Spatially filtered ridge regression (SFRR): A regression framework to understanding impacts of land cover patterns on urban climate

Understanding the impacts of land cover pattern on the heat island effect is essential for sustainable urban development. Conventional model fitting methods have restricted ability to produce accurate estimates of the land cover-temperature association due to the lack of procedures to address two important issues: spatial dependence in proximal spatial units and high correlations among predictor variables. In this study, we seek to develop an effective framework called spatially filtered ridge regression (SFRR) to estimate the variations in the quantity and distribution of land surface temperature (LST) in response to various land cover patterns. The SFRR effectively integrates spatial autoregressive models and ridge regression, and it achieves reliable parameter estimates with substantially reduced mean square errors. We show this by comparing the performance of the SFRR to other widely adopted models using Monte Carlo simulation followed by an empirical study over central Phoenix. Results highlight the great potential of the SFRR in producing accurate statistical estimates, providing a positive step toward informed and unbiased decision-making across a wide variety of disciplines. (Code and data to reproduce the results in the case study are available at: https://github.com/cfan13/SFRRTGIS.git.).