Introduction

The main goal of this project is to convert the regional NIC ice charts to a format for scientific investigations that is easier to compare with existing formats such as the NSIDC ice concentration images from SSM/I, and that provides an Arctic-wide view of ice concentration.  This short report will explain the current status of the project and will give some examples, description of the methods, difficulties encountered, and the further work that will be needed.

Background

The National Ice Center produces ice charts approximately on a 7 day schedule, with some areas more often.  These charts are made available over the internet in gif format for 20 main geographic areas.  Arcinfo is used to generate these charts.  The charts that are distributed are vector drawn maps with "egg" codes associated with the different ice conditions.  These "egg" codes provide very detailed information about the ice conditions:  concentration, type of ice, age of ice, etc.  The data is also put into a GIS interchange format called e00 files.  The e00 files contain all the necessary information to recreate the ice charts.  There is an e00 file for each geographic area, and each e00 file may contain many areas of ice with differing ice parameters.

Processing Steps and Current Status

All of the e00 Arcinfo files that were found on the NIC ftp site were downloaded.  Depending upon the time of day, the NIC server would hang quite often during the transfer, requiring many restarts.  The downloading improved tremendously if the transfers were done during non working hours.

 A fairly complete format description for e00 files was found on the internet. The e00 files contain several sections, of which three were used.  Arcinfo, as the name implies, puts all GIS information into arc segments.  By combining these arcs, polygons can be generated with unique associated parameters.  The three e00 sections from the NIC files are the arc segments, information to combine the arcs into polygons, and the ice information for each polygon (This information is a SIGRID encodeing of the egg code for the polygon). Each arc segment consists of x and y positions in meters from the pole, based upon a polar stereographic map projection.  The first step in the processing at NSIDC is to convert these positions to latitude and longitude values and to generate complete polygons for each unique ice area.   Special handling was needed to handle certian  situations in the e00 files.  For example, where an area of one ice condition has islands of another type inside the first areas's polygon, smaller "doughnut" polygons are included as part of the description and must be taken care of differently.  Also, the polygons may be described as following a sequence of arcs clockwise or counter-clockwise.

The next step is to combine the separate geographic areas into a single scene for the entire Arctic.  An azmuthal equal-area (AEA) projection was chosen for the final product to correspond to the data now being used for most image data at NSIDC (the EASE-GRID).  The lat/lon positions are converted to locations on the 12.5 km AEA grid.  The grid cells inside the polygons are filled with values corresponding to the total ice concentration for that area.  All processsing at NSIDC for this work is done with the IDL programming language.

Total ice concentration images have been generated for 1998 and 1999.  The starting date is day 1 of each year, and continuing every 7 days for the rest of the year.  The 7 day time frame was chosen because this is approximately the cycle NIC uses for updating the ice charts.   For each Arctic-wide image, the most recent NIC chart day was used.  Therefore, as an example, if an image was done for January 14, the NIC charts closest in time prior to January 14 would be used.

Examples and Problems

There are e00 files for 1997, 1998, 1999, and 2000.  The major problem that was found is the missing, incomplete e00 files, or e00 files that appear to not be in e00 format and are therefore not readable.  The following figure shows the usable data that was found on the NIC web site.

The gap in temporal coverage for 1997 appears to be from 3 causes.  Some files are missing a section that contains the ice parameters.  However, most of the unusable files are not in ASCII and are thus unreadable as e00 files., and no way  could be found to work with this data.

Coverage was fairly complete for 1998 and 1999 until the middle of December.  The coverage for most of the areas ends in December.  The NIC web site was checked a couple of times, but the missing e00 files from December 1999 to the present could not be found.  The "missing data" may possibly be archived in a different location.

Overall, the quality of the information in the e00 files was very good.  Since this is an operational product for near real-time use, there are some discrepancies that appear when full Arctic Basin images are generated.  Because separate geographic areas may be done at different times and by different people, the area boundaries may not always match exactly (Next figure).

At times, some polygons may be missing from a region as shown in the next figure.  The e00 file for the HIGH Arctic West was compared with the online NIC gif chart.  The 2 charts look correct. The e00 file for the High Arctic West area is data merged from 2 NIC charts, HIGH ARCTIC 1 and HIGH ARCTIC 2.   A possible explanation is that once in a while, something is lost when the charts are combined.  The full images seem to correspond to the e00 files, so it appears that the e00 files are incorrect.

A polygon may appear to be mislabeled.  That is, it may have an obviously incorrect total concentration.  It may actually be mislabeled, or a unknown unique special case in the e00 file may occur.  This hasn't been fully investigated at this time.  An example of "mislabeling" is shown in the next two figures.  (1) south of the Davis Strait    (2) Beaufort Sea.  The online gif charts look correct, but the e00 file seems to have the wrong ice code for the upper Beaufort Sea area.

For a general feel of the quality of the final images in the NSIDC format, an animation of weekly images for 1998 and 1999 can be downloaded.  This is a fairly large file of about 3.2 megs and may take a little time to download.  It is an animated gif file.  This is just a rough stab at where the project is currently.
A link to the animation: ANIMATION
 

Project Continuation

Issues:

1.  Unusable or Missing data:  It may or may not be a major effort on the part of the NIC to redo the 1997 e00 files.  Are there current e00 files from December 1999 to the present, or, again, would it be a major effort by NIC to generate them.

2.  Re-projection of the data:  The algorithms used to re-project the NIC polar stereographic data into the azmuthal projection is not exactly correct.  A small file of a few geographic locations (tie points) would be needed from NIC to correct this problem.
The file would need to contain a few latitude and longitude pairs along with the corresponding Arcinfo x and y values for projection.  This would then be used to correct the geo-location problem.

3.  Missing and mislabeled polygons:  This may or may not be a major problem.  At this time, the problems appear to be in the e00 files.  In each case, the online NIC charts look correct and consistent.  Some specific cases will need to be investigated between the NIC and NSIDC to resolve this issue.

(As a last resort the 2 following options could be possible).
I showed the images to Jim Maslanik who has used the NIC charts previously in his research.  His suggestion was to not attempt to fill in missing or incorrectly labeled areas.  His reasoning was any researchers using the data would understand that this is basically an operational product, and they should treat it as such.  The small percentage of the time that there is missing or incorrect data would reinforce that fact.

Another possibility is to manually fill in missing polygons from nearby data, or from the next previous NIC chart.  This could involve a fair amount of manual labor, but is not impossible.

4.  Quality control:  A formal set of QA depends upon how the first 3 issues are resolved.

5.  Project continuation:  Since most of the current 1997 e00 files are unusable, and if e00 files will not be generated in the future, should this project continue with only 1998 and 1999 data.  Or, should it be ended with only these 2 years being done.

6.  An ancillary file will be generated giving the dates of the e00 files used for each location.  A third file may be created showing the complete egg codes for each area.  This would be used if an interactive system is developed.  By clicking in an area, all the NIC chart data would be available.  This would be used for informational purposes, but not  as a formal or supported product.