Multiscale applicability assessment of PM_2.5 datasets in Chinese urban agglomerations: Accuracy, spatiotemporal...

Various high-resolution PM_2.5 datasets have emerged with the improvement of satellite remote sensing techniques. However, the current performance evaluation of these datasets remains insufficient, hindering progress in urban pollution research. This study presents a comprehensive multiscale evaluation of 5 PM_2.5 datasets-including the China High-resolution Air Pollutants (CHAP), Full-cover High-resolution Air Pollutant, Long-term Gap-free High-resolution Air Pollutant, Satellite-derived Air Pollution, and Tracking Air Pollution in China (TAP) -across ten major Chinese urban agglomerations spanning a 20-year period (2000-2020). Our four-phase analysis included: (1) dataset validation against in-situ measurements, (2) pixel-level analysis of intra-urban variations, (3) dataset comparison at the urban-agglomeration scale, and (4) uncertainty quantification using the three-cornered hat method. The results revealed that all datasets effectively captured PM_2.5 variations across the ten urban agglomerations, but that significant variability is present in metrics among these datasets. Notably, a relatively high proportion of opposite trends were observed in intra-urban agglomerations, especially on the Northern Slope of Tian Shan urban agglomeration. Generally, CHAP demonstrated superior performance compared with the other datasets, exhibiting the highest correlation coefficient and the lowest root-mean-square deviation. An uncertainty analysis showed that CHAP performed the best, with over 50 % of the data from China exhibiting minimal uncertainty (<15 μg/m³), especially in the Pearl River Delta (as low as 5 μg/m³). In contrast, TAP showed the highest uncertainty (>15 μg/m³), covering only 5 % of China with low uncertainty and performing the worst in the Huaihai economic region. It was found that site density, spatial heterogeneity, model input, and inversion algorithms significantly affect dataset accuracy and consistency.