Urban Heat GIS Analysis Automation

I want a fully reproducible workflow that turns raw satellite and vector data into a complete Urban Heat Island story for Wroclaw and Bengaluru. The job is split in two tightly-linked stages: one Google Earth Engine script and one Python routine.

Stage 1 – Earth Engine
Create a single .js file that I can run once per city. It must
• pull daytime land surface temperature from Landsat 8 (collection 2, Tier 1 preferred) and export the raster plus a high-resolution PNG (Fig 8)
• generate a Local Climate Zone map (Fig 9) but with boundaries adjusted to the most recent OpenStreetMap/urban extent layers rather than the default rule set
• extract night-time land surface temperature from MODIS and export the corresponding raster/PNG (Fig 11)
• output a city-wide 100 m fishnet grid as GeoJSON/Shapefile so the Python step reads it directly

Stage 2 – Python
Using the grid and rasters above, a single script should:
• build temperature box-plots by LCZ (Fig 10)
• run Getis-Ord Gi* for both day and night scenes to flag statistically significant hot/cold spots, then map the z-scores (Fig 12)
• fit MGWR, OLS and a Random Forest model, assess them with both random and spatial cross-validation, and export a comparison table plus MGWR coefficient maps (Fig 13)

Deliverables
1. GEE_UHI_Wroclaw_Bengaluru.js and UHI_hotspots_MGWR.py, fully commented
2. A short README with run instructions and any package versions
3. The six figures and the model comparison table reproduced for each city

Everything has to be parameterised so I can add new cities later by changing a few variables. Let me know if you need sample data or prefer alternative libraries; I’m open as long as the outputs and file names stay consistent with the list above.