Report for Contract
State of Colorado
Dept. of Natural
Development of New Methodologies
for Determining Extreme Rainfall
William R. Cotton,
Ray L. McAnelly and
Dept. of Atmospheric
November 5, 2001
Since our last report and our meeting with DNR personnel in Denver
on 20 July 2001, we
have performed additional perturbation simulations of two cases that we
have previously described: the Fort Collins
flood event of 28 July 1997,
and the Big Thompson flood event of 31
July 1976. Because these cases and our initial simulations
have been described in some detail in earlier reports, these ensemble runs
are only summarized briefly below. However, their results will be
incorporated into our aggregate results and the final report.
We have also performed simulations of two additional
extreme precipitation cases: the Dallas Divide flood of 31 July 1999, and the Park Range
heavy rains over 18-22 September 1997.
Although both of these events occurred more recently than the cases
surveyed by McKee and Doesken (1997), Doesken (personal communication)
says that both events should be included in an updated "recommended
final list of storms for consideration in investigating extreme rainfall
potential in the Rocky Mountain region of Colorado" (their Table 5).
These cases are overviewed and the results of the simulations performed to
date are presented below.
2. Additional simulations of the Fort Collins event of 28
simulations of the Fort Collins event were performed with the parallel
version of RAMS 4.29. All of these simulations were initialized with the NCEP reanalysis
pressure level data from 1200 UTC 28 July 1997. The control simulation was initialized
with a horizontally and vertically homogenous soil moisture field
corresponding to 50% saturation (0.21 m3/m3). Four simulations utilized an elevation dependent
soil moisture field of 50% and 30%, 50% and 70%, 30% and 50%, and 70% and
50%, respectively, for mountain and plain regions. The elevation
separating these two regions was chosen to be 6000 ft. The purpose of
these simulations was to examine the effect of high- and low-elevation
soil moisture changes on simulated high- and low-elevation
accumulated precipitation maxima. The primary conclusion drawn from these
sets of simulations was that the soil moisture variations within an
elevation category most closely controls the maximum accumulated
precipitation within that category. However, the partitioning of total
domain precipitation between high and low elevation categories is more
heavily influenced by low-elevation or plains soil moisture
variations, particularly in the case of dry (30%) plains soil moisture.
The final simulation of the Fort Collins event was one in which the initial
relative humidity was raised to 98% below 700 mb.
simulation did not produce larger Grid 4 accumulated precipitation maxima. However, the initial low-level moisture field
was already nearly saturated over northeastern Colorado and this perturbation resulted in an
increased of less than 1 g/k of water vapor at the surface. Maximum total precipitation in these runs
ranged from 16.7 inches in the control run to 11.4 inches in the final run,
with their positions occurring at various locations over the pains and at
3. Additional simulations
of the Big Thompson event of 31
Three simulations were also performed for the Big Thomson
flash flood event, initialized with NCEP reanalysis pressure level data
from 1200 UTC 31 July 1976. The first utilized horizontally
homegenous 50% saturation soil moisture initialization. The second
simulation utilized 35% soil moisture initialization below 5600 ft and 50%
above 5600 ft. This was done in an attempt to improve the low-level
surface temperature and dewpoint forecast.
The third simulation incorporated surface observations in
the initialization as well as the NCEP reanalysis data, in an attempt
to improve the intial near-surface temperature and moisture fields.
Precipitation maxima in these three runs were 11.2, 7.0 and 5.9 inches,
respectively; the first value is close to the observed maximum value. However, none of these simulations
reproduced the flooding event over the Big Thompson drainage, and the
inclusion of surface observations did not improve this aspect of the
Divide flood of 31 July 1999
The Dallas Divide flood occured on the
afternoon of 31
July 1999. We classified this
storm as a Local Convective event according to McKee and Doesken's (1997)
classification scheme, and a Type IV event according to the synoptic
classification scheme of Maddox et al. (1980).
Observational information available to us for this event includes
a case study by National Weather Service forecasters in Grand Junction
(Avery et al., 2001), a detailed analysis of radar and lightning data of
the event by Henz (2000), and a survey of the hydrogeological effects of
the flood by Jarrett (personal communication).
Four simulations have been performed for the Dallas Divide
event. For this and other more recent cases, 40-km ETA model analyses are
available and can be utilized for initialization, providing much greater
horizontal and vertical resolution than the NCEP reanalysis data. The
four simulations consist of the following: (A) CONTROL -- initialized at
1200 UTC 31 July 1999 using the 40-km ETA model analyses for the
atmospheric fields and 70% homogenous soil moisture initialization; (B)
ETASOIL -- Same as "A" except initial soil moisture and
temperature is based on the ETA, 4-level soil model analysis; (C) DAL-PAR
-- Same as "A" except that the initial fields and Grids 3 and 4
were shifted northward to the Parachute, Colorado area; and (D) Same as
"B" except Grids 3 and 4 were relocated near Colorado Springs,
with no shifting of the initial fields. One further simulation with
increased atmospheric water vapor was performed, but increased cloudiness
and reduced solar heating prevented significant convection from
developing, so those results were discarded. Only the control simulation A
is described here, while the results of simulations B, C and D are also
included in our aggregate results.
The event occurred in a moist monsoon environment, in
deep southwesterly flow ahead of a weak mid-level shortwave in Utah.
Figure 1 shows the initial pressure, relative humidity, and wind fields at
about 4.5km MSL (about 600 mb), on the RAMS coarse
Figure 1. Relative humidity,
pressure and wind fields on Grid 1 at 4.5km MSL, at initial time of simulation
A for the Dallas Divide event. Nested Grids
2-4 are indicated by dashed boxes.
grid, along with the
nested grid locations, for simulation A. The observed storm system
consisted of a series of several cells on the upper reaches of the Dallas
Creek drainage. They translated and propagated northeastward down the
watershed, with subsequent cells forming successively further up the
drainage toward Dallas Divide and thus raining into the same watershed
(Henz, 2000). Avery et al. (2001)
reported that during the latter stages, the heaviest rain progressed
slightly southward. Maximum observed rainfall was almost 3 inches (Avery
et al., 2001), while Henz calculated peak rainfall amounts of 4-5 inches
based on radar data. We do not know whether frozen precipitation was
observed or diagnosed with radar data.
Figure 2 shows the total precipitation accumulation on Grid
4 in simulation A (heavy precipitation contours
overlaid on shaded topography).
The simulated storm system in
Figure 2. Total precipitation contours (heavy
contours beginning at 1mm and at 25mm or ~1-inch increments) overlaid on shaded
topography for simulation A of the Dallas Divide event.
"A" evolved in a very similar manner to
the observed evolution described by Henz (2000) and Avery et al.
(2001), except the similated scenario occurred 15-18 km to the southeast
of Dallas Divide (DDV), in the higher elevation region surrounding Mount
Sneffels. Several distinct
cells originated near Telluride (TEL), intensified as they tracked over Mt.
Sneffels, and subsequently weakened
as they moved downslope toward the upper Uncompahgre
River above Dallas (DAL)
and Ridgeway. Most of these had almost identical tracks, with the
multi-cell system eventually propagating southeastward away from the
previous dominant track. The resultant maximum total precipitation
simulated on Grid 4 (1.67-km grid spacing) is 19.2 cm (7.6 in.), just to
the lee (northeast) of Mount Sneffels. This
accumulation occurs in the 4.5 hour period between 2100 UTC 31 July and
0130 UTC 1 August, just a little later than observed.
Because the simulated event is at higher elevations than the
observed storm, the precipitation pattern in Fig. 2 includes both liquid
and ice precipitation. The rainfall pattern (not shown) has two distinct
maxima, one on the upstream side of Mt.
Sneffels where the cells
intensified, and a larger maximum of 14.5 cm (5.71 inches) several
kilometers to the northeast of the precipitation maximum in Fig. 2, at an
elevation of about 2800m. Between these two rain maxima, hail was the
predominant form of precipitation over the highest terrain, with a maximum
hail accumulation (liquid water equivalent) of 16.8 cm (6.6 in.). Graupel
accumulations were less than 0.1 mm. Thus the simulated rainfall maximum
was just a little more than that estimated by
Henz (2000), although its location is not over the main Dallas Creek
watershed as observed.
5. Park Range
heavy rain event of 18-22 September 1997
The Park Range storm system was
classified as a GLC event (General storm system with embedded Local
Convection) and a Type III event, respectively, using the classification
schemes of McKee and Doesken (1997) and Maddox et al. (1980). The several
day precipitation event was due to a synoptic scale wave that dug into Nevada,
closed off, and gradually moved eastward through Utah
and southern Wyoming. Like the
previous large-scale event we simulated that impacted the San Juans on 4-6
September 1970, this event also was influenced by moisture from a tropical
storm (Linda) in the eastern Pacific.
simulation for this fairly recent event was initialized with the 40-km ETA
analysis at 0000 UTC 18 September 1997,
for both the atmospheric fields and soil moisture and temperature fields.
Figure 3 shows the 500 mb relative humidity, height and wind fields on
Grid 1 at 1200 UTC on the 19th, 36h into the simulation, when the low was
digging into central Nevada. As in the San Juan simulation of 4-6 September
1970, we found that with a larger Grid 1 extending well into the Pacific,
the wave did not developed sufficiently; thus it was necessary to draw in
the western boundary onto the U.S. mainland in order to directly nudge
the wave into our domain. For the finest nested Grid 4 in this GLC event,
we used a 2km grid spacing, slightly coarser than
the 1.67km spacing we use in the LC events but finer than the 3km spacing
used in the General event over the San Juans. This compromise enables the
widespread rains affecting the entire Park Range to
be included on Grid 4, while still reasonably resolving convective
processes of the embedded convection.
We aren't aware of
any detailed precipitation analyses of this event. However, Doesken (personal communication)
provided us recording raingage data from two National Atmospheric
Deposition Program (NADP) precipitation sites in the Park Range
north of Steamboat Springs: the Tower site on Buffalo
Pass and the Dry
Lake site about 700m lower in
elevation and 8km west of the Tower site. These offer excellent
time-resolved data, and are supplemented by on-line daily rain amounts at
other NADP sites and
Figure 3. Relative
humidity, height and wind fields on Grid 1 at 500mb, at 36h into control
simulation for the Park Range event. Nested Grids 2-4 are indicated by
by routine daily and hourly
NWS and cooperative precipitation data. These data indicate five-day total
rains of at least 2-3 in. over most of the northwest quadrant of Colorado,
4-5 in. over intermediate elevations such as in the Flattops, and with
maximum observed amounts of 6.19 and 8.08 in., respectively, at the Dry
Lake and Tower sites in the Park Range. No data are available to indicate
whether any frozen precipitation occurred.
Only the control simulation of the Park Range
event has been completed thus far; another simulation with enhanced
lower-tropospheric moisture is underway. The control simulation was run
through 0000 UTC 24 September 1997. The maximum of the accumulated precipitation
field (Fig. 4) is 24.2 cm (9.53 in.), occurring approximately 20 km to the
north-northeast of Steamboat Springs (SBS) in the Mount Zirkel Wilderness
at about 3200m (10,240 ft.) elevation. Approximately 13-14 cm (~5.5 in.)
of this maximum is hail. The maximum rain accumulation (not shown) of 18
cm (~7 in.) occurs several hundred meters lower in elevation, about 2-3 km
to the north of the Dry Lake site (DLK). Graupel accumulation is only 3-7
mm at the location of maximum total precipitation and is less than 1-mm at
the location of the maximum rain accumulation. The time distribution of
the simulated precipitation is in reasonable agreement with the
observations, although the synoptic wave remained somewhat stronger and
produced precipitation in the Park Range for a day
longer than observed.
Figure 4. Total precipitation contours (heavy
contours beginning at 1mm and at 25mm or ~1-inch increments) overlaid on shaded
topography for control simulation of the Park Range event.
Avery, B.A., C.N. Jones,
J.D. Colton, and M.P. Meyers,
2001: A southwest Colorado mountain flash flood in
an enhanced monsoonal environment. Online paper at
http://www.crh.noaa.gov/gjt/science.htm, National Weather Service, Grand Junction, Colorado.
Henz, J.F., 2000:
Cloud-to-ground lightning relationaships to flash flooding in western Colorado monsoon thunderstorms. Poster presentation, Southwest Weather Symposium (Tucson, AZ),
National Weather Service, University of Arizona, and
Maddox, R.A., F. Canova,
and L.R. Hoxit, 1980: Meteorological
characteristics of flash flood events over the western United States. Mon. Wea. Rev., 108, 1866-1877.
McKee, Thomas B., and
Nolan J. Doesken, 1997: Colorado extreme storm
precipitation data study. Final Report. Summary of accomplishment and work performed February
through October 31, 1996. Climatology Report 97-1, CSU, Fort Collins,
CO, May, 107 pp.