A Simple and Universal Aerosol Retrieval Algorithm for Landsat Series Images Over Complex Surfaces

Abstract

Operational aerosol optical depth (AOD) products are available at coarse spatial resolutions from several to tens of kilometers. These resolutions limit the application of these products for monitoring atmospheric pollutants at the city level. Therefore, a simple, universal, and high‐resolution (30 m) Landsat aerosol retrieval algorithm over complex urban surfaces is developed. The surface reflectance is estimated from a combination of top of atmosphere reflectance at short‐wave infrared (2.22 μm) and Landsat 4–7 surface reflectance climate data records over densely vegetated areas and bright areas. The aerosol type is determined using the historical aerosol optical properties derived from the local urban Aerosol Robotic Network (AERONET) site (Beijing). AERONET ground‐based sun photometer AOD measurements from five sites located in urban and rural areas are obtained to validate the AOD retrievals. Terra MODerate resolution Imaging Spectrometer Collection (C) 6 AOD products (MOD04) including the dark target (DT), the deep blue (DB), and the combined DT and DB (DT&DB) retrievals at 10 km spatial resolution are obtained for comparison purposes. Validation results show that the Landsat AOD retrievals at a 30 m resolution are well correlated with the AERONET AOD measurements (R2 = 0.932) and that approximately 77.46% of the retrievals fall within the expected error with a low mean absolute error of 0.090 and a root‐mean‐square error of 0.126. Comparison results show that Landsat AOD retrievals are overall better and less biased than MOD04 AOD products, indicating that the new algorithm is robust and performs well in AOD retrieval over complex surfaces. The new algorithm can provide continuous and detailed spatial distributions of AOD during both low and high aerosol loadings.

An Improved High‐Spatial‐Resolution Aerosol Retrieval Algorithm for MODIS Images Over Land

“…The spatial locations and site information of each AERONET site are shown in Figure and Tables 2–5 in the supporting information. However, AERONET does not provide AOD measurements at 550 nm; therefore, they are interpolated with the Ångström exponent algorithm based on the available AOD measurements at the two nearest wavelengths among 440, 500, and 675 nm to compare them with the satellite AOD retrievals (Levy, Remer, & Dubovik, ; Sun et al, ; Wei et al, , ; Wei & Sun, ). Table shows a summary of the data sets used in this paper.…”

“…The most critical step is to accurately determine the surface reflectance. Previous studies showed that 1% estimation errors in surface reflectance could lead to approximately 10% errors in aerosol retrieval when the land surface reflectance (LSR) is less than 0.04; when the LSR increases, estimation errors increase by more than 15% with the same 1% inaccurate estimations for surface reflectance (Kaufman, Tanré, et al, ; Wei et al, ). For dark‐target areas (e.g., vegetation and ocean), surface reflectance can be more accurately estimated due to their homogeneous surfaces and low surface‐reflectance characteristics.…”

“…However, for bright surfaces (e.g., urban areas, arid/semiarid areas, and deserts) other than snow/ice, the sensitivity of aerosol change to top‐of‐atmosphere (TOA) reflectance decreases with an increase in the LSR. Additionally, diverse underlying surfaces complicate the accurate estimation of LSR and increase the uncertainty of aerosol retrievals (Hsu et al, ; Li et al, ; Sun et al, ; Wei et al, ).…”

“…MODerate resolution Imaging Spectroradiometer (MODIS) sensors were successfully launched onboard the Terra and Aqua satellites in December 1999 and May 2002, respectively, and the second‐generation operational DT algorithm (for land and ocean; Levy et al, ) and the Enhanced DB algorithm (only for land; Hsu et al, ) have been the main aerosol retrieval algorithms and have produced long‐term and global‐coverage daily DT and DB AOD products at a spatial resolution of 10 km since collection (C) 6 (MOD04_10K; Levy et al, ). MOD04_10K AOD products have been extensively evaluated over land and widely used in studies of atmospheric aerosols from local to global scales (Bilal et al, , ; Levy et al, ; Li et al, ; Wei et al, ; Wei & Sun, ). However, analyses of the spatial distributions and variations of atmospheric pollutants in small‐ and medium‐scale areas are limited due to their coarse spatial resolutions (Bilal et al, ; Li et al, ; Wei et al, ).…”

Improved Aerosol Retrievals Over Complex Regions Using NPP Visible Infrared Imaging Radiometer Suite Observations

“…During past 20 years, a series of sensors have been launched and used in aerosol related researches over land and ocean, such as National Oceanic and Atmospheric Administration Advanced Very High Resolution Radiometer (Geogdzhayev et al, ; Hsu et al, ), Multi‐angle Imaging SpectroRadiometer (Diner et al, ; Limbacher & Kahn, ), Advanced Along Track Scanning Radiometer (Grey et al, ; Guo et al, ; Mei et al, ; Xue et al, ), Polarization and Directionality of the Earth's Reflectance (Deuze et al, ), Sea‐Viewing Wide Field‐Of‐View Sensor (Melin et al, ; Sayer et al, ), Landsat Operational Land Imager (Tian et al, ), MODIS (MODerate‐Resolution Imaging Spectroradiometer; Levy et al, ), VIIRS (Visible Infrared Imaging Radiometer Suite; Jackson et al, ; Su et al, ; Zhang et al, ), and Chinese FengYun Medium Resolution Spectral Imager (Han et al, ; Tong et al, ). One of the main challenges for the AOD retrieval using satellites is to isolate aerosol particle scattering contributions from satellite recorded signals, which is the superposition of atmospheric path reflectance including atmospheric molecules and aerosol matters as well as surface reflectance signals (Wei et al, ). Aerosol optical properties are also key parameters to accurate retrieval of near surface particulate matter, which varies greatly in space and time (Li et al, ; Zheng et al, ).…”