Author(s): Sihui Dong; Zoe Avery; Rachel Devine; Kilisimasi Latu; Asaad Y. Shamseldin; Robyn Simcock; Aung Naing Soe; Conrad Zorn; Minghao Chen

Linked Author(s):

Keywords: Green Roofs; Climate Mitigation; Machine Learning; Sustainability; Urban Heat Island

Abstract: As urban temperatures continue to rise, affecting both quality of life and ecological stability, living roofs are increasingly recognized as a viable strategy for urban cooling. This study is part of a collaborative 'Living Roofs' project between the University of Auckland and Auckland Council, utilizing three different living roof test bed configurations to evaluate their potential in mitigating the Urban Heat Island (UHI) effect. Over a 10-month period, data on air temperature, humidity, wind speed and direction, solar radiation, and atmospheric pressure were collected to characterize the microclimatic benefits of living roofs. To improve the accuracy of UHI mitigation forecasting, a hybrid model integrating Seasonal Autoregressive Integrated Moving Average (SARIMA) with Long Short-Term Memory (LSTM) networks was employed. The SARIMA-LSTM model effectively captures seasonal and nonlinear variations in meteorological parameters, achieving a root mean square error (RMSE) as low as 1.1°C and a coefficient of determination (R2) reaching 0.95 for temperature predictions. Humidity forecasts also demonstrated high reliability, with RMSE ranging between 2.0% and 3.2%. These findings underscore the efficacy of advanced data-driven approaches for modeling urban microclimate interventions, providing valuable insights for climate resilience planning and the optimization of nature-based cooling strategies.

DOI: https://doi.org/10.64697/978-90-835589-7-4_41WC-P1863-cd

Year: 2025