a riffle stability index to evaluate sediment ...riffle particles smaller than the dominant large...

ABSTRACT: Riffles in moderately entrenched streamreaches with gradients of 2 percent to 4 percent thathave received excessive sediment from upstream havea distinctly different and higher proportion of smallermobile particles than riffles in systems that are indynamic equilibrium. The mobile fraction on the rifflecan be estimated by comparing the relative abun-dance of various particle sizes present on the rifflewith the dominant large particles on an adjacent bar.Riffle particles smaller than the dominant large parti-cles on the bar are interpreted as mobile. The mobilepercentile of particles on the riffle is termed “RiffleStability Index” (RSI) and provides a useful estimateof the degree of increased sediment supply to riffles inmountain streams. The RSI addresses situations inwhich increases in gravel bedload from headwatersactivities is depositing material on riffles and fillingpools, and it reflects qualitative differences betweenreference and managed watersheds. The RSI corre-lates well with other measures of stream channelphysical condition, such as V* and the results of fishhabitat surveys. Thus, it can be used as an indicatorof stream reach and watershed condition and also ofaquatic habitat quality.(KEY TERMS: bedload transport; sedimentation; peb-ble count; riffles; channel stability; monitoring.)

INTRODUCTION

In recent years there has been increasing interestin stream and watershed condition. For example, thebroad goals of the Clean Water Act (33 U.S.C. 1251-1387) to restore and maintain the chemical, physical,

and biological integrity of the Nation's waters is beingimplemented through watershed assessments at vari-ous scales. Although broad scale condition and trendassessments can be done using indirect indicatorssuch as USEPA's BASINS model, models cannot sub-stitute for long term monitoring (Halpern, 2000). Oneof the more elusive parameters to quantify isincreased sediment supply to streams. Erosion anddeposition can be increased profoundly by human-caused disturbance in a watershed. Increases in sedi-ment deposition can result in adverse effects tobeneficial uses of mountain streams by reducing habi-tat complexity and quality for cold water fisheries(Hicks et al., 1991).

This paper presents a Riffle Stability Index (RSI)method for measuring the response of sediment trans-port channel types (Schumm, 1977) to increases insediment supply. The strategy of the methodology isto determine the particle size percentile of the rifflethat is mobile and use it as an index of sedimentinput from upstream. RSI data collected from man-aged and reference watersheds in Idaho and Virginiaare compared to demonstrate the sensitivity and utili-ty of the technique for interpreting watershed condi-tion in these streams. RSI is also compared with othermethods for evaluating stream channel physical con-dition to validate the RSI methodology.

THEORETICAL BACKGROUND

A stable stream reach is one in dynamic equilibri-um. Over a time frame of several years, sediment sizeand sediment transport rates into a reach are similar

1Paper No. 01174 of the Journal of the American Water Resources Association. Discussions are open until February 1, 2003.2Hydrologist, USDA Forest Service, George Washington and Jefferson National Forests, 5162 Valley Pointe Parkway, Roanoke, Virginia

24019 (E-Mail: [email protected]).

JOURNAL OF THE AMERICAN WATER RESOURCES ASSOCIATION 1069 JAWRA

JOURNAL OF THE AMERICAN WATER RESOURCES ASSOCIATIONVOL. 38, NO. 4 AMERICAN WATER RESOURCES ASSOCIATION AUGUST 2002

A RIFFLE STABILITY INDEX TO EVALUATESEDIMENT LOADING TO STREAMS1

Gary B. Kappesser2

to those exiting the reach (Lane, 1954; Heede, 1980).When sediment transport rates into a reach exceedthose exiting, some sediment is deposited within thereach. Accelerated deposition is typically accompaniedby a textural change of the bed material (Lisle, 1982;Madej, 1982). If the imbalance persists, channels maywiden; pools may shorten, become less deep, andtransform into runs (Bisson and Sedell, 1982); andgeneral aggradation of the bed surface may occur(Lisle, 1982). Of these responses, textural change,defined as a change in the relative abundance of par-ticles of different sizes on the streambed, was foundby Lisle et al. (1993) to be the only hydraulic variablethat responds consistently to changes in sedimentload.

Streambed texture can be estimated using Wolmanpebble counts (Wolman, 1954). In the pebble countprocedure, at least 100 particles are picked from thebed on the basis of a grid system, the intermediateaxis is measured, and a frequency distribution isdeveloped. Sampling is best conducted within wad-able areas of the channel that have homogeneous pop-ulations of substrate (Kondolf, 1997), such as riffles orpools. In riverbeds with poorly sorted bed material, agreater number of pebbles may be needed to ade-quately capture the greater variability (Kondolf, 1997;Wolman, 1954).

Beschta (1987) presented a conceptual model ofsediment transport in streams in which most streamsfunction as integrated systems and where processesin individual reaches influence those in other reaches.For example, during periods of accelerated headwatererosion, large amounts of sediment become availablefor downstream transport. Generally, headwaterstreams have steep gradients and can route sedimentrapidly to downstream reaches. As this sediment isrouted downstream, it encounters decreasing channelgradients where deposition, particularly of coarsergrained materials, occurs.

In many mountainous regions of the United States,deposition of coarse grained bedload material is likelyto occur in stream reaches that classify as B and Fbchannel types (Rosgen, 1996). These Rosgen B3 andF3b channels are nearly identical in appearance, geo-morphology, function, and fish utilization. However,the entrenchment ratios of the F3b channels are lessthan 1.4; whereas the B3 channels have entrench-ment ratios between 1.4 and 2.2. Entrenchment ratiois defined as the ratio of the width of the flood pronearea to the surface width of the bankfull channel. Theflood prone area width is measured at the elevationthat corresponds to twice the maximum depth of thebankfull channel as taken from the established bank-full stage (Rosgen, 1996). Both channel types have amedian particle size (D50) of 64 to 256 millimeters,channel gradient between 2 and 4 percent, width/

depth ratio greater than 12, and moderate sinuosity(greater than 1.2). B3 and F3b stream reaches havestreambeds that are predominantly cobble with lesseramounts of boulders, gravel, and sand (Rosgen, 1996).Within the range of cold water fish in national forests,B and Fb channels represent some of the most impor-tant habitat (Dave Cross, personal communication).

Headwaters of natural gravel bed rivers have manygrain sizes, some of which can be transported down-stream under commonly occurring flows and somewhich cannot except under extremely high flows orfloods (Gessler, 1971). During flow events capable oftransporting bedload, sediment in stream systems indynamic equilibrium is scoured from pools anddeposited on riffles (Hirsch and Abrahams, 1981;Lisle, 1979). As flow decreases, finer particles arewinnowed out of the riffles and deposited in the nextpool downstream (Hirsch and Abrahams, 1981).When sediment supply from upstream exceeds thestream's competence to transport it, pools fill and rif-fles load with finer mobile sediment (Andrews, 1982).

Riffles represent concentrations of larger residualparticles in stream channels (Yang, 1971; Church andJones, 1982) and therefore offer the greatest contrastbetween residual particles and those that are mobile.Thus, riffles offer logical sites to determine channelbed texture.

To use riffle substrate as an index of streambedcondition, the size of the largest bedload particlesmobile during frequent flood events must be known.Point bars have been identified as sources of thisinformation. Point bars are composed of and formedby the coarse part of the total bedload (Leopold, 1992,1994), which is deposited primarily during near floodstages (Bridge and Jarvis, 1982). Rosgen (1996) foundthat core samples from point bars and central barsprovide good indicators of bedload sizes that aretransported at normal high flows.

Bar development usually occurs only in channelswith gradients of approximately less than 5 percent;above that gradient bars rarely form (Church andJones, 1982). Classic point bars are found in low gra-dient meandering streams with higher sinuosity (e.g.,sinuosity > 1.2). Smaller, less extensive bar depositscalled lateral bars are found in higher gradient, moreentrenched streams (Church and Jones, 1982) such asB or Fb type streams (Rosgen, 1996). Lateral bars aresimilar to point bars in that they contain deposits ofthe larger sizes of material transported as bedload.Therefore, a sample of the largest particles on a later-al bar can be used to estimate the largest particlesmobilized during high flow events. The largest parti-cles on a bar can be identified by systematic selection,as well as by random sampling (Bluck, 1982; Ash-worth and Ferguson, 1989). To characterize bardeposits, Bluck (1982) selected 30 of the largest clasts

JAWRA 1070 JOURNAL OF THE AMERICAN WATER RESOURCES ASSOCIATION

Kappesser

from bar deposits, and Ashworth and Ferguson (1989)measured the mean diameter of the 10 largest parti-cles on the barhead. The geometric mean of the 10 to30 largest particle sizes on a bar, termed the domi-nant bar particle size, can be compared with thecumulative size distribution of the range of particlesizes present on an adjacent riffle to estimate the per-centile of mobile riffle material.

The working hypothesis of this paper is that rifflebed material in Rosgen B and Fb channel types is amixture of smaller mobile materials that can movefrom one riffle to the next during frequent flood flowevents and larger residual particles that do not moveor move only slightly and stay within the same rifflefor many frequent flood flow events. As a riffleincreasingly is loaded with sediment from upstream,textural shifts occur because of the relative increasein the mobile component of the bed. Thus, the mobilepercentile in stable systems is distinctly differentfrom that in systems that have received excessive sed-iment from upstream. The mobile percentile, that is,the RSI, is a useful expression of the relative degreeof textural shift in the riffle. By extrapolation, sedi-mentation of a reach can be estimated, and the condi-tion of the watershed above this reach may beinferred.

METHODS

Study Sites

Idaho field data presented in this analysis are fromthe Coeur d'Alene and St. Joe Rivers and their tribu-taries within the boundary of the Idaho PanhandleNational Forests (IPNF) and east of Spokane, Wash-ington (Figure 1). Watershed elevations range from640 m above sea level to nearly 2,135 m at the Idaho/Montana border. Average annual runoff is 81 cm, withpeak runoff occurring during snowmelt in April andMay. The mountain watersheds support stands ofDouglas fir (Pseudotsuga menziesii), larch (Larixspp.), grand fir (Abies grandis), hemlock (Tsuga spp.),Western red cedar (Thuja plicata), and lodgepole pine(Pinus contorta) (Eyre, 1980). The area is underlainby Precambrian Belt series metasedimentary rocks,with inclusions of granitics and mica schists (Bayer,1984). The Belt geology is fairly homogeneous acrossthe study area in both lithology and structure.

Study watershed areas range from 13 to 39 km2and are classified as either reference or managedbased on their past management history (Table 1).Reference watersheds have no history of roading, log-ging, or mining. Stream channels in the reference

watersheds have visible attributes of stability as iden-tified by Pfankuch (1975). Managed watersheds onthe IPNF commonly had large clear cuts, sometimesextending for thousands of acres. A high density oflogging roads was necessary to support the past har-vest. Typically, logging roads were constructed on a122 m parallel spacing on the contour, or with a roaddensity sometimes exceeding 9 km of road per squarekilometer of watershed. Following harvest, the cut-ting units were burned to prepare the area for thenext crop of trees. Stream channels in managedwatersheds lacked visible attributes of stability.

Virginia field data are from headwater streams onthe George Washington and Jefferson NationalForests (GWJNF) in southwestern Virginia (Table 1).Two of the Virginia streams, North Creek and ClearCreek (Figure 2), are examined intensively in thispaper. North Creek is a 31 km2 watershed tributaryto the James River north of Roanoke, and Clear Creekis a 14 km2 catchment in the headwaters of the UpperTennessee River near Norton, Virginia. They lie with-in the Blue Ridge and Valley and Ridge provinces,respectively. The Blue Ridge province is underlain bya complex of metasedimentary and igneous rocks,whereas the Valley and Ridge province is underlainby highly folded and faulted Paleozoic sedimentarysandstones, limestones, and shales (Bayer, 1984).Watershed elevations range from less than 305 m tomore than 915 m at higher elevations. Average annu-al runoff ranges from 38 to 51 cm per year. March isthe month of highest streamflow. Peak flows occur inresponse to storm events throughout the year, and


A Riffle Stability Index to Evaluate Sediment Loading to Streams

Figure 1. Locations of Coeur d'Aleneand St. Joe Watersheds, Idaho.

runoff is dominated by rainfall inputs. The mountainwatersheds support stands of temperate deciduousforest, dominated by tall broadleaf trees including oak(Quercus spp.), beech (Fagus spp.), birch (Betula spp.),hickory (Carya spp.), walnut (Juglans spp.), maple(Acer spp.), and tulip poplar (Liriodendron tulipifera)(Eyre, 1980). Lands that are currently within theGWJNF previously were privately owned and wereextensively harvested and often grazed or farmedprior to 1940. Historic landuse resulted in greatlyincreased sedimentation to streams (Day, 1941).Channels are currently in various stages of recovery,and full recovery is expected to take several decades.

Due to the landuse history in this area, no true ref-erence streams or reaches were available in this area.However, North Creek is in better condition thanmost streams in Virginia, even though it has ongoingmanagement activity and a network of roads. Approx-imately 2.4 km of the 7.3 km reach sampled for RSIare within the only designated Outstanding ResourceWater in Virginia (Commonwealth of Virginia, 1997),and stream habitat surveys conducted by the Centerfor Aquatic Technology Transfer confirm the goodphysical condition of the channel (Andrew Dolloff,unpublished data).


Kappesser

TABLE 1. Idaho and Virginia Study Watersheds.

Idaho PanhandleReference Managed Virginia*

Canyon Creek Big Creek 2557 Tributary to Stony CreekEagle Creek Bird Creek Bad CreekEast Fork Comfy Ck Blacktail Creek Bark Camp BranchEast Trib Halsey Creek Bottom Creek Burns CreekFly Creek Boulder Creek Chestnut RunHalsey Creek Burnt Cabin Creek Clear CreekHell's Canyon Cascade Creek Colon HollowJorden Creek Cascade Creek Cornelius CreekLarkins Creek Cat Spur Creek Davyland BranchLong Canyon Creek Chilco Creek Davyland TributaryLong Canyon Creek Conie Creek Fox CreekMosquito Creek Dusty Creek Helton CreekRousch Creek E. F.Hayden Creek Jennings CreekRuby Creek Eagle Creek Joel BranchSawtooth Creek Flemming Creek Laurel ForkSkookum Creek George Creek Lewis ForkSpotted Louis Creek Grouse Creek McFalls CreekStraight Creek Hayden Creek Middle CreekTimber Creek Haystack Creek North CreekTinear Creek Heller Creek Opossum CreekU. St. Joe River Hungry Creek Patterson Creek Tributary

Jim Creek Phillips CreekLone Cabin Creek Poor Fork CumberlandLower West Branch Roaring BranchMokins Creek Robinson ForkMontana Creek South Prong Barbours CreekMosquito Creek Stoney CreekN.F.Hayden Creek Straight BranchN.F. Grouse Creek Whiterocks CreekRapids Creek Wolf CreekRocky Run CreekRutledge CreekSiwash CreekSmith CreekWalker CreekYellowbanks Creek

*No reference sites were sampled in Virginia, but North Creek is most like a reference stream.

Clear Creek watershed is much more typical ofother watersheds in the region than is North Creek.The Clear Creek watershed was extensively harvest-ed at the beginning of the 20th century and currentlyhas a network of roads and some ongoing manage-ment activities. Channel condition is more impactedby watershed activities, as evidenced by a lack ofvisual attributes of stability (Pfankuch, 1975) andfrom stream survey information (James Nutt, unpub-lished data).

Field Measurements and Calculation of RSI

Stream systems were segmented into uniformreaches using the classification system of Rosgen(1996). Reaches were premapped on 1:24,000 scaletopographic maps and validated in the field. Mini-mum reach length was 200 meters. Three riffles weremeasured within each uniform Rosgen B2, B3, B4,F2b, F3b, or F4b reach. Each riffle selected for mea-surement was judged to be representative or typicalwithin its larger reach after inspecting the entirereach. Riffles were located in straight sections ofreaches, had uniform depth in the cross section, andincluded a thalweg crossover. A typical riffle is shownin Figure 3.

Particle size distribution on each riffle was deter-mined by Wolman (1954) pebble count. At least 200 particles were measured per riffle. Sampling pointswere identified by establishing a sampling grid, withtransects extending across the bankfull width over

the entire length of riffle. Samples were taken at 0.3 m intervals along a measuring tape at each tran-sect. The number of transects needed was determinedby measuring average bankfull width and dividing by0.3 m to determine the number of samples per tran-sect, then dividing the number of samples per tran-sect into 200 and rounding up. Spacing betweentransects was determined by dividing the length ofriffle by the number of transects. Transects weretaken from downstream to upstream. The location ofthe first downstream transect was determined bymeasuring one-half of a transect spacing upstreamfrom the downstream end of the riffle.

The intermediate axis of each particle was mea-sured using a metric caliper and tallied by one phiinterval size class (Lane et al., 1947) in which the sizerange in millimeters doubles with each size class (2,4, 8, 16, 32, etc.) except that sand size and smallerparticles were grouped as less than 2 millimeters. Thepercent in each size class was determined, and thecumulative percent finer then was calculated for eachsize class. The cumulative percent finer is plottedagainst the upper value for each size class to generatea cumulative particle size distribution curve (Blatt etal., 1972).

A lateral bar or similar depositional feature wasidentified in close proximity to each measured riffle.The entire bar was inspected visually to identify the dominant large size of particles present. Approxi-mately 10 to 30 freshly moved dominant large parti-cles residing on the depositional feature wereidentified. The number of particles that were



Figure 2. Locations of North Creek and Clear Creek, Virginia.

measured depended upon the number of particlesavailable that were approximately the same dominantlarge size. Freshness of movement was evaluated byrelatively brighter color, lack of particle embedded-ness, lack of stain, and lack of algae or moss coatingthe particle. For each of the particles, the intermedi-ate axis was measured and recorded to the nearestmillimeter. The geometric mean of the sample wascalculated and compared with the plotted cumulativeparticle size distribution for the riffle. The percentileof the cumulative particle size distribution corre-sponding to the mean of dominant large particle sizeson the lateral bar is the RSI (Figure 4).

Statistical Analyses

Nonparametric Wilcoxon rank sum tests (Steel andTorrie, 1980) were used to compare RSI values fromreference and managed watersheds because the datawere not normally distributed. Wilcoxon Z-statistics,calculated from sums of ranks for RSI valuesand sample size, were used to determine whetherthe ranks were significantly different (α = 0.05). P-statistics are provided to show the actual levelof probability of significance. Box whisker plots

illustrate the range of RSI for each set of data. Thebold horizontal bar inside the box indicates the medi-an RSI value. Upper and lower limits of each box,respectively, indicate the upper and lower quartiles;the middle 50 percent of the data fall within this box.The upper limit of the whisker is the 95th percentile,and the lower limit of the whisker is the 5th per-centile. Regression analyses also are performed toillustrate relationships between different variables.The Wilcoxon statistical tests were performed usingSAS (SAS Institute Inc., 1988), and regression analy-ses were performed using Excel.

RESULTS

North Idaho

In the Coeur d'Alene and St. Joe watersheds, 21watersheds were determined to be reference and 36watersheds were identified as managed. The 21 refer-ence watersheds provided a total of 51 RSI samples,and the 36 managed watersheds provided a total of109 RSI samples. Distributions of RSI values for the


Kappesser

Figure 3. Example Riffle in North Creek, Virginia. Bankfull width, 13.5 meters; channel gradient, 0.021;median particle size, 140 millimeters; mean of dominant large particle sizes on bar, 200 mm; RSI, 60.

reference and managed watersheds are displayed inFigure 5. The reference watersheds had a median RSIvalue of 58, whereas the managed watersheds had amedian value of 80. The 75th percentile of thereference watershed RSI values was 72, which wasclose to the 25th percentile value of 68 for the man-aged watersheds. Wilcoxon test results comparing theIdaho managed and reference watersheds (Table 2)indicate that the two groups of watersheds had signif-icantly different RSI values (P < 0.0001).

A subset of both groups (n = 19), composed of sec-ond order B channels representing the tributaries tothe St. Joe River, were further evaluated to explorethe relationship between loading of mobile materialon the riffle and filling of pools. The subset was select-ed to reflect more uniform lithology and better hydro-logic similitude. In pools adjacent to measured riffles,residual depth of each pool tail crest was measured(Lisle, 1987) and multiplied by pool length and bank-full width to obtain a rudimentary estimate ofresidual pool volume. This value then was divided by

watershed area above the pool to allow channel com-parisons from watersheds of different sizes. Thistransformed residual pool volume was regressedagainst RSI values (Figure 6). The regression lineshows that as RSI values increase, pool volumesdecrease. The variability in the data (R2 = 0.52) isattributed to other factors that create and maintainpools in stream channels, such as large woody debrisand bedrock outcrops.

Virginia Data

Since 1994, RSI measurements have been madefrom at least 30 watersheds on the GWJNF. For allVirginia data sites, both the arithmetic mean andmedian of the samples are 70. The 25th percentile is58.5, and the 75th percentile is 80.5. Because no truereference watersheds were available to sample in theVirginia study, data from North Creek and ClearCreek are used to represent watersheds in variousstages of hydrologic recovery.



Figure 4. Determination of Riffle Stability Index. The meandominant large particle size on the bar is compared to thecumulative particle size distribution curve to obtain RSI.

Figure 5. Range of RSI Values for B Channel Reaches ofReference and Managed Watersheds, North Idaho.

TABLE 2. Wilcoxon Test Statistics for Idaho Reference Versus Managed Watersheds,and Idaho Reference Watersheds Versus North Creek, Virginia.

Statistic Idaho Reference Idaho Managed Idaho Reference North Creek

Sample Size 51 109 51 21

Median 58.3 80.0 58.3 58.0

Mean Rank 45.8 96.7 36.0 37.8

Sum of Ranks 2336.5 10543.5 1835.0 793.0

Z-Statistic -6.4779 0.3224

Probability

North Creek

Twenty-one RSI samples were taken in NorthCreek. They had a median RSI value of 58. The lowerand upper quartile values, respectively, were 53 and66 (Figure 7). Median RSI values and distributions ofRSI data from North Creek were similar to the 51 ref-erence watershed samples from northern Idaho (Fig-ure 7). Wilcoxon comparisons between RSI data forthe Idaho reference watersheds and North Creekshowed almost identical mean ranks (36.0 and 37.8,respectively) (Table 2). The probability that the twodata sets are the same was 0.37 (Table 2), indicatingthat North Creek RSI data were not significantly dif-ferent from the Idaho reference watersheds.

Clear Creek

RSI data from Clear Creek illustrate RSI’sresponse to localized channel aggradation. A logstructure called a “K-dam” was constructed for fishhabitat improvement approximately 25 years ago inClear Creek. The top of the log structure is approxi-mately 1 m above the existing channel bottom. Thestructure has functioned effectively as a sedimenttrap and has filled completely with sediment trans-ported from upstream. The channel immediatelyabove the structure is best described as a run. RSIsamples were taken immediately above the K-damand also at riffles 125 m upstream and 168 m down-stream from the K-dam.

The upstream riffle produced an RSI value of 77.The run immediately above the structure had an RSIvalue of 96, and the downstream riffle had an RSIvalue of 79. The RSI value at the structure reflectsthe local textural shift produced by bedload sedimenttrapped behind the structure. The RSI value was lessthan 100 only because of large boulder material in theadjacent stream banks that was included in the peb-ble count tally (i.e., within bankfull to bankfull mea-surements).

DISCUSSION

Idaho Streams

The data from northern Idaho show a significantdifference in RSI values between the reference andmanaged watersheds. The median RSI for the man-aged watersheds was 80, compared to 58 for the refer-ence watersheds (Figure 5), indicating gravel andsand loading of riffles in managed watersheds. Theincreased sediment supply is attributed to the highdensity of roads, associated road fill failures, and cul-vert failures and from release of sediment that hadbeen stored in intermittent and ephemeral headwaterchannels (Horner, 1993).

The range of RSI values in reference watersheds isa reflection of natural variability of these systemsthat includes disturbance regimes such as fire, insect,disease, and localized storm events. The range of RSIvalues in B channels of managed watersheds innorthern Idaho (Figure 5) may be attributed to anumber of factors, including differences in the per-centages of the watersheds that had been harvested,the location and spatial distribution of harvests, size of openings, location and density of roads, andthe use or effectiveness of Best Management Prac-tices. The lower end of the range of RSI data suggest


Kappesser

Figure 6. Plot of Residual Pool Volume/Upstream WatershedArea Versus RSI. St. Joe Tributary Watershed

Second Order B Channels.

Figure 7. Range of RSI Values for B Channel Reaches of NorthCreek, Virginia, and Idaho Reference Watersheds.

that watersheds with a moderate degree of welldesigned management activity (harvest and roads)would be expected to have RSI values similar to thoseof the reference watersheds.

Virginia Streams

The localized variability of RSI values in the NorthCreek channel (Figure 7) might be attributed to local-ized recent disturbances such as bridge washouts dur-ing flood events the year before the field data werecollected. Because of the landuse history of water-sheds on the GWJNF and also because of environ-mental regulations of the past two decades,watersheds with a magnitude of treatment compara-ble to those in northern Idaho do not exist. The reachof Clear Creek with its log K-dam is used as a substi-tute. The commonality is that the load of sedimententering the reach exceeds the load of sediment exit-ing the reach. The Clear Creek data suggest that theRSI value approaching 100 is valid in Virginia as wellas in Idaho in riffles where almost all bed material ismobile.

Data sets from North Creek and Clear Creek evalu-ate the applicability of RSI in a very different physio-graphic and climatic region from northern Idaho. Thesimilarity in RSI medians and ranges between NorthCreek and B channel reaches of reference watershedsin northern Idaho suggests that the RSI technique isrobust to the natural variations among differentregions and is able to distinguish between stablestreams and those that are loading with excess sedi-ment, at least within B and Fb channels.

Comparison of RSI to Other Metrics

The textural change on the riffle is only one mani-festation of channel response to increased sedimentload. As riffles load, pools fill with mobile materialand prolonged increases in sediment supply mayresult in reductions in pool lengths and increases inthe occurrence and frequency of runs as riffles elon-gate and become finer textured (Lisle, 1982).

Changes in the physical condition of a streamchannel may be reflected in aquatic communityresponses. Cross and Everest (1995) reported theimplications for bull charr, a fish currently listed asthreatened (U.S. Fish and Wildlife Service, 1998). In1992, they surveyed 20 reference and managed water-sheds tributary to the St. Joe River for bull charredds. Sixty-one redds were located in seven tribu-taries and the main stem of the St. Joe River. Withthe exception of two, all bull char redds were located

in reference streams with RSI values less than 65.They observed that in streams with high RSI values,pools essential to bull charr and other native fish hadreduced volume and relative abundance.

Correlations between RSI and measures of fishhabitat also are reported by Cross and Everest (1995).Their study was conducted concurrently with RSIdata collection and within these same watersheds.Using the habitat typing methodology described byMcCain et al. (1990), they measured the lengths,widths, and depths of each habitat unit within uni-form Rosgen reach types. They also measured themaximum depth of each pool at the pool tail crest ele-vation. Also called the riffle crest, the pool tail crestforms the downstream lip of the pool, that is, the poolthat would remain if there were negligible surfaceflow (Lisle and Hilton, 1992). Their results showedthat for Rosgen B reaches in tributaries to the St. JoeRiver, managed watersheds had less pocket water andpool habitat and increased run/glide and braid habitatrelative to reference watersheds. Mean residual poolvolume for second order B channels of managedwatersheds (mean RSI 82) was 51 percent less thanthe mean residual value of reference watershed pools(mean RSI 44).

Pool filling can be measured using a procedurecalled “V*” (Lisle and Hilton, 1992). A study conduct-ed in 1992 for the California North Coast RegionalWater Quality Control Board (Chris Knopp, unpub-lished data) evaluated 60 stream reaches betweenSan Francisco and the Oregon border. Most reachesclassified as B3 (Rosgen, 1996), but a few C3 reachesalso were included. RSI data, V* measurements, andother parameters were collected at each stream reach.The R2 value for the regression comparing RSI to V*was 0.453 (Figure 8). The regression line relates theRSI value of 58 (median of Idaho references andNorth Creek) to a V* value of 0.1 (stable channel),anda RSI value of 100 to a V* value of 0.9 (pool nearlyfilled with sediment).

RSI compares the largest mobile particles in astream channel with the cumulative particle size dis-tribution. Kaufmann et al. (1999) are pursuing a simi-lar approach called Relative Bed Stability, whichcompares the median particle size (D50) of bed materi-al with the largest mobile particles. Although the twoprocedures use different field methods, the conceptsbehind them are explored to identify conditions underwhich there was an advantage to using the medianparticle size (D50) instead of the cumulative particlesize distribution to generate an index of stability.

Data from 93 Virginia stream reaches from water-sheds listed in Table 1 were used to develop both RSIand the ratio of the median particle size on the riffle(D50) to the diameter of the largest mobile particle(Dm). For comparisons here, the geometric mean



diameter of the bar sample was used as an estimate oflargest mobile particle. The Wolman pebble count onthe riffle was used to determine the median particlesize (D50) of the bed material. A log transformationwas performed to produce [log (D50/Dm)] values forthe data set. The [log (D50/Dm)] values are plottedagainst RSI values in Figure 9.

The equation of the regression line (Figure 9)relates a RSI value of 50 with a [log (D50/Dm)] valueof 0. The two values are equivalent and have thesame interpretation; that is, the median particle sizeof bed material is equal to the largest mobile particle.

As RSI values increase and [log (D50/Dm)] valuesbecome increasingly negative, departure from theregression line increases. Above a RSI value of 70, thedata points split into two groups, one that falls abovethe regression line and another that falls below it.Values below the regression line and RSI greater than70 have very small median particle size of bed materi-al. Points with [log (D50/Dm)] values between -1.5 and -1.0 have D50 values from 2 to 16 mm. Points with [log (D50/Dm)] values between -1.5 and -2.5 have a D50of 2 mm. The riffles depicted by points below theregression line and RSI greater than 70 have a strongbimodal distribution, with one mode being sand andthe other cobble and small boulders. The channelscould be described as highly embedded. They probablyrepresent former B3 channels that have been smoth-ered with sand from upstream disturbance. Pointsabove the regression line with RSI values greaterthan 70 have D50 values in the gravel size class rangeand represent channels that have been loaded withgravel size bedload from upstream sources.

The lack of strong correlation between RSI and [log (D50/Dm)] for RSI values greater than 70 (Figure9) might be explained by differences in geology androck weathering. Study watersheds underlain byincompetent sandstone or decomposed granite haverelatively larger amounts of sand and very fine gravelavailable as bedload. When sedimentation from theseparticle sizes becomes extreme, channels have largernegative values of [log (D50/Dm)] without correspond-ing very high RSI values. Study watersheds underlainby competent hard quartzite or granite bedrock thatdecompose to gravel size particles or siltstones thatweather from gravel to silt without surviving as sandsize particles will produce higher RSI values andsmaller negative values of [log (D50/Dm)].

RSI is more sensitive in watersheds underlain bycompetent rocks that decompose to produce gravels.The [log (D50/Dm)] metric is more sensitive in watersheds with incompetent rocks that decompose tosand or very fine gravel particle sizes. The RSI and[log (D50/Dm)] metrics can be generated from thesame set of field data. In conjunction with each other,they can be used to evaluate a wider range of water-shed geologies and conditions.

RSI effectively provides a “real time” interpretationof the Wolman pebble count data. The range of RSIvalues from less than 50 to 100 can be used as anindicator of the magnitude of textural fining of the rif-fle in Rosgen B and Fb channel reaches. There are noclear breaks between the numbers, and the followingranges are suggested only as guidelines. RSI valuesfrom 40 to 70 approximate the middle quartiles forthe IPNF reference watersheds and suggest a rifflethat is in dynamic equilibrium. Sediment load enter-ing the riffle equals the sediment load exiting the


Kappesser

Figure 8. Plot of RSI Versus V* From Watersheds inNorthern California (unpublished data, used

by permission from Chris Knopp).

Figure 9. Plot of [log (D50/Dm)] Versus RSI. Data points for RSIgreater than 70 that fall below the regression line have riffleD50 of sand or very fine gravel. Those above the regression

line have a riffle D50 of larger gravel or cobble.

riffle over time. RSI values above 85 and approaching100 correspond to the upper quartile of IPNF man-aged streams and are above the range of IPNF refer-ence watersheds. They are indicative of riffles thatare loading increasingly with excess sediment. RSIvalues between 70 and 85 are intermediate and sug-gest that the riffle is somewhat loaded with sediment.RSI values less than 40 represent either riffles with ahigh bedrock component or riffles that have scoured.RSI values can be interpreted with respect to the rif-fle from which the data were collected, the reach thatincludes the riffle, and by extension to the watershedabove the reach.

In most instances, sediment loading on riffles inRosgen B and Fb channel reaches is the result ofincreased sediment produced from the watershedabove. Thus, the RSI values can indicate upstreamwatershed condition. Riffles in northern Idaho andVirginia with values less than 70 are indicative ofwatersheds in good condition. Values between 70 and85 indicate watersheds are in fair condition, andvalues greater than 85 indicate poor watershed condi-tion.

In all cases, these RSI values are merely suggestedguidelines based on available data and for the specificlocales of data collection. Interpretations should bevalidated for other locales by developing comparisonsof reference and treatment values or by correlatingRSI with other instream parameters. For example,the data of Cross and Everest (1995) that showedstrong preference of bull charr (Salvelinus confluen-tus) for stable stream channels in northern Idahowith RSI values less than 65 support the RSI inter-pretation that those streams are in dynamic equilibri-um and their watersheds are in good condition.Similarly, their data showing reduced pool volume formanaged watersheds with a mean RSI of 82 reinforcethe interpretation that these stream channels areloading with sediment and that their watershed con-dition is poor.

RSI values are indicators and should not be takenas absolutes. Marcus et al. (1995) and Kondolf (1997)discussed limitations of the Wolman pebble counttechnique. Sampling error and sampler bias aresources of error. Lateral bar samples carry similarlimitations, as does the use of three representativeriffles within a reach. Riffle texture may become finerduring years without significant high flow events andmay coarsen in years with major storms. Thus, RSIvalues are expected to fluctuate with the dynamics ofthe stream channel over a range of natural variabili-ty.

However, the RSI procedure is “forgiving” in thaterrors in determining the mean of the dominant largeparticle size on the bar will not result in great differ-ences in the index value. For example, one of the riffle

sites on North Creek had an RSI of 55 based on a barsample mean diameter of 197 mm. If the bar samplemean was overestimated by 20 percent to be 240 mm,the resulting RSI value would have been 62. If the barsample mean was underestimated by 20 percent, thenthe RSI value would have been 46. In all cases theinterpretation of the index would be the same. Simi-larly, riffles that are loaded with gravel size sedimentwill produce a high RSI value for a range of bar sam-ple means.

Riffles commonly have “patches” of finer materialand localized areas of coarse residual large bed ele-ments. A pebble count of 200 or greater was used tocapture these localized areas. The larger sample sizealso provides a more accurate representation of theparticle size distribution of the riffle (Bunte and Abt,2001). A one-phi size class interval was used in bothIdaho and Virginia. Changing to a one-half phi inter-val in future sampling (Harrelson et al., 1994) mayimprove the results by reducing the cumulativeparticle size distribution curve interpolation range.

Bar samples were used to determine the largestmobile particle size because they probably are a morereliable indicator than tractive force equations. Trac-tive force equations also must rely on one specificflow, usually bankfull. The size and relative abun-dance of particles on both riffles and lateral bars maybe influenced by recent flood history. Hallisey andBelt (1996) compared the results of the bar samplewith those using a tractive force equation for 123 rif-fle sites on the Idaho Panhandle National Forest.They found that the results agreed in only 16 percentof the sites but that the tractive force equation signifi-cantly over or underestimated on the remaining sites.

CONCLUSION

An evaluation of the condition and trend of streamchannels and watersheds requires direct measure-ments that are capable of showing change over timeand that can be related to beneficial use. The RSIprovides a useful estimate of the degree of texturalchange of riffles in mountain streams due toincreased sediment supply. It applies to watershedswhere increases in gravel size bedload from activitiesin the headwaters are depositing on riffles and fillingpools in these systems. RSI is easily accomplished insmall to medium wadable stream channels and doesnot require sophisticated equipment. It reflects quali-tative differences between reference and managedwatersheds. RSI correlates well with other measuresof stream channel physical condition, such as V*.Thus, it can be used as an indicator of stream reachand watershed condition. Filling of pools and loading



of riffles with sediment represents a reduction ofhabitat quality and complexity for aquatic species.Habitat degradation by loss of complexity is one of themost important limits on the diversity of fish commu-nities (Hicks et al. 1991). RSI correlates well with theresults of fish habitat surveys and provides a tool forquantifying aquatic habitat degradation in B and FbRosgen channel types. Interpretations can be madefrom one-time measurements, and changes can bemonitored over time through repeated measurements.Thus, stream channel and watershed condition andtrend can be measured over time and related to animportant component of instream beneficial use.

ACKNOWLEDGMENTS AND DISCLAIMER

I thank Bob Kasun, Loren Everest, Mike Owen, Jules Folnagy,Judy McHugh, Ken Heffner, Kelly Shanahan, Judy Hallisey, SteveFlood, Rob Harper, Jill Cobb, Ed Lider, Jim Nutt, Chris Owens,Wes Childress, Dawn Kirk, and Rebecca Kavage for their assis-tance in field data collection. Chris Knopp kindly provided fielddata from watersheds in California. The paper benefited greatlyfrom a review by Pam Edwards. Dave Cross, Ann Puffer, BrucePruitt, John Potyondy, and Kristin Bunte provided technical sup-port and encouragement. The use of trade, firm, or corporationnames in this publication is for the information and convenience ofthe reader. Such use does not constitute an official endorsement orapproval by the U.S. Department of Agriculture or the U.S. ForestService of any product or service to the exclusion of others that maybe suitable.

LITERATURE CITED

Andrews, E. D., 1982. Bank Stability and Channel Width Adjust-ment, East Fork River, Wyoming. Water Resources Research18(4):1184-1192.

Ashworth, P. J. and R. I. Ferguson, 1989. Size-Selective Entrain-ment of Bed Load in Gravel Bed Streams. Water ResourcesResearch 25(4):627-634.

Bayer, K. C., 1984. Generalized Structural Lithologic and Physio-graphic Provinces in the Fold and Thrust Belts of the UnitedStates; Exclusive of Alaska and Hawaii. U.S. Geological Survey,Reston, Virginia.

Beschta, R. L., 1987. Conceptual Models of Sediment Transport inStreams. In: Sediment Transport in Gravel-Bed Rivers, C. R.Thorne, J. C. Bathurst and R. D. Hey (Editors). John Wiley andSons Ltd., New York, New York, pp. 387-419.

Bisson, P. A. and J. R. Sedell, 1982. Salmonid Populations inStreams in Clearcut vs. Old Growth Forest of Western Washing-ton. In: Fish and Wildlife Relationships in Old-Growth Forests,W. R. Meehan, T. R. Merrall, and J. W. Matthews (Editors).Amer. Inst. Fish. Res. Bios., pp. 121-130.

Blatt, H., G. Middleton, and R. Murray. 1972. Origin of Sedimenta-ry Rocks. Prentice-Hall, Inc., Englewood Cliffs, New Jersey.

Bluck B. J., 1982. Texture of Gravel Bars in Braided Streams. In:Gravel-Bed Rivers, R. D. Hey, J. C. Bathurst, and C. R. Thorne(Editors). John Wiley and Sons Ltd., New York, New York, pp.291-339.

Bridge, J. S. and J. Jarvis, 1982. The Dynamics of a River Bend: AStudy in Flow and Sedimentary Processes. Sedimentology29:499-541.

Bunte, K. and S. R. Abt, 2001. Sampling Surface and SubsurfaceParticle-Size Distributions in Wadable Gravel and Cobble-BedStreams for Analysis in Sediment Transport, Hydraulics, andStreambed Monitoring. Gen. Tech. Rep. RMRS-GTR-74, USDAForest Service, Rocky Mtn. Research Station, Fort Collins, Col-orado, 428 pp.

Church, M. and D. Jones, 1982. Channel Bars in Gravel-BedRivers. In: Gravel-Bed Rivers, R. D. Hey, J. C. Bathurst, and C.R. Thorne (Editors). John Wiley and Sons Ltd., New York, NewYork, pp. 291-339.

Commonwealth of Virginia, 1997. State Water Control Board 9 VAC25-260-5 et seq. Water Quality Standards. Statutory Authority: §62.1-44.15(3a) of the Code of Virginia, 9 VAC 25-260-30,Antidegradation Policy.

Cross, D. and L. Everest, 1995. Fish Habitat Attributes of Refer-ence and Managed Watersheds with Special Reference to theLocation of Bull Charr (Salvelinus confluentus) Spawning Sitesin the Upper Spokane River Ecosystem, Northern Idaho. FHRCurrents, Fish Habitat Relationships Technical Bulletin No. 17,USDA Forest Service.

Day, J. F., 1941. Bloody Ground. Doubleday, Doran and Co., NewYork, New York.

Eyre, F. H. (Editor), 1980. Forest Cover Types of the United Statesand Canada. Society of American Foresters, Washington, D.C.,147 pp.

Gessler, J., 1971. Beginning and Ceasing of Sediment Motion. In:River Mechanics, H. W. Shen (Editor). Fort Collins, Colorado,Vol. 7, Chapter 7, pp. 7-1 to 7-21.

Hallisey, J. E. and G. H. Belt, 1996. Relationships Between ParticleMovement and Channel Morphology in Some Northern IdahoStreams. Water Resources Bulletin 33(2):383-391.

Halpern, R. A., 2000. The New Watershed Tools: Genuine Steel orChrome-Plated Plastic? Water Resources IMPACT 2(6):23-26.

Harrelson, C. C., C. L. Rawlins, and J. P. Potyondy, 1994. StreamChannel Reference Sites: An Illustrated Field Guide to FieldTechnique. Gen. Tech. Rep. RM-245, USDA Forest Service,Rocky Mountain Forest and Range Experiment Station, FortCollins, Colorado, 61 pp.

Heede, B. H., 1980. Stream Dynamics: An Overview for Land Man-agers. Gen. Tech. Rep. RM-72, USDA Forest Service, RockyMountain Forest and Range Experiment Station, Fort Collins,Colorado, 26 pp.

Hicks, B. J., J. D. Hall, P. A. Bisson, and J. R. Sedell, 1991.Responses of Salmonids to Habitat Changes. In: Influences ofForest and Rangeland Management on Salmonid Fishes andTheir Habitats, William R. Meehan (Editor). American FisheriesSociety Special Publication 19, Bethesda, Maryland, pp. 483-518.

Hirsch, P. J. and A. D. Abrahams, 1981. The Properties of Bed Sedi-ments in Pools and Riffles. Journal of Sedimentary Petrology51(3):757-760.

Horner, N. J., 1993. A Tale of Two Rivers. Idaho Wildlife 13(5):4-6.Kaufmann, P. R., P. Levine, E. G. Robinson, C. Seeliger, and D. V.

Peck, 1999. Quantifying Physical Habitat in Wadeable Streams.EPA/620/R-99/003, U.S. Environmental Protection Agency,Washington, D.C.

Kondolf, G. M., 1997. Application of the Pebble Count: Notes onPurpose, Method, and Variants. Journal of the American WaterResources Association 33(1):79-87.

Lane, E. W., 1954. River-Bed Scour During Floods. American Soci-ety of Civil Engineers Transaction 89, pp. 1069-1079.

Lane, E. W., Carl Brown, G. C. Gibson, et al., 1947. Report of theSubcommittee on Sedimentation Terminology. Transactions,American Geophysical Union 28(6):936-938.


Kappesser

Leopold, L. B., 1992. Sediment Size That Determines Channel Mor-phology. In: Dynamics of Gravel-bed Rivers, P. Billi, R. D. Hey,C. R. Thorne, and P. Tacconi (Editors). John Wiley and SonsLtd., New York, New York, pp. 297-311.

Leopold, L. B., 1994. A View of the River. Harvard UniversityPress, Cambridge, Massachusetts, 298 pp.

Lisle, T. E., 1979. A Sorting Mechanism for a Riffle-Pool Sequence.Geological Society of America Bulletin 90:1142-1157.

Lisle, T. E., 1982. Effects of Aggradation and Degradation on Riffle-Pool Morphology in Natural Gravel Channels, NorthwesternCalifornia. Water Resources Research 90-6:1643-1651.

Lisle, T. E., 1987. Using “Residual Depths” to Monitor Pool DepthsIndependently of Discharge. Research Note PSW-394, USDAForest Service, 4 pp.

Lisle, T. E. and S. Hilton, 1992. The Volume of Fine Sediment inPools: An Index of Sediment Supply in Gravel-Bed Streams.Water Resources Bulletin 28(2):371-383.

Lisle, T. E, F. Iseya, and H. Ikeda, 1993. Response of a ChannelWith Alternate Bars to a Decrease in Supply of Mixed-Size BedLoad: A Flume Experiment. Water Resources Research29(11):3623-3629.

Madej, M. A., 1982. Sediment Transport and Channel Changes inan Aggrading Stream in the Puget Lowland, Washington, inSediment Budgets and Routing in Forested Drainage Basins.Gen. Tech. Report PNW-141, USDA Forest Service. pp. 97-108.

Marcus, W. A., S. C. Ladd, J. A. Stoughton, and J. W. Stock, 1995.Pebble Counts and the Role of User-Dependent Bias in Docu-menting Sediment Size Distributions. Water ResourcesResearch 31(10):2625-2631.

McCain, M. E., D. D. Fuller, L. M. Decker, and C. K. Overton, 1990.Stream Habitat Classification and Inventory Procedures forNorthern California. FHR Currents No. 1, USDA Forest ServiceSouthwest Region, San Francisco, California.

Pfankuch, D. J., 1975. Stream Reach Inventory and Channel Sta-bility Evaluation. R1-75-002, USDA Forest Service, GovernmentPrinting Office, No. 696-260/200, Washington, D.C., 26 pp.

Rosgen, D. L., 1996. Applied River Morphology, Wildland HydrologyBooks, Pagosa Springs, Colorado.

SAS Institute Inc., 1988. SAS/STAT User's Guide, Release 6.03 Edi-tion. SAS Institute Inc, Cary, North Carolina, 1028 pp.

Schumm, S. A., 1977. The Fluvial System. John Wiley and Sons,New York, New York, 338 pp.

Steel, R. G. D. and J. H. Torrie, 1980. Principles and Procedures ofStatistics: A Biometrical Approach (Second Edition). McGraw-Hill Book Co., New York, New York, 633 pp.

U.S. Fish and Wildlife Service (USFWS), 1998. Determination ofThreatened Status for the Klamath River and Columbia RiverDistinct Population Segments of Bull Trout. Federal Register63(111): 31647-31674.

Wolman, M. G., 1954. A Method of Sampling Coarse River BedMaterial. Transactions of the American Geophysical Union35(6):951-956.

Yang, C. T., 1971. Formation of Riffles and Pools. Water ResourcesResearch 7(6):1567-1574.



a riffle stability index to evaluate sediment ...riffle particles smaller than the dominant large...

Documents