The Urban Heat Island (UHI) & Geospatial Analysis project focuses on analyzing satellite imagery and weather data to study urban heat patterns. The notebook utilizes geospatial datasets, machine learning, and remote sensing techniques to assess the impact of land cover and building density on temperature variations.

• Geospatial Data Processing: Reads and processes building footprints (.kml), weather data (.xlsx), and thermal imagery.
• Urban Heat Island (UHI) Index Computation: Analyzes temperature variations using satellite-based Land Surface Temperature (LST).
• Building Footprint Analysis: Extracts spatial features from .kml data.
• Machine Learning Model: Trains predictive models using Training_data_uhi_index.csv and generates benchmark results.
• Weather Data Integration: Incorporates meteorological observations (NY_Mesonet_Weather.xlsx) to enhance analysis.
• Submission Template: Provides Submission_template.csv for structured model predictions.

We are proud to announce that our team, DatAi, achieved significant recognition in the 2025 EY Open Science AI and Data Challenge: Cooling Urban Heat Islands (External Participants). Our project was ranked:

• 69th internationally out of 2076 participants.

• 11th in Malaysia, showcasing our strong regional performance.

  Our model achieved an accuracy of 96.78%, demonstrating its effectiveness in predicting urban heat patterns.

  This accomplishment highlights our team's dedication and expertise in utilizing AI and geospatial analysis to address urban heat challenges. We are excited to continue contributing to innovative solutions for sustainable urban environments.

• Team Members:

  • Ong Yi Yan - Team Leader
  • Tang Yan Qing
  • Koh Su Xuan

• Received a Participation Certificate for the completion of the challenge.

• Python: Primary language for geospatial and statistical analysis.
• geopandas & rasterio: Handles vector and raster data processing.
• xarray & rioxarray: Processes satellite-derived thermal imagery.
• pandas & numpy: Data wrangling and feature engineering.
• scikit-learn: Machine learning model training and evaluation.
• matplotlib & seaborn: Visualization of UHI patterns.

• Building_Footprint.kml: Contains spatial information of buildings.
• Landsat_LST.ipynb: Jupyter Notebook for processing Land Surface Temperature data.
• NY_Mesonet_Weather.xlsx: Weather dataset for atmospheric data integration.
• Sentinel2_GeoTIFF.ipynb: Notebook for processing Sentinel-2 imagery.
• Training_data_uhi_index.csv: Dataset used for model training and evaluation.
• Submission_template.csv: Format for submitting results.
• UHI Experiment Sample Benchmark.ipynb: Baseline notebook for UHI model evaluation.
• Model.ipynb: Jupyter Notebook containing the full implementation of data processing, feature extraction, and model training.

1. Set Up Environment: Ensure Python and Jupyter Notebook are installed.
2. Install Dependencies: Run the following command:

   pip install geopandas rasterio xarray pystac_client planetary_computer rioxarray scikit-learn pandas numpy matplotlib seaborn fiona rtree odc-stac

3. Run the Notebooks: Open and execute Landsat_LST.ipynb and Sentinel2_GeoTIFF.ipynb to process geospatial data.

• Satellite Imagery: Landsat & Sentinel-2 for analyzing land surface temperature.
• Building Footprint Data: Extracted from .kml file.
• Weather Data: Meteorological observations from New York Mesonet dataset.

1. Data Loading:
   • Imported geospatial data (.kml, .tiff, .csv, .xlsx).
   • Read weather and satellite imagery datasets.
2. Feature Engineering:
   • Computed UHI index using LST from Landsat.
   • Integrated building density and land cover statistics.
3. Model Training:
   • Built predictive models using RandomForestRegressor for temperature forecasting.
   • Validated model performance with benchmark comparisons.
4. Exporting Results:
   • Processed predictions stored in Submission_template.csv.
   • Results visualized using GIS and data plotting tools.