Applicability of Downscaling Land Surface Temperature by Using Normalized Difference Sand Index.

Area of Science:

• Remote Sensing
• Geospatial Analysis
• Environmental Science

Background:

• Coarse spatial resolution Land Surface Temperature (LST) from satellite data has limitations in remote sensing applications.
• Downscaling LST is crucial for detailed analysis in heterogeneous landscapes like arid oasis-desert ecotones.

Purpose of the Study:

• To improve LST downscaling accuracy in arid oasis-desert ecotones.
• To develop a more effective downscaling approach by integrating new indices and refining existing methods.

Main Methods:

• Developed a new Normalized Difference Sand Index (NDSI) to characterize desert regions.
• Modified a previous random forest approach by removing land cover datasets and incorporating SAVI, NDBI, and NDWI.
• Applied the method to downscale LST from Landsat 8 and MODIS data in Zhangye city.

Main Results:

• The NDSI effectively captures desert characteristics, and downscaled LST aligns with oasis-desert ecosystem patterns.
• Achieved high accuracy (R²=0.99, RMSE=1.25 K) compared to ground observations (HiWATER).
• The approach demonstrated superior accuracy and minimized LST retrieval errors in desert areas compared to other methods.

Conclusions:

• The enhanced downscaling approach is accurate and suitable for arid and semi-arid regions, particularly in spring and summer.
• The model shows optimal performance in vegetation and desert areas and can be applied across various spatial resolutions.
• While effective in arid zones, its accuracy is reduced in humid regions, suggesting further refinement may be needed.