Prediction of outdoor thermal comfort changes and uncovering mitigation strategies based on machine learning algorithm : a decision support tool for climate-sensitive design : a case study of Glasgow, UK

Abstract

Cities are becoming increasingly warm as a result of climate change and increasing population (Dimoudi et al., 2013). It forewarns severe weather and climate catastrophes happening considerably sooner than initially anticipated. There will be more heatwaves, longer warm seasons, and shorter cold seasons. It is predicted that the UK will experience 50% more hot days in the decades ahead (Arnell et al., 2021). The magnitude of the heat island in Glasgow can approach approximately 4 °C under certain climatic circumstances (Krüger, Drach and Emmanuel, 2018). Such thermal variation in Glasgow demands more action; otherwise, rising UHI intensity may have disastrous impacts on inhabitants' health.

In light of managing and controlling the outdoor environment to meet thermal comfort, this research developed a framework for predicting and simulating the thermal comfort proxies depending on the secondary and historical data. This framework will facilitate the decision-making process from a climatic perspective at the initial planning stage by forecasting the heat stress changes in urban settings. This study has considered Generalized Linear Regression (GLR), Exploratory Regression, Spatial Autocorrelation, Geographically Weighted regression (GWR), and Artificial Neural Network (ANN) algorithms to find the best fit model to predict the proxies of thermal comfort.

The ANN model achieved better performance predicting Mean Radiant Temperature (MRT) and Land Surface Temperature LST considering the 12 predictors. Among all the factors, SVF was the best factor in predicting MRT, while NDBI was recognised as a significant variable in LST forecasting. The framework was validated by predicting LST for vacant lands in Glasgow. Three scenarios were considered to evaluate the impact of greening strategies (100%-50%-0%).

The results from LST prediction depict the modest effects of urban greenery in decreasing heat stress at the surface level under a non-linear trend. The UTCI simulation in ENVI-met was applied in the framework to understand thermal comfort. The impact of shadowing from plants and buildings, for instance, could alter thermal comfort depending on the area's compactness and openness in Glasgow's central district. Heat mitigation measures at the level of lowering the surface temperature do not always meet human thermal comfort. The association between LST and MRT has made it clear that it is impossible to establish a direct connection between them. It is advantageous to provide such a tool that may be used by policymakers who are less proficient in climatology, to forecast changes in heat stress in urban environments swiftly.