SIFN Stream Network Classification Project

From SIFN Wiki
Jump to: navigation, search

Contents

Introduction

Welcome to the Southern Instream Flow Network's (SIFN) River Classification Project page. SIFN is a program of the Southeast Aquatic Resources Partnership (SARP). The southeastern United States is a global hotspot of aquatic biodiversity, including the greatest number of aquatic species of federal and state concern of any region in the U.S. Instream flows--including the full range of naturally varying flows of rivers and streams--are critical to the maintenance of these rich aquatic ecosystems. To support the development of state-of-the-science instream flow standards, 60 regional aquatic resource managers and national experts in instream flows (also referred to as environmental flows)met in Nashville, TN in December 2009 for the second annual SIFN Workshop to share information and prioritize research needs. The SIFN Resarch Agenda, which outlines five priority research topics, was a product of that meeting. Development of a regional river classification system was one of the five priority research topics identified in the SIFN Research Agenda.

Why classify?

Classification is a basic step in formulating an understanding of the natural world. Fundamentally, classification is a way of simplifying and imposing conceptual order on a complex, infinitely variable universe. The basic goal of classification is to group together similar objects—whether they be things (e.g., individual organisms or species), systems (e.g., habitat units or ecosystems), or processes (e.g., flooding or biogeochemical cycling)—and to distinguish them from other groups of objects. The fundamental assumption of any classification is that objects in the same class are more similar to each other than they are to objects in other classes, and hence objects in the same class are expected to behave similarly in some fashion. In the context of river/stream networks in general and ecological flows in particular, the objectives of classification are several:

  1. To simplify communication; a classification can serve as a shorthand description that conveys a particular set of characteristics that applies to objects in a given class.
  2. To facilitate generalization of results and extrapolation of information from locations where data exist (e.g., gauged or sampled streams) to locations where no data exist (e.g., ungauged or unsampled streams).
  3. To facilitate biological/ecological monitoring and research design.
  4. To improve statistical analyses by minimizing within-class variation while maximizing among-class variation.
  5. To identify river systems expected to respond similarly to flow alteration.

In their recent paper on the ecological limits of hydrologic alteration (ELOHA), Poff et al. (2010) provide the following rationale for river classification in the context of developing regional environmental flow standards:

River classification serves two important purposes in the ELOHA framework:
  1. By assigning rivers or river segments to a particular type, relationships between ecological metrics and flow alteration can be developed for an entire river type based on data obtained from a limited set of rivers of that type within the region (Arthington et al. 2006; Poff et al. 2006). For each river type there is a range of natural hydrologic variation that regulates characteristic ecological processes and habitat characteristics (Lytle & Poff 2004; Arthington et al. 2006), and that represents the baseline or reference condition against which ecological responses to alteration are measured across multiple river segments falling along a gradient of hydrologic alteration.
  2. Combining the regional hydrologic modelling with a river typology facilitates efficient biological monitoring and research design. Specifically, it is possible to strategically place monitoring sites throughout a region to capture the range of ecological responses across a gradient of hydrologic alteration for different river types. This is particularly valuable in regions with sparse pre-existing biological data or where monitoring and research resources are limited.

The 2010 SIFN Resarch Agenda provides the following rationale and objectives for developing a regional river classification system:

The use of limited ecological data can be extended with the assumption that ecosystems with similar streamflow attributes and geomorphic characteristics respond similarly to flow alteration. A regional river classification system would allow states to supplement their own limited data for flow-ecology relationships with information from other states. To support instream flow standards, a classification system for southern aquatic ecosystems is needed that is based initially on ecologically-relevant streamflow characteristics. It should support subclassifications based on other factors that influence how biota respond to hydrologic alteration, such as water temperature and channel form and materials. Ideally such a river classification system could be integrated with similar classification efforts in the northeast and the national data being developed under the National Fish Habitat Action Plan.

Why not just use covariates?

Where biological response variables of interest have a continuous functional relationship to environmental parameters, modeling those parameters as covariates is generally the most desirable approach. However, in many cases biological metrics show nonlinear or threshold responses to environmental parameters, or the environmental parameters themselves exhibit abrupt or categorical rather than gradational changes. For example, the transition from gravel-bed to sand-bed rivers is typically rather abrupt and is usually associated with abrupt changes in benthic fauna. Thus, a classification that distinguishes rivers on the basis of dominant substrate type (e.g., bedrock, gravel, sand, fine sediments) may be more useful for predicting biological response than a continuous measure such as median particle size (D50). As another example, while stream temperature varies continuously across both space and time, many organisms have fixed temperature range tolerances. Hence temperature classes (e.g., cold-, cool-, and warm-water rivers), which can easily be extrapolated at landscape scales, may be more useful predictors of expected fish assemblages than continuous temperature data, at least at broad spatial scales. A thoughtfully designed classification can serve as a stratification or blocking variable that can increase statistical power in data analyses and improve sampling efficiency in the design of field-based research or monitoring programs.

Classification Approaches

There are many different types of approaches to classification and many different ways one could go about constructing a typology of classification approaches. One useful distinction is whether a classification scheme is hierarchical (having multiple levels, with each class in the first n-1 levels having one or more subclasses, where n is the number of levels) or nonhierarchical (just one set of classes with no subclasses). Examples of nonhierarchical classification systems include the channel habitat unit classification of Bisson et al. (1982) and Grant et al. (1990) and the reach-scale channel morphology classification of Montgomery and Buffington (1997). (Montgomery and Buffington actually described a two-level hierarchy of channel reaches nested within valley segments, but while the channel reach classification has been widely adopted the valley segment tier has generally been ignored.) Hawkins et al. (1993) described a three-level hierarchical adaptation of the channel unit classification of Bisson et al. (1982). Rosgen’s (1994, 1996) stream classification is another example of a hierarchical classification scheme, with nine morphologically defined primary stream classes and up to six types per class based on channel substrate, for a total of 42 major stream types.

Another useful distinction for geographic classifications is whether the classification applies at a single spatial scale or at multiple scales. The examples mentioned in the preceding paragraph are all essentially single-scale classifications—the channel unit scale (~ 1-10 channel widths in length) in the case of Grant et al. (1990), channel subunit (~0.1-1 channel widths) to channel unit scale (not clearly distinguished) in the case of Bisson et al. (1982) and Hawkins et al. (1993), and the reach scale (~ 10-1000 channel widths) in the case of Montgomery and Buffington (1997) and Rosgen (1994, 1996). Frissell et al. (1986) described a spatially hierarchical stream habitat classification approach, but did not present a fully developed classification scheme. A more fully developed example is provided by Higgins et al. (2005), who describe a spatially hierarchical multiscale framework for classifying aquatic ecosystems (river/stream segments and their surrounding subcatchments) and provide an example application for the Willamette River basin in Oregon that has four spatial scales (large rivers, medium rivers, small rivers, and headwaters/creeks) with 1, 4, 8, and 14 classes, respectively.

River/stream classification systems can also be linear (i.e., channel- or network-based) or areal (i.e., watershed-based). Local scale classification approaches (e.g., the channel reach and habitat unit classification schemes discussed above) are typically channel-centric. Multiscale hierarchical approaches such as that of Higgins et al. (2005) are more likely to be watershed-based, recognizing that the importance of large-scale landscape influences becomes increasingly important at larger spatial scales. In both cases, however, the focus is on the river channel and adjacent riparian areas (including floodplains where present).

Classification vs. spatial partitioning

There is an important distinction to be made between geographically dependent vs. geographically independent classification schemes (sensu Detenbeck et al. 2000). Geographically dependent classifications, such as ecoregions (Omernik 1987), are geographically defined and spatially partition a landscape or river network into units such that each unit or category often contains only a single region. Geographically independent classification schemes partition the landscape or channel network into classes of units based on similarities in one or more characteristics without regard to geographic location, such that units within each class are not necessarily geographically contiguous but are dispersed across the landscape. Some geographically dependent classifications, such as ecoregions, have been used extensively as a stratification or categorical variable in lotic aquatic studies (e.g., USEPA 2000, 2006; Yuan et al. 2008). However, these are best described as a spatially hierarchical geographic partitioning scheme rather than a true classification scheme.

River and Basin Characteristics Used for Classification

A wide variety of characteristics have been used to classify rivers and streams, including size, geomorphology and channel form, hydrologic flow regime, biological characteristics, and water quality and chemistry, among others. Some examples are given below to illustrate the diversity of approaches.

Stream size

  • Strahler stream order (Strahler 1957)
  • Wadeable vs. nonwadeable (Stoddard et al. 2005) – operational definition, useful for field-based sampling design
  • Drainage area – e.g., headwaters/creeks, small rivers, medium rivers, large rivers (Higgins et al. 2005)

Geomorphology and channel form

Water quality and chemistry

  • Water temperature – e.g., cold-water, cool-water, and warm-water streams
  • Water chemistry – e.g., Merovich et al. (2007) stream classification based on water chemistry impacts from acid mine drainage

Hydrologic flow regime Because the SIFN river classification effort is motivated by a desire to better manage instream flows (aka environmental flows), hydrologic flow regime is necessarily a key factor to consider in developing a regional river classification scheme. Many hydrologic classifications have been proposed, based on both natural and altered flow regimes. Most share a common conceptual framework and analytical approach, with generally minor variations, along the lines of the following examples.

Based on 78 long-term streamflow records across the continental U.S., developed 11 summary statistics to characterize flow intermittency, flood frequency, flood predictability, and overall flow predictability. Identified 9 stream types and discussed implications for lotic community structure.
Presents a conceptual/theoretical framework for assessing environmental flow needs based on a consensus view of an international group of leading river scientists. Outlines an approach to develop a regional classification for river segments based on ‘ecologically relevant’ flow characteristics postulated to be associated with differing ecological characteristics, and the development and testing of hypotheses about ecological responses to flow alterations.
  • USGS Hydroecological Integrity Assessment Process (HIP)
http://www.fort.usgs.gov/Resources/Research_Briefs/HIP.asp
Software and methodology (referenced by Poff et al. 2010) to compute 171 hydrologic indices, identify 10 nonredundant primary indices to characterize 5 major flow regime components (magnitude, frequency, duration, timing, rate of change), and develop a stream classification specific to a state or large region.

Landscape characteristics (GIS-based approaches) In recent years, GIS-based approaches to classifying rivers and watersheds based on physical landscape and stream characteristics such as basin area, channel or valley slope, discharge, water temperature, geology or geomorphology, etc. have become increasingly popular. Examples include:

Hierarchical classification of river segments based on climate at the broadest spatial scale and on topography, geology, and land-cover at progressively finer scales.
Classified 43,931 small (~200 km2) watersheds or subbasins covering the entire U.S. into 20 noncontiguous HLRs based on topographic, geologic, and climatic data.
Developed abiotic classification of 10 river types in New South Wales, Australia based on data from 322 reference sites in combination with GIS analysis. Data used included topography, river size, river substrate, turbidity and water chemistry. Independent classifications based on fish and macroinvertebrate assemblages were also developed.

Combined landscape and biological characteristics

Presents a 4-level spatially hierarchical approach that combines information on species distribution and drainage basin boundaries to define regions at broad scales (“ecological drainage units” nested within “aquatic zoogeographic units” and GIS-based analysis of physical landscape characteristics (topography, hydrography, hydrologic regime, geology, etc.) at finer scales.
Combines physical attributes (catchment area and modeled mean July water temperature) with sampling-based fish assemblage data to classify river valley segments.

Methods of Determining Class Membership

Ad-Hoc

These are primarily descriptive utilitarian classification schemes. They may be purely subjective/qualitative, or they may be partially or wholly based on quantitative data with classes assigned based on arbitrary cutoff values. Quantitative criteria are generally preferred where possible to reduce subjectivity.
  • Subjective (including “expert opinion” approaches)
Examples include ecoregions (Omernik 1987) and stream habitat classifications (e.g., Bisson et al 1982; Hawkins et al. 1993)
  • Quantitative (quantitative criteria with arbitrary cutoff values)
Prime examples are Strahler stream order and the channel reach classification system of Rosgen (1994, 1996)

Conceptual/Theoretical

Conceptual/theoretical classification schemes may still involve a considerable degree of subjectivity, but differ from ad-hoc classifications in that they have a conceptual or theoretical basis. That is, class membership is hypothesized to reflect differences in system behavior or function based on theoretical considerations. The channel reach classification of Montgomery and Buffington (1997), Process Domains (Montgomery 1999), and the River Environment Classification (REC; Snelder and Biggs 2002) are examples of conceptual/theoretical classifications.

Statistical/Algorithmic

Statistical classifications are developed using a host of statistical and/or numerical techniques of varying degrees of sophistication, including multivariate regression, principal components analysis (PCA), Cannonical Correspondence Analysis (CCA), various clustering algorithms, classification and regression trees (CART), nonmetric multidimensional scaling (NMDS), and artificial neural networks (ANN). Hydrologic flow regime classifications (e.g., Poff et al. 2010; Snelder et al. 2009) and GIS-based landscape classifications (e.g., Wolock et al. 2004; Brenden et al. 2008; Besaw et al. 2009; Turak and Koop 2008) are often developed using these approaches.

River Classification Examples

River Environment Classification (REC) (Snelder and Biggs 2002; Snelder et al. 2004)

The REC is based on the assumption that climate, topography, geology, and land cover are dominant causes of spatial variation in physical and biological characteristics. Snelder et al. (2004) present an example application of the REC in New Zealand that comprises a hierarchical classification at 4 spatial scales:

1. Climate 103 – 105 km2 5 classes
2. Topography 102 – 103 km2 3 classes
3. Geology 10 – 100 km2 4 classes
4. Land cover 1 – 10 km2 3 classes

Classes were defined objectively on basis of above factors only, not on spatial proximity. That is, the REC is geographically independent (sensu Detenbeck et al. 2000). The number of classes & cutoff values set a priori (i.e., not statistically determined). Performance was assessed using macroinvertebrate sampling data. Authors report that this approach provided higher classification strength than other a priori classifications (ecoregions, climate regions), but still relatively low. The REC offers the advantages of being conceptually simple, objective, and straight-forward to apply.

Higgins et al. (2005) Approach

Higgins et al. (2005) provide an example of an explicitly hierarchical, catchment-based, approach with 4 classification levels at decreasing spatial scales:

  1. Aquatic zoogeographic units (104 – 105 km2)
    Expert opinion classification based on species presence/absence data
  2. Ecological drainage units (103 – 104 km2)
    Expert opinion classification based on species distributions plus physical & climate info, preexisting classifications (e.g., ecoregions)
  3. Aquatic ecological systems (102 – 103+ km2)
    Classification based on geospatial data including elevation, gradient, stream size, geology, inferred hydrologic regime. Can be defined at multiple spatial scales—e.g., 4 scales defined in Willamette River Ecological Drainage Unity (OR)
  4. Macrohabitats (river valley segments from NHD-Plus, typ. 1-10 km long)

This approach is conceptually simple and appealing, but classifications at larger scales (aquatic zoogeographic unit and ecological drainage unit) are subjective due to their reliance on expert opinion.

ELOHA Framework (Poff et al. 2010)

The Ecological Limis of Hydrologic Alteration (ELOHA) approach (Poff et al. 2010) utilizes statistical classification based on streamflow metrics computed from 'baseline' hydrographs. Baseline hydrographs are computed from gaging station data for locations and time periods where there exists minimal anthropogenic hydrologic alteration due to dams, flow diversions, urbanization, etc. ELOHA assumes that the flow regime is the dominant control on biological condition/response. However, it acknowledges that geomorphic sub-classification may be important because geomorphic setting can mediate effects of flow alteration. Key ecologically relevant aspects of flow regime in the ELOHA framework include flow magnitude, frequency, duration, timing, and rate of change. Several software tools exist to generate and select metrics for flow regime characterization (e.g., IHA, HAT). However, no single generally accepted method exists to select flow metrics or generate hydrologic classes. For more on ELOHA, see the ELOHA Toolbox.

Because the hydrologic flow regime classification in the ELOHA framework relies on gaging station records (i.e., point data), the ELOHA approach calls for hydrologic modelling to compute hydrographs at ungaged locations in order to apply the flow regime classification to the entire drainage network. Several models for accomplishing this task are briefly discussed or mentioned by Poff et al. (2010). Another alternative (not mentioned by Poff et al.) is to develop a statistical model to predict flow regime class based on landscape attributes and climate data. The predictive model could then be used to extrapolate flow regime classes from gaging station locations to the entire drainage network.

Kennard et al. (2010) show how this approach might work in their continental-scale application of the ELOHA approach to Australia, in which they used Bayesian mixture modelling to identify 12 hydrologic classes based on 120 flow metrics (subsequently reduced to 12 key metrics). They then used geospatial data (topography, geology and soils, areal cover by forest and grasses, and climate) to predict flow regime class at gaging station locations using classification and regression trees (CART). Kennard et al. did not take the next step of using the CART model to predict hydrologic classes throughout the stream network, but this would be a logical next step. Given that their best CART model correctly classified only 62% of stream gages in the "correct" class determined from the hydrologic classification, however, there would be significant uncertainty in such a spatially extrapoloated classification.

References

Arthington A.H., Bunn S.E., Poff N.L. & Naiman R.J. (2006) The challenge of providing environmental flow rules to sustain river ecosystems. Ecological Applications, 16, 1311-1318.

Besaw L.E., Rizzo D.M., Kline M., Underwood K.L., Doris J.J., Morrissey L.A. & Pelletier K. (2009) Stream classification using hierarchical artificial neural networks: A fluvial hazard management tool. Journal of Hydrology, 373, 34-43.

Bisson P.A., Nielsen J.L., Palmason R.A. & Grove L.E. (1982) A system of naming habitat types in small streams, with examples of habitat utilization by salmonids during low streamflow. In: Proceedings of a Symposium on Acquisition and Utilization of Aquatic Habitat Inventory Information, pp. 62-73.

Brenden T.O., Wang L. & Seelbach P.W. (2008) A river valley segment classification of Michigan streams based on fish and physical attributes. Transactions of the American Fisheries Society, 137, 1621-1636.

Detenbeck N.E., Batterman S.L., Brady V.J., Brazner J.C., Snarski V.M., Taylor D.L., Thompson J.A. & Arthur J.W. (2000) A test of watershed classification systems for ecological risk assessment. Environmental Toxicology and Chemistry, 19, 1174-1181.

Frissell C.A., Liss W.J., Warren C.E. & Hurley M.D. (1986) A hierarchical framework for stream habitat classification: Viewing streams in a watershed context. Environmental Management, 10, 199-214.

Grant G.E. & Swanson F.J. (1995) Morphology and processes of valley floors in mountain streams, Western Cascades, Oregon. In: Natural and Anthropogenic Influences in Fluvial Geomorphology (Eds J.E. Costa, A.J. Miller, K.W. Potter & P.R. Wilcock), pp. 83-101. Geophysical Monograph Series, Vol. 89. American Geophysical Union, Washington, D.C.

Grant G.E., Swanson F.J. & Wolman M.G. (1990) Pattern and origin of stepped-bed morphology in high-gradient streams, Western Cascades, Oregon. Geological Society of America Bulletin, 102, 340-352.

Hawkins C.P., Kershner J.L., Bisson P.A., Bryant M.D., Decker L.M., Gregory S.V., McCullough D.A., Overton C.K., Reeves G.H., Steedman R.J. & Young M.K. (1993) A Hierarchical approach to classifying stream habitat features. Fisheries, 18, 3-12.

Higgins J.V., Bryer M.T., Khoury M.L. & Fitzhugh T.W. (2005) A freshwater classification approach for biodiversity conservation planning. Conservation Biology, 19, 432-445.

Kennard M.J., Pusey B.J., Olden J.D., Mackay S.J., Stein J.L. & Marsh N. (2010) Classification of natural flow regimes in Australia to support environmental flow management. Freshwater Biology, 55, 171-193.

Lytle D.A. & Poff N.L. (2004) Adaptation to natural flow regimes. Trends in Ecology & Evolution, 19, 94-100.

Merovich G.T.,Jr, Stiles J.M., Petty J.T., Ziemkiewicz P.F. & Fulton J.B. (2007) Water chemistry-based classification of streams and implications for restoring mined Appalachian watersheds. Environmental Toxicology and Chemistry, 26, 1361-1369.

Montgomery D.R. & Buffington J.M. (1997) Channel-reach morphology in mountain drainage basins. Geological Society of America Bulletin, 109, 596-611.

Omernik J.M. (1987) Ecoregions of the conterminous United States. Annals of the Association of American Geographers, 77, 118-125.

Poff N.L. & Ward J.V. (1989) Implications of streamflow variability and predictability for lotic community structure: A regional analysis of streamflow patterns. Canadian Journal of Fisheries and Aquatic Sciences, 46, 1805-1818.

Poff N.L., Olden J.D., Pepin D.M. & Bledsoe B.P. (2006) Placing global stream flow variability in geographic and geomorphic contexts. River Research and Applications, 22, 149-166.

Poff N.L., Richter B.D., Arthington A.H., Bunn S.E., Naiman R.J., Kendy E., Acreman M., Apse C., Bledsoe B.P., Freeman M.C., Henriksen J., Jacobson R.B., Kennen J.G., Merritt D.M., O'Keeffe J.H., Olden J.D., Rogers K., Tharme R.E. & Warner A. (2010) The ecological limits of hydrologic alteration (ELOHA): a new framework for developing regional environmental flow standards. Freshwater Biology, 55, 147-170.

Rosgen D.L. (1996) Applied River Morphology. Wildland Hydrology, Pagosa Springs, Colo.

Rosgen D.L. (1994) A classification of natural rivers. Catena, 22, 199-169.

Snelder T.H. & Biggs B.J.F. (2002) Multiscale river environment classification for water resources management. Journal of the American Water Resources Association, 38, 1225-1239.

Snelder T.H., Cattaneo F., Suren A.M. & Biggs B.J.F. (2004) Is the River Environment Classification an improved landscape-scale classification of rivers? Journal of the North American Benthological Society, 23, 580-598.

Snelder T.H., Lamouroux N., Leathwick J.R., Pella H., Sauquet E. & Shankar U. (2009) Predictive mapping of the natural flow regimes of France. Journal of Hydrology, 373, 57-67.

Stoddard J.L., Peck D.V., Olsen A.R., Larsen D.P., Van Sickle J., Hawkins C.P., Hughes R.M., Whittier T.R., Lomnicky G., Herlihy A.T., Kaufmann P.R., Peterson S.A., Ringold P.L., Paulsen S.G. & Blair R. (2005) Environmental Monitoring and Assessment Program (EMAP) Western Streams and Rivers Statistical Summary. EPA 620/R-05/006, .

Strahler A.N. (1957) Quantitative analysis of watershed geomorphology. Transactions - American Geophysical Union, 38, 913-920.

Turak E. & Koop K. (2008) Multi-attribute ecological river typology for assessing ecological condition and conservation planning. Hydrobiologia, 603, 83-104.

USEPA (2006) Wadeable Streams Assessment: A Collaborative Assessment of the Nation's streams. EPA 841-B-06-002, U.S. Environmental Protection Agency, Office of Research and Development and Office of Water, Washington, DC.

USEPA (2000) Mid-Atlantic Highlands Streams Assessment. EPA/903/R-00/015, U.S. Environmental Protection Agency Region 3, Philadelphia, PA.

Wolock D.M., Winter T.C. & McMahon G. (2004) Delineation and evaluation of hydrologic-landscape regions in the United States using geographic information system tools and multivariate statistical analyses. Environmental management, 34, S71-S88.

Yuan L.L., Hawkins C.P. & Sickle J.V. (2008) Effects of regionalization decisions on an O/E index for the US national assessment. Journal of the North American Benthological Society, 27, 892-905.


Environmental Flows/Stream Classification Bibliography - Literature


River Classification Committee:

  • John Faustini, USFWS, Lead
  • Chris Konrad, TNC/USGS, Co-lead
  • Mary Davis, TNC/SARP





























Retrieved from "http://sifn.bse.vt.edu/sifnwiki/index.php/SIFN_Stream_Network_Classification_Project"