Fundamental knowledge on the determinants of air temperatures across
spatial and temporal scales is essential in climate change mitigation
and adaptation. Spatial-based statistical modelling provides an
efficient approach for the analysis and prediction of air temperatures
in human-modified environments at high spatial accuracy. The aim of the
study was firstly, to analyse the environmental factors affecting
extreme air temperature conditions in a coastal high-latitude city and
secondly, to explore the applicability of generalized additive model
(GAM) in the study of urban–rural temperatures. We utilized air
temperature data from 50 permanent temperature logger stations and
extensive geospatial environmental data on different scales from Turku,
SW Finland. We selected five temperature situations (cases) and
altogether 12 urban and natural explanatory variables for the analyses.
The results displayed that (i) water bodies and topographical conditions
were often more important than urban variables in controlling the
spatial variability of extreme air temperatures, (ii) case specificity
of the explanatory variables and their scales should be considered in
the analyses and (iii) GAM was highly suitable in quantifying and
visualizing the relations between urban–rural temperatures and
environmental determinants at local scales. The results promote the use
of GAMs in spatial-based statistical modelling of air temperature in
future.