Impacts of snow and cloud covers on satellite-derived PM2.5 levels

Abstract

Satellite aerosol optical depth (AOD) has been widely employed to evaluate ground fine particle (PM_2.5) levels, whereas snow/cloud covers often lead to a large proportion of non-random missing AOD values. As a result, the fully covered and unbiased PM_2.5 estimates will be hard to generate. Among the current approaches to deal with the data gap issue, few have considered the cloud-AOD relationship and none of them have considered the snow-AOD relationship. This study examined the impacts of snow and cloud covers on AOD and PM_2.5 and made full- coverage PM_2.5 predictions by considering these impacts. To estimate missing AOD values, daily gap-filling models with snow/cloud fractions and meteorological covariates were developed using the random forest algorithm. By using these models in New York State, a daily AOD data set with a 1-km resolution was generated with a complete coverage. The "out-of-bag" R² of the gap-filling models averaged 0.93 with an interquartile range from 0.90 to 0.95. Subsequently, a random forest-based PM_2.5 prediction model with the gap-filled AOD and covariates was built to predict fully covered PM_2.5 estimates. A ten-fold cross-validation for the prediction model showed a good performance with an R² of 0.82. In the gap-filling models, the snow fraction was of higher significance to the snow season compared with the rest of the year. The prediction models fitted with/without the snow fraction also suggested the discernible changes in PM_2.5 patterns, further confirming the significance of this parameter. Compared with the methods without considering snow and cloud covers, our PM_2.5 prediction surfaces showed more spatial details and reflected small-scale terrain-driven PM_2.5 patterns. The proposed methods can be generalized to the areas with extensive snow/cloud covers and large proportions of missing satellite AOD data for predicting PM_2.5 levels with high resolutions and complete coverage.

Keywords: AOD; Cloud Cover; Gap-filling; MAIAC; PM2.5; Random Forest; Snow Cover.

Fig. 1

Study areas. Latitude: [40.1°N, 45.6°N]; Longitude: [80.5°W, 71°W]. The area outside New York State served as the buffer. Red diamonds are EPA AQS stations. Blue triangles are NAPS stations.

Fig. 2

AOD spatial distributions in 2015: (a) annual distribution of Aqua AOD (original and gap-filled AOD); (b) annual distribution of gap-filled Aqua AOD; (c) annual distribution of Terra AOD (original and gap-filled AOD); (d) annual distribution of gap-filled Terra AOD; (e) differences between gap-filled and original Aqua AOD (gap-filled minus original AOD); (f) differences between gap-filled and original Terra AOD (gap-filled minus original AOD).

Fig. 3

Random forest modeling performance: (a) 10-fold cross-validation (CV) scatters with an R² of 0.82 and an RMSE of 2.16 μg/m³; (b) variable importance ranking. The PM_2.5 convolutional layer had the highest value and five of the top-seven important variables were land-use terms.

Fig. 4

PM_2.5 spatial distributions in 2015: (a) annual distribution of PM_2.5 with a 1-km resolution by Equation 2; (b) gap-filled PM_2.5 by Equation 3; (c) differences between full-model and no-AOD PM_2.5 in the snow season (full-model minus no-AOD PM_2.5); (d) differences between full-model and cloud-only PM_2.5 in the snow season (full-model minus cloud-only PM_2.5).

Fig. 5

PM_2.5 accumulation effect in the valleys of Upstate New York in winter. The areas are on the border of Chenango County and Otsego County (Latitude: [42.375°N, 42.75°N]; Longitude: [75.5°W, 75.15°W]). The widths of the valleys are ~3 km.