Kumar et al. (2025) High-Resolution Soil Moisture Mapping using MSAVI-LST based Triangle Method

Identification

• Journal: ISPRS annals of the photogrammetry, remote sensing and spatial information sciences
• Year: 2025
• Date: 2025-12-19
• Authors: Sonu Kumar, Rajendra Prasad, J. Sharma, B H PRAJAPATI
• DOI: 10.5194/isprs-annals-x-5-w2-2025-357-2025

Research Groups

• Indian Institute of Technology (BHU), Varanasi, India
• India Meteorological Department, New Delhi, India

Short Summary

This study developed an improved downscaling approach for SMAP soil moisture from 9 km to 1 km resolution using the 'Triangle method' by integrating the Modified Soil Adjusted Vegetation Index (MSAVI) as an alternative to NDVI. The 1-1 polynomial order with MSAVI significantly outperformed other combinations, achieving a correlation coefficient of 0.76 and an RMSE of 0.032 m³/m³.

Objective

• To develop and validate an improvised downscaling approach for SMAP soil moisture from 9 km to 1 km spatial resolution using the 'Triangle method', by integrating the Modified Soil Adjusted Vegetation Index (MSAVI) as an alternative to NDVI and evaluating different polynomial regression orders.

Study Configuration

• Spatial Scale: The study focused on the Varanasi district of Uttar Pradesh, India (25°18’45” N, 82°58’15” E). Input SMAP soil moisture was at 9 km resolution, MODIS Land Surface Temperature (LST) at 1 km, and MODIS spectral reflectance at 500 m (used to calculate vegetation indices, then aggregated to 1 km). The output downscaled soil moisture was at 1 km resolution. In-situ measurements were collected within 1 km x 1 km grids.
• Temporal Scale: In-situ observations were collected on 12 different dates between 2019 and 2025, covering various seasonal conditions. Satellite data (SMAP, MODIS LST, MODIS reflectance) were used at daily temporal resolution.

Methodology and Data

• Models used:
  • Triangle method (with polynomial regression relationships of 1-1, 1-2, 2-1, and 2-2 orders).
  • Single-Channel Algorithm (SCA) based on the tau-omega model for initial SMAP soil moisture estimation.
  • Mironov et al. (2012) dielectric model for soil dielectric constant calculation.
• Data sources:
  • Satellite:
    • SMAP L1 TB E product (9 km spatial resolution, daily) for brightness temperature.
    • MODIS MYD11A1 product (1 km spatial resolution, daily) for Land Surface Temperature (LST).
    • MODIS MOD09GA product (500 m spatial resolution, daily) for surface spectral reflectance (Bands 1 and 2 used for NDVI and MSAVI calculation).
    • SMAP L2 SM P E product for Vegetation Water Content (VWC).
  • Observation (in-situ):
    • HydraGo portable probe measurements of soil moisture, soil temperature, and dielectric constant.
    • Collected across different regions of Varanasi on 12 dates between 2019 and 2025.

Main Results

• The 1-1 polynomial order with MSAVI significantly outperformed other combinations and the traditional NDVI-based approach for soil moisture downscaling.
• For MSAVI with 1-1 polynomial order, the highest correlation coefficient (R) was 0.7603, the lowest RMSE was 0.0396 m³/m³, and the bias was 0.0029 m³/m³.
• In comparison, for NDVI with 1-1 polynomial order, the correlation coefficient (R) was 0.6152, the RMSE was 0.0377 m³/m³, and the bias was 0.0306 m³/m³.
• Lower-order polynomial models generally exhibited better performance (lower RMSE, higher R) than higher-order models for both NDVI and MSAVI.
• MSAVI consistently outperformed NDVI across most polynomial orders, demonstrating its robustness in heterogeneous landscapes.
• Downscaled soil moisture maps using MSAVI (1-1 order) showed improved spatial detail, particularly around geographical features like the Ganges River and vegetated areas, compared to the original 9 km SMAP product and NDVI-based downscaling.

Contributions

• Introduces and validates the integration of the Modified Soil Adjusted Vegetation Index (MSAVI) as an alternative vegetation index within the 'Triangle method' for downscaling SMAP soil moisture.
• Demonstrates that MSAVI significantly improves vegetation sensitivity and overall accuracy of soil moisture downscaling compared to traditional NDVI, particularly in heterogeneous landscapes with mixed land-use.
• Evaluates and identifies the optimal polynomial regression order (1-1) for the Triangle method when using MSAVI, leading to more robust and accurate high-resolution soil moisture estimation.
• Provides valuable insights for refining satellite-based soil moisture downscaling techniques, enhancing their utility for localized applications like precision agriculture and watershed hydrology.

Funding

Not explicitly mentioned in the provided text.

Citation

@article{Kumar2025HighResolution,
  author = {Kumar, Sonu and Prasad, Rajendra and Sharma, J. and PRAJAPATI, B H},
  title = {High-Resolution Soil Moisture Mapping using MSAVI-LST based Triangle Method},
  journal = {ISPRS annals of the photogrammetry, remote sensing and spatial information sciences},
  year = {2025},
  doi = {10.5194/isprs-annals-x-5-w2-2025-357-2025},
  url = {https://doi.org/10.5194/isprs-annals-x-5-w2-2025-357-2025}
}