Sbai Salah Eddine , Lalla Btissam Drissi , Nezha Mejjad , Jamal Mabrouki , Aleksey A. Romanov
{"title":"机器学习模型在非洲北部摩洛哥上空颗粒物时空模式预测和预报中的应用","authors":"Sbai Salah Eddine , Lalla Btissam Drissi , Nezha Mejjad , Jamal Mabrouki , Aleksey A. Romanov","doi":"10.1016/j.apr.2024.102239","DOIUrl":null,"url":null,"abstract":"<div><p>Atmospheric air pollution exposure raises morbidity and mortality rates and is a major cause of the world's illness burden. In this context, we explored spatial and temporal trends in particulate matter PM10 from 2003 to 2020 over Morocco to assess air pollution exposure. We use the capabilities of ML models to study PM10 trends using 26 predictor variables, including meteorological parameters, volatile organic compounds, atmospheric oxidants, and aerosol optical depth data from the Copernicus Atmosphere Monitoring Service (CAMS). For this purpose, three ML models were built: Multiple Linear Regression (MLR), Random Forest Regression (RFR), and Generalized Additive Model (GAM). To match and optimize these models, a set of ML algorithms has been coupled with each model. The results show all these models are highly accurate in predicting and forecasting PM10 total column trends. Cross-validation showed that GAM had better prediction ability for the PM10 total column with R<sup>2</sup> = 0.994 and a very low root mean squared error (RMSE) not exceeding 0.046 × 10<sup>−16</sup> kg/m<sup>2</sup>. The GAM model showed much higher predictive ability and lower bias than the other models. This finding can be explained by the advantages of GAMs, including their ability to capture complex and non-linear patterns in the data, making them particularly useful when relationships are not easily represented by linear models. This study has presented a comprehensive methodology for predicting the spatiotemporal variability of PM10. The proposed methodology holds potential applicability across all regions, facilitating the generation of high-resolution PM10 monitoring and the establishment of systems for the early detection of air pollution incidents in Morocco. Furthermore, the developed models exhibit versatility, enabling their application for estimating future trends of individual pollutants or making real-time predictions of air quality levels. This research contributes to advancing the understanding and proactive management of air quality in the context of Morocco, offering valuable insights for pollution control efforts.</p></div>","PeriodicalId":8604,"journal":{"name":"Atmospheric Pollution Research","volume":"15 9","pages":"Article 102239"},"PeriodicalIF":3.9000,"publicationDate":"2024-06-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"Machine learning models application for spatiotemporal patterns of particulate matter prediction and forecasting over Morocco in north of Africa\",\"authors\":\"Sbai Salah Eddine , Lalla Btissam Drissi , Nezha Mejjad , Jamal Mabrouki , Aleksey A. Romanov\",\"doi\":\"10.1016/j.apr.2024.102239\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"<div><p>Atmospheric air pollution exposure raises morbidity and mortality rates and is a major cause of the world's illness burden. In this context, we explored spatial and temporal trends in particulate matter PM10 from 2003 to 2020 over Morocco to assess air pollution exposure. We use the capabilities of ML models to study PM10 trends using 26 predictor variables, including meteorological parameters, volatile organic compounds, atmospheric oxidants, and aerosol optical depth data from the Copernicus Atmosphere Monitoring Service (CAMS). For this purpose, three ML models were built: Multiple Linear Regression (MLR), Random Forest Regression (RFR), and Generalized Additive Model (GAM). To match and optimize these models, a set of ML algorithms has been coupled with each model. The results show all these models are highly accurate in predicting and forecasting PM10 total column trends. Cross-validation showed that GAM had better prediction ability for the PM10 total column with R<sup>2</sup> = 0.994 and a very low root mean squared error (RMSE) not exceeding 0.046 × 10<sup>−16</sup> kg/m<sup>2</sup>. The GAM model showed much higher predictive ability and lower bias than the other models. This finding can be explained by the advantages of GAMs, including their ability to capture complex and non-linear patterns in the data, making them particularly useful when relationships are not easily represented by linear models. This study has presented a comprehensive methodology for predicting the spatiotemporal variability of PM10. The proposed methodology holds potential applicability across all regions, facilitating the generation of high-resolution PM10 monitoring and the establishment of systems for the early detection of air pollution incidents in Morocco. Furthermore, the developed models exhibit versatility, enabling their application for estimating future trends of individual pollutants or making real-time predictions of air quality levels. This research contributes to advancing the understanding and proactive management of air quality in the context of Morocco, offering valuable insights for pollution control efforts.</p></div>\",\"PeriodicalId\":8604,\"journal\":{\"name\":\"Atmospheric Pollution Research\",\"volume\":\"15 9\",\"pages\":\"Article 102239\"},\"PeriodicalIF\":3.9000,\"publicationDate\":\"2024-06-25\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"Atmospheric Pollution Research\",\"FirstCategoryId\":\"93\",\"ListUrlMain\":\"https://www.sciencedirect.com/science/article/pii/S1309104224002046\",\"RegionNum\":3,\"RegionCategory\":\"环境科学与生态学\",\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"Q2\",\"JCRName\":\"ENVIRONMENTAL SCIENCES\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"Atmospheric Pollution Research","FirstCategoryId":"93","ListUrlMain":"https://www.sciencedirect.com/science/article/pii/S1309104224002046","RegionNum":3,"RegionCategory":"环境科学与生态学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q2","JCRName":"ENVIRONMENTAL SCIENCES","Score":null,"Total":0}
Machine learning models application for spatiotemporal patterns of particulate matter prediction and forecasting over Morocco in north of Africa
Atmospheric air pollution exposure raises morbidity and mortality rates and is a major cause of the world's illness burden. In this context, we explored spatial and temporal trends in particulate matter PM10 from 2003 to 2020 over Morocco to assess air pollution exposure. We use the capabilities of ML models to study PM10 trends using 26 predictor variables, including meteorological parameters, volatile organic compounds, atmospheric oxidants, and aerosol optical depth data from the Copernicus Atmosphere Monitoring Service (CAMS). For this purpose, three ML models were built: Multiple Linear Regression (MLR), Random Forest Regression (RFR), and Generalized Additive Model (GAM). To match and optimize these models, a set of ML algorithms has been coupled with each model. The results show all these models are highly accurate in predicting and forecasting PM10 total column trends. Cross-validation showed that GAM had better prediction ability for the PM10 total column with R2 = 0.994 and a very low root mean squared error (RMSE) not exceeding 0.046 × 10−16 kg/m2. The GAM model showed much higher predictive ability and lower bias than the other models. This finding can be explained by the advantages of GAMs, including their ability to capture complex and non-linear patterns in the data, making them particularly useful when relationships are not easily represented by linear models. This study has presented a comprehensive methodology for predicting the spatiotemporal variability of PM10. The proposed methodology holds potential applicability across all regions, facilitating the generation of high-resolution PM10 monitoring and the establishment of systems for the early detection of air pollution incidents in Morocco. Furthermore, the developed models exhibit versatility, enabling their application for estimating future trends of individual pollutants or making real-time predictions of air quality levels. This research contributes to advancing the understanding and proactive management of air quality in the context of Morocco, offering valuable insights for pollution control efforts.
期刊介绍:
Atmospheric Pollution Research (APR) is an international journal designed for the publication of articles on air pollution. Papers should present novel experimental results, theory and modeling of air pollution on local, regional, or global scales. Areas covered are research on inorganic, organic, and persistent organic air pollutants, air quality monitoring, air quality management, atmospheric dispersion and transport, air-surface (soil, water, and vegetation) exchange of pollutants, dry and wet deposition, indoor air quality, exposure assessment, health effects, satellite measurements, natural emissions, atmospheric chemistry, greenhouse gases, and effects on climate change.