A spatial regression approach to modeling urban land surface temperature.

Area of Science:

• Earth Science
• Remote Sensing
• Environmental Modeling

Background:

• Land surface temperature (LST) is crucial for urban planning, thermal comfort, and understanding environmental impacts like climate change and rainfall.
• Satellite-derived LST data is often limited by cloud cover, necessitating robust modeling for accurate forecasting.
• Spatial regression models offer a potential solution for LST prediction by accounting for spatial dependencies.

Purpose of the Study:

• To model and compare the robustness of spatial lag and spatial error models in reproducing Land Surface Temperature (LST).
• To examine the contribution of various dependent variables (built-up area, water, albedo, elevation, vegetation) to LST.
• To provide a reliable method for LST forecasting essential for environmental and urban studies.

Main Methods:

• Employed spatial lag and spatial error models for LST modeling.
• Utilized Landsat 8 and Shuttle Radar Topography Mission (SRTM) data.
• Validated models using k-fold cross-validation, mean square error, and standard deviation.

Main Results:

• Successfully modeled LST using spatial regression techniques.
• Quantified the influence of NDBI, NDVI, MNDWI, albedo, and elevation on LST.
• Demonstrated the comparative robustness of the spatial models in LST reproduction.

Conclusions:

• Spatial regression models provide a viable approach for LST modeling and forecasting, especially when direct observations are scarce.
• Understanding the interplay between land cover, elevation, and LST is key for accurate environmental assessments.
• The validated models offer a reliable tool for researchers and urban planners relying on LST data.