Abstract
Outdoor climate and comfort influence the extent of exposure and activity performed by people. For a tropical island nation of Singapore, which made US$12.4 billion in 2016 as per World Travel & Tourism Council report from tourism and related industries and is expected to grow by 7.3% annually. The population mix of Singapore consists of 30% expats (Singstat 2018) and has an open tourist policy to welcome people from around the world. Singapore’s outdoor climate is highly influenced by sun and humidity. Shielding people from these through proper infrastructure helps facilitate better transition of people. Covered walkways have been devised and established across the island nation on a phased out manner. These walkways are devised to help shield public transport commuters and general public from harsh outdoor climate in a better way. A detailed analysis of these existing covered walkways helps us understand the performance and help evolve the design process, for future walkways. The ability to quantify the comfort offered by the walkways helps us to understand their working principles and build upon this knowledge to improve future designs. The aim of this thesis is to establish a thermal comfort based walkway performance analysis for covered walkways that embodies the effect of context and climate. This study combines the survey data (perceived comfort) from walkway users and thermal sensor data (actual thermal comfort) collected at various covered walkways across Singapore. We work with descriptive statistical measures on the (smaller) measured datasets and inferential statistic techniques on a larger dataset simulated via Rhinoceros and Grasshopper interface, to help better understand the ranges of thermal comfort offered by covered walkways. Firstly, this research highlighted that the comfort offered by current walkways to have no significance and the walkways were unable to reduce the heat stress into the moderate range at all times of the day. Secondly, the comfort distribution had a huge variance in the afternoon, when compared to other times of the day. Thirdly, outdoor thermal comfort simulated using Honeybee + Ladybug plugin had a prediction accuracy of 50%, where in it underestimated or overestimated by one UTCI scale (6 degrees) with a sample distribution of 48% due to uncertainties in 3 or more microclimate variables. Lastly, the accuracy of the simulation was improved by a regression fit with the smaller dataset (n = 414) to 68% for UTCI in same scale and to 100% for UTCI off by one scale. The learning from clustering of walkway locations highlighted the following findings. Firstly, urban context related features can be employed as predictors to quantify outdoor thermal comfort. Secondly, a structured regression spline approach can be employed on the simulation data to identify and rank urban feature based on their importance. Thirdly, the feature set related to time of the day and cluster ID were identified to be of utmost importance with cumulative data across various sites. Fourthly, cluster based feature importance across various clusters identified the feature set to be different for afternoon and all times of the day. Lastly, the method of clustering can be extended to include a new location and perform comfort modelling seamlessly. The research framework established through this research can be extended for solving similar urban infrastructural problems across various scales.