Local regression models for spatial interpolation of urban heat island—an example from Wrocław, SW Poland

Abstract

Geographically weighted regression algorithm (GWR) has been applied to derive the spatial structure of urban heat island (UHI) in the city of Wrocław, SW Poland. Seven UHI cases, measured during various meteorological conditions and characteristic of different seasons, were selected for analysis. GWR results were compared with global regression models (MLR), using various statistical procedures including corrected Akaike Information Criterion, determination coefficient, analysis of variance, and Moran's I index. It was found that GWR is better suited for spatial modeling of UHI than MLR models, as it takes into account non-stationarity of the spatial process. However, Monte Carlo and F3 tests for spatial stationarity of the independent variables suggest that for several spatial predictors a mixed GWR-MLR approach is recommended. Both local and global models were extended by the interpolation of regression residuals and used for spatial interpolation of the UHI structure. The interpolation results were evaluated with the cross-validation approach. It was found that the incorporation of the spatially interpolated residuals leads to significant improvement of the interpolation results for both GWR and MLR approaches. Because GWR is better justified in terms of statistical specification, the combined GWR+interpolated regression residuals (GWR residual kriging; GWRK) approach is recommended for spatial modeling of UHI, instead of widely applied MLR models.

Correlation or Causality between Land Cover Patterns and the Urban Heat Island Effect? Evidence from Brisbane, Australia

“…This finding indicates that local-based models provide more rationale and a realistic relationship between UHI and its driving factors. Such a finding was confirmed also in previous literature [41,42,72]. The spatial variation of R 2 can also be used to detect areas where other factors might be needed to better explain MUHI.…”

“…Urban planners, however, may not be able to apply such models due to their complexity. The application of the GWR model to assess the relationship between LUC and UHI is rather limited with a few exceptions [41,42]. Previous studies have shown that a GWR model outperformed an OLS model in terms of predicting the relationship between UHI and its associated driving forces, including LUC.…”

Seasonality of the Urban Heat Island Effect in Madison, Wisconsin

“…Regression kriging is well known to improve interpolation skill compared to simpler methods such as inverse distance weighting and ordinary kriging (Szymanowski and Kryza 2009), which do not utilize information about the land surface. Regression kriging is well known to improve interpolation skill compared to simpler methods such as inverse distance weighting and ordinary kriging (Szymanowski and Kryza 2009), which do not utilize information about the land surface.…”

A diagnostic equation for the daily maximum urban heat island effect for cities in northwestern Europe

“…Alternatively, statistical models have been utilized, usually linear regression (i.e. Bottyán and Unger, 2003;Hoffmann et al, 2012;Szymanowski and Kryza, 2012). These methods are relatively straightforward and easy to use, but not purely physically based and require retuning for each city.…”