Retrieving near‐global aerosol loading over land and ocean from AVHRR

Abstract

The spaceborne advanced very high resolution radiometer (AVHRR) sensor data record is approaching 40 years, providing a crucial asset for studying long‐term trends of aerosol properties regionally and globally. However, due to limitations of its channels' information content, aerosol optical depth (AOD) data from AVHRR over land are still largely lacking. In this paper, we describe a new physics‐based algorithm to retrieve aerosol loading over both land and ocean from AVHRR for the first time. The over‐land algorithm is an extension of our Sea‐viewing Wide Field‐of‐view Sensor and Moderate Resolution Imaging Spectroradiometer (MODIS) Deep Blue algorithm, while a simplified version of our Satellite Ocean Aerosol Retrieval algorithm is used over ocean. We compare retrieved AVHRR AOD with that from MODIS on a daily and seasonal basis and find, in general, good agreement between the two. For the satellites with equatorial crossing times within 2 h of solar noon, the spatial coverage of the AVHRR aerosol product is comparable to that of MODIS, except over very bright arid regions (such as the Sahara), where the underlying surface reflectance at 630 nm reaches the critical surface reflectance. Based upon comparisons of the AVHRR AOD against Aerosol Robotic Network data, preliminary results indicate that the expected error confidence interval envelope is around ±(0.03 + 15%) over ocean and ±(0.05 + 25%) over land for this first version of the AVHRR aerosol products. Consequently, these new AVHRR aerosol products can contribute important building blocks for constructing a consistent long‐term data record for climate studies.

Evaluation of NASA Deep Blue/SOAR aerosol retrieval algorithms applied to AVHRR measurements

“…The adaptation of the DB and SOAR algorithms to the AVHRRs is described by Hsu et al []. The physical principles behind the AVHRR application of the algorithms are the same as those behind the SeaWiFS, MODIS, and VIIRS versions.…”

“…Once the surface reflectance has been obtained, band 1 AOD is retrieved directly from the AVHRR measurement. The AOD at 550 nm is extrapolated from this using an assumed (regionally and seasonally dependent) Ångström exponent (AE) based on AERONET climatologies [ Hsu et al , ]. Thus, both 550 nm and band 1 AOD are provided within the data set, even though AVHRR has no band near 550 nm.…”

“…Each retrieval also has an associated quality assurance (QA) flag between 1 and 3. QA = 1 (“poor”) indicates that internal tests [ Hsu et al , ] suggest some potential problem, such as cloud contamination or an improper surface model, so the retrieval is likely to be quantitatively less reliable. These retrievals should not be used for most applications.…”

“…An expected error (EE) envelope is defined, intended to represent a one standard deviation confidence envelope around the retrieved AOD [e.g., Sayer et al , ], such that one standard deviation of retrievals (i.e., about 68%) match the ground truth AOD to within this level (and, following Gaussian statistics, approximately 95% within twice the EE envelope, etc.). For the initial AVHRR Deep Blue data set, the EE is taken [ Hsu et al , ] as ±(0.03 + 15%) over water and ±(0.05 + 25%) over land (with AOD defined relative to the Sun photometer values, i.e., a diagnostic rather than prognostic measure), for all wavelengths considered. These may be refined further in the future.…”

“…Additional information and documentation is provided at https://deepblue.gsfc.nasa.gov. The specific implementation of these algorithms to the AVHRRs is described in a companion paper, Hsu et al []. Note that the data set is referred to as “AVHRR Deep Blue,” although it is composed of both the distinct DB and SOAR algorithms.…”