Location

The Boston Mountains are in north-central Arkansas and northeastern Oklahoma, extending approximately 200 miles from Independence County, Ark., westward to Muskogee County, Okla., and approximately 40–50 miles from the Ozark Highlands southward to the northern margin of the Arkansas Valley. The Boston Mountains are located within the Boston Mountains physiographic section (fig. 1; Fenneman, 1938). Generally, the Boston Mountains physiographic section lies within the Boston Mountains U.S. Environmental Protection Agency (EPA) Level III Ecoregion with inclusions in the Ozark Highlands and Arkansas Valley (Omernik, 1987). The EPA has further divided the study area into six Level IV Ecoregions (fig. 1B; EPA, 2010; Omernik and Griffith, 2014).

The Boston Mountains in Arkansas are contained within nine watersheds, or hydrologic units, that are part of the Arkansas-White-Red regional (2-digit code) hydrologic unit (table 2; fig. 2). The Boston Mountains are within two subregional hydrologic units—the Upper White and the Lower Arkansas. The percentages of the Boston Mountains in Arkansas within specific hydrologic units (table 2) were determined by extracting the hydrologic unit code (HUC) shapefile (U.S. Geological Survey, 2016) using extract by mask in ArcGIS Pro to the area contained within the Boston Mountains physiographic section (Fenneman, 1938; Esri, 2023). Areas of the extracted HUCs were divided by the total area of the Boston Mountains physiographic section (fig. 2). The Boston Mountains are primarily within the Beaver Reservoir (HUC 11010001; 23 percent), the Buffalo (HUC 11010005; 20 percent), the Little Red (HUC 11010014; 19 percent), and the Dardanelle Reservoir (HUC 11110202; 12 percent) hydrologic units.

Topography

The Boston Mountains in Arkansas are the eroded and uplifted southern extent of the Ozark Plateau. This mountain range spans an area that is about 35 miles north to south and about 200 miles east to west (Purdue, 1907). The relief, or differences in elevation from the valley floors to the ridgetops, generally ranges from 300 to 1,500 feet (ft) (Maxfield, 1964). The maximum elevation of the range is at Wahzhazhe Summit (formerly Buffalo Lookout, 2,561 ft, North American Vertical Datum of 1988 [NAVD 88]; USGS, 1967). The Boston Mountains are a plateau that is dissected by numerous streams that have cut deep, narrow stream valleys with large streambeds ranging in elevation from 260 to 850 ft, NAVD 88 (McKeown and others, 1988). Along the northern border of these mountains is an escarpment with many valleys eroded by northward-flowing streams. “This escarpment is highest in its middle portion and gradually falls off eastward and westward to the borders of the area” (Purdue, 1907, p. 1). To the south, the mountains gradually slope into the Arkansas Valley (Purdue, 1907).

Geology

All the rocks exposed within the Boston Mountains are of sedimentary origin and consist of sandstones, shales, and limestones. The sandstones and shales were formed of muds, sands, and organic material that were transported from adjacent land areas by streams and deposited over the bottom of the sea when the Boston Mountains and neighboring regions were beneath sea level. Deposited fine-sediment muds containing a large amount of carbonaceous matter subsequently became consolidated, forming the shales. Coarse material (loose sand) became consolidated, forming the sandstones. The limestones were formed largely of shells and other parts of animals that lived in the seas at that time. The rocks are consolidated but not metamorphosed. No igneous or volcanic rocks occur at the surface within the Boston Mountains. The formations that occur at the surface in the Boston Mountains are all Carboniferous. The northern edge of the Boston Mountains is Mississippian, and most of the southern area is predominantly Pennsylvanian (fig. 3; Purdue, 1907). The Arkansas Geological Survey 1:500,000-scale geologic map of Arkansas, including a stratigraphic column of the geological formations underlying Arkansas, as well as the formation age, geologic history, distribution, and formation description, is provided in Haley and others (1993) (fig. 4).

The rocks of the northern and middle parts of the Boston Mountains are mainly horizontal. The structure of the southern portion of the Boston Mountains is monoclinal, with the rocks dipping southward, generally at a low but perceptible angle. Faulting occurs along the east-west line where the Boston Mountains extend into the Arkansas Valley. The downthrow of these faults is on the south side of the mountains (Purdue, 1907).

The endurance of the Boston Mountains’ elevation relative to the erosion observed along their northern boundary and the southern boundary of the Springfield-Salem Plateaus is because of the flat anticline structure of the Boston Mountains. The geologic structure of the mountains determined the location of the drainage divide, and the principal agents of erosion were headwater streams. In some areas of the mountains, massive beds of sandstone that were resistant to erosion also contributed to the preservation of the mountains. Notably, the thick ledges of the Atoka Formation formed steep slopes to the north and south of the divide. Steep slopes in the Boston Mountains caused the streams that flow northward and southward from the divide to be swift, resulting in entrenched floodplains and channels without substantial lateral cutting or meandering. Alternating layers of hard (resistant) and soft (erosive) rock have resulted in numerous waterfalls and rapids with outcropping of the harder layers. Streams primarily developed in the direction of the dip and strike of rocks with few streams developing in other directions (Purdue, 1907).

Soils

Soils develop horizons that form because of persistent physical and chemical weathering processes acting on the parent material. Soils that form in similar parent material, age, topography, and climate have soil horizons that are similar in texture, structure, colors, and thickness. The soils in the study area are primarily Ultisols that are intensively weathered and characterized by low fertility. Soils in this order are acidic because of long periods of weathering during the Pleistocene and Holocene Epochs (Hoelscher and others, 1975). These soils form in humid climates under pine-hardwood forests and are generally moist throughout the year. The soils are strongly leached and are generally of medium texture and moderate permeability (Steila and Pond, 1989). Within the Boston Mountains, soils are of the suborder Udults (Nofziger, 2000). They are stony and nonstony, with medium texture and siliceous or mixed mineralogy. Ridgetops, benches, and upper slopes are well drained, shallow, and moderately deep (Mountainburg and Linker series). Middle and lower slopes and concave inter-ledge positions are well-drained, deep Paleudults (Nella series) and Hapludults (Enders series). Stream floodplains are Udifluvents (Ceda series) and Hapludults (Spadra series), and valley terraces are Fragiudults (Leadvale and Taft series) and Hapludults (Pickwick series) (Nofziger, 2000).

Land Cover and Population

Oak and hickory forests are the major land cover, with pasture and hay lands within broader stream valleys. Northern red oak (Quercus rubra), southern red oak (Quercus falcata), white oak (Quercus alba), and hickories (Carya spp.) typically dominate the uplands, but shortleaf pine (Pinus echinata) grows on drier, south- and west-facing slopes underlain by sandstone. Pasture or hay lands occur on nearly level ridgetops, benches, and valley floors (EPA, 2010). About 14 percent of the Boston Mountains study area has been cleared of natural vegetation for agricultural use. The major agricultural farming activities are pasture and hay land according to the National Land Cover Database (NLCD; Dewitz, 2023). From 1973 to 2000, net forest land cover was reduced by 1.7 percent, mechanically disturbed (clear cut or cleared) increased by 1 percent, and agriculture increased by 0.7 percent (Karstensen, 2009).

The Boston Mountains are generally sparsely populated; the largest population center, the Fayetteville metropolitan area (547,000 population), is located along the northwestern edge of the Boston Mountains (fig. 1B; U.S. Census Bureau, 2020). Boone, Carroll, Independence, Madison, Pope, and Washington Counties (fig. 2B) increased in population from 2010 to 2020, while the population in Cleburne, Conway, Newton, Searcy, Stone, and Van Buren Counties (fig. 2B) either remained stable or decreased (U.S. Census Bureau, 2020).

Climate

The mean annual temperature (1895–2013) for Fayetteville, Ark., was 58 degrees Fahrenheit (°F). The mean temperature for April through September was 71 °F, and the mean temperature for October through March was 45 °F (1895–2013). Extended warm and humid periods were common in summer (Office of the Arkansas State Climatologist, 2014).

Annual precipitation totals (1895–2013) ranged roughly from 45 to 55 inches (in.) (Office of the Arkansas State Climatologist, 2014). Precipitation results from middle latitude cyclones (lows), with warm, cold, and other frontal situations; tropical lows from the Gulf of America; and thunderstorms, or orographic uplift, caused by hills and mountains. Rainfall was generally abundant throughout the year. The dry months in Fayetteville were January and February, averaging 2.5 and 2.4 in. of precipitation, respectively. The number of days with measurable precipitation averaged about 100 per year. Most of the precipitation fell as rain. Heavy local storms that result in precipitation totals from 5 to 10 in. over extensive areas are common. During fall, winter, and early spring, precipitation events are usually less intense and of longer duration than during the summer. The mean annual snowfall total was 5 in. Snowfall was generally light and remains on the ground only briefly, but rare winter storms do occur with accumulations of as much as 10 in. during a 24-hour period. Ice storms were also infrequent but can be severe. On average, 26 tornadoes were reported each year (1950-2013) and generally occurred during the spring months (Office of the Arkansas State Climatologist, 2014; U.S. Climate Data, 2024).

Reach Selection

At various times over the past 80 years, the USGS has maintained and operated approximately 40 streamgages throughout the Boston Mountains. Based on the criteria listed below, 14 study reaches associated with these streamgages were selected for analysis (table 1; fig. 1).

Land Cover and Use Analysis

Land cover is the vegetation and physical material covering the Earth’s surface. Data from the NLCD 2021 (Dewitz, 2023) were used to determine the land cover and use of each of the watersheds above the selected stream reaches. The watersheds of the selected study reach streamgages were clipped from the NLCD shapefile by using the Extract by Mask tool in ArcGIS Pro (Esri, 2023). The percentage of each landcover class was determined by dividing the area of each class by the watershed area. These data allow for the extrapolation of our findings to similar land cover and land use areas within the Boston Mountains.

Longitudinal Profile and Cross-Section Surveys

Topographic surveys of stream longitudinal profiles and cross sections were conducted at each study reach to obtain information on the thalweg and bankfull slopes and cross-sectional hydraulic geometry. Reaches were surveyed at 3 to 12 cross sections, with a mean and median of approximately 6 cross sections ranging in length from 163 to 7,540 ft. Distances from the first to the last cross section ranged from 5.2 to 83 bankfull widths, with a median of 13.3 and a mean of 16.8 bankfull widths. Surveys at smaller streams were conducted by using traditional optical level, rod, and tape surveying equipment (±0.005 ft accuracy) with elevations determined by using geodetic surveying equipment (±0.1 ft accuracy; Rydlund and Densmore, 2012). Larger stream channels were surveyed by using a combination of traditional optical survey equipment and a kayak- or canoe-mounted sonar-measured bathymetry with Global Positioning System horizontal positioning (±0.3 ft vertical and 1.5 ft horizontal accuracy) and 1-meter (m) digital elevation model lidar surveys (±0.5 ft vertical accuracy; Arkansas GIS Office, 2017; Kroes and Ruhl-Whittle, 2025), and elevations at cross sections were determined from geodetic surveys. Each topographic survey measured the location and elevation of points along the thalweg and bankfull profiles and along selected riffle and pool cross sections. Longitudinal profiles were acquired above and (or) below the streamgage location for a total distance of at least 20 bankfull widths. All bankfull indicators that could be located and surveyed were measured and included points on both the left and right banks. Stage elevations associated with streamflows at the 1.5-year recurrence interval were used to aid in the identification of bankfull indicators during stream surveys. Field identification of bankfull indicators was cross referenced to cross-sectional plots and compared with multiple cross sections from each reach for final bankfull determinations (Kroes and Brinson, 2004). Cross-sectional surveys were acquired up to an elevation high enough to include the flood-prone elevation (twice the maximum bankfull depth; Leopold and others, 1964; Rosgen, 1996). The following channel geomorphic metrics and measurements were used to classify the studied stream reaches.

The particle size transported at bankfull discharge was compared with the measured streambed material to determine the percentage of the streambed that may be mobilized during a bankfull, 1.5-year flood discharge.

Streambed Material Particle Sampling

Streambed material sampling was conducted on eight streams to develop particle-size distribution plots from which particle-size quantile values, descriptive statistics of particle-size distributions, and particle-size distribution percentages of substrate type were calculated to determine the shapes of the individual particles composing the streambeds for each study reach. Grain-size ranges given for streambed material particle-size ranges and percentages of substrate type were based on the modified Wentworth scale (American Geological Institute, 1982). Information on streambed material particle-size distribution is a parameter used in the Rosgen (1994) stream reach classification system. Streambed material particle sizes were measured by using two methods: (1) a modified Wolman pebble count (Wolman, 1955) and (2) a sieve analysis of bar samples.

The modified Wolman pebble counts were conducted across the riffles and pools within each study reach (Harrelson and others, 1994). An observer waded the stream by using a step-toe procedure to collect and measure approximately 100 streambed material samples at each riffle and pool. Materials from the active streambed, defined as the area between the toes of the left and right bankfull terraces, were measured. For each sample selected, the longest axis (length, denoted “a-axis”), intermediate axis (width, denoted “b-axis”), and shortest axis (thickness, denoted “c-axis”) were measured and recorded (fig. 5). From the pebble count data, the bedrock tallies were removed, and cumulative frequency curves were developed. Bedrock was defined as any exposure of native solid rock in the streambed or along the streambanks. The median (D50) and one standard deviation from the median (D16 and D84) particle sizes were determined. Particle counts, cumulative frequency, descriptive statistics, and percent by substrate type for each study reach are available from Kroes and Ruhl-Whittle (2025).

The second streambed material particle-size sampling procedure was a sieve analysis of bar samples. A 5-gallon pail (approximately 55 pounds or 25 kilograms) of bar gravel was collected from the downstream face of a point bar at an elevation approximately two-thirds of the distance between the bankfull and thalweg elevations. The sample was collected by first removing the armored layer of gravel and then collecting all the particles from an area approximately 0.3 m in diameter and 0.4 m in depth. The sample was dried and weighed to the nearest 0.1 gram (g) to determine the total sample weight. The sample was placed in a nest of sieves (refer to table 3 for listing of sieve sizes used) and shaken to separate the sample. The sample from each sieve was then removed and weighed. The final total weights retained on each sieve were summed and compared to the original total weight before sieving. From the weight retained on each sieve, cumulative frequency curves were developed, and D16, D50, and D84 particle sizes were determined.

Streambed material particle-shape analysis can provide information about the particle transport history and aid facies differentiation and characterization of depositional environments. Particles were classified into four basic shapes according to the ratios of the three particle axes: the a-, b-, and c-axes (fig. 5). Sneed and Folk (1958) classified particle shapes in terms of compactness, platyness, bladedness, and elongatedness (fig. 6). Triangular diagrams (fig. 6), based on ratios of the three-orthogonal particle axis were used for unbiased presentation of primary particle-shape data (Graham and Midgley, 2000). The Sneed and Folk (1958) descriptive particle-shape data and summaries for the Wolman pebble count data are provided in Kroes and Ruhl-Whittle (2025).

Streambed particle-shape analyses were not used in the regional analysis in this report. Particle shape affects the amount of area a particle has exposed to the forces of flow, drag, and lift acting on it. This difference in shape affects particle entrainment, transport, and deposition. Consequently, two particles having the same weight and b-axis lengths but with different a- and c-axis lengths (different shapes) will respond differently to streamflow. These data were collected and computed in Kroes and Ruhl-Whittle (2025) as a means of archiving the dataset as a building block until a sufficiently larger dataset exists to further analyze streambed material particle shapes.

Stream Reach Classification

Streams of similar drainage areas may differ in width, depth, and sinuosity because of climate, geology, valley type, slope, sediment load, and (or) streambed and bank materials; however, because bankfull dimensions can characterize stream channels, streams with similar drainage areas can be classified and compared. Rosgen (1994) developed a stream reach classification system dividing streams into seven major types and dozens of subtypes, each denoted by a letter and number based on stream form and pattern. Because streams may vary in character over relatively short distance, the Rosgen (1994) classification system describes individual reaches, not the entire stream system.

Each reach was classified to the Rosgen level II stream type (Rosgen, 1994) based on the average of stream-channel metrics collected at measured cross sections and profiles. Level I classification (types A through G) describes the geomorphic characteristics at a coarse scale and is based on the entrenchment ratio and width-to-depth ratio. Level II classification (subtypes A1 to A6, B1 to B6, and so forth) provides a more detailed morphological description of the stream through additional examination of the stream pattern, profile, and streambed materials based on measured cross-section geometry, water-surface slope, and median size of the streambed material (Rosgen, 1996). Rosgen (1996, 2006) provided a means for describing deviations of measured values from the average level II values by using the following suffixes. The suffix “a” designation indicates that streams classified as a B type have a slope that is between 4 percent and 9.9 percent. The suffix “c” designation indicates that streams classified as a B type have a slope that is less than 2 percent. The suffix “b” designation indicates that streams classified as a C type have a slope that is between 2 percent and 3.9 percent. The suffix “c” designation indicates that streams classified as a C type have a slope that is less than 0.1 percent. The suffix “/1” designates the presence of bedrock within the study reach as noted during pebble counts.

Statistical Comparisons

Sediment particle-size distributions of streambed and bar material were compared for reaches where sediment was analyzed by using regression equations and correlation coefficients with the geomorphic parameters listed in the “Longitudinal Profile and Cross-Section Surveys” section, as well as 1.5-year-interval flood discharge and drainage basin characteristics available from USGS StreamStats (USGS, 2024) (drainage area, mean basin elevation, 10–85 stream slope, longest flow path, basin shape, estimated 2-year flood discharge, mean annual rainfall, mean annual runoff). Stream classifications were compared to the geomorphic and hydraulic parameters listed above using t-Tests: two- sample assuming unequal variance, one-tailed P, to determine significant (α ≤ 0.05) parameter differences between stream classifications.

Photographs

Digital photographs were taken at all cross sections. The photographs include views looking along the centerline of the cross section at the left and right banks and looking upstream and downstream from the cross-section thalweg. Representative stream reach images are available in Kroes and Ruhl-Whittle (2025).

General Geologic Characteristics of Selected Stream Reaches

Geology is the primary framework upon which fluvial processes operate, largely governing the landforms observed today. The geologic formations underlying each of the selected stream reaches are listed in table 4. Because of the differential erodibility of the rock types underlying the Boston Mountains, the ridges consist largely of sandstone and shale, while the valleys are mostly underlain by mixed sandstone, shale, and limestone. Most of the study reaches are underlain by the Atoka, Bloyd, or Boone Formations. The Atoka Formation, consisting of a sequence of sandstones and shales, underlies Mikes Creek Tributary near Ozone, Jack Creek near Winfrey, Frog Bayou at Winfrey, and Mulberry River near Mulberry (sites 1, 5, 8, and 14, respectively; fig. 1B; table 4; McFarland, 2004). The Bloyd Formation, consisting of a mix of sandstones, shales, and limestone, underlies Jones Creek at Winfrey, Big Piney Creek near Dover, and Middle Fork of Little Red River at Shirley (sites 7, 12, and 13, respectively; fig. 1B; table 4; McFarland, 2004). The Boone Formation, consisting of limestone, chert, and shale, underlies Maxwell Creek at Kingston, Smith Creek near Boxley, Buffalo River near Boxley, and Bear Creek near Silver Hill (sites 4, 6, 9, and 11, respectively; fig. 1B; table 4; McFarland, 2004). The Batesville Sandstone, Fayetteville Shale, and Pitkin Limestone underlie Trace Creek Tributary near Marshall and Richland Creek near Witts Spring (sites 2 and 10, respectively; fig. 1B; table 4). The Cane Hill Member of the Hale Formation, consisting of silty sandstone and siltstone, underlies Tick Creek near Leslie (site 3; fig. 1B; table 5; McFarland, 2004).

General Soil Characteristics of Selected Stream Reaches

Soils provide insight into the evolution, age, and stability of the landform upon which they develop. The soil series mapped by the U.S. Department of Agriculture, Natural Resources Conservation Service are arranged according to the soil parent material type for the 14 selected stream reaches (table 4). Complete descriptions of the soil series are available from Hoelscher and others (1975) and Ditzler (2017).

Residuum soils are residual soil material formed in place by weathering (Hoelscher and others, 1975; Ditzler, 2017). These soils are on the ridgetops and side slopes and constitute the upland valley floors in the larger valleys at an elevation above the floodplain or oldest alluvial terrace. Allen and Summit soils are shale residuum on upland terraces. Enders soils are shale-sandstone residuum on valley uplands. Enders and Nella soils are shale residuum on ridgetops and side slopes. Allen soils are sandstone residuum on ridgetops and side slopes. Newnata soils are limestone, shale, and siltstone residuum on hilltops and side slopes. Mountainburg soils are sandstone on hills, ridges, mountaintops, and mountainsides. Steprock soils are sandstone residuum on very steep sides of hills, mountains, and ridges.

Colluvial soils are unconsolidated sediments that have moved downslope because of gravitational forces (Hoelscher and others, 1975; Ditzler, 2017). Enders soils are shale-sandstone colluvium found on uplands, mountaintops, ridges side slopes and footslopes of mountains. Nella soils are colluvium from interbedded limestone, alkaline shale, and siltstone found on hillsides, benches, and footslopes. Newnata soils are colluvium from interbedded limestone, alkaline shale, and siltstone found on hillsides and ridges. Summit soils are colluvium on interfluves, divides, and hillslopes. Noark soils are colluvium from cherty limestones found on steep uplands. Arkana and Moko soils are residuum from limestone found on steep uplands.

Alluvium or alluvial deposits represent the most recent deposition of sediment within a watershed and are present at each of the selected stream reaches in the channel and on the floodplains (Hoelscher and others, 1975; Ditzler, 2017). Terraces are the oldest alluvial deposits, representing abandoned floodplains, and are present at most selected stream reaches. Bruno, Ceda, Iuka, Kenn, Leadvale, Leesburg, Razort, and Spadra soils are Quaternary alluvium on the floodplains and terraces.

Land Cover Above Selected Stream Reaches

Data from the NLCD of 2021 (Dewitz, 2023) were used to determine the land cover within the Boston Mountains. Forest (deciduous forest, evergreen forest, and mixed forest) is the dominant land cover comprising 78 percent of the area. Undeveloped open land (shrub/scrub, grassland/herbaceous, or pasture/hay) represented 18 percent of the area. Developed land (developed, open space; developed, low intensity; developed, medium intensity; or developed, high intensity) was the third greatest land cover representing 0.86 percent of the area.

These data (Dewitz, 2023) were also used to determine the land cover within each of the watersheds above the selected stream reaches (table 5). Forest is the dominant land cover in the selected watersheds (table 5). On average, the watersheds were 84 percent forest, ranging from 48 to 99 percent. The second largest land cover category within the selected watersheds was undeveloped open land. On average, the selected watersheds were 12 percent undeveloped open land, ranging from 0.56 to 46 percent. The third largest land cover category within the selected watersheds was developed. On average, the selected watersheds were 3.1 percent developed, ranging from 0.75 to 5.8 percent. These three land cover categories—forest, undeveloped open land, and developed—accounted for greater than 99 percent of the land cover within all selected watersheds.

No trends in watershed development, forest prevalence, or undeveloped open land were identified, but developed area of the watersheds ranged from 0.75 to 5.8 percent. The watersheds above sites 2, 5, 10, and 12 had greater than 4 percent developed area, and watersheds above sites 11, 13, and 14 had less than 2 percent developed area. The watersheds above sites 3, 6, and 14 had the greatest forest cover (>93 percent) and the watersheds above sites 5 and 12 had less than 65 percent coverage. The watersheds above sites 3, 6, and 14 had less than 3.3 percent undeveloped open land, and those above sites 2, 5, and 12 had greater than 18 percent undeveloped open land.

Streambed Material Analysis

An example of the results of particle data analysis is shown in figure 7 for Smith Creek near Boxley, Ark. (07055650; site 6; fig. 1; table 6). The computed streambed material particle-size classes (D16, D50, and D84 percent), along with the percentage of bedrock for all sampled cross sections, were used for subsequent statistical analyses relating geomorphic characteristics to basin characteristics and in the determination of the stream reach classification. Streambed material particle-size distributions were determined for sites 1, 2, 3, 6, 7, 8, 10, and 13 (table 6). The same analysis was conducted for the point bar material particle-size distributions. Particle-size quantile values of percent finer, descriptive statistics of particle-size distributions, and the particle-size distribution percentages of substrate type for cross sections and point bar material are available in Kroes and Ruhl-Whittle (2025).

Bedrock was identified as streambed material in seven of the eight streams sampled for particle-size distribution. Streambed mean particle size ranged from 30 to 99 millimeters (mm), with an overall mean of 73 mm (Kroes and Ruhl-Whittle, 2025). On average, 68 percent of the measured streambed sediment was at or below the size that may be mobilized during a bankfull discharge. Although analysis was limited by small sample sizes (N = 8), streambed material particle size decreased nonsignificantly with mean watershed elevation (coefficient of determination [R²] = 0.26; probability value [p-value] = 0.25) and exhibited finer sediments with increased entrenchment (R² = 0.33; p-value = 0.18). Observed streambed particle size correlated significantly with calculated shear stress (R² = 0.80; p-value = 0.006) and calculated width of particle transported at bankfull discharge (R² = 0.80; p-value = 0.007).

By using Sneed and Folk (1958) particle-shape classification, we determined that bed materials at sites 1, 2, and 7 were primarily very platy and very bladed (Kroes and Ruhl-Whittle, 2025).

Other sites had more varied particle shapes, but the data formed a centroid between bladed and very bladed. This distribution of particle shapes is likely due to sediment origin as eroded sedimentary layers.

Measured bar sediment particle means ranged from 0.6 to 16 mm, with an overall mean of 12 mm. Bar sediment size correlations were also limited by small sample size (N = 8) but exhibited decreasing size with increasing sinuosity (R² = 0.32; p-value = 0.14) and entrenchment (R² = 0.31; p-value = 0.16). Increases in bar material size were correlated with increasing streambed material size (R² = 0.45; p-value = 0.067). However, these results were not statistically significant.

The calculated particle sizes transported at bankfull ranged from 19 to 256 mm (table 6). These calculated values correlated significantly with observed bed sediment size distributions (D50 [R² = 0.73; p-value = 0.007] and D84 [R² = 0.86; p-value = 0.0009]). Of the eight reaches where bed sediment was measured, four were dominated by cobble. Those four cobble-dominated sites had calculated particle sizes transported at bankfull greater than 100 mm. Extrapolation of these observations and regressions to sites where bed sediment was not measured suggests that gravel may be the dominant particle size at sites 4 and 13 and cobble may be dominant at sites 5, 9, 11, and 14.

Drainage basin land cover classes generally did not correlate well with mean streambed particle sizes; however, the strongest nonsignificant correlation was between increasing drainage basin percentages of emergent herbaceous wetlands and decreasing mean streambed particle size (R² = 0.29; p-value = 0.17). Land cover correlated well with mean bar particle sizes. The percentage of area in the watersheds that was emergent wetland increased nonsignificantly with finer distributions of bar particles (R² = 0.46; p-value = 0.06). Likewise, increases in finer bar material were nonsignificantly related to increasing percentages of open water (R² = 0.47; p-value = 0.06). Mixed forest (R² = 0.38; p-value = 0.10) and developed medium intensity (R² = 0.35; p-value = 0.12) also nonsignificantly increased as bar materials became finer.

Geomorphology at Selected Stream Reaches

The channel shape, or geomorphology, was measured at each of the selected stream reaches. The geomorphological attributes measured or calculated at each of the selected stream reaches include the bankfull top width and mean depth, the flood-prone width, and the water-surface and valley slopes. The ratios of the bankfull top width to mean depth (width-to-depth ratio), the flood-prone width to bankfull top width (entrenchment ratio), and the water-surface slope to the valley slope (sinuosity) allow comparisons to be made between watersheds of different sizes. The width-to-depth ratio from the selected stream reaches ranged from 14.8 to 78.4, averaging 32.1. The entrenchment ratio ranged from 1.58 to 9.60, averaging 3.35. The sinuosity from the selected stream reaches ranged from 1.05 to 2.25, averaging 1.19. The stream reach width-to-depth ratio, entrenchment ratio, sinuosity, and level II classification (Rosgen, 1994) are presented in table 7.

Classification of Selected Stream Reaches

According to the Rosgen (1994) level II stream system, half of the sites were classified as B-type streams, and half were classified as C-type streams. Comparisons of geomorphic parameters were made using two-sample, one-tailed t-tests assuming unequal variance to determine differences in stream geomorphology between the two classifications. B-type streams had lower upstream watershed (10-85) slope (means of 79 and 252 ft/mi, respectively; p-value = 0.048), greater hydraulic radii (means of 3.01 and 4.99 ft, respectively; p-value = 0.044), and depths (means of 3.26 and 5.14 ft, respectively; p-value = 0.054) than C-type streams. C-type streams had finer D50 bar material size fractions (means of 6.04 and 15.15 mm, respectively; p-value = 0.051) and D85 (means of 22.7 and 55.7 mm, respectively; p-value = 0.048) than B-type streams.

These stream types were also compared for drainage basin land cover. B-type streams had less barren land (rocks, sand, and clay (means of 0.002 and 0.081 percent, respectively; p-value = 0.038) than C-type streams. B-type streams also had less woody wetlands (means of 0.015 and 0.050 percent, respectively; p-value = 0.098) and emergent herbaceous wetlands (means of 0.002 and 0.017 percent, respectively; p-value = 0.036) than C-type streams.

Regional Hydraulic Geometry Relations

Regional hydraulic geometry curves were constructed by plotting bankfull discharge and measured bankfull geometry dimensions (cross-sectional area, mean depth, and top width) from stable reaches and the associated bankfull streamflow against the contributing drainage area (table 7; fig. 8). Regression equations were derived from these hydraulic geometry curves and express the mathematical relation (power functions, Y = aX^b) between the bankfull channel dimensions (Y) and the contributing drainage areas (X). The regression equations and corresponding 95-percent confidence and prediction intervals are presented on the regional hydraulic geometry curves (fig. 8). The 95-percent confidence intervals define a range of values that have a 95-percent probability of encompassing the results for other B- or C-type streams within the Boston Mountains physiographic section. The prediction intervals predict the 95-percent probability ranges for estimates of channel dimensions for a single stream of a given drainage area in the Boston Mountains physiographic section. The drainage areas of the studied stream reaches ranged from 0.19 to 373 square miles.

The regression equations for bankfull streamflow (eq. 3), bankfull channel cross-sectional area (eq. 4), mean depth (eq. 5) and top width (eq. 6) as a function of the contributing drainage area for streams in Boston Mountains in Arkansas are