Sorry, you need to enable JavaScript to visit this website.

ATBD Introduction

Introduction

Atmospheric particulates otherwise known as aerosols, have a significant impact on the Earth’s radiative budget, climate change, hydrological processes, and the global carbon, nitrogen and sulfur cycles. To understand the wide-ranging effects of aerosols, it is necessary to measure aerosol characteristics globally with high spatial and temporal resolution. The aerosol retrieval from the polar-orbiting MODerate resolution Imaging Spectrometer (MODIS-Salmonson et al 1989) on board NASA’s Terra (launched in 1999) and Aqua (launched in 2002) satellites has taken these much needed global measurements and turned them into usable data products. Due to its high spatial resolution (as fine as 250 m at nadir), wide swath (2330 km, so that most everywhere on the globe is observed at least once daily), and large spectral range (36 channels between 0.412 to 14.2 μm), the impact of MODIS’ aerosol products has far exceeded nearly everyone’s imagination. Figure 1 shows that the number of publications referencing “MODIS AND AEROSOL” has grown nearly exponentially since launch. Not only have MODIS aerosol products been used to answer scientific questions about radiation and climate (e.g. IPCC, 2013; Yu et al., 2005), but they have also been used for a plethora of applications. Most prominent among them is the monitoring of surface air quality for human health (e.g. Chu et al., 2003, Al-Saadi et al., 2005) .

Figure 1: The image shows the near-exponential rate of publication from the ISI citation web site on "MODIS and aerosol" starting around 1992 when operational algorithms and data centers made the data available to the broad international science community. MODIS aerosol data (Image inspired by Y. Kaufman and ATBD-09).

Aerosol Products

The main aerosol products from MODIS include total spectral ‘Aerosol Optical Depth’ (AOD or τ) and ‘Fine Aerosol Weighting’ (FW or η). In the literature τ is sometimes referred to as ‘Aerosol Optical Thickness’ (AOT) but AOD will be used in this document. The concept of FW is generally defined in the literature as the fraction of optical depth contributed by the fine mode aerosol. In this document, η refers to the fractional contribution of fine (radius between 0.1 and 0.25 μm) aerosol to the total τ, and is reported at a particular wavelength (0.55 μm). The MODIS dark target team continually evaluates the products. One of the evaluation steps is “validation”, where we define an “expected error” envelope (EE) to describe how the products compare with ground-truth observations provided by the AEROsol Robotic NETwork (AERONET). We define a mean EE such that for N “valid” collocations (e.g. retrievals from MODIS and AERONET which are spatially and temporally contiguous), 66% of all MODIS data values are within a ±delta envelope when compared with AERONET values. (Note that there is an assumption here that AERONET retrievals represent the "true" values of AOD.) For example, Levy et al. (2010) provided C5 AOD validation over land using data from 2000 through 2008, and Remer et al. (2008) provided validation over ocean, using data from 2000 to 2007. These papers demonstrated that the C5 AOD product (at 0.55 μm) was comparable to ground truth (e.g. AERONET) with EE envelopes of ±(0.03 + 5%) over ocean and ±(0.05 + 15%) over land, respectively. EE therefore sets out an expectation for the overall performance of the product rather than for an individual retrieval. Although defined for AOD at 0.55 μm, the same EE characterized AOD in other bands. Note that these EE envelopes represent the set of valid MODIS/AERONET matchups, which are not “global”. AERONET site sampling is not global, and neither is MODIS. However, based on the EE definition, these and other validation exercises help determine conditions for which EE is not met. For example, Shi et al. (2011), Kleidman et al. (2012) and Schutgens et al. (2013) noted that C5 MODIS ocean AOD error increases as wind speed increases. Non-spherical dust over the ocean was known to lead to errors in retrieving spectral τ (e.g. Levy et al., 2003). Hyer et al., (2011) found the snow contaminations within the C5 over land products. In general, the over land C5 MODIS retrievals overestimate AOD at brighter and elevated AERONET stations (Levy et al., 2010).

The EE definition also allowed for characterizing changes over time. For example, Levy et al. (2010) showed that while differences between MODIS-Aqua and AERONET did not change over time, differences between MODIS-Terra and AERONET did change. This finding along with other studies led to developing updated calibration for MODIS-Terra for C6.

The other retrieved parameter, aerosol fine weighting (η), was separately compared over land and ocean with sunphotometer data, on multiple scales (e.g. Kleidman et al., 2005; Chu et al, 2005; Anderson et al., 2005). Breon et al. (2011) showed that 53% of MODIS-AERONET collocations over ocean have a fine mode AOD within the total AOD defined uncertainty bounds, which shows a degree of skill for FW over ocean. Levy et al. (2010) found that the MODIS-retrieved FMF over land had too little skill to derive a meaningful EE envelope.

After the validation efforts of the C5 products, the combined algorithm went through a number of iterative revisions leading to the formulation the C6 algorithm, which went into operational production in early 2014, beginning with the MODIS-Aqua instrument. We also describe a new MODIS aerosol product produced at a 3 km resolution using the dark target algorithms in section 6. Although the Deep Blue retrieval (Hsu et al., 2006, Hsu et al., 2013, Sayer et al., 2013) is a component of the MOD04/MYD04 product, it is not detailed in this ATBD. However, a description of a new merged deep blue – dark target variable that is part of the C6 release is given in Appendix A5. As we continue to evaluate and validate the operational C6 products the results will be published on the dark target website (https://darktarget.gsfc.nasa.gov) and within the literature.

Retrieval Algorithm

The MODIS dark target aerosol algorithm is actually comprised of two independent algorithms, one for deriving aerosols over land and the second for aerosols over ocean. Both algorithms were conceived and developed before the launch of Terra and described in depth in Kaufman, et al. (1997b), Tanré, et al. (1997) and ATBD-96. Based on sensitivity tests (small perturbations from theoretical values), these two papers estimated the uncertainty for aerosol retrieval, and defined estimated error (EE) envelopes over each surface. The theoretical basis of the algorithms remained the same until the second generation land algorithm was developed, described in Levy et al. (2007a,b) and ATBD-09. The mechanics and details of the algorithms have continually evolved from the pre-launch algorithm design through the current algorithm. MODIS data is organized by collections. A collection consists of products that were generated by similar, but not necessarily the same versions of the algorithm. The algorithms and products of Collection 4 (C4) are the foundation upon which we have built C6 and were described by Remer et al., (2005). An overhaul to the land algorithm was implemented in Collection 5 (C5), which included surface reflectance parameterizations, aerosol optical models, and assumptions relating to how the 2.11 μm channel relates to surface and surface properties (Levy et al. 2007a, b). The C5 ocean algorithm update is described in Remer et al. (2008). A complete history of changes to the operational algorithm over the course of the MODIS mission can be found at http://modis-atmos.gsfc.nasa.gov/MOD04_L2/history.html, and at the MODIS dark target algorithm website (https://darktarget.gsfc.nasa.gov/). C5, and a modest update called C5.1, was a complete dataset from both Terra and Aqua, spanning from Terra first light in February 2000 through 2015.

The purpose of the remainder of these web pages is to describe in detail the MODIS Collection 6 Aersol retrieval algorithm(s) and to provide the necessary background science and technical information to understand the retreival process.