Enhancing Soil Organic Carbon Mapping with Remote Sensing-Derived Soil Moisture Indices

Obtaining accurate spatial information on soil organic carbon (SOC) is essential for understanding the global carbon cycle. Digital soil mapping (DSM) has emerged as an effective approach for SOC mapping, where the selection of influential environmental covariates plays a critical role. Soil moisture (SM), which influences soil water status and the decomposition of SOC, holds great potential as a covariate for SOC estimation, particularly due to its ability to be assessed at large spatial scales using remote sensing. Previously, the normalized shortwave-infrared difference bare soil moisture indices (NSDSIs), derived from Landsat SWIR bands during bare soil periods, have been employed in SOC mapping. However, since soils are often covered by vegetation, there is a need to develop new SM indices suitable for vegetated areas and to evaluate their performance across regions with varying vegetation densities.

In this study, we introduced a novel SM index by integrating NSDSIs into the Optical TRApezoid Model, creating the OPTRAM-NSDSI. This index was compared against the original OPTRAM based on shortwave infrared transformed reflectance (OPTRAM-STR) and NSDSIs. SM indices were generated for two study areas in China: Zhuxi, Fujian (104 samples across 43.93 km², with forestland and farmland as dominant land uses) and Heshan, Heilongjiang (106 samples across 60 km², primarily farmland). The Integrated Nested Laplace Approximation combined with the Stochastic Partial Differential Equation approach was applied as the SOM prediction model.

Our results demonstrate that incorporating SM variables into commonly used environmental covariates significantly enhances prediction accuracy. The NSDSIs achieved the highest accuracy improvement of 26.8% in terms of Lin's concordance correlation coefficient in Zhuxi, while the OPTRAM-NSDSI achieved the highest improvement of 56.7% in Heshan. This suggests that OPTRAM-NSDSI is particularly effective in regions with higher vegetation density, whereas NSDSIs perform better in areas with lower vegetation density. Additionally, the optimal image acquisition dates for SM estimation appear to coincide with the vegetation "green-up" stage.

This study offers valuable insights into leveraging SM information to enhance SOC mapping, particularly in vegetated areas.