WMOD00311

• Title: Harold D. Orville Symposium - Forty Years of Modeling Clouds and Weather Modification
• Date: 4/26/2003
• Weather Modification - Doc Type: Report

efficient lookup tables, but this is probably not
important in the overall performance of the
scheme. As other investigators had done, ice
multiplication by rime splintering was included
(Hallett and Mossop, 1974; Heymsfield and
Mossop, 1984; Mossop, 1985). Finally, the
deposition and condensation freezing
parameterization of Meyers et al. (1992) was
incorporated into the scheme, increasing the
number concentrations of ice crystals at middle
levels.

The challenge was not in simulating the radar
reflectivity structure of strong continental
convection, but rather that of weaker oceanic
convection. I spent most of my efforts on looking
at a couple of cases from GATE (tropical East
Atlantic) and TOGA COARE (tropical Western
Pacific). We concentrated most of our efforts at
GSFC on 2D simulations of squall lines, in which
the convection was initiated by convergence and
lifting from initial low-level cold pools. Maximum
radar reflectivities in GATE convective cores
rarely exceeded 50 dBZ, and yet early versions of
the 4ice scheme produced echoes approaching
65 dBZ! Substantial improvement in the
convective echo structure was obtained when I
realized that conserving particle number
concentrations when converting between different
hydrometeor species could cause spurious
changes (or lack of conservation) in the higher
particle moments. Accurate calculations of radar
reflectivities, precipitation rates, and other higher
moments required preserving the width of the
particle spectra, as represented by the slope of an
exponential distribution.

Likely the most controversial aspect of this
scheme was a crude representation of the ice
enhancement results from the University of
Washington (Hobbs and Rangno, 1985, 1990;
Rangno and Hobbs, 1991). It helped to initiate
large numbers of ice particles during the early
stages of squall-line evolution when convective
updrafts were strong enough to carry some rain
above the freezing level. The difficulty with rime
splintering, as I saw it, was that it required
precipitation ice ("rimers") and cloud water to
coexist in a relatively narrow temperature range
(particle surface temperatures from -30C to -80C).
Ironically, including the Hobbs-Rangno ice
enhancement provided a sufficient number of
rimers to form in the updrafts and allow the rime-
splintering mechanism to become effective. Again
looking back at the situation, maybe part of the

problem was our methods of convective
initialization? I tried runs where convection
propagated through a pre-existing ice cloud and
with imposed larger-scale vertical motion, but my
concern was that there was a lack of observational
support for bog using in such effects.

I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I

I seriously considered including the number
concentrations of cloud water and rain, much as
what was originally designed by Ziegler (1985) or
using the more recent two-moment approach of
Meyers (1995), but decided to wait until the ice
scheme was formulated first. Meyers (1995)
developed a comprehensive two-moment
microphysical scheme that calculated number
concentrations for all forms of condensate.

Another double-moment bulk scheme has
been developed for use in MM5 (Reisner et al.,
1998). A version of the Reisner microphysics
where the number concentrations of ice are
calculated has been running operationally in the
Rapid Update Cycle (RUC) for the past several
years. An early version scheme produced too
much graupel in the 40-km RUC model, but this
has sense been modified to yield more reasonable
results now running in the 20-km RUC (Brown et
al., 2000). At NCAR and at Forecast Systems
Laboratory (FSL), plans are to calculate the
number concentrations of cloud water, rain, and
snow in the RUC over the next few years, primarily
for aviation interests.

There are several drawbacks in using multiple-
moment schemes. First, they are computationally
more expensive than single-moment bulk
schemes. Second, they require an accurate,
positive-definite advection scheme to prevent
decoupling of mixing ratio from number
concentration. Third, they sometimes behave
unpredictably and are more sensitive to small
perturbations in the initial conditions than simple
bulk schemes.

6. BIN MICROPHYSICAL APPROACHES

Biri microphysical models explicitly calculate
particle number concentrations in different size
categories. Early formulations focused on the
evolution of the drop spectrum from the onset of
cloud droplets by condensation through large drop
self collection (e.g., Berry and Reinhardt, 1974;
Soong, 1974; Takahashi, 1981), while more recent
formulations have also included size categories for
cloud condensation nuclei (e.g., Kogan, 1991).

8