This study sought to determine the peak discharge for a flood that took place in the headwaters of the Prairie Dog Town Fork of the Red River, in Randall County, in the vicinity of Canyon, Texas and Palo Duro Canyon State Park in May of 1978. Although the height of the floodwaters was recorded at several points along the reach between the City of Canyon and Palo Duro Canyon State Park, no estimates of the magnitude of the peak discharge, cubic meters per second (cms) or cubic feet per second (cfs), have been made. This study provides the first estimate of the peak discharge for this historic flood.
On the afternoon of May 26, 1978, a line of storms formed in the western Texas Panhandle, and two of these storms converged west of the town of Canyon near Buffalo Lake in southwestern Randall County (Burnett 2008). The heaviest rain occurred between the towns of Hereford and Canyon. Official rainfall totals for weather stations in the region measured at Canyon, Hereford and Umbarger are given in Table 1, and locations of the towns are shown in Figure 1. However, unofficial rainfall totals suggest that as much as 225–250 mm (9–10 in) may have fallen in parts of the watershed (Keck 1978; Burnett 2008).
A significant portion of the runoff on Tierra Blanca Creek was captured by the dry Buffalo Lake, which filled with 3.7 million cubic meters (3000 acre-ft) of water. In the early morning hours of May 27, the floodwaters reached the town of Canyon and the flood peak reached the dam at Lake Tanglewood, approximately 16 km (10 mi) downstream from Canyon at about 5 AM, where it flowed 2m (6 ft) above the spillway (Burnett 2008). Floodwaters began affecting campgrounds in Palo Duro Canyon State Park about the same time.
Methods and Results.–The Prairie Dog Town Fork of the Red River begins at the juncture of Palo Duro Creek and Tierra Blanca Creek on the northeast edge of the town of Canyon (Fig. 1). The channel enters a series of canyons 3 km from the juncture, collectively called Palo Duro Canyon, carved into the Tertiary Ogallala Formation, and underlying Triassic and Permian sedimentary rocks (McGowen et al. 1979).
The starting point for the peak discharge calculation is a historical marker at the first water crossing in Palo Duro Canyon State Park that marks the height of the water during the peak of the flood. The historical marker states that the flood peak occurred when the water level reached 7.3 m (24 ft) above the channel and 6.4 m (21 ft) above the flood plain. At the first water crossing, the cross-section of the channel and floodplain is asymmetrical. A broad floodplain extends to the west - this floodplain is occupied, in part, by campsites that are part of the Hackberry Campground; to the east of the channel there is a steep bank. At the time that the 1978 flood occurred, the water crossings were low-water crossings. The pavement and concrete of the roadway acted as temporary base level and prevented incision of the channel below the level of the water crossings. The bottom of the channel lay approximately 1 m (3 ft) below the level of the flood plain. The flood plain in the vicinity of the first water crossing today is essentially at the same elevation as in 1978. Because it is occupied by the Hackberry Campground, many of the picnic tables and trees have remained in their locations since 1978. However, in 2015 the low-water crossing was removed and replaced with a bridge. Since that time the stream has incised a narrow channel into the alluvial sands and gravels of the flood plain. The challenge for this study was to approximate the channel cross-section and gradient at the time of the 1978 flood.
The basis for the determination of peak discharge is the slope-area methodology used by the United States Geological Survey to determine peak discharge values from high water marks (Benson & Dalrymple 1967). This method is based upon the Manning Equation, Q = (AR2/3 S1/2))/n, where Q = peak discharge (cms); A = cross-sectional area (m2); R = hydraulic radius, which is A/wetted perimeter; S = slope of the energy surface; and n = Manning's roughness coefficient.
Wetted perimeter is the perimeter of a river channel cross-section which is covered by water at a specific stage of flow. The cross-sectional profile at the first water crossing was surveyed using a hand level, 4-m rod and 60-m measuring tape. Measurement of the profile provided measurements used to calculate A, wetted perimeter and R. The value of S was approximated by using the gradient of the channel measured using a hand level on a tripod, a 4-m rod and a 150-m steel tape. The location of the line of section is shown in Figure 2. Initial results of the survey are shown in Figure 3. The advantage of the slope-area method over the basic Manning Equation is that it allows for subdivision of the channel profile into separate vertical segments. Each segment is then assigned its own appropriate roughness coefficient. During the initial survey of the channel profile the section was divided into five segments and each segment was assigned a roughness coefficient by comparison with similar flood plains in Arcement et al. (1998).
The initial cross-section shown in Figure 3 has a total area of 751 m2 with a total width of 181 m (of which 57 m occur in segment 1 are judged to be non-contributing, i.e. the water in that segment was probably not moving downstream). This part of the section is on the inside of a bend and the abundance of trees in this segment would also have inhibited water movement downstream. The active part of the cross-section is approximately 124 m wide and contains approximately 652 m2. The active portion of the cross-section is divided into four segments and each segment is assigned a roughness coefficient, n. The water line is determined by what we know about the height of the flood at its peak (6.4 m above the flood plain). Because the bottom of the channel today is approximately 2.4–3.4 m below the floodplain (see Figs. 3 & 4), the total height of the water above the bottom of the channel is 9.8 m. Historical photographs of this water crossing on Google Earth Street View in 2019 show that the channel in 2013 prior to the construction of the bridge (and probably in 1978) was only about one m below the elevation of the flood plain. The elevation of the bottom of the channel relative to its floodplain is also suggested by the historical marker which lists a difference of 0.9 m between the elevation of the channel and the floodplain. The corrected cross section, shown in Figure 4, represents what I hypothesize to be the most likely channel cross-section for the 1978 flood. This corrected cross-section has an active cross-sectional area of 646 m2, and maintains the high water mark at 6.4 m above the floodplain. All of the above parameters (A, R2/3, and n) go into the Manning equation to calculate the conveyance K, as calculated in Table 2. Conveyance is defined as the carrying capacity of the channel or a segment of the channel. The hydrologic properties of the corrected cross-section are tabulated in Table 2 and results in a total conveyance, K, of 33,780.
The slope of the energy surface – S, the slope of the water surface at peak discharge – was approximated from the channel slope. The slope of the channel was calculated using a hand level on a tripod and stadia rod (see Fig. 5). This measurement over a distance of 122.5 m yielded an elevation change of 0.396 m, which yields an energy slope, S, of 0.00322. Calculation of the slope of the channel using the 7.5 minute Fortress Cliff Quadrangle (20 ft contour interval) gave a slope of 20/5600 = 0.00357. The two calculated values of S yield a weighted average of 0.00339.
Using the values of A, R, and n from the corrected profile as shown in Figure 4, the peak discharge can be calculated using the values of S derived from direct measurement of the channel, as well as from the topographic map. The calculation using the weighted value of S gives a peak discharge (Q) of about 1970 cms (70,000 cfs) (Table 3).
Sources of error in the calculation of discharge.–This study followed a standard technique used by federal water resource agencies to estimate discharge, but several possible sources of error in the calculation reduce the precision of the estimate of discharge. These sources of error, include:
Possible changes in the channel cross-section. Although the ends of the cross-section rest on bedrock of Permian age, most of the channel today is cut into alluvial sediments that could be readily reshaped by major floods. However, the elevation and cross-section of the flood plain has remained approximately the same.
Changes in the roughness coefficient due to changes in the floodplain vegetation. Although the density of large trees in the floodplain have probably remained about the same, the density of shrubs may be different today than in 1978. However, the shrubby vegetation probably has a lesser impact on roughness than the trunks and branches of mature trees.
Changes in the slope of the channel. As has been stated previously, the channel has incised into the alluvial sediments since the low water crossings that maintained the channel at a constant elevation were removed. However, the measurement of the channel today and calculation of the channel gradient using the USGS 1:24,000 Fortress Cliff Topographic Quadrangle yield similar results.
Conclusions.–The elevation of the flood plain was used as the most reliable benchmark for this calculation. However, the depth of the channel below the floodplain was modified to a value consistent with decades-old photographs of the area. The United States Geological Survey gaging station (#07297910) at Hwy 207, approximately 32 km (20 miles) below the First Water Crossing in the State Park, recorded a peak discharge on May 27 of 1300 cms (46,400 cfs). Location of this gaging station is shown on Figure 1. Although some additional runoff occurred from rainfall downstream of the State Park, the observed peak at the gaging station is believed to be almost entirely due to the runoff that entered the channel upstream from the State Park. This would represent a transmission loss of about 30% and is consistent with observed losses of stream channels in arid regions (Lange 2005). The transmission losses are interpreted to be the result of infiltration of water into the subsurface in overbank areas.
The Killgore Research Center at WTAMU provided research space for this project. The initial survey of the cross-section at the first water crossing was done with the assistance of members of the Applied Hydrogeology class of the Spring of 2018. Finally, I wish to thank the reviewers for their constructive comments on this manuscript.