Exploratory Rainfall Simulation of 2004 Halloween Eve Flash Flood
2006-2007 Engagement Grant Report
Duane Stevens and Jax (Guangxia) Cao
Introduction
In the evening of 30 October 2004, a heavy thundershower on Oahu Island, Hawai‘i induced a severe flash flood in the Manoa valley, causing a severe property loss. The flood damaged and pile up many cars, ripped through houses and buildings, and inundated the UH Hamilton Library ground floor, a loss of $50 millions occurred for the UH alone. This commonly called Halloween Eve Flash flood, among many others, again raises a key forecast question of how to correctly predict the abnormal rainfall for property protection for both civilian and military. The best tool to handle the problem is through numerical modeling. This project uses 2004 Halloween flash flood as the case, to explore the issue of improving the simulation of heavy rainfall event. We use the Weather Research and Forecast (WRF) model for the exploration.
The project has been conducted in a period of January – August 2007, with a support from the 2006-2007 MHPCC/HPC engagement program, and has the following objectives:
1) to test different physical options to identify the best approximation for the Halloween even case,
2) to change the initial and run-time files through a pseudo 3DVAR and 4DVAR approach to identify key initialization strategy for a better forecast.
Methods
WRF is a state-of-the-science model, jointly developed by the National Center for Atmospheric Research (NCAR) with several federal agencies. This project uses three nested domains, with the resolution of 25:5:1 km, to configure the WRF model. The overall simulation strategy is shown in Table 1. The simulation is performed in MHPCC JAWS, and runs over 32,000 hours.
Table 1. WRF model simulation strategy, 1. dry run, i.e. model run without soil moisture; 2. dry and nudging runs; 3. dry and the runs with the changed initial condition; 4. wet runs, i.e. model run with soil moisture; 5 runs with changed initial condition, and the soil moisture replaced with an one-month run results, and modified run-time model conditions. MP = microphysics, and Cu = cumulus parameterization.
MP1 Kessler / MP2 Lin et al.. / MP5 Ferrier / MP6 WSM6 / MP8 ThompsonCu Kain-Fritsch / 1 / 1, 2, 3, 4, 5 / 1, 2, 3, 4, 5 / 1, 2, 3, 4, 5 / 1, 2, 3, 4, 5
Cu Betts-Miller-Janji / n/a / 1, 2, 3, 4, 5 / 1, 2, 3, 4, 5 / 1, 2, 3, 4, 5 / 1, 2, 3, 4, 5
Cu Grell-Devenyi / n/a / 1, 2, 3, 4, 5 / 1, 2, 3, 4, 5 / 1, 2, 3, 4, 5 / 1, 2, 3, 4, 5
Results
Our simulation indicates that Lin et al. and Thompson microphysics and Betts-Miller-Janjic cumulus paramterization performs better (Figure 1 and 2), in that more rainfall is simulated over the Oahu Island.
The surface soil moisture affects the rainfall output in the model, and runs with the soil moisture produce more precipitation than the runs without soil moisture (Figure 3 and 4). However, the nudging technique does not change the model simulation result. As expected either changing the initial condition or modifying the soil moisture leads to an improved rainfall simulation (Figure 5 and 6). Combining above techniques with the modified run-time model environment enables us to have total rainfall simulation around 280 mm per day matches that of the observation 270 mm per day (Figure 7). However there is still the problem of correctly simulating the location of the simulation. Our results do suggest that in the future a combined idealized simulation is necessary to generate guidelines for a further pseudo-4DVAR strategy.
Acknowledgement: This study is supported by the 2006-2007 MHPCC/HPC Engagement grant.
Figure 1. Comparison of the simulation among four microphysics options, upper left: Lin et al. scheme, upper right: Ferrier scheme, low left: WSM6 scheme, and low right: Thompson scheme.
Figure 2. Comparison among three cumulus parameterization under Lin et al. microphysics; upper left: Kain-Fritsch, upper right: Betts-Miller-Janjic, low left: Grell-Devenyi ensemble, low right: Betts-Miller-Janjic but with Thompson microphysics.
Figure 3. Model simulation without soil moisture.
Figure 4. Model simulation with the soil moisture. Contour unit: mm.
Figure 5. The rainfall simulation with the soil moisture modified by a month-long simulation output.
Figure 6. Model simulation with modified initial condition on wind and air moisture. Contour unit: mm.
Figure 7. The total rainfall simulation after a period of 24 hours till 18 ZZ 31 October 2004. Unit: mm.