High Resolution Terrain Data Example
This page has some examples of how, by using higher resolution terrain height data, the grid cells generated by the WRF Preprocessing System (WPS) seem to follow the actual terrain more closely. For all of these pictures, Google Earth was set to draw terrain height with a 3X exaggeration. To find out more about getting and incorporating the high resolution, 1/3 arc second data into WPS, go to my page that more or less describes the process: High Resolution Terrain Height Data in WRF. If you would like the KML format data that was used to generate these pictures, it can be downloaded from my page about WRF data displayed in Google Earth: WRF Data in Google Earth.
The pictures on this page need to be clicked on to make them larger. The important details require them to be viewed a nearly full screen size.
This is an example of the grid cells when 30 arc second data is used to create them.That is the highest resolution that comes with WRF as provided on the regular download page. The cells are artificially drawn as flat, using the center point for the terrain height of the entire cell. When drawn this way, I would expect about half of each cell to be above the terrain shown by Google Earth and half below. If you click on the picture to the left to zoom in, you can see that is definitely not the case. The grid cells are 1km in size, which is approximately the same resolution as the 30 arc second data, but after interpolation is performed, much of what little detail contained in the data is smoothed out. With some cells, the actual terrain is completely above or below the translucent surface of the single height grid cell.
The picture to the right shows the same location but with 200m grid cells created from 30 arc second data. The same problem is even more evident at this resolution. As mentioned, drawing the grid cells this way is artificial because WPS also provides terrain height at the centers of all four sides. But by drawing a grid cell at a single height, the step up or down from one cell to the next is evident. Further down, I have pictures of grid cells drawn as I believe WRF represents them internally.
|
Here are the same two pictures but with grid cells generated with higher resolution terrain height data. This data is 1/3 of an arc second, which is approximately 10 meters in horizontal resolution. In both pictures, the Google Earth terrain is more likely to be half above and half below the flat, translucent grid cell surface. That is what I would expect and hope for in grid cells I am using for a weather forecast. Even some of the cells that seem to not be half and half, become so when zooming in closer. That has to do with the way Google Earth handles the resolution of the terrain it draws on the screen. (Zooming in on these PICTURES of the Google Earth screen does not change anything.)
|
These are the same four pictures as above, but a different method was used to draw the surface of the grid cells. For these, I used the terrain height given for the staggered grid points, which are on the middle of each of the four sides of a grid cell. From these, I calculated a height for each corner based on the closest four staggered points. This method probably best presents the grid cells as used within WRF; all the edges match up, giving a continuous surface. The flat surfaces shown above help me evaluate the quality of the terrain height data I am using. These sloped surfaces help me see how WRF is going to look at things.
Even with the higher resolution terrain height data, it seems apparent that a lot of details are lost when using a 1km horizontal resolution. Reality has so much more detail that seems like it would affect high resolution wind forecasts.
|
|
|