WMOD00493

• Title: Nested Model Simulations of Regional Orographic Preocipitation
• Date: 1/17/1993
• Weather Modification - Doc Type: Report

260-km domain. Small dots show each MM4 grid point and
large dots indicate the CCM (Community Climate Model) R-
15 (rhomboidal spherical harmonic truncation at wave
number 15) 500-km resolution grid points. 1be CCM has
been used widely by modelers to examine the effects of
doubled C02 on global climate (Washington and Meehl,
1984). Clearly, nested regional and local-scale models are
required to examine precipitation processes at the watershed
leveL

a

AT p. 1'00.000 Of( &3OTZOOQ. i883 lilt.

b ~4H pcp AT S- 1000 ON ft,.W720 (CtnJ

Figure 1. Monsoon flow and precipitation over the West.
MM4 simulations of the monsoOIi flow at 700 mb on July
20, 1983, at 0000 UTC (figure la). Vectors indicate airflow
with moisture sources in the Gulf of Mexico and Pacific
marked by bold arrows. MM4 grid points are marked by
small dots and wind vectors are plotted at every 3 cd point.
Large dots ( .) show the very coarse resolution of the general
circulation model grid with - 500- km resolution. The
nested Clark model domains for 10-km (solid bold square)
and 5-km resolution (dashed square) are marked over the
Gunnison Basin. Precipitation predicted by MM4 from
1200 UTC on July 19 to 1200 UTC July 20, 1983, shows a
5.25-cm maximum over Colorado (figure Ib).

Reclamation is using two models to simulate local

orographic precipitation: Rhea and Clark models. Medina
(1992) has adapted the steady-state Rhea model (Rhea; 1977)
to simulate local precipitation over the Gunnison River
Basin in Colomdo aud has adapted the model to use initial
conditions from soundings derived from MM4. This
approximates a one-way, local-scale nesting. Comparisons
between Rhea modE:I simulations from mwinsondes and
MM4-derived soundings are encouraging; however,
convective precipitatilon is not accurately simulated in either
the MM4 or Rhea models.

Therefore, we are also using the more sophisticated,
three-dimensional, time-dependent, local-scale Clark model
to simulate the local evolution of convective clouds and
precipitation over complex terrain in the central Colomdo
Rockies. The Clark model is used extensively throughout
the research community to simulate phenomena ranging
from airflow over complex terrain to deep convection (Clarlc,
1979). The model is an anelastic, finite difference model
with one- and two-wny intemctive nesting described by Clark
(1977), and Clarlc and Hall (1991). The Clark model will be
used to understand the complex three-dimensional airflow
and evolution of clouds and precipitation in a limited sample
of cases. It will contribute to better parameterizations of
airflow structure for lthe Rhea model, which may be roo on a
large number of casE:S. Initial Clark model simulations are
examining model sensitivity to initial conditions and
increased resolution .of complex terrain in western Colorado
using mwinsondes aud MM4 soundings.

3. INITIALIZATION AND ASSESSMENT OF
MODEL SIMULATIONS

Giorgi and Bates (1989) began a validation of the
CCM-MM4 nested modeling approach using observations
for 40 months from January 1982 to December 1,983, and
January 1988 to April 1989. They used initial conditions
from ECMWF (European Center for Medium-range Weather
Forecasts) analyses of observed data gridded to a spatial
resolution of 2.8 by 2.80 and having temporal resolution of
12 h. MM4's lateral boundary conditions were adjusted
every 12 h using the ECMWF gridded data. These initial
conditions were located at grid points that matched the CCM
T-42 (triangular spherical harmonic truncation at wave
number 42) used in climate simulations.

Giorgi et al. (1992a) have validated mean monthly
precipitation predict1ed by the MM4 model during the 1982-
83 and 1988-89 peliods. They found that the correlation
coefficient between observed and modeled monthly
precipitation varied from 0.53 to 0.94 among each of six
subregions of the Western United States. Over the Central
Rocky Mountains, cold season precipitation was reasonably
well simulated with a bias of 40%, however, the wann
season bias was 157% in the base runs. The revised
horizontal diffusion method lowered this bias to 39% in the