Quantitative aerosol retrieval from space grew out of the need for atmospheric correction, that is the process of removing atmospheric effects on satellite remote sensing of earth's surface. Important satellite measures of land surface vegetation and ocean color were confused because of changing aerosol above the scene. The physical concept is that over "dark" surfaces (vegetated land and dark ocean), aerosol appears to brighten the scene, and the by-product of atmospheric correction is the retrieval of aerosol properties. As the scientific community became interested in quantifying the aerosol (e.g. for climate purposes), the dark target concept was refined. This Dark-Target (DT) algorithm was originally developed by Tanre et al., (1997) and Kaufman et al., (1997), has evolved over time, and is currently operational on the MODIS satellite instrument. Following is a short "history".
Yoram Kaufman of NASA Goddard and Didier Tanré of the Laboratoire d’Optique Atmospherique at the University of Lille were two young scientists who were already close collaborators and good friends in 1988 when NASA announced the formation of a MODIS Science Team. Both Yoram and Didier saw opportunity to produce the first comprehensive global data set of aerosol products that would help narrow the uncertainties in estimating climate change. Both wrote and submitted proposals articulating this vision, making them competitors for the same “aerosol seat” on the Science Team. Amazingly both proposals were selected, allowing them to work together. The first step that Yoram and Didier took in their effort was to divide the world. Didier would take the over ocean retrieval and Yoram would take the over land retrieval. Didier teased Yoram telling him, “My part is more important because I cover ¾ of the Earth.” Yoram replied, “You can have the ocean. Nobody lives there, so who cares.” Thus, began a work and partnership that they would continue together until Yoram’s untimely death in a bicycle accident in 2006.
Yoram and Didier were not working in a vacuum. There had been a history of deriving aerosol from space sensors. These had started with Landsat satellites and migrated to AVHRR. The early algorithms all assumed values for surface reflectance and aerosol optical properties, modeled the expected reflectance at top of the atmosphere for a variety of aerosol loadings and then backed out the aerosol optical depth by comparing the actual reflectances to those that were modeled. This was done one wavelength channel at a time. Because it was much easier to constrain the assumption of surface reflectance when surface reflectance was low, those algorithms became known as “dark target” algorithms.
In principle, Didier had the easier problem because the ocean away from glint was always dark and its reflectance was well-characterized unless affected by chlorophyll absorption by phytoplankton. Glint could be masked from geometrical considerations, and wavelengths affected by chlorophyll could be avoided. Retrieving the aerosol optical depth (AOD) at a channel from the red through the near-infrared was a fairly straightforward proposition. The NOAA team had begun to produce an AOD from AVHRR channel 1 (660 nm) over the global ocean by the mid 1990s. Didier knew that MODIS was better designed for aerosol retrieval than AVHRR and he wanted to push the MODIS retrieval to beyond simple AOD, but how much information was really available in MODIS spectral measurements? Didier used principal component analysis (Tanré et al., 1996) to identify that the 6 channels of MODIS resulted in only 2.5 pieces of information. He could retrieve AOD and information about size, but not the specifics of the size distribution and nothing about particle shape or absorption. There would be two modes in the aerosol model and each mode would have some flexibility, but the main free parameters would be the aerosol loading and the relative size of the two modes.
Meanwhile, Yoram was working on the more difficult problem of deriving aerosol quantities over land, where the problem was characterizing a highly variable surface. There had to be a way to characterize surface reflectance without a priori knowledge. In the mid-infrared, specifically channel 3 (3.75 µm) on AVHRR, fine mode particles were invisible. The surface looked the same in this channel no matter the aerosol loading above it, which meant he could estimate surface reflectance in the visible using an empirical relationship. Finally, he could derive the AOD above this characterized surface in visible wavelengths. Field experiments provided the data set to derive the empirical relationship and experience shifted the mid-IR channel from 3.75 µm to 2.1 µm.
There was a third scientist who was working in parallel with Yoram and Didier during this time. Brent Holben of NASA Goddard organized a network of automated sun and sky scanning radiometers that would take the place of the feeble hand held devices that had been standard ground-based AOD instruments for decades. As the MODIS aerosol concept progressed, Brent was able to install small networks of these radiometers during field experiments in Brazil and along the U.S. east coast in 1992 and 1993, respectively. These small networks became the nucleus that has grown into the well-established AERONET federated system of today. Brent’s early networks provided measurements of sky radiance that could be inverted into volume size distribution (Nakajima and King, XXXX) and a variety of other aerosol optical parameters (Dubovik and King, XXXX). These inverted products helped to define the aerosol assumptions in the MODIS DT retrievals, and the measured AOD could be used for validation.
The airborne imagery that provided the test bed for the Dark Target proto-algorithms was obtained from the MODIS Airborne Simulator (MAS) and A Visible InfraRed Spectrometer (AVIRIS) flying on board NASA’s ER-2 during specific NASA sponsored field campaigns. These campaigns included the Sulfate/Smoke Clouds Aerosol and Radiation (SCAR) experiments in the mid-Atlantic, in California and in Brazil, and the Tropospheric Aerosol Radiative Forcing Observational eXperiment (TARFOX) that spanned the years 1993 – 1997. These field campaigns produced data sets used in algorithm development for a variety of MODIS products besides aerosols, and also used for direct scientific inquiry that advanced our understanding of aerosols, clouds, radiation, surface properties and fires.
Right from the beginning a team was formed in support of Didier and Yoram’s efforts. Shana Mattoo who had been working for Yoram previously became the lead programmer for the algorithm, delivering tens of thousands of lines of code for operational processing. Lorraine Remer signed on early in the process to support Yoram’s science efforts and to help manage the group as it grew. Bo-Cai Gao, Richard Kleidman, Rong-Rong Li, Allen Chu and Andrew Wald were all support scientists who were involved in developing the algorithm from the earliest stages. As we neared Terra launch Robert Levy, Charles Ichoku and Vanderlei Martins joined what had begun to be self-identified as the MODIS aerosol group. The MODIS aerosol group was characterized by the joy they found in their work and the fun to be had in working together.
Then at the end of 1999, after a decade of preparation and development, Terra launched with MODIS on board and the MODIS aerosol algorithm was put to test in an operational environment for the first time. There was a flurry of excitement in the first year as the products and algorithm were evaluated. Eric Vermote, PI of the land atmospheric correction algorithm worked closely with the group during the first half year until the initial validation looked promising. Then Eric split off to fine-tune the basic Dark Target concepts for his community’s needs. Early on the MODIS aerosol group discovered that the standard MODIS cloud mask had to be abandoned and a cloud mask specific to the aerosol algorithm developed and implemented by Vanderlei Martins. Within 2 years, with mostly peripheral modifications to the ‘at launch’ code, the group declared the MODIS aerosol product to be ‘fully validated’ within specific uncertainty bounds, with isolated lingering problems defined.
After the algorithm had matured and was applied to both Terra and Aqua, Yoram and Didier’s attention began to drift to other scientific problems. Didier became immersed in retrieving aerosol from the French POLDER satellite sensors and in creating a data center at his institution at the Laboratoire d’Optique Atmospherique. Yoram began to use the MODIS products to address geophysical questions such as intercontinental aerosol transport, aerosol-cloud interaction and aerosol fertilization of ocean ecosystems. In 2003, Yoram handed leadership of the MODIS aerosol group to Lorraine Remer, and in his words, “to become a data user, not a data producer”. For the next two years Yoram enjoyed some of the most productive and creative years of his career, often collaborating with a young post-doc named Ilan Koren.
Tragically, we lost Yoram’s creative energy and guidance in 2006 when he died from injuries acquired in a bicycle accident just outside the gates of the NASA Goddard campus. The entire global community of aerosol and remote sensing scientists were shaken to their core upon his death. NASA’s Earth Observing satellite sensors dropped a minute of data collection in salute to an individual who had been instrumental in not only the MODIS aerosol algorithms and products, but in NASA’s Earth Science mission as a whole.
Lorraine Remer continued leading the MODIS aerosol group after Yoram’s death. Under Lorraine’s leadership the Dark Target over land algorithm was significantly modified by Robert Levy who evaluated and updated each assumption and abandoned the single channel retrieval in favor of a simultaneous multi-channel inversion. Other modifications included the development of a 3 km product, a wind speed dependence Look Up Table for the ocean retrieval and other small changes. In 2012, Lorraine stepped away from the MODIS aerosol team and Robert Levy took over the leadership.
Today the MODIS aerosol team is tackling the challenges of aging sensors and calibration drift. Led by Robert Levy and consisting of Shana Mattoo as lead programmer, and Leigh Munchak, Falguni Patadia, Pawan Gupta and Richard Kleidman as support scientists, the team is dedicated to maintaining the high level of quality of the original product to the point that the 15-year MODIS aerosol product can be used as a Climate Data Record, with no artificial trends and well-constrained uncertainties. This is a higher standard than originally envisioned by Yoram and Didier in the 1990s. Also unanticipated by Yoram and Didier is the adaption of the product for air quality applications. This has increased the pressure on the MODIS aerosol team to produce a product with finer resolution and greater accuracy in urban locations.
In the future the MODIS aerosol team is transforming into the Dark Target aerosol team. The Dark Target algorithm used successfully on MODIS for 15 years is being applied to Suomi-NPP VIIRS and to a wide variety of aircraft imagers such as AMI and eMAS. A consistent well-understood algorithm, with a consistent set of aerosol products, that crosses platforms allows for intercomparison between sensors and continuation of the Climate Data Record begun by MODIS. As long as broad spectrum imaging radiometers are flying, the Dark Target algorithm conceived by a pair of friends in the 1980s, will be working to bring to the community the familiar and reliable aerosol characterization that has become a community benchmark.