Study Of Aerosol Optical Depth Retrieval Algorithm For Long Term Aerosol Dataset From 1981-2000 Over China Based On AVHRR

The Advanced Very High Resolution Radiometer（AVHRR）onboard NOAA series satellites is the only sensor that is capable of providing continuous global earth observations over 40 years.Due greatly to lack of near infrared channels and onboard calibration facilities,there are few attempts to develop aerosol optical depth（AOD）retrieval algorithm over land based on AVHRR data.Aiming at building long term AOD dataset over China,an Aerosol optical depth retrieval algorithm over Vegetated Land based on AVHRR（AVL）was developed based on AVHRR data in this study.First,AVHRR GAC（Global Area Coverage）data was calibrated using new released time fitness coefficients for calibration to produce a global AVHRR re-calibration dataset.A strict cloud detection was utilized by combining CLAVR（Clouds from AVHRR-Phase I）and a thresh in 3×3 window for screening cirrus and cloud shadow.Regarding the cloud-free pixels,gas absorption and Rayleigh scattering correction were made using a developed aerosol-surface reflectance coupling calculation model.Second,a surface reflectance estimate model was developed for mainland China.In this model,TOA reflectance at 3.75μm is assumed to be not affected by aerosols.A regression relationship between two channels was developed using MODIS surface reflectance product.Before used to regress the relationship,MODIS data were spectrally transferred to AVHRR channels,diminishing the error caused by different spectral response reflectance.For building a stable regression relationship,NDVI was utilized in regression.However,NDVI was sensitive to aerosols.Thus,a time frame（in previous 30 days of the first half year or next 30 days of the second half year）was used to determine a clearest NDVI.After surface reflectance was estimated,a VZA（Viewing Zenith Angle）correction was made to restore the observation geometry of surface reflectance.Third,in order to validate AVL dataset,sun photometer observations from AERONET,CARSNET as well as BEM（Broadband Extinction Method）AOD retrieve from solar radiation were collected.Satellite AOD products were also collected,including ADL,MODIS,and AVHRR Deep Blue（DB）datasets.Compared to ground-based“true”values,1)ADL generally underestimated AOD,especially over 0.6;2)AVHRR DB can be used for comparing AVL attributing to 51%data points falling into an EE of±(25%τ_BEM+0.05);3)BEM AOD shows high consistency with AERONET and MODIS,being capable of validating AVL AOD as a reference.Fourth,a 20-year AOD dataset for China region（15°～60°E,70°～140°N）was developed using AVL,covering 1981 to 2000.Compared to AVHRR DB,AVL shows high consistency in AOD spatial pattern for daily product,especially in high AOD regions,such as Eastern China,Sichuan Basin,Guanzhong Basin,and northern India.Due partly to data quality control,AVHRR DB often doesn’t include high AOD retrievals over vegetated surfaces.AVL shows obvious overestimation over sparsely vegetated lands in Inner Mongolia and Mongolia.The issues in daily products with each product were magnified in monthly,seasonal and annual AOD.Finally,the skill of AVL in retrieving AOD over China was tested using AVHRR and BEM AOD.Result shows that 61%AVL AOD data points fall into an expected error（EE）of±(25%τ_BEM+0.05)relative to BEM AOD,which is slightly superior to AVHRR DB.A monthly AOD time series comparison of AVL with BEM and AVHRR DB illustrates that AVL is capable of tracing seasonal AOD variations and AOD peaks in summer,which is much better than ADL.While AVHRR DB shows the largest data coverage especially in winter,it commonly underestimated AOD at all radiation sites in winter.