A Novel Evolutionary Deep Learning Approach for PM<sub>2.5</sub> Prediction Using Remote Sensing and Spatial–Temporal Data: A Case Study of Tehran

Forecasting particulate matter with a diameter of 2.5 μm (PM<sub>2.5</sub>) is critical due to its significant effects on both human health and the environment. While ground-based pollution measurement stations provide highly accurate PM<sub>2.5</sub> data, their limited numb...

Full description

Saved in:
Bibliographic Details
Main Authors: Mehrdad Kaveh, Mohammad Saadi Mesgari, Masoud Kaveh
Format: Article
Language:English
Published: MDPI AG 2025-01-01
Series:ISPRS International Journal of Geo-Information
Subjects:
Online Access:https://www.mdpi.com/2220-9964/14/2/42
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849718447612624896
author Mehrdad Kaveh
Mohammad Saadi Mesgari
Masoud Kaveh
author_facet Mehrdad Kaveh
Mohammad Saadi Mesgari
Masoud Kaveh
author_sort Mehrdad Kaveh
collection DOAJ
description Forecasting particulate matter with a diameter of 2.5 μm (PM<sub>2.5</sub>) is critical due to its significant effects on both human health and the environment. While ground-based pollution measurement stations provide highly accurate PM<sub>2.5</sub> data, their limited number and geographic coverage present significant challenges. Recently, the use of aerosol optical depth (AOD) has emerged as a viable alternative for estimating PM<sub>2.5</sub> levels, offering a broader spatial coverage and higher resolution. Concurrently, long short-term memory (LSTM) models have shown considerable promise in enhancing air quality predictions, often outperforming other prediction techniques. To address these challenges, this study leverages geographic information systems (GIS), remote sensing (RS), and a hybrid LSTM architecture to predict PM<sub>2.5</sub> concentrations. Training LSTM models, however, is an NP-hard problem, with gradient-based methods facing limitations such as getting trapped in local minima, high computational costs, and the need for continuous objective functions. To overcome these issues, we propose integrating the novel orchard algorithm (OA) with LSTM to optimize air pollution forecasting. This paper utilizes meteorological data, topographical features, PM<sub>2.5</sub> pollution levels, and satellite imagery from the city of Tehran. Data preparation processes include noise reduction, spatial interpolation, and addressing missing data. The performance of the proposed OA-LSTM model is compared to five advanced machine learning (ML) algorithms. The proposed OA-LSTM model achieved the lowest root mean square error (RMSE) value of 3.01 µg/m<sup>3</sup> and the highest coefficient of determination (<i>R</i><sup>2</sup>) value of 0.88, underscoring its effectiveness compared to other models. This paper employs a binary OA method for sensitivity analysis, optimizing feature selection by minimizing prediction error while retaining critical predictors through a penalty-based objective function. The generated maps reveal higher PM<sub>2.5</sub> concentrations in autumn and winter compared to spring and summer, with northern and central areas showing the highest pollution levels.
format Article
id doaj-art-12706d70fe79430bad2cc8a0f51561d6
institution DOAJ
issn 2220-9964
language English
publishDate 2025-01-01
publisher MDPI AG
record_format Article
series ISPRS International Journal of Geo-Information
spelling doaj-art-12706d70fe79430bad2cc8a0f51561d62025-08-20T03:12:22ZengMDPI AGISPRS International Journal of Geo-Information2220-99642025-01-011424210.3390/ijgi14020042A Novel Evolutionary Deep Learning Approach for PM<sub>2.5</sub> Prediction Using Remote Sensing and Spatial–Temporal Data: A Case Study of TehranMehrdad Kaveh0Mohammad Saadi Mesgari1Masoud Kaveh2Faculty of Geodesy and Geomatics, K. N. Toosi University of Technology, Tehran 19967-15433, IranFaculty of Geodesy and Geomatics, K. N. Toosi University of Technology, Tehran 19967-15433, IranDepartment of Information and Communication Engineering, Aalto University, 02150 Espoo, FinlandForecasting particulate matter with a diameter of 2.5 μm (PM<sub>2.5</sub>) is critical due to its significant effects on both human health and the environment. While ground-based pollution measurement stations provide highly accurate PM<sub>2.5</sub> data, their limited number and geographic coverage present significant challenges. Recently, the use of aerosol optical depth (AOD) has emerged as a viable alternative for estimating PM<sub>2.5</sub> levels, offering a broader spatial coverage and higher resolution. Concurrently, long short-term memory (LSTM) models have shown considerable promise in enhancing air quality predictions, often outperforming other prediction techniques. To address these challenges, this study leverages geographic information systems (GIS), remote sensing (RS), and a hybrid LSTM architecture to predict PM<sub>2.5</sub> concentrations. Training LSTM models, however, is an NP-hard problem, with gradient-based methods facing limitations such as getting trapped in local minima, high computational costs, and the need for continuous objective functions. To overcome these issues, we propose integrating the novel orchard algorithm (OA) with LSTM to optimize air pollution forecasting. This paper utilizes meteorological data, topographical features, PM<sub>2.5</sub> pollution levels, and satellite imagery from the city of Tehran. Data preparation processes include noise reduction, spatial interpolation, and addressing missing data. The performance of the proposed OA-LSTM model is compared to five advanced machine learning (ML) algorithms. The proposed OA-LSTM model achieved the lowest root mean square error (RMSE) value of 3.01 µg/m<sup>3</sup> and the highest coefficient of determination (<i>R</i><sup>2</sup>) value of 0.88, underscoring its effectiveness compared to other models. This paper employs a binary OA method for sensitivity analysis, optimizing feature selection by minimizing prediction error while retaining critical predictors through a penalty-based objective function. The generated maps reveal higher PM<sub>2.5</sub> concentrations in autumn and winter compared to spring and summer, with northern and central areas showing the highest pollution levels.https://www.mdpi.com/2220-9964/14/2/42PM<sub>2.5</sub>aerosol optical depthgeographic information systemsremote sensinglong short-term memoryorchard algorithm
spellingShingle Mehrdad Kaveh
Mohammad Saadi Mesgari
Masoud Kaveh
A Novel Evolutionary Deep Learning Approach for PM<sub>2.5</sub> Prediction Using Remote Sensing and Spatial–Temporal Data: A Case Study of Tehran
ISPRS International Journal of Geo-Information
PM<sub>2.5</sub>
aerosol optical depth
geographic information systems
remote sensing
long short-term memory
orchard algorithm
title A Novel Evolutionary Deep Learning Approach for PM<sub>2.5</sub> Prediction Using Remote Sensing and Spatial–Temporal Data: A Case Study of Tehran
title_full A Novel Evolutionary Deep Learning Approach for PM<sub>2.5</sub> Prediction Using Remote Sensing and Spatial–Temporal Data: A Case Study of Tehran
title_fullStr A Novel Evolutionary Deep Learning Approach for PM<sub>2.5</sub> Prediction Using Remote Sensing and Spatial–Temporal Data: A Case Study of Tehran
title_full_unstemmed A Novel Evolutionary Deep Learning Approach for PM<sub>2.5</sub> Prediction Using Remote Sensing and Spatial–Temporal Data: A Case Study of Tehran
title_short A Novel Evolutionary Deep Learning Approach for PM<sub>2.5</sub> Prediction Using Remote Sensing and Spatial–Temporal Data: A Case Study of Tehran
title_sort novel evolutionary deep learning approach for pm sub 2 5 sub prediction using remote sensing and spatial temporal data a case study of tehran
topic PM<sub>2.5</sub>
aerosol optical depth
geographic information systems
remote sensing
long short-term memory
orchard algorithm
url https://www.mdpi.com/2220-9964/14/2/42
work_keys_str_mv AT mehrdadkaveh anovelevolutionarydeeplearningapproachforpmsub25subpredictionusingremotesensingandspatialtemporaldataacasestudyoftehran
AT mohammadsaadimesgari anovelevolutionarydeeplearningapproachforpmsub25subpredictionusingremotesensingandspatialtemporaldataacasestudyoftehran
AT masoudkaveh anovelevolutionarydeeplearningapproachforpmsub25subpredictionusingremotesensingandspatialtemporaldataacasestudyoftehran
AT mehrdadkaveh novelevolutionarydeeplearningapproachforpmsub25subpredictionusingremotesensingandspatialtemporaldataacasestudyoftehran
AT mohammadsaadimesgari novelevolutionarydeeplearningapproachforpmsub25subpredictionusingremotesensingandspatialtemporaldataacasestudyoftehran
AT masoudkaveh novelevolutionarydeeplearningapproachforpmsub25subpredictionusingremotesensingandspatialtemporaldataacasestudyoftehran