Mapping Study Reach

Changes in channel characteristics and planform were assessed with repeat geomorphic mapping. This study component focused on a 97.2-km-long reach of the Siletz River from Elk Creek to Millport Slough (rkm 104.3–7.1) for 2 periods (calendar years 1939 and 2016) with additional mapping periods (1994, 2000, and 2009) for 7 smaller sub-reaches between Palmer Creek and the City of Siletz (rkm 87.2–68.5; fig. 3; table 1). The upstream drainage area of Siletz River is 224 km² at Elk Creek, 298 km² at Palmer Creek, and 782 km² at Millport Slough. Tributaries in the study area with drainage areas of more than 30 km² include Sunshine (31 km²), Mill (34 km²), Rock (111 km²), Sam (39 km²), Euchre (35 km²), and Cedar (35 km²) Creeks (U.S. Geological Survey, 2019).

Mapping Methods

Landforms greater than 250 m² in area within the active channel were mapped at a scale of 1:2,500 in a geographic information system (GIS; ArcGIS™ 10). The active channel is defined as the area typically inundated during annual high discharge events and includes the low-flow main channel, secondary water features (for example, ponds and alcoves), bedrock, and gravel bars (both vegetated and unvegetated). Channel maps were repeatedly verified to ensure consistent delineation of features among years and throughout the study area following the protocol of Wallick and others (2011). These maps (Gordon and others, 2021) provide an inventory of the landforms and channel features from which to assess spatial and temporal changes in the location and areal coverage of landforms, such as gravel bars.

Geomorphic maps for the Siletz River from Elk Creek to Millport Slough (rkm 104.3–7.1; fig. 3; table 1) were developed using aerial photographs from 1939 and 2016. Scanned copies of the black and white photographs collected in 1939 were acquired from the University of Oregon Map and Aerial Photography Library. These photographs were geo-referenced using the U.S. Department of Agriculture’s National Agriculture Imagery Program one-meter resolution digital orthophotographs acquired in 2016 for the study area. The rectified and mosaicked photographs from 1939 are published by Gordon and others (2021). The 2016 orthophotographs are available from the National Agriculture Imagery Program (2016).

Geomorphic maps were also made using orthophotographs collected in 1994, 2000, and 2009 (National Agriculture Imagery Program, 1994, 2000, 2009; National Agriculture Imagery Program, 2009) for seven mapping sub-reaches (about 0.5–1.0-km-long) between Palmer Creek and the City of Siletz (rkm 87.2–68.5; fig. 3; table 1). Sub-reaches were numbered 1–7 increasing upstream. Sub-reaches were selected based on where the Siletz River channel moved laterally between 1939 and 2016. Sub-reaches, except for the sub-reach near rkm 85, encompass gravel bars where particle measurements were made for this study (fig. 3).

Channel Mapping Uncertainty

Digital channel maps have many sources of uncertainty, including the quality of underlying photographs, errors introduced by geo-referencing and digitizing processes, and differences in discharge conditions at the time of aerial photograph acquisition that influence how much of the channel is underwater when mapping (Gurnell, 1997; Mount and Louis, 2005; Hughes and others, 2006; Walter and Tullos, 2010). The photographs from 1994, 2000, 2009, and 2016 were generally of adequate resolution and free of glare and shadow to facilitate detailed mapping of bars and channel features. Aerial photographs from 1939 had some glare, shadow, or low resolution along the mapping corridor. Uncertainty associated with the geo-referencing process resulted in an average root mean square error of 3.7 m for the overall study area (ranging from 0.35 to 12 m for each photograph; Gordon and others, 2021). A qualitative inspection of the location of static features (for example, bedrock) within the 1939 mosaic relative to those in the 2016 imagery indicate offset is smallest near the channel where most control points were placed (coinciding with active channel mapping) and increases towards the edge of the mosaic.

Water levels at the time of photograph collection may influence bar exposure and wetted channel width, influencing the mapping area of bar and wetted channel features. Overall, discharge was generally lowest during the 2000 photography collection (mean daily discharge of 3.0–3.4 m³/s) and then slightly greater during the 1939 (5.3 m³/s), 2016 (6.7–7.7 m³/s), and 2009 (7.5 m³/s; table 5) photograph collection. Discharge during the 1994 photography was substantially greater (13 m³/s). Discharge differences between the photographs likely have some effects on year-to-year comparisons of mapped features. In particular, the channel maps made from the 1994 photographs likely have greater wetted areas and reduced bar areas compared to channel maps from other years due to the elevated discharge during the collection of the aerial photographs in 1994. The effects of discharge and varying levels of inundation of mapped features for different discharges were not systematically quantified for this study but could be evaluated in future studies using methods described in Wallick and others (2011) or results of the 1D hydraulic model produced by this study. For comparison, the stage difference between the maximum and minimum main discharge (table 5) as determined from the stage-discharge relation (number 16.0) is about 0.28 m (0.92 ft) at the Siletz River streamgage (USGS station 14305500). The combination of water levels and tide in the Tidal reach may have also affected the mapping. Limited tidal records overlap with mapping periods for Siletz Bay, so the effect of tides on mapping was not systematically evaluated. However, the difference between mean lower low water and mean higher high water recorded over a 3-month period in 2017 for USGS station 9435992 (Chinook Bend, Siletz River) was 1.683 m (National Oceanic and Atmospheric Administration, 2023).

Mapping Results

Active Channel Characteristics, 2016

The Siletz River flows through six geomorphic reaches with varying channel characteristics (table 4) as the river traverses the steep, confined reaches to the tidally affected, low gradient reach (rkm 104.3–7.1). In 2016, the mapped area of the primary wetted channel generally increased downstream (fig. 6A). Accordingly, the mapped areas of gravel bars and bedrock fluctuated between these reaches (fig. 6B–C). Mapped bar area per rkm was the greatest in the Sam Creek, Upper Sandstone, and Lower Canyon reaches. Mapped bedrock area per rkm was the greatest near rkm 60.0 in the Mill Creek reach and throughout the Upper and Lower Canyon reaches. Mapped secondary water features were few and mostly low elevation portions of bars that were submerged at the time of the aerial photograph collection (fig. 6D). The mapped area of secondary water features did not include some side channels around islands (such as near rkm 95.8, 94.6, 88.8, 88.2, and 74.9) that were mapped as part of the main channel.

Channel characteristics from the detailed 2016 mapping are summarized here by the six geomorphic reaches along the Siletz River:

• • Upper Canyon reach from Elk to Sunshine Creeks (rkm 104.3–97.9; fig. 7A–B; table 6). From Elk Creek to the area downstream from Holman Creek (rkm 104.3–99.8), the active channel was narrow (10–55 m), had few gravel bars along the inside of channel bends and downstream from Holman Creek, and flowed over and along bedrock outcrops. From rkm 99.8–98.9, the active channel widened, ranging from 20 to 80 m. This stretch contained several large gravel bars (up to 5,360 m²) on the inside of channel bends (such as at rkm 99.7) and flanking the channel, and the only secondary water feature mapped in this reach. From rkm 98.9 to 97.9, the width of the active channel decreased, ranging from 20 to 40 m.

• • Lower Canyon reach from Sunshine Creek to downstream from Wildcat Creek (rkm 97.9–88.6; fig. 8A–C; table 6). Downstream from Sunshine Creek to rkm 94.6, the active channel widened (ranging from 20 to 80 m). Gravel bars were larger (ranging from about 250 m² up to 7,800 m²) and more abundant downstream from Sunshine Creek. This section had gravel bars on the inside of channel bends (such as at rkm 96.4) and flanking the channel as well as a mid-channel bar at rkm 96.9. In rkm 94.6–88.6, bedrock outcrops lined the relatively narrow channel (active channel width of 10–75 m), and some gravel bars were present at rkm 89.6 downstream from Wildcat Creek.

• • Upper Sandstone reach from downstream from Wildcat Creek to Rock Creek (rkm 88.6–80.4; fig. 9A–B; table 6). From rkm 88.6 to 84.2, the width of the active channel widened (ranging from 25 to 100 m). The largest bars in the reach (ranging up to nearly 18,000 m²) were present between Palmer Creek at Moonshine Park (rkm 87.4) and Baker Creek (rkm 84.6). Downstream from Baker Creek, the active channel narrowed (ranging from 10 to 50 m) and had relatively smaller lateral gravel bars (up to about 3,000 m²). Bedrock was mapped intermittently throughout the Upper Sandstone reach.

• • Sam Creek reach from Rock Creek to the City of Siletz (rkm 80.4–68.8; fig. 10A–C; table 6). The channel was relatively confined for most of the reach with bedrock intermittently visible in and along the channel. The width of the active channel ranged from 35 to 100 m. Gravel bars (ranging up to about 17,500 m²) were on the inside of channel bends, such as between rkm 80.4–79.8 downstream from the Rock Creek, rkm 76.0–75.5 near Twin Bridges downstream from Scott Creek, rkm 73.6–72.7 at Klamath Grade bar downstream from Sam and Bentilla Creeks, and rkm 70.0–69.7 at Windchime bar.

• • Mill Creek reach from the City of Siletz to the head of tide at Cedar Creek (rkm 68.8–39.5; fig. 11A–D; table 6). Active channel widths in the Mill Creek reach ranged from 10 to 70 m. The reach had intermittent gravel bars at channel bends and flanking the channel between rkm 68.9–45.1 and 42.7–39.5. Bars were generally less than 7,350 m² and relatively smaller than those mapped in the upstream reaches. Bars were generally located immediately downstream from tributary confluences, such as near Mill Creek (rkm 67.1), Thompson Creek (rkm 55.3), Euchre Creek (rkm 45.4), and Hough Creek (rkm 42.6). Bedrock was intermittently mapped along this reach.

• • Tidal reach from the head of tide at Cedar Creek to Millport Slough (rkm 39.5–7.1; fig. 12 A–G; table 6). Active channel widths in the Tidal reach ranged from 20 to 120 m. This reach had fewer gravel bars mapped in 2016 than the upstream reaches. Gravel bars were primarily located in the upper portion of the reach near the Bulls Bag area between Jaybird Creek (rkm 39.5) and rkm 33.6. Mapped gravel bars did not exceed areas of more than about 5,740 m². No bedrock was visible in this reach in the 2016 photographs.

The relative abundance of in-channel gravel along reaches of varying lengths on the Siletz River and other gravel-bed rivers of western Oregon can be compared by normalizing mapped reach-aggregated gravel bar area by reach length to determine unit bar area (total area of bars per meter of channel length [square meters per meter m²/m]). In 2016, unit bar area along the Siletz River ranged from 3.2 to 9.0 m²/m in the geomorphic reaches upstream from the head of tide and was 1.5 m²/m in the Tidal reach (table 6). These unit bar values for the Siletz River were similar to values for other rivers along the northwestern Oregon coast, such as the Kilchis, Trask, and Tillamook Rivers, that also drain Coast Range sedimentary and volcanic rocks (table 7; O’Connor and others, 2014). However, Siletz River unit bar areas were considerably less than some reaches of rivers along the southwestern Oregon coast that drain the gravel-rich Klamath Mountains, such as the Chetco, Applegate, Illinois, and Rogue Rivers where the maximum unit bar areas are 7.0–10.2 times the maximum unit bar area of the Siletz River.

Table 6.    

Summary of mapped bar characteristics as delineated by Gordon and others (2021) from orthophotographs taken in 2016 (National Agriculture Imagery Program, 2016) for river kilometers 104.3–7.1 along the Siletz River, western Oregon.

[Number of bars: Gravel bars spanning reach bounds were split and are counted separately in this table. The total number of gravel bars mapped for 2016 along the Siletz River is 173 (see tables 7 and 8). Mean mapped bar area calculated by dividing Total mapped bar area by Number of bars. Abbreviations: rkm, river kilometer; m², square meters; m, meters; m²/m, square meters per meter]

Table 6.    Summary of mapped bar characteristics as delineated by Gordon and others (2021) from orthophotographs taken in 2016 (National Agriculture Imagery Program, 2016) for river kilometers 104.3–7.1 along the Siletz River, western Oregon.

Reach	River kilometers (rkm)	Number of bars	Smallest mapped bar (m2)	Largest mapped bar (m2)	Mean mapped bar area (m2)	Total mapped bar area (m2)	Unit bar area (m2/m)
Upper Canyon	104.3–97.9	13	204	5,360	1,690	25,400	4.0
Lower Canyon	97.9–88.6	25	2.6	7,790	2,790	78,100	8.4
Upper Sandstone	88.6–80.4	24	182	17,000	2,940	73,400	9.0
Sam Creek	80.4–68.8	32	256	17,000	2,300	73,700	6.4
Mill Creek	68.8–39.5	55	261	7,350	1,720	99,600	3.4
Tidal	39.5–7.1	20	342	5,740	2,390	40,600	1.3

Table 7.    

Unit bar area mapped for predominantly fluvial (non-tidal) reaches from this study and prior studies along Oregon coastal rivers.

[Rivers are presented from north to south along the Oregon coast. Unit bar area: Area of bar in square meters per channel length in meters and is determined using the National Hydrography Dataset centerline for the Siletz, Umpqua, and Chetco Rivers whereas other studies used centerlines digitized from orthophotographs. Abbreviation: km²: square kilometers]

Table 7.    Unit bar area mapped for predominantly fluvial (non-tidal) reaches from this study and prior studies along Oregon coastal rivers.

River	Basin area (km2)	Unit bar area for predominantly fluvial reaches	Year of orthophotographs used for mapping bar area	Source
Nehalem River	2,207	18.3	2009	Jones and others (2012c)
Miami River	94	8.1 and 27.9	2009	Jones and others (2012c)
Kilchis River	169	9.2	2009	Jones and others (2012c)
Wilson River	500	12.6 and 18.6	2009	Jones and others (2012c)
Trask River	451	9.7	2009	Jones and others (2012c)
Tillamook River	156	7.1	2009	Jones and others (2012c)
Siletz River	970	3.4 to 9.0	2016	This report
Umpqua River	12,103	5.0 to 17.6	2005	Wallick and others (2011)
Coquille River	2,745	0.4 to 12.6	2009	Jones and others (2012b)
Rogue River	13,390	10.6 to 63.1	2009	Jones and others (2012a)
Applegate River	1,994	4.3 and 71.5	2009	Jones and others (2012a)
Illinois River	2,564	91.8	2009	Jones and others (2012a)
Hunter Creek	44	19.1 and 19.7	2009	Jones and others (2011)
Chetco River	914	9.3 to 77.5	2005	Wallick and others (2010)

Changes in Channel Planform and Features, 1939–2016

Between 1939 and 2016, the Siletz River from Elk Creek to Millport Slough (rkm 104.3–7.1) was primarily laterally stable in terms of planform and bar locations. Overall, the total mapped bar area, including vegetated and unvegetated bars, was similar in both 1939 and 2016 (less than 2 percent change and within mapping uncertainty; table 8). However, the number of mapped bars increased by 25 percent while the average size of mapped bars decreased by 21 percent (fig. 6B; table 9). The mapped area of secondary water features was greater in 2016 (10,500 m²) than in 1939 (8,500 m²), though it is possible that those differences were related to higher discharge depicted in the 2016 aerial photographs relative to 1939 (table 5). However, the loss of a large side channel (6,000 m²) near rkm 95.0 between 1939 and 2016 was the result of planform change; here, the Siletz River had a side channel flowing around a partially unvegetated island bar in 1939, but the river was concentrated into a single channel along the right bank in the 2016 aerial photographs.

Some of the largest net increases in mapped gravel bar area detected between 1939 and 2016 were at rkm 8.0, 30.0, 34.0–38.0, 51.0, 85.0–87.0 (fig. 6B) where bar growth or the deposition of new bars occurred in locally wide areas of the active channel. Long, thin bars mapped along the channel margins between rkm 10.0–8.0 in 2016 but not 1939 likely are evident because of tide differences between those photographs (fig13A–C). Near the south side of the Bulls Bag area and upper portion of the Tidal reach (rkm 38.0–34.0), mapped bar area increased as channel shifted laterally between rkm 36.4–34.5 (fig. 13D–F) and 39.0–38.0 (fig. 13G–I). At rkm 51.0, near the confluence of Ojalla Creek, the channel shifted laterally, and mapped bar area increased as the number of bars decreased (fig. 13J–L). Locations of historical channel migration documented by Derouin (2015; specifically, between 1939 and 2002 at rkm 51.0, between 1958, 1988, and 2012 for rkm 39.0–38.0, and between 1958 and 2012 for rkm 36.4–34.5) coincide with the lateral and planform found in this study. Results for rkm 87.0–85.0 are summarized in the following section for sub-reach 1 at Moonshine Park and sub-reach 2 upstream from Baker Creek.

Some of the largest reductions in mapped bars were at rkm 99.0, 96.0, 95.0, 79.0, 39.0, and 33.0 (fig. 6B). At rkm 99.0, long gravel bars flanked the Siletz River in 1939, but these bars became smaller and were partially submerged in 2016 (fig. 14A–C). At rkm 96.0, vegetation establishment appeared to reduce bar area mapped based on the 2016 photographs (fig. 14D–E). At rkm 95.0 and 39.0, the reduction in mapped gravel bar area is associated with vegetation establishment and the channel splitting around mid-channel bars in 2016 (figs. 14D–F, 15A–C). At rkm 33.0, bar area declined as the channel moved north slightly (fig. 15D–F). This shift in channel position was caused by the toe of a landslide on the left riverbank that pushes the channel toward the inside of the meander bend along the right riverbank (Harden, 2013).

Table 8.    

Total bar area as mapped by Gordon and others (2021) from photographs taken in 1939, 1994, 2000, 2009, and 2016 (National Aerial Photography Program, 1994, 2000; National Agriculture Imagery Program, 1994, 2000, 2009, 2016; Gordon and others, 2021) for seven mapping sub-reaches along the Siletz River, western Oregon.

[Average bar area: Average bar area calculated by dividing the total bar area for each subreach by the number of mapped bars. Abbreviations: m, meter; m², square meters; --, no data; %, percent]

Table 8.    Total bar area as mapped by Gordon and others (2021) from photographs taken in 1939, 1994, 2000, 2009, and 2016 (National Aerial Photography Program, 1994, 2000; National Agriculture Imagery Program, 1994, 2000, 2009, 2016; Gordon and others, 2021) for seven mapping sub-reaches along the Siletz River, western Oregon.

Mapping area	Sub-reach number and general location	2016 reach length (m)	Total bar area (m2)
			1939	1994	2000	2009	2016	Net % change
Full mapping area	--	97,200	399,637	--	--	--	390,765	−2%
Sub-reaches	1—Windchime bar	654	10,658	9,180	13,498	12,080	12,300	15
	2—Klamath Grade	572	9,832	6,503	8,709	6,919	7,766	−21
	3—Twin Bridges	549	15,850	10,602	15,628	13,596	14,273	−10
	4—Rock Creek	685	24,903	21,656	24,751	18,896	20,326	−18
	5—Logsden	457	4,440	2,110	5,524	4,639	5,101	15
	6—upstream from Baker Creek	615	12,162	14,360	13,978	11,255	13,090	8
	7—at Moonshine Park	753	12,370	17,902	18,192	15,660	17,269	40

Table 9.    

Number of bars and mean bar area as mapped by Gordon and others (2021) from photographs taken in 1939, 1994, 2000, 2009, and 2016 (National Aerial Photography Program, 1994, 2000; National Agriculture Imagery Program, 1994, 2000, 2009, 2016; Gordon and others, 2021) for seven mapping sub-reaches along the Siletz River, western Oregon.

[Refer to figure 3 for the location of sub-reaches. Mean bar area: Average bar area calculated by dividing the total bar area for each subreach by the number of mapped bars. Abbreviations: m, meter; m², square meters; --, no data; %, percent]

Table 9.    Number of bars and mean bar area as mapped by Gordon and others (2021) from photographs taken in 1939, 1994, 2000, 2009, and 2016 (National Aerial Photography Program, 1994, 2000; National Agriculture Imagery Program, 1994, 2000, 2009, 2016; Gordon and others, 2021) for seven mapping sub-reaches along the Siletz River, western Oregon.

Mapping area	Sub-reach number and general location	2016 reach length (m)	Number of bars						Mean bar area (m2)
			1939	1994	2000	2009	2016	Net % change	1939	1994	2000	2009	2016	Net % change
Full mapping area	--	97,200	139	--	--	--	173	24	2,875	--	--	--	2,259	−21%
Sub-reaches	1—Windchime bar	654	3	2	6	5	5	67	3,550	4,590	2,250	2,420	2,460	−31
	2—Klamath Grade	572	2	3	3	3	4	100	4,920	2,170	2,900	2,300	1,940	−61
	3—Twin Bridges	549	1	1	2	2	2	100	15,850	10,600	7,810	6,800	7,140	−55
	4—Rock Creek	685	3	2	3	4	4	33	8,300	10,800	8,250	4,720	5,080	−39
	5—Logsden	457	2	2	3	2	2	0	2,220	1,060	1,840	2,320	2,550	15
	6—upstream from Baker Creek	615	2	2	3	4	4	100	6,080	7,180	4,660	2,810	3,270	−46
	7—at Moonshine Park	753	3	3	4	5	6	100	4,120	5,970	4,550	3,130	2,870	−30

Changes in Channel Planform and Features for the Short Repeat Mapping Sub-Reaches, 1939–2016

Although the overall channel location and area of mapped bars remained similar in 1939 and 2016, distinct patterns of geomorphic change were more apparent when evaluating repeat mapping data within the short (0.5–1.0-km long) sub-reaches where channel characteristics were evaluated for multiple periods (fig. 16A–G; tables 8-9). Total mapped bar area decreased between 1939 and 2016 in sub-reaches 4 (at Rock Creek; net reduction of 18 percent), 3 (at Twin Bridges; net reduction of 10 percent), and 2 (at Klamath Grade; net reduction of 21 percent), with bars increasing in number and decreasing in average size. Total mapped bar area increased in sub-reaches 7 (at Moonshine Park; net increase of 40 percent), 6 (upstream from Baker Creek; net increase of 8 percent), and 1 (at Windchime bar; net increase of 15 percent), with bars increasing in number and decreasing in average size. Sub-reach 5 (at Logsden) had a 15 percent increase in total bar area.

Modest changes in the locations of the channels and bars were found over the 77-year mapping period for seven mapping sub-reaches (figs. 17A–J–23A–J). Channel planform adjustments were noted in relation to sediment deposition and bar growth in sub-reaches 7, 6, and 1. In sub-reach 7 (at Moonshine Park), the channel split around an island at rkm 87.4 in 1939, but then the channel on the left bank was mostly filled in with sediment by 1994, leaving a secondary channel along the left bank (fig. 17A–J). Downstream in sub-reach 6 (at Baker Creek), one bar changed from a single bar on the right bend near rkm 85 to two smaller bars—one still on the river bend and a mid-channel bar (first evident in the 2000 aerial photographs; fig. 18A–J). Sub-reaches 5 (at Logsden; fig. 19A–J), 4 (at Rock Creek; fig. 20A–J), 3 (at Twin Bridges; fig. 21A–J), and 2 (at Klamath Grade; fig. 22A–J) changed little over the mapping period. Sub-reach 1 at Windchime bar had the most substantial planform and bar change of all the mapping sub-reaches; here the thalweg shifted position several times from 1939 from 2016 (fig. 23A–J).

The repeat mapping also is helpful for identifying when vegetation (primarily woody shrubs such as willow or cottonwood) is scoured and stripped from gravel bars by high flows or when vegetation colonizes or expands on gravel bars during periods of lower peak discharge. All sub-reaches displayed evidence of scour (such as flood-swept surfaces and vegetation removal) along most bar surfaces in 2000 (figs. 17A–J–23A–J). Large swaths of woody shrubs appeared to be scoured from bar surfaces in sub-reaches 7 at Moonshine Park, 4 at Rock Creek, and 3 at Twin Bridges between the collection of the 1994 and 2000 aerial photographs (figs. 17 A–J, 20A–J–21A–J). This scour was likely associated with the largest event of record on November 26, 1999 (1,520 m³/s, 53,800 ft³/s, AEP of about 0.002). Scour appeared more limited to gravel bar margins in 2009 and 2016.

Specific-Gage Analysis Methods

Two approaches were used to evaluate long-term patterns and relatively episodic adjustments in local channel bed variability at USGS streamgage 14305500 (Siletz River at Siletz Oregon; U.S. Geological Survey, 2021b; fig. 1):

Specific-Gage Analysis Uncertainty

The estimates of changes in channel bed elevation using specific-gage analyses have multiple sources of uncertainty. Uncertainty and error related to discharge measurements (Turnipseed and Sauer, 2010) and stage measurements (Sauer and Turnipseed, 2010) at streamgages may stem from the method or equipment used to make the measurements, changes in methods or equipment over time, and changes in site conditions. However, the USGS Office of Surface Water sets accuracy standards for these field measurements; the standard for stage measurements, for example, is plus or minus (±) 0.01 ft (0.003 m, or 0.2 percent of effective stage; Sauer and Turnipseed, 2010). Additionally, the stage-discharge rating curve is used to interpolate discharges between measurements. Moreover, the creation of new ratings curves is subject to the hydrographer’s judgement; some hydrographers may create new ratings more readily than others. As such, ratings might not be changed for years and inaccurately represent stability at a site, whereas ratings that might be changed frequently could imply more instability than is warranted. This source of uncertainty primarily pertains to method 1 used in the specific gage analysis. Another source of uncertainty that primarily pertains to method 2 of the specific gage analysis is that periods of backwater effects from beaver dams, large wood, or human activities may cause a field measurement to deviate from a rating. Uncertainty relating to specific-gage analyses was not evaluated for this study. Findings presented here are one line of evidence in understanding historical and recent bed-elevation changes on the Siletz River near USGS streamgage 14305500. However, streamgage locations are commonly chosen based on stability, and results from the specific gage analyses may not be representative for the whole river within the study area. The general interpretations of channel bed elevation that follow are likely representative of real patterns in the channel bed elevation given the long period of record and multiple measurements evaluated.

Specific-Gage Analysis Results

Comparison of Stage at Specific Discharges across Multiple Rating Curves

USGS streamgage 14305500 on the Siletz River has been at its current location (rkm 71.4) and 31.19 m (102.32) ft above the National Geodetic Vertical Datum of 1929 since October 1938. Prior to that time, the station was installed at different locations and different datums within 4.0 km of the current location. These locations had similar stream characteristics. As early hydrographers described the site, “The bed is of coarse gravel and sand, free from vegetation and shifting. One channel at all stages.” (R.W. Davenport, U.S. Geological Survey, written commun., April 19, 1912). From 1905 to 1912, a non-recording gage shows a relatively stable stage-discharge relation except for a short period during 1910 where low discharges were affected by backwater from a dam built to collect lamprey (fig. 24; circle A). The streamgage was at three different locations between January 1924 and September 1938 and had a maximum change in stage of 0.09 m at the 5 percent exceedance probability (fig. 24; circle B).

Since USGS streamgage 14305500 was moved to its current location and datum in September 1938, the relation between stage and discharge has varied with specific discharge and time (fig. 24; circle B). Between October 1938 and October 1941, the rating was relatively stable. In October 1941, increases in the rating ranged between 0.01 m at the lowest discharges to 0.22 m at slightly greater discharges. Between October 1941 and January 1970, net decreases in stage were observed for all discharges, though stage appeared to decline more at the highest discharges evaluated (0.48 m at 159 m³/s) compared to lower discharges (0.07 m at 2.3 and 2.9 m³/s). From January 1970 to February 1996, minor stage changes were observed for all analyzed discharges with net changes ranging from 0.0 to 0.03 m. In February 1996, stages increased by 0.06 m at 152 m³/s to 0.13 m at 2.3 m³/s (fig. 24; circle C); this sharp increase in stage (and likely local channel aggradation) was probably caused by channel changes associated with the fifth largest peak discharge event on record that occurred on February 7, 1996. In January 2002, stage decreased by 0.07–0.12 m (back to levels similar to what was observed before February 1996) for the 75 and 95 exceedance discharges (fig. 24; circle D), whereas more gradual decreases in stage occurred through March 2007 for the 5, 25, and 50 exceedance discharges with net changes ranging from 0.07 to 0.08 m (fig. 24; circle E). From March 2007 to April 2021, stage has been relatively stable through the current rating (21.0 as of April 12, 2021).

Comparison of Stage Field Measurements to a Recent Rating Curve

More than 200 field measurements of stage (n=236) have been recorded at discharges less than 152 m³/s since 1979 (fig. 25). Residuals between field measurements and stages predicted from the current rating curve (21.0) range from −0.15 to 0.14 m, where negative values indicate stages lower than current rating prediction and positive values indicate stages higher than the current rating predication. Removing seven outliers classified as “poor” quality measurements during field visits (not shown in fig. 25) shifts the minimum residual to −0.05 m and has a median difference from rating curve 21.0 of 0.02 m (absolute differential values). From 1979 to 1996, most measurements fall within ±0.08 m (about the size of a cobble) of the current rating. Following the peak flow event on February 7, 1996 (fig. 5), stage residuals are as much as 0.11 m greater than predicted from the current rating. From 1996 to about 2007, stage residuals declined. From 2007 to 2019, stage residuals approached zero, indicating bed stability. This finding aligns with the recent bed stability indicated by the comparison of the multiple stage-discharge rating curves over time (fig. 24). Some measurements made in 2020 were affected by debris and backwater (J. Flinn, U.S. Geological Survey, written commun., April 21, 2020), likely accounting for deviations from the current rating of as great as 0.13 m.

Specific-Gage Analysis Discussion

The relation between stage and discharge at USGS streamgage 14305500 on the Siletz River has varied over time. The relatively small increases in stage associated with October 1941 possibly resulted from incorporation of more discharge measurements. The prominent changes in 1996 and 2002 were likely caused by floods. A sharp increase in stage (and likely local channel aggradation) in February 1996 is probably attributable to the February 7, 1996, peak flood of 982 m³/s (34,700 ft³/s; exceeding 0.1 AEP), which is the fifth largest annual peak discharge event on record (figs. 5, 24–25). This event also caused aggradation at USGS streamgages on the Wilson (U.S. Geological Survey streamgage 14301500; Wilson River near Tillamook, Oregon; Jones and others, 2012c) and Chetco Rivers (U.S. Geological Survey streamgage 14400000; Chetco River near Brookings, Oregon; Wallick and others, 2010). The January 2002 shift in the stage-discharge relation (and potential lowering of local bed elevations) follows moderate floods in WY 2022 (the largest of which was 515 m³/s [18,200 ft³/s] on January 8, 2002), but the updated rating could also indicate some adjustments related to earlier peak flood events that had occurred since the previous rating update in October of 1998.

However, other changes in the stage-discharge rating curve do not appear to coincide with individual floods, even though several large floods occurred during the period of record. For example, the stage-discharge relation was relatively stable between 1970 and 1996 (figs. 24–25). During this time, the maximum annual peak flood was 900 m³/s (31,800 ft³/s) on January 11, 1972, followed by 750 m³/s (26,500 ft³/s) on December 25, 1980. Additionally, the stage-discharge relation was relatively stable from March 2007 to April 2021 despite several large peak flow events, including the third and eighth largest events on record (1,070 m³/s [37,700 ft³/s] on January 19, 2012, and 912 m³/s [32,200 ft³/s] on December 7, 2015). The lack of response in the stage-discharge relation to floods between 2007 and 2021 may be attributable to differences in flood duration and local deposition dynamics (as hypothesized for the Wilson River streamgage [14301500; Wilson River near Tillamook, Oregon] by Jones and others, 2012c), shape of the hydrograph, event timing, or available sediment supply. More recent deviations between the stage-discharge rating curve and field measurements in 2020 are likely attributed to large wood accumulations in the channel creating backwater conditions (J. Flinn, U.S. Geological Survey, written commun., April 21, 2020) and are not indicative of aggradation.

Bed-Material Particle Size Methods

The USGS and the Confederated Tribes of Siletz Indians of Oregon collected surficial particle size data at a total of 11 sites along the Siletz River in August 2017 and July 2018 (“surficial bar particle counts”; fig. 3; Jones and Keith, 2021). A single transect was collected at nine different bar sites (eight in 2017 and one in 2018). Additionally, two transects were collected at the Neighborhood bar (rkm 79.7; fig. 20I) in 2017 to assess distributions of coarse and fine sediment deposits where spawning of Chinook salmon and Pacific lamprey, respectively, has been observed by the Confederated Tribes of Siletz Indians of Oregon. At eight of these surficial bar sites, subsurface bulk samples were collected. Near five of the surficial bar sites, in-channel (wetted) particle counts were also collected (“surficial channel particle counts”). Point, lateral, and island bar sites were selected based on bar size (greater than about 800 m²) and accessibility and chosen to maintain consistency with previous bed-material studies in Oregon coastal rivers (Wallick and others, 2010, 2011; Jones and others, 2011, 2012a–c). These sites were likely formed by recent or historical deposition events and have been scoured by ongoing flood events as indicated by the absence or minimal coverage of vegetation. The data collected by this study were supplemented by surficial particle size measurements data collected by the Confederated Tribes of Siletz Indians of Oregon in 2014 at 35 gravel bars between the City of Siletz and the Bulls Bag area (rkm 68.4 and 33.4; fig. 3; Stan van de Wetering, Confederated Tribes of Siletz Indians of Oregon, written commun, August 8, 2017). Slight differences in sampling methods may introduce some error into the data, but data were deemed comparable at the reach scale.

Surficial bar particle counts systematically sampled particles at evenly spaced increments along two transects (Bunte and Abt, 2001) parallel to flow direction of the main channel. Two hundred clasts were measured with a gravel template (Federal Interagency Sediment Project, U.S. SAH–97™ Gravelometer), which standardizes the measurement of sediment clasts greater than 2 mm in diameter. Surficial particle counts collected by the Confederated Tribes of Siletz Indians of Oregon followed a similar protocol using a gravelometer template. Surficial channel particle counts followed a more traditional Wolman (1954) pebble count approach. One hundred randomly selected particles are measured following a zig-zag pattern across the low-flow wetted channel, primarily perpendicular to flow, within a single geomorphic unit (riffle) of about 50-m long. Subsurface samples were taken at the mid-point of the surficial bar particle counts after the surficial gravel was removed from the bar surface over an area about one square meter and to the depth of about the largest particles. Approximately, one cubic meter of substrate material (50.8–71.5 kilograms [kg]) was excavated and sent to the USGS Cascades Volcano Observatory Sediment Laboratory in Vancouver, Washington, for subsurface particle size analyses.

The particle size distribution data, descriptive statistics, and armoring ratios are used to describe the longitudinal variation in bed-material particle sizes along the Siletz River. Descriptive statistics, including the 16th, 50th or median, and 84th percentiles, of the particle count diameters sizes were calculated from the distributions for all surface and subsurface particle data. Armoring ratios, comparing the median particle size of the surficial bar particle counts to the subsurface measurements, were calculated for all sites where subsurface data were available. The armoring ratio is typically close to 1 (meaning surface and subsurface particles are of similar sizes) for locations where sediment supply exceeds transport capacity, and approaches or exceeds 2 for locations where transport capacity exceeds sediment supply (Bunte and Abt, 2001).

Particle Size Measurements Uncertainty

Surficial particle count uncertainty can source from operator sampling bias, sampling location, and collection technique (for example, Wohl and others, 1996; Bunte and Abt, 2001; Olsen and others, 2005; Daniels and McCusker, 2010). For example, one sampling transect on a gravel bar is not representative of sediment particle distributions across the entire bar surface. We did not systematically evaluate uncertainty relating to particle counts for this study.

Bed-Material Particle Size Results

Upstream from the City of Siletz (rkm 68.5) in the Upper Sandstone and Sam Creek reaches (rkm 88.6–68.8), the median diameter (D₅₀) of surficial particles at the 11 sites on bar surfaces ranged from 18 to 144 mm (table 11). Reported D₁₆ and D₈₄ values range from 3 to 48 and 41 to 242 mm, respectively. Although particle sizes were highly variable between sampling locations, surficial particle distributions generally fell into two categories: bars with finer textures (typical D₅₀ less than 22 mm) and bars with coarser textures (D₅₀ greater than about 50 mm; figs. 26–27A–J). The coarsest bar textures (D₅₀ ranging from 52 to 144 mm) were sampled at the Logsden, Twin Bridges, Rock Creek, Neighborhood (coarser transect), Mid-channel, Bentilla Island, and Klamath Grade bars. The finest bar textures (D₅₀ ranging from 18 to 21 mm) were sampled at the Upper Moonshine, Neighborhood (finer transect), Downstream Moonshine, and Windchime bars. Compared to the finer bars, the coarser bars had a wider range of D₅₀ and D₈₄ values and fewer smaller particles (less than about 24 mm).

Downstream from the City of Siletz in the Mill Creek and Tidal reaches (rkm 68.8–7.1), surficial D₅₀ values at 35 bars sampled by the Confederated Tribes of Siletz Indians of Oregon ranged from 17 to 74 mm (fig. 28). Reported D₁₆ and D₈₄ values range from 7 to 38 and 31 to 118 mm, respectively. Overall, the 2014 measurements made downstream from the City of Siletz were slightly finer than the 2017 counts made upstream from the City of Siletz.

Longitudinal assessment of particle size data from Moonshine Park to the Bulls Bag area (measurements made from rkm 87.5–33.4) show that overall particle distributions tended to be coarsest in the upstream reaches and finer downstream. Median and D₈₄ values tended to be greater in the Upper Sandstone and Sam Creek reaches and then generally decline downstream in the Mill Creek and Tidal reaches (fig. 28; tables 10–11). Each geomorphic reach also had local areas of with relatively high D₅₀ and D₈₄ values, commonly coinciding with tributary confluences and bedrock outcrops. In the Upper Sandstone reach, particle sizes generally increased in the downstream direction from the Logsden bar (rkm 81.1). In the Sam Creek reach, particle sizes generally increased in the downstream direction at the Rock Creek (rkm 80.1), Twin Bridges (rkm 75.9), and Klamath Grade (rkm 73.3) bars downstream from Rock, Scott, and Sam Creeks, respectively. In the Mill Creek reach, particle sizes were relatively high downstream from Thompson and Euchre Creeks at rkm 55.1 and 45.4. In the Tidal reach, particle sizes were relatively high at Cedar Creek bar (rkm 39.4). Locally high D₅₀ and D₈₄ values were also observed in steep sections with bedrock outcrops, such as at Below Mill Launch bar (rkm 60.5), and sections where the active channel is relatively confined, such as Old River Road bar (rkm 57.8). At the most-downstream particle count (rkm 33.4) made on a bar downslope of the landslide pushing the Siletz River northward, D₅₀ and D₈₀ increased again.

Measurements at the seven sites upstream from the City of Siletz where subsurface bed-material samples were collected revealed subsurface textures that were similar to or finer than bar surfaces. The median size of subsurface samples ranged from 17 mm at Downstream Moonshine bar (rkm 86.1) to 76 mm at Logsden bar (rkm 81.1; fig. 27A–J; table 11). At three sites (Upper Moonshine, Downstream Moonshine, and Windchime bars), the particle size distributions of subsurface and surface samples were very similar, and armoring ratios varied from 1.0 to 1.1, indicating that sediment supply for these particle sizes is approximately balanced by transport capacity (table 11). However, at the four other sites where surface and subsurface samples were collected (Logsden, Rock Creek, Mid-channel, and Twin Bridges bars), subsurface particle size distributions were substantially finer than samples of bar surfaces, having armoring ratios approaching or exceeding 2, which may indicate excess transport capacity relative to the supply of sediment sizes on the bar surface. The most heavily armored bars (armoring ratios of 4.8 to 5.1) were observed at Rock Creek and Twin Bridges bars (respectively).

Surficial particle sizes sampled at five sites within the low-flow wetted channel generally showed similar distributions as surficial particle size measurements made on adjacent bar surfaces except for at the Twin Bridges bar (fig. 27A–J). The median diameter of surficial particle counts in the wetted channel at those five sites ranged from 27 mm at Upper Moonshine bar to 67 mm at Mid-channel bar (table 11). In-channel D₅₀ values were slightly greater than bar surface D₅₀ values at the Upper Moonshine and Mid-channel bars (difference of 6–14 mm) and slightly less than bar surface D₅₀ values at the Bentilla Island and Klamath Grade bars (difference of 10–31 mm). The exception was Twin Bridges (rkm 75.9) where the D₅₀ value in the wetted channel was substantially finer than the adjacent bar surface measurement (difference of 71 mm).

For the measurements made for this study in the Upper Sandstone and Sam Creek reaches, the percent sand (percent of particles with a diameter less than 2 mm) for all surficial particle counts on gravel bars and in the wetted channel was typically less than 4 percent but exceeded 10 percent in measurements made at Downstream Moonshine bar (figs. 26–27A–J; table 11). Sand content was considerably greater in the subsurface samples, ranging from 12 percent at Downstream Moonshine bar to 31 percent at Windchime bar.

Bed-Material Particle Size Discussion

The surficial particle distributions on sampled gravel bars between Moonshine Park and the Bulls Bag area (rkm 87.2–33.4) were highly variable but generally display an overall pattern of downstream fining as the river approaches its mouth at the Pacific Ocean. This pattern of downstream fining is likely caused by a combination of selective transport (whereby the ability of rivers to transport large particles decreases downstream with diminishing gradient) and attrition (whereby sediment particles lose mass through prolonged fluvial transport [for example, Parker, 1991; Lisle, 1995]). Despite this overall trend of downstream fining, the bars along the Siletz River have highly variable local bed-material textures, with finer textured bars (D₅₀ ranging from 18 to 21 mm) and coarser bars that were generally well-armored and mantled in larger cobbles (D₅₀ ranging from 52 to 144 mm; fig. 28; tables 10–11). Similar variation in bed texture has been documented for other mixed-bed and alluvial rivers in western Oregon, such as the Umpqua and Wilson Rivers (Wallick and others, 2011; Jones and others, 2012c).

Although causal linkages between particle size and hydrogeomorphic factors influencing bar textures were not systematically evaluated, bed-material textures along the Siletz River appear to demonstrate local influences, such as channel morphology (channel width and gradient) and sediment inputs from tributaries, and basin-scale geology. Bars tended to be coarser above the City of Siletz because of the steeper channel gradient (fig. 4) and narrower channel in these reaches (fig. 6A), which increase transport capacity. This section of the Siletz River is also likely coarse because the main-stem channel and tributaries drain the Coast Range volcanic rocks lithologic province (fig. 1). This province is associated with sediment that is more resistant to breakage during fluvial transport and greater bed-material yield compared to the Coast Range sedimentary rocks lithologic province (O’Connor and others, 2014). Some increases in particle sizes between Euchre and Cedar Creeks (rkm 45.4–39.5) are caused by sediment contributions from tributaries entering the Siletz River along the right side of the channel that are also draining the Coast Range volcanic rocks lithologic province.

The coarser bars upstream from the City of Siletz (figs. 26, 28) also tend to be highly armored (table 11). Such conditions indicate that transport capacity may exceed sediment supply at these locations. The highly armored bars sampled on the Siletz River had some of the greatest armoring ratios calculated for Oregon coastal rivers (table 13) and were comparable to bars sampled along the Umpqua River, where transport capacity substantially exceeds sediment supply (Wallick and others, 2011). Bars with finer particle distributions, such as Upper Moonshine and Downstream Moonshine bars, were less armored, indicating sediment supply exceeds transport capacity at these locations. The results emphasize that factors, such as local hydraulics and channel morphology, allow for the deposition of finer sized particles in the active channel at select locations along this high energy river that typically has sufficient transport capacity to entrain available bed-material sediment (as evidenced by extensive in-channel bedrock). Additionally, the coarsely armored bars and in-channel bedrock suggest that systematic and substantial bed-level lowering (or incision) is likely to be minimal along much of the Siletz River upstream from the City of Siletz.

As would be expected for this semi-alluvial river with numerous armored gravel bars, many of the bars sampled between Moonshine Park and the City of Siletz did not have large percentages of sand (figs. 26–27A–J; table 11). Sand is supplied to the Siletz River, as evident by the greater percentages of sand in the subsurface particle distributions and observations of fine-grained deposition in overbank areas. However, the transport capacity of the Siletz River is generally sufficient to transport sand and finer particles downstream.

One-Dimensional Hydraulic Model Methods and Inputs

HEC-RAS 1D models require topographic and bathymetric data to define the river and floodplain geometry, hydrologic inputs for boundary conditions, and assignment of hydraulic parameters that govern the calculation of energy losses and channel routing. Model results were calibrated from water pressure transducers that measured water depth between 2017 and 2018 (Leahy and others, 2024) and validated by comparing model results of stage to the rating curve of USGS streamgage 14305500.

Discharge Scenarios and Boundary Conditions

The 1D model of the Siletz River was run as a series of mixed-regime flow simulations for eight distinct steady-state discharge scenarios to characterize hydraulics at elevated discharge when geomorphic change is anticipated. The flow scenarios were selected to span the range of elevated discharge events observed during the period of record and potential future major floods and were described using AEPs calculated from the long-term discharge record at USGS streamgage 14305500 (table 14). Flow scenarios were determined by calculating the peak flows for six AEPs (0.995, 0.5, 0.1, 0.02, 0.01, 0.002) and by reducing the discharge associated with the 0.995 AEP to two lower flows (25 and 50 percent of the discharge of the 0.995 AEP). These flow estimates are valid for the reach surrounding USGS streamgage 14305500.

Boundary conditions at key tributary confluences along the Siletz River upstream and downstream from the station were determined from flow estimates provided in the NHDPlus V2 hydrologic dataset (McKay and others, 2012). Each flow scenario value was scaled according to the estimated mean annual flow (from the gage-adjusted Enhanced Unit Runoff Method; McKay and others, 2012) at 13 ungaged locations and the upstream boundary of the model (table 14). Figure 30 shows the location of the tributaries, the longitudinal profile of the modeled reach (derived from the minimum channel elevation at each cross section), and the location of bedrock exposed in the channel. The downstream boundary condition was established using a normal depth for a slope of 0.0006 m/m, calculated using the average bed-gradient between the downstream cross sections of the river. A sensitivity analysis was run to evaluate the sensitivity of results to changes in the normal depth, which revealed only minor changes when increasing or decreasing the normal depth by 10 percent.

Table 14.    

Discharge values for 6 annual exceedance probabilities (AEPs) and 2 lower discharge scenarios at 14 locations (13 ungaged, 1 gaged) in the one-dimensional hydraulic model reach on the Siletz River, Oregon.

[Modeled AEPs are in cubic meters per second (m³/s). Abbreviations: rkm, river kilometer; AEP, annual exceedance probability; %, percent; NA, not applicable]

Table 14.    Discharge values for 6 annual exceedance probabilities (AEPs) and 2 lower discharge scenarios at 14 locations (13 ungaged, 1 gaged) in the one-dimensional hydraulic model reach on the Siletz River, Oregon.

River station	Rkm	AEP
		25% of 0.995	50% of 0.995	0.995	0.5	0.1	0.02	0.01	0.002
		Recurrence interval, in years
		NA	NA	1.005	2	10	50	100	500
		Discharge, in cubic meters per second
69914	87	34	68	137	472	574	729	790	925
61746	84.6	35	70	140	483	588	747	810	949
58822	83.7	35	71	141	485	591	752	815	954
51148	81.3	38	76	151	521	634	806	874	1,020
47738	80.4	46	93	186	640	779	989	1,070	1,260
44506	79.4	47	93	187	644	783	995	1,080	1,260
33207	76	47	94	188	648	788	1,000	1,090	1,270
30913	75.3	49	99	197	681	828	1,050	1,140	1,340
29611	74.9	50	100	200	689	838	1,070	1,160	1,350
27300	74.2	50	100	200	691	840	1,070	1,160	1,360
25893	73.8	50	101	202	695	846	1,070	1,170	1,360
19532*	71.9	53	105	211	728	885	1,120	1,220	1,430
16413	70.9	53	106	213	733	890	1,130	1,230	1,440
8247	68.5	53	106	213	733	890	1,130	1,230	1,440

^*

River station 19532 location coincides with the river segment at U.S. Geological Survey streamgage 14305500 (Siletz River at Siletz, OR; U.S. Geological Survey, 2021b).

Model Calibration and Verification

The 1D model was calibrated by comparing simulated water surface elevations (WSEs) with measured WSE obtained from 10 pressure transducers (PTs) located along the model reach (figs. 3, 9A–B–10A–C; Leahy and others, 2024). PTs were installed in vented PVC tubing, which was bolted to the downstream end of large features that were unlikely to move at high flow, such as bedrock, bridge piers, or large boulders. The PTs were installed in autumn 2017 and retrieved in summer of 2018, collecting continuous water level measurements at 30-minute increments over this period. Most PTs were placed in the low-flow channel, and thus collected water level and temperature data at a wide range of flow conditions throughout their deployment; however, three PTs (Bentilla, Downstream Neighborhood, and Downstream Moonshine bars) were installed after the 0.50 AEP event in 2017 and were dewatered at lower flows during summer 2018. This missing data from summer of 2018 affect model calibration at the 25 percent of the 0.995 AEP at Bentilla and Downstream Neighborhood bars and at the 0.50 AEP at all 3 PT locations. Due to this lack of data, results at these flows and locations are subject to higher uncertainty. To convert the pressure readings of the PTs to relative water level, or stage, atmospheric pressure was collected near the middle of the reach. Atmospheric pressure was then subtracted from PT pressure measurements, and the resulting value was then converted to stage using equations provided by PT manufacturers (Onset Computer Corporation, 2022). To convert the relative water level values for each PT to WSE values that could be compared with model results for model calibration, the elevation of each PT was surveyed at time of deployment and retrieval using an RTK GPS and added to the water level values collected by PTs (complete description of the PT data are provided in the metadata for this publication; Leahy and others, 2024).

Water surface elevations for four discharge scenarios, all equal or less than the magnitude of the 0.50 AEP, were used for model calibration. During model calibration, several model parameters, such as the Manning’s n value, ineffective flow areas, and cross section placement were systematically inspected and adjusted to improve model performance. Manning’s n values were adjusted based on reasonable roughness values that minimized model error (observed WSE minus model WSE). Final roughness values ranged from 0.03 to 0.045 for the low-flow channel and 0.037 to 0.1 for vegetated bars, islands, and overbank areas. In the channel, roughness values closer to 0.03 were assigned to reaches of exposed bedrock, gravel, and (or) simple channel geometries whereas higher roughness values (closer to 0.045) were assigned to cobble and (or) boulder reaches or reaches with complex channel geometry. For the overbank areas and gravel bars, lower roughness values (typically ranging from 0.037 to 0.06) were assigned to sparsely vegetated bars, and cultivated overbank areas, whereas densely vegetated islands and bars and heavily wooded overbank areas were assigned roughness values ranging from 0.07 to 0.1. Ineffective flow areas were added to locations in the cross sections that were inundated by water, but likely did not convey flow, such as disconnected side channels and at mouths of tributaries. Cross section length and spacing were also inspected and adjusted to ensure each cross section fully contained the water at the highest simulated discharges and the convergence ratio was typically not less than 0.7 or greater than 1.4.

At all reference locations and for the calibration flows, model error was less than 0.6 m (2 ft), and in most cases, less than 0.3 m (1 ft; fig. 31). Individual calibration locations had varying level of bias and trends. For example, model error at the Moonshine bar location increases with higher discharges, and modeled WSE is consistently higher than measured WSE. However, there are no clear trends or bias in the calibration results reach-wide. Of the 10 calibration locations, 3 showed modeled WSE exceeding measured WSE at all flows, while 2 showed modeled WSE beneath measured WSE at all flows. The other eight locations showed a mix of over/under estimating WSE, depending on discharge. There also doesn’t appear to be consistent bias at a given magnitude discharge. Overall, the final model is a reasonable correspondence to measured water-surface stages across the range of modeled discharges and reasonable and consistent values of energy-loss parameters specified in the model.

Following calibration of the 1D model, model results were compared to WSE observations not used as part of the calibration process. Unlike the calibration WSE dataset that relied on data from PT measurements, the validation WSE dataset was created by taking the most recent stage-discharge rating curve for the USGS streamgage 14305500 (rating 21.0) and adding the gage datum (elevation) to the relative stage values. Hence, while the calibration dataset represented discharges from autumn 2017—summer 2018 and did not include flows over the 0.5 AEP, the validation dataset included WSEs for larger magnitude floods. Validation results indicated that the model was within 0.6 m (2 ft) of the 0.01 and 0.002 AEPs, lending confidence to the model results for large flows, at least near the location of the USGS streamgage 14305500 (fig. 32).

One-Dimensional Hydraulic Model Limitations

Because the 1D model was not calibrated to flows greater than the 0.50 AEP, results are most reliable for flows less than or equal to the 0.50 AEP for most the modeling domain except near USGS streamgage 14305500 where the model was validated to much higher flows using the stage-discharge rating curve. However, at these uncalibrated higher flows, the model underestimates energy losses, resulting in higher WSE model results than represented in the rating curve. This could be due to many factors, the most likely of which is that actual hydraulic roughness at these elevated discharges is greater than what is parameterized in the model. The model is used to compute hydraulic information (such as water depth, velocity, and energy slope) at each cross section and interpolated between cross sections based on the bounding cross sections and the terrain model. 1D models assume that all flow is uniformly directed downstream and are typically best suited for single thread reaches where channel planform is relatively simple (for example, few gravel bars) and where changes in channel or overbank geometry (such as channel widening) occur gradually. Hence, the Siletz River 1D model was mainly applied to single thread, confined reaches, but results from local areas with more complex channel morphology may be less accurate.

One-Dimensional Hydraulic Model Results

Following calibration and validation, the final 1D Siletz River model was used to simulate hydraulic conditions for eight discharge magnitudes (table 14), and results were analyzed to evaluate patterns that relate to local and river-scale hydraulic controls and variation with discharge. Longitudinal profiles of model output for discharge, top width of the active channel, hydraulic radius, energy slope, and velocity are shown in figure 33A–F. The largest step increase in discharge across all flows occurred at rkm 80.3 (fig. 33A) because of the confluence of the Siletz River with Rock Creek, the largest tributary sub-basin in the modeling reach (11 km²; 43 mi²) that is heavily forested (fig. 3; U.S. Geological Survey, 2019). The top wetted width of the active channel (henceforth, ‘top width’) ranged from a minimum of about 25 m (82 ft) at the lowest flow near rkm 72 to a maximum of about 105 m (344 ft) at high flows near rkm 80 (fig. 33B). At most modeled flows, the top width varied from about 30 and 60 m (98 and 196 ft). Local bars and floodplains were largely inundated by flows exceeding the 0.995 AEP, thus top width changed very little as flows increased. Depths varied from less than 1 m (0.3 ft) at low flows to over 15 m (49 ft) at the highest modeled discharges (fig. 33C).

Results from the 1D steady flow model show that velocities (fig. 33D) ranged from less than 0.3 m/s (1 ft/s) to about 3.7 m/s (12 ft/s) for discharges at or below the 0.995 AEP (211 m³/s [7,452 ft³/s] at USGS streamgage 14305500) but increased to a maximum velocity of about 4.9 m/s (16 ft/s) for the 0.002 AEP (1,428 m³/s [50,423 ft³/s]; table 15). Once the flow in the river approached the 0.50 AEP, the maximum velocity increased very little (less than about 0.6 m/s [2 ft/s]) for flows up to and including the 0.002 AEP. The change in maximum velocity of the flows up to the 0.50 AEP was about 1.5 m/s (5 ft/s). In general, stream velocities were the highest overall in the upstream-most 2 kms of the model where the Siletz River exits a steep and narrow bedrock canyon upstream from the modeled reach. Considering all cross-sections in the 1D model, mean channel velocity for the 25 percent of the 0.995 AEP was 1.18 m/s (3.88 ft/s) and increased to 1.86 m/s (6.10 ft/s) at the 0.995 AEP (table 14). The mean channel velocity of the 0.50 AEP was 2.82 m/s (9.25 ft/s) and only increases to 3.56 m/s (11.67 ft/s) at the 0.002 AEP.

The hydraulic radius, a metric used to describe the shape of the channel, varied from about 0.5 m (1.6 ft) near rkm 80 at the lowest flow to almost 10 m (33 ft) near rkm 73 at the 0.002 AEP (fig. 33E). Reaches with lower hydraulic radii had more water in contact with the channel bed and were subject to more frictional forces than reaches with higher hydraulic radii. In general, reaches with higher hydraulic radii tended to have higher flow velocities. Although hydraulic radius varied considerably along the modeled reach, hydraulic radius generally increased from upstream to downstream for flows greater than or equal to the 0.50 AEP. The energy slopes for flows below the 0.50 AEP tended to be much higher (maximum of 0.023 near rkm 80) than flows at or above the 0.50 AEP at many locations along the modeled reach (fig. 33F). With few exceptions upstream from rkm 80, flows greater than or equal to the 0.50 AEP had energy slopes less than about 0.005. Generally, the highest energy slopes were upstream from about rkm 80 in the modeled reach. The channel in the reach upstream from rkm 80 is steeper than the downstream reach and includes bedrock knickpoints in the channel which locally steepen the energy slope at low flows. These bedrock knickpoints tended to get inundated at higher flows (greater than about 481 m³/s [17,000 ft³/s]), lowering the local energy slope.

One-Dimensional Hydraulic Model Discussion

The Siletz River 1D model provides a basis for evaluating hydraulic conditions for different discharge magnitudes along the 19-km model reach of the Siletz River between Moonshine Park and the City of Siletz. The results can help characterize this reach of the Siletz River that has spawning habitats used by Chinook salmon and Pacific lamprey and understand expected hydraulic conditions now and in the future. Results show that there is a notable change in hydraulic character between the 0.995 and 0.5 AEPs, but smaller changes between the 0.5 and 0.002 AEPs (figs. 33A–F; 35A–F). For example, there is a notable increase in wetted width at flows greater than the 0.995 AEP, but very small increases in wetted width at higher AEPs (figs. 34B, 35A), suggesting that the Siletz River has inundated local bars and floodplains at the 0.5 AEP, but widens little at flows greater than the 0.5 AEP. This is also shown in figure 34A–F, where all but one gravel bar is inundated at flows between 0.995 and 0.5 AEPs. While wetted width stays similar at discharges above 0.5 AEP, depth and channel velocity continue to increase between the 0.5 and 0.002 AEPs (figs. 35B-C). This again suggests that most of the available floodplains have been inundated at 0.5 AEP, resulting in swifter, deeper water.

Two-Dimensional Hydraulic Model Methods

The 2aD model used the same terrain model and WSE calibration dataset used in the 1D model development, whereby model terrain was developed utilizing a blend of lidar and boat-based bathymetry and the WSEs were measured using PTs. Bathymetry surveys in this reach were conducted specifically for 2D modeling, requiring a higher point density compared to 1D modeled reaches (refer to the “Topographic Inputs” sub-section under the “One-Dimensional Hydraulic Model Methods and Inputs” section).

Two-Dimensional Computational Mesh

Whereas the 1D model performs hydraulic computations at a series of defined cross-sections, the 2D model utilizes a computational mesh that encompasses the model domain. The HEC-RAS 2D modeling capability uses an Implicit Finite Volume algorithm (U.S. Army Corps of Engineers, 2016). This allows for the use of a structured or unstructured mesh with computations occurring at each cell boundary, which may range from three to eight cells (U.S. Army Corps of Engineers, 2016).

A mesh with a computational point spacing of 6.1 m (20 ft) was generated for a flow area that included the active channel, floodplain, and upland areas (fig. 36). Breaklines were added within the 2D flow area to represent significant barriers to flow, such as roads and natural embankments. Because smaller cell sizes are typically required to define substantial changes in geometry and rapid changes in flow dynamics, point spacing along the breaklines was reduced to between 1.5 m and 3 m (5 and 10 ft). Where the water surface gradient was flat or gradually changing, such as in the active channel and most floodplain areas along this reach, larger grid cell sizes were appropriate. The mesh had a total of 61,086 cells. The average cell size was about 29 m² (309 ft²).

Model Calibration

The same WSE and associated discharge data used to calibrate the 1D model were used to calibrate the 2D model. Two simulation flow hydrographs were used to calibrate the model. The first hydrograph included the 25 and 50 percent of the 0.995 AEPs and the 0.995 AEP and the second hydrograph included the 0.50 AEP. Each of the two simulations were ramped up to each of the AEPs used for the calibration and held steady for 5–7 hours to ensure the model was stable before ramping up to the next calibration flow. For the 0.50 AEP run, the model slowly ramped up to the calibration flow, held steady for 5 hours, then decreased. The 2D model domain included the location of three pressure transducers described in Modeling and Calibration section for the 1D model. Therefore, measured WSEs could be compared to modeled WSEs for the four calibration flows. Time steps, Manning’s n values, breaklines, and cell size were adjusted based on the model output. The difference between the measured WSE and the modeled WSE at the three pressure transducer locations for the four calibration flows ranged from −0.01 to 0.61 m (−0.04–2.01 ft) with a mean error of 0.15 m (0.51 ft) and a standard deviation of 0.27 m (0.88 ft). In general, the least amount of error was at low flows, and the 0.50 AEP had about 0.6 m (2 ft) of error (fig. 37). Model error can be attributed to several factors, including errors in the underlying topography and errors in measuring the WSE from the pressure transducers. Uncertainty is also associated with the discharge boundary condition for the model, which uses estimated steady flows, while the calibration data from the pressure transducers measured WSE during an actual flow event. The final model parameters represent what we determined to be a reasonable correspondence to measured water-surface stages across the range of modeled discharges and reasonable and consistent values of energy-loss parameters specified in the model. Potential application of model results should consider if this level of uncertainty is acceptable for the intended application of the results.

Two-Dimensional Model Limitations

Hydraulic models, like all models, make assumptions and simplifications to simulate physical conditions, which can introduce uncertainty or errors into simulations. Perhaps the most notable uncertainty in the 2D hydraulic model is that discharge conditions were modeled as steady-state flows, whereby each discharge was modeled for 99 simulation hours. However, storm events on the Siletz often result in hydrographs that rise and fall rapidly, potentially resulting in different hydraulic conditions at the time of peak (simulated) flow. Additionally, hydraulic calculations are performed at the cell level, and therefore features smaller than the cell are not explicitly accounted for in the results. For example, if a computational cell is 5×5 m, the hydraulic effects of a 1×1 m boulder within that grid cell would not be explicitly accounted for in the results. HEC-RAS 5.0.7 leverages sub-grid bathymetry, thus enabling the ability to capture some features (at higher resolutions than computational cells, but many smaller features are not captured. This is important when considering applications of model results, such as evaluating aquatic habitat suitability, where aquatic organisms may be relying on hydraulic conditions at scales smaller than those modeled. Additionally, the 2D model was calibrated with in situ PTs, as described above, which recorded WSE at discharges up to the 0.5 AEP. Thus, model results at higher discharges are uncalibrated and thus have higher, but unquantified, uncertainty. Finally, hydraulic models are sensitive to the underlying bathymetric data. Thus, errors and uncertainties in the DEM can propagate into model results. In this study, the DEM was created by combining single-beam sonar data collected along long profiles and cross-sections throughout the study reach with lidar data (refer to “Topographic Inputs”). This DEM represents a simplified surface of the channel and floodplain, and thus smaller features, such as boulders, are not represented in the DEM.

Two-Dimensional Hydraulic Model Results

The highest velocities in the 2D model reach were in four locations: (1) between rkm 76.85 and 76.10, (2) between rkm 79.95 and 75.75 where the main channel flows around Twin Bridges bar, (3) between rkm 75.40 and 75.10, and (4) downstream from rkm 74.75. Considering the whole 2D model reach and all the modeled flows, the maximum simulated water velocity values were about 3.0 m/s (9.8 ft/s) for the 25 percent of the 0.995 AEP, 4.0 m/s (13.0 ft/s) for the 50 percent of the 0.995 AEP, 4.7 m/s (15.3 ft/s) for the 0.995 AEP, and about 5.5 m/s (18.0 ft/s) for the 0.50 AEP. These local maximums occur in the main channel as water flows around Twin Bridges bar for flows at or less than the 0.995 AEP. For the 0.50 AEP, maximum velocities occur near rkm 76.25 just upstream from Twin Bridges. Flow depths approach 11 m (36 ft) at Twin Bridges, the deepest location in the model reach, at the 0.50 AEP.

Two-Dimensional Hydraulic Model Discussion

The 2D hydraulic model was developed to produce high-resolution hydraulic results in a high priority area for salmon spawning within the 1D reach and demonstrate the utility of 1D and 2D models for future studies along the Siletz River. Comparisons of the hydraulic model output from the 1D and 2D models reveal fundamental differences in the underlying computational framework for each model, especially in areas of complex channel morphology, such as where the channel has large gravel bars near Twin Bridges (figs. 38A–H). For this comparison, we examined the results from the 1D model at the cross sections as well as the interpolated data between cross sections. In the 1D model, hydraulic outputs (such as velocity, depth, and WSE) are computed at each cross section then interpolated based on the topography and water conveyance between cross sections. In the 2D model, hydraulic parameters are computed for each grid cell in the computational mesh. The underlying bathymetry and topography dataset (DEM) is the same in both the 1D and 2D models; the 1D model discretizes this terrain into cross-sections as well as for the interpolation results between cross sections, while the 2D model discretizes this terrain into cells. Therefore, the similarity of model results is not surprising.

At all simulated flows less than or equal to the 0.50 AEP, the magnitude and spatial distribution of velocities as computed by the 1D and 2D models for in-channel and overbank areas were similar (figs. 38A–H–39A–B). For discharges less than or equal to the 0.50 AEP, the maximum simulated velocity was higher in the 2D model (about 5.5 m/s or 18 ft/s) compared to the 1D model (3.7 m/s or 12.1 ft/s) along the same reach. The location of the velocity maximums occurred in the same general area (upstream from Twin Bridges starting at about rkm 76.1 and along the outside of the slight bend in the river near rkm 74.15) in both models (as shown in figure 38 at the 0.995 and 0.50 AEPs). Overall, the velocity of both models in this reach is about ±1 m/s (3 ft/s) of each other, except for the outside bend around Twin Bridges bar at the 0.50 AEPs and small areas in a few other locations. The 1D and 2D models produce similar velocity patterns at the 0.995 AEP, with the largest difference in velocity occurring near the Twin Bridges bar.

The general similarities between the 1D and 2D models in this reach are not surprising, given that the channel is a single-thread and laterally confined and both models utilize the same model terrain. Although the model differences are modest, interpolated 1D results throughout the rest of the model domain do not necessarily produce similar results as a detailed 2D model. The considerably coarser terrain data in the rest of the 1D model extent are not suitable for 2D modeling purposes, and simulated hydraulics in the segment of rkm 77–74 may not be representative of hydraulic conditions elsewhere in the 1D domain.

Hydraulic models are used to inform a range of management and restoration design questions. The benefits and limitations of 1D and 2D models generally depend on the questions asked with them. For example, where the Siletz River is typically a single channel and the direction of flow is largely downstream (for example, minimal lateral flow over banks), 1D models can provide accurate results to questions, such as inundation extent, water depth and velocity, and bed shear stress. However, 1D models are typically inadequate to answer questions that rely on hydraulic properties at scales finer than river cross sections or questions pertaining to more complex hydraulics, such as eddies and backwaters. In such instances, 2D models are advantageous because they calculate hydraulics at a user-defined resolution, and flow is simulated in both x- and y-directions. Such 2D models can be used to evaluate fish habitat or detailed flow around in-water structures, such as bridges or engineered log jams. Further, although 1D model results can be interpolated between cross-sections to achieve pseudo-2D results, these areas have much higher uncertainty than the same areas calculated in the 2D model. Thus, 2D models may be more appropriate if higher confidence is required at sub-cross-section scales.

Bedload Transport Methods

Implementation of Bedload Transport Equations and Inputs

Bedload transport capacities were calculated using the particle size data collected by the USGS and Confederated Tribes of Siletz Indians of Oregon in 2017–18 and hydraulic results from the 1D HEC-RAS model (fig. 33A–F; White and others, 2025) for the each of the sediment transport equations (Keith and Jones, 2025). Calculations using the Parker (1990a, b), Wilcock and Crowe (2003), and Recking (2013) equations were completed using scripts developed in the R statistical programming language (R Core Team, 2021) to compute bedload transport capacity at each cross-section location within the 1D hydraulic model created for this study, similar to approaches by Anderson and others (2017).

Particle Size

Surficial particle size data used in the Parker (1990a, b), Wilcock and Crowe (2003), and Recking (2013) bedload transport equations are from particle counts made along the Siletz River in 2017–18 (figs. 26–27A–J; tables 10–11; Jones and Keith, 2021). Co-location of particle counts with cross-sections in the HEC-RAS model were limited, requiring assignment of particle size distributions to some cross sections. Because the surficial bar-based bed-material particle counts showed considerable variability in particle sizes (D₅₀ ranged from 18 to 144 mm; figs. 26–27A–F; tables 9–10), two particle size distributions were used as the sediment inputs for estimating bedload transport:

• • Reach composited sediment input—All surficial particle-count distributions from the three gravel bar measurements within the Upper Sandstone reach (rkm 88.6–80.4) were composited and applied to cross sections upstream from the reach boundary at Rock Creek (rkm 80.4; fig. 40). Similarly, all surficial particle-count distributions from the eight gravel bar measurements within the Sam Creek reach (rkm 80.4–68.8) were composited and applied to cross sections downstream from Rock Creek. The absence of a clear longitudinal trend in particle size distributions indicates that composited particle size distributions should be approximately valid for assessing reach-scale patterns of bedload transport even when applied to cross sections several kilometers away from the measurement location.

• • Fine and coarse composited sediment input—Surficial particle-count distributions from the 11 gravel bar measurements in the Sam Creek and Upper Sandstone reaches classified as either “fine” or “coarse” distributions (fig. 26) were composited according to their classification and applied to all 1D model cross-sections. By separately simulating bedload transport using the fine and coarse composited sediment inputs, the sensitivity of particle size on modeled transport capacity could be systematically evaluated (fig. 40). Modeling transport capacity for fine and coarse bars also helps identify the frequency with which some gravel bars may be mobilized under different discharge magnitudes.

Hydraulic Parameters from One-Dimensional Hydraulic Modeling

Hydraulic parameter outputs from the 1D simulations for the eight distinct steady discharges (fig. 33A–F; table 14; White and others, 2025) were used as input to the bedload transport equations. Hydraulic parameters included hydraulic radius and energy slope (fig. 33E–F).

Development of Bedload Transport Rating Curves and Application to Specific Water Years

To support the study goal of evaluating bedload sediment transport dynamics over a broad range of discharge conditions and future climate, rating curves relating bedload transport capacity to discharge were developed for each cross-section. Interpolated transport capacity rates between calculated rates were applied to hydrographs, which allowed the calculation of integrated bedload transport capacity over multiple water years. Altogether, 9 bedload transport rating curves were developed for each of the 180 cross-sections in the 19-km model reach, using 3 transport equations (Parker, 1990a, b; Wilcock and Crowe, 2003; and Recking, 2013) and three particle size scenarios (reach-, fine-, and coarse-composited).

The bedload transport rating curves at each cross-section were applied to a suite of water year types to evaluate temporal and annual fluctuations in bedload transport caused by different hydrological conditions. The four water years used in the bedload transport simulations were WYs 1996, 2000, 2010, and 2018. WYs 1996 and 2000 represent relatively high magnitude discharge years, whereas WYs 2010 and 2018 represent relatively low magnitude discharge years. For each WY, we calculated the number of days discharge was exceeded the 0.995 and 0.5 AEPs for comparison.

• • WY 2000 had some of the highest flows modeled in this study and includes the peak discharge event on record (1,523 m³/s [53,800 ft³/s] on November 26, 1999), which is estimated to about the 0.002 AEP. Flows exceeded the 0.995 and 0.5 AEPs for 10.3 and 1.3 days, respectively.

• • WY 1996 also had numerous high flow events, but the peak event did not exceed the 0.02 AEP (as occurred in WY 2000). Flows in this year exceeded the 0.995 and 0.5 AEPs for 18.6 and 3.1 days, respectively, or more days than in WY 2000.

• • WY 2010 was a relatively low-magnitude discharge year; no peak flow events exceeded the 0.5 AEP, and discharge only exceeded a 0.995 AEP for about 0.7 days throughout the year.

• • WY 2018 was also a relatively low-magnitude discharge year and was modeled because it spanned the period in which pressure transducers were deployed from autumn 2017 to summer 2018 and captured all the moderate and high flows (and sediment transport events) during the water year. Peak discharge in WY 2018 was 736 m³/s (26,000 ft³/s), just lower than about a 0.2 AEP (table 2).

Recognizing that computed bedload transport is highly variable both spatially (even between adjacent cross-sections with similar morphology) and temporally (indicating even modest differences in discharge and associated hydraulics), bedload transport for each water year was summarized by taking the median bedload transport for all cross sections within each reach (Upper Sandstone and Sam Creek).

Bedload Transport Uncertainty and Limitations

Calculations of bedload transport are sensitive to the selected equation and input data (including particle size, channel gradient, depth, and discharge; Wilcock and others, 2009). Using multiple equations to calculate bedload transport results provides ranges of estimates that can be used to assess uncertainty. These equations have been applied to other gravel-bed rivers (Wilcock and others, 2009; Wallick and others, 2010; Anderson and others, 2017) with varying degrees of success. Particle size measurement error can include multiple sources (discussed in the “Uncertainty in Particle Size Measurements” section); modeling multiple reach averaged distributions (reach-average, fine, and coarse) likely encompasses the uncertainty resulting from measurements. Validation of the hydraulic model steady discharge simulations (discussed in the “Model Calibration and Verification” section for the 1D model) indicates that the model results are within 0.6 m (2 ft) of the 0.01 and 0.002 AEPs. In the absence of measurements of bedload transport to address uncertainty and identify the most appropriate equations and grain size distributions for the Siletz River, multiple equations and inputs (including various composited particle size distributions and steady discharge simulations) can be used to bracket a range of plausible transport capacities and identify longitudinal or discharge-based trends in transport capacity.

Bedload Transport Results

Calculated transport capacity rates are highly variable along the Siletz River between rkm 87.2 and 68.5. Calculated rates vary with bedload transport equation used, discharge magnitude, and input parameters (particle size and hydraulic model outputs; fig. 41A–C). Eight steady discharge simulations (table 14) were selected to span the range of historical discharge records and were carried through all transport capacity analyses. Results presented here focus on the 0.995, 0.50, 0.02, and 0.01 AEPs (fig. 41A–C), recalling that (1) AEP is the probability of the largest flood of the year being equal to or greater than a given value for any given year and (2) discharge increases as the AEP decreases.

Modeled Annual Bedload Transport Capacities for Water Years 1996, 2000, 2010, and 2018

The bedload transport rating curves were used to calculate annual bedload transport capacity (or “annual bedload transport”), which is an estimate of total mass of bedload sediment that may have been carried by the river in a particular water year scenario under conditions of unlimited coarse sediment supply. Annual bedload transport for the entire modeled area for the Upper Sandstone and Sam Creek reaches was evaluated for four water years (WYs 1996, 2000, 2010, and 2018). The annual bedload transport computed by the Parker (1990a, b) equation and reach-composite sediment inputs ranged from <0.001 to 3,140 thousand metric tons (kt) across all cross-sections within the study area and the four water year scenarios (fig. 42A–C). Median annual transport values across all cross sections were considerably lower (15, 12, 1, and 3 kt for WYs 1996, 2000, 2010 and 2018, respectively; table 18).

To better characterize the range of potential annual transport rates, annual transport rates were also computed for: (1) a high-discharge scenario represented by WY 2000 using reach-composite sediment inputs and the three transport equations (fig. 42B), (2) a low-discharge scenario represented by WY 2010 using reach-composite sediment inputs and the three transport equations (fig. 42C), and (3) for both years using the Parker (1990a, b) equation but using either the fine- or coarse-composite sediment inputs (figs. 42B–C). Compared to the Parker (1990a, b) equation, the Wilcock and Crowe (2003) and Recking (2013) equations produced greater reach-median transport capacities. For WY 2000, reach-median transport capacities computed with the Wilcock and Crowe (2003; 43 kt) and Recking (2013; 24 kt) equations were 3.7 and 2.0 times greater (respectively) than those computed with the Parker (1990a, b; 12 kt) equation (fig. 42B; table 18). For WY 2010, reach-median transport capacities computed with the Wilcock and Crowe (2003; 3 kt) and Recking (2013; 5 kt) equations were 4.0 and 5.8 times greater (respectively) than those computed with the Parker (1990a, b; 1 kt) equation (fig. 42C; table 18).

Even greater changes in annual bedload transport values were observed when varying input particle sizes with the Parker (1990a, b) equation. For both WYs, using the fine-composite sediment input increased the reach-median annual transport capacity, whereas using the coarse-composite sediment input decreased reach-median annual transport capacity (figs. 42B–C). Compared to the values computed using the reach-composite sediment input (fig. 42B–C), reach-median annual transport increased from 12 to 166 kt in WY 2000 and 1 to 21 kt in WY 2010 using the fine-composite input (table 18). Reach-median annual transport decreased from 12 to 3 kt in WY 2000 and 1 to negligible transport in WY 2010 using the coarse-composite input (table 18).

Comparing the modeled median annual transport rates for the Upper Sandstone and Sam Creek reaches reveals that the Upper Sandstone reach may have transport capacities that are about 2–4 times greater than those calculated for the Sam Creek reach (table 18). For water years with relatively high discharges (1996 and 2000), median annual transport capacities ranged from about 19–22 kt in the Upper Sandstone reach and 6–7 kt in the Sam Creek reach (table 18). Conversely, in water years with relatively lower flows (2010 and 2018), median annual transport capacities for both reaches ranged from 0 to 4 kt (table 18).

Table 18.    

Median bedload transport capacities for the Upper Sandstone reach (rkm 87.2–80.4) and the Sam Creek reach (rkm 80.4–68.5) calculated by Keith and Jones (2025) with the Parker (1990a, b), Wilcock and Crowe (2003), and Recking (2013) equations and a variety of sediment inputs (reach-, fine-, and coarse-composite sediment inputs) for relatively high- (1996 and 2000) and low- discharges (2010 and 2018) water years along the Siletz River, western Oregon.

[Transport capacity is negligible for rates of <0.001. Values in thousands of metric tons. Water year extends from October 1 through September 30 and is named for the year in which the period ends.]

Table 18.    Median bedload transport capacities for the Upper Sandstone reach (rkm 87.2–80.4) and the Sam Creek reach (rkm 80.4–68.5) calculated by Keith and Jones (2025) with the Parker (1990a, b), Wilcock and Crowe (2003), and Recking (2013) equations and a variety of sediment inputs (reach-, fine-, and coarse-composite sediment inputs) for relatively high- (1996 and 2000) and low- discharges (2010 and 2018) water years along the Siletz River, western Oregon.

Reach	River kilometer	Number of cross sections	Composited sediment input	1996	2000	2010	2018
Parker (1990a, b)
Upper Sandstone	88.6–80.4	82	Reach	22	19	2	4
			Fine	278	190	35	63
			Coarse	5	4	<0.001	1
Sam Creek	80.4–68.8	98	Reach	7	6	<0.001	1
			Fine	146	105	14	30
			Coarse	2	2	<0.001	<0.001
Total modeled study area	88.6–68.8	180	Reach	15	12	1	3
			Fine	223	166	21	52
			Coarse	3	3	<0.001	<0.001
Wilcock and Crowe (2003)
Upper Sandstone	88.6–80.4	82	Reach	120	83	18	30
			Fine	312	215	82	113
			Coarse	9	7	<0.001	1
Sam Creek	80.4–68.8	98	Reach	14	11	<0.001	2
			Fine	184	121	41	60
			Coarse	3	3	<0.001	<0.001
Total modeled study area	88.6–68.8	180	Reach	63	43	3	13
			Fine	278	188	54	83
			Coarse	6	5	<0.001	1
Recking (2013)
Upper Sandstone	88.6–80.4	82	Reach	49	33	10	15
			Fine	175	118	54	67
			Coarse	10	8	<0.001	1
Sam Creek	80.4–68.8	98	Reach	18	13	1	3
			Fine	111	73	34	44
			Coarse	4	4	<0.001	1
Total modeled study area	88.6–68.8	180	Reach	34	24	5	9
			Fine	157	106	39	55
			Coarse	6	6	<0.001	1

Inspection of discharge hydrographs for the four water year scenarios and their associated patterns of daily and cumulative bedload transport reveal the linkages between the magnitude, duration, and frequency of discharge events capable of transporting bed-material sediment and the cumulative impact of these events when considering total bedload transport each year. The following results focus on reach-median transport capacity computed using the Parker (1990a, b) equations and reach-composite sediment input.

• • The WY 1996 was a relatively high-discharge year with relatively high modeled median bedload transport capacity (table 18). Multiple events exceeding the 0.995 AEP were recorded at USGS streamgage 1430550 (Siletz River, at Siletz, OR) in WY 1996 and occurred between November and April (fig. 43A). Additionally, discharges exceeded the 0.5 and 0.01 AEPs in November-December and February, respectively. During WY 1996, increases in modeled transport capacity for the Upper Sandstone reach with the reach-composite sediment input and both reaches with fine-composite sediment input begin accumulating during a relatively small discharge event in October, whereas the Sam Creek reach with the reach-composite sediment input and both reaches with coarse-composite sediment input begin accumulating during a peak discharge event exceeding the 0.995 AEP in November. Substantial increases in modeled bedload transport throughout the water year for all reach and sediment input combinations coincide with greater discharge and longer periods at higher discharges.

• • The WY 2000 was also a relatively high-discharge year with the peak of record in November, which is around the 0.002 AEP (fig. 43B). However, the second greatest discharge event during this water year peaked near the 0.5 AEP in December. Four additional events exceeding the 0.995 AEP occurred in November, January, February, and June. During this water year, modeled bedload transport capacity begins increasing during an early November event and increases notably during the peak of record for all simulations.

• • Discharge in WY 2010 was relatively low compared with other modeled years; only one event in January exceeded the 0.995 AEP (fig. 43C). Although all reach and sediment input combinations show some increases in modeled bedload transport throughout the year, only the fine-composite sediment inputs and Upper Sandstone reach with the reach-composite simulations ended the water year with modeled median annual bedload transport exceeding 1 kt.

• • The WY 2018 was also a relatively low-discharge period with the peak event exceeding the 0.5 AEP (October) and a few smaller events exceeding the 0.995 AEP between November and April (fig. 43D). These peaks result in modeled median annual bedload transport that was about twice that for WY 2010.

Bedload Transport Discussion

The Siletz River is a mixed bed river with exposed bedrock and alluvium making up the channel bed. Bedload transport modeling was used to identify locations where the deposition of gravel supplied from upstream sources would most likely occur and evaluate the integrated transport capacity across select high- and low-discharge water years to evaluate the range of coarse sediment that may be transported along the Siletz River between rkm 87.2 and 68.5. Areas of low transport capacity could preferentially deposit sediment that may be important for salmon or lamprey spawning. Understanding the range of likely transport over different water years and discharge events can provide insights as to how future changes in discharge may impact bedload transport. Multiple transport equations, discharge simulations, and input particle sizes were selected to evaluate the range of predicted transport capacities.

In general, these results indicate that bedload transport between Moonshine Park and the City of Siletz is minimal for discharges at or less than the 0.995 AEP (fig. 41A–C). As discharge increases from the 0.995 to 0.5 AEP, transport capacity and sediment transport increase (table 17). Additionally, cobble and smaller particle sizes (less than 64 mm) are likely mobilizing at the 0.5 AEP, whereas coarser particle sizes require larger floods for transport and are mobilizing less frequently. Additionally, single high-magnitude events (as recorded in WY 2000) can transport substantial amounts of bed-material sediment; however, cumulative annual sediment transport can be greater in years with multiple events (as recorded in WY 1996; fig. 43A–B).

Overall longitudinal trends in bedload transport capacity rates showed substantial spatial variability among cross sections. Although relative differences in computed bedload transport capacity between adjacent cross sections (fig. 41A–C) may not be representative of actual bedload transport, some of the local variability aligns with observed channel morphology, hydraulics, and bed-material particle sizes. Sections with locally high transport capacity rates tended to coincide with areas of extensive in-channel bedrock or coarse bars, which both indicate that typical transport capacities are sufficiently great to preclude the deposition of sands to gravel-sized clasts or at locations downstream from tributary inflows that increase discharge in the Siletz River (for example, Miller and Rock Creeks). Conversely, some cross sections with relatively low transport capacity rates occurred near locations where relatively finer particle measurements were measured (for example, Windchime bar), indicating that lower transport capacity in these areas enables smaller gravel particles to deposit.

Uncertainties in the bedload transport capacity estimates are likely large. Although not explicitly quantified for this study, approximate ranges of uncertainty can be inferred from comparing transport capacity model runs. The change in integrated transport capacities for WY 2000 at sequential cross sections provides a generalized estimate of uncertainty and is typically (75 percent of cross-sections) about a factor of five, although locations where very low and high transport values were modeled at adjacent cross sections can have much higher uncertainties. Annual loads would not actually vary much over distances of several hundred meters. As a result, the variability among cross sections provides a rough estimate of the uncertainty due to variation in local hydraulics, and therefore transport capacity, along the study reach. Median loads calculated for the Upper Sandstone and Sam Creek reaches generally should be representative of reach-scale bedload transport conditions compared to using results for a single location.

Modeled median annual bedload transport applied to water years suggest that the magnitude of that transport depends on sediment inputs and whether discharges were relatively low or high. Overall annual bedload transport is relatively low for the reach- and coarse-composite sediment inputs, particularly for low-flow water years (reach-based median transport capacity ranges from 0 to 22 kt; table 18). Modeling the fine-composite sediment input produces greater magnitudes of modeled median annual bedload transport, particularly for high-discharge years (reach-based median transport capacity ranges from 14 to 278 kt). Normalized by total basin area (970 m²), the range of median annual bedload transport ranges from about 14 to 287 tons/km² for the four modeled water years.

The Siletz River median annual bedload transport values are similar to values for the Chetco River (Wallick and others, 2010) and less than values for the Umpqua River (Wallick and others, 2011). Wallick and others (2010) calculated mean annual transport at seven sites along the Chetco River that ranged from about 8 to 217 kt (WYs 1971–2008; converted from volumes reported in Wallick and others [2010] using a bulk density of 2.1 kg/m³). Normalized by area (914 km²), these values for the Chetco River translate to about 9–237 tons/km², which is still similar to the Siletz values normalized by basin area. Wallick and others (2011) calculated mean annual transport capacity along the south and main-stem Umpqua Rivers that ranged from 0 to 623 kt for 42 sites (WYs 1951–2008). Normalizing these mean annual transport capacities by basin area (12,103 km²) results in smaller values of 0–51 tons/km² than for the Siletz River.

Regional bed-material models created for western Oregon (O’Connor and others, 2014), provide an indication of mean annual sediment supply which can be compared with the bedload transport capacities modeled for this study. The model of O’Connor and others (2014) indicates that the long-term, average annual volume of coarse sediment entering the Siletz River within the modeled reaches ranges from about 8 to 11 kt and generally decreases in the downstream direction. Decreasing sediment flux modeled along the Siletz River suggests particle attrition (or sediment particles breaking down into finer sized particles during fluvial transport) and conversion to suspended sediment exceeds the input of coarser bedload sediment gravel supplied by tributaries, which is characteristic of Coast Range sedimentary rocks (O’Connor and others, 2014). Although sediment supply fluctuates annually, we can compare these values to modeled transport capacity for this study during high- and low-discharge years to roughly evaluate relative magnitudes of supply and transport (fig. 42A–C). In a low-discharge water year (such as WY 2010), median annual transport capacity with reach-composited sediment inputs is about 0–2 kt (table 18), which when coupled with mean annual bed-material sediment supply of about 10 kt, would likely result in some deposition. Conversely, a high-discharge water year (such as WY 2000), may produce reach median annual transport capacity of 6–19 kt (with reach-composited sediment inputs; table 18), which when coupled with 10 kt of bed-material sediment supply, would likely transport most incoming sediment downstream.