Mapping Soil Organic Carbon by Integrating Time-Series Sentinel-2 Data, Environmental Covariates and Multiple...

Area of Science:

• Remote Sensing
• Soil Science
• Environmental Monitoring

Background:

• Sentinel-2 (S-2) data is widely used for soil organic carbon (SOC) mapping.
• The full potential of time-series S-2 data for SOC mapping remains underexplored.

Purpose of the Study:

• To develop an innovative approach for mining time-series S-2 data for SOC mapping.
• To identify the optimal monitoring time window for SOC estimation.
• To precisely map SOC in arid regions using integrated data and models.

Main Methods:

• Analysis of temporal variation patterns in SOC and time-series S-2 data correlations.
• Integration of environmental covariates with multiple ensemble models (stacking, weight averaging, sample averaging).
• Inclusion of soil properties and S-2 texture information in prediction models.

Main Results:

• The correlation between SOC and S-2 data shows interannual and monthly variations, with July-August identified as the optimal monitoring window.
• Soil properties and S-2 texture information significantly improved SOC prediction model accuracy (8.85% and 61.78%, respectively).
• The stacking ensemble model demonstrated superior prediction performance compared to weight and sample averaging models.

Conclusions:

• Mining spectral and texture information within the optimal monitoring time window is crucial for accurate SOC mapping.
• The integration of environmental covariates and advanced ensemble models enhances SOC prediction capabilities.
• This approach offers high potential for precise soil organic carbon mapping in arid environments.