Evaluating MODIS cloud-free snow cover datasets using massive spatial benchmark data in the Tibetan Plateau.

Area of Science:

• Earth Observation
• Climate Science
• Remote Sensing

Background:

• Accurate snow cover data is critical for climate change research, water resource management, and model calibration.
• Existing Moderate-resolution Imaging Spectroradiometer (MODIS) snow cover datasets lack systematic evaluation due to varied benchmarks and parameters.
• Traditional validation methods using station data face spatial and temporal scale mismatches and underrepresentation.

Purpose of the Study:

• To establish a scale-matched spatial benchmark dataset for evaluating MODIS snow cover products.
• To systematically assess the performance of seven MODIS cloud-free snow cover datasets across different conditions.
• To provide objective assessment methods for satellite snow cover data and guide future algorithm development.

Main Methods:

• Compiled a large-scale benchmark dataset using over 18,000 Landsat and 11,000 Sentinel-2 images spanning two decades.
• Evaluated seven MODIS cloud-free snow cover datasets based on seasons, elevation, land cover, and subregions.
• Compared clear-sky, cloud-removed, and integrated spatiotemporal cloud removal approaches.

Main Results:

• NIEER_MODIS_SCE demonstrated superior performance for clear-sky conditions, especially with optimized NDSI thresholds per land use.
• Spatiotemporal cloud removal datasets outperformed other methods for cloud-removed data.
• The best overall dataset achieved 0.82 overall accuracy and 84.56% snow retrieval accuracy, with weaknesses noted in forest areas.

Conclusions:

• Regional customization of snow mapping algorithms is crucial for accuracy.
• Spatiotemporal cloud removal techniques offer significant improvements over single-step methods.
• The developed benchmark and assessment methods provide a robust framework for evaluating satellite snow cover products and informing future strategies.