 0.000008, RMSE of 0.01147 ± 0.00037, and R2 of 0.94615 ± 0.00342, reported as mean ± SD over five independent random seeds, outperforming LSTM, BiLSTM, Transformer, Duet, THATSN, PPGF, SimpleTM, and TimeKAN. These results indicate that the proposed framework can improve estimation accuracy and temporal stability under discrete soil moisture observation conditions, providing a useful data-driven tool for point-scale retrospective soil moisture estimation and environmental monitoring.

Authors

Institutions

Publication Details

Journal
PLoS ONE
Published
2026-11-20
DOI
https://doi.org/10.1371/journal.pone.0353409
Primary Topic
Soil Moisture and Remote Sensing
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Soil moisture retrospective estimation and interpretability analysis integrating meteorological drivers and environmental factor representations

Jian Hua, Weiya Ge, Xiang He, Wei Xu et al.
6 links
PLoS ONE
Soil Moisture and Remote Sensing
article

Soil moisture retrospective estimation and interpretability analysis integrating meteorological drivers and environmental factor representations

Jian Hua, Weiya Ge, Xiang He, Wei Xu, Aiju Li, Kai Yang
article en

Abstract

Retrospective soil moisture estimation is essential for understanding land-surface hydrological processes, vegetation water regulation, and drought-related environmental responses. However, in-situ soil moisture observations are often discrete and irregular, while their temporal evolution is jointly affected by meteorological forcing and soil-depth conditions. To address this problem, this study proposes a Mamba-based retrospective soil moisture estimation framework that integrates Meteorological-Driven and Environmental-Factor Collaborative Gating (MDECG) with Multi-Scale Feature Rearrangement and Progressive Fusion (MFRPFM). The MDECG module jointly incorporates meteorological drivers and soil-depth information through a data-driven collaborative gating mechanism to adaptively enhance informative external forcing, while the MFRPFM module captures short-term fluctuations and long-term trends through multi-scale temporal representation. Experiments were conducted on the 2009-2016 soil moisture dataset of a north subtropical evergreen and deciduous broad-leaved mixed forest, with estimation targets at 10, 20, 30, 40, and 50 cm soil depths. For each target date, the model uses a 30-day window of ERA5-Land meteorological variables, including target-day records, together with soil-depth information. The estimation is performed after the complete target-day reanalysis inputs become available, representing retrospective same-day estimation rather than advance forecasting. The proposed method achieved the best average performance across the five soil depths, with an MAE of 0.00916 ± 0.00032, MSE of 0.000133 ± 0.000008, RMSE of 0.01147 ± 0.00037, and R2 of 0.94615 ± 0.00342, reported as mean ± SD over five independent random seeds, outperforming LSTM, BiLSTM, Transformer, Duet, THATSN, PPGF, SimpleTM, and TimeKAN. These results indicate that the proposed framework can improve estimation accuracy and temporal stability under discrete soil moisture observation conditions, providing a useful data-driven tool for point-scale retrospective soil moisture estimation and environmental monitoring.

PLoS ONEVol. 21(11)
China Geological Survey (CN)
Climate action, Life on land
Soil Moisture and Remote Sensing
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon