基于多尺度地理加权回归的建成环境对通风潜力的影响研究——以武汉市主城区为例

Effect of Built Environment on Ventilation Potential Based on Multi-scale Geographically Weighted Regression:a Case Study of the Main Urban Area of Wuhan

大都市常通过优化建成环境来改善城市通风潜力,以缓解城市热岛效应并提高区域环境质量.基于风洞试验等方法模拟城市通风潜力缺乏对大尺度的连续监测,传统回归方法忽视了建成环境影响的空间非平稳性和尺度效应.本文应用电路理论识别武汉主城区通风潜力,基于3Ds(密度、多样性和设计)指标体系描述建成环境,并详细对比了普通最小二乘法(OLS)、地理加权回归(GWR)和多尺度地理加权回归(MGWR)模型的结果.研究证实使用考虑空间异质性的局部模型是必要的.MGWR放宽了GWR模型的固定带宽假设,进一步考虑了建成环境因子影响的尺度效应,有效避免了过拟合和异常符号反转.相关研究应考虑驱动机制的空间异质性和尺度效应.

Metropolises tend to improve urban ventilation potential through optimizing built environment(BE)to alleviate urban heat island and promote regional environment quality.The simulation of the urban ventilation potential based on wind tunnel tests and other methods lack continuous monitoring on a large scale,and traditional regression methods overlook the spatial non-stationary and scale effects of built environment effects.This paper applies circuit theory to identify the ventilation potential of the main urban area of Wuhan,forms an indicator system to describe the BE based on 3Ds(density,diversity,design)theory,and compares the results of ordinary least squares(OLS),geographically weighted regression(GWR)and multi-scale geographically weighted regression(MGWR)models detailedly.The result has confirmed the necessity of using a local model that considers spatial heterogeneity.MGWR relaxes the fixed bandwidth assumption of the GWR model,further considers the scale effect of built environment factors,effectively avoids overfitting and abnormal symbol reversal.Relevant research should consider the spatial heterogeneity and scale effects of driving mechanisms.