Superior PM2.5 Estimation by Integrating Aerosol Fine Mode Data from the Himawari-8 Satellite in Deep and Classical Machine Learning Models

Artificial intelligence is widely applied to estimate ground-level fine particulate matter (PM2.5) from satellite data by constructing the relationship between aerosol optical thickness (AOT) and surface PM2.5 concentration. However, size properties, such as mode fraction (FMF), are rarely considered in satellite-based modeling, especially machine learning models. This study investigated linear non-linear relationships AOT (fAOT) over five AERONET stations China (Beijing, Baotou, Taihu, Xianghe, Xuzhou) using fAOT 5-year (2015–2019) data. Results showed that separated FMF (fAOT = × FMF) had significant with PM2.5. Then, Himawari-8 V3.0 V2.1 (FMF&AOT-PM2.5) were tested input a deep model four classical The results FMF&AOT-PM2.5 performed better than (AOT-PM2.5) modelling estimations. was then retrieval during 2020, found have agreement AOT-PM2.5 on dust haze days. correlation both days (dust days: R 0.82; 0.56) compared 0.72; 0.52) partly contributed superior accuracy of FMF&AOT-PM2.5. demonstrates importance including improve estimations emphasizes need for more accurate product enables retrieval.