Visualization of Polar Ocean Parameters Using Vis5D


Gary Page and Harold Geller


A paper, computer, and video presentation made in partial fulfillment of requirements for CSI 803, Scientific and Statistical Visualization, George Mason University, Spring 1997, Professor Carr.




Using high resolution model data generated during an eight hour run on the Arctic Region Supercomputer Center (ARSC) Cray Y/MP, we demonstrate how scientific visualization techniques can be used to provide a clearer understanding of the physical oceanographic characteristics of the Arctic Ocean currents. We use model output data provided to us by one of the developers of the Barrow Canyon model, Signorini et al., whom incorporated the so-called "semi-implicit ocean circulation model using a generalized topography following coordinate system" developed by Song and Haidvogel, 1994. The Barrow Canyon research paper is to be published later this year in the Journal of Geophysical Research and it examines the results from the modeling of the flow dynamics of the Arctic Ocean canyon called Barrow Canyon. It also examines in situ measurements of data collected in the summer of 1993 in the Barrow Canyon region and seeks to compare the model output results with these in situ measurements. The results to be published in the Journal of Geophysical Research were provided to us by the principal author and we provide an examination of his visualization techniques and offer an alternative to the visualization of the data using the scientific visualization package known as Vis5D, which is available free from the University of Wisconsin, via ftp or the World Wide Web. The following figure depicts our general approach to this project.




According to the most recent climate forecasts, global warming is predicted to occur at the rate of some 3.5 degrees Celsius per century (Manabe and Stouffer, 1993). This does assume that carbon dioxide continues to increase at the rate of 1% per year as it has over the past decade. However, such a global increase is actually representative of an even larger increase in temperature in the Arctic region (IPCC, 1995). This is due to the fact that any increase in temperature is expected to be followed by a reduction of sea ice cover which causing a reduced surface albedo will absorb more heat in the region. As it turns out, one of us has done research which does not agree with this amplified temperature rise in the Arctic because it has been discovered that an increase in the temperature may increase the amount of precipitation which will in turn increase the amount of sea ice (Geller, 1996). Nonetheless, this makes the study of the Arctic region waters a key factor in the forecasting of the rates of temperature rise over the coming decades.


We were fortunate to have the opportunity to examine the recent physical oceanographic research conducted in the Barrow Canyon region of the Arctic Ocean. This research was led by Dr. Sergio Signorini and revolved around the modeling of the ocean currents using the Arctic Region Supercomputer Center (ARSC) Cray Y/MP in addition to in situ measurements in the region. Dr. Signorini was kind enough to produce an eight hour supercomputer run of his model and provide us with the model output. Dr. Signorini also provided us with crucial details to understand the model output and help visualize the model output data using more advanced visualization techniques than was available to him at the time of the writing of the original research. Dr. Signoriniís paper will be published in the Journal of Geophysical Research this year. However, he was kind enough to provide us with a hardcopy preprint of the paper as well as graphical files of some of the figures to appear in the published paper.


Physical Oceanography Research in the Barrow Canyon Region


Dr. Sergio Signoriniís paper is titled "Flow Dynamics of a Wide Arctic Canyon" and is co-authored with Dr. Munchow and Dr. Haidvogel. Dr. Signorini is now a scientist at SAICís General Science Corporation in Laurel, Maryland where he works with one of us on a Japanese space agency (NASDA) funded program to establish an International Arctic Research Center (IARC) which will be physically located on the campus of the University of Alaska Fairbanks (UAF).


The research involved implementation of an oceanographic model referred to as "a semi-implicit ocean circulation model using a generalized topography following coordinate system." This model was developed by Yuhe Song and Dale Haidvogel and first published in the Journal of Computational Physics in 1994 [Song and Haidvogel, 1994]. The model is described here in Dr. Signoriniís own words:


"We used a three-dimensional, rigid lid, primitive equation model with orthogonal curvilinear coordinates in the horizontal and a new general coordinate in the vertical (SPEM5.1). The stretching of this new vertical coordinate is achieved by using a nonlinear function of z which allows high resolution in the upper ocean while maintaining the bathymetry-following properties of the sigma-coordinate (Song and Haidvogel, 1994). This particular feature is ideal for the application to Barrow Canyon where both steep topography and strong upper ocean vertical stratification are present. The model was initialized from rest (u=0, v=0). The initial temperature (T) and salinity (S) fields were specified by choosing analytical expressions which emulate the observed vertical stratification. The equation of state is non-linear (Jackett and McDougall, 1994) and uses the potential temperature to calculate the in-situ density. The grid has open boundaries at the west, north, and east sides of the domain which must be specified. The model is forced at the southwest boundary by a 40 km wide barotropic jet representing the mean Bering Strait transport of 1 Sverdrup. At this boundary, the along canyon velocity, (u), is specified, the cross-canyon velocity, (v), is set to zero, and constant vertical stratification is imposed. In the interior, the density is initially stratified using analytical expressions, as in the open boundary, but it is allowed to evolve temporally and spatially under the influence of advection and mixing. The jet is modulated sinusoidally with a period of 1.5 days and an amplitude of 0.5 Sverdrup in order to emulate the time dependancy observed in the data. Click here to see a picture from model output which shows velocity vectors at a depth of 30 m, about 15 days after model start. The inset in the picture contains a time series (in days) of the transport (in Sv) imposed at the southwest open boundary; it indicates the time (black circle) at which the velocity vectors snapshot was taken from the model output. At this time, the magnitude of the down canyon velocity field approaches that of the September 25 ADCP survey when the along-canyon volume transport reached 1.1 Sv. Note the presence of a 10-km wide cyclonic recirculation feature on the western side of the canyon. It represents the entrainment of Chukchi shelf water into the canyon from depths less than 100 meters."


In the conduct of his research Dr. Signorini compares the developed model output to measurements made in the fall of 1993, by scientists on the Canadian icebreaker CCGS Henry Larsen (Signorini et al., 1997). Measurements included the standard CTD (conductivity, temperature and density) measurements as well as measurements using the Acoustic Doppler Current Profiler which were described by one of Dr. Signoriniís colleagues in the Journal of Atmospheric and Oceanographic Technology (Munchow et al., 1995).


Dr. Signorini provides the reader in his upcoming Journal of Geophysical Research paper a view of the Barrow Canyon region within which the measurements were made by Munchow et al., 1995. We provide this visualization in Figure 1. Note the use of vectors to indicate velocity. These vectors are actually averaged over the depths of the measurements. Yellow stars are used to depict the relative location of the CTD measurements along the icebreaker track. Green diamonds and white dots are used to depict the locations of the ADCP towed array measurement path for the two days on which these measurements were made.


Figure 1 [Source: Signorini et al., 1997]


Note in Figure 1 how numerous data are depicted simultaneously. From this single figure the reader can determine the local bathymetry, the velocity of the currents measured, the location of the measurement points and the track of the icebreaker and its towed array. However, had the authors used a visualization package such as Vis5D, this figure might have been able to provide the reader with a somewhat clearer picture of the ocean currents since the depiction in this planar view has to average the measurements made at various depths. With a software package such as Vis5D, the individual velocity vectors at various bathymetric depths might be depicted.


Another view of the region is depicted for the reader in Figure 2. This figure allows the reader to locate his position in the Arctic region somewhat better than the previous Figure 1. However, in this case, a small inset of the actual geographic location for the general reader might have added to the information conveyed in this figure. Note the use of the longitudinal and latitudinal gridlines with a red colored boundary depicting the study region of the physical oceanographic model. This may have been improved had the authors included the icebreaker track within the red colored boundary, again for reference for the general reader, thus making Figure 1 a "blowup" of the study region with additional details available.

Figure 2 [Source: Signorini et al., 1997]


While the Signorini et al., 1997 paper is geared towards the professional researcher in the physical oceanographic field, the general reader may benefit from additional information about ocean currents and their influence on climate, providing a better understanding of why this work is important even outside of its field of specialty. Scientists consider the distribution and flow of water on this planet as a powerful indicator of the state of the atmoshpere-ocean climate system. In the late 18th century, Count Rumford was one of the first to observe and understand that masses of cold water, which were observed even in regions where the surface water never gets cold, were part of a global ocean overturning circulation cell. In fact, today with better models of this phenomenon now referred to as the "great ocean conveyor belt," it is apparent that the driving force is the convection currents in the polar regions of the world. A view of the "great ocean conveyor belt" is depicted in Figure 3, taken from the University of Victoria, Canada Climate Modeling Group.


Figure 3 [Source: University of Victoria, Canada, Global Modeling Group]


Model Output Results


Here we present and review some of the products used in the upcoming published paper to assist the reader in the visualization of the model output results. In Figure 4, the reader is presented a visualization of the ocean water density along the central transect of Barrow Canyon. The horizontal axis depicts the distance from a reference point in kilometers while the vertical axis depicts the depth in meters. The reader is also presented with a contour visualization of the velocity field that developed from the running of the model, and although not depicted, it was compared to the velocity field derived from the in situ measurements. The authors also used a color representation in which blue is lighter surface waters and red the deeper denser waters. This color depiction, a standard rainbow metaphor is viewed dimly by some cartographers [Tufte, 1997]. Edward Tufte notes in his latest book that such color codes are "often found in scientific publications" and that "such a visually naïve color-scale would be laughed right out of the field (or ocean) of cartography [Tufte, 1997]. The depiction is called "unnatural and unquantitative" causing the reader to lose some of the data "in the soup" [Tufte, 1997]. Tufte offers an alternative which consists of various shades of blue vice the complete rainbow of colors. Nonetheless, the information displayed by Signorini et al., if not intuitively obvious as Tufte would lead you to believe, is available for the readerís benefit. Perhaps, incorporating a third dimension in the figure would further relieve the criticism that Tufte might have of such a visualization.


Figure 4 [Source: Signorini et al., 1997]


Figure 5 depicts the cross-section of the modeled velocity field superimposed on the modeled density field. Again the criticism of Tufte to such rainbow scale would apply here as well as in Figure 6. In Figure 6, the reader is presented a depiction of the cross-section of the modeled density field with the modeled thermal wind velocity field superimposed.


Figure 5 [Source: Signorini et al., 1997]


Figure 6 [Source: Signorini et al., 1997]


Finally, we present Figure 7 from Signorini et al. Which depicts the time average (1.5 day cycle) of the velocity and density fields along four cross sections of Barrow Canyon. Sections 1 through 4 are respectively 30, 15, 0 and -15 kilometers southward of the location of the central ADCP transect depicted in Figure 1. In Figure 7 the authors depict the density field in the rainbow colors where red is denser and blue less dense. Superimposed upon the density field is the vertical profiles of the total (barotropic plus baroclinic) velocity within the four regions. The authors conclude that the major driving mechanism of the up-canyon current flow is the "non-linear interaction of the variable barotropic flow with the steep bottom topobraphy." [Signorini et al., 1997]


Figure 7 [Source: Signorini et al., 1997]


The authors note the "the model and thermal-wind-derived cross-canyon velocity fields are very similar." This allows them to conclude that the developed model and approach taken are validated to the degree presented numerically. While the authors do provide some side by side comparisons, in numbers and visualizations, a more visual rendition of this comparison in a single figure would have highlighted the differences better for the general reader.


Discussion of Vis5D


While the Signorini research team produced a number of quality output graphics for the Journal of Geophysical Research for publication, none of the published graphics make use of three dimensional data viewing. Three dimensional displays simply appear to appeal to human eyes and brains as providing a "clearer" view of the data being visualized, which in this case represents an analogue to a physical reality. To produce displays in three dimensions on screen and in hardcopy, we turned to a visualization package known as Vis5D, which was being used by scientists at the University of Alaska Fairbanks, with whom one of us work on a regular basis.


Simply put, Vis5D is a scientific visualization software package available free from the ftp archives of the University of Wisconsin. For brevity and to ensure an accurate overview of the software, we provide here a description of the software as provided by its authors in the "README" file that accompanies the software when downloaded from the University of Wisconsin ftp archives.


"Vis5D is a software system for visualizing data made by numerical weather models and similar sources. Vis5D works on data in the form of a five-dimensional rectangle. That is, the data are real numbers at each point of a "grid" which spans three space dimensions, one time dimension and a dimension for enumerating multiple physical variables. Of course, Vis5D works perfectly well on data sets with only one variable or one time step (i.e. no time dynamics). However, your data should have some depth in all three spatial dimensions.


"The Vis5D system includes the vis5d visualization program, several programs for managing and analyzing five-dimensional data grids, and instructions and sample source code for converting your data into its file format. We have included the Vis5D source code so you can modify it or write new programs. We have also included sample data sets from the LAMPS model and from Bob Schlesinger's thunderstorm model, so you can work through our examples.


"Vis5D version 1.0 was written by Bill Hibbard and Dave Santek of the University of Wisconsin Space Science and Engineering Center, supported by the NASA Marshall Space Flight Center, and by Marie-Francoise Voidrot-Martinez of the French Meteorology Office. Later version enhancements were written by Bill Hibbard, Brian Paul, and Andre Battaiola. Dave Kamins and Jeff Vroom of Stellar Computer, Inc. provided substantial help and advice in using the Stellar software libraries. Simon Baas and Hans de Jong of the Netherlands ported Vis5D to HP workstations. Pratish Shah of Kubota Inc. ported Vis5D to the Kubota Alpha/Denali workstation. Mike Stroyan of Hewlett Packard added PEX support.

"Vis5D is offered under the terms of the GNU General Public License, which you can find in the file "NOTICE". As the notice states, there is no warranty for the Vis5D system, but we would be interested in hearing about your questions and problems."



Discussion of Data Formats and Vis5D


A recent text on graphics files formats [Brown and Shepard, 1995] provides the reader with details on no less than 47 different formats. The various formats differ in their implementation of data compression, color table representation, and other parameters associated with the details of a graphics file format. In our specific case, Dr. Signorini provided us with a data file that consisted of columns of model output data in a form which Vis5D could not handle. Vis5D makes use of its own file format which users must convert their data to, in order to make use of the visualization package. The authors of Vis5D provide all comers with a skeleton of a program to perform the file format conversion of the usersí data into the Vis5D file format. In our effort, two different files were required to be converted into Vis5D format. The first file to be converted was the Department of Defense ETOPO5 file consisting of bathymetry and topography of the world. The second file to be converted was the Signorini model output file. The approaches used are summarized in the figure incorporated into the abstract above with some additional discussion provided the reader in the discussion of the model output data, in a section of this document below.


Displaying Topographic/Bathymetric Data in Vis5D


As noted in the previous section, topography and bathymetry data from the Department of Defenseís ETOPO5 dataset had to be extracted and formatted into a Vis5D readable file format. Figures 8 and 9 depict the Vis5D visualization of the bathymetry of the Barrow Canyon model study region. Figures 10 and 11 depict the more general view of bathymetry and topography of the Arctic region.

Figure 8


Figure 9


Figure 10


Figure 11


Converting Model Data into Vis5D Format


Dr. Sergio Signorini provided us the output data from the eight hour model run in an ASCII formatted file that had a unique structure. The first column provided the time in hours from the start of the model run. The second and third columns provided the longitude and latitude, respectively. The fourth column consisted of the vertical coordinate, in the model coordinate system expressed in meters. The fifth column consisted of the depth expressed in meters. The sixth column consisted of the east component of the ocean current velocity vector at model level k expressed in centimeters per second. The seventh column consisted of the north component of the ocean current velocity vector at model level k expressed in centimeters per second. The eighth column consisted of the vertical component of the ocean current velocity vector at model level k expressed in centimeters per second. The ninth column consisted of the water density parameter expressed in grams per cubic centimeter. Finally the tenth and eleventh columns contained the model coordinates for the respective rows. Furthermore, there were 25 levels of data provided each with 2127 grid points per level.


Converting this dataset into a dataset that could be used by Vis5D consumed the largest portion of the effort for this project. To accomplish this feat, one of us developed a C program which performed a polymonial interpolation of the model depths. This C program output was used in a shell script which allows the user to define a latitude/longitude grid for producing a file output of times and depths extracted form the output from the C program. Next, the data required the development of an output file which was uniformly distributed throughout the level and the depths. The software package Splus was used to perform this uniform gridding, which itself was based upon an algorithm first published by Akima, 1978.


Finally, the Splus file output required re-formatting to an ASCII tabular form that could be converted into a proper Vis5D file format. Once in Vis5D file format, the model output data could finally be properly visualized with the downloaded software.


Results of Whatever Model Data we can format in Vis5D


This section awaits the final completion of the conversion of the model data provided by Dr. Signorini. It is hoped that graphics displays will be available by 7 May 1997.


Generating an Animation in Vis5D


The generation of an animation in Vis5D is a simple matter if your data is already in the Vis5D file format. The file format utilized by Vis5D supports different time samples of the same data. Thus, if properly formatted, any Vis5D data file can be animated by a simple click of the proper control buttons.


Generating a Video Using Vis5D Visualizations


In preparation for the generation of a video depicting our experiences and efforts using Vis5D in the visualization of the scientific modeling and data made available by scientists and data distribution centers, we developed what we call a script. This script listing depicts the scene sequence that we hope to capture on screen using the equipment made available to us by Professor Carr and the "holodeck" visualization laboratory of the SITE center of George Mason University.


In the first scene we display a copy of the HTML file developed from the paper that we produced for this project. In the second scene we review the work of Dr. Sergio Signorini as published on the World Wide Web at the Arctic Region Supercomputing Center. Scene three highlights the software package with which we are working in this project, that is, Vis5D. We use the Vis5D website at the University of Wisconsin which contains some static displays developed with Vis5D. Scene five switches to our own compiled version of the Vis5D software and demonstrates the visualization power of the software using demonstration datasets provided with the downloaded software. Scene six depicts the Vis5D bathymetry dataset that we extracted from a Department of Defense product known as ETOPO5. Scene seven demonstrates how a resampled subset of the ETOPO5 dataset depicts the study area of the Signorini research team. Scene eight demonstrates the model output data that this effort developed using the techniques discussed earlier in this paper.


We had hoped to have additional scenes in the video, but time did not permit their completion. These scenes were to depict the recently released Russian datasets from the National Snow and Ice Data Center which we procured (at a $50 cost) for this project. Also, the additional scenes were to depict the Russian data in the Vis5D visualization software package and how this measured dataset compared to the Signorini teamís model output data. Finally, we had originally hoped to include in the video a depiction of how Vis5D could be demonstrated over the World Wide Web using any web browser as the front end to the software on a remote server.




We found that the most difficult portion of the process of utilizing a three dimensional visualization software package is the proper formatting of the data file itself. With the use and availability of freeware or public domain visualization software, any user with the properly formatted dataset can view their data in a myriad of three dimensional perspective views.




Based upon our experiences during the conduct of this project we would recommend that commercial ventures investigate the development of file format conversion software which would allow the transfer of data from one software application to another. Furthermore, we have found the software application developed at the University of Wisconsin, known as Vis5D is a powerful visualization software package which anyone can obtain free from the university. Naturally, we have numerous recommendations for the improvement of the software, however, the power of this software and the availability of the source code make it a very attractive package. Furthermore, the university has a project underway in which the Vis5D package is being used to develop a web-based Java application called VisAD which not only will allow for the visualization of data but also for the further analysis of the data, just with the use of a web browser on any platform on the World Wide Web.




Akima, H. (1978). A Method of Bivariate Interpolation and Smooth Surface Fitting for Irregularly Distributed Data Points. ACM Transactions on Mathematical Software, Volume 4, pp. 148-164.


Brown, C.W. and B.J. Shepherd (1995). Graphics File Formats Reference and Guide, Manning Publications Company, Greenwich, Connecticut.


Munchow, A., C.S. Coughran, M.C. Hendershott and C.D. Winant (1995). Performance and calibration of an acoustic Doppler current profiler towed below the surface, Journal of Atmospheric and Oceanographic Technology, Volume 12, pp. 435-444.


Signorini, S.R., A. Munchow and D. Haidvogel (1997). Flow Dynamics of A Wide Arctic Canyon, Journal of Geohpysical Research, to be published.


Song, Y. and D. Haidvogel (1994). A Semi-implicit Ocean Circulation Model Using a Generalized Topography Following Coordinate System, Journal of Computational Physics, Volume 115, pp. 228-244.


Tufte, E.R. (1997). Visual Explanations, Images and Quantities, Evidence and Narrative, Graphics Press, Cheshire, Connecticut.