Nie et al. (2026) Soil Moisture Retrieval Using Multi-Satellite Dual-Frequency GNSS-IR Considering Environmental Factors

Identification

• Journal: Remote Sensing
• Year: 2026
• Date: 2026-03-18
• Authors: Shihai Nie, Yongjun Jia, Peng Li, Xing Wu, Yuchao Tang
• DOI: 10.3390/rs18060917

Research Groups

• National Satellite Ocean Application Service, Beijing, China
• State Key Laboratory of Remote Sensing Science and Digital Earth, Beijing Normal University, Beijing, China
• School of Land Science and Technology, China University of Geosciences (Beijing), Beijing, China
• School of Geographic Information and Tourism, Chuzhou University, Chuzhou, China
• Beijing Special Engineering Design Institute, Beijing, China

Short Summary

This study developed a dual-frequency GNSS-IR framework for soil moisture retrieval, integrating multi-satellite observations and environmental factors. It found that retrieval performance converges with 5-6 satellites per constellation, and that dual-frequency fusion and environmentally informed nonlinear models significantly enhance accuracy and stability.

Objective

• To quantify the optimal number of satellites for single-constellation multi-satellite GNSS-IR soil moisture retrieval and assess the combined effects and relative contributions of multiple environmental factors (Normalized Difference Vegetation Index (NDVI), temperature, precipitation) on retrieval accuracy and stability.

Study Configuration

• Spatial Scale: Two stations from the Plate Boundary Observatory (PBO) network in the USA: P031 (Colorado, inland plateau semi-arid region, 39.51550° N, 107.90867° W, 1657.6 m altitude) and P387 (Oregon, coastal humid region, 44.29675° N, 121.57446° W, 963.0 m altitude). Antennas were installed approximately 2.0 m above the ground surface. The terrain around both sites is relatively flat with few obstructions.
• Temporal Scale: 232 consecutive days of GNSS data (Day of Year (DOY) 071–302, 2017), covering the growing season and its transitional stages. Data were sampled at 15 s intervals. Temporal robustness was assessed using three sub-windows: W1 (DOY 71–150), W2 (DOY 151–230), and W3 (DOY 231–302).

Methodology and Data

• Models used:
  • Entropy-based fusion (EFM) for adaptive weighting of dual-frequency phase-delay observations.
  • Univariate Linear Regression (ULR) model.
  • Random Forest (RF) model (nonlinear, incorporating environmental factors).
  • Lomb-Scargle (L-S) spectral analysis for dominant frequency extraction.
  • Least squares for phase-delay parameter estimation.
• Data sources:
  • GNSS Observation Data: Multi-constellation (GPS, GLONASS, Galileo, BDS) dual-frequency Signal-to-Noise Ratio (SNR) data from the Plate Boundary Observatory (PBO) network (https://www.unavco.org/). Low elevation angles (5–30°) were used.
  • Environmental Factor Data:
    • Precipitation and Temperature: In situ sensor measurements from the International Soil Moisture Network (ISMN) (https://ismn.earth/en/).
    • Normalized Difference Vegetation Index (NDVI): Moderate-Resolution Imaging Spectroradiometer (MODIS) NDVI product (500 m spatial resolution, 16-day temporal resolution), resampled to daily values using a Piecewise Cubic Hermite Interpolating Polynomial (PCHIP) (https://ladsweb.modaps.eosdis.nasa.gov/).
  • Reference Data: In situ Soil Moisture Content (SMC) measurements.

Main Results

• Optimal Satellite Number: Multi-satellite combinations significantly improve SMC retrieval stability, but the gain exhibits diminishing returns, converging when approximately 5–6 satellites within a single GNSS constellation are jointly used. A five-satellite configuration was identified as optimal for balancing accuracy and computational efficiency.
• Dual-Frequency Fusion Performance: EFM-based dual-frequency fusion consistently outperforms single-frequency schemes across all GNSS constellations, enhancing the response of phase-delay observables to soil moisture variations while suppressing noise.
  • At station P031, average improvements for ULR were 14.81% in R², 12.44% in RMSE, and 16.96% in MAE. For RF, improvements were 4.30% in R², 11.61% in RMSE, and 12.79% in MAE.
  • At station P387, average improvements for ULR were 16.71% in R², 16.70% in RMSE, and 17.94% in MAE. For RF, improvements were 3.63% in R², 13.19% in RMSE, and 12.63% in MAE.
• Environmental Factors and Nonlinear Modeling: The RF model, incorporating NDVI, temperature, and precipitation, consistently outperforms the linear ULR model in retrieval capability and statistical stability across all GNSS constellations and multi-satellite schemes.
  • At station P031, the average improvements of RF over ULR were 20.04% in R², 17.93% in RMSE, and 19.57% in MAE.
  • At station P387, the average improvements of RF over ULR reached 28.63% in R², 18.74% in RMSE, and 35.61% in MAE.
• Dominant Environmental Drivers: The relative contributions of environmental factors are site-dependent. At P031 (inland semi-arid), temperature is the dominant factor (56.28%), followed by NDVI (35.03%). At P387 (coastal humid), NDVI is the dominant factor (61.09%), followed by temperature (33.64%). Precipitation consistently showed the smallest contribution at both sites.
• Temporal Robustness: The key findings regarding the convergence of multi-satellite gain (around 5–6 satellites), the superiority of dual-frequency fusion, and the improved performance of the environmental-factor-informed RF model were found to be temporally consistent across different sub-windows of the observation period.

Contributions

• Proposed a marginal threshold criterion to quantitatively characterize the diminishing returns in retrieval performance with increasing satellite participation and to determine the optimal number of satellites for single-constellation GNSS-IR soil moisture monitoring.
• Established an integrated single-constellation, multi-satellite, dual-frequency fusion framework that incorporates multiple environmental factors (NDVI, temperature, precipitation) to quantify their relative contributions to retrieval accuracy and stability, thereby enhancing the robustness of GNSS-IR soil moisture retrieval under complex environmental conditions.

Funding

• Open Fund of State Key Laboratory of Remote Sensing and Digital Earth (grant number OFSLRSS202519)
• Major Program of the National Natural Science Foundation of China (grant number 42192561)

Citation

@article{Nie2026Soil,
  author = {Nie, Shihai and Jia, Yongjun and Li, Peng and Wu, Xing and Tang, Yuchao},
  title = {Soil Moisture Retrieval Using Multi-Satellite Dual-Frequency GNSS-IR Considering Environmental Factors},
  journal = {Remote Sensing},
  year = {2026},
  doi = {10.3390/rs18060917},
  url = {https://doi.org/10.3390/rs18060917}
}