A significant concept to numerical weather prediction is resolution. The notion is that at a higher resolution, more features can be represented by a model. However, this increased resolution comes at additional computational expense.
Take, for example, a tropical cyclone. Below is an image of Hurricane Fran, at 1km resolution.
Note the structure that is visible: the details of the eye, for example.
I am envisioning an applet that allows the user to choose the resolution via some sort of slider. The image will be 'radar' like image. This image, at 1km resolution, covering a 512km by 512km domain, will clearly show the structure of the eye-wall and of the outer rain bands.
Using the slider, one can choose to degrade the resolution. Choices for resolution will be 1km, 2km, 4km, 8km, 16km, 32km, 64km, and 128km. Thus, at the coarses resolution, there are only 16 pixels, whereas at 1km resolution, there are 262144pixels (512**2)
When the slider is moved from one resolution to another, the "radar" image changes to that resolution. I believe that images can be made using IDV, but just changing how many pixels are shown. One can choose to use them all, every 2nd, every 4th, etc. (Kelvin demo for LEAD).
The changing image will show the consequences of resolution reduction: the eye wall structure will decay, and at somepoint, the eye will not be discernable.
In addition to the image changing, there will be a text readout of the number of pixels. For example, for a resolution of 4km, nx=128, ny=128, and the total number of pixels is 16834.
Beside this number of pixels, there will be a 'model computing time'. For example, we could assume that at 128km resolution, the model will take 1 minute to run. Assuming that a halving the resolution leads to 4x gridpoints, and making the time step 1/2 as long, a 64km run would then take 16 minutes. In the limit, the 1km resolution would take 16384 minutes, (over 11 hours). This model computing time would be expressed as a common clock. The number of time would be 'shaded'. For instance, if the model would take 1 hour to run, the clock would read 1pm, with the area of the clock from Noon to the hour-hand being shaded in.
The two dramatic visuals would be the image being sharper or more degraded, and the clock.
The idea is to incorporate, but improve upon, images such as this:
An important point is to make sure that the false idea (misconception!) that a 1km model will resolve a 1km feature is not communicated. Not sure how to do that...