The assessment and management of soil moisture (SM) in agricultural landscapes have become increasingly critical for ensuring global food security. However, traditional methods often lack accuracy in large-scale predictions due to limited field measurements and the absence of effective remote sensing integration. To address this gap, we propose pedotransfer functions (PTFs) for predicting SM at field capacity (SMFC) based on readily measurable soil properties (RM-SPs) and spectral indices data. The study was conducted in agricultural lands across three Iranian provinces: Damavand in Tehran, Bukan in West-Azerbaijan, and Baneh in Kurdistan. A total of 137 surface soil samples (0–10 cm) were collected, and RM-SPs were measured. Concurrently, Landsat 8 and Sentinel-2 satellite imagery was fused to generate soil (SSIs), vegetation (VSIs), and moisture (MSIs) spectral indices. The SMFC was predicted using four PTF models: RM-SPs, RM-SPs + SSIs, RM-SPs + VSIs, and RM-SPs + MSIs. Results demonstrated that incorporating spectral indices significantly enhanced SMFC prediction accuracy. The RM-SPs + MSIs model achieved the highest performance (MSEp = 0.008, Cp = 3.8, and R²p = 0.86). Among the different models, the combination of RM-SPs and MSIs yielded the most accurate predictions, with the model demonstrating strong reliability across all tested sites. Furthermore, the study highlighted that the fusion of Landsat 8 and Sentinel-2 data proved to be highly effective in capturing spatial variability in soil moisture, particularly in areas with varying land uses and crop types. This study highlights the effectiveness of combining field-measured soil properties with remote sensing data for large-scale SM assessment, offering a promising tool for improved agricultural water management. Graphical Abstract This is a visual summary that serves as a pivotal entry point into the research, offering a concise overview of the study's core findings and methodologies. The graphical abstract illustrates a comprehensive framework for estimating soil moisture at field capacity (SMFC) using fused Landsat 8 and Sentinel-2 satellite data, combined with pedotransfer functions (PTFs). It depicts the process of collecting soil samples from three Iranian provinces and analyzing readily measurable soil properties (RM-SPs) such as bulk density, texture (sand, silt, clay), and organic carbon. The use of the van Genuchten retention model for SMFC estimation is highlighted, alongside the extraction and classification of spectral indices into SSIs, VSIs, and MSIs. These indices are visually integrated into the soil data using PTFs to enhance prediction accuracy. The logical flow showcases the model development and evaluation using Adjusted R², Mean Square Error (MSEp), and Mallows’ Cp, culminating in the selection of four optimal PTF models. Among them, the RM-SPs + MSIs model is visually emphasized as the most accurate. The RM-SPs + MSIs model demonstrated the highest accuracy in predicting SMFC, significantly outperforming the other models. The validation phase confirmed the model's reliability across all study areas, highlighting its potential for large-scale soil moisture assessment and drought management, conveying the study’s findings at a glance.

Enhanced Surface Soil Moisture Prediction Through Dual-Satellite Spectral Fusion

期刊论文
发表日期：

2025年4月28日

查看全文

Kamal Khosravi Aqdam,
Amin Nouri,
Naser Miran,
Seyedeh Ensieh Faramarzi,
Maryam Akhlaghi

摘要

The assessment and management of soil moisture (SM) in agricultural landscapes have become increasingly critical for ensuring global food security. However, traditional methods often lack accuracy in large-scale predictions due to limited field measurements and the absence of effective remote sensing integration. To address this gap, we propose pedotransfer functions (PTFs) for predicting SM at field capacity (SM_FC) based on readily measurable soil properties (RM-SPs) and spectral indices data. The study was conducted in agricultural lands across three Iranian provinces: Damavand in Tehran, Bukan in West-Azerbaijan, and Baneh in Kurdistan. A total of 137 surface soil samples (0–10 cm) were collected, and RM-SPs were measured. Concurrently, Landsat 8 and Sentinel-2 satellite imagery was fused to generate soil (SSIs), vegetation (VSIs), and moisture (MSIs) spectral indices. The SM_FC was predicted using four PTF models: RM-SPs, RM-SPs + SSIs, RM-SPs + VSIs, and RM-SPs + MSIs. Results demonstrated that incorporating spectral indices significantly enhanced SM_FC prediction accuracy. The RM-SPs + MSIs model achieved the highest performance (MSEp = 0.008, Cp = 3.8, and R²_p = 0.86). Among the different models, the combination of RM-SPs and MSIs yielded the most accurate predictions, with the model demonstrating strong reliability across all tested sites. Furthermore, the study highlighted that the fusion of Landsat 8 and Sentinel-2 data proved to be highly effective in capturing spatial variability in soil moisture, particularly in areas with varying land uses and crop types. This study highlights the effectiveness of combining field-measured soil properties with remote sensing data for large-scale SM assessment, offering a promising tool for improved agricultural water management.

Graphical Abstract

This is a visual summary that serves as a pivotal entry point into the research, offering a concise overview of the study's core findings and methodologies. The graphical abstract illustrates a comprehensive framework for estimating soil moisture at field capacity (SM_FC) using fused Landsat 8 and Sentinel-2 satellite data, combined with pedotransfer functions (PTFs). It depicts the process of collecting soil samples from three Iranian provinces and analyzing readily measurable soil properties (RM-SPs) such as bulk density, texture (sand, silt, clay), and organic carbon. The use of the van Genuchten retention model for SM_FC estimation is highlighted, alongside the extraction and classification of spectral indices into SSIs, VSIs, and MSIs. These indices are visually integrated into the soil data using PTFs to enhance prediction accuracy. The logical flow showcases the model development and evaluation using Adjusted R², Mean Square Error (MSEp), and Mallows’ Cp, culminating in the selection of four optimal PTF models. Among them, the RM-SPs + MSIs model is visually emphasized as the most accurate. The RM-SPs + MSIs model demonstrated the highest accuracy in predicting SM_FC, significantly outperforming the other models. The validation phase confirmed the model's reliability across all study areas, highlighting its potential for large-scale soil moisture assessment and drought management, conveying the study’s findings at a glance.