Enhancing Soil Moisture Prediction in Drought-Prone Agricultural Regions Using Remote Sensing and Machine Learning Approaches
The North China Plain is a crucial agricultural region in China, but irregular precipitation patterns have led to significant water shortages. To address this, analyzing the high-resolution dynamics of root-zone soil moisture transport is essential for optimizing irrigation strategies and improving...
Saved in:
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2025-01-01
|
| Series: | Remote Sensing |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2072-4292/17/2/181 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | The North China Plain is a crucial agricultural region in China, but irregular precipitation patterns have led to significant water shortages. To address this, analyzing the high-resolution dynamics of root-zone soil moisture transport is essential for optimizing irrigation strategies and improving water resource efficiency. The Richards equation is a robust model for describing soil moisture transport dynamics across multiple soil layers, yet its application at large spatial scales is hindered by its sensitivity to boundary conditions and model parameters. This study introduces a novel approach that, for the first time, employs a continuous time series of near-surface soil moisture as the upper boundary condition in the Richards equation to estimate high-resolution root-zone soil moisture in the North China Plain, thus enabling its large-scale application. Singular spectrum analysis (SSA) was first applied to reconstruct site-specific time series, filling in missing and singular values. Leveraging observational data from 617 monitoring sites across the North China Plain and multiple spatial covariates, we developed a machine learning model to estimate near-surface soil moisture at a 1 km resolution. This high-resolution, continuous near-surface soil moisture series then served as the upper boundary condition for the Richards equation, facilitating the estimation of root-zone soil moisture across the region. The results indicated that the machine learning model achieved a correlation coefficient (R) of 0.92 for estimating spatial near-surface soil moisture. Analysis of spatial covariates showed that atmospheric forcing factors, particularly temperature and evaporation, had the most substantial impact on model performance, followed by static factors such as latitude, longitude, and soil texture. With a continuous time series of near-surface soil moisture, the Richards equation method accurately predicted multi-layer soil moisture and demonstrated its applicability for large-scale spatial use. The model yielded R values of 0.97, 0.78, 0.618, and 0.43, with RMSEs of 0.024, 0.06, 0.08, and 0.11, respectively, for soil layers at depths of 10 cm, 20 cm, 40 cm, and 100 cm across the North China Plain. |
|---|---|
| ISSN: | 2072-4292 |