Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Field-scale soil moisture estimated from Sentinel-1 SAR data using a knowledge-guided deep learning approach

Yi Yu, Patrick Filippi, Thomas F. A. Bishop

TL;DR

The paper tackles the problem of field-scale soil moisture estimation from Sentinel-1 SAR, where traditional semi-physical models like the Water Cloud Model (WCM) often fail to generalize across diverse landscapes. It introduces a knowledge-guided deep learning approach that embeds WCM physics into a long short-term memory (LSTM) network via a dual-component loss to enforce physical consistency, using VH SAR, Landsat-derived vegetation metrics, and auxiliary surface data. Across a Yanco, Australia study with four-fold cross-validation against OzNet 0–5 cm SM, WCM alone yields RMSE $0.08$–$0.10$ and $R$ $0.26$–$0.34$, while the knowledge-guided LSTM achieves RMSE $0.06$–$0.08$ and $R$ up to $0.64$, with a stable vegetation parameter around $A\approx 0.02$. This hybrid framework improves accuracy and physical plausibility for field-scale SM monitoring in heterogeneous landscapes, suggesting potential for broader deployment and cross-site generalization in operational SM monitoring.

Abstract

Soil moisture (SM) estimation from active microwave data remains challenging due to the complex interactions between radar backscatter and surface characteristics. While the water cloud model (WCM) provides a semi-physical approach for understanding these interactions, its empirical component often limits performance across diverse agricultural landscapes. This research presents preliminary efforts for developing a knowledge-guided deep learning approach, which integrates WCM principles into a long short-term memory (LSTM) model, to estimate field SM using Sentinel-1 Synthetic Aperture Radar (SAR) data. Our proposed approach leverages LSTM's capacity to capture spatiotemporal dependencies while maintaining physical consistency through a modified dual-component loss function, including a WCM-based semi-physical component and a boundary condition regularisation. The proposed approach is built upon the soil backscatter coefficients isolated from the total backscatter, together with Landsat-resolution vegetation information and surface characteristics. A four-fold spatial cross-validation was performed against in-situ SM data to assess the model performance. Results showed the proposed approach reduced SM retrieval uncertainties by 0.02 m$^3$/m$^3$ and achieved correlation coefficients (R) of up to 0.64 in areas with varying vegetation cover and surface conditions, demonstrating the potential to address the over-simplification in WCM.

Field-scale soil moisture estimated from Sentinel-1 SAR data using a knowledge-guided deep learning approach

TL;DR

The paper tackles the problem of field-scale soil moisture estimation from Sentinel-1 SAR, where traditional semi-physical models like the Water Cloud Model (WCM) often fail to generalize across diverse landscapes. It introduces a knowledge-guided deep learning approach that embeds WCM physics into a long short-term memory (LSTM) network via a dual-component loss to enforce physical consistency, using VH SAR, Landsat-derived vegetation metrics, and auxiliary surface data. Across a Yanco, Australia study with four-fold cross-validation against OzNet 0–5 cm SM, WCM alone yields RMSE – and –, while the knowledge-guided LSTM achieves RMSE – and up to , with a stable vegetation parameter around . This hybrid framework improves accuracy and physical plausibility for field-scale SM monitoring in heterogeneous landscapes, suggesting potential for broader deployment and cross-site generalization in operational SM monitoring.

Abstract

Soil moisture (SM) estimation from active microwave data remains challenging due to the complex interactions between radar backscatter and surface characteristics. While the water cloud model (WCM) provides a semi-physical approach for understanding these interactions, its empirical component often limits performance across diverse agricultural landscapes. This research presents preliminary efforts for developing a knowledge-guided deep learning approach, which integrates WCM principles into a long short-term memory (LSTM) model, to estimate field SM using Sentinel-1 Synthetic Aperture Radar (SAR) data. Our proposed approach leverages LSTM's capacity to capture spatiotemporal dependencies while maintaining physical consistency through a modified dual-component loss function, including a WCM-based semi-physical component and a boundary condition regularisation. The proposed approach is built upon the soil backscatter coefficients isolated from the total backscatter, together with Landsat-resolution vegetation information and surface characteristics. A four-fold spatial cross-validation was performed against in-situ SM data to assess the model performance. Results showed the proposed approach reduced SM retrieval uncertainties by 0.02 m/m and achieved correlation coefficients (R) of up to 0.64 in areas with varying vegetation cover and surface conditions, demonstrating the potential to address the over-simplification in WCM.
Paper Structure (11 sections, 13 equations, 3 figures)

This paper contains 11 sections, 13 equations, 3 figures.

Table of Contents

  1. Introduction
  2. Study area and data
  3. Study area and in-situ data
  4. Sentinel-1 SAR data
  5. Landsat-resolution surface reflectance
  6. Auxiliary information
  7. Methods
  8. Water cloud model
  9. Knowledge-guided long short-term memory
  10. Results and discussion
  11. Conclusion

Figures (3)

  • Figure 1: (a) The spatial distribution of Sentinel-1 backscatter coefficient on 01 Feb 2019; (b) the synthesised NDVI derived from the fusion of MODIS and Landsat data on 01 Feb 2019; and (c) the land cover classification based on the ESA WorldCover 10 m v200 dataset zanaga_esa_2022.
  • Figure 2: Scatterplots of WCM-predicted SM against in-situ SM in across the four-fold cross-validation. The unit is m$^3$/m$^3$. The red dashed line represents the 1:1 line.
  • Figure 3: Scatterplots of SM predicted by the WCM knowledge-guided LSTM against in-situ SM across the four-fold cross-validation. The unit is m$^3$/m$^3$. The red dashed line represents the 1:1 line.