2.1 Data streams

For large-scale climate-related studies the snow-cover and sea-ice database can be conveniently divided into long-term (historical) data sets, shorter term data sets derived primarily from satellite analysis that exist in map form, and those data that are available in digital format. Recent discussions of snow-cover and sea-ice data have been presented by Barry (Reference Barry and Untersteinerin press [a], Reference Barry and Hecht[b]) and some discussion of cryospheric data in relation to climate modeling has been given by Reference Kotlyakov, Krenke and EaglesonKotlyakov and Krenke (1982). Both sea-ice and snow-cover data are reported by synoptic observation, aerial reconnaissance, satellite and occasionally, for sea ice, by submarine remote sensing. The satellite data are primarily from scanning visible and infrared (IR) radiometers and single or multichannel microwave radiometers. While the satellite data are received in digital form, they are usually processed to give some form of imagery, and are then transferred to chart format (Figs.1 and 2). Much attention is now being given to the digitization of such charts. An exception to the data interpretation-digitization process is found in the case of sea-ice information derived from the microwave radiometers. These data are presented in the form of gridded brightness temperatures and algorithms are available to convert these into ice concentrations, which are then directly available in a digital format.

Fig. 1

Snow-cover data streams.

Fig. 2

Sea-ice data streams (Reference Barry and UntersteinerBarry in press [a]).

2.2 Historical data

Snow depth is an indicator of storm frequency, moisture availability, and local factors such as redistribution by wind, compaction, and ablation. The synoptic code has provision for measurements of depth of snow on the ground. These data are collected daily by synoptic stations and reported through the global telecommunications network. These records are not easily retrieved from the high volume synoptic meteorological data archives, but some convenient extracts are available in map form (e.g. US Department of Agriculture 1935-onwards).

The water content of the snowpack is commonly measured by ground surveys for streamflow prediction and flood forecasting. Such data are collected twice-monthly in the United States of America by the Soil Conservation Service on a state-by-state basis. More frequent snowpack data are now available for the western states through the snow telemetry (SNOTEL) network. Data are obtained from approximately 500 remote automatic stations via ionospheric meteor burst telemetry to a central site in Portland, Oregon (Reference Barton and BurkeBarton and Burke 1979). For Canada, mean snow depth and water equivalent are published for approximately 1 300 snow courses (Canada Atmospheric Environment Service 1961-onwards). Other regional snow-cover data sets have been produced for Austria (Hydrographischer Dienst in Österreich 1962), Great Britain (Reference CraneCrane 1979) and the Soviet Union (Reference CraneCrane 1979). Snow-cover data are also collected in many mountain ranges in connection with avalanche prediction and hydroelectric programs, as well as for programs of winter orographic cloud seeding. These records, however, are not usually available in accessible archives.

Some historical sea-ice records were mentioned in the introduction. Monthly ice charts for April to the end of the ice season (August or September) for the years 1901 to 1939 and 1946 to 1956 have also been published for the Arctic Ocean by the Dansk Meteorologisk Institut (1901(?)–1959). These data are available in digital form from 1901 (Reference KellyKelly 1979). Several regional charts are available based usually on mixed satellite, aircraft and some surface observations. Such local data sets are produced by Canada, Norway, Finland, West Germany, Sweden, and Japan. Ice charts for the Soviet Arctic have been prepared since 1937 (Reference CraneCrane 1979). Little information is known, however, concerning the availability of the Soviet products.

2.3 Satellite data

Satellite data have been available since the early 1960s. Sensors and analysis systems, however, have been continuously upgraded since this time. Consistent coverage has only been available since the early 1970s with the launch of the National Oceanic and Atmospheric Administration (NOAA) series of satellites, and with the introduction of a passive microwave radiometer on board Nimbus-5. Being unaffected by clouds or the polar night, this sensor allowed, for the first time, the possibility of continuous all-weather synoptic analyses of sea-ice distribution.

The National Environmental Satellite Data and Information Service (NESDIS) has mapped northern hemisphere snow and ice boundaries since 1966. Three classes of surface reflectivity are mapped on a weekly basis. Before 1970 the analysis was based on visible data from the Advanced Vidicon Camera System (AVCS) and the Automatic Picture Transmissions (APT) on the ESSA satellites. Thermal IR data were added with the introduction of the scanning radiometer data from ITOS-1, and increased resolution was obtained from the Very High Resolution Radiometer (VHRR) with the launch of NOAA-2 in October 1972. The most consistent data are those obtained from 1974 onwards (Reference Kukla and RobinsonKukla and Robinson 1981). All of the weekly snow and ice charts from 1966 to 1980, as well as monthly summaries, are now available in a digital format (Reference DeweyDewey 1981).

NOAA VHRR data have also been used by the US Navy to map sea-ice extent and concentration since 1972 for the northern hemisphere, and from 1973 for the southern hemisphere. Electrically Scanning Microwave Radiometer (ESMR) data from Nimbus 5 and 6 were routinely incorporated into the analyses from December 1972 to September 1976. Since October 1977 DMSP visible and IR data have also been used when available. Production of these charts was taken over by the US Navy-NOAA Joint Ice Center in June 1979.

The microwave radiometer data from Nimbus 5 and 6 have also been used experimentally to map continental snow cover (Reference Hall, Foster, Chang and RangoHall and others 1979). Techniques are being developed to use both passive and active microwave data to estimate snow depth or snow water equivalent, wetness and snowpack structure (Reference RangoRango 1980). In this regard, research is presently being conducted using Scanning Multichannel Microwave Radiometer (SMMR) data from Numbus 7.

Several years of SMMR data will also be available for mapping sea-ice concentration, the fraction of multiyear ice, and the surface ice temperature on a six-day basis. Sensor systems have been developed that could provide information on many of the secondary sea-ice characteristics (Table I). Detailed information on ice morphology, ridging intensity, leads and fractures, and ice drift are becoming more vital for model validation as the complexity of cryospheric and atmospheric models increase. However, few data of this type have been collected.

Table 1

Sea-Ice Observation Requirements Identified for NASA Ice and Climate Experiment (ICEX) (from NASA 1979)

2.4 Digital data products

Several sets of digitized snow and ice data are now available. Reference WalshWalsh (1978) has extracted northern hemisphere ice concentration values for 1953 to 1977 on a 110 km grid, using a variety of sources (Reference CraneCrane 1979). From 1972 the source used was primarily the weekly Navy-NOAA ice charts. The same grid was used by Reference KellyKelly (1979) for the digitization of the Danish Meteorological Institute ice charts.

Reference Lemke, Trinkl and HasselmannLemke and others (1980) have digitized the monthly ice charts of the British Meteorological Office for the Arctic, 1966–1976, and the weekly US Navy charts for Antarctica, 1973–1979. For the northern hemisphere, the parameter given is the area of ice per 10° longitude × 1° latitude sector, and, for Antarctica, it is the latitude of > 5/8 ice concentration per 5° longitude sector. Arctic and Antarctic ice concentrations and northern hemisphere snow cover have also been digitized, in geographic sectors, for various time periods by Reference Kukla and GavinKukla and Gavin (1979).

A further digital sea-ice data set exists in the form of ESMR microwave brightness temperatures. The most continuous period for which data are available is from September 1973 to December 1974. In the northern hemisphere, the set consists of gridded brightness temperatures on a 293 × 293 matrix, overlying a polar stereographic projection centred on the pole. For the southern hemisphere, the format is the same, but the data are also given as concentration classes in 4% intervals. An atlas of four years of Antarctic sea-ice concentrations has been prepared for 1973–1976 (see Reference Zwally, Comiso and ParkinsonZwally and others 1981).

3.1 Problems of data collection

Reference Kukla and RobinsonKukla and Robinson (1981) compared the operational snow- and ice-cover charts produced by NOAA-NESS and by the US Navy-NOAA Joint Ice Center with ground station reports and original satellite imagery. From 1974 the NOAA-National Earth Satellite Service (NESS) reflectivity charts were found to be sufficiently accurate (in terms of the location of cryospheric boundaries) for large-scale climate studies. The reflectivity classes were also adequate for hemispheric studies, but not for those at the regional scale. Before 1974, however, large discrepancies were found in the snow-cover charts for autumn, and reflectivity classes were found to be inadequate throughout the year. Much of this error is attributed to errors occurring in regions of persistent cloud cover, a situation frequently found in middle and high latitudes in autumn. Problems in determining the areal snow extent are also found in mountainous regions, and in forest. Reference Kukla and RobinsonKukla and Robinson (1981) point out that a dense coniferous forest with a dark canopy may show very little difference between summer and winter, even with 30 cm of snow on the ground. The Navy-NOAA ice charts were found to be a good representation of ice extent and concentration, although thin ice was frequently not differentiated from open water.

Problems in image interpretation also occur when using conventional visible and IR satellite data due to the difficulty in distinguishing between snow/sea ice and cloud cover. Differentiation between snow/ice and cloud is complicated by the similarity of their radiative temperatures (particularly in the case of of low cloud), and by the similarity of their spectral reflectance in the visible wavelength. Experimental work with laboratory and both aircraft and satellite observations show that automated discrimination between snow or ice and cloud is possible in the near IR (Reference Bunting and EntremontBunting and Entremont 1982, Reference Scharfen and AndersonScharfen and Anderson 1982), but sensors such as these will not be in operational use until the next generation of meteorological satellites is launched.

Microwave radiometer data from Nimbus 5 and 6 are available during the polar night and are essentially unaffected by cloud cover; as such they have proved very useful in determining the position of the ice edge. The microwave brightness temperature received by the satellite, however, is a function of the surface temperature, the amount of open water in the field of view, and the emissivity of the near surface layer of the ice. The emissivity varies primarily as a function of the salinity of the ice, which is inversely correlated with age. The inability to resolve these separate effects quantitatively may lead to large errors in the estimate of ice concentration, and illustrates the major problem with the use of the single channel data (two channels in the case of Nimbus 6). In this respect, considerable improvement is anticipated from the multi-channel (SMMR) microwave data from Nimbus 7. These data, available from 1979, are now being processed.

In attempting to construct a long-term data set suitable for climatological analysis, a major problem often occurs due to the inhomogeneity within the data source. Reference WalshWalsh (1978), for example, has produced a 25-year digital set of monthly sea-ice data on a 110 km grid, showing ice concentration at 1/10 intervals. The data sources, their spatial and temporal resolution, and their reliability, vary considerably over the 25-year period (Table II). Table III shows the satellites and sensor systems used in the NOAA-NESS snow and ice boundary analyses since 1966. In this case the data set is based entirely on satellite observations, but again sensors employed have undergone considerable development with time.

Table II

Description of Data Sources Used for the Walsh (1978) Sea-Ice Data Set (from Crane 1979)

Table III

Satellites and Sensors used in Mapping Northern Hemisphere Snow and Ice Cover (from Reference CraneCrane 1979)

3.2 Problems of data compilation

Some of the difficulties involved in the production of climatological archives arise from the fact that most sea-ice/snow-cover analyses are performed on an operational basis. Several consequences of this approach are apparent. For example, because of time constraints not all of the potential data sources are necessarily included in the analysis, and there is little opportunity for quality checking of the data or products. Where different data sources are used, there is usually no indication in the final product of which data sources were used, nor to how much of the analysis they contributed. An example can be seen in the case of the US Air Force snow depth model (SNODEP). The model outputs global snow depths for a grid superimposed on a polar stereographic projection (grid spacing is 25 nautical miles (46 km) at 60°N and S). Snow depths are given in six categories from <5 to >25 cm and the analysis is updated on a daily basis. The model inputs include station observations of snow depth, precipitation, and temperature, ice cover from the weekly Navy-NOAA charts, and satellite analysis. Missing data are filled from climatology. No information is given, however, on the data source for any given grid point and time. Reference Kukla and RobinsonKukla and Robinson (1981) find that the model tends to overestimate snow-cover extent, and suggest that these data may be less suited for climate-related studies involving albedo than the NOAA charts for most of the year. They also note, however, that the model may provide the best operational source of snow-extent data in autumn, and the quality of the data has improved over time (data are available from 1967 to the present).

A related problem is found in maps composited over several days. In this case it is usually impossible to determine which regions were mapped on any particular day. Most of the satellite-derived operational snow and ice charts fall into this category, as do the three-day composited images produced from the ESMR data. This may not pose a problem for hemispheric-global scale studies, where a weekly time step may be sufficient; it would become increasingly important, however, for analyses at smaller regional scales.

3.3 Problems of data management

Aside from the problems of data compilation that have already been discussed, several problems can be mentioned that are common to many data management tasks. For example, cryospheric data are fragmented by region with no single agency being responsible for gathering them together. In many cases, data are not put out in a form that can be readily used by the researcher. Snow and ice data are frequently presented in chart format, whereas most research purposes would require some form of numerical gridded data.

The user community often has great difficulty in determining what data are available and what information exists concerning their quality and possible limitations. Frequently, there are also long delays between the receipt of data on the ground and the delivery of preprocessed data to the user. An example of this is the SMMR data for 1979, which are only now becoming generally available.

A growing demand is also being felt for the development of level 3 data sets (Figs.1 and 2). Very few merged data sets are available at present. Exceptions can be found in the Arctic Ice Dynamics Joint Experiment (AIDJEX), and the current Marginal Ice Zone Experiment (MIZEX). In both cases large amounts of concurrent sea-ice, atmospheric and oceanic data have been collected, but both experiments are of a limited time span and data are only for relatively small geographic regions. Merged data sets are also necessary at the hemispheric or global scale. Considering all these problems, there is an obvious need for the development of a comprehensive data management system that would provide a multi-disciplinary polar research data set.