Drivers of Temporal Variability of Urban Heat Island Intensity: A Case Study of Phoenix, Arizona

Abstract

Urban heat islands (UHIs), in which land use modifications and human activities cause increased urban temperatures compared to local rural counterparts, present a critical threat for cities. The urban heat island intensity (UHII) is known to vary significantly in time, wherein synoptic meteorological conditions and temporal patterns influence its temporal variability. While there are some patterns that are understood, there exists limitations in the understanding of short-term temporal variability. Using the case study of Phoenix, this thesis defines the city’s canopy-level UHI, its diurnal behavior, and uses regression analysis through decision-tree and linear-regression algorithms to develop attributive models to identify drivers of variability among meteorological parameters at the hourly and daily resolution, and across different diurnal contexts. Decision-tree-based algorithms were able to produce relatively accurate attributive models, especially for analyzing the total UHI and nighttime UHI, wherein hysteretic dynamics related to radiative parameters and relative humidity were found to be critical drivers of short-term variability.

Keywords: Phoenix, urban heat islands, canopy urban heat island intensity, meteorological parameters, temporal variability, machine learning, attributive models