WATER AVAILABILITY MODELING IN THE STATE OF TEXAS   David Mason CE 394 K.2 - Surface Water Hydrology University of Texas at Austin

Project Objectives
Data Collection and Development
        Digital Elevation Model
        RF3 Coverage
        Water Right and Diversion Locations
        Primary Control Points
        Precipitation and CN Grid
        Flow Length Grid
Analysis of Data
        CRWR-Prepro Procedures
        Control Point Location Review
        Building the Database
Results
Reasons for Discrepancies
Conclusion
Future Work

The objective of this project was to show how GIS and GIS utilities could be used to create a spatial water rights database for a few of the major river basins in Texas.  This work has been the main focus of my research for the Center for Research and Water Resources (CRWR) at the University of Texas at Austin.  The basins which I focused on in this project included the Nueces, San Antonio and Guadalupe river basins.

• Total upstream drainage area
• Mean annual precipitation across the drainage area
• Mean SCS curve number across the drainage area
• Downstream flowlength to basin outlet

DATA COLLECTION and DEVELOPMENT:

The first step in the construction of the database was the data collection and development.  This required obtaining all of the necessary maps and information on the basin and projecting this information into the proper coordinates.  The data needed for this project included: 1) Digital Elevation Model (DEM) 2) RF3 Coverage 3) Water Right and Diversion Locations 4) Primary Control Points 5) Precipitation and CN Grids 6) Flow Length Grid.  The following steps will describe the process used for the development of each of the data sets.
 

• Digital Elevation Model: The files used in the development of the DEM were initially downloaded from the USGS website. These files are 1° by 1° maps with 90 m grid cells.  The maps were then merged into one giant DEM, and clipped with the basin boundary using ArcInfo.  The following figure is the end result of this process for the Guadalupe basin:

• RF3 Coverage: The RF3 coverage is a file containing all the rivers, streams and water features in the basin. This file was downloaded from the EPA Basins website. The necessary files were acquired by selecting them from a clickable .gif of all the HUC8 regions in Texas.  Before being used though, much of the clutter and unnecessary features in the files had to be edited out.  The goal is to create a single line stream network for the entire basin.  The steps in this editting process are listed below:
 
• Using the map query option in Arcview, only the R, S and T stream types were chosen in the attribute table.   These letters denote regular reach, start reach and terminal reach, respectively.  The new shapefile created was now a much simpler network, but had gaps where the lake feature once existed.  Previously, these gaps were filled one by one using the editing features in Arcview (as you can imagine, this was very time consuming). For this network though, a file of transport paths through the open water features (created by the USGS and EPA) was merged with the newly created shapefile. This served to fill a majority of the gaps while the rest were completed by hand.
 
• The next step was to check the stream network against a coverage of water right locations.  Where there were water rights not on a stream, a digital map called a DRG (discussed below) was used as a backdrop to digitize a stream for that point.
 
Below is a sample of the the final stream network for the Guadalupe basin.

Figure 3: Stream network of the Guadalupe basin

• Water Right and Diversion Locations:  The water right locations are the main source of control points in the basin for which parameters must be calculated.  Initially, these water right points were selected out of a large database by the basin number.  Once the file was created, the data was sent off to TNRCC for review.  Along with this review, TNRCC also added additional diversion locations that may be associated with each of the water rights.  This can result in quite a substantial difference in the size of the file upon its return.  For example, a file of 488 points for the Guadalupe basin was sent to TNRCC and was sent back with 788 points!!

• Primary Control Points: The primary control points are the second set of points to be used for delineating watersheds.  These points are acquired from whichever contractor is currently working on the basin.  For all three cases, this contractor was HDR, inc.  These points are mainly USGS stream gage locations which are used to distrubute naturalized flows throughout the basin.  The location of these points (lat/long) were received in an excel spreadsheet, along with copies of maps showing their exact locations on the streams.  Using Arc/Info, a control point coverage was created and projected on to the stream network.

• Precipitation and Curve Number Grids:  The precipitation grid was downloaded from the Oregon State Prism website while the CN grid was acquired from CRWR.  The first step in the processing was to clip the large grids to the applicable basin boundary.  Once completely, the files were then resampled in Arc/Info so that the grid would be exactly coincident with that of the DEM.  Finally, a grid calculation was performed in Arcview which produced a file with the average CN or precipitation over the upstream drainage area of each cell.

• Flow Length Grid: The flow length grid is a file containing the flow length of each grid cell to the outlet of the basin.  This grid was produced using the Hydrologic extension in Arcview.  The purpose of these values is for the contractor on the project to perform calculations of the channel losses due to evaporation.

Once the data had been collected, the next step was to prepare the data for use in constructing the database. This step included running the DEM and stream network through the CRWR-prepro process, performing a final review of the control point locations, and running a script to compile all the data into one coverage.
 

• CRWR-Prepro Procedures: The step in preparing the data for final use was to process the stream network and the DEM in CRWR-Prepro. This process consisted of "burning" the stream network into the DEM, filling the sinks of the DEM, and computing the flow direction and flow accumulation grids.  The followings steps show how this was done:

• Burning into the DEM: Essentially, the "burning in" process entails raising all the cells of the DEM not on the stream network by a specified amount. This creates a trench which "traps" the water until it flows out of the outlet at the bottom of the stream. Performing this process ensures an exact match between the vector and raster stream network.

• Filling the sinks of the DEM: Although most of the DEM data is accurate, there are still some sinks that form in the terrain that must be corrected. The fill sinks command in CRWR-prepro raises the elevations in these small pits to match that of the surrounding terrain. If this step is not performed, problems could arise in the flow paths that Prepro calculates in the future steps.
 
• Computing the flow direction grid: The next grid derived from the DEM was the Flow Direction Grid. This grid stores a number in each cell which corresponds to the direction of the steepest descent (according to the eight direction pour point model). From this grid, a basic flow path was defined for the entire basin.

• Computing the flow accumulation grid: The flow accumulation grid uses the flow direction grid to calculate the number of cells upstream of each individual cell. Therefore, as you move downstream along the network, the value in the cells on the stream will increase, hitting a maximum at the outlet of the basin.

• Control Point Location Review: Before moving on to the calculation of the parameters, a review must be made of all the control points to ensure their proper location.  If not, accurate values will not be found.  Therefore, this step becomes one of the most tedious and time consuming processes of the entire database construction.  Basically, each point must be checked closely to make sure it is on the apppropriate cell in the flow accumulation grid.  Currently, a script is being written which could automate this process, but for now it is done by hand.  Each point was checked individually and moved on to the flow accumulation cell, and a unique control point ID number (provided by TNRCC) was also added.  An example of this problem is shown below.  If  this point was not moved onto the stream, it would have returned a flow accumulation value on the order of 10 cells, rather than its true value of close to 5000 cells!!  This would make a huge difference in its drainage area.

• Building the Database: Once all the points were in their place, it was then time to begin building the database.  In the past, this was done by using more of the CRWR-prepro procedures, such as converting the control points to an outlet grid, and then merging this grid with the rest of the grids.  I began using these procedures and immediately ran into the problem of the ID field being too long.  Therefore, Arcview would not let me convert the points to a grid.  My solution to that problem was to add another column of unique ID numbers (i.e. 1,2,3,,,etc).  I again tried the convert to grid command, and was successful in attaining a grid.  However, as I examined the grid more closely, I discovered that the program would arbitrary select one ID number in the case where two points fell on the same grid cell.  This problem was solved by trying a completely different approach, which was a script written by Patrice Melancon and editted for our use by Brad Hudgens.  This script has a looping function that takes all the cell values of any number of grids "below" the control point and attributes them to a table by the ID number.  The following is a sample of this table:

After much time and effort, the final product was an Arcview point coverage with fields for control point ID, flow accumulation cell count, average precipitation and average curve number across the drainage area, and flow length to the outlet.  Since the file only gave the amount of cells upstream in the flow accumulation grid, an equivalent drainage area in square miles had to be calculated.  This was done under the edit field command in Arcview by multiplying the number of cells by the cell area, and converting square meters to square miles.  Also, the flow length was calculated in meters and had to be converted to miles.  The completed file was then evaluated for quality control.  The computed drainage areas were compared to established USGS values for each stream gage location in the basin.  The following is an analysis of the results found in each of the three basins studied:

• Nueces - The results from the Nueces database were very encouraging.  As you can see from the spreadsheet below, the average error in drainage area was only 1.36%!  Most of the points had % errors much below that, falling almost exactly in line with the USGS values.  Also, not a single point was above a 3% error, which is well within accepted accuracy levels.

• San Antonio - The results from the San Antonio database were somewhat worse than that of the Nueces.  This time, the average error in drainage area was about 3.44%.  Even though this value is a little higher than the 1.36% found in the Nueces, it is skewed upward by the large error in only a couple points.  Otherwise, most of the values are pretty close to USGS figures.  Upon further study of the errant points, it was found that one of the streams erroneously crossed the basin boundary and captured some extra area.  With this problem rectified, the results were much better.  Although, there still seems to be a problem on the lower end of the basin.  This will be discussed more later.

• Guadalupe - The results from the Guadalupe database were found to be quite poor.  Although the error was only slightly higher on average as compared to the others (3.57%), we see that many of the points are consistantly higher than those of the USGS.  Also, upon closer analysis done by both myself and HDR, it was found that even the drainage areas that are close to USGS values mask large discrepancies found in smaller areas within the basin.  Again, just as with the San Antonio, the error seems to increase as you move towards the outlet of the basin.  A possible reason for this may be that the basins outlet to the coast, where terrain relief is less severe.  This fact has been proven to cause similar errors in other studies done at CRWR.

There are several possible reasons for the errors in the drainage areas.  The first problem could be found in the resolution of the DEM.  In this project, 90m data cells were used for the analysis.  Recently, 30m DEM data was made available by the USGS.  This type data has been proven to produce more accurate results in drainage area calculations.  The higher resolution picks up the more subtle changes in the terrain which helps in representing the land surface.  This would be especially effective in the lower end of the Guadalupe and San Antonio were accurate representation of the terrain becomes very important.

A second reason could be that not enough of the smaller tributaries were represented in the river reach file.  Through some of our quality control methods, we have found that going back and digitizing some of the smaller streams around the control points help to produce better results.  However, this is another very time consuming process and we're hoping that the 30m DEM alone will take care of this.

A final reason for the disprepancy may simply be that some of the USGS figures are wrong!  Many of those drainage areas were delineated by hand from old USGS quadrangles, which my not be representative of the current land surface.  From talking to the folks at TNRCC, there have been specific examples where USGS drainage areas crossed directly over burms and such that would obviously prevent the flow of water.  Therefore, it may be that the CRWR values are actually the accurate figures and the USGS values are the ones that need to be corrected!

The procedure set forth in this project was used to produce a spatial water rights database for three river basins (Nueces, San Antonio and Guadalupe) in south central Texas. For each water right and stream gage location, the resulting database was a compilation of the total upstream drainage area, the mean annual precipitation and the average SCS curve number in the drainage area, as well as the flow length from the point to the outlet of the basin.

The eventual use of such data will be in the WRAP (Water Rights Analysis Program) model, which is currently being developed by Texas A&M for the WAM (Water Availability Modeling) project. Since the water rights are linked in the database, they can then be analyzed in the correct sequence by the model to obtain the naturalized flow at each water right from the corresponding flow at the appropriate stream gage. This research was initiated by TNRCC, which must develop new river basin simulation models in order to determine available water in accordance with Senate Bill 1.

Aside from performing the same analysis for the rest of the basins in Texas, much of the future work associated with this project will focus on the use of the 30m DEM data sets.  Currently, the files are much too large for use. Smaller cells mean many more cells to cover the same area.  An effort has been made to perform some of the same procedures outlined above, but so far, the computers at CRWR have not been able to handle the files.  Hopefully, for my sake, this problem will be resolved soon! J