A SOUTHWEST COLORADO MOUNTAIN FLASH FLOOD IN AN ENHANCED
Brian A. Avery *, Christopher N. Jones, Jeffery D. Colton, and Michael P. Meyers
National Weather Service, Grand Junction, Colorado
On 31 July 1999 thunderstorms with heavy rain caused widespread damage to the small county of Ouray, Colorado's tenth least-populated county. Rainfall exceeding the 100-year event criterion contributed to the damage of nearly two miles of County Highway 24, and Colorado State Highway 62 was undercut in three places by floodwater. Dallas Creek, a tributary of the Uncompahgre River, swelled to 51 times its normal volume, resulting in the destruction of the waterway's only river gaging station. Four public bridges were damaged, and a fifth was rendered impassable, isolating several families on Log Hill Mesa and hampering access of emergency vehicles. Total public damage from these storms totaled over $750,000, half of the county's highway budget and nearly a quarter of the entire county annual budget. Private property damages included $100,000 to bridges and structures, and $20,000 to livestock and hay. No fatalities were associated with this event, but seven people had to be evacuated from an inundated barn by a local mountain rescue team. This study identifies the unique meteorological, hydrological and geographical factors that worked in tandem to create a devastating and financially-crippling flood in this sparsely populated location.
2. SOUTHWEST MONSOON BACKGROUND
Increased moisture and more frequent thunderstorms in the southwest United States are attributed each summer to the seasonal shift of lower and middle tropospheric winds to a southerly direction. Locally this wind shift and moisture increase is referred to as the "southwest monsoon." The monsoonal moisture advects north around the western periphery of a Bermuda High. Several studies (e.g. Reitan 1957, and Hansen 1975) have shown the primary moisture source of the southwest monsoon to be the Gulf of California and eastern Pacific Ocean. The seasonal northward transport of moisture has been partly ascribed to a low-level jet (Tang and Reiter 1984), with the farthest push north observed during the months of July and August (Hales 1974). Several historical flash floods of note along the Front Range of southern Wyoming and northern Colorado (Big Thompson 1976, Cheyenne 1985, and Fort Collins 1997) have occurred during late July and early August, with each strongly tied to the southwest monsoon. A study by Weaver and Doesken (1990) showed that the recurrence probability for a catastrophic severe weather event along the Front Range was greatest during this period. This is especially likely during very active monsoon seasons. Monsoonal "bursts", the convectively unstable periods of the event, can be particularly wet and enduring during La Niña years. During cold phases in the Eastern Pacific (JMA SST index is 0.5C below average for six consecutive months), enhanced tropical activity and intensity associated with La Niña often increases the available moisture for transport into the southwest United States (Bove et al. 1998).
3. SYNOPTIC ENVIRONMENT
The 1999 Colorado monsoon season began early in the second week of July and daily convective activity was prevalent throughout the southwest corner of the state for the remainder of the month. By the end of July, a prolonged wet period had resulted in near-saturated soils and above-normal streamflows throughout the San Juan mountains.
On 31 July 1999, the 1200 UTC (times hereafter UTC) 500 hPa pattern showed sprawling high pressure over the Gulf Coast states with the western edge extending to west Texas . Broad zonal flow was present over the northern tier of states, with a diffuse pattern over the desert Southwest (Fig. 1).
Fig. 1. NGM 500 hPa height analysis (solid line), and 700 hPa dewpoint analysis (degree C - solid line). Large arrow denotes 250 hPa jet maximum. Dashed bold line denotes 700 hPa shortwave. Note: click on image to enlarge.
Careful examination of infrared satellite (IR) imagery indicated an east-moving shortwave extending from central Wyoming to southern Utah, not evident in analyses of the 700 hPa or 500 hPa height fields. Additionally, tropical moisture was observed moving north from the eastern Pacific Ocean per water vapor imagery, and through analyses of the 700 hPa isodrosotherm field. This synoptic environment displayed characteristics of both the Type I and IV scenarios of flash flooding described by Maddox (1980) for the western United States. The surface synoptic pattern for the same time indicated no discernable front in the Great Basin or surrounding region. Surface dew points of 14C to 17C were in place across southwest Colorado, with surface temperatures running only a few degrees higher. The dew point values observed in this case compared favorably to those documented in other devastating western United States flash floods (e.g. Corrigan and Vogel 1993, Knievel and Johnson 1998, and Runk and Kosier 1998).
A combination of observing systems were utilized to interrogate the unusual convective environment over southwestern Colorado on 31 July 1999. Regional soundings and wind profilers provided valuable information regarding the highly unstable and weak shear environment. High-resolution satellite imagery enabled the detection of an important convergence boundary, and the WSR-88D revealed the training of slow-moving storms over the Dallas Creek drainage.
4.1 Thermodynamic Structure
Grand Junction, Colorado (KGJT), located 120 km northwest of the flash flood site, provided the nearest upper-air sounding data for this case. The 31 July 1200 KGJT sounding indicated unusually high morning instability, evidenced by convective available potential energy (CAPE) of 1683 J kg-1 and a lifted index (LI) of -5C. Convective inhibition (CIN) was 0 J kg-1. Utilizing the 26C maximum temperature observed at Ridgway, Colorado (7 km east of the thunderstorm genesis region) on 31 July yielded a modified LI of -6C and CAPE in excess of 2500 J kg-1. Monsoonal moisture pushed precipitable water (PW) values to 28.7 mm, approximately 170% of normal. Studies have shown that other devastating flash floods occurring west of the Continental Divide have exhibited similar PW values. A southern Nevada flash flood documented by Runk and Kosier (1998) displayed nearby PW values of 26.4 mm, while a flash flood near Opal, Wyoming revealed representative PW readings of 29.0 mm (Corrigan and Vogel 1993). A comparison of the 31 July 0000 and 1200 KGJT soundings showed that moisture below 700 hPa had increased. East winds were present through 700 hPa on the 1200 KGJT sounding, a likely result of contamination from local drainage winds at KGJT. Above this level, winds depicted on the 1200 KGJT sounding were representative of the larger-scale flow. The wind was uniformly from the southwest through 250 hPa, with speeds increasing from 10 kts to 45 kts between 600 hPa and 250 hPa. Weak wind shear of 8 m/s was present within the 0 to 6 km layer. A wind profiler at Aztec, New Mexico (150 km south of the flood site) confirmed that the shear within this layer remained nearly unchanged through 1 August 0000. Forecasts of storm motion and the mean wind vector indicated the convective environment was favorable for backwards propagation. Typical of the mountain-valley wind circulation, it is likely that a diurnal northeast upslope wind would have been present by the time of initial convection; however, a lack of local wind observations precludes definitive identification.
The Summer of 1999 occurred during a La Niña episode, and was an active period for convection in the eastern Pacific Ocean. Three tropical systems had formed west of Mexico during the week prior to 31 July. Early morning water vapor imagery on 31 July indicated that the Four Corners region was on the western edge of an advancing tropical moisture plume associated with these systems.
The tail of the afore-mentioned shortwave was seen on satellite images moving east from central Utah to the Continental Divide between 31 July 1800 and 1 August 0200. Minimum convective cloud top temperatures over Ouray County fell in the warm top range of -47C to -52C, and identified closely with the single-clustered thunderstorms described by Spayd and Scofield (1983). The most important feature ascertained through satellite examination was that of a quasi-stationary thunderstorm outflow boundary elongated along the Uncompahgre Plateau. This feature was evident in 1 km resolution visible imagery at 1715 (Fig. 2), and was oriented northwest-southeast on the south side of the plateau.
Fig. 2. 1 km visible satellite photo depicting
boundary in southwest Colorado at 1715 UTC.
The tail of the boundary extended southeast to the genesis region of the flash flood producing thunderstorms, and became the focus for the initial convection. A similar quasi-stationary boundary was found by Runk and Kosier (1998) to be a mechanism that helped initiate convection. In the Dallas Creek case, a large area of subsidence was located west of the genesis region in conjunction with the terrain-anchored outflow boundary. This cloud-free area lacked cloud development until the late afternoon hours, well after the onset of nearby convection. Visible imagery at 2 km resolution indicated that thunderstorm training occurred across the quasi-stationary boundary east of the subsidence field during the time period 1815 to 2315. In addition to the outflow boundary and mesohigh, overshooting tops were seen between 2045 and 2245 at the 2 km resolution. This approximates the time that the shortwave identified earlier had tracked through western Colorado.
The evolution of the precipitation estimation pattern from the KGJX WSR-88D radar (located at a height of 3048 m MSL, 105 km north of the flood site) confirmed that the thunderstorms were regenerating and moving over the same area of western Ouray county. From thunderstorm initiation until approximately 2200 the heaviest precipitation was oriented over the Pleasant Valley drainage. Peak reflectivity of 56 dBZ was observed at 2131 over this basin. The axis of heaviest precipitation and maximum reflectivity shifted south and east to the West Fork Dallas Creek drainage after 2200. The maximum reflectivity noted over this basin was 44 dBZ at 2247. The times of peak reflectivity corresponded very well with observed maximum rainfall rates over the respective drainage basins. The south and east shift of the heavy rain axis occurred on the right rear flank of the thunderstorm cluster. Chappell (1986) described this as the most active section of a storm as it becomes stationary, and the region most likely to cause heavy rain and subsequent flooding. Reflectivity data also indicated that the storms gradually weakened as they moved away from the higher terrain. Low-level boundaries were not discernable through radar reflectivity data due to the elevation of the KGJX WSR-88D and its distance from the Dallas Creek watershed.
5. NGM FORECAST CONDITIONS
A comparison of the NGM and ETA models indicated that the NGM provided the best forecast of the monsoonal flow and the weak shortwave west of the region. The ETA typically exhibits a tendency to underpredict the amount and coverage of precipitation in the southwest United States during the monsoon season (Hydrometeorological Prediction Center staff - personal communication). This bias was frequently noted by operational forecasters at Grand Junction during the summer of 1999; therefore only NGM forecasts from the 31 July 1200 model run are presented here.
Observed 850 hPa dew points at 1200 varied from 14C to 16C over southwest Colorado. The NGM depicted a tight 850 hPa e gradient across western Colorado between 31 July 1800 and 1 August 0000. Previously, Shi and Scofield (1987) have shown that flash flood producing thunderstorms often form along the e gradient. Six-hour forecast total precipitation amounts (valid 1 August 0000) were near one inch in southwest Colorado, but a four inch bull's-eye was forecast over south central Colorado. However, forecast PW values of 19.0 to 25.4 mm were lower than the 28.7 mm observed from the 31 July 1200 KGJT sounding. Moderate instability was forecast for the afternoon, with projected LI values of -2C to -4C. This range was less than the LI of -6C, estimated by modifying the 31 July 1200 KGJT sounding for the actual high temperature at Ridgway. The weak shortwave evident in IR satellite imagery was reflected by the NGM model initialization in the 700 to 600 hPa layer. This feature was forecast by the NGM to traverse western Colorado between 31 July 1800 and 1 August 0000. It is important to note that the often-used method of examining the 500 hPa vorticity field failed to display this feature. A 700-300 hPa or 700-500 hPa layer analysis would likely have yielded better identification of the shortwave.
6. DALLAS CREEK BASIN CHARACTERISTICS
6.1 Topography and Vegetation
Dallas Creek drains 252 square kilometers of land, much of which is in a valley bordered by the San Juan Mountains to the south and Dakota sandstone mesas to the west and north. Basin elevations range from approximately 2073 m near the confluence of the Uncompahgre River to 4314 m at the peak of Mt. Sneffels. Vegetation in the higher terrain consists mainly of Ponderosa pine, spruce, pinyon and juniper. The lower portion of the basin rests upon a large formation of Mancos shale, a slick rock with a high clay content that has a tendency to swell when it gets wet. This results in little natural vegetative growth and the valley is used primarily for ranching and the agricultural production of hay.
Dallas Creek flows into the Uncompahgre River just upstream from Ridgway Reservoir, and contributes about 20% of the total reservoir inflow. The main stem of Dallas Creek is supported by five tributaries: Pleasant Valley Creek, Cottonwood Creek, West Fork Dallas Creek, East Fork Dallas Creek and Beaver Creek (Fig. 3).
The latter four tributaries drain the higher terrain of the northern San Juan mountains and merge to form Dallas Creek, while Pleasant Valley Creek lies along the lower elevations of the valley floor and joins Dallas Creek farther downstream.
Stream measurements are made on the Dallas Creek
main stem by a United States Geological Survey
(USGS) gage located downstream from Pleasant
Valley Creek and 2.4 km upstream from the mouth.
According to the 1998 USGS Water Resources Data
publication, the mean annual flow at this site is 40.2
cfs, with the highest monthly average (76.6 cfs)
occurring in July.
Fig. 3. Dallas Creek and its tributaries.
6.3 Flood History
The gaging of Dallas Creek began in 1922 and continued until 1927. The gage was re-established in 1956 and has operated to the time of this writing. USGS records indicate that flooding has been rare on this stream, with most (78%) annual peak flows cresting below 500 cfs. Until 1999, the flood of record had been a flow of 1120 cfs, occurring in August 1923. This was the only previously recorded crest above 1000 cfs.
Flash flooding and mudslides are common to this region, especially in the higher terrain south of the Dallas Creek basin. According to Chronic (1980), there are four major factors that contribute to the area's high slide potential:
1. Unstable volcanic rocks in the area often contain thick layers of poorly consolidated volcanic ash.
2. These volcanic rocks lie above even less stable Mancos shale.
3. Valley walls were over-steepened by glaciers.
4. This region is subject to thunderstorms and heavy snows which saturates the soil and underlying rock formations.
The capability for flash flood monitoring in this area is good. Radar coverage is adequate and the United States Bureau of Reclamation (USBR) has established an extensive rain gage network in this basin for the support of Ridgway Reservoir. Ten automated meteorological stations with attached DCPs (Data Collection Platforms) are located between the reservoir and the headwaters of Dallas Creek and the Uncompahgre River.
7. 31 JULY 1999 FLOOD EVENT
The 1999 Summer Monsoon season began in western Colorado during the second week of July. In the San Juan mountains, thunderstorm activity increased dramatically afterwards. Due to an increased flow of moist air from the eastern Pacific Ocean, precipitable water values were above normal for much of the month. Cooperative observing stations around the Dallas Creek area reported measurable precipitation nearly every day during the last half of the month. Total July rainfall was in excess of 200% above normal at several sites in southwest Colorado; Telluride, just south of the Dallas Creek basin, measured a monthly total of 159.8 mm (250% of normal), setting a July record. The period of heaviest July rainfall in the Dallas Creek basin occurred during the last week of the month. Thunderstorm activity with heavy rain was observed in or near the basin each day and numerous mudslides were reported in the adjacent San Miguel River basin to the southwest.
The heaviest rain in this area fell on 30-31 July, a period with precipitable water values in excess of 130% of normal. Placerville, located a few miles to the southwest of the Dallas Creek headwaters, measured 79.5 mm on 30 July alone. Two of the ten USBR DCP meteorological station rain gages in the Dallas Creek drainage also reported heavy rain on 30 July, with even heavier rain the following day. Pleasant Valley Meteorological Station, at an elevation of 2296 m in western Ouray County, is located in the drainage of Pleasant Valley Creek. The West Fork Dallas Creek Meteorological Station, located in southwest Ouray County at an elevation of 2823 m, is representative of rainfall that falls into both Cottonwood and West Fork Dallas Creeks.
The Pleasant Valley Meteorological Station received 95.8 mm of rain from thunderstorms on 30-31 July. Rainfall during this 48 hour period accounted for 72% of the July total. This site was located closest to the axis of heavy precipitation on 31 July, based on radar images and local reports. With nearby soils at or near saturation, over 25 mm of rain fell between 2000 and 2100, followed by an additional 25 mm the following hour. By 2200, rainfall rates had fallen to around 12.5 mm per hour, then ended shortly before 0000. Between 2000 and 2300, a total of 69.9 mm was recorded, an amount in excess of a 100 year rainfall for that location. The diminishing rainfall rates at Pleasant Valley was a result of a slight southern progression of the heavy rain band. The West Fork Dallas Creek Meteorological Station, approximately 11 km south of Pleasant Valley, recorded only light amounts of rain through 2200. Heavier rain totaling 33.3 mm had fallen at this gage on 30 July, and surrounding soils were also at or near saturation. After 31 July 2200 heavy rain again began to fall, with 30.7 mm recorded at the site by 2300. Lesser amounts of less than 6.5 mm fell between 2300 and 0000. The meteorological stations located near the easternmost tributaries, Beaver and East Fork Dallas Creeks, reported less than 6.5 mm during the afternoon of 31 July.
With rainfall rates of over 25 mm an hour for two hours in Pleasant Valley, combined with already saturated soils, Pleasant Valley Creek came out of its banks. Local reports and damage patterns to surrounding hayfields indicate that a large volume of overland flow swept across the valley and inundated the normally tiny stream. The post-flood debris field near the Route 24D bridge indicated that the creek, normally only a meter across, swelled to over 30 m wide. The bridge, about 3.7 m above the stream bed, sustained moderate damage when, according to eyewitnesses, it was overtopped by water levels that extended to nearly a meter above the bridge surface.
As the heavy rain shifted south, Cottonwood and West Fork Dallas Creeks also became inundated and came out of their banks. Their combined water was forced into the Dallas Creek main channel, which merges with Pleasant Valley Creek about 3 km downstream. Downstream from this confluence, the main stem Dallas Creek, normally 3 meters wide, expanded to 60 to 90 meters across. A concrete bridge on route 24A that traverses the creek about one mile downstream from the Pleasant Valley Creek confluence was undercut and damaged.
At the onset of this flood event, flows on Dallas Creek were averaging between 130 and 140 cfs, nearly double the monthly normal due to recent thunderstorm activity. By 2200 small rises on began to be recorded at the USGS gage. Telemetered data indicated that the flow had risen to around 250 cfs. One hour later, water volumes doubled to over 500 cfs. By 1 Aug 0000, Dallas Creek had increased to nearly 1900 cfs, exceeding the old 1923 record of 1120 cfs. Water levels continued to rapidly increase until 0045 when all data from the gage ceased. Estimates by the USGS through floodmarks indicated that Dallas Creek peaked around 3960 cfs, at a stage of 2.3 m (7.58 feet). This volume of water was 354% greater than the previous 1923 flood of record, and over 5100% of the July average flow (Fig. 4).
Fig. 4. Dallas Creek near Ridgway (DCKC2) hydrograph and basin rainfall.
Operational forecasters need to be astutely aware that conventional analysis of standard pressure surfaces may fail to detect significant meteorological features. Model forecasts of 500 hPa heights and vorticity completely overlooked the shortwave that enhanced the convective environment on 31 July 1999. Identification of these weak features becomes increasingly important in a moist environment, especially when PW values are very high. The utilization of cross-sections and time-height sections available to the forecaster is vital. Also, a thorough examination of satellite data was needed to detect an early morning thunderstorm outflow boundary that was the focus for initial convective development. It is imperative to carefully monitor the location of these boundaries and other mesoscale convergence zones in order to forecast the onset of convection (and possibly strong thunderstorms). The influence of the complex terrain surrounding the Dallas Creek Drainage cannot be underestimated. The Uncompahgre Plateau played a significant role in stalling the lingering thunderstorm outflow boundary. Additionally, with weak southerly flow aloft, the thunderstorms rolled continuously north off the San Juan mountains (located south of the drainage) where the thunderstorms kept regenerating over the higher terrain. Knowledge of the local topography can aid greatly in accurately forecasting movement and effects of thunderstorms, especially within small basins.
Post-storm analysis also suggests that the rainfall band axis orientation, basin orientation, area geology, and rainfall rates all played a critical role in this event. The mesoscale environment was one of new thunderstorm formation spreading downstream through the basin, in itself a classic flash flood scenario. Three of the five main tributaries of Dallas Creek each experienced similar rainfall amounts of 25 to 50 mm in a short period of time. While this amount of rain in any one of the streams probably would not result in the actual amount of damage, the volume of the combined water of three individual merging streams, however, did.
This event exemplified the need to monitor basins as a whole during potential flash flood scenarios. It is not enough to take into account rainfall over a main channel or waterway; rather, all affected tributaries and drainage channels must be regarded. In this case, it appeared that heavy rain over multiple waterways that combined into one played a large role in the overall flooding. Also, the Dallas Creek basin floor is composed of a rock surface well known for its association with flash flooding and debris movement during very wet episodes. In mountainous terrain, where flash flood guidance is rarely available, it behooves the flash flood forecaster to be aware of the local geology and its particular flash flood characteristics.
As Maddox (1980) points out, flash flood events in the western United States usually occur with considerably less rain amounts than occur in the eastern portion of the country. Due to the sparsity of flash flood guidance in mountainous terrain, western forecasters often use the "one inch or more of rain per hour" rule-of-thumb to quantify flash flood potential. However, this rate of rainfall does not always result in flash flooding. Factors such as geology, slope, vegetation, antecedent conditions and affected basin drainages need to be considered in addition to rainfall rate and a general knowledge of meteorological characteristics that produce localized flash flooding.
The authors would like to thank Liz Page of COMET/NWS for providing the archived data, and Jeff Evans of the Storm Prediction Center for the Grand Junction upper-air sounding data and image.
Bove, M.C., J.J. O'Brien, J.B. Elsner, C.W. Landsea, and X. Niu, 1998: Effect of El Niño on U.S. landfalling hurricanes, revisited. Bull. Amer. Meteor. Soc., 79, 2477-2496.
Chappell, C.F., 1986: Quasi-stationary convective events. Mesoscale Meteorology and Forecasting, P.S. Ray, Ed., Amer. Meteor. Soc., 289-310.
Chronic, H., 1980: Roadside Geology of Colorado. Mountain Press Publishing Co., 334 pp.
Corrigan, P., and J.L. Vogel, 1993: Meteorological analysis of a local flash flood near Opal, Wyoming 16 August 1990. 13th Conference on Weather and Forecasting, Vienna, VA, Amer. Meteor. Soc., 375-377.
Hales, J.E. Jr., 1974: Southwest United States summer monsoon source - Gulf of Mexico or Pacific Ocean? J. Appl. Meteor., 13, 331-342.
Hansen, E.M., 1975: Moisture source for three extreme local rainfalls in the southern Intermountain region. NOAA Technical Memorandum, NWS Hydro-26, 57 pp.
Hydrometeorological Prediction Center, 2000: Personal Communication.
Knievel, J.C., and R.H. Johnson, 1998: The 28 July 1997 Fort Collins flood: synoptic and mesoscale analysis. Eighth Conference on Mountain Meteorology, Flagstaff, AZ, Amer. Meteor. Soc., 396-399.
Maddox, R.A., L.R. Hoxit, and F. Canova, 1980: Meteorological characteristics of heavy precipitation and flash flood events over the western United States. NOAA Technical Memorandum, ERL APCL-23, 87 pp.
Reitan, C.H., 1957: The role of precipitable water vapor transport and the water balance in Arizona's summer rains. Technical Report on the Meteorology and Climatology of Arid Regions. No. 2, Inst. Atmos. Phys., The University of Arizona, 18 pp.
Runk, K.J., and D.P. Kosier, 1998: Post-analysis of the 10 August 1997 southern Nevada flash flood event. Nat. Wea. Dig., 22, 10-24.
Shi, J., and R.A. Scofield, 1987: Satellite observed Mesoscale Convective System (MCS) propagation characteristics and a 3-12 hour heavy precipitation forecast index. NOAA Technical Memorandum NESDIS: 20, 43 pp.
Spayd, Leroy E. Jr., and R.A. Scofield, 1983: Operationally detecting flash flood producing thunderstorms which have subtle heavy rainfall signatures in GOES imagery. Fifth Conference on Hydrometeorology, Tulsa, OK, Amer. Meteor. Soc., 190-197.
Tang, M., and E.R. Reiter, 1984: Plateau monsoons of the northern hemisphere: A comparison between North America and Tibet. Mon. Wea. Rev., 112, 617.
Weaver, John F., and N.J. Doesken, 1990: Recurrence probability - a different approach. Weather, 45, 333-339.