Flood Forecasting for the Buffalo Bayou Using CRWR-PrePro and HEC-HMS

Seth Ahrens

Written for CE394K (GIS in Water Resources) taught by Dr. D. R. Maidment

15 May 1998

Table of Contents


The Buffalo Bayou, which forms west of Houston, TX, passes through downtown Houston, and exits into the Houston Ship Channel, is responsible for draining approximately three hundred sixty square miles of flat southeastern Texas terrain. This flatness, especially when combined with the urban sprawl of Houston, leaves the bayou and the surrounding area particularly susceptible to severe, widespread flooding events. For example, a strong thunderstorm system struck the area from 16 to 18 October 1994, dropping ten to twenty inches of rain and causing massive flooding around the city. Since then, this storm has served as a benchmark example of the damage heavy rains can incur upon the Houston metropolitan area. In addition, it has drawn attention to the need for a flood forecasting system that will help officials make decisions when severe weather strikes. Hence, the overall goal of this project involves establishing such a forecasting system for the region along the Buffalo Bayou.

This project will be completed through the use of a surface-water analysis package linked to a geographic information system (GIS). The Hydrologic Engineering Center (HEC), located in Davis, California, recently updated its widely-accepted surface-water model HEC-1 to the Hydrologic Modeling System (HMS), which will be used to perform the runoff modeling aspect of this project. A new GIS tool recently developed at the University of Texas at Austin will link the GIS with the HMS. This program, named CRWR-PrePro, at this point possesses the capabilities to link to HMS. In the future, CRWR-PrePro will have the ability to link to other modeling packages as well. For linking with HMS, CRWR-PrePro uses as its primary inputs a digital elevation model (DEM) and a curve-number coverage and builds in a GIS environment an input basin model for HMS. More information about CRWR-PrePro is available via a paper written for the 1998 ESRI User's Conference by Dr. Francisco Olivera as well as documentation by Ferdi Hellweger, one of the original PrePro authors, and a web page which contains a class exercise involving the use of CRWR-PrePro.

Funding for this project is from the HEC through the United States Army Corps of Engineers (USACE) office in Galveston, TX. Many thanks to USACE and the HEC, especially James Doan and Tom Evans.

Data Requirements

The first step on the way to completing this project involved the collection of a wide variety of data sources for creation of a regional GIS database for the system. This database covers the following coordinates and includes the Houston metropolitan area:

Northwest corner 30 deg, 15 min N, 96 deg W
Northeast corner 30 deg, 15 min N, 95 deg W
Southeast corner 29 deg, 30 min N, 95 deg W
Southwest corner 29 deg, 30 min N, 96 deg W

To help the reader more easily determine what area of Texas this rectangle represents, consider the following image.

The project falls into an area located in the triangle formed by Highway 290, Interstate 10, and the Harris County border line, which passes through Katy, TX and heads north.

In all cases, each data type covers completely the rectangle enclosed by the coordinates in the above table. This was done so that any studies on the Houston area in the future could use this complete database as a starting point for future analyses.

Several different types of data were collected to complete this database. Briefly, the different types of data collected were thirty-meter digital elevation models (DEM), NEXRAD-based rainfall data, STATSGO soils data, digital line graph data (DLG), stream coverages, and Anderson land use/land cover data. As part of this project, a CD-ROM containing all of these types in their raw forms was created. For a more complete description of these data, please refer to the README file documenting the contents of this CD-ROM.

Data Development

Since the CD-ROM of the GIS database contains only raw data, most data types required some level of preprocessing before they could be viewed in a GIS. For example, the DEM data originally were acquired in a format which more or less was a text-based grid. To convert this into a usable format, the spatial analyst in ArcView was invoked. From here the DEM grids could simply be imported into the ArcView environment via the "Import Data Source" command under the "File" heading when in an ArcView view. Note that the "Import file type" is "USGS DEM". Though the DEM is readable after itís being imported, it has not yet been converted to a grid format. To do this, one merely needs to save this theme as a grid using the "Convert to grid" command from the "Theme" menu. This process was carried out for all forty-eight DEMís in the study area. Then, using the "merge" command in the GRID portion of Arc/Info, the forty-eight DEMís were combined into a single DEM covering the entire area. Unfortunately, this merging revealed several gaps in the DEM data mostly made up of strips of individual cells. An example of one of these gaps is in the image below.

In most cases, this gap was six cells wide. To fix this problem, it was suggested that the data be transferred to a spreadsheet so that each of the six-cell gaps could be filled via manual interpolation. To accomplish this, the merged DEM was split into seventeen vertical strips such that each strip contained less than 256 columns, which is the widest spreadsheet Excel could hold. Each strip was then fixed in Excel. These strips were then merged again using Arc/Info. The resulting DEM now contains no gaps in the data.

A second type of data source for this project is NEXRAD-based rainfall data. This data arrived compressed into many different spreadsheet-readable text files. At this point, instantaneous rainfall intensities are available for ten thousand points across the study area (one-kilometer resolution) approximately every six minutes. Future development of this data will involve establishing regular rainfall increments of five minutes over the entire storm. For inclusion in the HMS rainfall/runoff model, these data need to be converted into a Hydrologic Rainfall Analysis Project (HRAP) grid, which is the standard projection for rainfall as per the National Weather Service. According to the manual from the HEC describing this system, the grid is a square-celled map based on a Polar Stereographic map projection. Parameters for this projection are as follows:

Units Meters
Datum Sphere with radius = 6371.2 km
Standard Parallel 60 0' 00"
Central Meridian 105 0' 00"

Each cell in the grid is 4.7625 km on each side and the gridís Y-axis is aligned parallel to the central meridian. Finally, the grid is registered so that the north pole lies exactly 400 cells in the X direction and 1600 cells in the positive Y direction from the grid origin. At this point, the rainfall data is located in a grid with rectangular sides, each approximately one kilometer in length. Tom Evans of the HEC has graciously offered a program that will be able to take this grid and convert it into a HRAP grid for use in this project. Future work will perform this data preparation.

As mentioned in the introduction, the storm being modeled dropped ten to twenty inches of rain over the Houston area from 16-18 October 1994. To get an idea of the magnitude of this event, which was estimated as being in the range of a twenty-five- to one-hundred-year event, consider the following data from a USGS stream flow gauge in the Houston area over October and November 1994.

An additional type of data for this project is STATSGO soils data. It was hoped that STATSGO data would be available, but this higher-resolution type of soils data has not yet been catalogued for the entire state of Texas. Conveniently, the only preprocessing required for these data was projection conversion.

The next type of data tracked down for this project were the DLG data. Three different coverages are typically available when working with a DLG--political boundaries, hydrographical features (rivers, pipelines, levees, etc), and transportation routes (roads and railroads). The DLG data will be helpful in this project for they will provide an additional method of determining the stream coverage besides through flow direction/flow accumulation calculations and the EPAís RF1 stream coverage. This is important because it may be difficult to determine the locations of the streams throughout the study area via the DEMís due to the areaís flatness. In addition, the location of the Addicks and Barker reservoirs within the system can be pinpointed with these data.

To preprocess the DLG data for this project, an AML needed to be written. Fortuitously, someone else had already written this AML, and only minor changes were required to edit the AML so it could be used with the coverages for this project. The image below includes an example of the hydrography coverage over part of the DEM around the Addicks reservoir, which is demarcated by the "J"-shaped black lines. Note how at least at this level, the streams are located in a reasonable position in relation to the DEM.

As mentioned above, EPA RF1 data will be used to aid in determining the precise location of the stream network throughout the DEM. As with the STATSGO data set, this information was available in a format which only required projection conversion as part of the preprocessing.

The final data set retrieved for this project was the Anderson land use/land cover classification scheme. These files were originally available in ArcView Export (*.e00) format so they had to be converted first using the Import71 tool that is part of the ESRI ArcView package. After implementation of this utility, the data were able to be read into the ArcView environment. For more information about working with land use/land cover files, please see this reference page.

Where to from Here?

Obviously, a substantial amount of work for this project remains uncompleted at this point. Thus, significant quantities of time will be devoted to completing this project this summer before the 1 September deadline. Below is an outline of the steps required for project completion.

  1. Prepare the rainfall data set for input into the HEC-HMS program. As mentioned earlier, Evansís program will be very useful for completing this setup. However, other software will be required. The HEC has a program called GridParm which takes the output from Evansís program and overlays it on top of the delineated watersheds in the study area. The output from GridParm is a parameter file which HMS requires for its ModClark modeling option.
  2. Since a large part of the functionality of GridParm is already present in the HEC-PrePro code, the code will be altered so that the unique functions of GridParm are assimilated into HEC-PrePro.
  3. Delineate the watersheds and determine the steam network over the study area using the fill pits/flow direction/flow accumulation functions in ArcViewís Hydrologic Modeling extension. This will allow for comparison among this network, the DLG stream network, and the RF1 stream network. If it is determined that the calculated stream network is insufficient, the network from one of the other data sets will have to be burned into the DEM. It is possible that a new type of data source called LIDAR data will be available for some or all of the study area over the summer. LIDAR data provides an extremely detailed view of an areaís topology; therefore, it could prove to be invaluable in determining the stream network in this flat study area.
  4. Run HEC-PrePro to create the basin input model for the HMS.
  5. Using the ModClark rainfall data set created by GridParm, calibrate the HMS model with the 16-18 October 1994 storm. Compare the results with USGS fifteen-minute stream flow data.
  6. Adjust the model as necessary to reflect better the actual flow values from the USGS gauges.
  7. Submit a report to the USACE documenting the successes and limits of the model.

Back to my home page.