Reassessing the Reliability of WRF-CMAQ-Based PM_2.5 Prediction from a Generalization Perspective

Methane is a potent greenhouse gas, yet existing satellite and ground observations in China are too sparse to accurately quantify emissions, highlighting the need for an expanded ground network. Using Bayesian inversion and simulated annealing, we identify optimal locations for new sites. Results show that prioritizing stations in eastern and southwestern China would be most effective. Expanding the network with 50 stations could nearly double the current emission constraint capability.

Black carbon's climate impact is highly uncertain because its radiative effect depends on particle size, mixing state, and column loading.  This study combines satellite data, physics-based simulations, and machine learning to estimate black carbon forcing across contrasting regions. The same black carbon amount can warm or cool the atmosphere depending on local aerosol properties. The machine learning framework provides a fast, transferable tool for regional climate assessment.

There are always gaps in satellite records of air pollution because of factors such as clouds and operating conditions. We used a physics-aware deep learning method, producing a gapless 0.05° hourly dataset of tropospheric nitrogen dioxide over Asian pollution hotspots from 2019 to 2024. The dataset can capture sub-daily changes, revealing how traffic, industry, and clean-air policies shaped pollution, supporting health and environment studies.

This study examines how black carbon particles from the coal industry influence regional climate by absorbing sunlight. Based on ground measurements and modeling, we find that conventional approaches, which oversimplify particle size and structure, which could leading a bias in terms of their warming effect. Our results highlight that more realistic particle characterizations are crucial for improving climate predictions in industrial regions.

This study provides TROPOMI (TROPOspheric Monitoring Instrument) with a new methane emission estimation method that can accurately identify emission sources. Our results generate non-negative emission datasets using objective selection and filtering methods. The results include lower minimum emission thresholds for all power grids and fewer false positives. The new method provides more robust emission quantification in the face of data uncertainty, going beyond traditional plume identification and background subtraction.