Forecasting UHI Intensity in Sydney Using LSTM Models

Patrick Duhirwe

Data Analyst
ML Engineer
Python
R
TensorFlow
  1. Data Collection and Processing: Hourly air temperature data over 18 years from 8 sites around Sydney was collected, with quality control measures applied to ensure data integrity.
  2. Temporal UHI Variations Analysis: Investigated diurnal, monthly, seasonal, and annual variations of UHI intensity using the difference in temperature between urban areas and a suburban/rural reference location.
  3. LSTM Model Development: Two types of LSTM models were developed; one using ambient temperature (LSTMT) and the other incorporating both air temperature and wind speed (LSTMW), to forecast UHI intensity.
  4. Model Settings and Performance Validation: Model parameters were optimized through trial and error, and performance was validated using statistical metrics such as RMSE, MAE, and R^2.
  5. Analysis of Results: The UHI trends and patterns were analyzed, including the impact of diurnal and seasonal changes, and spatial distribution across the different monitoring stations.



Partner With Patrick
View Services

More Projects by Patrick