FLOOD07101

• County: Statewide
• Basin: Statewide
• Title: A Numerical Modeling Study to Investigate the Assumptions Used in the Calculation of Probable Maximum Precipitation
• Date: 3/1/1999
• Prepared By: Water Resources Research
• Floodplain - Doc Type: Educational/Technical/Reference Information

794

ABBS: INVESTIGATION OF PROBABLE MAXIMUM PRECIPITATION ASSUMPTIONS

niqll<: for areas of 50-70 km2, For areas of 500 km2 the model
produces between 5% and 15% more precipitation than the
maximization relationship of the current PMP technique.
Hcn:e the precipitation is not linearly related to the precipi-
tabk water.
The 2-hour PMP values for the Oberon and Chifley Dam
catchments and the 24-hour PMP values for the catchments of
the Chifley and Warragamba Dams and the Hawkesbury-
Nepean catchment [Pearce et a/.. 1992, 1993] have been in-
cluded on Figure 8 for comparison, The 2-hour PM? values
have been determined using the GSDM [Bureau of Meteoro/-
ogy, 1984], The 24-hour PMP values have heen determined
osing the GSAM IMinO' et ai"~ 1996J, For both durations the
model has been able to simulate a small art a storm approach-
ing the PMP values for the Oberon and Chifley Dam catch-
ments. For larger areas the 24.hoUT PM? \ialues for the War-
ragamoa Dam and the Hawkesbury~Ncpc,ln catchments lies
well above the depth-area CUNes for the east coast low of
August 1986. This indicates that for this plrticular storm the
model hns not been able to be maximized to a level approach-
ing ;1 PMP storm. However, it may be possible to maximize
other storms to the PMP values for larger areas.

4. Discussion and Conclusions

4.1. Discussion

~11 this paper we have shown that numerical models can be
used for quantitative precipitation forecasting (QPF) ofsnuth-
cast':rn Australian stor01s and may provide results at a hori-
zontal resolution of 5-10 km. This requires the use of an
appropriate mesoscale model (such as RAMS) coupled with <l
convective parameterization scheme that is suitahle for use at
a horizontal resolution of 5-10 km, The importance of this
approach has-not been dealt with here but should be consid-
ered if numerical models are to be used for QPF in the future.
Abns and Lee [1997] discuss this requirement in greater detail
and provide examples for two case studie~. The major app1i~
cation has been to use numerical models as a tool to assess
some of the assumptions used in the current method of esti-
mating PMP,
These techniques have been tested on four case studies: (1)
an east coast low, (2) an upper-level cutoff low, (3) a northwest
Australian tropical cyclone, and (4) a northeast Australian
tropical cyclone IAbbs alld Ryan, 1997], Only the results of
study 1 are presented in this paper. For eae h of the case studies
we have investigated the effects that increases in the moisture
avaiJahility have on the precipitation produced hy the storm,
on the precipitation efficiency of the storm, and on the DDA
analyses for the Morm. We have also investigated the effect of
terrain on the distribution of the precipitation produced by the
ca~t coast low and upper-level cutoff low.
The increased moisture simulations havl~ been performed by
initializing the model from the ECMWF "nalyses. but in Ihcse
ca~cs the moisture values have heen increased hy uniformly
increasing the temperatures of the atmosphere everywhere
while maintaining the relative humidities In this way the sys-
tem under investigation is still in dynamic balance, but tbe
specific humidity, and hence the surface d,~wpoint temperature
and precipitable water, has been increased. This method is
comparahle to the technique used 10 maximize storms in the
current PMP approach. A difference between this technique
. :\nd thnt currently employed is that the nitiill atmosphere in
,~ il': modeling studies is not necessarily sat'lrated; in the current

method of estimating PMP the atmosphere is assumed to he
saturated throughout its depth.
The maximization factors that result from the technique
used in this report lie between 1.1 and 1.3 and are limited hy
the ability of the numerical model to handle large changes to
the model's initial conditions, In contrast, MillO' et ai, [1996]
calculate values >2 but impose an upper bound of 1.8 on the
maximization factors that are used to estimate the PMP of a
storm. The inability of the numerical model to simulate these
events with such large maximization factors raises questions as
to the validity of the large maximization factors for these par~
ticular storms. Although the dewpoint temperatures associated
with these maximization factors may occur at that particular
location and time of year, it remains to be determined whether
they could actually occur immediately preceding such extreme
precipitation events. Despite the difference in the upper limit
of the maximization factor that is used for a particular storm it
is still possible to investigate some of the assumptions (section
I) that arc used in the calculation of the PMP,

4.2. Conclusions

Five conclusions, with implications for the estimation of
PMP, may be drawn from this study. These conclusions arc
based on the results from the four cases studied, rather than
from the single case discussed here to illustrate the results.
1. As the moisture availability is increased, the precipita-
tion efficiency of the storms does not change significantly. For
each case study investigated the production of heavy rainfall
(rainfall rates >25 mm hr-I) is hetween 80% and 100% efli-
cient. This supports the simple model that assumes implicitly
that extreme precipitation storms have the highest efficiency.
2. As the moisture availability is increased, the duration or
the heavy rainfall increases, begins earlier, and is more con-
tinuous. Life cycles are not considered in the simple model.
although the results presented here and in the recent paper of
Zhao et al. [1997J suggest that the duration 'of the storm in-
creases as the moisture availability increases. An increase in
the duration of heavy rainfall will result in higher total rainfall.
Long~lived moderate rain processes are also important con~
trinutors to the total precipitation produced by the storm.
3. As the moisture availability changes. the spatial dislri~
but ion of the area over which more than 50% of the total
rainfall falls as heavy rainfall changes, Zhao et ai, (19971 also
found that the areal coverage of rainfall varies nonlinearly with
the precipitable water.
4. If the depth-area CUlVe for the increased moisture sim-
ul:.ltion lies above that of the control simulation, then the
maximization relationship of the CUrrent PMP technique un-
derestimates the precipitation compared with that simulatcll
by the model. The simulations reported here indicate that this
may occur; hence the precipitation is not linearly related to the
precipitable water. For the case studies presented in section 3
the model produces between 15% and 35% more precipitation
than the current storm maximization technique for areas of
50-70 km2 For areas of 500 km2 the model produces between
5% and 15% more precipitation then the current storm max-
imization technique.
5, Terrain effects have an effect hoth on the amount of
rainfall that nccurs over the higher terrain and on the distri-
bution of the rainfall due to the convergence component of the
storm. This is due to changes in the movement of the storm or
changes in the low-level wind field (e,g" hloeking) when terrain
effects arc neglected. The temporal variation of the precipita-

I
I
,
t
r
,
I
I
i
!
I
I
t
I
[
,
i
I
I
I
I
I
f
I
f
,
t
I