To test the EH model system, we looked for a real case of a storm/flash flood event. The event chosen hit the Chicago area in July of 1996. The Illinois USGS has a summary of the storm. A rainfall map and satellite sequence is also available from an Illinois Water Survey report on the storm.
Bob Wilhelmson and his group (Brian Jewett and Crystal Shaw) did the survey of potential cases. Brian prepared the initial data sets for MM5.
Doug Johnston prepared the topographic data set for the region where the precipitation was heaviest. This data set covers about 50 sq km near Aurora, IL. The entire data set is included in the flow2d tar file.
We (Brian Jewett and Rob Jacob) first tried to simulate the storm using just MM5. Two of the planned 5 grids were used. A few combinations of start time, FDDA, and parameterizations were tried but none could get the storm to rain in the right place at the right time.
To test flow2d with computed precip and real topography, the RAINC and RAINNC fields from the innermost grid of one of the above runs was pulled off of the MM5 history file and saved in a seperate, smaller, file. The program presend reads this file and sends a total precip field to the coupler which interpolates it and sends it to the runoff model.
Since MM5 could not get the rain in exactly the right place, the coupler is told that the location of the smaller grid (the runoff model) is located under the rain. The location of the rain and where the SW corner of the terrain should be was determined from viewing a vis5d file of the MM5 run.
The two-grid simulations still took about an hour of wallclock time on 60 nodes of the Origin 2000. To show MM5 communicating with the coupler in real time at the SC98 demo, a faster configuration was necessary. When starting MM5 for communication with the model system, a 1 grid case with a two minute timestep and a 12Z start time is used.
All of the pieces, MM5, coupler and flow2d, have been run together but the exact arguments for the coupler (the placement of the topo under the rain) still have to be adjusted.
