High-resolution air temperature mapping in urban areas: A review on different modelling techniques

Abstract

In this study, the importance of air temperature from different aspects (e. g., human and plant health, ecological and environmental processes, urban planning, and modelling) is presented in detail, and the major factors affecting air temperature in urban areas are introduced. Given the importance of air temperature, and the necessity of developing high-resolution spatio-temporal air-temperature maps, this paper categorizes the existing approaches for air temperature estimation into three categories (interpolation, regression, and simulation approaches) and reviews them. This paper focuses on high-resolution air temperature mapping in urban areas, which is difficult due to strong spatio-temporal variations. Different air temperature mapping approaches have been applied to an urban area (Berlin, Germany) and the results are presented and discussed. This review paper presents the advantages, limitations, and shortcomings of each approach in its original form. In addition, the feasibility of utilizing each approach for air temperature modelling in urban areas was investigated. Studies into the elimination of the limitations and shortcomings of each approach are presented, and the potential of developed techniques to address each limitation is discussed. Based upon previous studies and developments, the interpolation, regression and coupled simulation techniques show potential for spatio-temporal modelling of air temperature in urban areas. However, some of the shortcomings and limitations for development of high-resolution spatio-temporal maps in urban areas have not been properly addressed yet. Hence, some further studies into the elimination of remaining limitations, and improvement of current approaches to high-resolution spatio-temporal mapping of air temperature, are introduced as future research opportunities.

Machine learning-assisted mapping of city-scale air temperature: Using sparse meteorological data for urban climate modeling and adaptation

“…However, when the neighborhood-scale land cover (LC_2000) and social-economic (SE) data are introduced (v3 and v4), no consistent improvement is achieved. This may be a result of collinearity introduced by the increasing number of variables, even though those variables met the VIF requirement [25] . The smallest (0.92℃) is achieved by Model6_v2 (RK_MLR) with an of 0.959.…”

“…Recently, numerical simulation and physical models have been used [25,26] to obtain high-resolution air temperature at local scale (10 1 -10 3 m) by simulating the exchange of energy and momentum between urban canopy ows and the atmosphere [27,28] . Although numerical techniques can produce high spatiotemporal resolution results, appropriate parameterizations and huge computing power are still needed for their further applications [29] .…”

The study of effects of greenery on temperature reduction in urban areas

“…S989-S1000 creases and air quality worsens, affecting directly and indirectly people's health and comfort [2][3][4]. The major factors, which affect air temperature, can be categorized in three groups: temporal effect variablessuch as land surface temperature wind speed and cloud cover; permanent effect variables (spatial variables), such asland use/land cover urban morphology and building material and albedo; and, cyclic effect variables such assolar radiation and anthropogenic heat [5]. The research aimed at reduction of negative effects of UHI, are mostly focused on materials and formation of urban area.…”

Machine learning-assisted mapping of city-scale air temperature: Using sparse meteorological data for urban climate modeling and adaptation

“…Recently, numerical simulation and physical models have been used [25,26] to obtain high-resolution air temperature at local scale (10 1 -10 3 m) by simulating the exchange of energy and momentum between urban canopy ows and the atmosphere [27,28] . Although numerical techniques can produce high spatiotemporal resolution results, appropriate parameterizations and huge computing power are still needed for their further applications [29] .…”