9A.3
Graphics vs. Images for Display of Model Data

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner
Thursday, 6 February 2014: 12:00 AM
Room C106 (The Georgia World Congress Center )
Frederick R. Mosher, Embry-Riddle Aeronautical Univ., Daytona Beach, FL

Model and observational data have traditionally been displayed as graphics. Initially these data were displayed as contoured fields, but the use of filled graphics has become popular for many types of model fields. Satellite and radar data have traditionally been displayed as images. Initially these data were displayed as black and white images, but the use of colored enhancement tables has become popular for many types of satellite and radar data. All of these data fields are originally just numbers. The display techniques are used to help people make sense of the numbers by putting them into a picture format. The big difference between the graphical displays and image displays has been the original data resolution. Graphical display techniques are typically used for data sets on the order of 100 or so points in each direction. Image display techniques are typically used for data sets on the order of 1000 or so points in each direction. Weather forecast models have traditionally been displayed with graphical techniques. However, the size of the model data sets has been increasing, and is starting to approach the resolutions of traditional image data sets. For instance the National Weather Service (NWS) North American Model (NAM) output grid has 428 rows by 614 columns. The NWS Global Forecast System (GFS) has 361 rows by 720 columns for the ˝ degree output grids. The traditional method of contouring a grid to make a picture is to draw lines of equal (iso) values. The interpolation between grid points for the isoline allows the resultant picture to have a higher spatial resolution than the original gridded data. The traditional method of making an image is to transform each data pixel into an image pixel. Most computer display systems have 8 bit displays, so the data value is scaled to fit within the 0 to 256 brightness range. If the resultant image picture does not fill the screen, the pixel values are typically replicated (or sampled if the data size is larger than the screen) so the image will fill the screen. For contoured graphics, the area between contours can be colored the same color as the contour. Most computer displays support polygon objects, so the filled contour can be presented as polygons to the screen display. Typically filled graphics will support around 32 different colors of graphics. Converting model data into images rather than graphics allows for the user to see a wider range of structure in the data as well as being able to display fields with widely varying values. Consider the following example of the NAM field of surface Turbulent Kinetic Energy (TKE) for May 23, 2013 at 18Z. TKE is the kinetic energy of turbulent eddies. The turbulent eddies are caused by strong surface winds and an unstable lapse rate in the boundary layer. The model grid is first converted into an image format, remapped into a standard US projection used for satellite images, and then an enhancement table is used to convert the data values into colors. The display shows considerable horizontal variability which appears to be meteorologically correct. For instance the “hole” of low values of TKE in Oklahoma shows up in the verifying time visible satellite image as cloud free with thunderstorms to the east. The higher values of TKE near the Great Lakes are associated with strong horizontal winds. The higher values of TKE do not go over the lakes because the water is quite cold making a stable boundary layer which is not conducive to turbulence. When the original model grid was displayed as a contoured graphic, there were so many lines that one could not understand either the large scale process, or the small scale local processes. Making the model data into images allows for better understandings of the meteorological processing going on. Even “smooth” fields, such as sea level pressure, show local variations when displayed as images. Images of sea level pressure allow one to see slight variations in pressure caused by local heating or cooling superimposed on the larger scale synoptic field (the images also showed false pressure variations in the mountains which appear to be caused by the reduction to sea level of the pressure data). The use of image display technology for model data shows considerable promise for better understanding small scale structures resolved by the higher resolution models of today.