The coding for all of the base functionality in AV-ThreshR  is complete.   This page is part of draft documentation that has been prepared and is available on the Internet (http://hsp.nws.noaa.gov/oh/hrl/gis/threshrhome.html).

With the AV-ThreshR software, each RFC will receive a database of topographic data sets and flood frequency data layers.  To enable ThreshR Scripts work in all RFCs and make different types of computations depending on the location of interest, strict ThreshR database specifications must be met.  A contractor has collected the raw input data sets required to derive the initial flow direction grid used for subbasin delineation, collected  and reformatted nearly all of the data layers required to automatically compute flood frequency values in any of the 50 United States and Puerto Rico, and created data tables containing regression equation coefficients.  Some pre-processing procedures, cleaning, and checking must be done on this database before data packets can be delivered to the RFCs.  This includes pre-processing data delivered by the contractor to create a 400-m flow direction grid (approximately a 1 day task per RFC) and cleaning up the flood frequency data layers and tables (also roughly a 1 day task per RFC).  Flow direction pre-processing has been completed for 3 RFCs (ABRFC, MBRFC, and MARFC) while cleaning the flood frequency database has been completed at ABRFC and MBRFC.   Preliminary threshold runoff results for ABRFC and MBRFC are discussed below.

5.2  Future Plans

Future tasks include the following (1) complete database setup for each of the RFCs, (2) beta test software and implement suggested improvements, (3) improve the database and make minor software modifications to accomodate database improvements, and (4) evaluate results.  Planned database improvements and some methods for evaluating threshold runoff results are discussed here.

Database Improvements

AV-ThreshR will make deriving gridded threshold runoff estimates easier; however, the results will need to be critically evaluated as the software gets used.  Even without this evaluation, there are three areas where it is known that the current database is limited and can be improved.

(1)  Updates to the regression equations in USGS WRIR 94-4002 are expected from more recent work at USGS offices.  These updates should be incorporated into the AV-ThreshR database for the RFCs that are affected.  In most cases, no changes to the software will be required.  A related improvement can be made to the data table that stores regression equation coefficients and exponents.  Columns should be added to this table to reflect the range of values which are appropriate for use in the regression equations.

(2)  The vertical and horizontal resolution of the DEM in the current database is inadequate to resolve basins less than 20 mi2 and delineated areas of any size may be inaccurate in areas of low relief.  The most straightforward approach to alleviate this problem is to use a better DEM to derive subbasin boundaries and topographic parameters.  Fortunately, a 1 arc-second digital elevation model (National Elevation Dataset (NED)) is now available with national coverage.  The drawbacks associated with using the NED data are disk storage and computation time.  If the DEM and all of the required hydrologic derivatives are considered, the size of a 1 arc-second national database will be on the order of 1.2 terabytes.  A project to delineate flash flood basins nationwide (down to 2 mi2 in size) using this database has been undertaken at the National Severe Storms Lab (NSSL) with funding from NWS.  NSSL is cooperating with the EROS data center to facilitate this task.  The basins resulting from this project will be used in the Flash-Flood Monitoring and Prediction program (FFMP).  HRL is working with NSSL to facilitate pre-computing the topographic parameters required for threshold runoff calculations in each of these subbasins.  This will substantially increase the accuracy of area and length calculations over those that could be derived from the existing database.  Although there are still some parts of the country where the NED data will not accurately resolve basin boundaries, these small basin boudaries should be better than anything else currently available.  By having certain parameters pre-processed, much of the increased processing time associated with using higher resolution gridded data will be eliminated.

(3)  Gridded threshold runoff calculations require unit hydrograph estimates for many small, ungaged basins throughout the country.  Using a synthetic unit graph method is the only practical way to derive these estimates.  Carpenter et al. (1999) explore the use of the Snyder and the geomorphological unit graph methods.  Both of these methods required local knowledge or regional regression equations to implement.  In both the new AV-ThreshR software and the older GRASS-based ThreshR, the user is required to specify unit graph parameters for the location of interest.  It would be helpful to compile a database of parameters or equations that would make it easier for an end user to specify unit graph characteristics on a regional basis.

Evaluating Results

Threshold runoff values are used along with a rainfall-runoff model to derive the Flash-Flood Guidance (FFG) values that are used by forecasters at Weather Forecast Offices (WFO's).  The usefulness of FFG values can be evaluated based on reports of actual floods; however, since threshold runoff values are only one component of FFG, it may be difficult to assess the validity of the threshold runoff values themselves using this approach.

A validation approach for threshold runoff values would be to evaluate the two inputs, bankfull flow and unit hydrograph peak, independently.  At locations where flow data is available, synthetic unit hydrographs could be compared with manually derived unit graphs (as done by Carpenter et. al. (1999)) or evaluated using modeling techniques.  The Q2 values used to approximate flood flow should be compared with flood stage at locations where rating curves are known.  This will provide some guidance as to whether a 2 year return period is the most appropriate duration for a given region.