Scientometric and multidimensional contents analysis of PM_2.5 concentration prediction

Area of Science:

• Environmental Science
• Data Science
• Atmospheric Chemistry

Background:

• Precise prediction of fine particulate matter (PM2.5) concentration is crucial for environmental policy and air quality management.
• A comprehensive review of PM2.5 prediction studies over the past two decades (2000-2021) is needed to understand development trends and research hotspots.
• Bibliometric analysis provides a quantitative approach to map the evolution and key themes in PM2.5 prediction research.

Purpose of the Study:

• To conduct a comprehensive and quantitative review of PM2.5 concentration prediction studies from 2000-2021.
• To identify distinct research phases, keyword clusters, and emerging trends in PM2.5 forecasting.
• To outline future research directions for improving PM2.5 prediction accuracy and practical application.

Main Methods:

• Bibliometric analysis using CiteSpace software for visual mapping of research trends.
• Quantitative assessment of PM2.5 prediction studies over a 20-year period.
• Keyword clustering to identify core themes and representative research areas.

Main Results:

• PM2.5 prediction research has evolved through three distinct phases, entering rapid growth after 2017.
• Five major keyword clusters were identified, with forecasting data and methods representing key areas.
• Analysis highlighted the importance of dataset construction, prediction methodologies, and spatial-temporal scale determination.

Conclusions:

• Future research should focus on multi-source data fusion for PM2.5 prediction across various spatial-temporal scales.
• Technological integration and innovative applications in forecasting models are essential.
• Optimizing deep machine learning methods holds significant potential for enhancing prediction accuracy and practical implementation.