National Weather Hydrology Laboratory

Threshold Runoff Plan NOAA/National Weather Service Office of Hydrology Hydrologic Research Laboratory Silver Spring, MD 20910 April 1999 Threshold Runoff Plan Table of Contents Page 1.0 INTRODUCTION. . . . . . . . . . . . . . . . . . . . . . . .1 2.0 BACKGROUND. . . . . . . . . . . . . . . . . . . . . . . . .1 3.0 THE threshR METHOD. . . . . . . . . . . . . . . . . . . . .2 3.1 Basin Delineation and Parameter Estimation . . . . . .2 3.2 Threshold Runoff Computation . . . . . . . . . . . . .3 3.3 Interpolation. . . . . . . . . . . . . . . . . . . . .4 4.0 LIMITATIONS of threshR93. . . . . . . . . . . . . . . . . .4 5.0 AN UPDATED APPROACH . . . . . . . . . . . . . . . . . . . .5 5.1 IHABBS Database. . . . . . . . . . . . . . . . . . . .5 5.2 Data Resolution. . . . . . . . . . . . . . . . . . . .6 5.3 The New threshR. . . . . . . . . . . . . . . . . . . .7 6.0 REQUIREMENTS. . . . . . . . . . . . . . . . . . . . . . . .8 7.0 TIME LINE . . . . . . . . . . . . . . . . . . . . . . . . 10 8.0 FUTURE DIRECTION. . . . . . . . . . . . . . . . . . . . . 12 9.0 DEVELOPMENT RESOURCES . . . . . . . . . . . . . . . . . . 12 10.0 IMPLEMENTATION. . . . . . . . . . . . . . . . . . . . . . 13 THRESHOLD RUNOFF PLAN 1.0 INTRODUCTION With the deployment of the Weather Surveillance Radars - 1988 Doppler (WSR-88D) and the ability to display radar information on the Hydrologic Rainfall and Analysis Project (HRAP) grid, the spatial resolution for monitoring a flood-threatened area is greatly reduced from a county sized area to an area approximately 4 km on a side. To increase the utility of the radar information by comparing it to flash flood guidance, gridded flash flood guidance is needed for the same 4 km grid. This document describes the plan to develop and derive a parameter, gridded threshold runoff, required to compute gridded flash flood guidance. 2.0 BACKGROUND The Weather Forecast Offices (WFO) use flash flood guidance in the decision process to issue flash flood watches and warnings as part of the national flash flood watch/warning program. The River Forecast Centers (RFC) compute gridded, zone/county, and headwater flash flood guidance for the WFOs. Flash flood guidance is derived from an algorithm using the current soil moisture condition and the runoff needed to fill a stream channel at a particular location. Soil moisture conditions vary according to rain events and seasons of the year. Current soil moisture conditions are maintained by the rainfall-runoff models used by the RFCs. The runoff needed to fill a stream channel does not vary and is computed from the geophysical properties of the stream channel and stream network. For flash flood guidance purposes threshold runoff is the ratio of the stream flow at bankfull and the unit hydrograph peak flow for the watershed as shown in the following equation: R = Qp / qp* A (1) where R is threshold runoff in inches, Qp is the bankfull discharge in cubic feet per second (cfs), qp is the unit hydrograph peak flow in cfs per unit area in square miles (cfs/sq mi), and A is the area in square miles. Threshold runoff values are desired for thousands of small watersheds across the country. An automated, GIS-based procedure is necessary to process such a large amount of information. 3.0 THE threshR METHOD In 1993 OH developed an objective and hydrologically-based procedure to derive threshold runoff values on the HRAP grid for the entire country. Input parameters included 3 arc-second digital elevation data (approximately 90-m spacing), land use land cover (LULC) data, and EPA river reach files (RF3). The GRASS Geographic Information System (GIS) software was used to delineate small watersheds and to display input, intermediate, and output data. This GRASS-based procedure is known as threshR and will be referred to as threshR93 in this plan. The following paragraphs describe the three major steps required to derive gridded threshold runoff values: basin delineation and parameter estimation, threshold runoff computation, and interpolation. 3.1 Basin Delineation and Parameter Estimation In threshR93 the r.watershed program in GRASS determines the geographical boundaries for drainage areas. The program works with a hydrologic unit (HU) as defined by the USGS. Starting at the outlet of the HU (lowest elevation cell), the basin delineation algorithm looks up-slope (higher elevation cells) for ridges that define drainage areas to all the stream junctions. The hydrologic unit is subdivided into any number of sub-basins, which vary in size and shape, depending on a minimum drainage area criteria specified by the user. In addition to defining a set of sub-basin boundaries and their drainage areas, geometric parameters (such as length and slope) required to estimate unit hydrograph peak values and bankfull flow are determined for individual sub-basins and stream segments and as aggregate areas, lengths, and slopes accumulated along the stream paths. In threshR93, two methods are available for estimating unit hydrograph values and two methods are available for estimating bankfull flow. 3.2 Threshold Runoff Computation As defined by Equation 1, threshold runoff is the ratio of the stream flow at bankfull and the unit hydrograph peak flow for the watershed. Bankfull flow is derived by using one of two methods: (1) Mannings flow equation with observed site and channel characteristics or (2) the two-year return period flow derived from USGS regression equations. The unit hydrograph peak flow is derived using one of two methods: (1) Snyder's synthetic unit hydrograph or (2) the geomorphologic unit hydrograph. There are four methods available for computing runoff: Method 1: Bankfull Flow and Geomorphologic Unit Hydrograph Method 2: Bankfull Flow and Snyder's Unit Hydrograph Method 3: Two-Year Return Period Flow and Geomorphologic Unit hydrograph Method 4: Two-Year Return Period Flow and Snyder's Unit Hydrograph The method selected depends on the availability of data. With the usual case of sparse data, the 2-year return period flow regression equations are used to approximate bankfull flow and the Snyder synthetic unit hydrograph method is selected to determine the unit hydrograph peak flow. A single threshold runoff value is assigned to each headwater sub-basin. Moving downstream from the headwaters, each interbasin area is assigned a single threshold runoff value based on the total area upstream of its outlet point. Zero runoff values are assigned to sub-basins when (a) the accumulated area is greater than approximately 770 sq mi (2000 sq km), (b) the computed runoff is considered an outlier greater than 3 inches, or (c) the area is covered by or drains directly to a lake. 3.3 Interpolation The runoff values assigned to individual sub-basins at the stream junctions are interpolated to the HRAP grid for use with FFGS to compute gridded flash flood guidance. In many cases this will result in smoothing the runoff values. 4.0 LIMITATIONS of threshR93 Considerable time was required by a hydrologist to retrieve voluminous geophysical data from the USGS and EPA. Then, a significant number of hours of CPU time was required to process the data before computing threshold runoff values. Because the effort to use the threshR93 procedure was both computer and labor intensive (CPU, disk space, and a hydrologist), resources at the RFCs were not adequate to support threshR93 in addition to maintaining operational forecasts, and implementing new computer platforms. In an effort to improve operational speed a task was undertaken in 1997 to port the computer intensive portion of threshR93 (basin delineation using the GRASS program r.watershed) to a Cray computer. With the improved threshR93 the raw data is transferred from the RFC to the Cray where basin delineation is completed. The intermediate output files are transferred back to the RFC's workstation to complete the runoff and interpolation steps in the improved threshR93. Basin delineation on the Cray computer did not enhance performance compared with the RFCs workstations. On the contrary, the significantly longer run times on the Cray (three times longer than on the HP workstation) discouraged use of the Cray computer for the improved threshR93. The use of USGS 2-year return period flow equations to approximate bankfull flows cannot be used nationally since threshR93 is limited to only four of 33 regression equation parameters. Regression equation parameters that can be used include area, slope, stream length and one other parameter. The Corps of Engineers who developed and supported GRASS for several years have recently withdrawn their support. Now, NWS seems to be adopting the GIS software ArcInfo/ArcView as a matter of choice. Hence, the updated threshR will use ArcView rather than GRASS. 5.0 AN UPDATED APPROACH 5.1 IHABBS Database The National Operational Hydrologic Remote Sensing Center (NOHRSC) has developed an integrated hydrologic terrain database as input for their Integrated Hydrologic Automated Basin Boundary System (IHABBS). The IHABBS software and database are currently used to support RFC forecast basin delineation for water equivalent data, which NOHRSC derives seasonally for several RFCs. NOHRSC has invested substantial effort in developing IHABBS to support their mission. For use in the threshR93 software, 3 arc-second elevation values are projected and resampled to a 90-m grid spacing. The grids created at NOHRSC have a spatial resolution of 15 arc-seconds and these data will be used to support the updated threshR tool. This means that the number of data points in the North-South and East- West directions is reduced by a factor of 5, so that the amount of data that a tool like threshR analyzes is reduced by a factor of 25. 5.2 Data Resolution Experience using the NOHRSC 15" data in the Arkansas-Red Basin RFC (ABRFC) has shown that data at this resolution can be used to resolve drainage areas accurately down to about 20 mi2 (about the size of an HRAP cell). For many purposes, this resolution is adequate and, therefore, 15" data from the IHABBS database will be used to support the new threshR tool. The functionality of the new threshR will not preclude the use of finer resolution digital elevation model data, if desired. In some areas local government has contracted for high resolution digital terrain data which could be available to use in threshR. For an initial implementation it is expected that the differences in the resulting threshold runoff values derived using the 15" data resolution versus using the 3" data resolution will not be significant considering that flash flood guidance is used as a tool in issuing flash flood watches and warnings. It is worth noting that flash flood guidance is based on threshold runoff (a single variable) and soil moisture conditions in a forecast basin as maintained in NWSRFS. 5.3 The New threshR The motivation for developing the new threshold runoff software is to provide a more efficient and usable tool for the RFCs to compute threshold runoff values. The proposed new threshR software will be implemented using data generated at the NOHRSC. In the new threshR tool, threshold runoff will be computed using the same hydrologic theory used in threshR93. Geometric characteristics of stream segments and catchments and stream connectivity will be derived using the IHABBS database. The updated threshold runoff procedure will be less labor and computer intensive for the RFCs primarily because of reduced data volume and also because updated algorithms and data management tools will be more efficient. The new threshold runoff routines will be developed within the ArcView GIS software environment. The object-oriented scripting language called Avenue that comes with ArcView will be used to automate calculations in the new threshR and provide a customized GUI. The new threshR will require the Spatial Analyst extension to ArcView GIS to display and manipulate raster (gridded) data. There are many advantages to using ArcView GIS with the Spatial Analyst for hydrologic analysis. Built in functions and existing Avenue scripts in the public domain will facilitate quick development of the codes required to estimate threshold runoff and to create a customized GUI. If functionality external to ArcView is desired, then program calls can be made to external programs. In addition to improved usability, another advance that will be made for the new threshR application is the development of a GIS database that can be used to estimate bankfull flow in all 50 states and Puerto Rico. Bankfull flow can be approximated by the 2-year return period flow. In some areas the best approximation to bankfull flow is the 1.7 year return period and in others the best approximation is the 5-year and sometimes the 10-year return period. The USGS publishes a collection of the regression equations for 2- to 100-year flows for all 50 states and Puerto Rico. The equations use various parameters and there is no standardization between states. There are from 1 to 5 equations that apply to different zones in each state a total of about 150 2-year equations for the entire country. At least 33 different parameters are used in the 2- year regression equations. All the equations use drainage area while a few parameters are only used in one equation. The new threshR program will automatically read the required return flow equation parameters from a database and compute bankfull flow values when a working area is selected. 6.0 REQUIREMENTS A carefully designed GUI to execute the new threshR is critical to its degree of acceptance as a tool and its eventual use. The GUI must provide default values and still require a minimum of user interaction to compute gridded runoff values for a desired area. The functions that will be performed by the new threshR tool are: 1. Select a geographical area in which gridded runoff values will be computed. The topography and size of drainage basins must be considered when selecting the work polygon. Possible work polygons are the RFC area, WFO area, or a smaller area consisting of only a few sub-basins. Several options for specifying the work polygon will be available in the ArcView GUI. 2. Using flow direction values from the database, delineate watersheds at stream junctions greater than or equal to a user-specified minimum area. 3. Using flow direction and the delineated basin boundaries from step 2 and elevation and slope grids from IHABBS , compute geometric parameters for each sub-basin including but not limited to drainage area, stream length, and channel slope. 4. Use the parameter values from step 3 to compute unit hydrograph peak estimates. Options to use Snyder's synthetic unit hydrograph or the geomorphologic unit hydrograph are retained from threshR93. A suggestion was made to include the SCS unit hydrograph. 5. Using a nationwide GIS database built for determining the USGS return period flow regression parameters, compute the 2-, 5-, and 10-year return period flows for each sub-basin. 6. Compute the threshold runoff in each sub-basin (at each stream junction) using the results of steps 4 and 5. Runoff for each sub-basin (local runoff) and for the total area above and including the sub-basin (accumulated runoff) is computed. Thus, runoff values are computed over a range of sub-basin sizes. 7. Interpolate threshold runoff values from stream junctions to HRAP coordinates (as done in threshR93) so these values can be used in the current Flash Flood Guidance System. The ArcView GUI will be customized to facilitate execution of the above tasks with a minimum number of user responses. Minimally, the user must specify the geographic extent of interest, a unit hydrograph duration, a runoff calculation method, and an interpolation method. The customized ArcView GUI will facilitate the viewing of intermediate and final output, e.g. drainage areas, slopes, threshold runoff, etc. One method for evaluating the new threshR will be to compare results with output from threshR93 for Oklahoma.. 7.0 TIME LINE The approximate time line illustrated in Table 1 shows the planned development and the projected date of completion or release of each item. Figure 1: Threshold Runoff Development and Implementation Timeline Design regression equation database 2-6/99 Procure data maps 3-11/99 Digitize data maps 3-12/99 Code database object 4-7/99 Test database object 5-8/99 Deliver initial database to OH 8/99 Deliver update database to OH 12/99 Deliver final database to OH 2/00 Design projected database, basin delineation, characteristics 2-3/99 Design runoff parameter objects 3-5/99 Design interpolation object 4-5/99 Code db, delineation, characteristics 2-4/99 Code runoff parameter objects 4-5/99 7-8/99 Code interpolation object 4-5/99 7-8/99 GUI Design 2-5/99 7/99 GUI Coding 3-5/99 7-8/99 Prototype Demonstration 8/99 Feedback and Modifications 8-9/99 Internal OH Testing 8-10/99 External Beta Testing 10-12/99 Fix Bugs 1-2/00 Documentation User 7-8/99 10-01/00 System 11-01/00 Training 2-4/00 Deliver Executable and Documentation 2/00 RFCs derive gridded threshold runoff using threshR 2-7/00 8.0 FUTURE DIRECTION Currently, the Flash Flood Monitoring and Prediction (FFMP) application of WHFS uses gridded flash flood guidance to compare with gridded estimates of rainfall from the WSR- 88D radars. Forecasters at the WFOs must assess the flood potential in flood-prone sub-basins from gridded information displays on Advanced Weather Interactive Processing System (AWIPS) because sub-basin flash flood guidance and sub-basin estimated radar rainfall is not available. A future change will provide sub-basin information to determine the flood potential. To implement the change, the RFCs will give the sub-basin boundaries (from the new threshR tool and using finer spatial resolution data) to the WFOs for use in the FFMP application. Instead of gridded flash flood guidance, the RFCs will compute sub-basin flash flood guidance for the WFOs. Then, using the sub-basin boundaries at the WFOs, the radar estimated rainfalls will be derived from gridded values for the same sub-basins. As a result, the FFMP application will use actual sub-basin rainfall estimates and sub-basin flash flood guidance. The WFO forecaster will have information mapped directly to real watersheds, based on sound hydrologic principles, to determine flood potential in flood-prone sub-basins. Although the change requires changes in applications as currently developed and implemented, the change supports the national plan to provide better support to the flash flood program administered at the WFOs. 9.0 DEVELOPMENT RESOURCES The new threshR tool that uses IHABBS data will be developed in HRL. The development of the nationwide GIS database required to apply 2-, 5-, and 10-year USGS return period flow regression equations will be completed by a contract with Dennis Johnson at Michigan Technological University. All other tasks will be completed in HRL with assistance from the same contractor. Development at HRL and with the contractor will be concurrent with completion about March 2000. 10.0 IMPLEMENTATION After development and testing has been completed at HRL, threshR will be installed at each RFC. The RFCs will run threshR to derive gridded threshold runoff values for their entire forecast areas. Initially, and for uniformity across the nation, Method 4 will be used. (Later, as data becomes available to support the other methods, threshR may be re-run.) As the runoff values are derived, they may be loaded into the Flash Flood Guidance System (FFGS) to replace the initial runoff values currently being used. An RFC may desire to derive the new runoffs for their entire service area before loading them (up to 40 minutes) into the FFGS. After gridded flash flood guidance products are defined (about a one minute task in FFGS), the RFC can begin issuing gridded flash flood guidance products. (It is expected that AWIPS communications will be available to deliver the gridded guidance products to the WFOs.) OH has recently purchased Unix versions of ArcView and Arc Spatial Analyst for RFCs that do not already have this software to implement the new threshR tool.