Selecting urban-rural pairs of stations for Urban Heat Island effect: the novel MTD metric applied to Turin, Italy

In this work, we propose the Mean Temperature Difference (MTD) as a novel metric for selecting urban-rural pairs of stations needed in the Urban Heat Island (UHI) quantification. UHI refers to the warmer temperatures experienced by a city than its rural surroundings and it is commonly quantified as the difference between urban and rural air temperatures. As a consequence, the definition of stations pairs is a crucial task to evaluate the UHI intensity. The novelty of our MTD is the capability to objectively classify a heterogeneous group of weather stations without the need for preliminary assumptions about their landscape, i.e., urban, rural or peri-urban. This metric employs monthly-averaged hourly temperature measurements and is based on two main steps. Firstly, by adopting a simple formulation, MTD highlights the thermal pattern typical of each station with respect to the average one of the area of interest. Afterward, the Principal Component Analysis (PCA) is used to cluster stations into urban, rural and peri-urban landscapes. To show the features of the proposed method, we apply it to the metropolitan area of Turin, Italy, which is characterized by a complex morphology and therefore it is suitable to test the MTD. As usually happens with climatology and meteorology data, we find that the first two principal components emerging from the PCA analysis are sufficient to describe the thermal behavior of the stations. In particular, the first principal component explains the typical pattern characterizing the Urban Heat Island phenomenon: at night times the urban area retains more heat and cools slower than the rural surroundings, showing higher temperatures. Furthermore, a comparison with validated methods existing in literature to select urban-rural pairs demonstrates that the proposed approach is reliable, easily interpretable and valuable also when the study area exhibits a non-trivial landscape categorization. Given the widely recognized difficulty in defining good pairs of stations, our metric provides an additional and rigorous tool in this research field.