摘要
针对传统近地面NO_2浓度空间模拟过程中NO_2浓度与其影响要素之间关系的复杂非线性机制解释不充分的缺陷,本研究基于随机森林(RF)算法、融合多源地理要素开展了近地面NO_2浓度空间分布模拟研究.以卫星OMI对流层NO_2柱浓度数据和多源地理要素(道路交通、气象因子、土地利用/覆盖、地形高程、人口数量)为输入变量,近地面NO_2浓度为输出变量,利用RF算法构建近地面NO_2浓度反演模型.通过对比地面观测数据与传统土地利用回归模型(LUR)检验RF模型的有效性,基于所构建的最优RF模型在不同时间尺度下模拟分析中国大陆地区近地面NO_2浓度空间分布特征.结果表明:(1)集成多源地理要素的RF回归模型精度高,月均模型整体拟合度R^2 0.85,RMSE 6.08μg/m^3,交叉验证的R^2 0.84,RMSE 6.33μg/m^3,显著高于LUR模型(拟合R^2 0.53,RMSE 10.48μg/m^3,交叉验证的R^2 0.53,RMSE 10.49μg/m^3);(2)地面NO_2浓度与预测变量呈现显著的复杂非线性与时间尺度依赖关系,卫星OMI柱浓度对模型影响程度最大,重要性指标IncMSE介于97.40%～116.54%,多源地理特征变量对RF模型同样具有不可忽视的贡献力(IncMSE在23.34%～47.53%之间);(3)中国大陆地区NO_2污染程度较高,年均模拟浓度为24.67μg/m^3,存在明显季节性空间差异,NO_2浓度冬季(31.85μg/m^3)>秋季(24.86μg/m^3)>春季(23.24μg/m^3)>夏季(18.75μg/m^3),呈现以华北平原为高值中心、向外围逐渐减轻的空间分布格局.较已有研究揭示对流层NO_2柱浓度宏观分布特征,本研究对近地面NO_2污染特征的研究成果对于合理制定污染防控策略、降低居民暴露健康损害具有指导意义.
In order to capture the complex and nonlinear relationship between ground-level NO2 concentrations and predictor variables, random forest (RF) models combined with multiple types of geographic covariates were developed to estimate ground-level NO2 concentrations. In this process, satellite-based OMI NO2 tropospheric columns and multi-source geographic covariates (i.e., road network, meteorological factors, land use/cover, DEM and population density) were used as potential predictor variables and ground-level NO2 concentrations were used as the dependent variable for RF models construction. The reliability of the RF models was validated by comparison with ground-measured NO2 concentrations and typical linear land use regression (LUR) models. Afterwards, the spatial distribution characteristics of NO2 concentration mapped by RF models across time scales in China's Mainland were assessed and analyzed. Results showed that RF modeling outperformed LUR modeling with obvious higher model fitting-based R^2 and lower RMSE, which were 0.85 and 6.08μg/m^3 for monthly RF models compared with 0.53 and 10.48μg/m^3 for LUR models. This was confirmed by the cross-validation-based R^2 and RMSE with values of 0.84 and 6.33μg/m^3, while those of LUR models were 0.53 and 10.49μg/m^3. The partial dependence of RF models suggested that the actual relationships between ground-level NO2 concentrations and predictor variables were nonlinear and time-dependent. OMI NO2 tropospheric columns contributed most strongly to the RF models of NO2 concentrations, which had largest percentage of IncMSE (ranged from 97.40% to 116.54%). Meanwhile, the importance of different geographic variables could not be disregarded, which had values of IncMSE between 23.34% and 47.53%. Additionally, the NO2 concentrations simulated by RF models showed that the annual average NO2 concentrations across China's Mainland during the study period were 24.67μg/m^3, which had significant seasonal variations with value of 31.85, 24.86, 23.24 and 18.75μg/m^3 in wint
作者
游介文
邹滨
赵秀阁
许珊
何瑞
YOU Jie-wen;ZOU Bin;ZHAO Xiu-ge;XU Shan;HE Rui(School of Geosciences and Info-physics, Central South University, Changsha410083, China;State Key Laboratory of Environmental Criteria and Risk Assessment, Chinese Research Academy of Environmental Sciences, Beijing 100012, China)
出处
《中国环境科学》
EI
CAS
CSCD
北大核心
2019年第3期969-979,共11页
China Environmental Science
基金
国家重点研发计划项目(2016YFC0206205)
国家自然科学基金资助项目(41871317)
中南大学创新驱动计划(20170019010005)
关键词
NO
2
时空分布
随机森林
卫星遥感
地理要素
NO 2
spatial-temporal distribution
random forest
satellite remote sensing
geographic covariates
