During the 1990s, the Medea program brought together environmental scientists and members of the intelligence community to apply classified assets and data to further the understanding of environmental change (Richelson 1998). Under Medea auspices, the global “fiducials” program was established whereby participating scientists could request collection of classified images at environmentally sensitive locations around the globe. The term “fiducials” refers to the fact that the classified images were to be kept “in trust” in classified archives, with the eventual goal of declassification and release to the broader scientific community for research purposes. In 1999, Medea scientists requested that the intelligence community begin collecting images of Arctic sea ice at four different locations in the Arctic Basin during the summer months (the melt season). Two additional locations were added in 2005. The request forwarded by Medea scientists included collections starting in May and ending in September at the approximate locations shown in Figure 2.1. Collection of images during the summer months at these six sites has continued until the present day
The rationale for selecting these sites was as follows:
Beaufort - The Beaufort Sea has been the site of many field studies since the International Geophysical Year 1957/58. The ice in this region is the most studied and best known. It was a focal point for the automatic data buoy program and many studies of the surface heat budget, as well as submarine sonar cross sections.
Canada - This region typically contains the oldest and thickest ice with the longest residence time in the Arctic Basin.
Fram - Fram Strait between Greenland and Spitsbergen is the dominant exit route of sea ice from the Arctic Basin into the Greenland Sea. The amount of low-salinity ice exported is an important component of the basin-wide ice balance and potentially impacts the global ocean circulation.
Siberia - this oceanic region produces most of the first-year ice and was judged to be most sensitive to interannual changes of oceanic and atmospheric forcing. This has been borne out by the extreme negative anomaly of ice extent in the autumn of 2007.
Two additional sites were added after 2005; they are:
Chukchi - While the other sites are in areas with generally thicker perennial ice (with the exception of 2007, when the East Siberian site was ice free in September), the Chukchi site was chosen to sample ice that is seasonal.
Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.
OCR for page 11
2 Potential Uses of the Medea Data Set During the 1990s, the Medea program brought together environmental scientists and members of the intelligence community to apply classified assets and data to further the understanding of environmental change (Richelson 1998). Under Medea auspices, the global “fiducials” program was established whereby participating scientists could request collection of classified images at environmentally sensitive locations around the globe. The term “fiducials” refers to the fact that the classified images were to be kept “in trust” in classified archives, with the eventual goal of declassification and release to the broader scientific community for research purposes. In 1999, Medea scientists requested that the intelligence community begin collecting images of Arctic sea ice at four different locations in the Arctic Basin during the summer months (the melt season). Two additional locations were added in 2005. The request forwarded by Medea scientists included collections starting in May and ending in September at the approximate locations shown in Figure 2.1. Collection of images during the summer months at these six sites has continued until the present day The rationale for selecting these sites was as follows: Beaufort - The Beaufort Sea has been the site of many field studies since the International Geophysical Year 1957/58. The ice in this region is the most studied and best known. It was a focal point for the automatic data buoy program and many studies of the surface heat budget, as well as submarine sonar cross sections. Canada - This region typically contains the oldest and thickest ice with the longest residence time in the Arctic Basin. Fram - Fram Strait between Greenland and Spitsbergen is the dominant exit route of sea ice from the Arctic Basin into the Greenland Sea. The amount of low-salinity ice exported is an important component of the basin-wide ice balance and potentially impacts the global ocean circulation. Siberia - this oceanic region produces most of the first-year ice and was judged to be most sensitive to interannual changes of oceanic and atmospheric forcing. This has been borne out by the extreme negative anomaly of ice extent in the autumn of 2007. Two additional sites were added after 2005; they are: Chukchi - While the other sites are in areas with generally thicker perennial ice (with the exception of 2007, when the East Siberian site was ice free in September), the Chukchi site was chosen to sample ice that is seasonal. 11
OCR for page 11
12 Scientific Value of Arctic Sea Ice Imagery Derived Products FIGURE 2.1 Approximate locations of the image collections requested by Medea scientists in 1999 (red hexagons). The Barrow and Chukchi sites (red squares) were added in 2005. Images were collected at the North Beaufort site (yellow pentagon) only during 1999 and are not part of the dataset considered in this report. SOURCE: Figure courtesy of USGS National Civil Applications Program. Barrow – Barrow is the site of extensive real time monitoring of fast ice by investigators at the University of Alaska and elsewhere; imagery acquired here complements these and other in situ data. Some products have already been derived from these data sets, in particular statistics and maps of melt pond distributions in the Arctic, disseminated by the National Snow and Ice Data Center (NSIDC, 2000; Fetterer et al., 2008) and used by the Arctic research community. However, these products are based only on images taken during 1999-2001. Furthermore, only surface type (e.g., ice or water) maps based on the imagery have been released. The literal imagery itself or a lower-resolution version of it has not been released. In the latter years of the Medea program, procedures were established whereby Literal Imagery Derived Products (LIDPs) could be produced from the classified fiducials data at a resolution deemed suitable for declassification. Several hundred LIDPs with a nominal resolution of 1 meter have been produced from the images collected at the six Arctic sites from 1999 to present. Below we discuss the many potential scientific uses of these LIDPs. USES OF THE LIDPs: SEA ICE PHYSICAL PROCESSES The derived images will lend themselves to a wide range of studies, leading to significant improvements in how sea ice physical processes are represented in climate
OCR for page 11
Potential Uses of the Medea Data Set 13 models. These images will also enable scientists to understand changes in ice habitat. In particular, the images are useful for studying snow distribution and its relationship to surface topography, the initiation and development of meltwater ponds and their profound effect on the surface energy balance and the melting of ice in summer, the relationship of stress and strain rate and its reflection in the pattern of cracks and other discontinuities in the ice, lateral oblation, and thickness evolution. The LIDPs will help scientists better understand these specific physical processes, especially if used in conjunction with data from operational civil satellite systems. The utility of data acquired by remote sensing depends on our ability to interpret them in terms of the actual state of the observed object, i.e. the ground truth. In many cases, understanding the effects of calibrations, and complex radiative properties of the observed object and the properties of the intervening medium, require a plethora of ancillary data sets and ground truth studies. These are particularly difficult to obtain in areas where field (ground-borne) measurements are logistically difficult or sometimes impossible. The one-meter resolution, panchromatic LIDPs are not precisely ground truth. Nevertheless, they offer exceptional details of the surface features compared to images derived from widely available passive and active microwave, and infrared sensors. Such details are exemplified by Figure 2.2, which is a sequence of LIDPs (500 m on a side) showing the transition of melt ponds to open water during the summer of 2006. This committee believes that the great value of the Medea data are their potential to augment the meaning and interpretation of data obtained by other, unclassified, lower-resolution sensors. Below we describe in more detail the physical processes governing the evolution of Arctic sea ice that will be better understood through the LIDPs Snow Distribution Snow depth is an important ingredient in all thermodynamic models of sea ice. Most of the snow on Arctic sea ice falls during the autumn. Redistribution by wind produces an extremely variable snow cover. Smooth ice in frozen leads is often swept bare, while a large fraction of the snow collects in drifts behind aerodynamic obstacles, such as pressure ridges and hummocks. During periods of clear, cold weather, which typically follow precipitation events, the steep temperature gradient in the snow causes an upward diffusion of water vapor that hardens the snow surface and often makes the drifts survive for the whole winter. In the returning daylight of spring, drifts can be identified in 1-meter resolution images. In the absence of any other method to observe the snow cover, the Medea data collection will provide valuable information about the morphology of the snow cover and its interaction with the surface topography and help to improve the interpretation of Ice, Clouds, Land Elevation Satellite (ICESat) laser altimeter records in terms of freeboard, ice thickness, and snow depth. This issue was addressed in a recent paper by Kwok and Cunningham (2008).
OCR for page 11
14 Scientific Value of Arctic Sea Ice Imagery Derived Products 5/26 6/01 6/04 6/12 6/21 7/03 FIGURE 2.2 Development of melt ponds at the Chukchi Site during the summer of 2006 as seen in the 1-m resolution LIDP imagery. Each image in this sequence covers an area of ~500 m by 500 m. The last panel is of open water. SOURCE: Figure courtesy of USGS National Civil Applications Program). Lateral Ablation The loss of multiyear ice (Figure 2.3) may be governed, in part, by lateral melting of ice floes. Open leads with an albedo of less than 0.1 absorb 5-7 times as much solar energy as flat ice, and convection transfers this energy to the side wall of the floes. The possible importance of the process has long been recognized (Steele, 1992), but observations require a field party to spend an entire summer on the ice and make the technically difficult measurements of ablation in many places. In the history of U.S. Arctic research, there have been only four all-summer drifting stations. Sequential one-meter-resolution LIDPs will be capable both of measuring the kinematic shifts of ice floe assemblies and at the same time tracking the surface area of each individual piece of ice. Such observations will help explain the contribution of lateral melting to the loss of multiyear ice.
OCR for page 11
Potential Uses of the Medea Data Set 15 FIGURE 2.3 Imagery from the European Remote Sensing Synthetic Aperture Radar (ERS SAR; left) and the Canadian Space Agency (CSA) Radarsat (right) indicating the reduction in Arctic multiyear ice cover over 15 years. On the image, the thick multiyear ice is bright; the thinner first year ice is dark. The figure shows the great reduction in multiyear sea ice over a 15 year period. Alaska is to the upper left. SOURCE: Figure constructed by Ron Kwok, JPL. Ice Topography and Albedo In the Arctic summer, a strong positive feedback exists between the absorbed downwelling short-wave radiation and the state of the ice surface. The melting snow and ice produce melt ponds whose low albedo further enhances the rate of melting. The meltwater tends to collect on topographically low (i.e. thin) ice. Besides an average thinning of the ice, this process truncates the thickness distribution g(h) at the thin end and produces open water. The close linkages between meltwater pooling and ice surface topography are also key to deriving information about ice albedo from independent estimates of ice roughness or ice age (Eicken et al., 2004), with LIDP images providing a means to improve such indirect derivation of information about ice albedo. There are virtually no sustained, systematic observations of the evolution of the spatial variability in ice albedo because Moderate Resolution Imaging (MODIS) does not have sufficient resolution and Landsat does not have sufficient spatial coverage. The Medea LIDP images will provide an unprecedented view of how the surface topography affects the initial formation and subsequent evolution of melt ponds and their effect on the albedo and hence the short-wave radiative energy balance. Thus the sequential high- resolution pictures will be instrumental in estimating the thermodynamic part of the
OCR for page 11
16 Scientific Value of Arctic Sea Ice Imagery Derived Products changes in g(h) during summer. To obtain the maximum benefit from these images, however, calibration information is needed if it can be supplied. Ice Thickness Evolution Until recently, the only source of information on the thickness distribution has come from occasional transects by submarines with upward looking sonar (Rothrock et al., 2007). It was recently shown by Kwok and Cunningham (2008) that observations from the ICESat laser altimeter can be interpreted to produce ice thickness distributions in numerous places within the Arctic Basin (Figure 2.4). As shown in Figure 1.3, dynamic ice models carry the thickness distribution g(h) as an internal variable, controlled by the energy balance and by the mechanical deformation. In view of the large effect that the albedo has on the computed loss of ice during summer, the models have to assign a certain albedo to the different categories of ice thickness. There are no observational data documenting a relationship between different categories of ice thickness and their albedo throughout the summer. Hence, the designers of models have no choice but to parameterize a relationship between albedo and ice thickness, which makes it a powerful tuning parameter. Ice thickness distributions from ICESat and albedo from the Medea LIDPs should be invaluable to improve these parameterizations. FIGURE 2.4 Sea ice thickness from ICESat. (a) Spatial field of ice thickness from ICESat data acquired over a 35-day period between October and November of 2005 (ON05). (b) Same as (a) but of data acquired in February and March of 2006 (FM06). The start day and duration of each campaign are shown above. (c) Overall ice thickness distributions of the Arctic basin in ON05 (black) and FM05 (red). The quantities in the plot are the mean and standard deviation (in brackets) of the thickness distributions. (d) Thickness distributions of the multiyear sea ice zone. (e) Thickness distributions of the first-year ice zone. SOURCE: Kwok and Cunningham, 2008; Modified by permission of American Geophysical Union.
OCR for page 11
Potential Uses of the Medea Data Set 17 Deformation The realistic representation of the relationship of stress and strain rate of the ice has been a persistent and seemingly intractable problem in the design of dynamic sea ice models (Hibler, 2003). In a recent study, Kwok et al. (2008) use Radarsat to analyze the ice deformation fields on a 10 km scale (Figure 2.5), and compare them to the output of four different dynamic ice models. Given the different physics and forcing functions in the models, the authors cannot state the reasons for the differences between observed and modeled results, but they show that the model outputs are significantly in error, both in terms of the velocity field and ice production. If it were possible to compare the 10-km ice velocity and deformation field with the high-resolution view of the Medea LIDP, we could expect new insight into the processes whose combined effect lead to the relationship of stress and strain rate at the larger scales. Shear and Crack Pattern The Arctic pack ice is crisscrossed by countless cracks and leads with a wide range of sizes. They are related to the wind and water stress fields and their gradients, and sometimes to the land boundaries of the ocean. Synoptic weather systems, eddies in the ocean, and inertial and tidal motions produce discontinuities in the ice on a scale of 100 to 105 meters. When water at the freezing temperature in winter is exposed to a cold sky and cold air, rapid ice growth results. It was shown by Kwok et al. (2003) that the semidiurnal openings and closing caused by tidal and inertial motions could enhance ice production by 10 percent. In summer, the reverse is the case. In either case, civilian satellite images cannot resolve the smaller scale of the spectrum of openings. The 1-meter resolution Medea LIDPs would be of significant benefit to the calculations of the mass balance of the ice cover. Melt Pond Recurrence An issue of importance for the thermodynamic modeling of multiyear ice is the question whether or not melt ponds have a tendency to recur in the same place in consecutive summers. As mentioned above, field observations in summer are sparse. In fact, only one station was maintained by a western country for two consecutive summers (during the IGY, 1957-1958), but the ice break-up in 1958 forced the team to move to a different ice floe nearby. When the surface of melt ponds freezes in autumn, liquid water remains for many weeks under the snow ice cover, acting as a source of latent heat and retarding the formation of new ice at the base. The ability to follow an individual piece of ice from freeze-up throughout the winter until the onset of melting in the following summer would shed light not only at the recurrence of melt ponds but several other processes addressed in the previous sections.
OCR for page 11
18 Scientific Value of Arctic Sea Ice Imagery Derived Products FIGURE 2.5 The top six figures show time-varying fields of ice deformation (magnitude of shear) from November 28, 1999 through January 3, 2000 derived from 100-m resolution RADARSAT Synthetic Aperture Radar (SAR) imagery. These fields cover a large part of the Arctic basin, but not all the details in the fracture patterns are resolved by the 100-m resolution data (Kwok et al., 2008; Modified by permission of American Geophysical Union). Bottom figure: A sample LIDP image (1 m spatial resolution) offers a significant improvement in the resolution of the open- water leads (their width and orientation), and surface morphology beyond that seen in widely available SAR imagery. The square shows the coverage of one SAR pixel. SOURCE: Figure courtesy of USGS National Civil Applications Program.
OCR for page 11
Potential Uses of the Medea Data Set 19 COMPLEMENTING CIVILIAN AND COMMERCIALLY AVAILABLE DATASETS The more recent LIDPs would be extremely useful in assessing the performance of relatively new civilian satellite sensors, particularly the Advanced Microwave Scanning Radiometer for the Earth Observing System (AMSR-E), which began operation in 2002, and ICESat, which began operation in 2003. AMSR-E is a passive microwave sensor and thus suffers from limitations in sea ice retrieval due to surface melt, melt-ponding, and other sub-pixel processes. The effects on passive microwave emissivity from these processes are still not well understood. The LIDPs, with dimensions roughly 15 km x 15 km, are roughly the same scale as the AMSR-E sensor footprint (5-15 km, depending on sensor frequency channel; Figure 2.6). Previous validation campaigns with high- resolution imagery from satellites and aircraft as well as in situ data have helped resolved some of the ambiguities in the passive microwave signature (e.g. Cavalieri et al., 2006; Maslanik et al., 2006). However, the sea ice surface is highly variable in space and time, and these limited campaigns cannot capture that full variability. The numerous scenes that could be released here - covering several years, spanning the full range of the melt season, and encompassing several different geographic regions of different ice regimes (e.g., first-year vs. multiyear ice) - will encompass the full variability of the sea ice passive microwave characteristics. Passive microwave sea ice data are crucial for monitoring the long-term changes in Arctic sea ice because they have a continuous, consistent, and near-complete 30-year record. The released imagery would help improve this long-term record. The imagery will also be very beneficial to ICESat freeboard estimates that are being developed (e.g., Kwok et al., 2007). The imagery will confirm visible surface features revealed in the ICESat freeboard data. For example, surface shadows in the imagery can allow calculation of sea ice ridge freeboard heights, which are generally below the resolution of ICESat. The imagery may also provide snow cover information, an important unknown in ICESat data. The accurate interpretation of lower-resolution visible/infrared data from MODIS will also benefit from this imagery. MODIS, with 500-m resolution, provides a reasonably detailed picture of sea ice conditions. However, the resolution is not fine enough to explicitly capture most meltpond features. There has been some development in calculating meltpond statistics from MODIS imagery (Tschudi et al., 2008), but this high-resolution imagery would be a tremendous help in further refining these efforts. One of the greatest values of the Arctic sea-ice imagery data set lies not only in its high resolution and image quality, but in the availability of imagery with very high repeat rates in all key regions. Given the high probability of cloud cover over summer Arctic sea ice (typically 80 percent or more, Beesley, 1999), any other type of comparable commercial or research-grade imagery would be available at time intervals of at best several weeks. Since these acquisition dates are constrained by the satellite, likely many such acquired images would be unusable due to cloud cover. In contrast, the Arctic IDPs are available at much higher time resolution, in some cases on a daily basis, allowing studies of seasonal progression at a scale hitherto extremely difficult or impossible to achieve. For example, in a previous study one of the committee members worked with specially acquired commercial (SPOT) imagery over a location at roughly 75˚ N between the months of May through September. Only one of the more than 100 scenes was not obscured by clouds and thus could be analyzed in the study. In contrast, on the order of
OCR for page 11
20 Scientific Value of Arctic Sea Ice Imagery Derived Products two dozen scenes are available for a similar time period in a single year at a comparable Medea fiducial Site. FIGURE 2.6 AMSR-E, which began operating in 2002, is a passive microwave sensor. Because the LIDPs are the same scale as AMSR-E, they would be extremely useful in assessing the performance of AMSR-E (Image courtesy of Matt Smith, Information Technology & Systems Center, University of Alabama at Huntsville).