16.11 Developing and Testing a 3D Visualization Workstation Application at FSL

Paula T. McCaslin*, Philip A. McDonald* and Edward J. Szoke *

NOAA Forecast Systems Laboratory   Boulder, Colorado

 

 

1. INTRODUCTION

 

Visualization is the process of transforming numeric data into a visual form, allowing users to observe the information. Scientific three-dimensional (3D) visualization is the ability to display, analyze, manipulate and interact with 3D data in 3D space. NOAA Forecast Systems Laboratory (FSL) has supported development of 3D visualization software and software applications since 1990. Until recently, the emphasis of this development has been on research applications. Using commercial visualization software called Application Visualization System (AVS5), FSL displayed and investigated both analysis and forecast 3D gridded data. The software was used for the visual analysis of data and for scanning data for the presence of desired features. For example, plotting station observations with forecast-generated data allows visual comparison of the two. In addition to using AVS5 for research, the software is continually used to automatically create images of Local Analysis and Prediction System (LAPS) analysis and forecast output on an hourly basis for the World Wide Web.

 

Custom software modules were developed at FSL to enhance and extend AVS5 to more appropriately display meteorological data. One of these modules displays contours on vertical and horizontal cross sections or fitted to the topography. Another plots wind barbs and arrows. A third renders topography and political boundary information. These modules were developed to improve existing visualization methods, create new ways to represent the data visually, and to add location and orientation cues. With these tools researchers examined data displays to detect regions of specified data values, such as high, medium, low, their shapes and their locations.

 

FSL's 3D images and applications containing atmospheric data have been demonstrated on numerous occasions. They have also been used in publications and presentation materials. With the success of 3D visualization in the research context, a next step was to determine if there was value added by using 3D visualization in an operational forecast setting. To accomplish this it was necessary to develop a 3D visualization workstation application, allow forecasters to use the application in an exercise, and evaluate feedback from the exercise.

 

The Display-3D, known as D3D, was developed to investigate both the complexities and 3D structure of atmospheric parameters and the potential value added by its use in an operational forecast setting. It was designed as an advanced and experimental workstation to be used with WFO-Advanced D2D (MacDonald, 1996). The scope of this paper covers the development of the application and preliminary results of the D3D real-time, RT98, exercise.

 

2. BUILDING DISPLAY-3D

 

Early in 1997 a decision was made to use a software product known as Vis5D, developed at the University of Wisconsin, as the core of D3D. Having had success with AVS5 in the past, our first strategy was to use AVS/Express, an object-oriented visualization development environment from Application Visualization Systems. Vis5D superseded AVS/Express as the core software for D3D due to its superior animation capabilities, and the fact that Vis5D is an application specific to atmospheric science, eliminating the overhead found in a multidisciplinary application. Animation and performance were weighted heavily because of the sensitivity of the human visual perception to motion and the ease with which Vis5D (Hibbard, 1991) could handle these tasks. We explored alternatives including writing our own software or customizing an available visualization package before choosing Vis5D. The Vis5D development at FSL is performed in collaboration with the University of Wisconsin. This allows general purpose enhancements developed at FSL to be integrated into standard releases of Vis5D and minimizes the divergence between the two versions.

 

2.1 Graphical Software Development

 

Because D3D is intended to be used as a component of WFO-Advanced, it is imperative that it coexist with the other WFO-Advanced applications, principally D2D, with a minimum impact on workstation resources. To meet this requirement, Vis5D needed to be modified. The first step was to modify Vis5D to be able to access the WFO-Advanced database. This allowed access to all of the D2D basic and derived gridded data without necessitating the creation of intermediate Vis5D files. In a research environment, Vis5D is typically used to read one data file at a time. However, in D3D, Vis5D would need to be reloaded with model data without being restarted. Although Vis5D has this capability, its memory management and reinitialization functions needed to be refined. Since the D3D user interface was designed to operate as a process separate from Vis5D, the problem of communication between the two processes needed to be addressed. This was solved by incorporating an event-driven interprocess communication scheme in both the interface and Vis5D.

 

Once the issue of compatibility with WFO-Advanced had been addressed, attention was directed toward refining and enhancing Vis5D's display capabilities. FSL has made every effort to accomplish these enhancements by using Vis5D's Application Programmer's Interface (API) whenever possible and minimizing changes to Vis5D's core. Some of these refinements and enhancements include:

 

 

2.2 User Interface Software Development

 

A major development effort was made to replace the original Vis5D user interface designed for research meteorologists with one more suitable for forecast meteorologists that closely resembled the interface that forecasters would already be familiar with. To this end, a new D3D user interface was developed in Tcl/Tk. Though the D2D and D3D applications focus on 2D and 3D displays respectively, it was suggested that where possible the applications have the same look, feel, and function. The design approach included a significant amount of prototype development which was the foundation for further development (Grote, 1997).

 

An efficient result of having both graphical user interfaces written in Tcl/Tk allowed source code for some of D2D's user interface to be directly used in D3D. To illustrate, a tool created to control the attributes of, say, the animation function (called Loop properties) is invoked from the toolbar menu in both D2D and D3D. The Loop properties graphical interface has the same look, feel, and function in both applications because the source code that creates the user interface differs only in the resource call where D3D sends API messages to Vis5D to effect changes in the looping speed, direction, and delay intervals. Thus by reusing source code, where possible, the progress of software development was increased and the approach aided users familiar with D2D in learning the system.

 

The D3D user interface (Fig. 1) consists of three basic areas: a narrow menu bar at the top of the screen, a large context display window to the right, and the D3D Volume Browser.

 

Figure 1. D3D User Interface showing the menu bar, the context display window, and the D3D Volume Browser.

 

The Volume Browser, invoked from the menu bar, provides access to numerical models and gridded data sources (currently limited to volume radar). Through the Browser interface, the user can select the data source, fields, and rendering techniques to generate a customized list of graphics to display.

 

The choice of rendering techniques sets the D2D Browser apart from D3D's Browser. The rendering techniques available in D3D include: isosurfaces, vertical and horizontal cross sections contours and images, surface contours and images, volume visualization, a vertical sounding plot and a virtual data probe. A key interface design decision for the D3D Browser was to pair field selection with rendering technique selection. This eliminated the implicit redundancy in the customized list of graphics available to display (Fig. 2). With the amount of information present in 3D visualization it is important to avoid automatically adding products not explicitly requested by the user (Fig. 3).

 

Figure 2. D2D Volume Browser showing redundancy in customized list of graphics.

 

Figure 3. D3D Volume Browser showing no redundancy in customized list of graphics.

 

Additionally, as each selected item is added to the customized list, a property editor is created that can be accessed by a click of the right mouse button. This feature allows a user to modify default data values, value ranges and colors.

 

3. REAL-TIME EXERCISE

 

A limited real-time forecast exercise, RT98, to evaluate D3D was successfully conducted at FSL in July and August. Prior to the plans for a formal exercise, an effort began in the fall of 1997 to make D3D part of the FSL Daily Weather Briefing, using volunteers from the briefing staff. The volunteers worked closely with the D3D meteorologists to create displays for the briefing. This effort helped to introduce D3D capabilities to FSL meteorologists, as well as expose visitors and other attendees of the Weather Briefing to some new workstation capabilities. While this effort was fruitful and useful in the development of some new products to display in 3D, it was difficult to get other meteorologists to spend sufficient time with D3D to make significant evaluations of both products and the interface. Because of this it was decided that an exercise would be a next productive step in D3D development.

 

Although the development of the D3D Tcl/Tk interface was only in its initial stages, the time was right for meteorological evaluation of D3D capabilities. Developers worked closely with the meteorologists who suggested specific areas of improvement and evaluated the changes that were being made. FSL's Evaluation Team was brought in to develop detailed evaluation plans, evaluation metrics, and schedules for the exercise.

 

Training participants was an important aspect of the exercise. There was an initial meeting to introduce the application followed by personalized training sessions for each of the participants. The D3D developers took considerable strides in providing D3D capabilities that ran reliably throughout the exercise. They were called upon during the exercise to provide follow-up training and to give detailed explanations of the inner workings of D3D. The main focus of RT98 was to evaluate D3D, though there were new 3D datasets that participants were able to work with (such as a high-resolution local model and a 3D radar volume product).

 

4. RT98 RESULTS

 

A goal of RT98 was to provide developers with feedback on D3D displays and the user interface. The forecasters were generally impressed with the D3D capabilities. They generated nearly 50 pages of feedback on D3D related issues. These are being compiled and summarized in order to prioritize future development of 3D capabilities as well as improve the user interface. Some preliminary impressions follow.

 

Nearly all of the 3D display mechanisms were used for enhancing forecasters' understanding of the data. However, forecasters agreed that considerably more time would be needed in order to understand and use 3D display mechanisms in an efficient and meaningful manner in an operational context. Atmospheric processes are inherently three-dimensional, but most meteorologists have been accustomed to examining meteorological data on a surface (500 mb, for example).

 

Although isentropic analyses are more in line with thinking three-dimensionally (for example, warm and moist air upgliding over sloping cold air, displayed as pressure levels on an isentropic surface) especially for the RT98 participants it was still viewed by many as a fairly significant leap to look at model output with the 3D isosurface-type displays. While we believe that such 3D isosurfaces can provide a quick way of examining a great deal of model output, a common theme from the exercise was that most of the participants felt it would take a good deal of experience using 3D isosurfaces to determine what fields were most meaningful with these types of displays, as well as ways to display them. For example, one can color the isosurface of one variable by another variable, say relative humidity isosurface at 80% colored by height, to add more information to the 3D display. We purposely did not include a set of ready-made 3D type isosurface displays of meteorological fields that we considered interesting, as we did not want to influence the participants. However, it became apparent that such displays (known in workstation lingo as "bundles" or "procedures"), would certainly be valuable as training tools. Training was an issue raised by a number of participants, most of whom felt that more involved and lengthy training before the exercise was advisable. (This would likely be even more necessary in operations, as it should be noted that all of the participants, except the two forecasters from the Denver Forecast Office, had exposure and some use of 3D for over six months through the FSL Daily Weather Briefing program.)

 

There were a number of positive responses regarding some of the components of D3D that are actually more 2D in nature, but are more powerful in their capabilities within D3D. These include the horizontal and vertical cross section capability in D3D. Of course horizontal levels form the basic viewing structure in D2D, and the user can also create vertical cross sections. Within D3D these capabilities were expanded through the use of an easy to access (and very popular) slider bar that enabled one to quickly move the cross sections through the data, stopping at any arbitrary level. When used with an isosurface displayed, such functionality added a level of quantitativeness to examining the data (that is, one would know more precisely at what level the isosurface existed, as well as its value).

 

Another feature that the forecasters in particular found exciting was the sounding/parameter plot. Soundings at any location can be done in D2D, and a parameter can be plotted in a time-height cross section, only for one location at a time. In D3D a sounding (temperature and dewpoint plot with wind barbs also) can be chosen, and up to three additional meteorological parameters plotted, and then the sounding site can be easily moved to any location with an instant update to the readout, quickly perusing (in a very quantitative manner) a great deal of data. The model data can also be looped through time when stopped at a particular location. Again, it is presumed that maximum benefit might arise from using such features in conjunction with more purely 3D displays like an isosurface. However, we did not get responses noting this in particular, and this most likely goes back to the idea that it would take more than a few shifts using D3D to fully understand and utilize its capabilities.

 

The level of use and complexity of fields displayed varied among the participants. Some of the participants fairly quickly arrived at some rather complex but meteorologically meaningful displays of model output. For example, one participant quite familiar with looking at an isentropic model fairly quickly found some limitations in the 3D isosurface displays when trying to display fields of potential vorticity, whose value increases near the tropopause, rendering the lower portion of the display obscured. This user found an innovative way to display the information by examining the image from below with the topography removed, but we realized that a more useful method could be achieved with a more effective clipping mechanism, such as will be found in VIS5D Version 5.0. Another user familiar with local scale models was quickly able to display a very interesting field using isosurfaces of cloud water and ice, along with isosurfaces of positive and negative vertical velocity colored by height. When this product was displayed at the FSL Daily Weather Briefing it was quite well received. In fact, D3D was probably most easily used by most participants to display forecasts from the local model. This may have in part arisen from the type of forecast variables available from the local model but not from some of the larger scale models, such as cloud ice and water, and reflectivity. These variables are easily thought of in three-dimensions because we can see them in that manner simply by looking out the window at clouds or thunderstorms.

 

Further examination of the D3D usage log analysis will help determine which mechanisms were most commonly used. In terms of access to the data, the participants considered the D2D look and feel incorporated into the D3D user interface as a good approach that aided in learning and using the system. This was an important result in that the learning curve for using D3D appeared not to be hindered much by the user interface, so that more time could be spent assessing the application to the meteorology. In this regard having the interface look like the familiar one on D2D was important.

 

Among some of the suggested additional datasets were surface data that could be displayed with the 3D fields, the ability to plot winds and perhaps contour values on isosurfaces (this in particular arises from the connection to isentropic analysis, where one commonly looks at winds plotted at an isentropic level with a contour plot of pressure; the corollary in D3D would be an isosurface of a theta surface, which would show the slope, with wind barbs plotted on it, for example). Satellite data were also suggested as a potential 3D display, and indeed this has been attempted by the University of Wisconsin.

 

A 3D radar reflectivity display was also made available, created by combining radar data from 3 area WSR-88D radars, then using an interpolating technique to create a gridded dataset. For the exercise we were only able to achieve 15 min resolution for this display, which we think is a major hindrance to any forecasting use for this dataset (a radar volume is updated on D2D every 6 min). Nonetheless, this was viewed as a potentially valuable dataset that would require more exploration to determine its usefulness.

 

Even after only a peripheral analysis of RT98 exercise, it is apparent that 3D holds a good deal of potential for its use in an operational setting. However, many questions remain relating to how best to use it, training issues, potential display techniques and variables, etc. We hope to address some of these issues in a second D3D exercise, currently planned for the spring of 1999.

 

5. REFERENCES

Grote, U. H. and C. S. Bullock, 1997: User Interface Design of the WFO-Advanced Workstation.: Preprints Twelfth International Conference on Interactive Information and Processing Systems for Meteorology, Oceanography, and Hydrology , Long Beach, Amer. Meteor. Soc., 320-323.

Hibbard, W. R. and D. Santek, 1991: The Vis5D system for easy interactive visualization: Preprints Seventh International Conference on Interactive Information and Processing Systems for Meteorology, Oceanography, and Hydrology , New Orleans, Amer. Meteor. Soc., 129-134.

MacDonald, A. E., and J. S., Wakefield, 1996: WFO-Advanced: An AWIPS-like Prototype Forecaster Workstation.: Preprints Twelfth International Conference on Interactive Information and Processing Systems for Meteorology, Oceanography, and Hydrology , Atlanta, Amer. Meteor. Soc., 254-258.

Forecast Systems Laboratory, 1998, D3D User Guide http://www.fsl.noaa.gov/~osborn/d3d/D3DUG_TC.html

Forecast Systems Laboratory, 1998, LAPS homepage http://laps.fsl.noaa.gov

* Joint collaboration with the Cooperative Institute for  Research in the Atmosphere, Colorado State University, Fort  Collins, Colorado 80523. Corresponding author address:  Paula McCaslin, NOAA/ERL/FSL R/E/FS1, 325 Broadway,  Boulder, CO 80303-3328; e-mail: mccaslin@fsl.noaa.gov