Interactive visualization for high-resolution weather forecasting at the
1996 Centennial Olympic Games
Lloyd A. Treinish
lloydt@watson.ibm.com
Zaphiris D. Christidis
zaphiri@watson.ibm.com
IBM Thomas J. Watson Research Center
Yorktown Heights, NY 10598
Introduction
In general, operational meteorologists have a need to improve the accuracy
of their local forecasts. When faced with a highly public event like the
1996 Centennial Olympic Games composed of activities that are affected
by the weather, this requirement grows in importance. As a result, the
National Weather Service (NWS) of the National Oceanic and Atmospheric
Administration (NOAA) agreed to provide meteorological data, and warning
and forecast services for the Atlanta Committee for the Olympic Games (ACOG)
via a number of new technologies [Rothfusz et al, 1996].
One solution to this problem is the introduction of additional tools
for the forecaster. Typically, numerical models are run at relatively low
resolution over a large geographic region (e.g., 29 km for the continental
United States). Such models are used by the National Weather Service (NWS)
in this country and by the European Center for Medium Weather Forecasting
(ECMWF) in Europe. A meteorologist will employ this model output, their
knowledge of the region in question and local conditions derived from in
situ and remotely-sensed observations to arrive at a final forecast. However,
forecasts can be substantially improved with the application of regional
numerical modeling techniques, that provide predicted information at a
localized or mesoscale level. The goal is to enable forecasting the weather
at the Olympic games in Atlanta to be at an unprecedented level of precision.
Approach
This solution will allow for the inclusion of conditions unique to the
locale, and increase the accuracy of such forecasts by utilizing a forecasting
model that 1) operates at a higher spatial resolution (e.g., 2 km for the
Olympic Games), 2) provides effective and timely throughput, and 3) incorporates
visualization methods for analyzing the results more quickly. A key mesoscale
model, called Regional Atmospheric Modelling System (RAMS) developed by
NOAA's Forecast System Laboratory (FSL) will be used for this effort. RAMS
will supplement the lower-resolution model output by providing flexibility
for the forecaster in the selection of domains and level of detail. Since
RAMS focuses on the physics of cloud formation, for example, a much more
detail prediction of precipitation is possible. Hence, this model will
be used to help determine forecasts over four main Olympic venues, independently
at resolutions up to 2 km [Snook, 1996].
To provide timely production of model output for the Olympic Games,
a parallelized version of RAMS is operating on a distributed memory multi-processor
(30-node IBM RISC System/6000 Scalable POWERparallel or SP-2) installed
at the NWS field office in Peachtree City, GA. This configuration can provide
weather simulations for two venues simultaneously, with each simulation
using fourteen compute and one I/O processors, respectively. The RAMS model,
will be coupled to a synoptic weather model, which is simulating the weather
across North America on a shared-memory vector processor (Cray C-90). The
results of this broad range (approximately 15 km) simulation will be periodically
transmitted to the SP-2 in Peachtree City, which will then narrow the simulation
range to approximately two kilometers. The RAMS model will also accept
observational data used as the initial conditions for the weather forecasts.
It is planned to produce multiple, short (6 to 12 hours) high-resolution
(at 2 km resolution) and longer (18 hour), low- resolution (10 km) runs
per day during the Olympic Games.
This new capability of doing high-resolution, numerical weather prediction
requires a change in how the operational meteorologist approaches the analysis
of data to create a forecast. Since large volumes of complex data for each
RAMS run will be quickly produced, the use of traditional graphical representations
of data for forecasters will not suffice. Instead of static or simple flip-book
animations of two-dimensional techniques like contour maps, novel three-dimensional
visualization strategies will be employed. These methods will also be applied
to the observational and model data that comprise the initial and boundary
conditions for RAMS. These capabilities are being implemented with a general-purpose,
data analysis and visualization toolkit (IBM Visualization Data Explorer
or DX, [Abram and Treinish, 1995]).
The visualizations are being developed within a "natural" coordinate
system to provide a context for analysis, viewing and interaction, independent
of the original coordinate system of the data. This is derived from the
notion of correlative visualization, where each data set to be examined
is processed independently and merging takes places only at render time
[Treinish, 1994]. Two classes of visualization are being developed under
this strategy, both of which provide the forecasters interactive capabilities
with three-dimensional representations of the state of the atmosphere (e.g.,
thermodynamics, moisture, clouds) derived from the data, often emphasizing
redundant encoding schemes and registered with relevant terrain and political
boundary maps. The first is a set of simplified qualitative techniques.
They serve two purposes: 1) provide a browse product for gross assessment
of the data and 2) source material for dissemination of forecast information
for this very public event (e.g., media, World- Wide-Web, etc.). The second
class of visualizations is a set of quantitative techniques to support
analysis and development of more precise forecasts.
Results
The qualitative methods will be applied to the generation of key-frame
fly-over animations over specific Olympic venues and time-based animations
over fixed sites, examples of which are illustrated in the next five figures.
The first two are from a RAMS simulation over the full Olympics domain
from early December 1995.
This first figure shows a three-dimensional
representation of a cloud derived from a calculation of humidity. The data
are shown as a translucent, white isosurface of 95% humidity registered
with a pseudo-colored terrain map overlaid with coastline, river and state
boundary maps. This familiar representation can effectively show gross
atmospheric motion and potential distribution of moisture.
The next figure shows surface properties in
a three-dimensional fashion. Humidity is shown as pseudo-colored, filled
contour bands draped over a map similar to that used in the first figure.
Surface winds are shown via vector arrows of constant size, which are pseudo-colored
by speed. This representation illustrates orographic effects on the atmosphere.
In an effort to describe both upper air and surface characteristics
simultaneously, the ideas illustrated in the first two figures can be integrated,
and are shown in the next three images with data from a simulation from
late April 1996, when heavy rain showers occurred in northeastern Georgia.
Over a map similar to those used previously, hourly precipitation is shown
as pseudo-colored, filled contours deformed by topography in above.
The areas of heavy rainfall can be seen as a deep blue "puddle". The precipitation
data is overlaid with surface winds as before. A cloud "boundary" is represented
by a translucent white isosurface of cloud water density (both liquid and
ice). Inside the cloud isosurface is a translucent cyan isosurface, which
is derived from a simulation of the reflectivity that a weather radar sensor
would detect as if it were observing a real cloud. Such a surface corresponds
to a so-called rain shaft internal to the cloud, and shows direct correlation
with the region of relatively heavy precipitation.
Another approach to examining these same data is shown above
without the topography map, but with the boundary and river maps overlaid
on a precipitation surface. Cloud water density is now shown via direct
volume rendering, which is more amphorous in appearance but more realistic
compared to a somewhat artificial cloud boundary derived from an isosurface.
The surface winds are now precisely shown as streamlines, pseudo-colored
by speed with directional arrows. The geographic view of the region has
changed to show more of the vertical cloud structure. This can be shown
without the reflectivity surface as illustrated below.
The quantitative methods of the second class of visualization techniques
provide the forecasters with new tools to interact graphically with and
interrogate model output, track or steer the progress of the model as it
executes, and quickly and accurately assess results. Examples of these
methods are illustrated in the next several figures, which show the visualizations
and a prototype user interface presented to the forecaster. Most of these
images are full screen captures from a workstation. Click on any of the
low-resolution images to view the full image.
The main image in the figure above (upper
left) shows an isosurface of humidity at 65%, which is pseudo-colored by
temperature, which effectively segments the surface into the cold, but
moist upper air from the warmer air near the ground. The user has the ability
to probe the atmospheric volume with the mouse as a "virtual met-station".
The dialog box near the bottom center shows the interrogated coordinates
(latitude, longitude and pressure level) and humidity value. That same
probe can be used to define a virtual sounding, a graphical analogue to
a meteorologist placing a collection of instruments at a specific location
to observe the real atmosphere. In this case, one can derive the same information
from the simulated atmosphere (i.e., measurements of specific physical
quantities at locations of interest). The probe is extruded into an profile
tube, which is pseudo-colored by humidity in this image. Along the profile,
the computed winds are shown as vector arrows of velocity, which are pseudo-colored
and sized by speed. This information is analogous to what might be garnered
by a set of balloon-borne instruments launched from that location in a
real atmosphere. Based upon the three-dimensional representations, the
forecaster can determine areas of potential interesting activity, such
as updrafts, convective flow, etc. and place this virtual station accordingly.
The data being examined are further displayed by a conventional pressure
profile in the lower right of the image. A more detailed view of the the
main image is shown below, where streamlines
of wind velocity are visible, also pseudo-colored by speed. The seed points
for the streamline calculations are the same points where the vector arrows
are shown and those composing the humidity profile plot.
The next three images show another type of interaction, in which the
forecaster can select one or more different volumetric model outputs with
a choice of techniques in a similiar representation of the atmosphere as
used before. Two of the three also show simultaneously examination of various
surface parameters in a separate window with disparate methods. Control
panels for data and technique selection are visible as well as interactive
controls for the image windows (e.g., rotation, pan/zoom) and a VCR-like
widget for sequencing through time.
In the figure above, the image window shows
north-south vertical slices of volumetric temperature as pseudo-color,
filled-contour bands and a relative humidity isosurface at 75%.
In this next figure, an additional image
is visible, which shows two-dimensional representations of a single slice
from the volumetric data at a pressure level of or surface at 350 mb. Relative
humidity at that level is shown as pseudo-color, filled-contour bands.
Horizontal wind velocity is illustrated as vector arrows, pseudo-colored
by speed. These fields are overlaid with coastline, river and state boundary
maps. The upper image remains the same.
In this next figure, the lower image shows
surface temperature as pseudo-color overlaid with vector arrows of horizontal
winds, pseudo-color contours of humidity and coastline, river and state
boundary maps. The upper image still shows north-south vertical slices
of volumetric temperature as pseudo-color, filled-contour bands. But the
surface temperature from the lower image is deformed into pressure space
and registered in the three-dimensional atmospheric volume of the upper
image.
In this figure, only the surface data as from
the lower image of the previous figure are shown. The humidity is still
shown as pseudo-color, but now the winds are shown as streamlines with
directional arrows. Pseudo-color line contours of temperature are overlaid
now. All three parameters as well as the geographic maps are draped over
a surface, which is deformed by precipitable water.
Conclusion
While this approach to operational weather forecasting will be of immediate
value at this year's Olympics, enabling athletes and attendees to plan
around adverse weather conditions, these technologies could be applied
in other areas where precision forecasting shows promise. Such potential
applications include travel, agriculture, broadcast, pollution monitoring,
and fire control and management.
Acknowledgments
This effort is part of the Olympic Games weather forecasting system, a
joint project between IBM, and the National Weather Service and the Forecast
Systems Laboratory of the National Oceanic and Atmospheric Administration.
References
Abram, G. and L. Treinish. "An Extended Data Flow Architecture for Data
Analysis and Visualization". Proceedings of the IEEE Visualization 1995
Conference, October 29 - November 3, 1995, Atlanta, GA, pp. 263-270.
Rothfusz, L., J. Johnson, L. Safford, M. McLaughlin and S. Rinard. "The
Olympic Weather Support System". Proceedings of the American Meterological
Society 12th International Conference on Interactive Information and Processing
Systems for Meteorology, Oceanography, and Hydrology, January 28 - February
2, 1996, Atlanta, GA, pp. 1-6.
Snook, J. "Local Domain Forecasting Support to the 1996 Atlanta Olympic
Games". Proceedings of the American Meterological Society 12th International
Conference on Interactive Information and Processing Systems for Meteorology,
Oceanography, and Hydrology, January 28 - February 2, 1996, Atlanta, GA,
pp. 32-35.
Treinish, L. "Visualization of Disparate Data in the Earth Sciences".
Computers in Physics, 8, n.6, November/December 1994, pp. 664-671.
lloydt@watson.ibm.com
[ DX Home Page | Contact
DX ]
