Abstract

Heavy rainfall and flooding associated with Tropical Storm Hermine occurred on 7–8 September 2010 across central Texas, resulting in several flood-related fatalities and extensive property damage. The largest rainfall totals were received near Austin, Texas, and immediately north, with 24-h accumulations at several locations reaching a 500-yr recurrence interval. Among the most heavily impacted drainage basins was the Bull Creek watershed (58 km2) in Austin, where peak flows exceeded 500 m3 s−1. Storm cells were trained over the small watershed for approximately 6 h because of the combination of a quasi-stationary synoptic feature slowing the storm, orographic enhancement from the Balcones Escarpment, and moist air masses from the Gulf of Mexico sustaining the storm. Weather Research and Forecasting Model simulations with and without the Balcones Escarpment terrain indicate that orographic enhancement affected rainfall. The basin received nearly 300 mm of precipitation, with maximum sustained intensities of 50 mm h−1. The Gridded Surface Subsurface Hydrologic Analysis (GSSHA) model was used to simulate streamflow from the event and to analyze the flood hydrology. Model simulations indicate that the spatial organization of the storm during intense rainfall periods coupled with surface conditions and characteristics mediate stream response.

1. Introduction

In central Texas, the unique combination of flood-prone physiography and susceptibility to extreme meteorological events has produced some of the largest rainfall events and flood magnitudes in the United States and the world (Smith et al. 2000; Baker 1975). Moreover, Texas consistently leads the nation in flash flood–related deaths, the majority of which occur across central Texas in an area dubbed Flash Flood Alley (FFA; Zahran et al. 2008; Sharif et al. 2012). FFA is oriented from north to south across the central portion of the state, extending approximately from San Antonio to Dallas, Texas (TX). FFA holds several world-record rainfall rates on time scales less than 24 h (Slade and Patton 2003). Notable events include the 1921 storm near Thrall (965 mm in 24 h), the 1935 storm near D’Hanis (560 mm in 2 h 45 min), and the 1978 event near Bluff (790 mm in 24 h; Caran and Baker 1986; Caracena and Fritsch 1983). Numerous events over the past century in FFA have exceeded 24-h accumulations of 750 mm (Asquith and Slade 1995).

While FFA is prone to short-lived intense convective outbreaks in the absence of tropical disturbances, many of the largest events are of tropical origin. These storms are characterized by easterly waves of air masses containing enormous moisture amounts from passage over warm seas (Nielsen-Gammon et al. 2005). If the moist air masses are able to protrude inland far enough, they are met by the Balcones Escarpment, which is thought to serve as a barrier (Caran and Baker 1986). Orographic barriers act to force an air mass upward. If the upward lift forces the air mass to cool down enough, water vapor condenses to produce clouds and precipitation. Severe flash flood events have been known to occur in areas of orographically enhanced rainfall (Lin et al. 2001).

Once precipitation falls to the surface, the physiographic features of the landscape control hydrologic response. The Balcones Escarpment separates the limestone terrain of the Edwards Plateau from the gently sloping clay and sand terrain of the Blackland Prairies and Coastal Plain. Landscape features augmenting stream response along the Edwards Plateau and Balcones Escarpment include shallow stony soils underlain by bedrock, steep terrain, high rill densities on hillslopes, and sparse scrubby vegetation. At the bottom of the escarpment, clay soils permit little infiltration when saturated (Patton and Baker 1976). In addition to environmental factors, the region has experienced sizeable urbanization over the last several decades, which can exacerbate the flood potential through the increase of the impervious areas.

On 7–8 September 2010, persistent, heavy rainbands associated with Tropical Storm Hermine produced one of the heaviest rain events in nearly 30 years for Travis and Williamson Counties (central Texas). While the tropical storm–associated rainfall showed similarities to other extreme events in the region, it was different from most floods associated with dissipating tropical cyclones in that heavy rains developed well away from the remnant center. Predecessor rain events developing on the poleward side of recurving tropical cyclones have received recent attention (e.g., Schumacher 2011; Galarneau et al. 2010); however, rainfall from Hermine developed well behind the center. The 24-h rainfall totals of 250–400 mm resulted in a 500-yr event at several locations along the Balcones Escarpment. Flash flooding occurred across numerous area watersheds, with unit runoff values nearing the United States envelope curve at the North Fork San Gabriel River (52 m3 s−1 km−2) north of Austin, TX (Winters 2012). Among the most heavily impacted watersheds included the semiurbanized Bull Creek catchment (58 km2) in Austin. During the 2-day period, the 58 km2 catchment received over 290 mm of rain, sustaining large amounts of flood-related damage and a fatality.

In the present work, we examine the primary hydrometeorological controls leading to heavy rainfall across central Texas and flooding along Bull Creek. The manuscript is divided into three sections analyzing meteorological aspects of the storm, rainfall across the Bull Creek basin, and flood hydrology of the catchment. Specifically, we seek to characterize the structure, motion, and evolution of the storm along with determining the effect of terrain-aided forcing from the Balcones Escarpment. Gauge-adjusted radar is used to describe spatiotemporal rainfall patterns across the catchment and calculate recurrence intervals. Physics-based distributed hydrologic modeling is conducted to examine the important hydrologic mechanisms of the flood event. Models used include the Weather Research and Forecasting (WRF) Model (Skamarock et al. 2008) for storm simulation along with the Gridded Surface Subsurface Hydrologic Analysis (GSSHA) model (Downer and Ogden 2006) for hydrologic analysis.

2. Storm description

Features of the storm that produced flooding across the Bull Creek watershed are examined through analysis of synoptic- and mesoscale features leading to and evolving during the event. Hermine began as a depression south of Mexico’s southern Pacific coast. The depression moved northward over Mexico, becoming a remnant low as it crossed land. Once in the Gulf of Mexico, deep convection formed and thunderstorm activity became organized with the formation of cyclonic bands. As the storm moved away from land, it gained further organization and became a tropical storm at approximately 0600 UTC 6 September. Deep convection continued to develop near the storm’s center throughout the day as it trekked across the Gulf of Mexico at an average speed of 12 knots (kt; 1 kt = 0.51 m s−1; Avila 2010).

Hermine made landfall at around 0200 UTC 7 September, 40 km south of Brownsville, TX, along the northeastern coast of Mexico. A minimum pressure of 989 mb was sustained here with peak winds of 110 km h−1. Hermine tracked to the north-northwest across Texas, remaining a tropical storm for 16 h after making landfall. At around 0000 UTC 8 September, the storm was downgraded to a tropical depression. The system continued north and northeast through Oklahoma before becoming a remnant low and dissipating over Kansas (Avila 2010). Figure 1a displays the storm’s progression and track (NOAA 2010).

Fig. 1.

(a) Tropical Storm Hermine’s track through Texas. Times are given as four-digit hour (UTC) and day (in September 2010) for every other point. (b) Surface map displaying pressure and frontal locations (at 1200 UTC 7 Sep). (c) Precipitable water values (mm) from SuomiNet (at 1200 UTC 7 Sep). (d) Surface wind vectors and mixing ratios (g kg−1) map (0200 UTC 8 Sep). An outline of the Edwards Plateau ecoregion is shown.

Fig. 1.

(a) Tropical Storm Hermine’s track through Texas. Times are given as four-digit hour (UTC) and day (in September 2010) for every other point. (b) Surface map displaying pressure and frontal locations (at 1200 UTC 7 Sep). (c) Precipitable water values (mm) from SuomiNet (at 1200 UTC 7 Sep). (d) Surface wind vectors and mixing ratios (g kg−1) map (0200 UTC 8 Sep). An outline of the Edwards Plateau ecoregion is shown.

At 1200 UTC 7 September, Hermine was located in southern Texas with a stalled front positioned in the panhandle region of Texas (Fig. 1b). The 500-mb height field (not shown) indicates an area of high pressure behind the stalled front with a large gradient in atmospheric moisture present along its edge. Precipitable water values greater than 50 mm (central Texas September mean ≈ 36.6 mm, standard deviation = 12.0 mm) covered all of southern and central Texas with values at approximately half in the western portion of the state (Fig. 1c; Ware et al. 2000). The 1200 UTC 7 September radiosonde from Brownsville (data not shown) indicated the atmosphere was nearly saturated through the troposphere (precipitable water = 67.5 mm). CAPE values from this recording were in excess of 2000 J kg−1 indicating moderate instability. Precipitable water values measured from the Midland radiosonde, located near the edge of the front, were about one-third of those at Brownsville.

As Hermine moved farther into Texas, the precipitation shield began to change its structure and no longer appeared as the classic banding seen in tropical cyclones. At around 0000 UTC 8 September, as the remnant low center was nearing the Oklahoma border, areas of deep convection began developing farther south along a linear boundary, showing that Hermine had undergone extratropical transition. Advanced Research version of WRF (ARW) simulations (discussed below) show this linear boundary well in the surface wind field (not shown). Southeasterly winds advecting moisture-rich air from the Gulf were met by drier air wrapping around the western side of the storm, creating a strong boundary (Fig. 1d). The area of convergence roughly followed the Balcones Escarpment and resulted in the development of a mesoscale convective system (MCS). The MCS assumed a squall-line structure along the linear synoptic-scale boundary training over the area for approximately 6 h.

Figure 2 shows a series of infrared images displaying rapid convective growth along this boundary from 0015 to 0815 UTC 8 September. The development of deep convection well south of the remnant center was a unique feature of Tropical Storm Hermine. In general, a broadening of the rainfall region and decrease in mean rainfall rates can be expected after landfall. This pattern though, is sensitive to the presence of vertical wind shear, topography, and other factors that can result in heavy localized accumulations (Kimball 2008). Indeed, the well-documented reintensification of Tropical Cyclone Erin over Oklahoma and resultant flooding provide an example of the complicated land surface feedbacks that can lead to focused rainfall after landfall (Arndt et al. 2009). In the case of Tropical Cyclone Erin, abnormally moist soil conditions were hypothesized to provide the necessary environmental conditions for redevelopment over land (Evans et al. 2011). While increased soil moisture left by Hermine rain across the Texas Coastal Plain may have influenced the lower-tropospheric inflow region to the MCS, the location and pattern of the outbreak suggests the Balcones Escarpment provided an orographic effect.

Fig. 2.

(a)–(f) GOES infrared imagery with temperatures (°C) across Texas at 0015, 0215, 0400, 0500, 0630, and 0815 UTC 8 Sep 2010, respectively.

Fig. 2.

(a)–(f) GOES infrared imagery with temperatures (°C) across Texas at 0015, 0215, 0400, 0500, 0630, and 0815 UTC 8 Sep 2010, respectively.

The synoptic- and mesoscale setup for the storm shows similarities to other heavy orographic rainfall events in the United States. These events are characterized by unstable airstreams impinging on the barrier, a quasi-stationary synoptic system slowing the convective system, and very moist air masses sustaining the storm (Lin et al. 2001). To examine the role of terrain, WRF Model (Skamarock et al. 2008) simulations were conducted with and without inclusion of the Balcones Escarpment (hereafter referred to as terrain-included and terrain-removed simulations). In the terrain-removed simulations, topography over 100 m was smoothed to 100 m, removing effects from the escarpment.

Simulations were conducted on three nested domains with grid spacing of 27, 9, and 3 km (Fig. 3a). USGS 10-min terrain data were used on the largest domain, with 30-s terrain data used in the two smaller areas. Each of the domains contained 50 vertical levels. The European Centre for Medium-Range Weather Forecasts (ECMWF) interim reanalysis data were used for the initial and lateral boundary conditions (updated in 6-h intervals; Dee et al. 2011). Model simulations were conducted for 60 h, from 1200 UTC 6 September to 0000 UTC 9 September 2010. The analysis period captures the approximate 24-h maxima along the Balcones Escarpment (ending 1200 UTC 8 September 2010). Table 1 provides additional details on model setup.

Fig. 3.

Observed and WRF-simulated rainfall accumulations for the three domains. (a),(d),(g) Observed values (NWS RFC stage IV) for each of the domain areas. (b),(e),(h) Control runs. (c),(f),(i) Terrain-modified simulations. The spatial extent of the middle and inner domains is shown in (a); the Bull Creek watershed location is boxed in (g).

Fig. 3.

Observed and WRF-simulated rainfall accumulations for the three domains. (a),(d),(g) Observed values (NWS RFC stage IV) for each of the domain areas. (b),(e),(h) Control runs. (c),(f),(i) Terrain-modified simulations. The spatial extent of the middle and inner domains is shown in (a); the Bull Creek watershed location is boxed in (g).

Table 1.

WRF Model setup.

WRF Model setup.
WRF Model setup.

Figure 3 shows WRF simulations for each domain along with National Weather Service (NWS) River Forecast Center (RFC) precipitation estimates (multisensor stage IV). The observed rainfall isohyets aligned roughly parallel to the escarpment encompassing FFA. The heaviest accumulation was present as a comma-like feature from San Antonio to north of Austin (Fig. 3g). The rainfall event produced totals in excess of 250 mm over approximately 900 km2 from Austin and to the north. A large accumulation gradient was present normal to the line of heavy rainfall. At 75 km to the east and west of the heaviest rainfall, 24-h totals were generally less than 65 mm. The highest recorded 24-h rain gauge total (370 mm) occurred 30 km north of Bull Creek near Georgetown, TX. Interestingly, the 1921 Thrall, TX, storm (965 mm in 24 h), which held the U.S. record for over 50 years (Smith et al. 2000), occurred 30 km east of the maximum accumulation from Hermine.

The terrain-included simulations match the overall observed precipitation pattern well, with a large area of increased accumulations from north to south through the central portion of the state. The terrain-included run also captures increased values across southern Oklahoma and Arkansas, with minimal precipitation through the Texas Panhandle. Maximum rainfall from the terrain-included simulations was approximately 130 km north of observed maxima. The peak value from the 3-km domain was 323 mm, which agreed very well with the maximum stage IV bin of 315 mm. Considering the relative coarseness of the simulations, WRF terrain-included runs performed well in simulating broad precipitation patterns as well as localized areas of enhanced precipitation along the Balcones Escarpment. However, it should be noted that the terrain-included WRF simulations placed the areas of heaviest rainfall in different watersheds than were observed.

Terrain-removed simulations showed a marked difference when compared to the terrain-included runs. The simulations were not as effective in capturing the general precipitation patterns as terrain-included runs, but they did place maximum accumulation values near observed maxima. Minimal precipitation across the Texas Panhandle was correctly modeled, while elevated values in Oklahoma and Arkansas were not. In addition, while the linear boundary seen in the observed radar reflectivity fields (section 3) was modeled correctly in the terrain-included runs, the same feature was not reproduced in the terrain-removed simulations. The primary difference with the terrain-removed simulations was the lack of a narrow, elongated area of elevated accumulations along the Balcones Escarpment. This resulted in much lower rainfall totals across the inner domain for the terrain-removed simulations. Average precipitation across the entire inner domain (~200 000 km2) was 68, 58, and 23 mm for observed values, terrain-included runs, and terrain-removed runs, respectively. The difference in magnitude and location of precipitation between WRF simulations suggests terrain may have had an effect in the high rainfall totals across central Texas. Analyses of multiple storms and additional model output would provide a clearer understanding of the escarpment’s effect on precipitation.

3. Rainfall analysis

a. Bull Creek precipitation dataset

Bull Creek is a 58-km2 semiurbanized catchment on the northern side of Austin. Rainfall estimates across the catchment, as well as model forcing, relied on gauge-adjusted radar (GAR) data provided by the City of Austin [from 0500 UTC 7 September to 0500 UTC 9 September (from 0000 CDT 7 September to 0000 CDT 9 September)]. The rainfall data were used by the city as part of an operational flash flood forecasting system. Spatial resolution of radar bins across the study area was approximately 1 km, with accumulation values recorded in 15-min increments. The NWS S-band Weather Surveillance Radar-1988 Doppler (WSR-88D) 80 km south of Bull Creek (station KEWX) provided the reflectivity measurements. The radar beam center height (0.5° tilt) is around 1.11 km (beamwidth = 1.30 km) over the study area.

Figure 4 displays a series of KEWX base reflectivity measurements (0.5° elevation angle) corresponding to GOES infrared imagery times. The radar reflectivity images provide a spatiotemporal view of storm motion, which serves a central role in extreme accumulations (Doswell et al. 1996). The reflectivity images show the development and repeated training of storm cells along a linear boundary parallel to the escarpment intersecting Bull Creek. The well-defined, banded structure, along with its persistence, was an important characteristic of the event. A narrow band of reflectivity values greater than 50 dBZ remained over the watershed for approximately 4 h.

Fig. 4.

Base reflectivity images (dBZ; at 0.5° elevation angle) from NWS KEWX. Times correspond to GOES infrared images in Fig. 2. Outline of Bull Creek watershed is shown.

Fig. 4.

Base reflectivity images (dBZ; at 0.5° elevation angle) from NWS KEWX. Times correspond to GOES infrared images in Fig. 2. Outline of Bull Creek watershed is shown.

A heterogeneous network of telemetered rain gauges in and around the City of Austin provided point measurements for radar bias correction. Several rain gauges used in the bias correction were located within the Bull Creek watershed. Radar correction is based on a local bias approach applying a spatially variable ratio of gauge to radar accumulations. Variation of bias is distributed over the area on a 6-h window updated in 15-min increments. Mean bias corrections across the entire telemetered network were generally near 1.5 and below during the 24-h period of maximum rainfall. During the 6-h period of maximum rainfall over Bull Creek, the mean bias correction was 1.53. Figure 5 shows rain gauge values (used in the bias correction) from the Lower Colorado River Authority network along with GAR values for the bin holding the gauge. Accumulations at these gauges were within 5%. Additional detail concerning the construction of the precipitation dataset, bias correction methods, and performance during Tropical Storm Hermine across the City of Austin is provided by Looper and Vieux (2012).

Fig. 5.

Bull Creek precipitation accumulation maps for (a) event totals, (b) 6-h max, and (c) 3-h max. (d) A basin-averaged hyetograph is shown along with rain gauge–radar bin (containing rain gauge) comparisons. The location of the two rain gauges shown (d) is indicated in (c).

Fig. 5.

Bull Creek precipitation accumulation maps for (a) event totals, (b) 6-h max, and (c) 3-h max. (d) A basin-averaged hyetograph is shown along with rain gauge–radar bin (containing rain gauge) comparisons. The location of the two rain gauges shown (d) is indicated in (c).

b. Rainfall accumulation

Rainfall totals across the Bull Creek watershed ranged from 230 to 320 mm, with a basinwide average of 292 mm (Fig. 5a). The maximum 24-h basin total on a sliding scale was 275 mm (ending at 1000 UTC 8 September). Maximum 1-, 3-, and 6-h basin totals were 58, 124, and 162 mm, respectively (Figs. 3b,c). A basinwide-average hyetograph is shown in Fig. 5d. Rainfall isohyets for storm totals were oriented in a north–south manner running parallel to the general direction of streamflow. The largest accumulations occurred across the western central portion of the basin (>300 mm) with rainfall decreasing from west to east. Rainfall totals across the eastern half of the watershed ranged from 225 to 300 mm.

Winters (2012) calculated 24-h recurrence intervals for the storm at selected rain gauges across central Texas. Several rain gauges across the region experienced 24-h recurrence intervals of 500 years or greater. In the present analysis, we calculate recurrence intervals for eight durations of 1 day and less following the methods of Asquith (1998). The method employs a regionalization approach for the state of Texas utilizing a generalized logistic (GLO) distribution. The storm’s point annual nonexceedance probability F is estimated by

 
formula

ξ, α, and κ are parameters of the GLO distribution estimated from L moments. Variable ξ describes the location along the GLO distribution and can be interpreted as a median depth for a given duration. Variables α and κ describe the scale and shape of the GLO distribution, respectively, and variable Xd refers to the precipitation depth for a given duration. Variable κ is dimensionless while ξ, α, and Xd have units of inches. Asquith (1998) contains isomaps of the state displaying κ, ξ, and α values, which vary by duration and location.

Areal reduction factors (ARFs) are often used as a means to reduce the amount of precipitation from a design storm for a point to an effective depth over an entire watershed. Using a network of 108 Austin area rain gauges (~250 000 daily precipitation values), Asquith (1999) calculated a 1-day areal reduction factor of 0.80 for a 58 km2 basin. Since catchment-wide depth values are available from radar estimates, the ARF inverse was used to transform catchment-wide values to a point estimate (hereafter ARF−1 method) for nonexceedance probability calculations. For example, the 24-h point estimate is calculated by multiplying the maximum 24-h basinwide accumulation (275 mm) by the ARF−1 (275 mm × 0.80−1 = 344 mm). It should be noted the ARF produced by Asquith (1999) is for a 1-day duration event.

In addition to the ARF−1 method, point estimates were also collected from individual radar bin values. For radar bin calculations, accumulations were collected from the bin containing the largest depth for a given duration (i.e., bin location was allowed to vary). Table 2 shows recurrence intervals for a point T [where F = 1 − (1/T)] from ARF−1 and radar bin methods.

Table 2.

Recurrence intervals for durations of 1 day and less for individual radar bins and ARF−1 point estimates.

Recurrence intervals for durations of 1 day and less for individual radar bins and ARF−1 point estimates.
Recurrence intervals for durations of 1 day and less for individual radar bins and ARF−1 point estimates.

The 24-h recurrence interval for a radar bin and catchment to point value was 440 and 692 years, respectively. Recurrence intervals increased monotonically with duration for both methods. At durations less than 3 h, recurrence intervals were less than 100 years, indicating the importance of storm persistence. Comparison of the two recurrence intervals also suggests the large areal coverage of the storm was a unique factor. The radar bin recurrence intervals provide the most direct means of spatially distributed point estimates, and their maximum recurrence interval neared 500 years, which is consistent with the analysis of rain gauge data by Winters (2012). Recurrence intervals for the ARF−1 method were much greater than recorded by radar bin or rain gauge values. This suggests the storm had a broader spatial coverage than the data used by Asquith (1999) to arrive at the 0.80 ARF for a basin this size. In a critical examination of ARFs, Wright et al. (2014a) indicate ARFs are not representative of extreme rainfall in part because of their formulations mixing different storm types. The authors show that rainfall events in North Carolina produced by tropical storms tend to be larger with longer durations than those from organized thunderstorm systems. For Hermine, the ARF suggested by Asquith (1999) would need to be increased by 10% (0.88) to produce a 24-h recurrence interval matching the maximum from individual radar bin estimates (440 years). It is worth noting that the 1-day envelope curve for a basin of this size within this region is almost 3 times the amount experienced (~750 mm; Asquith and Slade 1995).

c. Rainfall intensity

High-intensity rainfall rates characteristic of the Edwards Plateau and Balcones Escarpment are a major factor in the region’s chronic flooding problem (Caran and Baker 1986). While short-duration rates were elevated in the present storm (0.25, 0.5, 1, and 2 h), they were not particularly rare from a recurrence interval standpoint (<70 years). Figure 6 shows average basinwide rainfall rates (mm h−1) for the event, percent of precipitation delivered by rainfall rate, and cumulative frequency plots of maximum intensities for each radar bin. Hourly rainfall rates are shown for both 15-min and 1-h intervals. The 1-h intensity intervals are calculated by summing 15-min intervals at the beginning of each hour (00). About 60% of the storm was delivered at basinwide rates exceeding 20 mm h−1. The 15-min intensity rates indicate 35% of the precipitation totals fell at basinwide rates greater than 40 mm h−1. Smith et al. (2000) recorded slightly larger percentages (40%–60%) delivered at 50 mm h−1 in larger flood events along the Texas Coastal Plain and Edwards Plateau. Maximum rainfall rates at individual radar bins varied by up to threefold across Bull Creek, with half of the bins exceeding 80 mm h−1 and one-quarter exceeding 100 mm h−1 (15-min measurements).

Fig. 6.

(a) Event rainfall intensity. (b) Percent of total basinwide accumulation by rainfall rate. (c) Cumulative frequency of max intensity for each individual radar bin. Values are shown using a 15-min and 1-h averaging period.

Fig. 6.

(a) Event rainfall intensity. (b) Percent of total basinwide accumulation by rainfall rate. (c) Cumulative frequency of max intensity for each individual radar bin. Values are shown using a 15-min and 1-h averaging period.

Roughly 60% (175 mm) of the total rainfall from the 2-day event fell during two intense periods lasting a combined 5 h. The first period of heavy rainfall occurred at 2200–2300 UTC 7 September when basinwide intensity values reached approximately 35 mm h−1. Rainfall rates at individual radar bins peaked near 80 mm h−1 over this time period. The second period of intense rainfall was much longer (0230–0630 UTC 8 September) and resulted in a large flood wave (discussed below). From 0230 to 0630 UTC precipitation rates averaged over 30 mm h−1, with mean values in excess of 50 mm h−1 persisting for an entire hour (0500–0600 UTC). Basinwide rates greater than 60 mm h−1 were sustained for half an hour (0500–0530 UTC), and peak rates neared 175 mm h−1 during this time period.

4. Flood hydrology

a. Bull Creek environmental setting

The Bull Creek watershed could be viewed as a microcosm for regional flood issues along the Edwards Plateau and Balcones Escarpment. Most, if not all, of the dominant physiographic features resulting in flash flood responses across the region are present in the watershed. Soils are primarily Brackett and Tarrant associations consisting of thin gravelly clay loam and stony clay soils, respectively (Fig. 7, right). Both soils have moderately low permeability when saturated and are underlain by limestone (beginning at 0–0.5 m in depth) throughout the catchment. Adjacent to stream courses, thicker Volente series soils can be found consisting of silty clay loam. Vegetation in the undeveloped areas of the watershed consists primarily of scrub oak and Ashe juniper stands. Topography within the basin varies greatly, with slopes ranging from nearly flat to 30% near the outlet. Table 3 provides some physical details of the basin and stream network.

Fig. 7.

(left) Land use and (right) soil maps for the Bull Creek catchment.

Fig. 7.

(left) Land use and (right) soil maps for the Bull Creek catchment.

Table 3.

Bull Creek geographic and land-use information.

Bull Creek geographic and land-use information.
Bull Creek geographic and land-use information.

Land use is roughly split evenly between developed and undeveloped lands (Fig. 7, left). Developed regions (mostly single-family residential) primarily occur along the northern and eastern region of the catchment, with open lands (mixed forest) occurring through the central region of the basin. In 2000, population in the basin was estimated at approximately 44 000, with expected increases to near 70 000 by 2030 (City of Austin 2010). The effect of urbanization on stream response to precipitation is unknown. While the traditional belief that increased urbanization has resulted in flashier streams across the region has permeated (e.g., Veenhuis and Gannett 1986), Sung and Li (2010) present a contrary hypothesis in their analysis of 10 area watersheds including Bull Creek. The authors suggest terraced landscapes produced through land grading practices during home construction have suppressed the amount and timing of surface runoff reaching stream courses. While their hypotheses were not explicitly tested in the current work, the study highlights the importance of surface processes in mediating storm response at Bull Creek.

b. GSSHA rainfall–runoff simulations

GSSHA is a fully distributed physically based watershed model with the capability to model the full hydrologic cycle (Downer and Ogden 2004, 2006). Parameters influencing hydrologic response are distributed across equally sized cells encompassing the watershed. Physics-based partial differential equations describing runoff processes are solved for each grid cell and channel reach to route rainfall through the landscape and stream network. The model and its earlier form [Cascade of Planes, Two-Dimensional (CASC2D)] have been used in the analysis of several flooding events across diverse geographic settings (e.g., Ogden et al. 2000; Sharif et al. 2006, 2010a,b, 2013; Chintalapudi et al. 2012, 2014; Elhassan et al. 2013; Wright et al. 2014b).

Sharif et al. (2010a) used the model to analyze the effect of varying precipitation inputs and land surface features on a smaller storm event in 2004 at Bull Creek. Precipitation totals from the 2004 event ranged from 10 to 20 mm across the basin, and peak flows were approximately one-quarter the size of the 2010 storm. The authors calibrated the model manually and were able to achieve an RMSE, error in peak, and error in volume of 12.4%, 5.1%, and 1.6%, respectively. The present results reflect the parameterization from Sharif et al. (2010a) with some slight modifications to landscape retention. Model setup is briefly described below; however, the reader is asked to refer to Sharif et al. (2010a) for a further discussion of model preprocessing.

1) Model setup

Hydrological processes simulated for the event include infiltration, landscape retention, overland flow, and stream routing. Evapotranspiration and deep aquifer contributions to streamflow were not included. Model preprocessing was conducted using ArcGIS and Aquaveo’s watershed modeling system. Watershed terrain was constructed from USGS 10-m DEMs filled using the Cleandam algorithm distributed with the GSSHA model. The USGS gauging station on Bull Creek (USGS 08154700) served as the outlet. Land-use, land-cover, and soil type data were obtained from the City of Austin (City of Austin 2014), National Land Cover Dataset (Homer et al. 2015), and the Natural Resources Conservation Service (Soil Survey 2013), respectively.

Infiltration calculations were conducted using Green and Ampt with redistribution (Ogden and Saghafian 1997) and saturated hydraulic conductivity values provided by Rawls et al. (1983). Saturated hydraulic conductivity for developed soil texture classifications (single or multifamily and commercial) were reduced by an amount proportional to the assumed fraction of impervious cover. Streets were assumed to have infiltration rates near zero. Grid cells were assigned to one of nine land-use classes for retention and overland roughness parameterization (Fig. 7, left). Table 4 shows surface parameters used in the model run. Low-intensity rainfall of approximately 20 mm occurred across the basin in the days leading to the event, and initial soil moisture was set at 0.2. The role of initial soil moisture mediating flood response is discussed below.

Table 4.

GSSHA infiltration and overland flow parameters.

GSSHA infiltration and overland flow parameters.
GSSHA infiltration and overland flow parameters.

Stream channels were modeled using irregular cross sections containing the main channel and flood plain (allowing for control of floodplain simulation). Stream cross-sectional data and reach-specific Manning’s roughness coefficient values were obtained from a City of Austin HEC-2 model. No cross-sectional data were available for the tributaries and they were simulated as a uniform trapezoidal profile. Tributary dimensions along with floodplain delineation were estimated from lidar data available from the City of Austin. Routing was calculated using the diffusive wave equation in 1D for streams and 2D for overland flow. Sharif et al. (2010a) demonstrated the large effect that stormwater detention facilities in the basin have on event flows. Major detention facilities were included in the model by disconnecting all or part of the structure catchment area if modeled inflows were less than basin capacity. The model was run on a 30-m grid with a 5-min time step. Errors in peak flow, volume, and RMSE percentage were calculated to assess model performance. Error in peak flow or volume is expressed as a percentage and given by

 
formula

where observed (obs) corresponds to measured peak flow and total volume while estimated (est) indicates GSSHA results. RMSE is also expressed as a percentage and given by

 
formula

where i = index denoting individual hydrograph ordinates and N refers to the total number of hydrograph ordinates.

2) Hydrologic simulation results

The heavy rainfall over Bull Creek resulted in a large, steep hydrograph peaking at 510 m3 s−1 in the early morning hours of 8 September (Fig. 8a). The peak flow was the highest recorded in the gauge’s 34-yr history, topping the previous high of 388 m3 s−1 in 1982. Peak flow was about half of the discharge envelope curve for watersheds this size near Austin (Asquith and Slade 1995). Mean stream velocity estimated by Winters (2012) using the slope area method was 4.1 m s−1. Peak unit discharge from the contributing areas was estimated at 10.97 m3 s−1 km−2 with a runoff ratio of 0.42. The unit discharge was the third highest among 37 area gauging stations analyzed by Winters (2012). Given the soil characteristics and rainfall intensity, it is likely almost all of the runoff is generated from infiltration-excess mechanisms. Figure 8a displays the GSSHA-generated hydrograph along with measured flow. The model simulated the event very well with error in peak, error in volume, and RMSE of 2.8%, 3.9%, and 22.1%, respectively. In their description of a real-time flood forecasting system in Austin, Looper and Vieux (2012) show similar results for a stage hydrograph. No model performance results are reported in their work.

Fig. 8.

GSSHA model output time series. (a) Simulated and observed streamflow. (b) Basinwide-average precipitation and infiltration rates (mm h−1) and soil saturation (%). (c) Basinwide cumulative precipitation and infiltration (mm). (d) Percent of developed and open areas receiving >20 mm h−1 of rainfall.

Fig. 8.

GSSHA model output time series. (a) Simulated and observed streamflow. (b) Basinwide-average precipitation and infiltration rates (mm h−1) and soil saturation (%). (c) Basinwide cumulative precipitation and infiltration (mm). (d) Percent of developed and open areas receiving >20 mm h−1 of rainfall.

Further insight into the hydrometeorological controls on the flood event can be gained through examination of watershed response to the two precipitation peaks previously discussed (2200–2300 UTC 7 September; 0230–0630 UTC 8 September). The first pulse of heavy precipitation, 34 mm in total, produced only a modest hydrologic response (from ~3 to 30 m3 s−1). The time between the first rainfall peak and the modest local discharge maximum was near 3 h, which was equal to the overall lag time (defined here as time difference between peak discharge and the time centroid of basin-averaged rainfall). The second, larger peak in basin-averaged precipitation (142 mm) was separated from peak discharge by only 70 min. The second peak in precipitation contained accumulations around 4 times larger than the first, but stream response was nearly 14 times greater.

Figure 8b displays time series of basinwide averages for soil saturation (%), infiltration rate (mm h−1), and rainfall rate (mm h−1). Cumulative values for infiltration and rainfall (mm) are shown in Fig. 8c. The plot indicates that, leading up to the first period of enhanced rainfall, infiltrability was high enough to accept nearly all of the rainfall. Soils became saturated across the basin after receiving approximately 35 mm prior to the onset of the first precipitation peak. Modest overland flow was generated during the first precipitation pulse. During this time, and through the heaviest precipitation, basinwide infiltration rates operated at or near saturated hydraulic conductivity.

Despite similar soil moisture conditions, the two periods of rainfall exhibited vastly different flow responses. This can be partially explained by the spatial organization of rainfall along with landscape characteristics. Distributed rainfall intensities were overlaid with land-use data to calculate percentage of open areas (49% of basin) and developed regions (51% of basin) experiencing rainfall in excess of 20 mm h−1 (estimated from 15-min rainfall accumulations). During the first peak in precipitation, about 90% of open areas received greater than 20 mm h−1 compared to 75% of urban areas (Fig. 8d). The heaviest rainfall cells were centered over the western side of the basin in the uplands portion of the watershed (Fig. 9a). Mixed forest land use and a large detention facility draining urban areas received much of this rainfall. During the second precipitation peak, rainfall was heavy through the central portion of the basin, with the largest 1-h accumulations occurring within a couple of kilometers from the outlet (Fig. 9b). This area received lower event totals than other regions of the basin, but it served a large role in controlling the timing and shape of the hydrograph.

Fig. 9.

During the two peaks in precipitation, 1-h max accumulations: (a) 2200–2300 UTC 7 Sep and (b) 0430–0530 UTC 8 Sep. The max 1-h accumulation occurs in (b).

Fig. 9.

During the two peaks in precipitation, 1-h max accumulations: (a) 2200–2300 UTC 7 Sep and (b) 0430–0530 UTC 8 Sep. The max 1-h accumulation occurs in (b).

In their 2010 Bull Creek investigation, Sharif et al. (2010a) note a significant overestimation in peak (80%) and volume (100%) when the mixed forest land-use type is modeled as grassland. The authors attribute this in part to the higher retention values for the forest land use. The authors used a retention value of 10 mm whereas we arrive at 15 mm for the mixed forest land use (Table 4). To further understand the effects of mixed forest surface retention, the event was modeled with seven different retention values ranging from 0.25 to 2 of the 15-mm value. To clarify how the changes affect the initial stream response and peak flows, results are presented for the initial response (small peak; from 1845 UTC 7 September to 0300 UTC 8 September 2010), peak flows (large peak; at 0300–0500 UTC 8 September 2010), and the full hydrograph. Figure 10 displays model performance statistics shown in Eqs. (1) and (2) for mixed forest retention perturbations.

Fig. 10.

Model performance statistics for small peak, large peak, and full hydrograph under (left) various mixed forest retention and (right) basinwide initial soil saturation representations.

Fig. 10.

Model performance statistics for small peak, large peak, and full hydrograph under (left) various mixed forest retention and (right) basinwide initial soil saturation representations.

Results indicate retention from this land use has a large relative effect in mediating the initial stream response, but the effects are greatly subdued when examining peak flows. The initial, smaller pulse in streamflow has a high relative change to retention perturbations, while RMSE and error in peak and volume are all within 10% of original modeled values for peak flows and overall hydrograph. The high retention factor for this land use may be related to losses to the karstic landscape. In the nearby Onion Creek basin, Looper and Vieux (2012) suggest losses to the karstic Edwards Aquifer may be responsible for reduced accuracy in simulations. There are numerous scattered karst features across the area, such as sinkholes and dissolution-enhanced fractures, which can provide rapid recharge (Jones 2003).

Regardless of the physical mechanism mediating stream response from the first peak in rainfall, retention adjustments to the mixed forest land use had little effect on the accuracy of the overall hydrograph. Several studies indicate land surface details are less important for extreme events, and accuracy of the hydrograph is highly dependent upon rainfall representation (e.g., Andréassian et al. 2004). Chintalapudi et al. (2012) showed the effect of land cover on hydrologic response decreases as the size of the rainfall event increases. Similarly, Sharif et al. (2015) demonstrated that the effects of land use on flood peaks are higher for storms with smaller recurrence intervals. To further highlight this notion, initial soil moisture was adjusted in a similar manner to mixed forest retention values (Fig. 10). GSSHA and CASC2D simulations have shown high sensitivities to initial soil moisture in other studies (Marsik and Waylen 2006; Senarath et al. 2000). Results indicate a similar pattern to soil retention perturbations where initial watershed response shows large relative change, but the overall effect on the hydrograph is minimal.

5. Summary and conclusions

An MCS formed from the remnants of Tropical Storm Hermine, producing torrential rainfall along a narrow swath of central Texas bordering the Balcones Escarpment. The system was able to persist through nighttime hours (7–8 September 2010), bringing nearly 300 mm of rainfall to the 58 km2 Bull Creek watershed in Austin, TX. The storm resulted in the largest flow event in the gauge’s 34-yr history. Flooding along the main stem of the creek resulted in one vehicular death as the car was swept from a low water crossing. The primary observations and findings from the study are as follows:

  1. The large-scale MCS was able to develop as a result of very moist Gulf air meeting a strong boundary of drier air from the northwest. WRF simulations with and without Balcones Escarpment terrain suggest geography aided in the increased rainfall pattern along FFA. Similar analyses of multiple events would provide a better understanding of orographic effects in the region. The well-defined, narrow rainbands and their persistence were key features that led to flooding.

  2. The Bull Creek watershed received nearly 300 mm for the event, with 24-h recurrence interval calculations from radar bins indicating a near 500-yr storm. Recurrence interval calculations less than 3 h indicate events smaller than 100-yr storms, suggesting duration was the major factor in event accumulations. Approximately 60% of the total accumulation fell during two periods lasting a combined 5 h. Average intensity rates across the basin were greater than 50 mm h−1 during maximum 1-h accumulations. Peak intensity from individual radar bins neared 175 mm h−1.

  3. Streamflow showed a nonlinear response to two periods of intense rainfall. This can be partially explained by the spatial organization of rainfall across the basin along with landscape characteristics. Model simulations using various representations of mixed land-use retention and initial soil moisture indicate the initial watershed response is quite sensitive to these parameters. The effect of these parameters on peak streamflow was minimal.

Acknowledgments

This work was partially funded by U.S. Army Corps of Engineers Engineer Research and Development Center Contract W912H2-16-P-0160. The authors thank Bob Rose, chief meteorologist for the Lower Colorado River Authority for his review of the storm description. In addition, the writers extend their thanks to three anonymous reviewers for making important suggestions that improved the manuscript.

REFERENCES

REFERENCES
Andréassian
,
V.
,
A.
Oddos
,
C.
Michel
,
F.
Anctil
,
C.
Perrin
, and
C.
Loumagne
,
2004
:
Impact of spatial aggregation of inputs and parameters on the efficiency of rainfall–runoff models: A theoretical study using chimera watersheds
.
Water Resour. Res.
,
40
,
W05209
, doi:.
Arndt
,
D. S.
,
J. B.
Basara
,
R. A.
McPherson
,
B. G.
Illston
,
G. D.
McManus
, and
D. B.
Demko
,
2009
:
Observations of the overland reintensification of Tropical Storm Erin (2007)
.
Bull. Amer. Meteor. Soc.
,
90
,
1079
1093
, doi:.
Asquith
,
W.
,
1998
: Depth–duration frequency of precipitation for Texas. USGS Water Resources Investigations Rep. 98-4044, 107 pp. [Available online at http://pubs.er.usgs.gov/publication/wri984044.]
Asquith
,
W.
,
1999
: Areal-reduction factors for the precipitation of the 1-day design storm in Texas. USGS Water Resources Investigations Rep. 99-4267, 81 pp. [Available online at http://pubs.er.usgs.gov/publication/wri994267.]
Asquith
,
W.
, and
R.
Slade
,
1995
: Documented and potential extreme peak discharges and relation between potential extreme peak discharges and probable maximum flood peak discharges in Texas. USGS Water Resources Investigation Rep. 95-4249, 63 pp. [Available online at http://pubs.usgs.gov/wri/1995/4249/report.pdf.]
Avila
,
L.
,
2010
: Tropical Storm Hermine 5–9 September 2010. Tropical Cyclone Rep., National Hurricane Center, 17 pp. [Available online at http://www.nhc.noaa.gov/data/tcr/AL102010_Hermine.pdf.]
Baker
,
V.
,
1975
: Flood hazards along the Balcones escarpment in central Texas. Geological Circular 75-5, Bureau of Economic Geology, 22 pp. [Available online at http://www.lib.utexas.edu/books/landscapes/publications/txu-oclc-1967634/txu-oclc-1967634.pdf.]
Caracena
,
F.
, and
J.
Fritsch
,
1983
:
Focusing mechanisms in the Texas Hill Country flash floods of 1978
.
Mon. Wea. Rev.
,
111
,
2319
2332
, doi:.
Caran
,
S. C.
, and
V.
Baker
,
1986
: Flooding along the Balcones Escarpment, Central Texas. The Balcones Escarpment, Central Texas. P. Abbot and C. Woodruff, Eds., Geological Society of America, 1–14.
Chintalapudi
,
S.
,
H. O.
Sharif
,
S.
Yeggina
, and
A.
Elhassan
,
2012
:
Physically based, hydrologic model results based on three precipitation products
.
J. Amer. Water Resour. Assoc.
,
48
,
1191
1203
, doi:.
Chintalapudi
,
S.
,
H. O.
Sharif
, and
H.
Xie
,
2014
:
Sensitivity of distributed hydrologic simulations to ground and satellite based rainfall
.
Water
,
6
,
1221
1245
, doi:.
City of Austin
,
2010
: Bull Creek watershed: Summary sheet. Watershed Protection Department, 12 pp. [Available online at http://www.austintexas.gov/sites/default/files/files/Watershed/eii/Bull_EII_ph2_2010.pdf.]
City of Austin
,
2014
: Land Use 2012. City of Austin GIS Department, accessed 6 October 2014. [Available online at ftp://ftp.ci.austin.tx.us/GIS-Data/Regional/coa_gis.html.]
Dee
,
D. P.
, and Coauthors
,
2011
:
The ERA-Interim reanalysis: Configuration and performance of the data assimilation systems
.
Quart. J. Roy. Meteor. Soc.
,
137
,
553
597
, doi:.
Doswell
,
C.
,
H.
Brooks
, and
R.
Maddox
,
1996
:
Flash flood forecasting: An ingredients-based methodology
.
Wea. Forecasting
,
11
,
560
581
, doi:.
Downer
,
C. W.
, and
F. L.
Ogden
,
2004
:
GSSHA: Model to simulate diverse stream flow producing processes
.
J. Hydrol. Eng.
,
9
,
161
174
, doi:.
Downer
,
C. W.
, and
F. L.
Ogden
,
2006
: Gridded Surface Subsurface Hydrologic Analysis (GSSHA) user’s manual. ERDC/CHL SR-06-1, Engineering Research and Development Center, USACE, 207 pp.
Elhassan
,
A.
,
H. O.
Sharif
,
T.
Jackson
, and
S.
Chintalapudi
,
2013
:
Performance of a conceptual and a physically-based model in simulating the response of a semi-urbanized watershed in San Antonio, Texas
.
Hydrol. Processes
,
27
,
3394
3408
, doi:.
Evans
,
C.
,
R. S.
Schumacher
, and
T. J.
Galarneau
,
2011
:
Sensitivity in the overland reintensification of Tropical Cyclone Erin (2007) to near-surface soil moisture characteristics
.
Mon. Wea. Rev.
,
139
,
3848
3870
, doi:.
Galarneau
,
T. J.
,
L. F.
Bosart
, and
R. S.
Schumacher
,
2010
:
Predecessor rain events ahead of tropical cyclones
.
Mon. Wea. Rev.
,
138
,
3272
3297
, doi:.
Homer
,
C. G.
,
J. A.
Dewitz
,
L.
Yang
,
S.
Jin
,
P.
Danielson
,
G.
Xian
,
J.
Coulston
,
N. D.
Herold
,
J. D.
Wickham
, and
K.
Megown
,
2015
:
Completion of the 2011 National Land Cover Database for the conterminous United States-Representing a decade of land cover change information
.
Photogramm. Eng. Remote Sens.
,
81
,
345
354
.
Jones
,
I.
,
2003
: Groundwater availability modeling: Northern segment of the Edwards Aquifer, Texas. Texas Water Development Board Rep. 358, 75 pp. [Available online at https://www.twdb.texas.gov/publications/reports/numbered_reports/doc/R358/Report%20358%20Northern%20Edwards.pdf.]
Kimball
,
S. K.
,
2008
:
Structure and evolution of rainfall in numerically simulated landfalling hurricanes
.
Mon. Wea. Rev.
,
136
,
3822
3847
, doi:.
Lin
,
Y.
,
S.
Chiao
,
T.
Wang
,
M.
Kaplan
, and
R.
Weglarz
,
2001
:
Some common ingredients for heavy orographic rainfall
.
Wea. Forecasting
,
16
,
633
660
, doi:.
Looper
,
J. P.
, and
B. E.
Vieux
,
2012
:
An assessment of distributed flash flood forecasting accuracy using radar and rain gauge input for a physics-based distributed hydrologic model
.
J. Hydrol.
,
412–413
,
114
132
, doi:.
Marsik
,
M.
, and
P.
Waylen
,
2006
:
An application of the distributed hydrologic model CASC2D to a tropical montane watershed
.
J. Hydrol.
,
330
,
481
495
, doi:.
Nielsen-Gammon
,
J. W.
,
F.
Zhang
,
A. M.
Odins
, and
B.
Myoung
,
2005
:
Extreme rainfall in Texas: Patterns and predictability
.
Phys. Geogr.
,
26
,
340
364
, doi:.
NOAA
,
2010
: Tropical Storm Hermine: September 5–8, 2010. NWS WFO, accessed 10 October 2014. [Available online at http://www.srh.noaa.gov/crp/?n=tshermine#hermine.]
Ogden
,
F. L.
, and
B.
Saghafian
,
1997
:
Green and Ampt infiltration with redistribution
.
J Irrig. Drain.
,
123
,
386
393
, doi:.
Ogden
,
F. L.
,
H. O.
Sharif
,
S.
Senarath
,
J.
Smith
,
M. L.
Baeck
, and
J.
Richardson
,
2000
:
Hydrologic analysis of the Fort Collins, Colorado, flash flood of 1997
.
J. Hydrol.
,
228
,
82
100
, doi:.
Patton
,
P. C.
, and
V. R.
Baker
,
1976
:
Morphometry and floods in small drainage basins subject to diverse hydrogeomorphic controls
.
Water Resour. Res.
,
12
,
941
952
, doi:.
Rawls
,
W. J.
,
D. L.
Brakensiek
, and
N.
Miller
,
1983
:
Green–Ampt infiltration parameters from soils data
.
J. Hydraul. Eng.
,
109
,
62
70
, doi:.
Schumacher
,
R. S.
,
2011
:
Distant effects of a recurving tropical cyclone on rainfall in a midlatitude convective system: A high-impact predecessor rain event
.
Mon. Wea. Rev.
,
139
,
650
667
, doi:.
Senarath
,
S.
,
F.
Ogden
,
C.
Downer
, and
H. O.
Sharif
,
2000
:
On the calibration and verification of two-dimensional, distributed, Hortonian, continuous watershed models
.
Water Resour. Res.
,
36
,
1495
1510
, doi:.
Sharif
,
H. O.
,
D.
Yates
,
R.
Roberts
, and
C.
Mueller
,
2006
:
The use of an automated nowcasting system to forecast flash floods in an urban watershed
.
J. Hydrometeor.
,
7
,
190
202
, doi:.
Sharif
,
H. O.
,
A. A.
Hassan
,
S.
Bin-Shafique
,
H.
Xie
, and
J.
Zeitler
,
2010a
:
Hydrologic modeling of an extreme flood in the Guadalupe River in Texas
.
J. Amer. Water Resour. Assoc.
,
46
,
881
891
, doi:.
Sharif
,
H. O.
,
L.
Sparks
,
A. A.
Hassan
,
J.
Zeitler
, and
H.
Xie
,
2010b
:
Application of a distributed hydrologic model to the November 17, 2004 flood of Bull Creek watershed, Austin, Texas
.
J. Hydrol. Eng.
,
15
,
651
657
, doi:.
Sharif
,
H. O.
,
M.
Hossain
,
T.
Jackson
, and
S.
Bin-Shafique
,
2012
:
Person–place–time analysis of vehicle fatalities caused by flash floods in Texas
.
Geomatics Nat. Hazards Risk
,
3
,
311
323
, doi:.
Sharif
,
H. O.
,
S.
Chintalapudi
,
A.
Elhassan
,
H.
Xie
, and
J.
Zeitler
,
2013
:
Physically-based hydrological modeling of the 2002 floods in San Antonio, Texas
.
J. Hydrol. Eng.
,
18
,
228
236
, doi:.
Sharif
,
H. O.
,
F. H.
Al-Juaidi
,
A.
Al-Othman
,
I.
Al-Dousary
,
E.
Fadda
,
S.
Jamal-Uddeen
, and
A.
Elhasssan
,
2015
: Flood hazards in an urbanizing watershed in Riyadh, Saudi Arabia. Geomatics Nat. Hazards Risk, doi:,
in press
.
Skamarock
,
W. C.
, and Coauthors
,
2008
: A description of the Advanced Research WRF version 3. NCAR Tech. Note NCAR/TN-475+STR, 113 pp., doi:.
Slade
,
R. M.
, Jr.
, and
J.
Patton
,
2003
: Major and catastrophic storms and floods in Texas. USGS Open-File Rep. 03-193. [Available online at http://pubs.usgs.gov/of/2003/ofr03-193/cd_files/USGS_Storms/index.htm.]
Smith
,
J. A.
,
M. L.
Baeck
,
J. E.
Morrison
, and
P.
Sturdevant-Rees
,
2000
:
Catastrophic rainfall and flooding in Texas
.
J. Hydrometeor.
,
1
,
5
25
, doi:.
Soil Survey
,
2013
: Web Soil Survey. NRCS, USDA, accessed 6 October 2014. [Available online at http://websoilsurvey.nrcs.usda.gov/.]
Sung
,
C. Y.
, and
M.
Li
,
2010
:
The effect of urbanization on stream hydrology in hillslope watersheds in central Texas
.
Hydrol. Processes
,
24
,
3706
3717
, doi:.
Veenhuis
,
J. E.
, and
D. G.
Gannett
,
1986
: The effects of urbanization of floods in the Austin Metropolitan Area, Texas. USGS Water Resources Investigations Rep. 86-4069, 73 pp. [Available online at http://pubs.er.usgs.gov/publication/wri864069.]
Ware
,
R. H.
, and Coauthors
,
2000
:
SuomiNet: A real-time national GPS network for atmospheric research and education
.
Bull. Amer. Meteor. Soc.
,
81
,
677
694
, doi:.
Winters
,
K.
,
2012
: Floods in Central Texas, September 7–14, 2010. Tex. Water J., 3, 14–25.
Wright
,
D. B.
,
J. A.
Smith
, and
M. L.
Baeck
,
2014a
:
Critical examination of area reduction factors
.
J. Hydrol. Eng.
,
19
,
769
776
, doi:.
Wright
,
D. B.
,
J. A.
Smith
, and
M. L.
Baeck
,
2014b
:
Flood frequency analyses using radar rainfall fields and stochastic storm transposition
.
Water Resour. Res.
,
50
,
1592
1615
, doi:.
Zahran
,
S.
,
S.
Brody
,
W.
Peacock
,
A.
Vedlitz
, and
H.
Grover
,
2008
:
Social vulnerability and the natural and built environment: a model of flood casualties in Texas
.
Disasters
,
32
,
537
560
, doi:.