Embedded Files
Before the weather computer would do anything, it needed to know the initial values for temperatures and winds at all levels at all grid point locations. How could Frank possibly have that information? Fortunately, his father works for the National Weather Service, and he kindly gave him access to the National Weather Service's analyses of temperature and wind. Then it was just a matter of interpolating the analysis values to the particular grid points that Frank had selected. At last all was set, and Frank started the weather computer.
By now the weather computer was computing winds and temperatures at lots of levels and lots of grid points, and it was taking much longer to make its forecast. But finally it finished. Frank examined the results. It proved to be a dull forecast. Because nothing was changing along the edges of the domain, 100 miles from the sixteenth green, not much was happening elsewhere, either. The temperature rose and fell just a little bit, but the winds stayed fairly constant.
Frank knew that if he wanted to forecast the winds, he would have to do better. So he asked his father for access to the National Weather Service gridded forecasts. His idea was that he could use those forecasts to update the data along his 100-mile perimeter, rather than keeping those values constant. That way, as his perimeter changed, the pressure, wind, and temperature variations would make their way inside the domain and affect the forecasts of wind and temperature at the sixteenth green.
It was a good idea, but his father wouldn't release the forecast data. (The National Weather Service was more proprietary with its data in those days.) So Frank came up with another idea: he would extend his forecast domain so that it was 1000 miles in all directions. He knew that typical weather systems would take a day or so to cover that distance, and that as long as he started with accurate weather information, the winds and temperatures ought to evolve fairly accurately within his model domain, at least for the first twenty-four hours. Also, to forecast the evolution of weather systems, he extended his domain upward to a height of 10 miles, to include the jet streams and steering currents.
So he got the analysis over the larger area, interpolated it to his grid point levels, and started the forecast. The weather computer immediately spitted and told him "Error 47". Shocked, Frank got the manual and looked up Error 47. He read:"Error 47: a ridiculous number of grid points. Reduce the number of points, or the computer will never be able to do all the calculations." Frank checked some books, and found that the horizontal spacing of his grids, 20 miles, was small but reasonable, but that the vertical spacing was a bit excessive. So he cut back to a vertical spacing of 1000 feet for his levels, which was much more reasonable. Even so, that left him with about 50 vertical levels (actually 55, since he kept a higher resolution near the ground where it counted) and 100 grid points in the north-south and east-west direction, for a total number of grid points of 100 x 100 x 50 = 500,000. A long, long, time ago, Frank was doing a forecast for a single point, and now he was doing a forecast for 500,000 points! No wonder it took longer! But at least with the new grid spacing, the weather computer was working again. And eventually it gave him his forecast.
Report abuse