1. Data Set Preparation
The following describes the various data processing and assembly tasks that were performed during Year 1. The domains for data coverages discussed below are illustrated in Figure 1. Our web site hosting many of these data is discussed in Section 2.
Figure 1. Areas of coverage of data sets being developed for this project. The figure at left shows the full AVHRR Polar Pathfinder area of coverage, with the selected SHEBA model-grid domain indicated by the red-dashed box. The SHEBA satellite-data subset is indicated by the blue box. The graphic on the right highlights the model-grid subset using ETOPO5 digital terrain data. Also being provided, but not shown here, are data for 55-km x 55-km locations corresponding to the SHEBA drift track.
1.1 Satellite-Derived Data Sets
Gridded, twice-daily AVHRR Polar Pathfinder (APP) data available for 1981-1998 were processed to fill in data gaps during the SHEBA year, include a previously missing year of coverage, and to apply revised calibration coefficients and an improved cloud detection algorithm. The characteristics of the APP data are described in the online documentation available via NSIDC ((http://nsidc.org/NSIDC/CATALOG/ENTRIES/nsi-0066.html for the 5-km products). The processing was extended to include data through December 2000. We thus have a full APP data set available for July 1981 - December 2000, which in addition to documenting conditions during the SHEBA period, provides the ability to contrast 1997-1998 with other years. A subset of these data covering the western Arctic has been extracted from the time series and is available online. The area of coverage is 3250 km x 2625 km (or 650 pixels by 525 pixels at 5-km resolution). Data are presently available at 5-km and 25-km resolution, with 1.25-km subsets also available for selected days. (The full APP data set, which covers the northern and southern hemispheres above approximately 50 degrees latitude, is archived at NSIDC). The APP product set placed online for SHEBA investigators includes a subset of the full APP product suite. The SHEBA data sets provided are clear-sky skin temperature and albedo, viewing angle and observation time, surface type masks, and cloud detection flags using 3 alternative algorithms. The data set also includes SSM/I-derived ice concentration, ice type and ice motion vectors. In addition to the SHEBA grid subsets, we have generated a Lagrangian subset for a 55km x 55km area surrounding the SHEBA site along its drift route.
To complement the standard APP products, a suite of APP-derived data sets have been prepared by J. Key at the University of Wisconsin. This data set consists of AVHRR retrievals of surface and cloud properties as well as radiative fluxes for September 1, 1997 to August 3, 1998. Cloud and surface properties were estimated from the five-kilometer APP data. The files are provided as 307 individual daily images and monthly means. All images are centered on 1400 local solar time (high sun), though the pixels within any given image may have been acquired 1-3 hours before or after that time. The area covered by these products coincides with the "satellite data subset" area shown in Figure 1. The retrieval algorithms of the Cloud and Surface Parameter Retrieval (CASPR) system were applied. The following parameters are available:
Surface and Clouds:
Figure 2. Time series of CASPR all-sky surface (skin) temperatures for the SHEBA region, as estimated using the AVHRR Polar Pathfinder data (dashed line), compared to measured in-situ temperatures (solid line).
We have also obtained the ISCCP D2 total cloud amount product. These data extend only through December 1993, but provide an additional frame of reference for the AVHRR- and surface-observed cloud conditions during SHEBA. The ISCCP data are in the process of being converted to our common map grid.
1.2 Other Data Sets
The set of model forcing fields prepared by the POLES project have been re-gridded and subsetted to correspond to the model grid in Figure 1. The original data and documentation are available at http://psc.apl.washington.edu/POLES/model_forcings/ModelForcings.html
A bathymetry and topography data set (shown earlier in Figure 1) mapped to the SHEBA modeling grid was created from the ETOPO5 digital data base of land and sea floor elevations obtained from the National Geophysical Data Center. (see http://www.ngdc.noaa.gov/mgg/global/etopo5.HTML for details on the original ETOPO5 data).
1.3 C-130 Data and High-Resolution Surface Characterization
Basic statistics of lead characteristics were derived from SHEBA C-130 airborne imaging microwave radiometer (AIMR) data by Dr. Mark Tschudi. These data include lead widths, orientation and other characteristics summarized in tabular form for the SHEBA area. These summaries are available on our web site (see Section 2). The value of the SHEBA aircraft data for surface mapping can be further exploited by georeferencing and assembling images and profile data from the various flight tracks into single coverages for each flight number. These data can then be combined with other data sets such as RADARSAT imagery, APP products, etc. to provide detailed spatial maps of surface conditions. Standard remote sensing packages are well suited to georeferencing and merging relatively small numbers of images. However, we have found that processing the numerous image subsets and profiles from SHEBA is best done using a tailored set of routines that we have developed in IDL. Assembly of the C-130 products using this approach is now underway. The effort to combine these data with other data sets and to extract of statistics on surface conditions is being shared with a separate NASA grant. The results will be useful for surface characterization during the SHEBA spring and summer period and will be provided to SHEBA investigators. Lead statistics for other portions of the year will be obtained from RADARSAT and AVHRR imagery.
2. Data Visualization, Selection and Delivery
As part of our effort, we are delivering our prepared data sets to NCAR JOSS. Most data sets are provided in NetCDF format as recommended during the last SHEBA meeting. However, while JOSS provides a valuable access route for these data, we have found in our own work integrating data analysis and modeling that a need exists for users to have the ability to visualize products in simple forms, as well as having access to some basic analysis steps and data set selection tools. This is particularly the case for our project, which seeks to greatly improve the usefulness of a variety of data sets for SHEBA investigators. These investigators have a wide range of skills and familiarity with data sets and processing. Basic provision of data files may be adequate for some of these users, but others may require additional support. In particular, since SHEBA projects focus on the study of processes and interactions, investigators need access to multiple data types. In the past, most of such data have been available in different formats, arrangements, etc., which complicate and discourage their use. Our goal is to simplify the applications of these data by presenting the data sets in a consistent and easy-to-access form. We also intend to encourage the use of the data by providing an interface that allows investigators to see examples, create plots, and select temporal subsets of all our provided data. Good examples of this type of system for climate data sets is the existing on-line access tools provided by NOAA's Climate Diagnostics Center (CDC) (http://www.cdc.noaa.gov) and the ISCCP pages at http://isccp.giss.nasa.gov. We do not have the resources to provide quite this level of sophistication, but a simpler interface that follows the general philosophy of such sites is feasible.
Toward this end, we have begun work on a web-based access route for our data sets. The site is located at http://polarbear.colorado.edu/sheba.html. This site currently provides access to the pathfinder-gridded AVHRR-derived products and the re-gridded and subsetted POLES data. Other data are being added as they become available. The site also provides various plots and summaries of monthly means, seasonal statistics, etc., as well as links to other related sites. We also intend to provide tools to allow users to generate on-demand plots and images.
3. Modeling Activities
For our project, modeling serves as a tool to provide information not readily obtained from direct observation for large areas, and as a data assimilation method to generate improved estimates of spatial and temporal variations in parameters such as ice thickness distribution, turbulent fluxes, and net radiation. Our original intent was to use a 3-D ice-ocean model developed in house for our work. However, discussions with other investigators suggested that a switch to the NCAR Community Sea Ice Model (CSIM) would be a better approach for the long term. The CSIM represenats a community-wide effort to bring about a sea ice model which can be used and compared by various groups. It also represents the state of the art in basin-scale sea ice modeling. CSIM provides a good testbed for testing and refining our data assimilation algorithms. By carrying out our assimilations using a CSIM clone, we can readily show other modelers that the methods we are working on can be included in the CSIM and what costs/ benefits their inclusion will yield. Furthermore, it will be easy to include CSIM updates as they become available. Also, the same model is also being included in ARCSyM, so coupled-model testing will be possible in the future.
We have obtained the latest version of CSIM from C. Bitz and have configured the model for the equal area (EASE-grid) domain shown in Figure 1. This includes 25, 50, and 100km versions of the model. As a initial test set, we are using UW POLES sea ice forcing dataset noted in Section 1.2, which includes 1979-1998 atmospheric forcing (downwelling shortwave and longwave radiation, air temperature, specific humidity, and wind speed) [based on NCEP and buoy measurements and annual cycle of ocean (ocean heat flux and currents) [based on Zhang's ocean-ice simulation. The model has been upgraded with the Alternating Direction Implicit VP dynamic solver (Zhang and Rothrock 2000), which operates much faster than LSOR (Zhang and Hibler 1996). We have also added simple melt pond and surface blowing snow parameterizations to provide more realistic surface physics (Arbetter et al. 2001).
Testing of the model, including the new pond and snow parameterizations, is still underway. Problems have been encountered in obtaining stable simulations. These are due to primarily to inadequacies in the atmospheric forcing (poor assumptions for specific humidity) and in the snow physics. We are attempting to address this using several approaches, including revised treatment of snow sublimation.
4. Climatological Analyses
The goal of this component of the research is to document the representativeness of the SHEBA region in relation to other years and portions of the Arctic. Using the data sets described above, we are calculating basic statistics such as long-term means and variability, and then contrasting these statistics with conditions during the SHEBA year. These results will help place the SHEBA experiment in the broader climate context, and should assist other investigators in assessing the degree to which model processes and parameterizations are likely to be applicable to other regions and climatic conditions.
Figure 3 shows an example of our analysis of AVHRR Polar Pathfinder-derived cloud fraction, using APP data for 1981-1998 (Drobot et al., 2001). Our results suggest a general increase in cloudiness over much of the Arctic Basin (including the SHEBA region), in conjunction with a general decrease in sea level pressure. This decrease in SLP has been noted previously as a component of the NAO and AO, but the apparent correlation with cloud cover illustrates the broader range of effects associated with these large-scale circulation modes. For the SHEBA period in particular, cloudiness was generally above normal, particularly in spring 1998 (Figure 4). Additional analyses using these data suggest that monthly mean cloud fraction in the SHEBA area in spring is negatively correlated with the AO.
Figure 3. Trends in monthly mean cloud fraction for
the Arctic for Aoril, as estimated using the AVHRR Polar Pathfinder cloud
product.
Figure 4. Anomalies in AVHRR Polar Pathfinder-derived monthly mean cloud fraction for the SHEBA sub-region (Figure 1) for October 1997 - July 1998.
In terms of ice conditions, our work to date has examined ice motion, ice concentration, and ice type. The comparison of mean cold-season (October - April) ice motion and multiyear ice concentration (e.g., Figure 5) suggests that while the basic patterns of ice drift and distribution of the old ice pack in the SHEBA period are similar to the long-term means, some notable differences exist. In 1997-1998, the east-to-west drift in the Beaufort Sea is more prominent, with faster drift speeds and less southward drift along the eastern portion of the Beaufort Gyre. Less old ice is present in 1997-1998, with a considerable reduction in old ice concentration in the Chukchi Sea. This area corresponds to an area where ice extent was much less than normal in the summer of 1998.
Figure 5. Passive microwave-derived ice motion (white vectors) and multiyear ice coverage (green through pink tones) for cold-season months (October - April). Long-term means (1978-2000) (left). Means for the SHEBA period (right).
References:
Arbetter, T.E., J.A. Curry, A.H. Lynch, A. Alam, and J.J. Cassano, 2001. Improved treatment of surface processes in a dynamic-thermodynamic model. . Proc. 6th. Conf. on Polar Met. and Oceanogr., San Diego, CA, 14-18 May.
Drobot, S.D., J.A. Maslanik, and C. Fowler, 2001. Spatial and temporal variations in monthly averaged cloud cover based on AVHRR Polar Pathfinder data. Proc. 6th. Conf. on Polar Met. and Oceanogr., San Diego, CA, 14-18 May.