Introduction
To the untrained eye streams and river systems may appear to be a simple network of natural open channels. In reality, these networks are a complex system consisting of multiple parameters used to represent dimension, flow, sediment, etc. Stream classification is the process of trying to analyze these parameters and label them in a descriptive manner. The ultimate goal is to be able to take these parameters and classify the river system as a whole. There is not currently a single stream classification system used to classify streams worldwide. There are a number of different methods currently in use. The Strahler, Leopold and Wolman, and the Montgomery and Buffington are three common methods along with arguably the most used Rosgen method (Ward et al., 2008). This paper will take an in-depth look at these four methods. Stream classification plays an important role in river restoration. In order to restore an ecosystem to a more natural setting it is necessary to be able to accurately classify the open channel to the desired setting.

Engineers, geomorphologists, and biologists have been trying to classify streams for the past century. The first recorded method of stream classification is a system set up by Davis in 1899 (Harman, 1999). There are a number of different reasons for classifying a stream. In order to avoid oversimplifying the process of classification, Rosgen listed four reasons for classifying streams: 1) To be able to predict the behavior of the river in regard to its physical aesthetics. 2) To develop relationships for given stream types in regard to hydraulics and sediment. 3) To extrapolate data specific to the site and apply them to similar rivers. 4) To classify a river is to be able to provide a consistent reference for describing the river’s morphology for those working in various disciplines (Rosgen, 1994).

The purpose of developing a consistent classification reference plays a significant role in restoration. River restoration is a communal effort and therefore can only be achieved through the work of professionals across a multitude of disciplines. In this rational, Rosgen’s fourth reason for classification becomes key. In order to perform the specific engineering steps to restore the river to its previous state, classification is required. Classification can aid in describing a specific parameter of the river in which restoration is desired. Prior to restoration certain aspects of the river have been changed to unnatural settings in regard to its natural state. Proper classification allows a river that is under restoration to be compared to a river that has retained its natural setting. This enables the parties involved to execute the restoration project which better emulates the natural setting of the river.

A current challenge in classifying a river is deciding which classification system to use. Currently Rosgen is arguably the most commonly used classification system (Ward et al., 2008). This paper will focus on this method along with another frequently used system developed by David Montgomery and John Buffington in 1997. In addition to these methods this paper will also briefly analyze techniques developed by Strahler, Leopold and Wolman, and Whiting and Bradley. This paper provides comparison and contrast of the featured methods.



Strahler’s method
One of the first methods developed for stream classification was developed by Strahler in 1952. This method simply describes the order of streams. This method starts with the smallest tributaries are considered 1st order. When two of these tributaries meet, the resulting tributary is considered 2nd order. When two 2nd order streams meet, the result is a 3rd order and so on (Strahler, 1952). Although this type of classification is fairly vague, it is an important indicator of stream size and drainage (Ward et al., 2008). Long 1st and 2nd order streams are often characteristic of manmade or severely altered natural channels. Figure 1 depicts an example of Strahler’s method.



figure1_Strahler.png
Figure 1: Strahler stream order classification method (Strahler, 1952)

Leopold and Wolman’s method
Another early method to be considered is Leopold and Wolman’s method developed in 1952 describing streams as braided, meandering, or straight. Leopold and Wolman are particularly concerned with the plan view of river system and describe channels as one of the three categories listed above. The method looks at specific reaches of the system as opposed to the whole system due to the river system often changing from straight, to meandering, to braided, etc. This early method was developed in order to attempt to “understand the mechanisms by which these laws operate in a river,” (Leopold et al., 1952). For more information on straight, meandering, or braided stream types, see the Channel Evolution Model wiki.

Leopold and Wolman describe braided rivers as channels that flow around alluvial islands developed in the system. This can include two or more channels. Braided channels are typically seen as wider, shallower and steeper than undivided channels of similar flow (Leopold et al, 1952). Velocity, cross sectional area, and roughness can all be factors resulting in the development of a braided channel. Leopold and Wolman performed a study in the lab by developing a braided channel by depositing a central bar consisting of coarse particles that could not be transported under current conditions. The coarse material acted as a catalyst for a subsequent island that developed and was maintained naturally by the system, thus creating a braided channel.

Meandering channels are different than braided channels in that they typically exhibit a single channel that is characterized by multiple turns or curves through the floodplain as it flows downstream. Sediment is deposited on the outside banks of the floodplain often leading to erosion. Meandering reaches are the most common classification of the three discussed by Leopold and Wolman.

Straight channels are classified in the Leopold Wolman method but are often rare in a natural setting. Straight natural reaches are most often shorter than ten times the width of the channel (Leopold et al, 1952). Straight reaches longer than this is often altered by man. Studies by Leopold and Wolman indicate that straight reaches also exhibit pools and riffles as do meandering channels. Further studies by Leopold and Wolman also indicate that although the banks of “straight channels” may be linear, the flow within the banks often exhibit a sort of meandering quality.

Whiting and Bradley’s method
Peter Whiting and Jeffery Bradley developed a stream classification system in 1993 that differed from the more common techniques used. Other common classification techniques focus more on large river systems whereas this technique was developed for headwater channels of small size. The variables considered in this process are hillslope gradient, channel gradient, channel gradient, valley bottom width, channel width, and sediment size. This technique uses a process based approach based on the above parameters (Whiting et al., 1993).

This technique uses three panels or graphs to classify the channel based on the physical parameters mentioned earlier. The first panel plots channel gradient against the hillslope stability and assigns an alpha variable of AD, MD, OD, SD, DD, DE, or NF (Whiting et al, 1993). AD, MD, OD, SD are all plotted together above a channel stability value of 0.8 and a channel gradient of less than 10 -1. NF falls in the area directly below that. DD falls in an area of slopes between 0.06 and 0.17. DE falls in an area of slope steeper than 0.17.

The next panel distinguishes between the four modes of AD, MD, OD, and SD. This panel relates valley width and channel width with four distinct areas (Whiting et al., 1993). The values start with SD having the highest valley width value and decreasing in order of OD, MD, and then AD.

The final panel relates median sediment size to the product of channel depth and channel gradient (Whiting et al., 1993). A numerical value of 0, 1, 2, 3, 4, or 5 is assigned based on areas broken up on the graph. These values are combined with the alpha value given previously to categorize the stream. There are a total of 42 possible categories for a river to fall in. See Figure 2 below for the three panels.

figure_2.png
Figure 2: Whiting and Bradley’s classification system (Whiting et al, 1993)

In order for this technique to be applied, field technicians are required to take measurements in the field. Generally this is necessary because headwater streams are too small to be measured using topography maps (Whiting et al., 1993). This technique was developed primarily for use in the Pacific Northwest.

Rosgen’s method
Analysis of Rosgen will be considered next. For the purpose of easily classifying rivers, Rosgen has broken the process into four levels. A river starts by being classified using Level I. The river is then further classified in Level II, by describing the river in the next sub-genre of classification. The river is then further classified in Levels III and IV. Each level deals with a different topic of characterization. Level I begins with geomorphic characterization. Level II deals with morphological descriptions. Level III characterizes the streams state. Finally, Level IV addresses validation of process characteristics (Rosgen, 1994). For the purpose of clarity, Rosgen primarily describes Level I and II in detail, and only briefly describes Level III and IV. Therefore, this paper will also focus mainly on Level I and II.

Level I provides a broad geomorphic characterization to start the classification process. Landform and fluvial characteristics are described and combine channel relief, shape, and dimension profiles (Rosgen, 1994). There are 8 categories that a stream can be classified as in Rosgen’s method. Streams are broadly classified as A, B, C, D, DA, E, F, or G. These categories are used to describe a variety of characteristics.

The first distinguishable characteristics in Level I are the longitudinal profiles used to represent slope. The slopes start with Aa+ being very steep at >10% gradually decreasing to DA at <0.5% and then increasing in slope to 4% at G. Slope can be related to bed features and can be described as pools, riffles, rapids, cascades, and steps (Rosgen, 1994). Riffle/pool streams are represented with CE, and F streams. Rapids are found in B and G streams, while steps and cascades are found in A and Aa+ streams. See Figure 3 for details.


figure_3.png
Figure 3: Longitudinal, cross-sectional and plan views of major stream types. (Rosgen, 1994)

Cross section morphology is also described in Level I of Rosgen’s method. The cross sections differ greatly in the 9 categories ranging from deep and narrow to wide and shallow. The cross section morphology also describes the flood plain ranging from well-developed flood plains to virtually no flood plain.

Finally, Level I discusses plan view morphology. The nine categories describe the sinuosity of the river system in question. River A types represent relatively straight streams, B represents low sinuosity, C represents meandering streams, E represents high meandering, and D/DA represent complex braided systems (Rosgen, 1994). This form of classification often uses the meander width ratio to describe the sinuosity. Plan view morphology is also very important for proper river restoration. Rosgen’s method can be used for “describing the most probable state of channel pattern in stream restoration work,” (Rosgen, 1994). See Figure 4 for plan view morphology.

figure_4.png
Figure 4: Rosgen’s plan view morphology classification (Rosgen, 1994)

Level II represents the morphological description of the channel. The next level of classification further describes the stream system in a more specific manner. This level breaks the channel into discreet slope ranges and introduces particle sizes of channel material. The stream types are given numbers to represent particle size diameter of the material with 1 representing bedrock, 2 is boulder, 3 is cobble, 4 is gravel, 5 is sand, and 6 is silt/clay. This generates 42 major stream types. The morphological description can only be applied to a limited length of river channel. This is due to the fact that morphology of stream systems often changes in a relatively short distance. Level II is therefore applied to only individual reaches, as opposed to being averaged over the entire basin (Rosgen, 1994).

The continuum concept is also applied to Level II. As stated before, stream systems are often changing throughout its length. Some parameters change while others stay the same and therefore only one or two of the variables that define a stream classification will be outside of the presented values. “This level recognizes and describes a continuum of river morphology within and between stream types,” (Rosgen, 1994). This application allows stream parameters such as slope to be sorted in sub-categories as opposed to slope. For example, if the majority of variables of a stream fit in the classification of C4 but has a slope of less than 0.001, the stream can be classified as C4c- (Rosgen, 1994).

Other variables considered at this level are entrenchment, width/depth, and sinuosity. For entrenchment, the entrenchment ratio can be defined as “the width of the flood-prone area to the bankfull surface width of the channel,” (Rosgen, 1994). The entrenchment ratios are given numbers for classification where 1 to 1.4 are significantly entrenched streams, 1.41 to 2.2 can be described as moderately entrenched, and greater than 2.2 are slightly entrenched. Width/depth ratio can be described as “the ratio of bankfull channel width to mean depth,” (Rosgen, 1994). A small ratio can be considered less than 12 while a moderate to high ratio is considered greater than 12. Sinuosity is defined as “the ratio of stream length to valley length,” (Rosgen, 1994). Sinuosity is often linked to slope and particle size of the channel and leads into our next topic of consideration.

Level II also addresses channel materials and slope. Channel materials play important roles in sediment transport as well as the development of the form, plan, and profile of the channel (Rosgen, 1994). Channel materials are classified using the pebble count method. Water surface slope plays an important role in channel morphology. Slopes, like other variables, can delineate from the expected values of the channels classification and therefore can be addressed with the continuum concept.

Level III describes the state of streams and helps measure existing conditions in response to channel change. This level acts as a method to propose prediction methodologies and can be used to aid in restoration efforts. Important variables in order to apply Level III include riparian vegetation, depositional patterns, meander patterns, confinement features, fish habitat indices, flow regime, river size category, debris occurrence, channel stability index, and bank erodibility (Rosgen, 1994).

The last level of classification in the Rosgen method is Level IV, which describes verification. This level provides specific information on stream processes used to verify various parameters. This level helps “provide sediment, hydraulic and biological information related to specific stream types,” (Rosgen, 1994). Classification at this level requires measurements and observations of sediment transport, bank erosion, channel geometry, biological data, and riparian vegetation data (Rosgen, 1994). See figure 5 for the breakdown of Rosgen’s classification.

Figure_5.jpg
Figure 5: Key to Rosgen Classification

Rosgen’s method is currently the most used classification system. Rosgen also discusses applying the system to restoration efforts. Historical data has shown that streams have been changing character due to imposed anthropogenic alterations in order to provide things like flood control, hydro-electric power, and allocation of water rights. These variables used to classify a river are often changed due to these alterations. Therefore, “to restore the “disturbed” river, the natural stable tendencies must be understood to predict the most probable form,” (Rosgen, 1994). Stream classification aids in providing the restoration team with knowledge of how a system’s variables naturally behave.

Montgomery and Buffington
An alternative form of stream classification was developed by David Montgomery and John Buffington in 1997. This method differs from Rosgen’s in that it primarily addresses mountain stream systems with a steeper gradient. This method recognizes three primary channel-reach substrates consisting of bedrock, alluvium, and colluvium. Alluvium substrates can be broken into 5 additional categories consisting of dune ripple, pool riffle, plane bed, step pool, and cascade. This creates a total of 7 different classification categories in which any mountainous river can be described. Montgomery and Buffington do not apply the continuum concept to their method. (Montgomery et al., 1997) states that for simplification of using a classification system, the multiple channel types between each category shall not be considered.

Cascade channels are the first alluvial channels to be considered. Cascade flow is usually described as tumbling or rolling flow dictated by large clasts. These channels typically exhibit steep slopes and are typically laid with cobbles and large boulders for bed material, creating a somewhat aggressive flow (Montgomery et al., 1997). Cascade channels are typically confined by valley walls and are narrow in nature. Cascade channels transport sediment at high rates with great efficiency and deliver transport to channels with smaller gradient.

Step-pool channels can be described as longitudinal steps separating multiple pools and reaches of the river as the flow travels downstream. The channels alter from supercritical to subcritical flow and contain pool spacing of approximately one pool to every four channel widths downstream (Montgomery et al., 1997). This type of channel can be associated with steep gradients and are considered deep in comparison with depth. They are usually confined by valley walls which keep them deep in nature. Like cascade channels, they consist of cobble and boulder bed material. They differ from cascade channels in that they exhibit a cyclic process of a steep step to relatively calm flow, while cascade channels are defined by their constant tumbling flow.

Plane-bed channels differ from the previous channels in that they typically have long stretches of beds that contain little to no features (Montgomery et al., 1997). Low width to depth ratios are typically found in these channels and can be confined or unconfined. These channels usually exhibit moderate to steep gradients. Plane-bed channels typically are dominated by gravel or cobble bed materials. Plane-bed channels act as both a supply and transport channel for sediments.
Pool-riffle channels exhibit a wave-like bed floor that helps define a sequence of bars, pools, and riffles (Montgomery et al., 1997). These channels typically occur at moderate to small slopes. These channels have large flood plains due to unconfined embankments with a width to depth ratio that varies from small to moderate. Gravel is the typical bed material found in these channels. Like plane-bed channels, pool-riffle channels can act as both the supply and transport for sediment.

Dune ripple channels differ from the other categories of channels in that they are typically low gradient with slower velocity. Small dunes of sand make up the bed floor and have a tendency to create small ripple like waves on the water surface. Dune ripple channels are classified as transport channels in regards to sediment. Bed particle size and the efficient ability to transport sediment separate this from the similar pool-riffle channel (Montgomery et al., 1997).

Colluvial channels consist of small headwater streams that contain little to ephemeral fluvial transport and flow over a colluvial valley fill (Montgomery et al., 1997). Research involving colluvial channels is lacking and therefore is more difficult to classify. Bed material varies in these channels and cannot be specifically classified. Vegetation, debris, and bedrock steps all reduce the energy available for sediment transport in these channels.

Bedrock channels are the last considered category in this classification system. These channels are defined by their lack of an alluvial bed. These channels are usually confined by some sort of valley wall. These channels consist of rock beds and are typically steeper in nature than alluvial channels. Deep flow and steep channel gradients prevent these channels from having an alluvial bed (Montgomery et al., 1997). See Table 1 for an overview of Montgomery and Buffington’s classification technique.

Table1: Overview of Montgomery and Buffington’s classification system
table_1.png


In comparison to Rosgen’s method, this system differs greatly. Rosgen’s system allows hundreds of options for classifying a river while Montgomery’s method only allows you seven options for describing a river. Rosgen’s method has a broader application to all climates and geographical regions while Montgomery’s method was developed specifically for mountainous regions of steep gradient. Rossgen suggests that his classification method can aid in restoration efforts. This is not the case for Montgomery’s method as stated by (Montgomery et al., 1997), “restoration designs requires further information on reach-specific characteristics.” This suggests that Montgomery’s method may be lacking in substance to properly classify a river to establish design criteria. This may be why Rosgen’s method is more commonly used.

Stream Classification with GIS
The classification of steams into perennial, intermittent, and ephemeral classes is an important part of stream classification and can be applied to many different fields. Currently, field techniques are primarily used to establish a stream being perennial, intermittent, or ephemeral. For more information on ephemeral streams, see the Ephemeral Streams wiki. While field techniques can be a reasonable way of obtaining data they are often riddled with measurement errors and can also be economically inefficient (Restrepo et al.). Research in stream classification using GIS (geographic information system) is currently ongoing and appearing to be more accurate and significantly cheaper.

The use of GIS to classify streams as perennial, intermittent, and ephemeral is currently being considered to replace field techniques. (Restrepo et al.) is currently researching the effectiveness of using GIS in place of field techniques. Restrepo primarily compares the GIS technique with commonly used field technique developed by the North Carolina Division of Water Quality (NCDWQ). NCDWQ technique requires field technicians to keep records of field observations and measurements using a thorough form to be filled out to report data. The form contains information to present on alluvial deposits, bankfull bench, active or relic flood plain, a continuous bed and bank, levees, sinuosity, and biological indicators. Field techniques such as these require a significant amount of experience and often exhibit a large amount of subjectivity.

The most important part of GIS stream classification is the allocation of accurate data. Digital data used for GIS classification is currently available for soil texture and elevation. The National Hydrography Dataset (NHD) from the EPA and Elevation Derivatives for National Applications (EDNA) from the USGS are two datasets that can be easily accessed for stream delineation (Restrepo et al.). The majority of the biological parameters that are measured using the field technique are unable to be portrayed using GIS; however, a number of other important parameters are accurately measured including: hydrologic soils, sinuosity, land use, groundwater, and baseflow.

(Restrepo et al.) performed a case study of the Upper Neuse River Basin to compare to field techniques. The study utilized high resolution LIDAR data. LiDAR is an aircraft-based remote sensing, using laser-driven pulses of light and multispectral cameras to scan and process digital information about a landscape. A shaded relief was projected using Albers spatial reference. Land use land cover data and high quality soil texture data was then obtained from the USGS dataset. Manual editing was performed to create the DEM (digital elevation model) data. GIS information was then generated and compared to the LIDAR data. The streamlines correlated well with the LIDAR data. Physical field data was not available at the time the paper was written. The study is still in progress.

Limitations of Classification Systems
All of the classification techniques discussed above do not come without flaws. One of the reasons why there are so many classification techniques and not one universal technique is because none of them are applicable to every stream. Most of the classification techniques listed were developed for a certain region and can only accurately classify a stream located in a similar climate amongst similar vegetation and at a similar elevation. For example, Whiting and Bradley’s method was developed for small headwater channels in the Pacific North West. Not only is this method only applicable to smaller scale systems, it’s also limited to a similar climate and topography of the North West (Whiting et al., 1993). This technique could not be applied to a more arid headwater system like one in New Mexico for example. Another example of this can be seen in Montgomery and Buffington’s method which primarily focuses on mountain systems with steep gradients. Another limitation to this method is that it can only classify a river into 7 different classifications. While this can make the classification process easier, it leaves a very broad description of the river and ultimately doesn’t tell you a lot about the river. Leopold and Wolman’s method also leaves a very broad description of the system. Describing a river as braided, meandering, or straight has its advantages but it does not give any other information. Also, river systems typically only continue with one of these descriptions for a brief time and therefore can only be classified for short reaches. Many rivers exhibit all three of these characteristics at some point. Many of these methods also require costly and sometimes inaccurate data measurement from field technicians. Using GIS data can solve this problem. Unfortunately there are still a number of biological variables that cannot be determined with GIS.

Although Rosgen’s method is currently the most widely used, it doesn’t go without criticism. Simon et al. takes a critical look at the Rosgen method of classification and addresses what could be considered a number of critical flaws (2007). Simon et al. primarily addresses the issue of Rosgen’s analysis of bankfull dimensions (2007). When considering sediments for classification, Rosgen suggests that particle counts should be considered from one bankfull level to the opposite bankfull level. Simon et al. suggests that this mixes two different alluvial materials requiring different forces and processes while depositing at different times (2007). Classification related to this issue can be seen when trying to describe two channels classified as C. One channel can have gravel bed and silt-clay banks, while the other containing a sand bed and sandbanks. These two channels could have the same median diameters of particle size. This would put both channels in the C5 type even though they would have completely different sediment transport regimes (Simon et al., 2007). For more information on the limitations of Rosgen’s method for classification see the below section on Stream Classification and Restoration and also the Channel Design wiki page.

Stream Classification and Restoration
River restoration in North America has played an important role in the recent decades. Classification allows a framework to be set for the proceeding restoration efforts. Classification paints a picture of how the system should look and operate and aids in organization of initial restoration plans. Although classification is important for the initial process of a restoration project, they are often misleading and taken too literally. Relying too heavily on classification techniques has led to many failures in restoration projects in North America. This section will discuss the proper use and misuse of classification systems when used for restoration.

Classification can play an important role for parameterizing systems over large areas where detailed data can only be taken in small areas (Kondolf, 1995). For example, it can often be difficult to take data throughout a long reach and therefore data may be taken only in a small area. Classification systems can aid in describing the rest of the channel with minimal measurements taken downstream. This can used as an initial framework for planning a restoration project by gaining some knowledge of what the channel should look like as a whole.

Formulating goals for river restoration projects often includes stream classification. Definition of biological and physical parameters of the system must be determined prior to restoration planning (Boon, 1998). Determining these parameters provides goals and gives guidance to what the restoration project would like to achieve. Flood control, water rights, and budget often require these goals to change and therefore the initial framework often changes (Kondolf, 1998).

Certain types of classification techniques may not be accurate for proper design, but they can provide what you could consider a “red flag” for certain design techniques. For example, Kondolf found that habitat enhancement structures very rarely survived a unit stream value of 35 Watts -2 or greater (1998). The National Rivers Authority classification estimates stream power. Using this information, restoration projects can red flag certain designs based on classifying a stream with greater than 35 Watts -2. For purposes such as this one, proper river classification is crucial for developing pre-project planning and design (Boon, 1998).

Although classification systems are important in the planning stage of a restoration project, they are often misused and abused for the complete project. Classification techniques can often be the easy way out of a critical design situation. Instead of dynamically studying the channel and what is really natural in regards to restoration, the system is often lumped into a specific classification even if it does not resemble every parameter that particular classification has. For example, a 1994 restoration project in Sierra Nevada was developed to return the channel to many of its natural settings. This project also included the filling of a natural pool that had resided in the channel for many years. This plan was proposed because the pool did not fit into a Rosgen type B class even though it was a natural pool (Kondolf, 1995). Classification systems act as a solid preliminary measure to restoration, and provide a sense of uniformity amongst professionals of different professionals.

Doyle suggests that the abuse of classification systems has occurred due to “ the ease of learning and applying classification schemes and public agencies’ fully adopting classification schemes and holding suspect, or even considering proposals or the findings of approaches that do not include reference to or use of a classification scheme.” (1999). It is suggested that classification techniques be used as an initial educational process as opposed to be used for the actual design of a restoration project. For greater analysis of the use and misuse of classification techniques for natural channel design see the Channel Design wiki.

Conclusion
Stream classification acts as the first step in proper stream management. To properly classify a stream, parameters such as dimension, flow, sediment, etc must be understood. Rosgen’s method is currently the most used classification technique in the scientific community. Using four levels of classification, Rosgen’s method yields a total of 94 possible types (Simon et al., 2007). In contrast to Rosgen’s method, Montgomery and Buffington’s method calls for only 7 classification classes and is used primarily to describe mountainous systems with steep gradients. Strahler’s early method of stream order has played an important role in how we look at stream systems today. Leopold and Wolman’s method of describing streams as braided, meandering, or straight have been implemented in virtually every classification system to follow. Whiting and Bradley’s method successfully allows classification of small headwater streams into 42 different classes. Stream classification plays an important role in river restoration. In order to properly restore a system to its natural setting, every parameter must be understood to predict changes either naturally or manmade. Although classification is often misused to perform the majority of the project, it can act as an excellent preliminary measure to establish a framework and a set of goals to the restoration project.

References

Boon, P. (1998). River restoration in five dimensions . Aquatic Conserv: Mar. Ecosyst. 8, 257-264.

Doyle, M. W., Miller, D. E., & Harbor, J. M. (1999). Should river restoration be based on classification schemes or process models? Insights from the history of geomorphology. ASCE International Conference on Water Resources Engineering , (pp. 1-9). Seattle .

Harman, W. A., & Jennings, G. D. (1999). Application of the Rosgen Stream Classification System to North Carolina . North Carolina Cooperative Extension Service .

Kondolf, M. G. (1995). Geomorphological stream classification in aquatic habitat restoration: uses and limitations. Aquatic Conserv: Mar. Ecosyst. 5, 127-141.

Leopold, L. B., & Wolman, M. G. (1957). River channel patterns: braided, meandering and straight. United States Geological Survey Professional Paper 282 - B, 39-84.

Montgomery, D. R., & Buffington, J. M. (1997). Channel-reach morphology in mountain drainage basins . GSA Bulletin 109(5), 596-611.

Restrepo, M., & Waisanen, P. (n.d.). Strategies for Stream Classifications. Sioux Falls : USGS EROS Data Center.

Rinaldi, M., & Johnson, P. A. (1997). Characterization of stream meanders for stream restoration . Journal of Hydraulic Engineering, 567-570.

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

Simon, A., Doyle, M., Kondolf, M., Shields Jr., F., Rhoads, B., & Mcphillips, M. (2007). Critical evaluation of how the Rosgen Classification and associated "Natural Channel Design" methods fail to integrate and quantify fluvial processes and channel response. Journal of the American Water Resources Association, 1117-1131.

Strahler, A. N. (1952). Hypsometric (area-altitude) analysis of erosional topography . Geological Society American Bulletin 63, 1117-1142.

United States Army Corps of Engineers and United States Envirionmental Protection Agency. (2004). Physical Stream Assesment: A Review of Selected Protocols for Use in the. Washington DC.

Ward, A., D'Ambrosio, J., & Mecklenburg, D. (2008). Stream Classification . Ohio State University .

Whiting, P. J., & Bradley, J. B. (1993). A process-based classification system for headwater streams. Earth Surface Processes and Landforms 18(77), 603-102.