Dhinakaran et al. (2026) Forecasting Soil Moisture Dynamics from SMAP Observations via Signal Decomposition

Identification

• Journal: Earth Systems and Environment
• Year: 2026
• Date: 2026-04-10
• Authors: D. Dhinakaran, V. Vijayalakshmi, S. Palpandi, R. Rajesh, D. Selvaraj, A. Rehash Rushmi Pavitra, Ahmad M. Salah
• DOI: 10.1007/s41748-026-01171-x

Research Groups

• Department of Computer Science and Engineering, Vel Tech Rangarajan Dr. Sagunthala R&D Institute of Science and Technology, Chennai, India
• Department of Data Science and Business Systems, SRM Institute of Science and Technology, Kattankulathur, Chennai, India
• Department of Electronics and Communication Engineering, Panimalar Engineering College, Chennai, India
• Edreesy Geospatial Solutions, LLC, Draper, Utah, USA

Short Summary

This paper proposes a Decomposition-Guided Forecasting Framework of Soil Moisture (DGF-SM) that integrates SMAP satellite observations with Seasonal Trend Decomposition by Loess (STL) and ARIMA-based forecasting, demonstrating high predictive accuracy and improved interpretability across diverse South Asian climatic regimes.

Objective

• To develop and evaluate a Decomposition-Guided Forecasting Framework of Soil Moisture (DGF-SM) that integrates satellite-based measurements with time-series decomposition and statistical predictions to improve soil moisture forecasting accuracy and interpretability across heterogeneous climatic regimes.

Study Configuration

• Spatial Scale: Country-level aggregated soil moisture for India, Bangladesh, Nepal, and Pakistan.
• Temporal Scale: Monthly soil moisture data from January 2016 to December 2021 (6 years).

Methodology and Data

• Models used: Seasonal Trend Decomposition by Loess (STL), Autoregressive Integrated Moving Average (ARIMA) model (specifically ARIMA(4,0,3) for the trend component), Exponential Smoothing (ETS) for comparison, and a univariate recurrent neural model as a learning-based benchmark.
• Data sources: NASA-USDA Enhanced SMAP Global Soil Moisture product (Level-3, Version 6) from the NASA National Snow and Ice Data Center, processed using Google Earth Engine (GEE).

Main Results

• The DGF-SM framework achieved a high predictive accuracy, reaching 96.9% explained variance in India, with a Mean Absolute Error (MAE) of 0.80 m³/m³ and Root Mean Squared Error (RMSE) of 0.95 m³/m³.
• The decomposition-guided strategy improved prediction stability, resulting in a 36.5%–70% reduction in MAE and up to 91% reduction in Mean Squared Error (MSE) compared to conventional direct forecasting approaches.
• The framework demonstrated consistent performance across heterogeneous climatic regimes, with lower forecasting errors in monsoon-dominated regions (India, Bangladesh) and slightly higher but stable errors in arid environments (Pakistan).
• Rolling-horizon validation confirmed robust short-term forecasting behavior, with gradual and controlled error growth from 1-month to 6-month prediction horizons, indicating reliable medium-term forecasting capability.
• STL decomposition confirmed additive seasonal behavior in all regions and revealed substantial regional differences in temporal dynamics, with India and Bangladesh showing strong seasonal and smooth long-term patterns, while Nepal and Pakistan exhibited weaker seasonality and less pronounced trends.

Contributions

• Development of a decomposition-guided forecasting framework tailored for satellite-derived soil moisture time series.
• Spatiotemporal analysis of soil moisture dynamics of SMAP over four South Asian nations.
• Statistical assessment of forecasting performance by both rolling-horizon and the static strategy.
• Demonstration of improved robustness and interpretability compared to conventional direct forecasting approaches.

Funding

• The authors received no specific funding for this study.

Citation

@article{Dhinakaran2026Forecasting,
  author = {Dhinakaran, D. and Vijayalakshmi, V. and Palpandi, S. and Rajesh, R. and Selvaraj, D. and Pavitra, A. Rehash Rushmi and Salah, Ahmad M.},
  title = {Forecasting Soil Moisture Dynamics from SMAP Observations via Signal Decomposition},
  journal = {Earth Systems and Environment},
  year = {2026},
  doi = {10.1007/s41748-026-01171-x},
  url = {https://doi.org/10.1007/s41748-026-01171-x}
}