Christopher N. Dunn, PE
Cameron T. Ackerman
James Doan, PE
Tom Evans
The U.S. Army Corps of Engineers’ Sacramento District and Hydrologic Engineering Center (HEC) are developing hydrologic models of the Sacramento and San Joaquin River basins. Because of the size of the study and a tight deadline, HEC incorporated GIS technology to expedite the modeling effort. GeoHMS is used to develop the hydrologic elements and their connectivity for HEC’s Hydrologic Modeling System (HEC-HMS). Additionally, GeoHMS calculates the river and watershed characteristics used for estimation of parameters of ungaged basins. This paper will discuss the benefits of using GeoHMS technology and the ease of its application.
Introduction
Due to several large and damaging flood events in the state of California during the 1980’s and 1990’s, most notably the January 1997 event, a Flood Emergency Action Team was assembled. The team was to develop recommendations on how the impacts of future flood events could be reduced and ecosystems restored. One of their recommendations was to perform a detailed comprehensive study of the entire Sacramento and San Joaquin River systems. The U.S. House of Representatives funded the study through the 1998 Energy and Water Appropriations Bill and directed the U. S. Army Corps of Engineers to develop comprehensive plans for flood control and to develop hydrologic and hydraulic models of the river systems.
To meet the goals of the Comprehensive Study, the Sacramento District of the Corps of Engineers had to develop hydrologic models for both the Sacramento and San Joaquin River basins. The watersheds for these two river systems comprise nearly 60,000 square-miles, bound California’s central valley floor and receive much of the runoff from the Sierra Nevada mountain range, Figure 1.0. Developing hydrologic models for 60,000 square miles with extremely variable terrain and to do it within a 10-month time period was a formidable task. The Sacramento District understood that the hydrologic models would be useful not only for planning purposes but, would also assist them in making basin wide flood operation decisions on a real-time basis during flood episodes. The District contacted the Hydrologic Engineering Center (HEC) to assist them with the development of the hydrologic models.
Within the Corps, HEC may have been uniquely qualified to conduct such a large study. HEC not only had their existing Hydrologic Modeling System program, HEC-HMS, but at the beginning of the study was actively working with Esri under a cooperative research and development agreement to develop the ArcView extension HEC-GeoHMS. HEC used GeoHMS to develop many of the physical parameters required by HEC-HMS and to develop the hydrologic models in an expeditious manner. HEC has developed HEC-HMS models for all of the main tributaries to the Sacramento and San Joaquin systems which includes over 32,000 square miles.
Figure 1. Sacramento and San Joaquin River Basins
HEC-GeoHMS
The purpose of HEC-GeoHMS is to facilitate the development of hydrologic models through GIS. GeoHMS uses functions from Spatial Analyst to analyze terrain models and develop hydrologic modeling inputs for HEC-HMS. The physical characteristics of the subbasins and reaches, which can be used to estimate rainfall-runoff parameters necessary for HEC-HMS, are directly calculated from the terrain data by GeoHMS. In short, GeoHMS significantly reduces the effort and time required to develop the physical characteristics of the subbasins that are necessary for the HEC-HMS models.
In GeoHMS, two views and a number of customized menus, tools, and buttons are introduced. The two views are the Main View and the Project View. In the Main View, the digital elevation model (DEM) is used as input for pre-processing steps such as flow direction and accumulation. The flow direction grid, developed by GeoHMS, and the DEM are used to create the grid-cell parameter file for the ModClark rainfall-runoff transformation. The flow direction and DEM grids are also used to create a new grid having the Standard Hydrologic Grid (SHG) parameters. The SHG that was developed at HEC has parameters that result in an equal-area projection (Evans, 1998). The resultant grid aggregates the DEM cells to a user specified grid-cell size and calculates an averaged travel distance from the flow direction grid.
In the Project View, the user can divide the basin into headwater subbasins, local subbasins with routing reaches, and other HMS elements. After the basin is appropriately delineated, GeoHMS estimates the watershed and stream characteristics needed by the HEC-HMS models and stores this information in theme attribute tables. Finally, GeoHMS develops the HMS schematic for the HMS model. (Doan, 2000)
HEC determined that in order to complete the hydrologic modeling in the short amount of time available, it would have to use GeoHMS to develop the inputs necessary for the HEC-HMS models. The rest of this paper will discuss how the HEC staff developed the hydrologic models using GeoHMS.
DEM Development
To begin the GeoHMS process, a terrain model must be developed. GeoHMS relies almost exclusively on topographic information to perform its tasks. GeoHMS procedures are not used to create the DEM. Instead, terrain tools found in ArcInfo and ArcView were used to assemble and correct the digital elevation models. A short description of the DEM development process for the Sacramento and San Joaquin Study follows. (McPherson, 2000)
The Sacramento/San Joaquin Comprehensive Study began before the National Elevation Dataset (NED) was available from the USGS, so the terrain data for the study was assembled from 1:24,000 scale DEMs. These DEMs were downloaded from the USGS at edcftp.cr.usgs.gov:/pub/data/DEM/7.5min. Using a series of scripts and ArcInfo macros written for the purpose, HEC staff processed roughly 2,900 individual DEM quadrangles from their SDTS file format into ArcInfo grids. Before the grids could be merged into a single terrain model, several inconsistencies had to be corrected. These included unit conversions, vertical and horizontal datum changes, and a change in the horizontal coordinate system conversion from Universal Transverse Mercator (UTM) in two zones into a single coordinate system for horizontal positions. Many of these data conversion steps would be unnecessary today using the NED.
About 85% of the DEMs had elevation values in meters, while the remainder were in feet. In all cases, the DEMs used integer values, so the meter grids had roughly one-third the vertical resolution of the foot grids. To combine the two unit systems, the grids were converted to centimeters to maintain integer format and avoid a reduction in accuracy.
Horizontal positions of all the DEMs were in UTM coordinates, with about half the quadrangles falling in UTM zone 10 and half in UTM zone 11. All the zone 11 DEMs and the vast majority of the zone 10 DEMs were referenced to the NAD27 horizontal datum, and a small number of the zone 10 grids were referenced to NAD83. For use in the study, all the grids were resampled to 30-meter postings in an Albers Equal-Area projection similar to the one used by California’s Teale Data Center for statewide maps.
Once the DEMs were assembled into a single consistent form, a preliminary drainage analysis was performed to test their suitability for use in hydrologic modeling. This analysis revealed a number of flaws requiring further corrections. For example, the DEMs representing Folsom Lake include a corner where four quadrangles meet, see Figure 2.0. In three of the quadrangles, the elevations were recorded in meters and the elevation of the lake was given as 142 meters, but in the fourth quadrangle the elevations were given in feet. After conversion, the elevation in that part of the lake was 142.04 meters. Because the lake’s outlet was located in the portion of the terrain model with the higher apparent elevation, any drainage analysis from this DEM would place the reservoir’s outlet at the wrong location. This error and about a dozen similar problems around reservoirs were corrected using the ArcInfo GRID editing tools.
Figure 2. DEMs at Folsom Lake
After the master DEM was developed and corrected, eleven basin coverages were cut from the master DEM and were later provided to the individual modeling teams assembled at HEC. Each of the eleven smaller DEMs included one to four actual river basins which were later subdivided during the GeoHMS process. It is important to stress that in order for GeoHMS to provide reasonable estimates of the physical parameters, the DEMs must be assembled carefully and conscientiously.
Data Acquisition
Before GeoHMS was used to delineate the basins and subbasins, a massive data collection effort was undertaken at HEC. Data for the two events to be modeled and for the period-of-record was collected and assembled. The two events used for hydrologic model calibration were the March 1995 event and the December 1996/January 1997 event. Other events such as the 1983 and 1986 events were also desired, but the data for those events was not as readily available or complete.
Over 50 agencies were contacted for data and over 4500 time series were processed for this study. For delineation, optimization, and calibration purposes, approximately 200 flow gages and nearly 75 reservoirs were used. Flow (hourly and daily), stage and reservoir data were collected, along with other data, to assist the modelers in the delineation of their models. The District requested that the basins be delineated at all flow gages with records for one or both of the events and it was agreed that the basins should be divided at all reservoirs with at least 10,000 ac-ft of storage.
Because of the shear number of gages and desired subbasin delineation points, it was important to identify the geographic location of gaging stations. By using the capabilities within ArcView, the hourly and daily flow gages were organized by type and provided to the modelers in separate ArcView coverages. The modelers could then import the ArcView data layers into GeoHMS and the subbasins could be accurately delineated. An example of a coverage displaying the gages is shown in Figure 3.0. To have the location of the gages spatially referenced was critical to the speed and accuracy of which the basin delineations could be performed.
Figure 3. Flow Gage Locations
Basin Delineation and Parameter Estimation
With the terrain data processing and data acquisition complete, GeoHMS could now be used to estimate the physical model components necessary for HEC-HMS. Using USGS hydrologic unit maps as templates, the eleven smaller DEMs were extracted from the master DEM as described above. Each of the eleven basins were then assigned to one of eleven two-person teams, composed of engineers from HEC and the Sacramento District. Using ArcView along with the Spatial Analyst and GeoHMS extensions, these teams performed the more detailed drainage analyses, dividing the basins into subbasins, and defining stream reaches and junctions. As discussed in the Data Acquisition section, where possible, subbasin outlets were placed at the locations of stream gages, so that flows recorded at those gages during the two events could be used to calibrate the models. Many of the pertinent steps required to develop the HEC-HMS models by using GeoHMS are described below.
General Watershed Delineation
In the Main View of GeoHMS, each of the modelers performed a number of tasks to prepare their basins for the GeoHMS processing. The pre-processing performed in GeoHMS creates a depression-less DEM which ensures that positive drainage will occur. Next, flow direction and flow accumulation grids are calculated based on the flow path of steepest decent. The drainage paths delineated from the DEMs were compared with existing hardcopy maps. In some cases, these comparisons revealed flaws in the DEMs, which had to be corrected before continuing onto model development.
For the Comprehensive Study, it became clear early that modeling consistency was critical. With eleven teams performing the analyses, the detail and direction that each team took had to be consistent. Therefore, because the study area was so large, HEC elected to target the minimum size of a subbasin to 50 square miles or greater. In reality, there were numerous occasions when smaller subbasins were used in the modeling. The Stream Definition step initiated the delineation process. Many of the modelers elected to perform the original delineation of their basins by using the percent of total area threshold option. Typically, the modelers chose one percent for their area threshold which meant that if their basin was 1000 square miles, the smallest subbasin automatically delineated based on the stream patterns would be 10 square miles. While 10 square miles is less than the 50 square mile threshold, the basins could be merged later within the GeoHMS process. In general, the automatic delineation process is useful, however, each team provided a critical review of the automatic delineations and either added additional delineations at gages or other hydrologically important points or combined subbasins when appropriate. The initial junction locations for all the models were then automatically developed during the Stream Segmentation step. Next, the various grids just created are converted into vector representations with polygons and lines. The pre-processing steps took only a few minutes for each of the teams. The automated basin delineation process within GeoHMS is very quick. In addition, the automated delineations are readily revised in subsequent steps of the GeoHMS process.
The final step of GeoHMS’s Main View is to identify the individual project or basin models. The eleven DEMs were subdivided so that each drainage basin could be modeled separately. For example, in Figure 4.0, one MainView exists but three project or basin models have been identified: Stony Creek, Cache Creek and Putah Creek. The project models are defined by their outlet. GeoHMS allows the user to select the outlet of the project model and then based on the results of the preprocessing steps, automatically cuts out the individual basin models. The physical characteristics of the individual basin models are automatically saved to an attributes table thus saving many hours of labor intensive manual measurements.
Figure 4. Basin Delineation
Subbasin Delineation and Processing
Now that the individual basin models are identified the true hydrologic modeling effort could begin. As mentioned before, the subbasins originally developed by GeoHMS in the preprocessing steps may or may not have been appropriate. In GeoHMS’s Project View, the original subbasins and hence the river reaches and junctions can now be more appropriately identified. Subbasins may be split or combined which obviously means that the river reaches and junctions for each of the new subbasins will also be redefined.
Each of the eleven modeling groups were given general guidelines on how to delineate their basin models with GeoHMS. These guidelines included:
Figure 5. Stream Profile and Subbasin Delineation
Armed with the guidelines, hard copy detailed maps of their basins, and ArcView layers of existing rivers, lakes, flow and stage gages, interstates and state highways, places (including regions, cities, and other points of interests), the modeling teams subdivided or combined the original subbasins using the tools within GeoHMS. One such tool within GeoHMS is the River Profile tool. It plots the streambed profile for visual inspection. If the streambed slope changed significantly at a certain location, the subbasin may have been subdivided at that location. The river profile for a selected subbasin and the entire delineated Putah Creek basin are displayed in Figure 5.0. What made the delineation process so expedient is that once the delineations were finalized, many physical parameters for the subbasins were automatically calculated. The division of the subbasins could have been done easily by hand but the computation of all of the physical characteristics, if done manually, would have taken an enormous amount of time.
The physical parameters developed by GeoHMS include the area of the subbasins, river length, river slope, subbasin centroid location and elevation, longest flow path for each subbasin, and the length along the stream path from the centroid to the subbasin outlet. Each of these parameters was then saved to one of a number of theme attributes tables. The drainage areas were automatically transferred to the HMS models via the HMSfile.basin file when the models were developed in a later step of GeoHMS. Other physical parameters were copied to a number of Excel spreadsheets where they were used to assist with the optimization models, the regression analysis, and the calibration process.
GeoHMS creates two other files that are directly importable into HMS. These files are: the mapfile.map, and the ModClark.mod file. The mapfile.map is the HMS schematic (or background map for HMS) and the ModClark.mod file includes the reach and travel lengths necessary for the ModClark rainfall/runoff transformation. The HMS schematic for Putah Creek is shown in Figure 6.0. GeoHMS allows the user to name the reaches and the basins for HMS in various attribute tables. GeoHMS uses the names to label the schematic. GeoHMS also allows the user to convert the map units which are predominantly in Metric units to automatically display in English units.
Figure 6. HMS Schematic for Putah Creek, CA
At this point, the physical HMS model has been developed and the hydrologic parameters such as precipitation, baseflow and losses are to be entered. Without GeoHMS, it is doubtful that HEC would have been able to develop the HMS models as quickly or accurately.
Optimization
Due to the fact that hourly flow gages did not exist at every subbasin outlet, a regression analysis was performed to consistently populate the HMS models with initial parameter estimates. In order to develop the physical parameters necessary for a regression analysis, HMS optimization models were developed at all of the hourly flow gages for the unregulated headwater streams. GeoHMS assisted with the optimization process by developing some of the physical parameters necessary for optimization and by allowing the modeling teams to easily construct additional HMS sub-models so that additional optimization models could be readily developed. For example, if a detailed HMS model included three subbasins contributing to an hourly gage and it was determined that the gage was an acceptable optimization location, GeoHMS allowed the modeling groups to define a new outlet point and then cut out a new HMS model at the headwater gage. GeoHMS then combined the three subbasins which could now be modeled as a separate HMS optimized model. GeoHMS allowed the modeling groups to develop as many unit hydrograph and loss rate parameter optimization models as possible within their basins. Along with the new basin model, the physical parameters for the new subbasin were estimated. Some of the physical parameters developed by GeoHMS that were used by the optimization process were the longest flow paths, slopes, stream lengths, and drainage areas. Since the detailed HMS models were only going to include subbasins of 500 square miles or less, only subbasins of less than 500 square miles were used for optimization. Once the physical parameters were available for the gaged basin models, the model parameters were estimated by optimization functions in HEC-HMS, which automatically adjusted the model parameters to find a best-fit solution for computed and observed runoff hydrographs.
Regional Regression Analysis
Based on the results of the optimization of the gage basins, a regression analysis was then performed. The purpose of the regression analysis was to provide a way for the modeling groups to populate the ungaged subbasins with physical parameters without having each parameter manually measured.
The regional regression analysis was used to estimate the time of concentration (Tc) and storage coefficient (R) for the Mod-Clark Unit Hydrograph rainfall/runoff transform method. GeoHMS calculates the following characteristics for each of the subbasins:
Each of these characteristics along with the basin factor, the product of the two flow path lengths divided by the square root of the subbasin slope (BF), were evaluated for correlation with variations in the Tc for the optimization models. The characteristics with predictive value were then used to estimate Tc for each subbasin in the detailed HMS models. HEC decided that the values of R for each subbasin would be estimated by assuming that the relation R/(R + Tc) is constant for hydrologically similar areas.
The results of the regression analysis demonstrated that two predictive equations for Tc and two relationships for R could be used for the entire study area. For all of the rivers north of the Merced River except for the Pit and McCloud the equation
Tc = 0.68(LLCA/S1/2)0.46 was used. For all other rivers, including the Merced, the equation used was Tc = 1.67(LLCA/S1/2)0.29. Similarly, the relationship for R/(R + Tc) was initially set to either 0.6 or 0.8 depending on the particular river. During the calibration process, the relationship for R/(R + Tc) may have been changed for a particular river so that the observed and computed hydrographs matched more appropriately. Again, GeoHMS was critical in developing the parameters for the regression analysis.
Routing Parameters
Initial estimates of routing reach travel times were estimated, in part, from a regression of field data. Field estimates of the Manning’s "n" values and the top and bottom widths and depths of the streams were taken at a number of key locations throughout the Sacramento System. From the estimates of the cross-sectional measurements, a hydraulic radius was computed. Using the tools within GeoHMS, HEC was able to determine a representative slope at those locations and compute the drainage area above these points. A relationship between the drainage area and the hydraulic radius was then estimated through a regression analysis. The purpose of the relationship was so that each modeler could obtain a hydraulic radius, knowing the drainage area at any given point along a stream. The drainage area is easily determined by GeoHMS.
Next, the field estimated Manning’s n values were compared to the Manning’s n values that are calculated by Jarrett’s equation for steep streams (Jarrett, 1984). Jarrett’s equation requires the modeler to know the slope and hydraulic radius. Once the Manning’s "n" values were calculated, the velocity along any stream could be computed. Placing the values into a spreadsheet and knowing the reach length, already developed by GeoHMS, the travel times for the reach lengths could be calculated. For the comprehensive study, the travel time in a reach was set equal to the Muskingum "K" value for the routing reach. The Muskingum "x" was set equal to 0.4 for the steep mountainous streams. The Muskingum routing method was prescribed by the District.
Calibration
Using the equations from the regression analysis to generate the values for the HMS models, the calibration process could begin. The physical parameters developed by GeoHMS were instrumental in the calibration process. The drainage areas were used directly while the routing reach lengths were used to help compute the routing times for each of the routing reaches. It was critical that the reach lengths be correct so that the timing of the hydrograph peaks would be accurate. At times GeoHMS’s area tool was used to estimate the area to the midpoint of a reach length so that representative hydraulic radii could be determined from the drainage area vs. hydraulic radius plot. Again, without GeoHMS, the development of these parameters would have been much more time consuming and the project would not have been completed in an efficient and consistent manner. Calibrations were compared to the observed hourly, or at times, daily hydrographs whenever available, see Figure 7.0. If daily gages were used, instantaneous peaks were retrieved from the USGS Web site to help the modelers approximate the actual hydrograph.
Figure 7. Hydrograph from Calibration of Cache Creek at Rumsey, CA for the March 1995 Event
Conclusion
The hydrologic modeling effort for the comprehensive study of the Sacramento and San Joaquin Rivers is HEC’s first large-scale attempt to develop new hydrologic models from GIS data sets. Because of the size of the project, approximately 60,000 square miles, and the need to work quickly, a 10-month schedule, to meet the needs of some of the other models used in the Comprehensive Study, the hydrologic models had to be developed efficiently and with consistent methodologies across a very large geographic region. The ability to define subbasins and routing reaches interactively based on terrain models allowed the hydrologic engineers to work quickly, and the ability to compare their intermediate results with existing maps increased their confidence in the validity of their model components. From the viewing and manipulation of the data, to the development of the physical parameters, to the optimization and regression analyses, it is clear that GeoHMS and its ability to calculate basin characteristics directly from terrain models allowed HEC to complete this regional analysis in a timely manner. It is also clear that GeoHMS will be used for future regional studies. GeoHMS with ArcView and Spatial Analyst were instrumental to completing this project. Finally, this project has provided a valuable testing ground for the GeoHMS extension, allowing its developers to discover and repair flaws that were not apparent in small sample projects, but which became evident on a project this large.
References
Evans, Thomas, and Pabst, Art (1998). "Standard Hydrologic Grid in Spatial Hydrologic Modeling," U.S. Army Corps of Engineers, Hydrologic Engineering Center, Davis, CA.
Doan, James (2000). "HEC-GeoHMS Tools for Hydrologic Modeling," Proceedings from the 20th Esri International User Conference, San Diego, CA. U.S. Army Corps of Engineers, Hydrologic Engineering Center, Davis, CA.
McPherson, Mathew, and Henneman, Heather (2000). "DEM Processing for Hydrologic Modeling Studies," Proceedings from the 20th Esri International User Conference, San Diego, CA. U.S. Army Corps of Engineers, Hydrologic Engineering Center, Davis, CA.
Jarrett, Robert D. (1984). "Hydraulics of High-Gradient Streams," U.S. Geological Survey, Lakewood, CO, Journal of Hydraulic Engineering, Vol.110, No.11, November 1984.
Author Information
Christopher N. Dunn, P.E.
Senior Hydraulic Engineer
U.S. Army Corps of Engineers
Hydrologic Engineering Center
609 Second Street
Davis, CA 95616
Telephone: 530-756-1104
Fax: 530-756-8250
christopher.n.dunn@usace.army.mil
Cameron T. Ackerman
Hydraulic Engineer
U.S. Army Corps of Engineers
Hydrologic Engineering Center
609 Second Street
Davis, CA 95616
Telephone: 530-756-1104
Fax: 530-756-8250
cameron.t.ackerman@usace.army.mil
Thomas A. Evans
Senior Hydraulic Engineer
U.S. Army Corps of Engineers
Hydrologic Engineering Center
609 Second Street
Davis, CA 95616
Telephone: 530-756-1104
Fax: 530-756-8250
thomas.a.evans@usace.army.mil
James H. Doan, P.E.
Hydraulic Engineer
U.S. Army Corps of Engineers
Hydrologic Engineering Center
609 Second Street
Davis, CA 95616
Telephone: 530-756-1104
Fax: 530-756-8250
james.h.doan@usace.army.mil