Deep learning PM2.5 concentrations with bidirectional LSTM RNN

Weitian Tong, Lixin Li, Xiaolu Zhou, Andrew Hamilton, Kai Zhang

Research output: Contribution to journalArticle

Abstract

A better understanding of spatiotemporal distribution of PM2.5 (particulate matter with diameter less than 2.5 micrometer) concentrations in a continuous space-time domain is critical for risk assessment and epidemiologic studies. Existing spatiotemporal interpolation algorithms are usually based on strong assumptions by restricting the interpolation models to the ones with explicit and simple mathematical descriptions, thus neglecting plenty of hidden yet critical influencing factors. In this study, we developed a novel deep-learning-based spatiotemporal interpolation model, which includes the bidirectional Long Short-Term Memory (LSTM) Recurrent Neural Network (RNN) as the main ingredient. Our model is able to take into account both spatial and temporal hidden influencing factors automatically. To the best of our knowledge, it is the first time of applying the bidirectional LSTM RNN in the spatiotemporal interpolation of air pollutants concentrations. We evaluated our novel method using a dataset that contains daily PM2.5 measurements in 2009 over the contiguous southeast region of the USA. Results demonstrate a good performance of our model. We also conducted simulations to explore the properties of spatiotemporal correlations. In particular, we found the temporal correlation is stronger than the spatial correlation.

LanguageEnglish
JournalAir Quality, Atmosphere and Health
DOIs
StateAccepted/In press - Jan 1 2019

Fingerprint

Long-Term Memory
Recurrent neural networks
Short-Term Memory
interpolation
Interpolation
learning
Learning
Air Pollutants
Particulate Matter
Epidemiologic Studies
Risk assessment
particulate matter
risk assessment
Long short-term memory
Deep learning
Air
simulation

Keywords

  • Air pollution
  • Bidirectional LSTM (Long Short-Term Memory)
  • Deep neural network
  • RNN (Recurrent Neural Network)
  • Spatiotemporal interpolation

ASJC Scopus subject areas

  • Pollution
  • Atmospheric Science
  • Management, Monitoring, Policy and Law
  • Health, Toxicology and Mutagenesis

Cite this

Deep learning PM2.5 concentrations with bidirectional LSTM RNN. / Tong, Weitian; Li, Lixin; Zhou, Xiaolu; Hamilton, Andrew; Zhang, Kai.

In: Air Quality, Atmosphere and Health, 01.01.2019.

Research output: Contribution to journalArticle

Tong, Weitian ; Li, Lixin ; Zhou, Xiaolu ; Hamilton, Andrew ; Zhang, Kai. / Deep learning PM2.5 concentrations with bidirectional LSTM RNN. In: Air Quality, Atmosphere and Health. 2019.
@article{9dc87db46bf24c38aaf27bc2a7c0f90e,
title = "Deep learning PM2.5 concentrations with bidirectional LSTM RNN",
abstract = "A better understanding of spatiotemporal distribution of PM2.5 (particulate matter with diameter less than 2.5 micrometer) concentrations in a continuous space-time domain is critical for risk assessment and epidemiologic studies. Existing spatiotemporal interpolation algorithms are usually based on strong assumptions by restricting the interpolation models to the ones with explicit and simple mathematical descriptions, thus neglecting plenty of hidden yet critical influencing factors. In this study, we developed a novel deep-learning-based spatiotemporal interpolation model, which includes the bidirectional Long Short-Term Memory (LSTM) Recurrent Neural Network (RNN) as the main ingredient. Our model is able to take into account both spatial and temporal hidden influencing factors automatically. To the best of our knowledge, it is the first time of applying the bidirectional LSTM RNN in the spatiotemporal interpolation of air pollutants concentrations. We evaluated our novel method using a dataset that contains daily PM2.5 measurements in 2009 over the contiguous southeast region of the USA. Results demonstrate a good performance of our model. We also conducted simulations to explore the properties of spatiotemporal correlations. In particular, we found the temporal correlation is stronger than the spatial correlation.",
keywords = "Air pollution, Bidirectional LSTM (Long Short-Term Memory), Deep neural network, RNN (Recurrent Neural Network), Spatiotemporal interpolation",
author = "Weitian Tong and Lixin Li and Xiaolu Zhou and Andrew Hamilton and Kai Zhang",
year = "2019",
month = "1",
day = "1",
doi = "10.1007/s11869-018-0647-4",
language = "English",
journal = "Air Quality, Atmosphere and Health",
issn = "1873-9318",
publisher = "Springer Netherlands",

}

TY - JOUR

T1 - Deep learning PM2.5 concentrations with bidirectional LSTM RNN

AU - Tong, Weitian

AU - Li, Lixin

AU - Zhou, Xiaolu

AU - Hamilton, Andrew

AU - Zhang, Kai

PY - 2019/1/1

Y1 - 2019/1/1

N2 - A better understanding of spatiotemporal distribution of PM2.5 (particulate matter with diameter less than 2.5 micrometer) concentrations in a continuous space-time domain is critical for risk assessment and epidemiologic studies. Existing spatiotemporal interpolation algorithms are usually based on strong assumptions by restricting the interpolation models to the ones with explicit and simple mathematical descriptions, thus neglecting plenty of hidden yet critical influencing factors. In this study, we developed a novel deep-learning-based spatiotemporal interpolation model, which includes the bidirectional Long Short-Term Memory (LSTM) Recurrent Neural Network (RNN) as the main ingredient. Our model is able to take into account both spatial and temporal hidden influencing factors automatically. To the best of our knowledge, it is the first time of applying the bidirectional LSTM RNN in the spatiotemporal interpolation of air pollutants concentrations. We evaluated our novel method using a dataset that contains daily PM2.5 measurements in 2009 over the contiguous southeast region of the USA. Results demonstrate a good performance of our model. We also conducted simulations to explore the properties of spatiotemporal correlations. In particular, we found the temporal correlation is stronger than the spatial correlation.

AB - A better understanding of spatiotemporal distribution of PM2.5 (particulate matter with diameter less than 2.5 micrometer) concentrations in a continuous space-time domain is critical for risk assessment and epidemiologic studies. Existing spatiotemporal interpolation algorithms are usually based on strong assumptions by restricting the interpolation models to the ones with explicit and simple mathematical descriptions, thus neglecting plenty of hidden yet critical influencing factors. In this study, we developed a novel deep-learning-based spatiotemporal interpolation model, which includes the bidirectional Long Short-Term Memory (LSTM) Recurrent Neural Network (RNN) as the main ingredient. Our model is able to take into account both spatial and temporal hidden influencing factors automatically. To the best of our knowledge, it is the first time of applying the bidirectional LSTM RNN in the spatiotemporal interpolation of air pollutants concentrations. We evaluated our novel method using a dataset that contains daily PM2.5 measurements in 2009 over the contiguous southeast region of the USA. Results demonstrate a good performance of our model. We also conducted simulations to explore the properties of spatiotemporal correlations. In particular, we found the temporal correlation is stronger than the spatial correlation.

KW - Air pollution

KW - Bidirectional LSTM (Long Short-Term Memory)

KW - Deep neural network

KW - RNN (Recurrent Neural Network)

KW - Spatiotemporal interpolation

UR - http://www.scopus.com/inward/record.url?scp=85059734996&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85059734996&partnerID=8YFLogxK

U2 - 10.1007/s11869-018-0647-4

DO - 10.1007/s11869-018-0647-4

M3 - Article

JO - Air Quality, Atmosphere and Health

T2 - Air Quality, Atmosphere and Health

JF - Air Quality, Atmosphere and Health

SN - 1873-9318

ER -