WMOD00582

• Title: Natural Variability of Thermodynamic Features Affecting Convetive Cloud Growth and Dynamic Seeding: A Comparative Summary of Three High Plain Sites from 1975 to 1977
• Date: 9/9/1981
• Weather Modification - Doc Type: Report

VOLUME 20

974

JOURNAL OF APPLIED METEOROLOGY

c. Predicted versus observed cloud occurrences

TABLE 2. Model-predicted cloud development versus observed
convective clouds in 1976 samples at MLS, GLD and BGS, and
1977 at GLD and BGS.

Model-predicted clouds
Site MLS GLD BGS
Observed
convective clouds Yes No Yes No Yes No
1976
All -Yes 61 0 64 I 62 0
clouds * -No 3 \ 0 10 0 4 0
1977
All -Yes 51 8 63 7t 66 2t
clouds * -No 6 2 I 2 10 0
1977
All -Yes 36 7 47 6t 57 2t
clouds** -No 21 3 18 2 19 0
* All convective clouds wit~in 250 km of the sounding .site.
** All convective ~Iouds within 75 km of the sounding site.
t Raob top below 200 mb.

those of observed clouds in the comparison of
individual clouds (Warner, 1970). A series of six
different initial cloud updraft radii were simulated
in all model runs~ These updraft radii, control the
amount of entrainment as simulated by Stommel's
(1947) empirical (0.2/ R) relationship, where R is the
updraft radiu's 'in meters. The radii used were 0.5,
1.0, 1.5,2.0,3.0 and '10.0 km. The larger radii (2.0,
3.0 and 10.0 kni) were assumed to simulate larger
cumulus congestus and cumulonimbus cloud sys-
tems; whereas the smaller radii (0.5-1.5 km)
simulated smaller convective clouds with larger
amounts of entr'ainment, which reduced cloud de-
velopment. The dynamics of a steady-state en-
training jet also limit the model by not explicitly
simulating the effects of cloud:top mixing (Scorer,
1958; Levine, 1959). The explicit effects of vertIcal
wind shear and interactions between adjacent cells
as in mesoscale lines are not simulated.'
The model objectively determines the buoyancy
of a parcel, then adds, in a parameterized form, the
effect of precipitation drag' and cooling effects of
mixing to determine the highest level to which a par-
cel of given radius may be expected to rise under the
initial conditions imposed by an observed rawin-
sonde, cloud-base height and updraft radius. Simp-
son and Wiggert (1971) in Florida, Dennis and
Schock (1971) in South Dakota and Matthews and
Henz (1975) in' Colorado used aircraft and radar
observations to (;onfirm the reliabilIty of these
models for diagnosing the potential for convective
development. However, caution should be used in
extrapolating results to other areas and periods be-
cause model results are highly dependent on the
rawinsonde input data.

The model results were compared with cloud
observations, as indicated from satellite imagery, to
determine how often the model over- and under-
predicted the presence of convective clouds. Pre-
vious studies (Weinstein, 1972a) failed to check the
existence of clouds.
Model reliability was analyzed using satellite
observations of clouds observed at 2300 GMT in
1976 and 1977. Results indicated that the model
could diagnose convective clouds on nearly all days
that they occurred; however, in 10-20% of the
cases the model diagnosed clouds when none were
observed within 250 km of the field site. Table 2
summarizes the model reliability in, terms of ob-
served convective clouds. In 1976 all convective
clouds within 250 km of the rawinsonde were. com-
pared. Note that the average separation between
National Weather Service rawinsondes is approxi-
'mately twice this distance in the High Plains. Results
in 1977 were divided into two classes-alll;touds
occurring within 75 km of the site and all l;louds
within 250 kmof the site. For instance, Goodland's
(GLD) 1977 re.sults showed that the model missed.
14% of the cases when all clouds within 250 km
were compared; however, it missed 36% ofthe cases
within 75 km of the site. These results indicate that
the model was able to diagnose the wide area poten-
tial for clouds (76-86% correct) better than the site
potential. This in part was, due to. the need for
mesoscale triggering to release the convective in-
stability as discussed by, Matthews and Silverman
(1980), using similar rawinsonde. data from t\1is
region. The GPCM model doe.s not explicitly ac-
co.unt for this release mechanism. Mesoscale cloud
lines were often observed bey~nd 75 km from G],.D,
while GLD remained in the clear, hence, poorer
verification there. The three~dimensional analysis
of mesoscale triggering through low-level conver-
gence, upper~level divergence, local orographic and
isentropic lifting should be performed. The lack of
this triggering is a critical limitation of one-dimen-
sional steady-state cloud models.
The skill ofthe GPCMmodel in predicting the
presence of clo.uds is primarily a function of its
ability to select cloud-base height. Initial cloud up-
draft diameter and the, initial impulse of2 m S-l deter-
mine the relative intensity or dep~h of convective
growth, but do. not affect the model's prediction of
cloud occurrence. Cloud-base height is computed
from the initial sounding at the CCL using a mixing
layer of the lowest 5 kPa. When the amount of sur-
face temperature rise, exceeds the climatological
extreme ranges of templerature , the model rejects the
sounding as requiring excessive surface heating for
convective cloud initiation. With this extreme
temperature range criteria, the model tended to

I
.I