Some interesting reflections from Laurier Williams about Model Grid points, the best time stamps for modelling, etc. This information was originally posted in the May 11, 2011 Cold Outbreak thread on Weatherzone Forums. This information may change but was relevant at the time of posting.
'The best models to get full information from are the time stamped 00z and 12z charts. The 06z and 18z models do not have as much raw data available, especially for the upper atmosphere where most countries (Australia included) only send up temperature/humidity sondes at 00 and 12z. (Wind sondes are better with most of the globe doing them 6-hourly.) Much of their assumptions are therefore based on satellite interpretation and the previous run of the model.'
'Models divide the world into a grid of points and on each run they calculate the interaction of weather variables from now into the future across this grid in a series of steps. The spacing between the points for the current public GFS is ~35km, ACCESS Regional is ~37.5km and ACCESS Global is ~80km. You can actually see the grid if you look at a forecast wind chart - the points that the wind barbs fly from are the grid points.'
'The elevation in the model for each grid point is the average elevation of its surrounding area. So if your town was on a mountain top at 1000m surrounded by flat country at 300m, the average for the grid point nearest your town would be, say, 500m. Some model outputs for point locations obviously make more of an effort to correct for this, both by interpolating between grid points and correcting for the difference between the grid elevation and the known point location elevation. If the point location model output is half-way decent, it will give you the elevation, latitude and longitude of the point for which it's forecasting, and you can tell a little from this as to whether it has made any of these corrections.'