A Framework to Predict High-Resolution Spatiotemporal PM<sub>2.5</sub> Distributions Using a Deep-Learning Model: A Case Study of Shijiazhuang, China
Air-borne particulate matter, PM<sub>2.5</sub> (PM having a diameter of less than 2.5 micrometers), has aroused widespread concern and is a core indicator of severe air pollution in many cities globally. In our study, we present a validated framework to predict the daily PM<sub>2.5...
Main Authors: | , , , , , , |
---|---|
Format: | Article |
Language: | English |
Published: |
MDPI AG
2020-08-01
|
Series: | Remote Sensing |
Subjects: | |
Online Access: | https://www.mdpi.com/2072-4292/12/17/2825 |
_version_ | 1797555004579512320 |
---|---|
author | Guangyuan Zhang Haiyue Lu Jin Dong Stefan Poslad Runkui Li Xiaoshuai Zhang Xiaoping Rui |
author_facet | Guangyuan Zhang Haiyue Lu Jin Dong Stefan Poslad Runkui Li Xiaoshuai Zhang Xiaoping Rui |
author_sort | Guangyuan Zhang |
collection | DOAJ |
description | Air-borne particulate matter, PM<sub>2.5</sub> (PM having a diameter of less than 2.5 micrometers), has aroused widespread concern and is a core indicator of severe air pollution in many cities globally. In our study, we present a validated framework to predict the daily PM<sub>2.5</sub> distributions, exemplified by a use case of Shijiazhuang City, China, based on daily aerosol optical depth (AOD) datasets. The framework involves obtaining the high-resolution spatiotemporal AOD distributions, estimation of the spatial distributions of PM<sub>2.5</sub> and the prediction of these based on a convolutional long short-term memory (ConvLSTM) model. In the estimation part, the eXtreme gradient boosting (XGBoost) model has been determined as the estimation model with the lowest root mean square error (RMSE) of 32.86 µg/m<sup>3</sup> and the highest coefficient of determination regression score function (R<sup>2</sup>) of 0.71, compared to other common models used as a baseline for comparison (linear, ridge, least absolute shrinkage and selection operator (LASSO) and cubist). For the prediction part, after validation and comparison with a seasonal autoregressive integrated moving average (SARIMA), which is a traditional time-series prediction model, in both time and space, the ConvLSTM gives a more accurate performance for the prediction, with a total average prediction RMSE of 14.94 µg/m<sup>3</sup> compared to SARIMA’s 17.41 µg/m<sup>3</sup>. Furthermore, ConvLSTM is more stable and with less fluctuations for the prediction of PM<sub>2.5</sub> in time, and it can also eliminate better the spatial predicted errors compared to SARIMA. |
first_indexed | 2024-03-10T16:41:08Z |
format | Article |
id | doaj.art-c5314cf0f5ee43f49bd7617b78d7de8e |
institution | Directory Open Access Journal |
issn | 2072-4292 |
language | English |
last_indexed | 2024-03-10T16:41:08Z |
publishDate | 2020-08-01 |
publisher | MDPI AG |
record_format | Article |
series | Remote Sensing |
spelling | doaj.art-c5314cf0f5ee43f49bd7617b78d7de8e2023-11-20T12:03:40ZengMDPI AGRemote Sensing2072-42922020-08-011217282510.3390/rs12172825A Framework to Predict High-Resolution Spatiotemporal PM<sub>2.5</sub> Distributions Using a Deep-Learning Model: A Case Study of Shijiazhuang, ChinaGuangyuan Zhang0Haiyue Lu1Jin Dong2Stefan Poslad3Runkui Li4Xiaoshuai Zhang5Xiaoping Rui6IoT Laboratory, School of Electronic Engineering and Computer Science, Queen Mary University of London, London E1 4NS, UKSchool of Earth Sciences and Engineering, Hohai University, Nanjing 211000, ChinaCollege of Resources and Environment, University of Chinese Academy of Sciences, Beijing 100049, ChinaIoT Laboratory, School of Electronic Engineering and Computer Science, Queen Mary University of London, London E1 4NS, UKCollege of Resources and Environment, University of Chinese Academy of Sciences, Beijing 100049, ChinaIoT Laboratory, School of Electronic Engineering and Computer Science, Queen Mary University of London, London E1 4NS, UKSchool of Earth Sciences and Engineering, Hohai University, Nanjing 211000, ChinaAir-borne particulate matter, PM<sub>2.5</sub> (PM having a diameter of less than 2.5 micrometers), has aroused widespread concern and is a core indicator of severe air pollution in many cities globally. In our study, we present a validated framework to predict the daily PM<sub>2.5</sub> distributions, exemplified by a use case of Shijiazhuang City, China, based on daily aerosol optical depth (AOD) datasets. The framework involves obtaining the high-resolution spatiotemporal AOD distributions, estimation of the spatial distributions of PM<sub>2.5</sub> and the prediction of these based on a convolutional long short-term memory (ConvLSTM) model. In the estimation part, the eXtreme gradient boosting (XGBoost) model has been determined as the estimation model with the lowest root mean square error (RMSE) of 32.86 µg/m<sup>3</sup> and the highest coefficient of determination regression score function (R<sup>2</sup>) of 0.71, compared to other common models used as a baseline for comparison (linear, ridge, least absolute shrinkage and selection operator (LASSO) and cubist). For the prediction part, after validation and comparison with a seasonal autoregressive integrated moving average (SARIMA), which is a traditional time-series prediction model, in both time and space, the ConvLSTM gives a more accurate performance for the prediction, with a total average prediction RMSE of 14.94 µg/m<sup>3</sup> compared to SARIMA’s 17.41 µg/m<sup>3</sup>. Furthermore, ConvLSTM is more stable and with less fluctuations for the prediction of PM<sub>2.5</sub> in time, and it can also eliminate better the spatial predicted errors compared to SARIMA.https://www.mdpi.com/2072-4292/12/17/2825PM<sub>2.5</sub>AODXGBoostpredictiondeep learningConvLSTM |
spellingShingle | Guangyuan Zhang Haiyue Lu Jin Dong Stefan Poslad Runkui Li Xiaoshuai Zhang Xiaoping Rui A Framework to Predict High-Resolution Spatiotemporal PM<sub>2.5</sub> Distributions Using a Deep-Learning Model: A Case Study of Shijiazhuang, China Remote Sensing PM<sub>2.5</sub> AOD XGBoost prediction deep learning ConvLSTM |
title | A Framework to Predict High-Resolution Spatiotemporal PM<sub>2.5</sub> Distributions Using a Deep-Learning Model: A Case Study of Shijiazhuang, China |
title_full | A Framework to Predict High-Resolution Spatiotemporal PM<sub>2.5</sub> Distributions Using a Deep-Learning Model: A Case Study of Shijiazhuang, China |
title_fullStr | A Framework to Predict High-Resolution Spatiotemporal PM<sub>2.5</sub> Distributions Using a Deep-Learning Model: A Case Study of Shijiazhuang, China |
title_full_unstemmed | A Framework to Predict High-Resolution Spatiotemporal PM<sub>2.5</sub> Distributions Using a Deep-Learning Model: A Case Study of Shijiazhuang, China |
title_short | A Framework to Predict High-Resolution Spatiotemporal PM<sub>2.5</sub> Distributions Using a Deep-Learning Model: A Case Study of Shijiazhuang, China |
title_sort | framework to predict high resolution spatiotemporal pm sub 2 5 sub distributions using a deep learning model a case study of shijiazhuang china |
topic | PM<sub>2.5</sub> AOD XGBoost prediction deep learning ConvLSTM |
url | https://www.mdpi.com/2072-4292/12/17/2825 |
work_keys_str_mv | AT guangyuanzhang aframeworktopredicthighresolutionspatiotemporalpmsub25subdistributionsusingadeeplearningmodelacasestudyofshijiazhuangchina AT haiyuelu aframeworktopredicthighresolutionspatiotemporalpmsub25subdistributionsusingadeeplearningmodelacasestudyofshijiazhuangchina AT jindong aframeworktopredicthighresolutionspatiotemporalpmsub25subdistributionsusingadeeplearningmodelacasestudyofshijiazhuangchina AT stefanposlad aframeworktopredicthighresolutionspatiotemporalpmsub25subdistributionsusingadeeplearningmodelacasestudyofshijiazhuangchina AT runkuili aframeworktopredicthighresolutionspatiotemporalpmsub25subdistributionsusingadeeplearningmodelacasestudyofshijiazhuangchina AT xiaoshuaizhang aframeworktopredicthighresolutionspatiotemporalpmsub25subdistributionsusingadeeplearningmodelacasestudyofshijiazhuangchina AT xiaopingrui aframeworktopredicthighresolutionspatiotemporalpmsub25subdistributionsusingadeeplearningmodelacasestudyofshijiazhuangchina AT guangyuanzhang frameworktopredicthighresolutionspatiotemporalpmsub25subdistributionsusingadeeplearningmodelacasestudyofshijiazhuangchina AT haiyuelu frameworktopredicthighresolutionspatiotemporalpmsub25subdistributionsusingadeeplearningmodelacasestudyofshijiazhuangchina AT jindong frameworktopredicthighresolutionspatiotemporalpmsub25subdistributionsusingadeeplearningmodelacasestudyofshijiazhuangchina AT stefanposlad frameworktopredicthighresolutionspatiotemporalpmsub25subdistributionsusingadeeplearningmodelacasestudyofshijiazhuangchina AT runkuili frameworktopredicthighresolutionspatiotemporalpmsub25subdistributionsusingadeeplearningmodelacasestudyofshijiazhuangchina AT xiaoshuaizhang frameworktopredicthighresolutionspatiotemporalpmsub25subdistributionsusingadeeplearningmodelacasestudyofshijiazhuangchina AT xiaopingrui frameworktopredicthighresolutionspatiotemporalpmsub25subdistributionsusingadeeplearningmodelacasestudyofshijiazhuangchina |