High Resolution Remote Sensing Based Aerosol Optical Depth Retrieval And Its Application Over Urban Areas

With the rapid growth of urbanization and industrialization in China during recent years,eco-environmental problems,especially air pollution issues,are becoming increasingly serious.Ultrahigh concentration of particulate matter and severe haze were reported to frequently occur in mega cites of China with widespread coverage,long duration and highly concentrated pollutant,which has gained worldwide attention.Owing to the low cost,rapid revisit period and wide coverage,satellite remote sensing is an important technique in monitoring the spatio-temporal distribution of air pollutants and their dynamic change.Although numerous satellite-based instruments,such as MOderate Resolution Imaging Spectrometer(MODIS),MISR(Multiangle Imaging SpectroRadiometer),Visible Infrared Imaging Radiometer Suite(VIIRS)and TOMS(Total Ozone Mapping Spectrometer),have provided massive aerosol related products at global scale over the past few decades,global aerosol products usually have a coarse spatial resolution ranging from a few kilometers to several tens of kilometers,most of them cannot cover all land surface types,their inversion algorithms are lack of regional optimization and validation,and the retrieval efficiency is limited under hazy days.On account of the reasons listed above,global aerosol products have significant limitation in the regional scale atmospheric environment related applications.The Gaofen-1(GF-1)satellite,launched by the Chinese government in April 2013,has four Wide-Field-of-View(WFV cameras onboard.Four WFV cameras provide multi-spectral images from visible to near-infrared(NIR)band,with a high spatial resolution of 16 m and a re-visiting period of 4 days.The characteristics of GF-1 WFV data make it very suitable to be applied in the high resolution Aerosol Optical Depth(AOD)retrieval.This thesis is focused on the AOD retrieval algorithm using GF-1 WFV data and the application of retrieved AOD data in the air pollution assessment.The main results achieved in the thesis were listed as follow:1.High resolution AOD retrieval algorithm using GF-1 WFV camera data Determination of regional aerosol types and estimation of surface reflectance are two key issues in the AOD retrieval algorithm.The aerosol types in the research region of Wuhan were determined by statistical analysis of aerosol optical properties derived from long-term CE-318 observations during 2008-2012.By introducing the MODIS AOD products to support the selection of cloudless and clean GF-1 WFV images and conduct the background aerosol correction,and then compositing the clean GF-1 WFV images seasonally by using minimum reflectance technique,high resolution surface reflectance database over Wuhan was finally established.Based on the regional aerosol types and surface reflectance database,combined with the Look Up Table(LUT)generated from 6S model,AOD from all available GF-1 WFV images can be derived.The proposed algorithm can work effectively over all land surface types under cloudless days and even hazy days in Wuhan area,with a spatial resolution of 160m×160m and a temporal resolution of 4 days.GF-1 WFV AOD presented a good relationship with MODIS AOD(R2 = 0.66;RMSE = 0.27)and ground measurements(R2 = 0.80;RMSE=0.25)in the validation,which indicated that the proposed algorithm had a good stability and high accuracy.2.The spatio-temporal distribution pattern of GF-1 WFV AOD over WuhanUtilizing the GF-1 WFV AOD retrievals over Wuhan during 2013-2016(a total of 157 images),we analyzed the spatio-temporal distribution pattern of AOD.According to the different spatio-temporal distribution pattern of localized and transported aerosols,the concept of Local Aerosol Optical Depth(LAOD)was proposed to indicate the local aerosol loading from anthropogenic sources.Annual and seasonal spatio-temporal distribution patterns of anthropogenic aerosol loading were then analyzed using LAOD.Annually,the LAOD over Wuhan peaks at 2013 and 2015,with a regional mean value of 0.42 and 0.40;the lowest LAOD appears at 2016,with an average value of 0.32.Seasonally,the highest LAOD appears in winter with an average value of 0.37,followed by autumn and spring with average value of 0.29 and 0.28.Summer is the cleanest season with an average LAOD of 0.23.The seasonal pattern of LAOD is consistent with the seasonal mean PM2.5 derived from ground measurements at the same period.In addition to the analysis of annual and seasonal patterns,LAOD was further used to identify the highly polluted regions over Wuhan.Regional mean LAOD over downtown and industrial parks were found to be 23%?34%higher than the mean LAOD over the entire Wuhan area,which were the significant sources of air pollution over Wuhan.Moreover,according to the phenomenon that LAOD over urban areas was obviously higher than the LAOD over surrounding areas,the concept of urban aerosol effect(similar to urban heat island effect)was proposed to indicate the contribution of human activities to the regional air pollution,and the difference of LAOD between urban and surrounding areas was used to represent the intensity of urban aerosol effect.The urban aerosol effect intensity was found to present a significant downward trend from 2013 to 2016(-0.009 year-1,P<0.05),which was consistent with the change trend of PM2.5 during the same period(-9.89?g/m3 year-l,P<0.01).3.The estimation of spatial distribution of PM2.5 over Wuhan using GF-1 WFV AOD dataUtilizing the GF-1 WFV AOD and ground-based PM2.5 data over Wuhan during 2013-2016,a Linear Mixed Effect model,which took the daily variation of PM2.5-AOD relationship into account,was established to estimate PM2.5 from GF-1 WFV AOD.The prediction accuracy of LME model was very high(R2=0.939,RMSE=14.0,MRE=11.7%).Meteorological data,including temperature,relative humidity,precipitation and wind speed,were taken into account in the LME model,but their contribution to the model accuracy was not significant.We also considered the site effect of LME model.It was found that prediction accuracy of LME model with site effect(R2=0.95)was slightly high than the LME model without site effect.Two cross validation methods,namely Leave One Out Cross Validation(LOOCV)and 10 fold Cross Validation,were further used to check the model's performance.Compared with the prediction accuracy,the cross validation accuracy showed a slight decline.Validation results demonstrated the stability of the LME model.Finally,the high resolution PM2.5 over Wuhan was estimated from GF-1 WFV AOD data using the LME model.Based on the PM2.5 maps derived from GF-1 WFV AOD data,we further analyzed the spatial distribution pattern of PM2.5 under different degrees of air pollution conditions,and evaluated the spatial distribution of mean PM2.5 during 2013-2016 over Wuhan.