The most comprehensive source of cloud information is the International Satellite Cloud Climatology Project (ISCCP) archive. This data base provides global coverage using NOAA, GOES, METEOSAT, GMS, and INSAT satellites. Because geostationary satellites provide useable data to only 60_ N, the NOAA polar orbiters are relied on to provide coverage of the study area. NOAA satellites provide 3 hour temporal sampling of the area. In addition temperature soundings are made by the TIROS-TOVS (Television and Infrared Operational Satellite-Operational Vertical Sounder). Continuous temporal coverage has been provided by the operational satellite system between July 1983 and December 1991.

Six data products are derived from satellite observations and computer analysis; (1) ISCCP-B1 data; original NOAA-AVHRR visible and infrared radiances, (2) ISCCP-B3 data; normalized coarse resolution NOAA-AVHRR visible and infrared radiances, (3) ISCCP-IS data; correlative ice and snow data sets compiled by the Navy/NOAA Joint Ice Center, (4) ISCCP-TOVS data; TOVS temperature and moisture profiles, (5) ISCCP-C1 data; cloud parameters derived from the ISCCP-B3 data, and (6) ISCCP-C2 data; monthly averages and summary statistics. B1, B3, IS, TOVS, and C1 data are available for the time period July 1983 - December 1991. C2 data are available from July 1983 - December of 1991. B1 data are expressed as 4 km x 4 km pixels on an equal area grid, B3 and IS data are expressed as 30 km x 30 km pixels on a equal area grid and C1 and C2 data are calculated from 30 km x 30 km B3 pixels and pixel statistics are reported over a 250 km x 250 km area. Additional information concerning data processing methods employed by ISCCP is available in the ISCCP Cloud Products User Guide (1989).

C1 and C2 data include the following cloud parameters; (1) cloud spatial statistics (number of cloudy and clear pixels, number of IR only pixels), (2) infrared radiance, (3) cloud top temperature, (4) cloud top pressure, (5) visible radiance, (6) reflectance, (7) visible and infrared optical thickness, (8) humidity, and (9) ozone column abundances. In addition the C2 data set provides a monthly statistical summary. It is important to note that visible optical thickness values are not available from September through March for latitudes greater than 60_ N due the to low sun angles.

B1, B3, IS, TOVS data (AVHRR radiances, ice and snow measurements, and atmospheric temperature and humidity 1983-1991), C1 data (cloud parameters, 1983-1991), and C2 data (monthly averages and summary statistics, 1983-1991) can be obtained directly from NESDIS-SDSD (National Environmental Satellite Data and Information Service-Satellite Data Services Division). These data are available on 9 track tapes or on IBM cartridges. Sixteen days of C1 data are contained on one tape; where as, one year of C2 data are contained on a single tape. Ninety five minutes of B3 data can be written to one tape. Specific regions are available on tape upon request for B1, IS and TOVS data. The cost per tape is $92 plus an $11 shipping and handling fee.

Naval Post Graduate School Climate Data

Additional cloud data are available from the Naval Post Graduate School. These data were collected by ground observers and include information on cloud amounts, cloud type and cloud radiance. Observations were taken for time periods ranging from 10 minutes to several hours for the duration of MIZEX and CEAREX. Climatological data which includes temperature, pressure and humidity are also available and have been archived. Unfortunately, cloud data have not been digitized, they are available in hardcopy format only. Additional climatological data on MS/DOS compatible optical disk are also available; however, much of the climatological data are still being digitized.

Air Force Data

The AIR FORCE Global Weather Central has archived cloud observations and other climatological data. The primary product is cloud amount. The data collection methods follow guidelines established by the World Meterological Organization (WMO). Observations collected in the study area are, however, very sparse. There are only several widely spaced observation stations on Greenland. Daily data collections have been made. At the time this document was prepared a detailed description of observation locations was ordered from NESDIS. Satellite observations (from the DMSP-Optical Line Scanner) have also been used during some periods to produce a nephanalysis (a cloud climatology produced by manual interpretation of satellite images). These data are available from the National Climatic Data Center on CCT's. The price per tape is $92 plus an $11 shipping and handling fee.

Cloud Analysis and Background

Clouds in the Arctic have many dynamic variations which can occur on diurnal, monthly, seasonal, and even yearly time scales. In order to accurately monitor this variability, careful attention must be given to the appropriate selection of cloud parameters, the spatial and temporal resolution over which cloud parameters are calculated and the selection of data products used to obtain cloud information. It is also important to note that visible optical parameters cannot be obtained during the night and during long periods of winter darkness. In this section the following topics will be discussed; (1) selected cloud parameters, (2) selected spatial resolution, (3) selected temporal resolution, and (4) data acquisition and organization.

Cloud Parameters

Climate researchers need to understand how the Sun's energy interacts with the Earth. Since clouds scatter and absorb visible and infrared radiation they play an important role in the Earth's radiation budget. Researchers are also interested in the spatial and temporal variability of cloud cover and how these variables relate to changes in snow and ice cover in polar regions. This variability is especially important near the ice-ocean boundary.

Researchers (Carleton, 1984; Rossow, 1985) have defined cloud parameters which they consider important in understanding cloud variability. They include: (1) cloud amounts, (2) cloud top temperature (degrees kelvin), (3) cloud top pressure (millibars), (4) reflectance (unitless), (5) visible (0.6 microns) optical thickness (unitless), and (6) humidity (precipitable centimeters). For the present project, primary concern was given to cloud amounts, visible and infrared radiances, reflectance, visible optical thickness, and cloud top pressure. These parameters were obtained from the ISCCP cloud data base. A list of ISCCP cloud parameters processed in the present work is given in Table C-1.

Spatial Resolution

In general the spatial resolution chosen for a cloud climatology depends on the size of the spatial variations of interest. For example, synoptic scale weather systems can produce large uniform cloud decks. Rossow (1989) suggest that significant variations in cloudiness may occur on scales as small as 30-100 km, especially in the tropics where air mass convection is likely.

Cloud generation mechanisms in polar regions (cyclogenesis, and advection) generally occur on synoptic scales, thus small scale changes in cloud structure are less likely in these regions. Variations in cloud structure cause related changes in cloud optical properties. Research (Rossow, 1989) has shown that optical properties tend to vary significantly on scales as large as 500 km, although, less significant variations in optical properties can occur on scales as small as 100 m. The ISCCP algorithm calculates optical properties (radiances and visible optical thickness) using 4 km x 4 km pixels. Cloud optical statistics are then compiled for pixels within a 250 km x 250 km area. This resolution is considered sufficient for studying cloud variability over most of the study area.

Temporal Resolution

Results reported by Hughes et al., (1983) reported that statistical variations in cloudiness in the high latitudes observed on weekly time scales are also present on monthly time scales. When cloud cover statistics are averaged on time scales larger than a month, statistical differences in cloudiness become difficult to interpret. Variations in cloud cover in the Arctic are generally related to seasonal variations in synoptic circulation (Carleton, 1984). Variations in cloud cover, however, can occur on short time scales in response to circulation anomalies and changes in the local ice edge. This study will focus on cloud statistics compiled on a monthly basis. For selected short time scale variations in cloudiness daily information were obtained for time frames of interest.

TABLE C-1  ISCCP Atmospheric Parameters
        PARAMETER                     SCALE FACTOR	
1       Mean Albedo                     1000
2       Mean Tau (visible optical thk)   100
3       Mean Cloud Amount (%)            100
4       *Mean Cloud Thickness (m)         10
5       Mean Cloud Base Height (m)        10
6       Cloud Height Above Ground (m)      1
7       Sigma Albedo                    1000
8       Sigma Tau                         10
9       *Sigma Thickness                  10
10      Time Sigma Albedo                100
11      Time Sigma Tau                    10
12      *Time Sigma Thickness            100
13      Low Level Cloud Amount (%)       100
14      Mid Level Cloud Amount (%)       100
15      High Level Cloud Amount (%)      100
16      Cumulus Cloud Amount (%)         100
17      Stratus Cloud Amount (%)         100
18      Altocumulus Cloud Amount (%)     100
19      Nimbostratus Cloud Amount (%)    100
20      Cirrus Cloud Amount (%)          100
21      Cirrostratus Amount (%)          100
22      Surface Temperature (deg K)      100
23      Sigma Temperature (deg K)        100
24      Snowice Cover (%)                100
25      Mean Solar Radiance             1000
26      Mean Temp Clouds (deg K)         100
27      Time-Sigma Temperature           100
28      Mean Spa-Sigma Temperature       100
29      Surface Temperature (TOVS)       100
30      Cloud Amount (%)                 100
*ERIM/CIESIN Derived Values

Data Acquisition and Organization

This analysis will rely mainly on ISCCP products, with the Naval Post Graduate School climatological data set being used to verify cloud characteristics. We will assume that the study period is coincident with MIZEX (1983, 1984, and 1987) and that we are interested in cloud parameters averaged on a monthly basis. Since monthly temporal resolution is of interest, C2 data will be purchased. Presently, C2 data are available only through 1991. For detailed case studies cloud parameter pixels (30 km spacing) used to generate C1 statistics will be obtained from Bill Rassow at NASA-Goddard. These data will be essential for understanding horizontal cloud structure and horizontal changes in cloud optical parameters. C1 data were acquired for the time periods corresponding to at least one of the MIZEX collections. In addition some stage B3 data (reduced resolution 30 km AVHRR data) might be utilized with the C1 data to infer pixel level cloud parameters.

Detection Algorithms

There is currently no single operational cloud detection algorithm that works well for all sections of the Earth (Rossow, 1987). Detecting and characterizing clouds in polar regions presents a particularly difficult challenge. In the following discussion a brief description of the ISCCP cloud detection algorithm is presented followed by a discussion of alternative detection methods proposed by other cloud researchers.

The ISCCP cloud algorithm is a robust method used to characterize clouds under a variety of time and space varying environmental conditions. The first and most crucial phase of the program is the cloud detection algorithm. Since the atmosphere is a highly variable medium, the detection algorithm uses statistical measures which relate spatial and temporal changes in the atmosphere to observed narrow band visible (0.6 um) and infrared (11 um) radiances. The initial portion of the algorithm is a first order attempt to isolate those pixels which have low spatial and temporal variability. This is accomplished with two statistical test, one to measure spatial variability and the other to measure time variability. These pixels are labeled "clear", and all other pixels are labeled "undecided" (status of pixel to be determined). Previous research has demonstrated that clear sky pixels, in general, have less spatial and temporal variability than cloudy pixels (Rossow, et al., 1985).

Following a first order analysis, a secondary analysis is conducted using three statistics. They are: (1) the difference in extreme radiances (visible and infrared) for consecutive 5 day periods, (2) the number and average values of the pixels labeled as clear for 5 day periods, (3) and the number and average values of the pixels labeled as clear by the time test and undecided by the spatial test (test described above). Large differences in extreme radiances generally indicate cloud contamination, while low populations of clear pixels suggest persistent cloudiness. If the statistics suggest little cloud contamination a clear sky radiance is formed from the 5-day-average of the pixels labeled as clear. If statistics suggest contamination, the clear sky radiance is formed from a 30-day-average. Finally if the number of clear pixels in the 30-day statistic is too small, the clear sky radiance is obtained from the 30-day radiance extreme. The final decision regarding the status of the image pixels is made by a bispectral threshold test. Using the clear sky radiance, all pixels with radiance values which differ significantly (3.5% (ocean) to 6.0% (land) for visible, and 3.0 K (ocean) to 8.0 K (land) for IR) are labelled as cloudy.

Detection of Clouds in Polar Regions

Detection of clouds is a difficult problem in polar regions over the snow and ice surface. Rossow (1987) estimated a 15% over prediction in cloudiness by the original (1983) ISCCP algorithm when compared with manual observations over the arctic ice pack. Over estimation of cloud cover seemed the most pronounced near the ice edge. There are several reasons for these difficulties; (1) the surface characteristics near the ice edge are very changeable in time and space, making it difficult to select a good clear sky value, (2) persistent cloudiness in the Arctic especially near the ice edge makes the selection of a clear sky value less certain, (3) snow covered surfaces are often as reflective as clouds (Key et al., 1989), and (4) the thermal structure of the troposphere is characterized by frequent isothermal and inversion layers (Key et al., 1989).

Some researchers (Key et al., 1989; Rossow, 1987; Maslanik et al., 1989) have suggested alternative methods for detecting and characterizing clouds in ice and snow covered polar regions. These methods rely on passive microwave data to characterize changes in the ice and snow surface. The ISCCP algorithm has recently been augmented to include correlative snow and ice data which describe spatial and temporal changes in surface characteristics. This avoids the selection of a clear sky radiance in areas where surface characteristics have changed. Other correlative data is also used to describe radiative properties of the atmosphere. Results reported by Key (1989) indicate that an augmented detection algorithm may provide a 5-10% improvement in computed cloud amounts when compared with results generated by the original algorithm, depending on the surface type and cloud proportions. Rossow (1990, personal communication) indicated that ISCCP data products (July 1983-1991) have been reprocessed using the augmented algorithm.

Radiative Transfer Analysis

Calculation of cloud optical properties is accomplished following classification of a pixel as clear or cloudy. Pixel radiances are compared to radiative transfer model calculations designed to simulate the measurements of the AVHRR channels. These comparisons are used to isolate the surface reflectances and temperatures from the clear sky radiances and the cloud optical thickness and top temperature from the cloudy radiances. Atmospheric properties that affect satellite measured radiances are specified from the correlative data sets (ISCCP, 1990). It is important to mention that visible optical thicknesses calculated over highly reflective surfaces, such as snow and ice, tend to be on the high side (Rassow, 1991, personal communication).

Data Processing

There are three principal components involved in processing cloud data. The first component involves the retrieval of information from the acquired data tapes. This may require the writing of software to read the computer tapes. Software for decoding the disk files is included with the ISCCP data sets.

The second component in the data processing phase deals with the issue of digitally mapping cloud parameters to a specific grid. At this phase a desired map projection should be defined. Cloud parameters must then be interpolated to conform to this mapping. Since the cloud data will be used in conjunction with snow and ice data (SSM/I sea ice data), the cloud data sets must be registered with this data. The choice of map grid, therefore, also depends on the format of accompanying snow and ice information. Research (Rossow et al., 1984) has shown that an equal area grid is more suitable for preserving information than an equal angle grid. All ISCCP data products have been mapped to an equal area grid.

A third component which is concerned with information storage and retrieval methods must be addressed. The choice of a method depends upon: (1) computer resources, (2) number of data points, and (3) the desired format of output data products (i.e., cloud optical parameters vs. lat-long.). This information system must allow easy access to data corresponding to a particular region and time period. Output formats should be compatible with the input formats of display programs and the optical models in use. A flow diagram for ISCCP data conditioning is shown in Figure C-1. Example processed monthly mean values for albedo and cloud amount are shown as Figures C-2 and C-3 respectively.

Data Analysis

Once cloud information has been mapped to the desired grid a statistical analysis will be performed. Specific attention will be given to variations in cloud optical properties over time and space. Important statistics that will be generated for this part of the analysis include; (1) spatial and temporal variances and means, and (2) a spectral analysis to isolate periodic variations in the data. In addition, temporal correlations will be investigated during the second component of the statistical analysis. The variances will allow a first order understanding of temporal variability in cloudiness and optical characteristics.

Spectral analysis can be helpful in isolating modes of temporal variability that are significant. These modes may occur on daily, weekly, monthly, seasonal, or yearly time scales. Short time scale temporal modes will be computed for the C1 data product (reported twice a day North of 60_ latitude) over a six month period. This analysis will be helpful in characterizing temporal variations on diurnal to monthly time scales.

The C2 statistic will also be utilized in the spectral analysis to isolate longer term temporal variabilities, such as those occurring on seasonal to yearly time scales. The autocorrelation function computed for the C1 and C2 statistics will be useful for determining the correlation time of cloudiness in the study area. In addition, spatial variability may be investigated in this analysis. Currently C1 and C2 statistics do not contain information regarding the spatial relationships of cloudy pixels. This information is necessary to understand cloud morphology (cloud shape and horizontal structure). If possible the spatial relationship of the pixels will be obtained for a month of C1 data and spectral analysis will be used to understand the spatial relationships and spatial correlations.

Some effort may be spent addressing the causes of cloud variability on observed spatial and temporal scales. Cloud variability is related to ice dynamics, boundary layer dynamics (short time scale), seasonal variations in solar insolation, and atmospheric and oceanic circulation anomalies.


Arctic Passive Microwave Database Grid

As briefly mentioned above, it is important to note that all of the data utilized was put onto the same grid in order to more easily do pixel-by-pixel comparisons of the data. The grid that was selected was the same grid on which the SSM/I and SMMR data available on CD-ROM from the National Snow and Ice Data Center (NSIDC) was archived.

The SSM/I brightness temperature and ice concentration polar grid as described in this section were defined in the PODS SSM/I Functional Requirements document (Bonbright, 1984). While both Northern and Southern hemisphere grids were defined, only the Northern hemisphere grids will be described here since that is the only data used in the current database. Each pixel on this grid has a corresponding latitude/longitude value which was calculated based on an ellipsoid model of the earth rather than a sphere. The ellipsoid used for this model was taken from a Hughes Aircraft Company Document (1980) which assumes a radius of 6378.273 kilometers and an eccentricity of .081816153. The polar stereographic formula used to convert X-Y to latitude/longitude and visa-versa can be found in Snyder (1982).

In order to minimize distortions, the plane tangent to the earth was placed at 70 degrees rather than at the poles. This will increase the distortion at the poles by about 3%. This does, however, cause little to no distortion in the Marginal ice zone which is an area of great scientific interest. While the original SSM/I northern hemisphere grid extends out as far as 31 degrees north in latitude, the grid used in this database was limited to all data north of 60 degrees. The grid which was used has an equal area spacing of 25 kilometers in each direction. In order to subset the area of interest, a square region was taken (270 records by 270 elements) and any points on the grid which had a latitude value less than 60 degrees were flagged as `out-of -bounds data'. A latitude and a longitude binary map of this area was generated. The latitude and longitude locations were written into binary files called LAT2.SUB and LONG2.SUB, respectively. These files contain the center latitude and longitude of each pixel on the 270 X 270 grid which was described above. The data values are 2-byte integers and are in .01 degree units (i.e. one must divide the values by 100 to obtain actual latitude and longitude values in degrees).

SSM/I Data

The Special Sensor Microwave / Imager (SSM/I) is a multichannel passive microwave radiometer. The SSM/I is on board the Defense Meteorological Satellite Program (DMSP) spacecraft and has been in operation since July, 1987. The instrument was designed to measure atmospheric/ocean surface brightness temperatures at four separate frequencies and two polarizations (vertical and horizontal). The system currently operates the following channels/polarizations: 19.3 V/H, 22.2 V, 37.0 V/H and 85.5 V/H GHz.

The sensor samples every 12.5 km along track for the 85.5 GHz channels and every 25 km at for the lower frequencies. The sensor provides global coverage except for areas north of 87 Degrees-North and south of 87 Degrees-South. The sensor is designed to collect data on a daily basis (which it has done with very little variation). We currently have all of the SSM/I data collected and put onto CD-ROM by NSIDC. This covers data starting in July 1987 and runs through December 1991.

The SSM/I data was used to obtain ice concentration estimates (both total and multi-year concentrations) and to obtain near surface wind speed (over the open ocean only). This data was included in the arctic database because ice concentration estimates can give us a great deal of information about global atmospheric and oceanic changes on the earth. Wind speed is a good indicator of atmospheric conditions and gives us information about the movement of the ocean surface.

Data Description

The SSM/I data used for this database was obtained from the National Snow and Ice Data Center (NSIDC). This data was made available to us on CD-ROM and was sent to us as soon as the data was processed and mastered onto the CD. As mentioned in the previous section, we currently have all of the SSM/I data put onto CD-ROM which runs through December 1991. As more CD's become available, the database will continue to be updated to remain as current as possible.

The SSM/I data was binned into 25 km grids for the 19.3, 22.2 and 37.0 GHz channels and into 12.5 km grids for the 85.5 GHz channels. The data was placed on the polarstereographic grid as described in Section 2.0. The method used to place the data on this grid is called the `drop-in-the-bucket' method. Briefly, this method involves taking the center position of each of the SSM/I footprints and placing the entire weight of that data point into the appropriate grid cell. While more sophisticated algorithms than this do exist, it was concluded that their benefits did not outweigh the overhead associated in the processing time (NSIDC, 1992).

The brightness temperatures are summed for each grid cell for a 24 hour period (midnight to midnight UTC) and then divided by the number of observations to obtain an average brightness temperature for the grid cell. The brightness temperatures (TB) were calculated from the Temperature Data Record (TDR)produced by the Navy's Fleet Numerical Oceanography Center (FNOC). The process from TDR is described in more detail in the NSIDC Special Report (1992). The TB's written on the CD-ROM are 2-byte integers in .1 degree Kelvin units. For each day the sensor collected daily (which was nearly every day) there exists a file for each of the SMM/I channels in operation (7 in all). Each of the files are 334 elements by 448 records. To read more information about the CD-ROM directory structure and image file format please see the NSIDC Report (1992), pages B1-B12.

Data Processing Methods

The NASA/Team algorithm was applied to the SSM/I data sets in order to obtain a total and multiyear ice concentration estimate. The NRL algorithm was applied to the data to calculate near surface wind speed. Each of these algorithms will be described briefly below along with references to obtain additional information about the algorithms.

Once the SSM/I data was processed three image files were generated on a daily basis (total ice concentration, multi-year ice concentration, and near surface wind speed). Each of these files were put on the polar stereographic grid as described in section 2.0. The images are 1-byte signed integer images which are 270 records by 270 elements in size (which gives a total of 72900 bytes per image). Each pixel represents a 25 km square on the ground.

The naming convention used for these images is based on the date in which the data was collected and on the type of data which was extracted. The naming format is as follows: yymmddS.TOT, yymmddS.MYC and yymmddS.NSW where yy are the last two digits of the year the data was collected, mm is the numeric month, and dd is the date. The extensions are as follows: TOT are the total ice concentration files, MYC are the multi-year ice concentration files and NSW are the near surface wind speed files.

Ice Concentration Algorithms

The ice algorithm which was developed by the NASA/Team is a multichannel concentration algorithm that generates both total and multiyear ice concentrations. The algorithm utilizes global tie points and utilizes both the polarization and spectral gradient ratios from the 19.3 and 37 GHz channels to determine ice concentration (see Cavalieri, 1984, Gloersen, 1986 and NSIDC, 1991 for further details).The multichannel technique was adopted from those developed for the SMMR sensor.

The validity of the algorithm has been tested by ERIM and a great number of other sources. ERIM's validation work involved comparing the SSM/I derived ice concentrations to the coincident P3-SAR manually interpreted data sets. The comparisons showed that this algorithm estimated total ice concentration to within 10% over all ranges of concentration (Jentz et al., 1991 and Russel et al., 1991). Other tasks to validate the NASA Team SSM/I total ice concentration algorithm involved comparing the SSM/I measurements with Landsat imagery (Steffen et al. 1991). This study found that globally the SSM/I underestimated sea ice concentration by 4% +/-7%. Larger errors were found during the summer melt periods. Cavalieri et al. (1991) compared the SSM/I algorithm with C-band SAR imagery and found that the mean difference of the SSM/I underestimated the SAR by 2.4% +/-2.4%.

Not as much research has been done on the validation of multiyear concentration estimates. A study was performed in the Beaufort Sea comparing this algorithm's SSM/I derived multiyear ice estimates to those derived for the P3-SAR and it was found that the SSM/I algorithm over-predicted multiyear ice concentration by up 25% in the low multiyear concentration areas and gave a good estimate of the high multiyear ice concentrations to within 15% (Jentz et al. 1991 and Russel et al. 1991). Cavalieri et al. (1991) also found that the SSM/I algorithm overestimated multiyear ice concentration by 12% +/-11%. These discrepancies have been found to be due (at least in part) to the large variability of multiyear ice emissivity observed over individual floes (Grenfell, 1991). A number of routines were run in order to derive both the total and multi-year ice concentrations. The flow of the routines and programs can be seen in the diagram below.

Wind Speed Algorithms

The wind speed algorithm used for this database was taken from the DMSP SSM/I Calibration / Validation Report (Hollinger, 1989). The wind speeds derived from the SSM/I brightness temperatures are ocean surface wind speeds defined in nearest meters / second only over the open ocean. This algorithm uses the TB's from the 19 V, 22 V 37 V and 37 H channels.

The wind speed algorithm works best in the presence of large storm systems, although the values may be contaminated by dense clouds and rain. Approximately 85% of the time the algorithm gives wind speed estimates within 2 m/s accuracy. The other 15% of the time the algorithm gives from +/- 2 m/s to > +/- 10 m/s accuracy. See Hollinger (1989) for a more detailed description of the algorithm and its accuracy.

The previous figure also describes the flow of the program for the wind speed algorithm. Both the concentration algorithms and the wind speed algorithms are run at the same time so the CD-ROM only needs to be accessed once for both operations.


The Scanning Multichannel Microwave Radiometer (SMMR) data was also obtained from NSIDC. The SMMR was launched on October 28, 1978 and was put on an alternating day collection mode on November 19, 1978 due to power limitations. The SMMR ceased operation in August 1987. This ten channel device collected data at five different, dual-polarized frequencies. These are 6.6 H/V, 10.69 H/V, 18.0 H/V, 21.0 H/V, and 37.0 H/V GHz. The SMMR sensor gives complete coverage north of 72 degrees each day the sensor was in operation. The SMMR data was used to obtain ice concentration estimates (both total and multi-year concentrations). This data was included in the arctic database because ice concentration estimates can give us a great deal of information about global atmospheric and oceanic changes on the earth.

Data Description

Like the SSM/I data, the SMMR data also was made available to us on CD-ROM on the same binned-polar stereographic grid developed by the NASA Ocean Data System (NODS). The SMMR brightness temperatures were mapped onto the SSM/I grid by using the `drop-in-the-bucket' technique on the brightness temperatures. See Gloersen et al. (1991) for more details concerning this procedure.

The current SMMR data that is available in this format covers all of the SMMR data collected while the sensor was in operation. This includes data on nearly an every-other-day basis starting in October 1978 and ending in August 1987.

Each CD-ROM contains individual channels of brightness temperatures (TB's). These brightness temperatures were derived from raw instrument voltage counts using a number of normalization and correction techniques. In depth descriptions of these techniques can be found in the Gloersen(1983,1987), Francis (1987), and Gloersen et al. (1991). Like the SSM/I the TB's written on the CD-ROM are 2-byte integers in .1 degree Kelvin units. For each day the sensor collected daily (which was nearly every other day) there exists a file for each of the SMM/I channels in operation (10 in all). Each of the files are 334 elements by 448 records. To read more information about the CD-ROM directory structure and image file format please see the NASA Report (1991).

Data Processing Methods

The ice concentration algorithm is an adaptation of the algorithm designed by the NODS Software Group (see the NODS User Handbook, Appendix CC, pp22-26). The algorithm uses the latest SMMR tiepoints which were determined by Don Cavalieri (Cavalieri et al, 1984). This algorithm computes both total and multi-year ice concentration estimates. This algorithm will be described briefly below along with references to obtain additional information about the algorithm and its validity.

Once the SMMR data was processed two image files were generated on nearly an ever-other-day basis (total ice concentration and multi-year ice concentration). These two files were each put on the polar stereographic grid as described in section 2.0. The images are 1-byte signed integer images which are 270 records by 270 elements in size (which gives a total of 72900 bytes per image). The naming convention used for these images is based on the date in which the data was collected and on the type of data which was extracted. The naming format is as follows: yymmddS.TOT and yymmddS.MYC (for total and multi-year ice concentration files, respectively).

Ice Concentration Algorithm

As mentioned above, the ice concentration algorithms used to obtain total and multi-year ice concentration estimates from SMMR brightness temperatures was one developed by NODS with tie points determined by Don Cavalieri. The algorithm is non-iterative and uses multiple channels and polarizations to calculate ice concentrations (i.e. 18 Ghz V & H and 37 GHz V). A weather filter (GR > .07) is applied to the 25km TB cell to correct for false concentrations computed in regions of open water.

A number of studies have been done to validate the concentration estimates from the SMMR using this algorithm. A comparison between SSMR derived total ice concentration and SIR-B ice concentration found the SMMR to give good estimates with a mean difference of 1.7 % and a standard deviation of 7.4% (Martin et al., 1987). Another comparison examined SMMR derived ice concentration estimates and Landsat imagery and found that the two concentrations were separated by less than 10% in nearly all cases (Steffen et al. 1988). Cavalieri et al. (1984) found that the precision in calculated total ice concentration was from 5-9% and multiyear from 13-25%. While this was an early study, further analysis regarding validation of this algorithm also appeared in Gloersen et al. (1986).


Brooks, D.R., E. Harrison, P. Minnis, and J. Suttles, 1986, "Development of Algorithms for Understanding the Temporal and Spatial Variability of the Earth's Radiation Balance," Reviews of Geophysics, Vol. 24, No. 2, 422-438.

Carleton, A.M., 1984, "Cloud-Cryosphere Interactions," Satellite Sensing of a Cloudy Atmosphere, Taylor & Francis, London, 289-325.

Coakley, J.A., and F. Bretherton, 1982, "Cloud Cover From High-Resolution Scanner Data: Detecting and Allowing for Partially Filled Fields of View," Journal of Geophysical Research, Vol. 87, No. C7, 4917-4932.

Davies, R., 1984, "Reflected Solar Radiances from Broken Cloud Scenes and the Interpretation of Scanner Measurements," Journal of Geophysical Research, Vol. 89, No. D1, 1259-1266.

D'Entremont, R.P., 1986, "Low- and Midlevel Cloud Analysis Using Nighttime Multispectral Imagery," Journal of Climate and Applied Meteorology, Vol. 25, 1853-1869.

Duvel, J.P., and R. Kandel, 1985, "Regional-scale Diurnal Variations of Outgoing Infrared Radiation Observed by METEOSAT," Journal of Applied Meteorology, Vol. 24, 335-349.

Harshvardhan, and J. Weinman, 1982, "Infrared Radiative Transfer Through a Regular Array of Cuboidal Clouds," Journal of the Atmospheric Sciences, Vol. 39, 431-439.

Henderson-Sellers, A., and N. Hughes, 1984, "Earth - the Water Planet," Satellite Sensing of a Cloudy Atmosphere, Taylor & Francis, London, 1-44.

Herman, G.F., and J. Curry, 1984, "Observational and Theoretical Studies of Solar Radiation in Arctic Stratus Clouds," Journal of Climate and Applied Meteorology, Vol. 23, No. 1, 5-24.

Hughes, N.A., and A. Henderson-Sellers, 1983, "The Effect of Spatial and Temporal Averaging on Sampling Strategies for Cloud Amount Data," Bulletin American Meteorological Society, Vol. 64, No. 3, 250-257.

Ingram, W.J., C. Wilson, and J. Mitchell, 1989, "Modeling Climate Change: An Assessment of Sea Ice and Surface Albedo Feedbacks," Journal of Geophysical Research, Vol. 94, No. D6, 8609-8622.

ISCCP, 1990, "ISCCP Cloud Products Users Guide," NASA's Climate Data System, New York.

Key, J.R., J Maslanik, and R. Barry, 1989, "Cloud Classification Using a Fuzzy Sets Algorithm: A Polar Example," International Journal of Remote Sensing, Vol. 10, No. 12, 1823-1842.

Key, J., and R. Barry, 1989, "Adaptation of the ISCCP Cloud Detection Algorithm to Combined AVHRR and SMMR Arctic Data," Proceedings of the 9th Annual International Geoscience & Remote Sensing Symposium, Vol. I., 188-191.

Kidwell, K.B., 1990 "International Satellite Cloud Climatology Project (ISCCP), Central Archive, Catalog of Data and Products," National Oceanic and Atmospheric Administration, National Environmental Satellite Data and Information Service, National Climatic Center, Satellite Data Services Division (SDSD), Washington, D.C.

Koffler, R., G. DeCothiis, and P. Rao, 1973, "A Procedure for Estimating Cloud Amount and Height from Satellite Infrared Data," Monthly Weather Review, Vol. 101, No. 3, 240-243.

Ledley, T.S., 1988, "A Coupled Energy Balance Climate-Sea Ice Model: Impact of Sea Ice and Leads on Climate," Journal of Geophysical Research, Vol. 93, No. D12, 15, 919-15, 932.

Maslanik, J.A., J. Key, and R. Barry, 1989, "Merging AVHRR and SMMR Data for Remote Sensing of Ice and Cloud," International Journal of Remote Sensing, Vol. 10, No. 12, 1691-1696.

Matthews, E., and W. Rossow, 1987, "Regional and Seasonal Variations of Surface Reflectances from Satellite Observations at 0.6 um," Journal of Climate and Applied Meteorology, Vol. 26, 170-202.

McKee, T.B., M. DeMaria, J. Kuenning, and S. Cox, 1983, "Comparison of Monte Carlo Calculations with Observations of Light Scattering in Finite Clouds," Journal of the Atmospheric Sciences, Vol. 40, 1016-1023.

Minnis, P., and E. Harrison, 1984, "Diurnal Variability of Regional Cloud and Clear-Sky Radiative Parameters Derived from GOES Data. Part I: Analysis Method," Journal of Applied Meteorology, Vol. 23, No. 7, 993-1011.

Rossow, W.B., 1989, "Measuring Cloud Properties from Space: A Review," Journal of Climate, Vol. 2, 201-213.

Rossow, W.B., C. Brest, and L. Garder, 1989, "Global, Seasonal Surface Variations From Satellite Radiance Measurements," Journal of Climate, Vol. 2, 214-247.

Rossow, W.B., L. Garder, and A. Lacis, 1989, "Global, Seasonal Cloud Variations from Satellite Radiance Measurements. Part I: Sensitivity of Analysis," Journal of Climate, Vol. 2, 419-458.

Rossow, W.B., 1987, "Application of ISCCP Cloud Algorithm to Satellite Observations of the Polar Regions," ISCCP Workshop on Cloud Algorithms in the Polar Regions, 1987.

Rossow, W.B., F. Mosher, E. Kinsella, A. Arking, M. Desbois, E. Harrison, P. Minnis, E. Ruprecht, G. Seze, C. Simmer, and E. Smith, 1985, "ISCCP Cloud Algorithm Intercomparison," Journal of Climate and Applied Meteorology, Vol. 24, No. 9, 877-902.

Rossow, W.B., and L. Garder, 1984, "Selection of a Map Grid for Data Analysis and Archival," Journal of Climate and Applied Meteorology, Vol. 23, 1253-1257.

Saltzman, B., and R. Moritz, 1980, "A Time-Dependent Climatic Feedback System Involving Sea-Ice Extent, Ocean Temperature, and CO2," Tellus, Vol. 32, 2, 93-118.

Saunders, R.W., 1986, "An Automated Scheme for the Removal of Cloud Contamination from AVHRR Radiances Over Western Europe," International Journal of Remote Sensing, Vol. 7, No. 7, 867-886.

Schiffer, R.A., and W. Rossow, 1985, "ISCCP Global Radiance Data Set: A New Resource for Climate Research," Bulletin American Meteorological Society, Vol. 66, No. 12, 1498-1505.

Schiffer, R.A., and W. Rossow, 1983, "The International Satellite Cloud Climatology Project (ISCCP): The First Project of the World Climate Research Programme," Bulletin American Meteorological Society, Vol. 64, No. 7, 779-784.

Schott, J.R., and A. Henderson-Sellers, 1984, "Radiation, the Atmosphere and Satellite Sensors," Satellite Sensing of a Cloudy Atmosphere, Taylor & Francis, London, 45-89.

Seze, G., and M. Desbois, 1987, "Cloud Cover Analysis from Satellite Imagery Using Spatial and Temporal Characteristics of the Data," Journal of Climate and Applied Meteorology, Vol. 26, 287-303.

Stephens, G.L., 1985, "Radiative Transfer through Arbitrarily Shaped Optical Media. Part II: Group Theory and Simple Closures," Journal of the Atmospheric Sciences, Vol. 45, No. 12, 1837-1848.

Stephens, G.L., 1984, "The Parameterization of Radiation for Numerical Weather Prediction and Climate Models," Monthly Weather Review, Vol. 112, 826-866.

The Polar Group, 1980, "Polar Atmosphere-Ice-Ocean Processes: A Review of Polar Problems in Climate Research," Reviews of Geophysics and Space Physics, Vol. 18, No. 2, 525-548.

Bonbright, D.I., 1984, "PODS SSM/I Functional Requirements (Version 1.0)," Jet Propulsion Laboratory Document 715-63.

Cavalieri, D.J., P. Gloersen, and W.J. Campbell, 1984, "Determination of Sea Ice Parameters with the NIMBUS-7 SMMR," Journal of Geophysical Research, Vol. 89, (D4), p. 5355-5369.

Cavalieri, D.J., J.P. Crawford, M.R. Drinkwater, D.T. Eppler, L.D. Farmer, R.R. Jentz, and C.C. Wackerman, 1991, "Aircraft Active and Passive Microwave Validation of Sea Ice Concentration from the Defense Meteorological Satellite Program Special Sensor Microwave Imager," Journal of Geophysical Research, Vol. 96, C12, pp. 21989-22008.

Francis, E.A., 1987, "Calibration of the NIMBUS-7 Scanning Multichannel Microwave Radiometer (SMMR), 1979-1984," Oregon State University, M.S. Thesis, 248p.

Gloersen, P., 1983, "Calibration of NUMBUS-7 SMMR: II. Polarization Mixing Corrections," NASA Technical Memorandum, TM-84976.

Gloersen, P. and D.J. Cavalieri, 1986, " Reduction of Weather Effects in the Calculation of Sea Ice Concentration from Microwave Radiances," Journal of Geophysical Research, Vol. 91 (C3), p 3913-3919.

Gloersen, P., et al. 1986, "Reduction of Weather Effects in the Calculation of Sea Ice Concentration from Microwave Radiance," Journal of Geophysical Research, Vol. 91, pp. 3913-3919.

Gloersen, P., 1987, "In-Orbit Calibration Adjustment of the NIMBUS-7 SMMR," U.S. National Aeronautics and Space Administration, Technical Memorandum 100678.

Gloersen, P., W.J. Campbell, D.J. Calvalieri, J.C. Comiso, C.L. Pakinson, and H.J. Zwally, 1991, "Arctic and Antarctic Sea Ice, 1978-1987: Satellite Passive Microwave Observations," NASA Special Publication.

Hollinger, J.P., and R.C. Lo, 1983, "SSM/I Project Summary Report," Naval Research Laboratory, NRL Memorandum Report 5055.

Hollinger, J.P., 1989, "DMSP Special Sensor Microwave/Imager: Calibration/Validation," Final Report, Vol. I.

Hughes Aircraft Company, 1980, "Special Sensor Microwave Imager (SSM/I), Computer Program Product Specification (Specification for FNOC)," Volume II, Sensor Data Processing, Computer Program Component (SMISDP).

Jentz, R.R., C.C. Wackerman, R.A.Shuchman, and R.G. Onstott, 1991, "NASA, Navy, and AES/York Sea Ice Concentration Comparison of SSM/I Algorithms With SAR Derived Values," IGARSS 1991.

Maritn, S.B. Holt, D.J. Cavalieri, and V. Squire, 1987, "Shuttle Imaging Radar B (SIR-B) Weddell Sea Ice Observations: A Comparison of SIR-B and Scanning Multichannel Microwave Radiometer Ice Concentrations," Journal of Geophysical Research, Vol. 92, C7, pp. 7173-7179.

National Aeronautics and Space Administration (Oceans and Ice Branch, Goddard Space Flight Center, 1991, "SMMR North Polar Radiances on CD-ROM," Special Report.

National Snow and Ice Data Center, 1992, "DMSP SSM/I Brightness Temperature and Sea Ice Concentration Grids for the Polar Regions on CD-ROM: User's Guide," Special Report -1.

Russel, C.A., Jentz, R.R. Wackerman, C.C. and Shuchman R.A., 1991, SSM/I Sea Ice Concentration Algorithm Validation," 2nd WMO Operational Ice Remote Sensing Workshop, Vol. 1.

Snyder, J.P., 1982, "Map Projections Used by the U.S. Geological Survey," U.S. Geological Survey Bulletin 1532.

Steffan, K. and J.A. Maslanik, 1988, "Comparison of Nimbus 7 Scanning Multichannel Microwave Radiometer Radiance and Derived Sea Ice Concentration with Landsat Imagery for the North Water Area of Baffin Bay," Journal of Geophysical Research, Vol. 93, C9, pp.10, 769-10, 781.

Steffen, K., and A. Schweiger, 1991, "NASA Team Algorithm for Sea Ice Concentration Retrieval from Defense Meteorological Satellite Program Special Sensor Microwave Imager: Comparison with Landsat Satellite Imagery," Journal of Geophysical Research, Vol. 96, C12 pp. 21971-21,987.