Guest Post By Professor Ben Herman of the University of Arizona. As written on the University’s website
Dr. Herman is primarily concerned with the optics of atmospheric aerosols, polarization and scattering, and the application of inversion techniques to analyze remote sensing data obtained from aircraft and satellites. Currently, he is working on several satellite based remote sensing projects to monitor ozone, temperature, water vapor and aerosols from space.
Following is Ben’s guest post on regional climate modelling
I have had several discussions here with various people concerning the problem of regional climate prediction using climate models to set up boundary conditions for a smaller, regional area in which a much smaller grid size is used. The problem is that if the boundary conditions are incorrect, obviously this will deteriorate any predictions made using those boundaries. Now with that said, there seems to be quite a lot of effort being put into regional prediction using global climate models. The climate models, at this time, are using a much larger grid than required for regional prediction, so this has led to the climate of the smaller regional areas of interest to be solved for separately, using the results of the global to be used as boundary conditions. I have compared it to having a set of numbers accurate to one decimal point, using those measurements in a series of mathematical operations, and providing answers to ,say three decimal points. You can do that, but, of course, the additional decimal points have no practical use.
Another way to look at this is to imagine that the global solution for a given time has been broken down into a Fourier series. This series will contain frequencies that are limited by the grid spacing of the global model. Any frequencies beyond this range are not present and thus, cannot have an impact within the smaller region, even though higher frequencies within this region may be solved for due to the higher resolution within this region. But, since these higher frequencies cannot be influenced by similar frequencies outside of the region in question, they will generally be in error. The only way that this error could be avoided would be if none of these higher frequencies were present during the valid time of the forecast, generally an unlikely occurence. It thus appears to me that undermost conditions, the regional forecasts, using present techniques, are operating with an incomplete set of initial conditions which will certainly limit their accuracy.