Operational prediction of flash floods produced by thunderstorm (convective
) precipitation in mountainous areas requires accurate estimates or predict
ions of the precipitation distribution in space and time. The details of th
e spatial distribution are especially critical in complex terrain because t
he watersheds are generally small in size, and small position errors in the
forecast or observed placement of the precipitation can distribute the rai
n over the wrong watershed. In addition to the need for good precipitation
estimates and predictions, accurate flood prediction requires a surface-hyd
rologic model that is capable of predicting stream or river discharge based
on the precipitation-rate input data. Different techniques for the estimat
ion and prediction of convective precipitation will be applied to the Buffa
lo Creek, Colorado flash flood of July 1996, where over 75 mm of rain from
a thunderstorm fell on the watershed in less than I h. The hydrologic impac
t of the precipitation was exacerbated by the fact that a significant fract
ion of the watershed experienced a wildfire approximately two months prior
to the rain event. Precipitation estimates from the National Weather Servic
e's operational Weather Surveillance Radar-Doppler 1988 and the National Ce
nter for Atmospheric Research S-band, research, dual-polarization radar, co
located to the east of Denver, are compared. In addition, very short range
forecasts from a convection-resolving dynamic model, which is initialized v
ariationally using the radar reflectivity and Doppler winds, are compared w
ith forecasts from an automated-algorithmic forecast system that also emplo
ys the radar data. The radar estimates of rain rate, and the two forecastin
g systems that employ the radar data, have degraded accuracy by virtue of t
he fact that they are applied in complex terrain. Nevertheless, the radar d
ata and forecasts from the dynamic model and the automated algorithm could
be operationally useful for input to surface-hydrologic models employed for
flood warning. Precipitation data provided by these various techniques at
short time scales and at fine spatial resolutions are employed as detailed
input to a distributed-parameter hydrologic model for flash-flood predictio
n and analysis. With the radar-based precipitation estimates employed as in
put, the simulated flood discharge was similar to that observed. The dynami
c-model precipitation forecast showed the most promise in providing a signi
ficant discharge-forecast lead time. The algorithmic system's precipitation
forecast did not demonstrate as much skill, but the associated discharge f
orecast would still have been sufficient to have provided an alert of impen
ding flood danger.