Assessing satellite based PM2.5 estimates against CMAQ model forecasts

摘要

In this work, we focus on estimations of fine particulate matter using MODIS AOD as part of a neural network scheme and compare this to both simple linear regressions and GEOS-CHEM products. In making this comparison it is well known the seasonal and geographical dependences observed in the PM2.5-AOD relationship; thus, to enhance our predictions, we apply WRF PBL information to our neural network method and assess its performance. As part of our analysis, we first explore the baseline effectiveness of AOD and PBL as strong factors in estimating PM2.5 in a local experiment using data collected at one site in New York City. Then, we expand our analysis to a regional domain where daily estimations are obtained based on site location and season. In our local test, we find the high efficiency of the neural network estimations when AOD, PBL and seasonality are primarily assessed (R~0.94 in summer). Later, we test our regional network and compare it with the GEOS-CHEM PM2.5 product. From this, we see better estimations from our experiment using urban/non-urban stations and applying different spatial schemes for training the neural network (R_(NN)~0.80, R_(GEOS-CHEM)~0.57 in an urban station with a distance radius of 0.1 degree; R_(NN)~0.74, R_(GEOS-CHEM)~0.69 in a non-urban station with a distance radius of 0.3 degree). Finally, we create regional daily PM2.5 maps and compare them to GEOS-CHEM outputs, evaluating the corresponding estimations using ground readings.