Under support from the Colorado Department of Natural Resources we have performed simulations of numerous historical heavy precipitation events in Colorado. One aim of this research is to develop an understanding of the variation in extreme rainfall events with altitude. Another is to develop a new methodology for extreme precipitation estimation. The simulations are performed with the non-hydrostatic cloud-resolving version of RAMS. The simulations are initialized from large-scale historical synoptic data and surface data. The coarse grid covers the western U.S. and using interactive nested grids, the grids are refined to permit cloud-resolving simulations of the respective storm events with grid spacing of 1.67 to 3 km depending on the scale of the particular extreme precipitation event. Following a baseline simulation, sensitivity runs are performed with various estimates of initial soil moisture, enhanced low-level atmospheric moisture, translation of the synoptic patterns relative to the underlying terrain.

The following storms have been simulated (the number after the event is the number of simulations performed:

1) July 28, 1997 Fort Collins Storm (6)

2) Aug. 31, 1976 Big Thompson Storm (6)

3) July 31, 1999 Dallas Divide Storm (5)

4) Sept. 18-22, 1997 Park Range Storm (2 so far)

5) Sept. 4-6, 1970 Southern San Juans Storm (At least 3)

6) July 26, 1999 Saguache Creek Storm (in preparation)

The following is a summary of what we have learned from this effort:

· The most accurate control simulations occur with the least convective, large-scale forced storms like the San Juan and Park Range storms. The least successful simulations occur with the older convective events like the Big Thompson storm. This is likely due to the coarse resolution of the initial NCEP reanalysis data used for the older events. Simulations of heavy convective events are highly sensitive to the specification of initial soil moisture fields.

· Precipitation maxima occurring at higher elevations generally have significant contributions from hail, which may diminish surface runoff rates due to prolonged melting.

· The simulated convective events produce extreme rainfall with typical spatial and timing errors of 10 to 50 km and one to several hours, respectively.

· The correlation between maximum accumulated liquid precipitation and elevation may be used to refine spatially interpolated estimates of maximum liquid rainfall through the use of a special interpolation technique called Kriging.