Real-Time Stormwater Management Using Depth, Duration, Frequency Thresholds

WEFTEC®.06

REAL-TIME STORMWATER MANAGEMENT USING DEPTH, DURATION, FREQUENCY THRESHOLDS

Susan Janek, P.E., George Oswald, P.E., Baxter Vieux, P.E., Ph.D. City of Austin Watershed Protection and Development Review Department 505 Barton Springs Road, Suite 1200, Austin, Texas 78704

City of Austin Watershed Protection and Development Review Department Vieux and Associates, Inc.

ABSTRACT

The City of Austin (City) operates a Flood Early Warning System (FEWS) to reduce public exposure to flash flood hazards. The system is operated and maintained by the Watershed Protection and Development Review Department, Watershed Engineering Division. Advanced information and warning of urban storm water flooding is achieved by the FEWS. A critical component of the system is rainfall detection and interpretation. Radar rainfall detection and forecasting also utilizes the existing ALERT (Automated Local Evaluation in Real Time) rain gauge network. An innovative web-based hydrologic information system built upon radar hydrology and implemented in 2004 has proved to be useful during storms producing heavy precipitation and flooding. This presentation will describe the development and use of Depth, Duration, Frequency (DDF) values in near real-time.

KEYWORDS Flood warning, DDF, real-time, internet, hydrologic information system, radar hydrology

BACKGROUND Central Texas is often called "Flash Flood Alley" because of its frequent, intense storms. While large events seem to happen every decade, lesser events also cause public safety concerns. Coordination between the Watershed Protection and Development Review Department (WPRDR) and Office of Emergency Management (OEM), both City of Austin agencies, results in hydrologists and emergency managers working together in the Emergency Operations Center (EOC) to warn of and respond to flooding along creeks and at major intersections in the City. Because of the challenge with identification and timely response during high-risk flood events, an integrated warning system is needed so that appropriate actions and deployment of resources can be effectively managed.

Operation of the Flood Early Warning System (FEWS) helps reduce public exposure to flash flood hazards. The system is operated and maintained by the WPDRD, Watershed Engineering Division. The FEWS is used to identify the development of flood hazard conditions that warrant public safety response for roadway closings and building evacuations. The FEWS consists of a variety of field sensors which transmit stream water level and rainfall data to a central data management computer system for decision-making.

The FEWS has been in operation since 1985 and is considered an ALERT system. Portions of the FEWS have been expanded and upgraded with growth in the service area and advancing technology. Currently, the field sensor network consists of 86 rain gauges, 41 stream depth gauges, and ten automated low-water crossing barricades. Information from the sensors is transmitted by a low frequency FM radio to a repeater station. From the radio repeater station, the information is then retransmitted on a different frequency to the base station – the Emergency Operations Center, where it is managed by the Hydrolynx Novastar software system. The FEWS provides rainfall and stream stage information for 26 major watersheds and it extends over 1400 square miles. The gauge components themselves are provided by both Hydrolynx Systems and High Sierra Electronics. Data received from the gauges is pushed to a File Transfer Protocol Site (.FTP) that is maintained by Hydrolynx Novastar. This information is monitored and used by the National Weather Service and the City of Austin’s publicly available internet site. The information is also sent to the City’s rainfall processing contractor Vieux and Associates, Inc. (VAI). Radar-based rainfall data services are provided via customized Internet access operated as a web service by VAI.

RADAR RAINFALL PROCESSING

Radar hydrology is the use of weather radar measurements for hydrologic applications. It deals with the estimation of current spatial variability and rainfall intensity and accumulation over a selected time interval for deriving estimates of runoff and stream flow. The FEWS includes the application of near real-time, bias-adjusted NEXRAD based rainfall estimates at square kilometer horizontal resolution to augment ground-based data for flood prediction. Initial deployment of a flood hazard prediction system supports real-time hydrologic modeling driven by precipitation values derived from radar and the heterogeneous rain gauge network observations at 15-minute increments.

The objective of the system is to quickly identify specific basins that are at high risk of flooding during a storm event. Multiple end-users need access to such information simultaneously and potentially at multiple locations. The information needs to be easy to interpret and appropriate actions need to be predefined and displayed unambiguously. The methodology used to achieve the objective for identification of flooding in specific basins is described below.

METHODOLOGY

Prior studies performed by the U.S. Geological Survey (USGS) for the Austin area led to identification of the rainfall depth, duration, and frequency (DDF) information. Over a period of time with numerous hydrologic and hydraulic models developed by and for the City of Austin plus the calibration of those models to real storm events has produced identification of the specific DDF information per each watershed that causes localized flooding in relatively small basins within the City and in areas that extend beyond the corporate limits. This experience along with existing DDF data was incorporated in a website that displays real-time rainfall information. The rainfall information is processed by combining National Weather Service (NWS) radar data with City rain gauge data. The processed rainfall is then displayed using a web site that was created by Vieux and Associates. Processed rainfall maps are generated every 15 minutes and displayed via automatic updates by the website. The rainfall is compiled as gridded maps with

1x1 km resolution of processed rainfall information displayed over the City and Travis County. In addition, gridded rainfall information may be displayed with pan and zoom capabilities to display rainfall information over a high-resolution street network, or aggregated within watershed basin boundaries.

The City developed the flood threat thresholds and action plans, critical basin response information, and requirements for information to be displayed by the website. With 15-minute updates, each basin is displayed with colors that are coded to show flood threat thresholds based on accumulated rainfall within specific time periods. The response time for a watershed can be characterized by the time of concentration, Tc. The time response depends on terrain slope, length, and channel characteristics. Table 1 presents selected watershed characteristics used to derive Tc. Smaller basins react more rapidly than large basins, though with smaller discharge.

Recommended actions are obtained from the website by dragging the cursor over any basin. When the basin is highlighted, it displays the basin DDF thresholds and the action plan for the current DDF status of the basin. Watershed basin colors advance from green to purple based on the status of the DDF as shown in Table 2. The ready display of color-coded information makes it easy provide specific responses during critical storm conditions, and saves valuable time for the Watershed Engineering Division employees to assess problems and recommend actions during a flooding event. In addition to the quantitative precipitation estimate, the Vieux and Associates website also implements a program that will forecast precipitation one hour in advance using radar that extends beyond the corporate limits of the City. Knowing whether more rainfall is expected to approach the City is useful for anticipating the appropriate response actions. The predictive rainfall is accomplished by advancing the radar imagery for the next hour while accounting for intensification and decay of precipitation.

Table 1-Data for selected basins used for Tc development

Stream- Plane U/s D/s Elev Avg. Tc INT(Tc) INT Critical L (mi) Length Elev (ft) Slope (hour) (Tc+0.5) Time (mi) (ft) (ft/mi) (hour) 51.433 25.693 1398.8 428.42 37.77 5.83 6 6 6 4.356 2.783 1010.6 842.17 60.53 1.34 1 2 1 3.984 3.009 847.32 483.19 121.02 0.91 1 1 1 17.59 11.339 1140.9 590.82 48.51 3.01 3 4 3 3.182 2.41 644.48 429.12 89.36 0.94 1 1 1 2.416 1.907 699.8 524.71 91.8 0.81 1 1 1 7.881 4.709 638.65 410.1 48.54 2.01 2 3 2 23.54 17.394 1107.3 652.37 26.15 4.74 5 5 6

A 10-year event or larger on a watershed scale will in most cases cause significant flooding so those events have been assigned a magenta color as seen in Table 2.

Table 2 - DDF rainfall amounts for a specific basin. The color shown is also displayed on the website for quick visual reference. Recurrence Interval 15 30 (YR) min min 1 hr 2 hr 3 hr 6 hr 8 hr 12 hr 1 day <2 2 0.98 1.32 1.72 2.16 2.32 2.67 2.87 3.06 3.44 2.5 1.05 1.42 1.86 2.35 2.53 2.91 3.12 3.33 3.84 3.33 1.14 1.54 2.04 2.58 2.79 3.19 3.42 3.64 4.33 5 1.26 1.71 2.28 2.89 3.13 3.56 3.82 4.07 4.99 10 1.47 1.98 2.68 3.42 3.71 4.21 4.51 4.81 6.10 25 1.76 2.36 3.28 4.20 4.55 5.14 5.52 5.90 7.64 50 2.01 2.68 3.79 4.88 5.28 5.94 6.40 6.86 8.87 100 2.29 3.04 4.37 5.66 6.11 6.85 7.41 7.96 10.20 250 2.73 3.57 5.26 6.86 7.38 8.24 8.96 9.67 12.00 500 3.11 4.02 6.06 7.94 8.51 9.47 10.34 11.20 13.50

Table 3 -Threat and category and corresponding action plan NO FLOODING PREDICTED. 1 May include light rain and drizzle < 1/2 inch per hour. Traffic impacts and localized drainage problems still possible. NUISANCE FLOODING 2 Low water crossings may be closed. Traffic slowed. Localized drainage problems possible. MINOR FLOODING 3 Near flood stage - minimal damage. Some flooding from storm drains, ditches and creeks. MODERATE FLOODING 4 Some evacuations may be required. Secondary roads blocked. Local drains overwhelmed - unpredictable localized flooding of streets/property.

FINDINGS

The first storm event to force operational testing of system occurred in the early morning of November 22, 2004. Rainfall that had occurred in the preceding 15-minutes is shown in Figure 2, and magnified in Figure 2. At 2:00 AM, Waller Creek, a critical urban area basin that frequently causes flooding, was displayed as Category 3 threat seen in Figure 4, which indicates that a critical rainfall threshold had been exceeded. A first responder was deployed to the basin for status verification and assessment. While flow in the creek was in-bank upon arriving at the site, flooding occurred shortly thereafter. In this event, the City had provided specific warning actions because they “beat” the flood to the location in the basin by use of their hydrologic information system.

Figure 1 – Gridded rainfall on November 22, 2004 2 AM

Figure 2 - Zoom to intense rainfall in gridded format (left), and basin averaged display (right)

Figure 3 - Waller Creek is orange in contrast to green -- indicating a Threat Category 3 risk of flooding

SIGNIFICANCE

Integration of rainfall monitoring technology resulted in an efficient and valuable tool for recognizing the threat of flooding with a detailed, site specific display system. Emergency managers have more accurate information about the location and severity of potential flooding through use of the information system. Internet accessibility and display using tools developed for the FEWS system results in a useful system that serves emergency managers and helps reduce the impact of flood hazards and increase public safety.

The FEWS tools make the development and improvement of future systems possible. PreVieux is a 1-hour predicted rainfall product that generates both image displays and numerical output. The PreVieux web interface is displayed in Figure 4. The lower picture demonstrates how the screen is captured during a rainfall event. Rainfall detected by radar and adjusted using rain a gauge network is translated forward in time allowing users to anticipate the movement and intensity of approaching rainfall. A notification system based on quantitative rainfall predictions supports staff awareness of impending high intensity rainfall.

Figure 4 - One hour predictive rainfall display called PreVieux. Upper panel (right and left) show the regional zoom and local subbasins affecting the City. Lower panel (left) shows approaching intense rainfall, and on the right, intense precipitation predicted to impact watersheds in the City.

Depth Duration and Frequency can be a very useful tool for identification of flood threat, as described. Figure 5 depicts a hydrograph with flood stage marked. The greater the time of warning or awareness of the flood threat, the more time is available to take action to protect life and property.

Figure 5 - Earlier threat awareness allows more time for action to be taken.

Limitations of the approach include the difficulty of verifying the time of concentration, Tc, for a particular event. Different responses to the same depth of rainfall may result due to the spatial distribution of rainfall that differs for each event and is rarely uniform. The effect of spatially variable rainfall is especially pronounced for larger basins, which poses a difficulty for identification of a flood threat from rainfall depth averaged over a basin. Finally, the current DDF approach does not incorporate the effect or variability of antecedent soil moisture conditions. While experienced users can temper pronouncements of an impending flood based on recent rainfall and soil moisture conditions, others with less training and/or experience may have difficulty with correct interpretation.

Running hydrologic models in near real-time is now technologically possible, and has been implemented in urban and in large scale river basins as described by Vieux and Bedient (2004); Vallabhaneni, et al. (2005); Vieux and Vieux (2005); and Vieux et al. (2003). Prediction accuracy of hydrologic models, whether lumped or distributed, benefits from comparison with observed stream flow where stream gauges exist. Technological advances in computing power and access to near real-time rainfall input have enabled near real-time flood forecasting. A demonstration of HEC-1, a USCAE model, modified and calibrated with real-time events was presented for Onion Creek by Williford (2005). Operational use of HEC-1 with radar rainfall input in an urban setting in Houston is described by Bedient et al. (2003). Currently, HEC-1 using radar rainfall input is running in near real-time for Onion Creek. A distributed hydrologic model that tracks rainfall runoff and soil moisture is being evaluated as a member of a model ensemble. Input and output are updated every 15-minutes. Strengths of the HEC-1 model include user familiarity and understanding, and calibration using actual events using radar rainfall as input. Neither DDF nor HEC-1 are able to model or update antecedent soil saturation, or represent spatially variable rainfall at scales smaller than the sub-basin. The sub-basin

descretization of the Onion Creek watershed in HEC-1 is presented in Figure 6 along with the location of USGS stream gauges.

Recharge to the Edwards Aquifer from creeks in the area is well documented. A portion of Onion Creek is underlain by the recharge area as shown in Figure 7, which is highlighted in light blue. This geologic feature contributes to a complex hydrologic challenge where channel losses by recharge to the aquifer reduce the flow arriving downstream.

Figure 6 - USGS gauge locations and HEC calibration points.

Figure 7 - Onion Creek with Edwards Aquifer Recharge area highlighted.

Vflo, a physics-based distributed hydrologic model was developed to integrate spatially variable radar rainfall input with distributed parameters to model runoff and track soil moisture. The system relies on the kinematic wave equations to describe how water flows downstream through the drainage network. Strengths of this type of model include the ability to track soil moisture, take the spatial variability of rainfall into account, and explicitly represent the variability in terrain, soils, landuse/cover and impervious area. The model is setup and parameterized using GIS data and calibration using real rainfall events. Recharge is modeled as loss from the channel where the creeks cross the Edwards Aquifer Recharge area. The Vflo model is setup at 250 m resolution, resulting in 13,365 cells for the drainage area of approximate 323 square miles representing Onion Creek as shown in Figure 8. Continuous modeling using radar rainfall input for a 5-year period improves event-based results, and is used to prepare the model for real-time operation. The model for Onion Creek is being evaluated for use in the FEWS system.

Figure 8 - Onion Creek displayed in the Vflo interface.

CONCLUSION

Public safety concerns about flooding resulted in the development of a system called FEWS that is used by the City of Austin in association with the Office of Emergency Management. The technological advances in radar rainfall monitoring, telemetered rain gauge networks, and web services result in a unique system that is capable of supporting hydrologists and emergency managers who respond to flooding along creeks and at major intersections in the City. Since its inception, the radar rainfall monitoring system has proven useful during many extreme events by providing information even before the flood arrives. Response times and threat thresholds for each basin are incorporated into the system so that basin-specific forecasts are made in terms of flood response. With rapid updates on measured and predicted precipitation during evolving storms, effective actions can be taken to help protect the public during flood events.

REFERENCES

Asquith, W.H., 1998, Depth-duration frequency of precipitation for Texas: U.S. Geological Survey Water-Resources Investigations Report 98-4044, 107.

Bedient, P. B., A. Holder, J. Benavides and B.E. Vieux, 2003. "Radar Based Flood Warning System - T.S. Allison." ASCE Journal of Hydrologic Engineering (Nov): 308-318.

Vieux, B.E., and J.E. Vieux, 2005. Rainfall Accuracy Considerations Using Radar and Rain Gauge Networks for Rainfall-Runoff Monitoring. Chapter 17 in Effective Modeling of Urban Water Systems, Monograph 13. W. James, K . N. Irvine, E. A. McBean & R.E. Pitt, Eds. ISBN 0-9736716-0-2.

Vallabhaneni, S., B.E. Vieux, and T. Meeneghan, 2005. Radar-rainfall technology Integration into Hydrologic and Hydraulic Modeling Projects. Chapter in Practical Modeling of Urban Water Systems, Monograph 12. Post published proceedings of the 2003 Stormwater and Urban Water Systems Modeling Workshops and Conference, Toronto Canada. Computational Hydraulics International, Guelph, Ontario.

Vieux, B.E., C. Chen, J.E. Vieux, and K.W. Howard. Operational deployment of a physics-based distributed rainfall-runoff model for flood forecasting in Taiwan. In proceedings, International Symposium on Weather Radar Information and Distributed Hydrological Modelling, IAHS General Assembly at Sapporo, Japan, July 3-11, 2003. eds. Tachikawa, B. Vieux, K.P. Georgakakos, and E. Nakakita, IAHS Red Book Publication No. 282: 251-257.

Williford, E., 2004. "Model development for a highly sloped watershed with attention to aquifer recharge to be used in a radar-based flood alert system." Submitted in partial fulfillment of Masters of Science, Civil and Environmental Engineering, Rice University, Houston, TX.