汾河流域灰绿空间对季节PM2.5 的非线性影响及阈值效应

doi:10.7522/j.issn.1000-694X.2026.00068

摘要/Abstract

摘要：

探究汾河流域灰绿空间对季节PM_2.5的非线性影响与阈值效应是制定流域生态环境治理措施的重要基础。本文以汾河流域为研究对象，基于Fragstats和形态学空间格局分析模型构建汾河流域灰绿空间8个综合指标，借助XGBoost-SHAP从变量重要性、非线性响应与阈值效应3方面，解析灰绿空间对不同季节PM_2.5质量浓度的影响。结果表明：（1）2000—2023年汾河流域四季PM_2.5质量浓度可分两个阶段：2000—2010年，四季PM_2.5质量浓度总体上升，到2010—2023年则呈显著下降趋势；空间上，PM_2.5质量浓度高值沿汾河河谷自北向南带状分布。（2）PM_2.5质量浓度呈冬季最高、夏季最低、春秋居中的季节演化特征，且春、夏、秋季超35 µg·m^-3阈值面积占比由2000年的93.4%降至2023年的37.1%。（3）灰绿指标空间分布具有明显的异质性，绿地主要分布于汾河两侧山地与台地，灰地则沿河谷低地分布。（4）在本文构建的灰绿空间指标体系内，灰地空间对PM_2.5总体上呈正向影响，绿地空间则呈抑制作用，但均有季节差异；进一步地，灰地形态结构因子（C_MSF）在四季均为影响最大，且对PM_2.5的影响表现出显著的季节性阈值差异。研究结果可为汾河流域分季节优化灰绿空间，制定差异化减污措施提供科学依据。

关键词: 季节PM_2.5, 灰绿空间, XGBoost-SHAP, 非线性特征, 汾河流域

Abstract:

Investigating the seasonal nonlinear effects and threshold responses of gray-green spaces on PM_2.5 in the Fenhe River Basin provides an essential basis for developing targeted ecological and environmental management strategies. In this study， eight composite indicators of gray-green space were developed using Fragstats and morphological spatial pattern analysis. An XGBoost-SHAP framework was then applied to quantify the effects of gray-green space on seasonal PM_2.5concentrations in terms of variable importance， nonlinear responses， and threshold effects. The results showed that：（1） From 2000 to 2023， seasonal PM_2.5 concentrations in the Fenhe River Basin exhibited two distinct phases， increasing overall from 2000 to 2010 and then decreasing markedly from 2010 to 2023. Spatially， high PM_2.5 concentrations were distributed in a north-south belt along the Fenhe River valley. （2） PM_2.5 concentrations showed clear seasonal variation， with the highest values in winter， the lowest in summer， and intermediate levels in spring and autumn. The proportion of areas exceeding 35 µg·m⁻³ in spring， summer， and autumn decreased from 93.4% in 2000 to 37.1% in 2023. （3） The gray-green space indicators exhibited substantial spatial heterogeneity. Green spaces were mainly distributed in the mountainous and terraced areas on both sides of the Fenhe River， whereas gray spaces were concentrated in the low-lying river valley. （4） Within the gray-green space indicator system developed in this study， gray-space indicators were generally positively associated with PM_2.5， whereas green-space indicators showed a mitigating effect， with both relationships varying seasonally. The gray-space morphological structure factor （C_MSF） showed the highest relative contribution in four seasons， and its association with PM_2.5 exhibited distinct seasonal threshold differences. These findings provide a scientific basis for season-specific optimization of gray-green spatial patterns and targeted pollution reduction in the Fenhe River Basin.

Key words: seasonal PM _2.5, grey-green space, XGBoost-SHAP, nonlinear characteristics, Fenhe River Basin

中图分类号:

P962

邓椿, 聂桐, 张仲伍, 孙维峰. 汾河流域灰绿空间对季节PM_2.5 的非线性影响及阈值效应[J]. 中国沙漠, 2026, 46(5): 78-91.

Chun Deng, Tong Nie, Zhongwu Zhang, Weifeng Sun. Seasonal variations and threshold effects of gray-green space on PM_2.5 in the Fenhe River Basin[J]. Journal of Desert Research, 2026, 46(5): 78-91.

图/表 10

图1 汾河流域注：基于国家测绘地理信息局标准地图服务网站审图号为 GS（2020）4619 号的标准地图制作，底图边界无修改。

Fig.1 Location of the Fenhe River Basin

表1 灰绿空间综合指标及含义

Table 1 Integrated indices of gray-green space and their definitions

指标大类	指标类型	指标含义
灰地空间	灰地覆盖强度（C_CI）	灰地空间数量特征
	灰地破碎复杂度（C_FC）	灰地空间形态的破碎程度与几何复杂性
	灰地聚合连通度（C_ACI）	灰地空间组织连续性
	灰地形态结构因子（C_MSF）	灰地空间的结构完整性与形态耦合特征
绿地空间	绿地覆盖强度（G_CI）	绿地空间数量特征
	绿地破碎复杂度（G_FC）	绿地空间形态的破碎程度与几何复杂性
	绿地聚合连通度（G_ACI）	绿地空间组织连续性
	绿地形态结构因子（G_MSF）	绿地空间的结构完整性与形态耦合特征

表2 7个机器学习模拟性能评估指标

Table 2 Evaluation metrics for seven machine-learning models’simulation performance

Model	Train_R²	Train_RMSE	Train_MAE	Test_R²	Test_RMSE	Test_MAE
Elastic Net（EN）	0.382	4.193	3.500	0.252	4.846	3.701
Multiple Linear Regression （MLR）	0.382	4.193	3.497	0.247	4.860	3.700
Support Vector Regression (SVR)	0.564	3.518	2.696	0.514	3.860	3.053
Multi Layer Perceptron(MLP)	0.536	3.612	2.945	0.507	3.872	3.145
Extreme Gradient Boosting （XGBoost）	0.826	2.017	1.591	0.572	3.589	2.835
Gradient Boosting Decision Tree（GBDT）	0.772	2.520	2.016	0.567	3.613	2.857
Extreme Trees（ET）	0.791	2.370	1.881	0.560	3.640	2.873

图2 2000—2023年间汾河流域不同季节PM2.5浓度变化趋势

Fig.2 Trends in seasonal PM2.5 mass concentrations in the Fenhe River Basin from 2000 to 2023

图3 2000—2023年汾河流域不同季节PM2.5浓度（μg·m-3）空间分布

Fig.3 The spatial variation of PM2.5 （μg·m-3） from 2000 to 2023

图4 灰绿空间综合指标的空间特征

Fig.4 Spatial characteristics of the integrated index for grey-green space

图5 灰绿空间对不同季度PM2.5浓度变化的重要性排序

Fig.5 Ranking of importance of grey-green space to the variation in seasonal PM2.5 concentrations

图6 灰绿空间对不同季度PM2.5浓度的影响蜂拥图

Fig.6 Swarm plot of the impact of grey-green space on seasonal PM2.5 concentrations

图7 典型灰绿空间指标对不同季度PM2.5浓度影响的偏依赖特征水平

Fig.7 Partial dependency feature level of typical grey-green space indicators on PM2.5 mass concentrations across different groups

图8 不同树模型下冬季灰绿空间指标SHAP重要性排序对比

Fig.8 Comparison of SHAP importance rankings of grey-green space indicators in winter across different tree-based models

参考文献 38

[1]	方云皓，赵丽元，窦碧莹，等.长江中游城市群建成区绿色基础设施对PM_2.5的消减效应［J］.生态学报，2024，44（1）：117-128.
[2]	董洁芳，邓椿，张仲伍.渭河流域PM_2.5时空演化及人口暴露风险［J］.生态环境学报，2023，32（6）：1078-1088.
[3]	Cao W， Wang L， Li R，et al.Unveiling the nonlinear relationships and co-mitigation effects of green and blue space landscapes on PM_2.5 exposure through explainable machine learning［J］.Sustainable Cities and Society，2025，122：106234.
[4]	谢舞丹，吴健生.土地利用与景观格局对PM_2.5浓度的影响：以深圳市为例［J］.北京大学学报（自然科学版），2017，53（1）：160-170.
[5]	欧维新，张振，陶宇.长三角城市土地利用格局与PM_2.5浓度的多尺度关联分析［J］.中国人口·资源与环境，2019，29（7）：11-18.
[6]	周丽霞，吴涛，蒋国俊，等.长三角地区PM_2.5浓度对土地利用/覆盖转换的空间异质性响应［J］.环境科学，2022，43（3）：1201-1211.
[7]	叶深，王鹏，黄祎，等.长三角城市群城市空间形态对PM_2.5与O₃污染空间异质性特征的影响研究［J］.生态环境学报，2023，32（10）：1771-1784.
[8]	刘迅，廖珊慧，李智恒，等.基于MSPA分析绿地形态空间格局对PM_2.5的消退作用：以广州市、深圳市为例［J］.生态科学，2024，43（3）：186-195.
[9]	马西娜，王梦媱.社区公园绿地空间形态对PM_2.5与CO₂的消减影响［J］.建筑科学与工程学报，2023，40（4）：179-190.
[10]	招建聪，陈亮，杨俊毅，等.数字技术-城市形态交互作用对城市PM_2.5的影响［J］.环境科学研究，2025，38（6）：1217-1228.
[11]	Zhu C， Zeng Y.Effects of urban lake wetlands on the spatial and temporal distribution of air PM₁₀and PM_2.5 in the spring in Wuhan［J］.Urban Forestry & Urban Greening，2018，31：142-156.
[12]	Bauer T J.Interaction of urban heat island effects and land-sea breezes during a New York City heat event［J］.Journal of Applied Meteorology and Climatology，2020，59（3）：477-495.
[13]	Miao C， Yu S， Hu Y，et al.Seasonal effects of street trees on particulate matter concentration in an urban street canyon［J］.Sustainable Cities and Society，2021，73：103095.
[14]	Abhijith K V， Kumar P， Gallagher J，et al.Air pollution abatement performances of green infrastructure in open road and built-up street canyon environments：a review［J］.Atmospheric Environment，2017，162：71-86.
[15]	Cao W， Zhou W， Yu W，et al.Combined effects of urban forests on land surface temperature and PM_2.5 pollution in the winter and summer［J］.Sustainable Cities and Society，2024，104：105309.
[16]	Chen M， Ren Z， Bi S.Impact of green space patterns on PM_2.5 levels：a local climate zone perspective［J］.Journal of Cleaner Production，2024，478：143975.
[17]	Chang B， Miao C， Xu S，et al.Assessing seasonal variations in the impact of urban forests and built-up areas on air pollution［J］.Sustainable Cities and Society，2025，135：107008.
[18]	Chen M， You D， Chen Z，et al.Synergistic and divergent effects of block morphology on carbon emissions and PM_2.5 pollution in Wuhan，China［J］.Sustainable Cities and Society，2025，134：106889.
[19]	Dao X， Wang Z， Yuan X，et al.Spatiotemporal characteristics of PM_2.5 components in the Beijing-Tianjin-Hebei region and its surrounding areas during the heating seasons from 2018 to 2023［J］.Journal of Cleaner Production，2026，538：147210.
[20]	史利江，刘敏，李艳萍，等.汾河流域县域经济差异的时空格局演变及驱动因素［J］.地理研究，2020，39（10）：2361-2378.
[21]	史利江，李前锦，薛曜祖，等.高质量发展视角下汾河流域县域生态效率的演变特征及其障碍因素［J］.经济地理，2025，45（8）：62-72.
[22]	刘海龙，王改艳，张鹏航，等.汾河流域新型城镇化与生态韧性耦合协调时空演变及协调影响力研究［J］.自然资源学报，2024，39（3）：640-667.
[23]	Huo W， Zhu Z， Sun H，et al.Development of machine learning models for the prediction of the compressive strength of calcium-based geopolymers［J］.Journal of Cleaner Production，2022，380：135159.
[24]	Chen T， Guestrin C.XGBoost：a scalable tree boosting system［C］//Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining.San Francisco，California， USA：ACM，2016：785-794.
[25]	Ahmed I A， Talukdar S， Islam A R M T，et al.Contribution and behavioral assessment of physical and anthropogenic factors for soil erosion using integrated deep learning and game theory［J］.Journal of Cleaner Production，2023，416：137689.
[26]	Lundberg S M， Lee S-I.A unified approach to interpreting model predictions［C］//Advances in Neural Information Processing Systems.New York，USA：Curran Associates，Inc.，2017：30.
[27]	Yang J， Huang X.The 30&thinsp；m annual land cover dataset and its dynamics in China from 1990 to 2019［J］.Earth System Science Data，2021，13（8）：3907-3925.
[28]	Fan M， He G， Zhou M.The winter choke：coal-fired heating，air pollution，and mortality in China［J］.Journal of Health Economics，2020，71：102316.
[29]	Jiang S， Kong S， Zheng H，et al.Winter-autumn air pollution control plan in North China modified the PM_2.5 compositions and sources in Central China［J］.Atmospheric Environment，2023，306：119827.
[30]	Zhang Y-L， Cao F.Fine particulate matter（PM_2.5） in China at a city level［J］.Scientific Reports，2015，5（1）：14884.
[31]	Guo Y， Wang C， Hong S，et al.Dynamic evolution of PM_2.5 removal by urban forests during rapid urbanization：from forest landscape pattern dominance to impervious surfaces［J］.Journal of Cleaner Production，2025，493：144930.
[32]	Nowak D J， Hirabayashi S， Bodine A，et al.Tree and forest effects on air quality and human health in the United States［J］.Environmental Pollution，2014，193：119-129.
[33]	Kokkonen T V， Xie Y， Paasonen P，et al.The effect of urban morphological characteristics on the spatial variation of PM_2.5 air quality in downtown Nanjing［J］.Environmental Science：Atmospheres，2021，1（7）：481-497.
[34]	Janhäll S.Review on urban vegetation and particle air pollution：deposition and dispersion［J］.Atmospheric Environment，2015，105：130-137.
[35]	Song A， Wang Z， Li S，et al.Comparative analysis of the impact of built environment and land use on monthly and annual mean PM_2.5 levels［J］.Atmosphere，2025，16（6）：682.
[36]	Xi M， Chen X， Fan J.Assessing the impact of urban form on PM_2.5 concentration along roads in a city with mobile monitoring and machine learning models［J］.Ecotoxicology and Environmental Safety，2025，303：118789.
[37]	Chen M， Dai F， Yang B，et al.Effects of neighborhood green space on PM_2.5 mitigation：evidence from five megacities in China［J］.Building and Environment，2019，156：33-45.
[38]	佘欣璐，高吉喜，张彪.基于城市绿地滞尘模型的上海市绿色空间滞留PM_2.5功能评估［J］.生态学报，2020，40（8）：2599-2608.