Project-25010 Understanding Urban Heat Island Effects through Geospatial Explainable AI and Visual Analytics

Project Title    

Understanding Urban Heat Island Effects through Geospatial Explainable AI and Visual Analytics

 Chuan  Chen, PHD Candidate,  Chair of Cartography and Visual Analytics, TUM
 chuan.chen@tum.de

Project Description    

The GeoVisX framework is an innovative integration of geospatial analysis, visual analytics, and explainable artificial intelligence (XAI), aiming to enhance the transparency and interpretability of spatial modeling. This project proposes the application of GeoVisX to analyze the Urban Heat Island (UHI) phenomenon in major metropolitan areas.

UHI is a significant urban environmental issue characterized by elevated temperatures in densely built-up areas relative to surrounding regions. It has direct implications for urban sustainability, public health, and climate adaptation planning. The GeoVisX framework allows for fine-grained spatial attribution of contributing factors (e.g., land surface characteristics, population density, NDVI, building patterns), visual decomposition of spatial model predictions, and interactive dashboards for stakeholder engagement.

Tasks and Responsibilities    

Key Requirements and Workflow:
1. Familiarization with GeoVisX Components
Understand the three pillars:
a) Geospatial Analysis (e.g., spatial regression, GWR, spatial autocorrelation)
b) Explainable AI (e.g., SHAP, LIME applied to XGBoost or deep learning models)
c) Visual Analytics (e.g., interactive heatmaps, geospatial dashboards)

2. Data Preparation
Acquire and preprocess urban-scale geospatial datasets:

Remote sensing imagery for LST and NDVI

Urban built-up and land cover maps

Socioeconomic and demographic data

Project datasets into a unified coordinate system
Generate grids or neighborhood units for analysis

3. Modeling and Explainable AI
Train machine learning models (e.g., XGBoost, Random Forest, or deep CNNs) to predict LST or UHI intensity

Use XAI to quantify feature contributions

Integrate spatial location as a feature for GeoSHAP analysis

Compare results with classic spatial statistics

4. Visual Analytics & GeoVisX Implementation
Visualize SHAP values spatially via heatmaps and spatial cluster detection

Implement interactive dashboards to explore variable contributions

Build temporal comparison modules to assess seasonal UHI variation

Highlight interpretable insights

5. Policy-Relevant Interpretation and Reporting
Identify priority zones for UHI mitigation based on explainable model results

Recommend urban interventions (e.g., greening, reflective surfaces)

Provide reproducible codebase and analytic pipeline using Python 

Minimum Qualifications     

Proficient in Python, ArcGIS and various geospatial analysis methods
 

Terms of the Project

 6 months

Key Words    

GeoVisX, Urban Heat Island (UHI), Explainable AI (XAI), Geospatial Analysis, Visual Analytics
Research Topics    Urban Climate, GeoAI, Spatial Explainability, Visual Analytics
Deliverables     Utilize the GeoVisX framework to identify, explain, and visualize the spatial drivers of urban heat island effects across selected cities using remote sensing and socio-environmental data, thereby informing policy interventions and adaptive planning.