Jove
Visualize
Contact Us
JoVE
x logofacebook logolinkedin logoyoutube logo
ABOUT JoVE
OverviewLeadershipBlogJoVE Help Center
AUTHORS
Publishing ProcessEditorial BoardScope & PoliciesPeer ReviewFAQSubmit
LIBRARIANS
TestimonialsSubscriptionsAccessResourcesLibrary Advisory BoardFAQ
RESEARCH
JoVE JournalMethods CollectionsJoVE Encyclopedia of ExperimentsArchive
EDUCATION
JoVE CoreJoVE BusinessJoVE Science EducationJoVE Lab ManualFaculty Resource CenterFaculty Site
Terms & Conditions of Use
Privacy Policy
Policies

Related Experiment Videos

A Spatial-Temporal Interpretable Deep Learning Model for improving interpretability and predictive accuracy of

Xing Yan1, Zhou Zang1, Yize Jiang1

  • 1State Key Laboratory of Remote Sensing Science, College of Global Change and Earth System Science, Beijing Normal University, Beijing, 100875, China.

Environmental Pollution (Barking, Essex : 1987)
|January 19, 2021
PubMed
Summary

Related Concept Videos

You might also read

Related Articles

Articles linked to this work by shared authors, journal, and citation graph.

Sort by
Same author

Stratospheric polar vortex shapes Arctic surface climate via a radiative pathway.

Nature communications·2026
Same author

PGMNO: A physics-Guided mamba neural operator framework for partial differential equations.

Neural networks : the official journal of the International Neural Network Society·2026
Same author

K-nearest neighbors approximation for the multiwavelength lidar retrieval of bimodal aerosol microphysical properties.

Applied optics·2026
Same author

Exploring Global Aerosol Size Dynamics from 2001 to 2024 Using the Pretrained Remote sensing pIxel-based Spatial-teMporal Deep Neural Network.

Environmental science & technology·2026
Same author

Climate impacts and future trends of hailstorms in China based on millennial records.

Nature communications·2025
Same author

Revolutionizing Satellite Real-Time Air Pollution Alerts through New On-Orbit System-on-Chip Technology.

Environmental science & technology·2025

A new Spatial-Temporal Interpretable Deep Learning Model (SIDLM) enhances satellite-based PM2.5 monitoring accuracy and interpretability. This advanced model offers improved predictions for air quality management and scientific research.

Area of Science:

  • Environmental Science
  • Remote Sensing
  • Machine Learning

Background:

  • Accurate monitoring of fine particulate matter (PM2.5) is crucial for understanding and mitigating air pollution.
  • Satellite-based remote sensing offers a promising approach for PM2.5 monitoring, but existing machine learning techniques lack interpretability and predictive precision.

Purpose of the Study:

  • To develop and evaluate a novel Spatial-Temporal Interpretable Deep Learning Model (SIDLM) for enhanced satellite-based PM2.5 measurements.
  • To improve both the interpretability and predictive accuracy of PM2.5 estimations derived from satellite data.

Main Methods:

  • Development of a 'wide' and 'deep' SIDLM architecture.
  • Comprehensive evaluation using diverse input data (top-of-atmosphere measurements, aerosol optical depth, meteorological data) and spatial resolutions (10 km, 3 km, 250 m) across China.
Keywords:
Deep learningInterpretabilityMODISPM(2.5)

Related Experiment Videos

  • Comparison with five existing machine learning inversion methods.

    • Main Results:

    • The top-of-atmosphere-based SIDLM achieved the highest predictive accuracy in China, with root-mean-square errors as low as 15.30 μg/m³ and R² values up to 0.70 at 10 km resolution.
    • Successful PM2.5 retrieval at a 250 m resolution over Beijing (RMSE = 16.01 μg/m³, R² = 0.62).
    • SIDLM outperformed other methods in accuracy, feature extraction, and interpretability, identifying key influencing districts and seasonal patterns of PM2.5 accumulation.

    Conclusions:

    • The SIDLM offers superior performance for satellite-based PM2.5 monitoring compared to traditional methods.
    • The model's interpretability provides valuable insights into PM2.5 influencing factors and spatiotemporal dynamics.
    • SIDLM is a promising tool for Earth observation data analysis, deep learning predictions, and spatiotemporal research.