USGS -- SMIG --
Surface-water quality and flow Modeling Interest Group

Two-dimensional Habitat Modeling in the Yellowstone / Upper Missouri River System

by Terry Waddle, Ken Bovee, and Zachary Bowen

USGS, Biological Resources Division
River Systems Management Section
Midcontinent Ecological Science Center
4512 McMurry Ave.
Fort Collins, CO 80525
Internet: terry_waddle@usgs.gov, ken_bovee@usgs.gov, zack_bowen@usgs.gov
Phone: (970) 226-9386, (970) 226-9320, (970) 226-9218

Editor's note:
This paper was presented at the North American Lake Management Society (NALMS) meeting in Houston, TX (December 2-4, 1997).

Contents

Abstract

This study is being conducted to provide the aquatic biology component of a decision support system being developed by the U.S. Bureau of Reclamation. In an attempt to capture the habitat needs of Great Plains fish communities we are looking beyond previous habitat modeling methods. Traditional habitat modeling approaches have relied on one-dimensional hydraulic models and lumped compositional habitat metrics to describe aquatic habitat. A broader range of habitat descriptors is available when both composition and configuration of habitats is considered. Habitat metrics that consider both composition and configuration can be adapted from terrestrial biology. These metrics are most conveniently accessed with spatially explicit descriptors of the physical variables driving habitat composition. Two-dimensional hydrodynamic models have advanced to the point that they may provide the spatially explicit description of physical parameters needed to address this problem. This paper reports progress to date on applying two-dimensional hydraulic and habitat models on the Yellowstone and Missouri Rivers and uses examples from the Yellowstone River to illustrate the configurational metrics as a new tool for assessing riverine habitats.

Introduction

Water resource managers and fish and wildlife agencies need a credible and practical method of evaluating impacts of modified flow regimes on riverine fauna. Historically, decisions related to instream flow requirements have been made by quantifying the effects of flow modifications on habitat availability for a few individual species. Many scientists and policy-makers suggest that the focus of instream flow assessments should now shift toward the description of critical ecosystem functions and how they are linked to various flow-dependent habitat types, at a variety of spatial scales (Sedell et al. 1990; Schlosser 1991; Lobb and Orth 1991; Hughes and Noss 1992; Aadland 1993; Orth 1995). As the research and development community moves toward this new focus, the picture that emerges is a description of habitat niches or patches utilized by a variety of species in different community assemblages. The emphasis of instream habitat management is shifting toward the management of the patch mosaic, and the development of biological interpretations and predictive capabilities thereof.

Our organization's first foray into this field is a study of habitat for communities of riverine fishes in the Upper Missouri and Yellowstone Rivers in cooperation with the U.S. Bureau of Reclamation; Montana Department of Fish, Wildlife, and Parks; and the Montana Cooperative Fisheries Research Unit (USGS/BRD). Species of concern in the upper Missouri/Yellowstone ecosystem include the pallid sturgeon (Scaphirhychus albus), the sturgeon chub (Hybopsis gelida), sicklefin chub (Hybopsis meeki), blue sucker (Cycleptus elongatus), and paddlefish (Polyodon spathula).

Recognizing the need for a more ecosystem-oriented approach for evaluating the effects of reservoir operations on the endemic fauna of the upper Missouri and lower Yellowstone River ecosystems, the U.S. Bureau of Reclamation is developing an interactive Decision Support System (DSS) for their water operations planning and day-to-day management activities. The purpose of the DSS is to allow reservoir operators and planners to include an assessment of the effects of alternative storage/release schedules on habitat for riparian and aquatic communities among the multiple objectives they must consider in water operations planning. The primary goal of our research is to assist Reclamation in completion of the aquatic biology component of the DSS. Figure 1 shows the overall study area and location of currently identified study sites.

(fig. 1)

Figure 1. Map of the upper Missouri/Yellowstone River study area showing locations of reservoirs and two-dimensional mapping sites.

The specific goal of this study is to identify indexes of riverine habitat that are related to the ecology of endemic Northern Great Plains fish communities and provide a means of incorporating that information into the DSS. Specifically, we are examining the basic question of how aquatic community characteristics and habitat in great plains streams are related. Three possible cause-effect relations are being considered:

  1. Community attributes and health of endemic populations primarily related to habitat variability over time.

  2. Community attributes and health of endemic populations primarily related to habitat variability over space.

  3. Community and population health related to habitat variability in both time and space.
These cause-effect relations have important ramifications with respect to the operation of Reclamation reservoirs. If the first condition dominates, then day-to-day reservoir operations may be important to the health of the downstream fish communities. In contrast, if the second condition dominates, the community structure and function may be dictated more by structural characteristics of the river than by short term temporal variability of streamflow. This scenario would suggest the necessity for periodic floods to maintain channel structure and a lack of sensitivity to daily reservoir operations. Should the third condition dominate, community characteristics would be expected to be related both to day-to-day and year-to-year flow conditions. Thus both spatial and temporal variability of habitat in conjunction with the various life stage processes of the biota must be considered.

The Spatial Component: Two-Dimensional Analysis

We have several reasons for attempting to use two-dimensional modeling to provide new means of habitat representation in the Yellowstone/Missouri River system, including:
  1. Better performance in complex channels. This consideration has components of hydraulic complexity, spatial representation of patch types, and issues of connectivity between channel and floodplain habitats.

  2. Spatially explicit representation of habitat characteristics and spatial distributions.

  3. Conveniently available spatial habitat metrics.

Performance in Complex Channels

The primary benefits of two-dimensional modeling, according to LeClerc et al. (1995) are superior accuracy in hydraulic modeling and a better quantification of microhabitat area. In our experience, many applications of the traditional one-dimensional habitat modeling approach embodied in PHABSIM (Milhous, et al., 1989) avoid important habitat areas in complex channels due to large field data collection and analytical modeling resource requirements. Two-dimensional models provide a means to overcome some of those difficulties.

When using one-dimensional models, the division of flow among the side channels must be quantified empirically to determine the simulation discharge for each of the parts, as a function of total discharge. Divided channels exhibit variable backwaters through a portion of the flow range so approximately twice the data and effort are needed for calibration and simulation when compared to single channels.

In contrast, two-dimensional models incorporate momentum calculations that allow the model to divide flow around an island with minimum intervention by the analyst. Figure 2 shows a map of the islands and tributaries at one of our Yellowstone River sites (Elk Island). At our mapping flow, this site had 22 islands. Due to the complex topography, as the flow increases there will be even more islands and divided channels until the stage gets high enough to inundate them entirely. We expect this site to be challenging to calibrate with a two-dimensional hydraulic model.

(fig. 2)

Figure 2. Planimetric map of channel and tributary boundaries at the Elk Island site on the Yellowstone River.

Flood Plains and Tributaries

Flood plains and tributary confluences may be extremely important habitat types during high flows. Flood plains are commonly dissected by secondary (flood) channels and dotted with hummocks of vegetation. Tributary mouths become extensive backwaters at high flows. We have documented backwaters several kilometers long in tributaries to the Yellowstone River during high flows. Backwater habitats are thought to be extremely important refugia for fish during high flow events, because they commonly provide conditions of slower current (main channel velocities may exceed 3 m/s), warmer temperatures, and abundant food resources.

Hydraulic conditions can be simulated on flood plains with one-dimensional models. However, one of the most serious challenges of one-dimensional models, is the difficulty in connecting the hydraulics on the flood plain with discharge in the main channel. Each secondary channel may have one or more inlets somewhere along the flood plain. Unless each inlet is located and its inflows determined empirically, it is virtually impossible to tell how much water should be flowing in the flood channels at discharges above or below measured conditions. The planimetric view of the flood plain used in two-dimensional models allows us to evaluate the connectivity of flood plain habitats with the main channel.

Spatially Explicit Representation: Two-dimensional Habitat metrics

Two-dimensional analyses allow spatially explicit representation of the physical characteristics of riverine habitat. Thus it makes application of both new and borrowed habitat metrics to instream problems more accessible. Two classes of habitat metrics (compositional and configurational) are commonly used in landscape ecology (McGarigal and Marks 1995). Compositional metrics refer to the relative amounts and abundances of different patch types in a landscape, without regard to their spatial distribution. Configurational metrics account for the relative proportions, but also describe their spatial characteristics, such as shape and juxtaposition with other patch types. Figure 3 gives an example of two hypothetical portions of a stream that have the same composition, but have very different configuration. Traditional instream habitat analysis using PHABSIM generates exactly one compositional habitat metric, weighted usable area for each target organism. In contrast FRAGSTATS, a spatial analysis system designed for applications to landscape ecology, conveniently calculates numerous patch type and landscape scale metrics (McGarigal and Marks 1995). Using this tool we will explore the possible significance of the configuration of habitat types as well as their abundance.

(fig. 3)

Figure 3. Hypothetical matrices of the distribution of shallow-water habitat in two river segments. Both matrices have the same proportion of shallow-water habitat but differ in the number, distribution, and size of habitat patches.

The Temporal Element

One of the elements that distinguishes rivers from other ecotypes is that habitat distributions change rapidly and dramatically as the discharge changes (Poff and Allan, 1995). In this sense, discharge is really an element of time with respect to habitat distributions because it has temporal attributes of periodicity, seasonality, and duration. To illustrate this concept, Figure 4 shows the spatial distribution of a single patch type (shallow-slow; i.e., < 0.5 m deep and < 0.35 cm/s) at three different discharges at the Fairview site on the Yellowstone River. These maps revealed differences in the area and spatial distribution of shallow-slow habitat among the three discharges. The total area of shallow-slow habitat was almost five times larger at 141.5 cms (13.50 ha) compared to 849. 0 cms (2.71 ha; Table 1). There were three large areas of shallow-slow habitat at 141.5 cms: one area was located along the southeastern shore of the large island, the second area was associated with a large sandbar near the northeastern boundary of the site, and the third area was located in the northern end of the back channel west of the large island. At 424.5 cms the back channel was filled with water and contained large patches of shallow-slow habitat. At 849.0 cms the only relatively large patches of shallow-slow habitat were located in the mouth of a small tributary flowing into the back channel west of the large island. The number of patches increased with increasing discharge while mean patch size decreased (Table 1) indicating greater fragmentation of shallow-slow habitat with increasing discharge. The mean nearest neighbor distance between patches of shallow-slow habitat was largest at 141.5 cms (5.4 m) reflecting the influence of isolated patches of shallow-slow habitat in the back channel west of the island. At low flows, these isolated patches will probably freeze from top to bottom and become death traps during a typical Montana winter.

(fig. 4)

Figure 4. Comparison of spatial distributions of shallow-slow habitat patches (black) at three different streamflows at the Fairview site, Yellowstone River. Grey stippled areas are wet, but not shallow and slow. White areas within channel boundaries are dry.

(table 1)

From a biological perspective, it is just as important to know when and how long a certain patch type or mosaic will be present as it is to know how abundant and well-connected it is. For example, flows greater than 850 cms only occur about 8% of the time in the Yellowstone River for during any year (Figure 5). Furthermore, such high flows occur primarily during the spring and early summer. Low flows (e.g., < 140 cms) occur about 20% of the time, usually during fall and winter. Therefore, seasonality and persistence of streamflow translate directly into a temporal distribution of habitat characteristics, a phenomenon known as a habitat time series (Bovee 1982).

(fig. 5)

Figure 5. Flow duration curve for the Yellowstone River at Sidney, MT for water years 1975-1995. Boxes indicate discharges simulated for habitat modeling.

As demonstrated in Figure 4, the composition and configuration of patch types are determined by streamflow. By modeling spatial attributes over a range of discharges, it is possible to derive a relationship between streamflow and any of numerous spatial metrics. It follows, then, that a habitat time series can be developed for these spatial metrics. This is an exciting prospect for ecological research, because it offers the opportunity to examine temporal and spatial variability simultaneously. To our knowledge, this analytical capability has not been used in prior ecological studies.

Data Collection and Modeling Methods

The general study design includes twelve or more sites (the final number to be determined as the study progresses), four each upstream and downstream of Ft. Peck Dam on the Missouri R., and four on the Yellowstone R.

Relating the physical habitat characteristics in these rivers to the structure and health of ecological communities requires data describing the location, abundance and health of the species making up that community. Not only is it significant to identify the species using a portion of the river, but to locate them accurately enough that the habitat conditions they are occupying can be identified. The Montana Department of Fish, Wildlife, and Parks; and the Montana Cooperative Fisheries Research Unit are collecting the target species; noting their life stage, age and other biological attributes and logging their specific location using Global Positioning System (GPS) equipment.

Physical data requirements of the two-dimensional hydraulic models being used include a three dimensional bathymetric map of each study site, a bed material map and certain flow related boundary conditions. To develop this data we are employing standard surveying techniques to establish elevation control and standard stream gaging techniques to determine discharge. We are obtaining planform locations using GPS equipment, depth by hydroacoustic sounding and bed material analysis by postprocessing analysis of the hydroacoustic signal.

For areas above the water surface (in particular the flood plain) we are employing a variety of techniques including surveying and digitizing of areal photographs. The precision of digitized photogrammetry is considerably lower than the standard surveying data, however, by echosounding the river at high flows we are able to obtain the greatest precision in the portions of the river channel that are inundated most of the year. We believe the greatly increased cost of a high quality survey of the flood plain would not be rewarded with increased accuracy in the final habitat representation.

As part of our purchase contract, the echosounder vendor provided specialized software that allows coarse delineation of bed material size from analysis of the echo signature. Preliminary work with this analysis indicates we can separate substrate materials into categories of coarse, medium and fine. This allow us to produce a map of approximate bed material roughness height for use in the hydrodynamic model.

To encompass the range of habitat conditions occurring throughout the year, we will simulate discharges over the full range of flows from the record. Figure 5 shows the flows we will simulate at each site on the Yellowstone River. Simulated discharges will be selected for the Missouri River both below and above Ft. Peck Reservoir to similarly cover the flow range.

At the beginning of the study, an informal survey of available finite element, depth averaged two-dimensional hydrodynamic models was conducted. Three candidate models emerged for consideration: FESWMS, RMA2, and CDG2D. The first two of these models, though developed by USGS and the Army Corps of Engineers, are available commercially. The third model is under development by the University of Alberta (Ghanem et al., 1994). CDG2D was selected due to its ability to simulate a wide range of flows (water surface changes of 6 meters are experienced in the Yellowstone R.) without rebuilding the finite element mesh and to simulate subcritical to supercritical transitions. Its ongoing use in instream flow applications in Canada was an additional selection factor.

Field bathymetric data is transformed into the finite element mesh for flow simulation in a two step process. A geographic information system (GIS) is used to develop a water surface elevation site map from our elevation data and the GIS is used to subtract depths measured with the echosounder to establish a point coverage of stream bed elevations. This process produces a bathymetric map that is passed to a mesh generation utility. The mesh generator then is used to sample from the bathymetric map and produce a suitable computational mesh to be used in simulating the selected discharges.

Application of the finite element model proceeds through a standard calibration and simulation sequence. Calibration consists of adjusting bed element roughness height and, where necessary, adding mesh nodes to achieve a good fit of observed water surface elevations and spot sampled velocities. The model is then used to simulate other flow conditions.

Status of the Study

We currently have hydroacoustic and elevation data for 6 sites, ground-truthing and digitizing of the areal photos for four sites will be accomplished this year, and fish data collection by the cooperating agencies is ongoing at all sites.

Currently hydrodynamic modeling is completed at one site except for high flows where the areal photogrammetry is needed to describe the flood plain. Preliminary work has begun on the second site (Elk Island). We will incrementally progress through hydrodynamic modeling of the other sites over the next 2 years.

As the hydraulic modeling produces sufficient information to begin assessing patches, we will begin addressing selection and correlation of habitat metrics with fish data. This activity will carry well into the next year as well. To ensure compatibility with other studies in the lower portions of the Missouri River, the Army Corps of Engineers RCHARC model will also be applied.

Discussion

If they did nothing else, the ability of two-dimensional models to provide a bird's eye view of habitat structure and dynamics would justify their use over one-dimensional models in the Yellowstone/Missouri River system. The simple ability to display patch distributions and fish sampling and catch data on a spatially accurate planimetric map will undoubtedly help us formulate and refine hypotheses. Additionally, two-dimensional models will allow us to expand the number of hypotheses that we can test. Using PHABSIM output in a habitat time series might result in a dozen or two habitat metrics, most of which examine temporal variation of a single spatial measure. The number of spatio-temporal metrics that can be potentially derived from two-dimensional modeling number in the thousands. Because the number of potential hypotheses could become overwhelming, we will initially investigate relations between patch dynamics and relatively simple biological indexes (e.g., IBI (Karr et al., 1996, Huges and Noss, 1992) or critical habitats for species of concern). We hope to use the results of the present study to improve our hypotheses and methods for future research.

From a practical standpoint, it is questionable whether or not we could actually model the Yellowstone and Missouri sites with a one-dimensional hydraulic model. It is certain that we could not have covered as extensive an area. For example, the Fairview site depicted in Figure 4 is one of our least complex and smallest. It is approximately 2.5 km in length, with one island and a single tributary confluence. This site would have required at least twice the field effort to obtain calibration data for a one-dimensional model. Assuming equal proficiency, it appears that the time required to calibrate the two-dimensional model at Fairview would be about the same as a one-dimensional model. The two-dimensional hydraulic model appears to have a clear superiority for a site like Elk Island, where one-dimensional modeling would be prohibitively costly.

Two-dimensional models are not without limitations, however. There is a severe trade-off between finite element mesh size and computational demand. Depending on the formulation, large mesh finite element models can require large execution times or large amounts of computer memory (RAM) to operate successfully.

Both the model's author and our group have found CDG2D to have a practical upper limit of approximately 5000 to 6000 mesh nodes to execute on a Pentium equipped personal computer with 128 Mb of memory. Similar results have been reported for Unix workstations. Memory requirements increase exponentially, making finite element meshes greater than 10,000 nodes impractical in the small computer environment . These limitations may force us to split a site like Elk Island (Figure 2) to enable us to provide needed mesh density in important habitat areas. In particular, representation of narrow back channels is vital to capturing possible high flow refugia. When building the mesh, great care must be taken to represent those areas while also maintaining sufficient density to capture the dynamics of the main channel.

References

Aadland, L.P. 1993. Stream habitat types: their fish assemblages and relationship to flow. North American Journal of Fisheries Management 13:790-806.

Bovee, K.D. 1982. A guide to stream habitat analysis using the instream flow incremental methodology. U.S. Fish and Wildlife Service FWS/OBS-82/26. 248 pp.

Ghanem, A., P.M. Steffler, F. Hicks, and C. Katopodis. 1994. Two-dimensional finite element modeling of physical fish habitat. Proceedings of the 1st International Symposium on Habitat Hydraulics. Aug. 18-20, Trondheim, Norway.

Hughes, R.M., and R.F. Noss. 1992. Biological diversity and biological integrity: current concerns for lakes and streams. Fisheries 17:11-19.

Karr, J.R., K.D. Fausch, P.L. Angermeier, P.R. Yant, and I.J. Schlosser. 1986. Assessing biological integrity in running waters: A method and its rationale. Special Publication 5. Illinois Natural History Survey, Champaign. 28 pp.

LeClerc, M., A. Boudreault, J.A. Bechara, and G. Corfa. 1995. Two-dimensional hydrodynamic modeling: A neglected tool in the Instream Flow Incremental Methodology. Transactions of the American Fisheries Society 124(5):645-662.

Lobb, M.D. III and D.J. Orth. 1991. Habitat use by an assemblage of fish in a large warmwater stream. Transactions of the American Fisheries Society 120: 65-78.

McGarigal, K., and B.J. Marks. 1995. FRAGSTATS: Spatial patten analysis program for quantifying landscape structure. U.S. Forest Service General Technical Report PNW-GTR-351. 59 pp + appendixes.

Milhous, R.T., M.A. Updike, and D.M Schneider. 1989. Physical Habitat Simulation System Reference Manual-Version H. U.S. Fish and Wildlife Service Biological Report 89(16). v.p.

Orth, D.J. 1995. Influence du compartiment trophique dans les responses des populations de poissons aux variations artificielles de dbit. Bull. Fr. Pche Piscic. 337/338/339:317-328.

Poll, N.L. and J.D. Allan. 1995. Functional organization of stream fish assemblages in relation to hydrological variability. Ecology 76:606-627.

Schlosser, I.J. 1991. Stream fish ecology: a landscape perspective. Bioscience 41:704-712.

Sedell, J.R., G.H. Reeves, F.R. Hauer, J.A. Stanford and C.P. Hawkins. 1990. Role of refugia in recovery from disturbances: modern fragmented and disconnected river systems. Environmental Management 14: 711-724.

Back to the SMIG Features Page

button bar

SMIG Home Mailing List Features Conferences Classes Reading Room Model Archives Feedback Home | Mailing List | Features | Conferences | Classes | Reading | Model Archives | Feedback

Stewart Rounds, SMIG coordinator <sarounds@usgs.gov>
U.S. Geological Survey
http://smig.usgs.gov/SMIG/features_0398/habitat.html
Last modified Wednesday, 17-Dec-2003 14:06:58 EST
Privacy Statement · Disclaimer · FOIA · Accessibility