Satellite And Ground-based FMF Data Fusion And Its Applications In PM_2.5 Remote Sensing Estimation

Fine particulate matter(particulate matter with aerodynamic diameters less than 2.5 ?m,denoted as PM_2.5,hereafter),which have the characteristics of small particle size,long duration of stay in the atmosphere and long transport distance,has significant impact not only on climate change but also on public health.Remote sensing methods can effectively monitor fine particulate matter pollution and regional transport.However,FMF?defined the ratio of fine mode aerosol optical depth to total aerosol optical depth,denoted as FMF,hereafter?used to distinguish anthropogenic aerosols and natural aerosols contributions in remote sensing has low accuracy,failing to meet the need of PM_2.5 remote sensing method?PMRS?.In this study,we merge MODIS FMF and ground-based FMF?AERONET&SONET?to obtain higher estimate accuracy of FMF based on the universal kriging method to provide input data to PMRS,thus improving the estimate accuracy of PM_2.5 near the surface.Firstly,we focus on the retrieval algorithms of FMF and discuss the similarities and differences of the three retrieval algorithms,namely mode combination algorithm?MODIS algorithm?,truncation radius method?AERONET algorithm,FMF@440nm?and spectral identification algorithm?AERONET algorithm,FMF@500nm?).The same mode classification and quite close FMF results on a same aerosol type demonstrate the reasonability of merging MODIS FMF and ground-based FMF,laying the foundation of data fusion based on the universal kriging?denoted as UK,hereafter?method.To obtain the spatiotemporal variability of FMF,variogram analysis is performed on MODIS FMF data from December 2015 to November 2016 over study area.We apply the UK method to merging FMF data in winter of 2016.Leave-one-out cross-validation results show that the maximum deviation between FMF fusion results and ground-based FMF observations at ground-based locations decrease from 0.552 to 0.198,reducing 54% compared with the original error.Then the MODIS FMF and FMF fusion results are applied to PMRS respectively.To validate the PM_2.5 mass concentration near the ground,in-situ measurements from Ministry of Environmental Protection,China are used.Results show that PM_2.5 mass concentration estimated from FMF fusion results is more closed to in-situ measurements.To meet the need of PMRS based on instantaneous remote sensing data,the study performs variogram analysis on MODIS FMF data from December 2010 to November 2016 over the study area and analyzes the seasonal variations of parameters in exponential functions.Aiming to quantify the impact of parameters in exponential variogram function on fusion results thus merging FMF simultaneously,we use the range parameter in winter of 2016?CRT?and the mean value of range parameter of 6 winters over 2011-2016?CMP?as initial values respectively to merge MODIS FMF and ground-based FMF,and then applied the fusion results to estimating PM_2.5 near the surface.Both leave-one-out cross-validation results and comparison results with in-situ PM_2.5 mass concentration measurements in CRT and CMP are closed to each other.It can be concluded that the seasonal average for many years can be a substitute for the same reason since the fusion results are insensitive to range parameter.