Valley Bottom Mapping and Analysis of Historical Channel Change

Historical and current channel maps, surveys, and aerial photographs provide a means for assessing planform and vertical changes to the Chetco River study area since the late 1930s, and also form a basis for future monitoring studies. In this study, planform changes to the morphology and land-cover types of the valley bottom are documented by analysis of multiple sets of aerial photographs dating back to 1939. Vertical changes to the channel and flood plain were assessed from sparse historical data, including 1939 and 1977 surveys, and the record of channel geometry documented at the USGS streamflow measurement station (14400000) at FPkm 15.2. Current information on topography, bathymetry, and vegetation was based on (1) LIDAR topography acquired in spring 2008, provided by the Oregon LIDAR Consortium (Oregon Department of Geology and Mineral Industries, 2009), (2) channel and estuary surveys from the summer 2008, and (3) 0.5 meter orthoimagery for 2005 developed from summer 2005 aerial photographs as part of the U.S. Department of Agriculture National Agriculture Imagery Program (NAIP).

Historical Changes in Channel Planform and Vegetation

Planview changes in channel morphology were quantified by mapping channel features from eight time periods using aerial photographs and the LIDAR topography. The time periods for channel mapping were selected to track channel change for the longest possible time period and to serve as a basis for assessing erosion and deposition for five time intervals: 1939–43, 1962–65, 1995–2000, 2000–2005, and 2005–08. These time intervals were selected on the basis 
of photograph availability and quality, and encompassed specific events possibly affecting channel morphology. During 1939–43 land use and gravel extraction in the Chetco River basin were minimal. The 1964 flood of record is included in the 1962–65 interval and represents an era of increasing land use throughout the basin, including navigational improvements near the mouth of the Chetco River and increased gravel extraction along the lower river corridor and timber harvest within the drainage basin. The three most recent time periods (1995–2000, 2000–2005 and 2005–08) postdate the era of most voluminous gravel extraction and timber harvest but encompass the two large floods of 1996 and 2006.

Acquisition and Rectification of Historical Aerial Photographs

Digital orthoimagery from 1995, 2000, and 2005 have been previously rectified and georeferenced and are in the public domain (table 2). By contrast, older sets of aerial photographs were available only as paper prints or negatives and required scanning, georeferencing, and rectification as part of this study (table 2). Coverage was complete for the entire study area for all photograph sets except for the photographs from 1939 which covered only as far as FPkm 13.5, leaving the upstream 2.5 km without coverage for 1939. The aerial photographs and LIDAR topography were acquired during flows less than 15 m³/s, well within the low-flow channel (tables 2 and 3).

Full details of georeferencing and rectifying are included in the metadata for the geographic information system (GIS) maps prepared in conjunction with this study (U.S. Geological Survey, 2009). The scanned historical aerial photographs were georeferenced in ArcGIS 9.2 using the orthophotographs from 2005 as a base layer and following the methodology of Hughes and others (2006). For the photographs from 1943, 1962, and 1965, we acquired 6–16 ground control points per photograph, and aimed to locate these control points as near as possible to the channel. A second order polynomial fit was applied to georeference the photographs, providing root mean square error (RMSE) values ranging from approximately 1 to 4.4 m. The photography from 1939 was difficult to register because of the small area covered by each photograph (approximately 1.5 × 2 km) and the small number of feature points present in both the 1939 and 2005 photographs. Consequently, the photographs from 1939 were georeferenced using only 3–6 ground control points per photograph and rectified using a first order polynomial, which resulted in RMSE values of 0.35–3.6 m. Once georeferenced, each photograph was rectified and then combined to create a seamless mosaic of images for each period.

Uncertainties and Limitations to Planimetric Mapping

Even with established protocols and spatial analysis techniques, uncertainty and error result from interpretive mapping of land-surface features from aerial photographs of varying quality and light conditions and from different periods (Gurnell, 1997; Mount and Louis, 2005; Hughes and others, 2006). For this study, the quality and resolution of the photographs varied spatially and temporally but were sufficient for most of the study mapping objectives. The major source of mapping error for most features in this study resulted from imprecise registration and rectification of historical aerial photographs, especially for the older photographs for which few features were available to use as control points. The RMSE values indicate that horizontal position uncertainties are less than 5 m; however, from test trials, positional errors for the historical aerial photographs resulting from the georeferencing and rectification process were determined conservatively to be almost everywhere less than 20 m. Positional errors associated with the publicly available orthophotographs for 1995, 2000, and 2005 are less than 6 m. Georeferencing errors have their greatest effect on analyses of photograph-to-photograph change, such as for quantitative estimates of channel movement and bar growth and erosion, but have little influence on measurements of total areas of features such as the for the channel and gravel bars.

Another consideration in comparing mapped features from several periods of time is differences in discharge between aerial photograph sets. Although all photography and LIDAR topography were acquired during low-flow periods (tables 2 and 3), small changes in discharge can influence delineation of channel and bar areas, particularly in areas where the channel is wide and shallow. We partly account for this in some analyses by determining the relations of bar and channel area to flow using a one-dimensional hydraulic model and the channel and flood-plain topography for 2008 (see complete model description in the Hydraulic Modeling section). This relation between bar and channel area to flow (fig. 7), indicates that as much as 15 percent of the total bar area is inundated within the range of flows in the analyzed photographs, and was used to normalize the channel width and bar area measurements for all analysis periods to a constant discharge of 2.8 m³/s, a discharge slightly less than the lowest discharge associated with any of the photograph sets.

Tide level has an especially large influence on the mapping within the Estuary Reach, particularly for gravel bars submerged with each tidal cycle. Because tidal stage varied between photograph sets, only the part of bars inferred to be above tidal range during low flow periods were mapped. Mapping of the bars was possible because bars subject to daily tidal inundation have substantial algal growth, giving them a distinctly darker tint in the photographs. The tidally inundated parts of the bars were included in the primary channel map unit.

Considering registration errors and digitizing precision, the horizontal uncertainty of the digital channel and flood plain maps was inferred to be less than 15 m for sharply defined features. For the maps from 1995 to 2008, positional uncertainty is probably less than 6 m as judged by the precise agreement between persistent features observable on this imagery. Flow variations between photograph sets add additional uncertainty, which can in part be accounted for by normalizing bar and channel area measurements to a reference discharge.

Mapping Channel Features, Flood-Plain Vegetation, and Bank Materials

The photograph mosaics provide the basis for systematic mapping of channels and bars as well as broad scale land-cover and vegetation characteristics. Geomorphic features were mapped for each of the seven photograph sets and for the LIDAR topography. The mapped features form a foundation for evaluating changes to channel and bar planform and support the analysis of depositional and erosional volumes. Land cover and vegetation were mapped for only the photographs from 1939, 1962, 1965, and 2005 to determine coarse patterns of change in vegetation cover and density within the geologic flood plain.

Mapping of geomorphic features was confined to the active channel, defined as the area typically inundated during annual high flows as judged by the presence of water and flow-modified surfaces (Church, 1988). Features within the active channel were divided into mapping units: (1) primary (low flow) channel, (2) gravel bars, (3) alcoves (side channels or other wetted areas connected to the primary channel), (4) tributaries, (5) jetties, (6) disconnected water features, and (7) the constructed boat basin. For each period of time, all such features larger than about 200 m² were digitized at a scale of 1:1,000. All linework was reviewed at a scale of 1:3,000 by another project team member to ensure consistency between time periods.

The primary channel was mapped by digitizing the wetted perimeter of the main channel as shown on aerial photographs and the LIDAR topography. Gravel bars, defined as gravel-covered surfaces with evidence of recent mobilization (bare or sparse vegetation) were separated into two categories: flood-plain bars (sharing a margin with the flood plain) and island bars (completely surrounded by water). Tributary channels and tributary fans also were mapped where these features were discernable; however, because of differences in photograph resolution and vegetation, tributary features present in some periods were not always apparent in other periods. Disconnected water features were defined as any water body within the active channel area completely separated from other water features, and mostly consisted of ponds in swales on flood-plain bars. Constructed features consisted of the boat basin, jetties, and the dike alongside the boat basin.

Although geomorphic features were mapped for only the active channel corridor, basic land-cover attributes, including vegetation, were mapped for the entire geomorphic flood plain, but for only four time periods. The geomorphic flood plain was defined for this study as the relatively flat surface formed of recent alluvium occupying the valley bottom, and was mapped on the basis of topography and field inspection. The flood-plain boundaries depicted here do not necessarily correspond to inundation levels of specific flood discharges or flood frequency. Choices of map units for the land cover and vegetation mapping were based on review of historical and recent aerial photographs to ensure that each of the land cover classes could be distinguished from each set of photographs, supplemented by field inspections during September 2008. Species information was compiled from field manuals and with assistance from silviculturist Robyn Darbyshire (U.S. Department of Agriculture, Forest Service, oral commun., September 12, 2008).

Eight mappable classes of land cover were defined, with three of these classes also assigned vegetation density ranges. Detailed descriptions of each mapping category are provided in the metadata accompanying the GIS files (U.S. Geological Survey, 2009) and are only summarized here. All wetted features, including the primary channel and alcoves, are mapped as Water, whereas rocky outcrops, including “Tide Rock” and “Morris Rock” (fig. 1), are mapped as Bedrock. Major paved roads, developments, and clusters of houses are mapped as Developed areas, although individual houses and small dirt roads are classified according to the surrounding land cover. Bare surfaces are nonbedrock terrestrial surfaces with less than 25 percent cover of discernable vegetation, typically appearing light colored on aerial photographs. Bare surfaces are chiefly gravel bars with recently disturbed surfaces (fig. 8). Sparse Vegetation is the designation for surfaces with 5–25 percent vegetative cover, and typically consists of isolated trees, grasses, and shrubs. These areas almost always are gravel bars vegetated with early successional species (fig. 8). Grasses, lawns, agricultural lands, and various herbaceous communities (including Vetch spp., Bacharis spp., and members of the composite family) are mapped as Herbaceous Vegetation, which has smooth texture and light brown or gray color in the aerial photographs (fig. 8). The Woody Shrub mapping unit is for areas with low canopies (chiefly less than 5 m) sufficiently dense to appear relatively smooth in the aerial photographs. Woody shrub cover is typically composed of willows (Salix spp.) and small (less than 5 m tall) alders (Alnus spp.). This type of cover is found almost exclusively on gravel bars, commonly growing in narrow groves or thickets aligned parallel to the channel (fig. 8). Clusters of large trees were mapped as Mature Trees, and typically included black cottonwood (Populous balsamifera), myrtlewood (Umbellularia californica), and tall alders on flood-plain surfaces outside of the active channel area (fig. 8). Although mature trees typically had a distinct size and texture when compared with willows and other shrub type vegetation in the aerial photographs, it was difficult to discern small trees from willows; hence, canopies associated with trees less than about 5 m tall were grouped together in the Woody Shrub category. Vegetation density values of moderate (25–75 percent cover) and dense (75–100 percent cover) were assigned to Herbaceous, Woody Shrub, and Mature Tree mapping units.

The bank materials along the Chetco River corridor were mapped in such a manner as to differentiate reaches bordered by erodible sediments from reaches flanked by more resistant bedrock or artificial revetment. Bank materials were defined as the natural or artificial material bordering the active channel and were mapped by walking the length of the study area and recording the condition and composition of the channel banks. Field observations were then compared with the recent orthoimagery and LIDAR topography to construct continuous maps of bank materials along both edges of the active channel at a scale of 1:5,000. The map units include the primary types of bank materials: (1) flood-plain risers formed of erodible sand and gravel contained in fluvial deposits flanking the active channel, (2) bedrock outcrops, and (3) artificial fill (primarily consisting of material used to fill the former tidelands near the present location of the boat basin at FPkm 0). Bank protection revetment, chiefly consisting of large angular boulders, was mapped as an overlay to the three primary categories of bank material.

Results of Channel Mapping

Overall trends for 1939–2008 for active-channel features of the study reach show evidence of a 34 percent reduction in gravel bar area and a slight decrease in channel sinuosity (fig. 9). Channel width has not changed systematically during this period. The reduction in bar area is much greater than can be attributed to differences in flow stage between photograph sets (figs. 9 and 10). These overall changes, however, reflect a temporally and spatially varied history of channel behavior. The largest change, the decrease in bar area, is almost entirely accounted for by the large reduction in flood-plain bar surfaces between 1965 and 1995. Prior to 1965 and subsequent to 1995, bar areas may have increased slightly for some reaches, especially between 2005 and 2008, although at smaller rates relative to uncertainties of mapping and the effects of different discharges on mapped areas (figs. 9 and 10).

Historical channel change for 1939–2008 for the Chetco River was greatest along the lower reaches where the valley bottom is wide and a greater percentage of the channel is bordered by more erodible flood-plain materials than the upper reaches (figs. 11–15). The North Fork (fig. 13) and lower Mill Creek (fig. 14) reaches have had the most planform change. For the North Fork Reach, the 1939 channel was relatively sinuous and narrow, with a sinuosity of 1.16 and an average width of 47 m. The maps from 1995 to 2008 show the channel to be straighter, with a sinuosity in 2008 of 1.05. In conjunction with sinuosity changes, the average water surface slope of the North Fork Reach has increased by about 10 percent between 1939 and 2008, from 0.000767 m/m to 0.000849 m/m. Low-flow channel width changes have been more variable; for example, reach average width along the North Fork Reach was 66 m in 1995, 41 m in 2000, 61 m in 2005, and decreased to 47 m by 2008 (fig. 9). Between 1939 and 2008, normalized (for flow stage) total bar area for the North Fork Reach diminished from 0.4 km² to 0.27 km² (fig. 9). Similarly, bar area for the Mill Creek Reach has been reduced from almost 0.6 km² in 1939 to about 0.3 km² in 2008 (fig. 9). The net changes for these reaches, however, do not reflect continuous trends because of episodic increases in sinuosity and bar area within the overall record (figs. 9 and 10).

Inspection of the individual photograph sets shows that the changes along the North Fork Reach took place in several steps. During 1943–62, channel migration at rates of as much as 14 m/yr between photograph sets created a large meander bend near the confluence of the North Fork Chetco River (FPkm 7.5). During winter 1969–70, a large bend near the confluence of Jack Creek¹ (FPkm 6) was cut off and abandoned (probably during the January 1970 flood of nearly 1,900 m³/s). Between 1969 and 1976, two similar avulsions resulted in abandonment of the North Fork bend (FPkm 7.5) and a smaller bend near FPkm 6.² The two avulsions at FPkm 6 and 7.5 most likely were during the 1970–72 period of large floods with peak discharges of 1,300–1,900 m³/s (fig. 2). These avulsions in the late 1960s and early 1970s account for the major decrease in sinuosity for the North Fork Reach between 1965 and 1995 (fig. 9). Partly as a consequence of these channel changes, bank revetment has been placed along these channel margins in the North Fork Reach, so that revetment and bedrock now border 47 percent of the reach in contrast to more than 75 percent of the North Fork Reach being historically bordered by erodible alluvial flood-plain materials (fig. 11). In recent decades, the lower Mill Creek Reach and North Fork Reach have been much less dynamic than for 1939–1965, shifting laterally at rates less than 6 m/yr with no major avulsions (figs. 13 and 14).

Along the Estuary Reach, the overall style of planform change from 1939 to 2008 has been lateral shifting of the channel between the confining valley walls in conjunction with substantial loss of net bar area (figs. 9, 10, and 12). For example, near FPkm 3, channel maps from 1939 to 1965 show the low flow channel against bedrock along right bank, and a large (150,000 m²) gravel bar (known locally as “Tidewater Estuary Bar”) along the left bank. Between 1965 and 1989³, the channel shifted south to erode much of this bar (fig. 12). Additionally, higher elevation areas of Tidewater Estuary Bar, which appear bare and recently active in the photographs from 1939 to 1965, were protected by revetment and developed for residential and commercial use by 1989. The cumulative result of these types of changes is that bar area for the Estuary Reach has decreased 36 percent between 1939 and 2008, although bar area has recently increased between 2005 and 2008 (figs. 9 and 10). Development along the Estuary Reach has resulted in extensive bank stabilization; 41 percent of the channel margin is now bordered by revetment (fig. 11).

Major changes to the mouth of the Chetco River are the result of 20th century development and navigational improvements that began in the 1950s. The aerial photographs from 1939 and 1943 depict the mouth of the Chetco River as about 200 m wide, with extensive sand bars and tidal lagoon. By 1962, a pair of jetties restricted channel width and closed off the former lagoon. By 1995, continued bank protection, jetty extension, and filling of former lagoon areas resulted in an overall straightening and narrowing of the channel so that channel width at the mouth presently ranges from 100 to 120 m; about one-half the width shown on the earliest maps and photographs (fig. 12).

Channel change along the middle and upper reaches of the study area has been much less than for the lower Mill Creek, North Fork, and Estuary Reaches. Within the Emily Creek and Upper reaches, and the upper part of the Mill Creek Reach, the channel crosses back and forth between the valley walls with intervening channel-flanking gravel bars. The general pattern and positions have remained generally stable with the most stable locations being where the channel abuts the bedrock valley walls (figs. 14 and 15). In isolated locations, the river has migrated laterally at rates as much as 6 m/yr where crossing from valley side to side. Where the valley bottom widens towards the lower part of the Mill Creek Reach (FPkm 7.5–8.5), the channel has been more active, particularly in the period from 1943 to 1962 when rapid migration resulted in the formation of a large meander bend near the North Fork confluence (fig. 14).

Results of Land Cover Mapping

The land cover and vegetation mapping shows that the dominant land cover for the geomorphic flood plain is Mature Trees, covering about 30 percent of the flood plain in 2005 and primarily consisting of flood-plain forests outside of the active channel (fig. 16). Water occupies about 20 percent of the flood plain at low flow. Developed area accounts for about 30 percent of the flood plain area along the Estuary Reach in 2005. The Mature Trees category systematically decreases as a percentage of flood-plain area downstream, as does Water except for the North Fork and Estuary Reaches. Developed area is only substantial in the North Fork and Estuary Reaches, and primarily for the photographs from 1962 and more recent photographs. The most dynamic classes are the Bare, and the Sparse, Herbaceous, and Woody Shrub vegetation categories, which cover the greatest relative area in the Mill Creek and North Fork Reaches. These cover-type vegetation classes are chiefly associated with gravel bars subject to colonizing vegetation. No obvious trends are evident for these classes except that the combined area of Water, Bare, and Sparse vegetation was greatest for all four reaches in 1965, mostly at the expense of Woody Shrub and Mature Trees categories, likely indicating flood-plain erosion and vegetation removal by the flood in 1964.

Vertical Changes in Channel Morphology and Bathymetry

Although lateral channel changes may have significant resource, habitat, and hazard consequences, changes in the vertical position of the bed are more indicative of riverwide changes in the balance between sediment input and export (Schmidt and Wilcock, 2008). Vertical changes are also difficult to detect without systematic surveys of the channel. For this study, we compared two previous detailed surveys—a U.S. Army Corps of Engineers navigational survey in 1939 for the Estuary Reach between FPkm 0 and 4.5, and a Soil Conservation Service (1979) survey in 1977 for a flood study of the upstream reaches between FPkm 4 and 15—with the LIDAR topography acquired in 2008 and our own surveys during summer 2008 made as part of this study. Additional local bed elevation information is from repeat surveys of isolated cross sections in the fluvial reaches as well as the detailed information on channel bed changes from streamflow measurements at the USGS streamflow-gaging station at FPkm 15.24.

Survey Data Used in Study

Of several early surveys near the mouth of the Chetco River (table 3), the most useful survey for characterizing channel morphology along the Estuary Reach is the navigational survey of 1939 (U.S. Army Corps of Engineers, 1939), in which closely spaced soundings and elevations in feet relative to Mean Lower Low Water (MLLW) are provided for FPkm 0 to 4.5. Details of digitizing, georeferencing, and datum conversion are included in the metadata for the accompanying GIS maps (U.S. Geological Survey, 2009). This survey included more than 1,000 points over the lower 4.5 km of channel. The survey from 1939 was compared to a USGS bathymetric survey completed in September and October 2008 between FPkm 0 and 3. This boat-based survey used a depth-sounding transducer mounted directly below a real-time kinematic (RTK) global positioning system (GPS) receiver. As the survey boat traversed the estuary at transects spaced at 30–50 m intervals, the depth-sounder recorded water depth, and the GPS recorded the boat position and GPS ellipsoid height for a total of nearly 200,000 points (complete metadata and GIS layers available in U.S. Geological Survey, 2009).

The bathymetric data of 1939 and 2008 for the Chetco River estuary were interpolated to three-dimensional surfaces using a modified version of the procedure of Merwade and others (2005), which transforms the data into a channel oriented coordinate system, interpolates a continuous surface using anisotropic kriging, and reprojects the surface back to the project coordinate system of UTM NAD83 (fig. 17). Once the bathymetric surfaces were created, longitudinal profiles of the channel thalweg from each time period were extracted and plotted against river kilometers for 2008.

To determine vertical channel changes along the upstream fluvial reaches of the study area between FPkm 4 and 15, longitudinal profiles and cross sections were compiled from a 1977 survey and compared to 2008 elevation data and surveys. In 1977, 42 cross sections across the entire valley bottom between FPkm 0 and 15.5 were surveyed as part of a flood hazard study by the Soil Conservation Service (1979). The location of each survey transect was depicted on orthophotographs from 1976 and as cross-section data shown by plots of distance (in feet from an arbitrary point) against elevation (in feet referenced to NGVD 29 datum). From this information, cross section locations and data were digitized by visually plotting survey transects shown in the orthophotographs from 1976 onto the orthophotographs from 2005. The elevations for 1977 were shifted from NGVD 29 datum to the NAVD 88 datum using the CorpsCon conversion routine (http://crunch.tec.army.mil/software/corpscon/corpscon.html, accessed January, 13, 2009) and by comparing elevations of benchmarks surveyed in 1977 and 2008 throughout the study area.

Nine of these cross sections from 1977 were approximately matched by (1) using RTK GPS and depth-sounder surveys from October 2008 of the active channel at the estimated locations of the cross sections from 1977, (2) merging these channel surveys from October 2008 with the LIDAR topography from May to June 2008 to extend the surveys for 2008 across the valley bottom, and (3), where required, shifting the cross section data from 1977 laterally so obvious and stable topographic features such as road beds and steep banks were aligned with those features on the cross sections for 2008. Such adjustments were necessary in a few cases because the cross section locations for 1977 were not precisely located. The survey in 2008 also produced a nearly complete longitudinal profile of the channel thalweg from FPkm 4 to 15 (fig. 4), which can be compared to the minimum elevation for each of the 42 cross sections surveyed in 1977. These surveys were supplemented by ancillary survey data for 1980–82 reported by Klingeman (1993; fig. 18).

The final source of vertical change information is from analysis of the history of stage-discharge rating curves at the USGS streamflow-gaging station at FPkm 15.2. Following the approach of Klingeman (1973) and Smelser and Schmidt (1998), we completed a specific gage analysis for the available record from October 1, 1969, to May 1, 2009. The specific gage analysis allows detection of changes in streambed elevation by assessing changes in water elevation (stage) through time for a set of discharge values. At USGS streamflow-gaging stations, discharge is related to stage by a stage-discharge rating curve, which is based on multiple simultaneous measurements of stage and discharge. If channel conditions change substantially (as shown by consistent offsets of newer measurements from established rating curves), or if a station is moved, a new rating curve will be developed. The specific gage analysis evaluates trends in bed elevation as indicated by the sequence of rating curves. For situations where channel width and roughness remain stable, the sequence of stages for a given discharge directly relates to changes in bed elevation. For the Chetco River, the analysis is straightforward because there have been no relocations or datum shifts for the station, although the record is shorter than for many USGS streamflow-gaging stations and 3 of the 39 ratings were unavailable.

Uncertainty and Limitations Associated with the Repeat Survey Data

The total uncertainty regarding the bathymetric surfaces created from the survey data of 1939 and 2008 is a function of the original data and the processing involved with creating digital maps and interpolated surfaces of the bathymetries. Although the accuracy of the original map from 1939 is unknown, the process by which the original map was registered, rectified, and digitized may have introduced uncertainty on the order of ±20 m for the horizontal positioning of points, but in most locations is substantially less. The interpolation procedure introduces additional error and uncertainty, thus the total accuracy of the bathymetry for 1939 is estimated to be ±20 m for the horizontal dimension and ±1 m in the vertical dimension as determined from distribution of differences between the digitized survey points and the gridded elevation data. Each of the points from the bathymetric survey in 2008 has a horizontal accuracy of ±0.015 m and a vertical accuracy of approximately ±0.05 m. The interpolated bathymetric surface in 2008 generally is within ±0.3 m of the original survey elevations.

The survey in 1977 by the Soil Conservation Service (1979) was in support of a flood hazard study and preparation of flood hazard maps. The survey is described (Soil Conservation Service, 1979, p. E-1) as a “third order field survey” using USGS base elevations. For such surveys, elevation tolerances (RMSE) are typically less than 0.15 m (American Society of Civil Engineers, 1999, p. 6). The conversion of the original sea level (NGVD 29) datum to NAVD 88 is straightforward and the converted data match resurveys in 2008 of benchmarks used in 1977 to within 0.05 m. Therefore, the primary source of uncertainty regarding the survey in 1977 is its horizontal positioning. The only available information for the precise location of the measurements for 1977 is the 1:4,800 photomosaic maps in the Soil Conservation Service (1979) report. On the basis of these maps, the cross section locations for 1977 were digitized onto the photomosaic map for 2005 used for this analysis by reference to stable features visible on both photograph sets. The uncertainty associated with the horizontal placement of the cross sections from 1977 on the maps for 2005 were determined to be less than 150 m. Such an offset in conjunction with the 0.001 average slope of the study reach would introduce vertical errors of less than 0.15 m attributable to uncertainty in horizontal cross section position for thalweg and water-surface elevations (assuming uniform slope and depth). The accuracy of the cross-section data surveyed in 2008 as a part of this study is approximately ±0.015 m, whereas vertical accuracy is approximately ±0.05 m. Discrepancies between the cross-section alignments in 1977 and 2008 cause some cross sections of 1977 to portray slightly different areas of the bar and flood plain than are depicted in the matching cross section of 2008; therefore, the cross sections are best viewed in terms of overall trends, especially for thalweg elevations, as differences in bank geometry do not necessarily indicate channel shifting.

Results of Repeat Surveys

Comparison of bathymetric surfaces within the Estuary Reach from 1939 and 2008 shows that the bed of the Chetco River was generally lower in 2008 than in 1939 (fig. 17). A difference calculation for the bathymetric surfaces for 1939 and 2008 between FPkm 0.5 and 3.5, corresponding to the reach between the Highway 101 Bridge to Morris Rock, indicates a net loss of 150,000 m³ of channel substrate between 1939 and 2008. This net loss corresponds to an average lowering of the entire channel bottom by about 0.5 m. Locally, however, channel shifting in three primary locations has resulted in much greater magnitudes of incision and aggradation (figs. 17 and 18). Near FPkm 3, the channel historically flowed against the right bank with bottom elevations of approximately 0.5 m (NAVD 88). By 2008, the channel had shifted toward the left bank and had deepened by 0.2 to 2.0 m, with the bed elevations in 2008 ranging from 0.3 to -1.5 m (NAVD 88). Near FPkm 1.7 a large alcove in 1939 extended nearly 0.5 km from the right bank. By 2008, this alcove had aggraded by approximately 1 m, and the main arm of the alcove currently is filled with sediments and partially vegetated. Near FPkm 1.0 and just upstream of the Highway 101 Bridge, the channel in 1939 flowed against the left valley wall, carving a deep channel with bed elevations ranging from -1.5 to -4 m (NAVD 88). By 2008, the channel had shifted to the right bank, and the thalweg from 1939 currently is an alcove with bed elevations of about -0.6 m. The thalweg of 2008 in this area is shallower (bed elevations of -1.5 to -2.5 m NAVD 88) and lacks the deep pool depicted in the survey from 1939.

For the short reach between FPkm 1.5 and 4.3, where all three surveys overlap, the longitudinal profiles from 1939, 1977, and 2008 indicate net lowering of the channel thalweg between 1939 and 2008 (fig. 18A). The magnitude of lowering is as great as 2 m, with the reach upstream of FPkm 2 showing the most consistent bed lowering. The resolution of the survey in 1977 is not sufficient to clearly indicate whether most of the channel incision in the estuary was before or after 1977, but the survey in 1977 does show that the channel had at least locally aggraded by nearly 1 m near the Highway 101 bridge at FPkm 0.85 between 1939 and 1977 before incising back to its 1939 elevation by 2008.

Upstream of the Estuary Reach and the extent of the bathymetric surveys, comparison of longitudinal profiles derived from surveys in 1977 and 2008 shows mainly bed lowering, especially between FPkm 4.5 and 6 in the North Fork Reach and between FPkm 8 and 12 in the Mill Creek and Emily Creek reaches. In these locations, the channel is consistently 1–2 m lower in 2008 than in 1977 (fig. 18B). This apparent lowering exceeds plausible uncertainties owing to survey accuracy. For the Upper Reach upstream of FPkm 12, net changes in bed elevation between the surveys in 1977 and 2008 have been small. In the Estuary Reach, the difference between the surveys in 1977 and 2008 indicates possible thalweg aggradation for the kilometer downstream of Tide Rock; however, the resolution of the survey in 1977 is poor compared to the bathymetric surveys (fig. 18A), which show net incision of about 1 m between 1939 and 2008.

Sparser measurements from 1980, 1981, and 1982, which were surveyed in relation to the survey in 1977 (Klingeman, 1993), indicate that a substantial part of the channel lowering in the Estuary, North Fork, and Mill Creek reaches occurred before 1982 at some locations (fig. 18B). Examination of the repeat surveys of the cross sections surveyed in 1977 and 2008 (fig. 19) indicate that channel lowering between FPkm 4 and FPkm 12 (corresponding to Rkms 5–13) was independent of the rest of the active channel, as bar elevations appear similar in 1977 and 2008 (particularly for cross sections E, F, and G in fig. 19).

Information collected during the course of flow measurements at the USGS streamflow-gaging station at FPkm 15.24 provides another source of quantitative information on channel change (fig. 20). The specific gage analysis (fig. 20A) encompasses 39 ratings over nearly 30 years. The large number of ratings is indicative of frequent changes in local geometry and substantial bed-material transport. For comparison, the South Umpqua River near Brockway has had only 11 ratings since 1942 (O’Connor and others, 2009). The ratings for the lower discharges are sensitive to scour and fill of low-flow pools and riffles near the measurement section, and consequently show more variation. For example, the rating for the 5.5 m³/s flow shows an overall trend of bed lowering after a period of slightly higher stages in the late 1970s, consistent with the ratings for all discharges, but with a total variation of 1.2 m. The ratings for the larger flows reflect more general reach scale channel and flood-plain conditions, including the volume of gravel in the bar flanking the left margin of the channel (fig. 20B), and indicate an overall lowering of flow-stage elevations since 1970, although with smaller magnitudes of change. Within the overall lowering trend, however, the high-flow ratings show evidence of aggradation and narrowing in the late 1970s and in 1997, after the 2,169 m³/s peak discharge in 1996. For all flow ratings, however, the overall trend has been a net decrease of stage associated with specific discharges, ranging from 0.86 m for the low discharges to 0.28 m for the high analyzed discharges. The series of ratings, especially for the larger discharges, also indicate aggradation of approximately 0.2–0.3 m culminating between 1976 and 1978, followed by nearly continuous decrease until an episode of aggradation in the late 1990s, interrupted by aggradation and narrowing after the 1996 flood. Since 2000, all ratings have decreased between 0.2 and 0.4 m (fig. 20A).

Historical Channel Change

The main observation from the planview mapping is a large decrease in bar (and bare gravel) area along the entire study area between 1939 and 2008. Historical changes in bar area, channel width, and sinuosity have been greatest near the confluence of the North Fork Chetco River, within the Mill Creek and North Fork Reaches, and downstream through the Estuary Reach. The largest changes were between 1965 and 1995, with the periods before and after showing little change or perhaps even opposite trends. The repeat surveys and specific gage analysis indicate that the overall historical vertical change has been bed lowering. Repeat surveys in the estuary show that the channel in 2008 was on average about 0.5 m lower than in 1939. Similarly, stretches of the Emily Creek, Mill Creek, and North Fork reaches appear to have channel thalweg elevations as much as 2 m lower in 2008 than in the surveys of 1977, with much of the lowering perhaps between 1977 and 1981. The specific gage analysis at FPkm 15.2 (Upper Reach) indicates episodes of aggradation in the late 1970s and late 1990s, but overall a long-term trend of bed lowering.

Many factors are likely responsible for these changes, including (1) direct physical alteration of the river corridor by bank stabilization and development, (2) bars evolving to flood plain by accumulation of overbank sediment and vegetation colonization, (3) changes in the volume of bed-material sediment brought into the study reach from upstream and tributary sources, either because of flow history or drainage basin conditions, (4) changes in the volume of sediment transported out of the study area by fluvial processes or by dredging and gravel extraction, and (5) floods, which are commonly a catalyst for change.

For the Estuary Reach, the channel and flood plain have been extensively modified by dredging, jetty construction and development between FPkm 0 and 2. Upstream within this reach, commercial aggregate removal may be a factor in decreased bar areas, but bank protection, fill, and development also has reduced bar area.

For the North Fork and Mill Creek reaches, the planview changes reflect the complicated interplay between the normal pattern of meander growth followed by cutoffs in wandering and sediment-rich rivers (Church, 1983; O’Connor and others, 2003), episodic tributary sediment input from the North Fork Chetco River and possibly Jack Creek, large mainstem floods triggering episodes of channel change, and the direct channel disturbance and indirect consequences of the long history of substantial gravel extraction in this reach. The channel lowering, decreased recent rates of channel migration, diminished bar area, and lesser amounts of bare gravel and sparse vegetation are all mutually consistent changes indicative of transformation from sediment surplus to bed-material deficit. Such transformations would promote the conversion of bars to flood-plain surfaces as illustrated in figure 21.

The Upper and Emily Creek reaches have had more stable planforms, reflecting the strong control imposed by the closer valley walls. Although bar elevations were similar in 1977 and 2008, the Emily Creek reach shows evidence of decreased bar area (fig. 9) and local bed lowering (fig. 18) between 1939 and 2008. Similarly, the specific gaging station analysis shows general trends of bed lowering and bar erosion near the gaging station in the Upper Reach since the late 1970s (fig. 20). The changes in these two reaches could be either the result of reduced supply from upstream relative to transport capacity or incision propagating from downstream areas where there has historically been substantial gravel extraction.

An important factor in the evolution of channel and flood plain of the lower Chetco River is the history of large flows, because they are probably responsible for bringing in large volumes of sediment and triggering channel change. The largest recorded flow was 2,155 m³/s on November 19, 1996, but a flood with an estimated discharge of 2,420 m³/s on December 22, 1964, (http://wdr.water.usgs.gov/wy2008/pdfs/14400000.2008.pdf), was of exceptional duration and is the largest known flood for the river (Soil Conservation Service, 1979). Anecdotal accounts describe substantial sedimentation along the Chetco River as a consequence of the 1964 flood (Maguire, 2001, p. 9), similar to that documented for several northern California drainages along the Pacific coast (Stewart and LaMarche, 1967; Kelsey, 1980; Madej, 1995). The flood of 1964 in particular caused major and persistent sedimentation in the Klamath Mountains area because of the great volumes of hillslope material eroded and delivered to the channels during the storm and ensuing flood (Hickey, 1969; Waananen and others, 1971; Lisle, 1981; Harden, 1995). For several southern Oregon and northern California coastal drainage basins, the large volumes of sediment transported to the main channels led to periods of aggradation for mainstem rivers, including the nearby Smith River, for as long as 15 years after the flood, followed by periods of channel incision (Lisle, 1981). Some changes on the lower Chetco River, such as the late 1970s aggradation at the USGS streamflow-gaging station and the subsequent channel lowering (and the attendant reduction in bar areas) may be a similar decadal time-scale response to this particularly significant flood.

Revised July 2012

First posted May 26, 2010