Mar 19: Scientists from Beijing Normal University and affiliated research centers have developed a novel method to improve soil-moisture monitoring in complex, vegetated landscapes, a critical factor for weather forecasting, drought assessment, and agricultural management. The study, published in the Journal of Remote Sensing (DOI: 10.34133/remotesensing.0981), demonstrates how integrating MODIS Bidirectional Reflectance Distribution Function (BRDF) information with CYGNSS satellite observations can overcome limitations of traditional vegetation-index-based approaches.
Soil moisture plays a key role in linking Earth’s water, energy, climate, and agricultural systems. However, dense vegetation can interfere with microwave-based satellite signals, reducing the accuracy of conventional retrieval methods that rely on indicators like NDVI and EVI. To address this, the research team developed Scheme A+, an optimized Random Forest model that selectively incorporates four key BRDF parameters alongside CYGNSS data to better represent vegetation structure and directional reflectance effects.
Key results of the study include:
-
High correlation with reference data: Scheme A+ achieved a correlation coefficient of 0.94 with SMAP soil-moisture observations.
-
Improved accuracy: Root-mean-square error (RMSE) was 0.024 cm³/cm³, 4.32% lower than NDVI-based schemes and 6.59% lower than EVI-based schemes.
-
Strongest gains in forested areas, where dense canopies weaken CYGNSS signals and make moisture retrieval particularly challenging.
-
Efficient feature selection: Using just 19% of BRDF variables retained most of the model’s predictive power, balancing accuracy with computational efficiency.
Feature-importance analysis highlighted the most influential variables in forested regions: b5 fiso (18.32%), DEM (13.68%), b2 fiso (8.32%), b5 fvol (6.8%), and b4 fiso (6.32%), demonstrating that shortwave-infrared isotropic and volumetric scattering carry critical information for vegetation-impacted moisture retrieval.
The researchers emphasized that the value of BRDF lies in capturing directional and structural information about vegetation, rather than simply adding more spectral bands. The combined optical–microwave approach provides a significant advancement for soil-moisture monitoring across forests, shrublands, grasslands, and croplands, especially where traditional index-based methods tend to underestimate moisture levels.
“This integration of BRDF and CYGNSS data opens a new pathway for accurate soil-moisture retrieval in densely vegetated environments,” said the study team. “It offers practical benefits for agriculture, water resource management, and climate monitoring in regions where vegetation complexity has historically challenged remote-sensing approaches.”
The study leveraged CYGNSS L1 V3 data from 2020, aggregated monthly to match SMAP Level 3 soil-moisture data, and included additional inputs such as surface reflectivity, roughness, elevation, slope, land-cover type, and tree height. Validation also incorporated in situ measurements from the International Soil Moisture Network (ISMN).
