基于广义线性模型的地表臭氧浓度的预测

近几十年来,随着城市规模的扩大及工商业的发展,城市光化学烟雾污染越来越严重。作为光化学烟雾的几种重要组分之一,臭氧是表征其严重程度和危害性的一个重要指标。由于高浓度臭氧对人体健康和生态系统的严重危害,如何高效、准确地提前预报高浓度臭氧成为了至关重要的问题。根据臭氧浓度分布的非Gauss特点,该文采用广义线性模型(generalized linear model,GLM)来对地表臭氧浓度进行预报。使用美国休斯顿及旧金山地区的监测数据分别建立了基于非Gauss分布假设的GLM模型。与普通线性模型的结果对比表明:广义线性模型不仅能够提高超标臭氧天数的正确预报率,同时也能降低高浓度臭氧的预报误差。

Surface ozone concentration predictions based on generalized linear models

With industrial and urban development,urban photochemical smog has become a growing environmental problem in recent decades.As a key component of photochemical smog,ozone indicates the severity of the photochemical smog.Since ozone has been demonstrated to be one of the most destructive pollutants causing great damage to human beings and biological system,forecasts of extremely high ozone levels days in advance are very important.This paper presents a generalized linear model to forecast surface ozone concentrations based on non-Gaussian features of the ozone concentration distribution.Monitoring data for the Houston and San Francisco Bay Areas are used for comparisons.This model improves the prediction rate relative to a linear regression model and reduces the prediction errors for extremely high ozone days.