The Truckee River in western Nevada has severely incised in response to a net 20 m of lake-level lowering of Pyramid Lake over the past 120 yr, leaving a suite of cut terraces that extend ∼15 km upstream from the lake where incision was arrested by bedrock in the channel. Channel planform changes and terrace development were mapped over a 70 yr period using rectified aerial imagery and light detection and ranging (LiDAR) data in ArcGIS. Discrete point elevations were extracted from the LiDAR data for channel remnants on each of the different-aged terraces and combined with channel distance measurements derived from the appropriate aerial imagery to derive true channel gradients for each of the photo years. These same point elevation measurements were also resolved to a common valley distance (CVD), as is often done when studying prehistorical terrace sequences. Comparison of the true channel distance (TCD) profiles to the CVD profiles demonstrates that the TCD profiles are longer and less steep than the CVD profiles. These differences exist because resolving terrace elevations to a CVD does not account for decreasing sinuosities and straighter channel planforms through time.
As base level fell to its historical low elevation, the channel became steeper, straighter, and smoother, which increased the rate of sediment transport and delivery to Pyramid Lake. Over the past 120 yr, ∼60,000,000 m3 of sediment have been removed from the bed and banks of the lower river and redeposited in the lake, temporarily increasing the sedimentation rate. Larger trenches through deltaic and lacustrine deposits occupied by rivers and streams flowing into Pyramid Lake and other closed basins may reflect similar rapid responses to lowering base levels at times in the past. Therefore, lowering lake levels at Pyramid Lake and elsewhere during relatively dry times may be at least partly responsible for transient increases in sedimentation rates in lacustrine archives.
Stream terraces are common features found along many fluvial systems, and correlated remnants represent the abandoned active channel and floodplain of the stream. Their formation is influenced by a variety of natural factors, including changing sediment loads, tectonics, and changing base levels (Born and Ritter, 1970; Bull, 1990; Merritts et al., 1994). Anthropogenic activities such as dam building, water consumption, and water diversion can also alter the dynamics of fluvial systems, leading to terrace formation and modification (Howard and Dolan, 1981; Hassan and Klein, 2002; Ben Moshe et al., 2008). The geometry of stream terraces is often used to infer rates of uplift or other types of deformation (Personius, 1995; Keller et al., 1998) and processes and rates of incision or base-level change (Leopold and Bull, 1979). The key to effectively using terraces for these purposes is being able to correlate isolated terrace remnants to one another so that their history and gradients can be reconstructed. This is often not simple or straightforward, and a variety of approaches have been used to correlate terraces (Knuepfer, 1988; Bull, 1990; Merritts et al., 1994). A factor that is often implicit in these analyses is that the gradient of the former stream channel matches the gradient of the correlated terrace remnants, but this does not account for changes in sinuosity through space and time. Davis et al. (2009) attempted to correct for differences in sinuosity amongst late Pleistocene and Holocene terraces when plotting long profiles, but their efforts were necessarily limited by an incomplete knowledge of the geometry of the channels that formed the terraces—an unfortunately all too common situation.
This study uses a historical suite of terraces that formed along the lower Truckee River in western Nevada in response to net lake-level lowering at Pyramid Lake to document processes of terrace formation. These terraces, formed by the upstream migration of headcuts and lateral planation (Born and Ritter, 1970), extend ∼15 km upstream from the lake where incision was arrested by bedrock outcrops in the channel and by a small dam. Georeferenced aerial photographs and imagery dating from 1938 to 2010 combined with light detection and ranging (LiDAR) topography have allowed detailed mapping of the formation of these terraces through time. This data set has also allowed long profiles of the lower Truckee River channel to be constructed for eight different photo years and compared to long profiles reconstructed using a common valley distance, which is the method commonly used when working with prehistorical terraces (Merritts et al., 1994). The purpose of this paper is to further document the development of the historical Truckee River terraces and their long profiles in response to base-level lowering and to place these results into a broader context by exploring the implications for sediment transport and delivery to the sedimentary archive accumulating in Pyramid Lake and in other closed basins.
BACKGROUND AND PREVIOUS WORK
The Truckee River in eastern California and western Nevada has its headwaters at the crest of the Sierra Nevada near Lake Tahoe, and it flows ∼170 km to the northeast until it terminates in Pyramid Lake (Fig. 1), an endorheic remnant of Pleistocene Lake Lahontan. The Truckee River basin encompasses ∼7050 km2, and its hydrology is largely driven by winter snows melting in the spring, punctuated by relatively brief, high-magnitude floods (Fig. 2).
The largest floods (>10,000 cfs; >283 m3s−1) over the last 100 yr on the lower Truckee River occurred in 1907, 1928, 1937, 1950, 1963, 1964, 1986, 1997, and 2005, with the flood of record occurring in 1997 at the Below Derby and Nixon gauges. These floods were primarily the result of heavy rains or rain-on-snow events. Although most of these floods only lasted a few days, a notable exception was the 1986 flood, which maintained a discharge above 5000 cfs (142 m3s−1) at the Nixon gauge from mid-February through mid-March of that year. More common spring snowmelt floods typically attain flows ranging from 2000 to 4000 cfs (57 to 113 m3s−1) at some point in April, May, and/or June. Relatively high spring snowmelt peaks ranging from 5000 to 7000 cfs (142 to 198 m3s−1) occurred in 1938, 1952, 1958, 1967, 1983, and 1996. In many years and sometimes in successions of years, such as 1928–1936 and 1987–1994, spring snowmelt flows did not attain even 2000 cfs (57 m3s−1).
Over the long term, the surface elevation of Pyramid Lake has been a major controlling influence on the geomorphology of the lower Truckee River. The surface of Pyramid Lake is base level for the lowermost river and controls aggradation and degradation processes as the lake rises and falls through time. This process has been ongoing throughout the Holocene, as indicated by multiple Holocene terraces flanking the active channel (Bell et al., 2005a). The oldest of these terraces (Qtn) stands ∼10–20 m above the modern river and ranges in age from ca. 6800 to younger than 10,800 14C yr B.P. Two younger terraces, standing ∼5 m (Qty3) and 3 m (Qty2) above the modern river, range in age from ca. 1300 to 2100 14C yr B.P. and 500 to 700 14C yr B.P., respectively (Bell et al., 2005a). Both of these late Holocene terraces are graded to an elevation near the historical highstand of ∼1182 m, which can be reasonably interpreted to mean that Pyramid Lake has probably not risen above this elevation in the past 2000 yr (Adams, 2008).
The reason that the surface elevation of Pyramid Lake maintains a relatively constant level around 1175–1180 m is because Pyramid spills into Winnemucca Dry Lake via Mud Lake Slough (Fig. 1) when lake level rises above ∼1177 m (Hardman and Venstrom, 1941). Further lake-level rise at Pyramid is then constrained until Winnemucca Dry Lake fills to a similar elevation. The volume of Winnemucca Dry Lake below 1177 m is ∼5 km3, and its surface area ∼245 km2, both of which act together as an effective buffer to rapid lake-level changes at Pyramid Lake in this elevation range.
The long-term implications of this arrangement are that deltaic deposits have repeatedly accumulated in the 1175–1180 m elevation range during the Holocene, forming a relatively large compound delta complex (Adams, 2008). Only in past extended periods of relatively dry climate has Pyramid Lake dropped to low levels and exposed the delta front, a situation mimicked over the past 120 yr by anthropogenic reductions in Truckee River flows.
The written record for Pyramid Lake began in 1844, when the surface of the lake was at 1176.5 m. It then rose to its historical highstand of ∼1182.3 m in 1868 and to a similar elevation in 1891 (Hardman and Venstrom, 1941; Harding, 1965). When Russell (1885) mapped the lower Truckee in 1882, the river was primarily a single-stem, meandering stream lined by cotton woods and willows. As the river approached Pyramid Lake, however, it bifurcated, with a portion flowing through Mud Lake Slough and the rest flowing into Pyramid. Because of large-scale diversions and consumption beginning in the late nineteenth century, the surface elevation of Pyramid Lake has dropped dramatically, reaching its historical lowstand of ∼1153.4 m in 1967 (Glancy et al., 1972). Since that time, the lake surface has undergone several fluctuations spanning several meters and is now (2012) at 1159.2 m (Fig. 2).
This lowering of base level initiated incision in the lowermost river (Born and Ritter, 1970) that has progressed ∼15 km upstream to Numana Dam, which was built in the 1930s (Fig. 1). Bedrock forms the bed of the channel just below the dam, suggesting that incision may not have progressed further upstream even if the dam was not present. Upstream from this small dam, the river has not been affected by base-level lowering, but it has been impacted by other anthropogenic activities, including channelization, clearing, widening, and damming, most of which were completed by the 1960s (Horton, 1997). Numana Dam probably does not greatly affect the sediment dynamics of the river through this reach because of its low height (∼3 m) and minimal storage capacity. Destabilization of the channel within and upstream from the study area, however, may have led to increased scour and widening during formation of the historical terraces. Marble Bluff Dam (Fig. 1) was built in 1975 to control further erosion and also does not have a large sediment or water storage capacity. Communications with U.S. Bureau of Reclamation personnel indicate that sediment does not accumulate behind the dam but instead passes over its relatively low lip.
Using repeat aerial photography, Born and Ritter (1970) documented the formation of a complex sequence of erosional terraces along the lowermost river over a period of 44 yr (1925–1969); these terraces were cut into modern through Pleistocene deltaic deposits that are primarily composed of sand, silt, clay, and gravel (Born, 1972). Although Born and Ritter (1970) were able to assign historical ages to a particular flight of terraces, they were unable to confidently correlate those individual terrace remnants to terrace remnants at different sites. Because of this uncertainty, Born and Ritter (1970) were unable to construct longitudinal profiles of the terrace remnants.
Glancy et al. (1972) further documented terrace formation by measuring the area and estimating the mass of bank erosion along the river in the 1969 water year (1 October 1968–30 September 1969) and utilizing aerial photographs from immediately before and after the 1969 water year. Although peak stream flows were not particularly high during the 1969 water year (<4500 cfs; <127 m3s−1), relatively high flows (2000–4500 cfs; 57–127 m3s−1) persisted for 127 consecutive days during the spring, causing Pyramid Lake to rise by ∼2 m (Glancy et al., 1972). Glancy et al. (1972) combined analysis of the air photos with field measurements to determine that ∼6,200,000 metric tons (MT) were eroded from the bed and banks of the Truckee River below Numana Dam and delivered to Pyramid Lake.
In a study of sedimentation rates in Pyramid Lake using automated samplers, Anderson (1977) reported large plumes of suspended sediment extending several kilometers north of the delta front in spring of 1975, when discharge reached a maximum of ∼3500 cfs (99 m3s−1). His periods of observation (2–3 wk), however, were too short to document annual or decadal sedimentation rates that were affected by a large range of flows.
The current study utilizes all of the aerial photography from Born and Ritter (1970) and Glancy et al. (1972) in addition to many generations of aerial photos collected in the years since those studies (Table 1). In 2008, high-resolution LiDAR data first became available for the Truckee River corridor, and these have allowed detailed mapping of the sequence of historical terrace formation along the lower river and assignment of each of the terrace remnants to a particular time interval and channel sinuosity.
This study uses rectified aerial imagery and LiDAR data from 2008 and 2010 viewed and mapped with ArcGIS® combined with field observations to document terrace development along the lower Truckee River over the past 70 yr. The primary data sources consisted of hard copies of aerial photographs, digital aerial imagery, LiDAR topography, and geologic maps (Bell et al., 2005a, 2005b).
The earliest aerial photographs cover the study area date from 1938, permitting the first detailed look at the lower river and providing a significant benchmark from which historical terrace development can be documented. Additional rectified imagery and LiDAR data from 2008 and 2010 (Table 1) provide a detailed time series of terrace development that facilitated mapping and assignment of each terrace remnant to a particular time period.
The color and black and white aerial photographic prints were scanned and digitized using a flat bed scanner. Scan resolution was generally 1200 dots per inch (dpi), and files were saved as TIFFs or JPEGs, mostly in grayscale, although some photos were scanned as color. Using the scan resolution, print dimensions, and digital image dimensions (picture elements or pixels), the nominal ground resolutions of the aerial photographs were calculated; for the 1938 1:20,000 scale prints, the ground resolution is better than 1 m, for the 1:60,000 scale photographs from 1954, the ground resolution is ∼3 m. The ground resolution for the digital data sets from 1994, 2000, 2006, 2008, and 2010 is 1 m or less.
The multidate, multiscale aerial photographs of the lower Truckee River were rectified to the 0.15 m 2000 and 1 m 2006 imagery using a standard polynomial-based image-to-map rectification process and the georeferencing tools in ArcGIS® v. 9.1 and 9.2. This process involved aligning the now raster-based photos to the UTM NAD83 Zone 11 coordinate system by shifting, rotating, scaling, skewing, and warping the images. Although orthorectification is more accurate, it was not attempted because the camera parameters required to build interior orientation (Thieler and Danforth, 1994) were not available for the older photographs.
The image-to-map rectification process involved the selection of ground control points (GCPs) common to both the scanned aerial photography and the ortho-imagery from 2000 and 2006. Several rule bases were developed for the point selection process in order to minimize potential errors that could accumulate and contribute to inaccurate results. Common GCPs were generally selected at a scale of 1:3000 or larger and commonly consisted of shrubs, rocks, and sometimes buildings, bridge abutments, road intersections, and other anthropogenic features. Care was taken to be cognizant of shadowing effects in the imagery when selecting GCPs, as these sometimes distorted the precise location of a feature. To avoid the introduction of spatial errors due to relief distortion, control points were preferentially selected in areas that were close to the elevation of the river. When possible, control points were also selected near the center of each photo to avoid errors due to lens distortion and camera tilt.
Once the images were georeferenced, they could be directly compared to one another and to the LiDAR topography in ArcMap. Although they have slightly different spatial extents, the 2008 and 2010 LiDAR data sets are essentially identical, except that the river channel bathymetry is better resolved in the 2010 data, but landforms are better resolved in the 2008 data set. Therefore, the 2008 LiDAR data set was used in most of the analyses in this paper, unless otherwise stated. The 2008 LiDAR data set was processed to remove vegetation (bald earth) and has a horizontal and vertical accuracy of better than ±0.75 m and ±0.2 m, respectively. The LiDAR digital elevation models (DEMs) were processed as slope maps and given an inverted monochromatic color ramp. Basically, this processing technique provides better contrast for subtle geomorphic features like small scarps than hillshaded DEMs in low-relief areas because all breaks in slope are highlighted by dark shades (Fig. 1). In this study, only the terrace surfaces were mapped, excluding the terrace risers, because these are cut terraces formed on older deltaic and other lacustrine deposits (Born and Ritter, 1970).
Mapping of the different terraces and assignment of ages proceeded through time beginning with the 1938 images and ending with the 2008 LiDAR data. A polyline shapefile was used to map the fluvial features on the 1938 images, including channels, bank margins, and scoured areas. Next, the same types of features were mapped on the 1954 images. The two polyline shapefiles were then merged and converted to a polygon shapefile. Next, the polygon shapefile of eroded areas was compared to the LiDAR slope map to locate terrace surfaces that had formed between 1938 and 1954 and that still remained (Fig. 3). Positive identification of surfaces was enabled by the presence of subtle, braided channel patterns and other geomorphic features that could be identified on both the 1954 image and on the LiDAR slope map (Fig. 4). The same procedure was used to identify terrace surfaces that formed between 1954 and 1961, 1961 and 1965, 1965 and 1968, and so on. To help keep track of the ages of the different terraces, a point shapefile was used to assign ages to each of the terrace surfaces as they were identified and mapped. These point attributes were later intersected with the polygons representing the different terrace surfaces to create a geomorphic map. Some adjustment of the terrace surface boundaries was necessary because of occasional slight offsets between the aerial photos and the LiDAR slope map. This did not, however, affect the assignment of ages to the different terrace surfaces.
Mapping the channel planforms for each of the photo years was a simple matter of using polyline shapefiles to map the boundaries and centerline of each of the channels. In addition, a midvalley axial line (common valley distance line) was mapped down the center of the river valley, generally bisecting the terraces on each side of the channel (Merritts et al., 1994). All of these lines were then divided into 100 m segments for measurement purposes. Numana Dam was designated the 0 m distance and values increase downstream.
The mapped channel planforms were then sequentially overlain on a semitransparent map of the terrace surfaces, which, in turn, was draped over the LiDAR slope map to identify channel remnants of specific ages that were still present in the 2008 landscape (Figs. 5A, 5B, and 5C). For a particular photo year, the elevations of points identified along the channel remnants preserved on terrace treads were then extracted from the LiDAR DEM. These elevations were resolved to the appropriate true channel distance, as measured downstream from Numana Dam, to reconstruct the gradient of the channel as it existed when the aerial photo was collected. These same points were also resolved to the common valley distance, as is typically done when studying prehistorical terraces (Merritts et al., 1994). All reconstructed profiles used the edge and elevation of Pyramid Lake at the time of the associated photos for the end points of each profile.
Sediment samples were collected from multiple terrace risers, beneath the thin (0.5–1 m) sand and gravel units capping the terraces, to determine their composition (Fig. 1). All samples were analyzed for particle-size distributions at the Soil Characterization Laboratory at the Desert Research Institute.
The total volume of sediment removed below Numana Dam in historical times was calculated by first creating a topographic surface using the 2008 LiDAR data that essentially mimicked the 2008 topography except that the new surface also extended across the trench. This was accomplished by first extracting elevation points from the LiDAR data near the rim of the historical trench and then augmenting this data set by adding additional points above the trench, for which elevations were interpolated from adjacent points on the rim of the trench. This synthetic point cloud of elevation data was then converted into a raster digital elevation model (synthetic DEM). The 2008 LiDAR DEM was then subtracted from the synthetic DEM using the Cut Fill tool within 3D Analyst, which returns the volume of removed sediment.
An attempt was made to calculate the amount of sediment removed between each photo year by using the length of shadows created by terrace risers on a given photo compared to the length of shadows created by buildings or natural features of known height. Due to the less than optimal photo resolution and unfavorable orientation of many terraces risers with respect to sun angle on each of the photo sets, this approach was deemed insufficiently accurate and was thus abandoned.
The BAGS sediment transport program (Pitlick et al., 2009) was used to examine the influence of a steepening slope on sediment transport rate during downcutting. For these scenarios, the 1961 and 1965 channels in the vicinity of the Born and Ritter (1970) terraces were used because of their similar width (∼500 m) and because of notable geomorphic changes during this short span. For each of the scenarios, the initial particle-size distribution of sandy fine gravel was also kept constant. These scenarios were designed to provide a first-order estimate of the magnitude of changes in sediment transport rate due to differences in slope and discharge. It is recognized that the details of this process were probably far more complicated.
A time series of images of the lower Truckee River is included in Figures 6A to 6M and as a slide show (Supplemental File1) that shows the development of the terrace sequence beginning in 1938 and extending to 2010. This time series was used in the manner described above to develop the geomorphic map of the lower river shown in Figure 7. For this map, the boundaries of the Qty2 and Qty3 units were slightly modified from Bell et al. (2005a) to better fit the detailed topography, which was not available at the time of their work. This new map presents the age and distribution of all historical and late Holocene terraces along the lower Truckee River. A similar map shows the changes to the channel centerlines through time (Fig. 8), another measure of the dynamics of this system.
The long profiles of the channels were plotted using both the true channel distance (TCD) and common valley distance (CVD) values. Figure 9 shows the long profiles of the 1938 channel plotted both ways, which emphasizes the differences in slope. Because the TCD was longer (more sinuous) than the CVD, it is not as steep as the profile constructed with CVD.
The evolving long profiles of the channel, sampled at discrete points in time and using the TCD, are presented in Figure 10A. Note that not all photo years were used to reconstruct long profiles. The profiles in Figure 10A prescribe a regular pattern where the channels first get shorter and steeper as lake level declined and approached its lowest historical elevation in 1967. After 1967, lake level began to rise, and the distal parts of the channels generally became longer and less steep. This pattern is somewhat complicated by the installation of Marble Bluff Dam in 1975, which is the cause of the abrupt step in the profiles from 1977 to the present. The downstream separation of TCD profiles is an indication of the amount of incision per unit channel length but does not necessarily reflect the elevation differences between adjacent terrace surfaces of different ages.
The elevation differences between adjacent terrace surfaces are better shown by the common valley distance profiles (Fig. 10B). These types of profiles are not nearly as different from one another, particularly after 1938 (Fig. 10B), and they appear to reflect much less incision through time. This is the type of plot that is typically generated when studying prehistorical terraces. What is also apparent from this plot is the amount of incision that has taken place since 1969 below Marble Bluff Dam, in places reaching nearly 10 m, which has led to a lower-gradient channel.
Figure 11 presents one of the more dramatic examples of the differences in long profiles due to using TCD versus CVD. The plots of the 1961 and 1965 stream profiles using TCD clearly show that the 1965 channel is shorter, steeper, and graded to a lower lake level than the 1961 channel, which was more sinuous. When plotted using a CVD, however, the two profiles appear to have the same gradient, only with the 1965 profile extending to a lower lake level (Fig. 11). The reason for these differences is that between 1961 and 1965, there were two relatively large floods (Fig. 2) that straightened, steepened, and in many places transformed the river from a single-stem, meandering channel to a braided channel (Figs. 6C and 6D). The important difference between the two types of profiles is that the true channel distance profiles accurately show the changes in stream gradient that occurred over the period of record, which is relevant to sediment transport calculations.
Based on results from the Cut Fill tool within ArcGIS, which was used to subtract the 2008 DEM from the synthetic DEM, the total volume of sediment removed from the historical trench since 1891 amounts to ∼60,000,000 m3. Assuming a bulk density of ∼1.4 g/cm3, which is the average for sediment samples collected from deltaic sediments along the lower Truckee River (Fig. 1; Table 2), the total mass of sediment eroded from the bed and banks of the lower Truckee River and delivered to Pyramid Lake in the past 120 yr is 84,000,000 metric tons (MT).
The cut or strath terraces along the lowermost Truckee River are typically covered by 0.5–1 m of fine gravelly sand to sandy fine gravel, which overlies much finer silt, clay, and sand deltaic and other lacustrine deposits. In places, there are a few beds and lenticular bodies of both fluvial and beach gravels exposed within terrace risers and cutbanks that are interbedded with the finer-grained deposits. Volumetrically, however, the fine-grained deposits are more important. Upstream from the historical highstand, terraces are more commonly composed of fluvial sands and gravels.
The effect of changing slope on sediment transport rate was investigated using the 1961 and 1965 true channel slopes near their distal ends where they emptied into Pyramid Lake. In 1961, the slope of the channel was ∼0.003 in its lower reaches but had steepened to ∼0.0047 by 1965 due to channel straightening, shortening, and incision (Fig. 11A). Keeping all other parameters constant, the sediment transport rate was estimated for a variety of discharge values (Fig. 12). Transport rates associated with the steeper slope are much higher and increase at a faster rate with increasing discharge than do rates associated with the gentler slope.
The detailed photographic and topographic record along the lower Truckee River makes this an ideal site to examine processes and rates of terrace formation and sediment delivery to a receiving basin. Granted, there is only a series of snapshots from a continuous movie (Supplemental File [see footnote 1]), but these images track the development and modification of this terrace sequence in far more detail than can usually be gleaned from studies of prehistorical terraces. Additionally, the LiDAR data have allowed the calculation of the total mass of sediment delivered to Pyramid Lake from incision by the river over the past 120 yr. In a way, the lower Truckee River can be thought of as a full-scale experimental flume that has been run with known discharge rates in real time over a period of 70 yr, with snapshots taken about every 10 years or less.
Based on observations and gauge records, Born and Ritter (1970) proposed a model for cut terrace formation along the lower Truckee River that included “(1) steady, although intermittent, lake-level decline; (2) degradation of the main river channels in response to lowering local base level; (3) periodic and extensive lateral erosion associated with high discharge and elevated river stage; and (4) stranding of lateral elements of the valley bottom as a result of renewed downcutting by the river.” In the 40 or so years since Born and Ritter’s (1970) original study, cut terraces have continued to form along the lower Truckee River, and there are now more data and observations that confirm their model.
When Pyramid Lake began to recede from its historical highstand of ∼1182 m, the Truckee River extended its length and incised into the former delta front, which had a convex-up slope of ∼0.0035–0.004, steepening to ∼0.014 as lake level approached 1168 m. Although few of the photos were taken during periods of high flows, the images from 1965 (1300 cfs; 37 m3s−1) show the lower Truckee River as a braided system filling the valley floor and actively eroding its banks (Figs. 5B and 6D). During larger flow events, not captured by the photos, it is likely that the lower Truckee River was also braided and covered the entire floor of its trench while both laterally eroding its banks and incising its bed.
Between 1938 and 1954, the width of the scoured trench in the vicinity of the Born and Ritter (1970) terraces more than doubled, from ∼250 m to ∼550 m, probably as a result of three moderately large flow events ranging from 5600 to 8600 cfs (159 to 244 m3s−1). During the same period, the channel also incised by ∼5 m at that location, as measured by the height of terrace risers formed between 1938 and 1954 (Fig. 13). Between 1954 and 1961, the river had incised an additional 1 m, probably during two flow events of ∼5000 and 6000 cfs (142 and 170 m3s−1). Two large floods of ∼11,600 cfs (329 m3s−1) and 8600 cfs (244 m3s−1) occurred between 1961 and 1965, causing an additional 2 m of dissection (Fig. 13). Between 1968 and 1991, the river further downcut ∼2 m in this area.
During periods when relatively low flows were recorded between photo dates, there was much less incision. Between 1969 and 1977, when flows only twice reached around 3000 cfs (85 m3s−1) (Fig. 2), only ∼0.6 m of incision was measured where 1969–1977 terrace remnants abut against 1968–1969 remnants. Similarly, less than 1 m of erosion was recorded between 1991 and 1994, and after 2006, both of which are periods when all flow events were ≤2500 cfs (<71 m3s−1) (Fig. 2).
The amount of incision appears to broadly correlate with the number and magnitude of large flow events, with more frequent and larger discharge events corresponding to deeper incision. During periods of lower flows, however, incision does still occur but at a lower rate. Ongoing incision is likely caused by the inherent instability of the system over the past 120 yr as the river continually seeks a new equilibrium.
Terrace formation occurred in virtually every time period bracketed by the aerial photographs listed in Table 1, even though these photo periods represent time intervals ranging from 1 to 16 yr. A notable exception is the time period from 1991 to 1994, when river flows did not exceed 2000 cfs (57 m3s−1) (Fig. 2). During some periods bracketed by the photo pairs, multiple terrace treads were formed in the same sequence as in the period from 1938 to 1954 (Figs. 4 and 13), offering further evidence of the dynamics of this system. Not all of the photo pairs that document terrace formation encompass large floods, suggesting that even moderate flows can be effective in vertical and lateral erosion, and this is a key to terrace preservation in this environment. An example is the 1969 water year, which did not have relatively high peak discharges (<4500 cfs; <127 m3s−1) but maintained a discharge of >2000 cfs (>57 m3s−1) for a period of 127 days. Lateral bank erosion was as high as 125 m along certain segments during the 1969 water year. Further downcutting of ∼0.6 m, induced by relatively low flows between 1969 and 1977 (Fig. 2), helped preserve the terrace treads formed during the 1969 water year (Fig. 7).
The long profile of the lower Truckee River has also evolved as the river has responded to lowering base levels at Pyramid Lake. In 1938, the true channel distance long profile was slightly concave-up for ∼18 km below Numana Dam. In the lower 5 km, however, the channel became convex-up, as incision had not kept pace with the fall in base level at Pyramid Lake (Fig. 9A). Through time, the channel became shorter, steeper, and developed a more pronounced convex-up shape, even though the southern shore of Pyramid Lake was receding until the mid-1960s, when the lake achieved its historical lowstand. Marble Bluff Dam was installed in 1975 to arrest incision, and since that time, the long profile of the lower Truckee River has evolved to prescribe a slightly concave-up shape, interrupted by the step of Marble Bluff Dam, as indicated by the 2008 profile (Fig. 10A).
The true evolution of the long profile of the lower Truckee River is not nearly as evident when examining the CVD profiles (Fig. 10B). These profiles were constructed using a common valley centerline to which the channels from the various time periods were resolved. Construction of long profiles in this way masks the dynamics of channel evolution because this approach does not account for changes in slope and sinuosity through time; this is particularly evident when comparing the 1961 and 1965 profiles constructed using CVD (Fig. 11).
During this brief 4 yr period, Pyramid Lake dropped by ∼4 m, and two relatively large floods occurred. As a result, along its lowermost reaches, the channel transformed from a single-stem meandering stream into a straighter, braided stream that had clearly incised the 1961 floodplain, producing a cut terrace (Figs. 5A, 5B, and 5C). The 1961 and 1965 CVD profiles, however, have the exact same gradient, and there is no apparent incision, which is clearly not the case. Comparison of the 1961 and 1965 TCD profiles documents that the 1965 TCD profile is steeper and shorter than the 1961 TCD profile (Fig. 11).
In summary, examination of the CVD profiles would suggest that the lower Truckee River underwent relatively minor incision as the channel lengthened because of receding lake levels. Instead, the channel actually straightened and steepened with time as lake level receded toward its historical lowstand.
It is probably rare for the sinuosity of a channel associated with a particular terrace to be effectively documented, let alone the evolution of channel sinuosity through time. This is why researchers necessarily use a common valley distance when constructing long profiles of prehistorical cut terrace remnants that were formed as a response to base-level fall. This restriction masks the true dynamics of the system, where the increase in slope from channel straightening through time greatly increases the sediment transport rate, thereby enhancing incision.
Figure 12 shows the changes in sediment transport rate due to an increase in slope. The rate is much higher in the steeper channel at all discharges and increases at a higher rate with increasing discharge than does the less steep channel. This order of magnitude increase in sediment transport rate underscores the importance of changing slope on the dynamics of stream channels responding to base-level lowering. Further, the increase in rate of sediment transport probably reflects the degree to which the system is out of equilibrium. Through time, and with a stable base level, the gradient of the channel would likely lesson, thereby decreasing the sediment transport rate as the system approached a new equilibrium.
The total volume of sediment removed from the historical trench amounts to ∼60,000,000 m3 over the past 120 yr. If those sediments had an average bulk density of 1.4 MT/m3, then the total mass of eroded sediment would be ∼84,000,000 MT. This is a minimum estimate of the total sediment load delivered to Pyramid Lake over the past 120 yr because it does not include estimates of suspended load and bed load derived from upstream of Numana Dam. It is likely that most of this sediment ended up in Pyramid Lake because evidence for eolian transportation out of the trench is minimal. If spread across the roughly 250 km2 of the flat floor of Pyramid Lake, the redeposited blanket of sediment would be ∼25 cm thick, assuming no change in density. Some of the sediment, however, was probably deposited at the front of the late twentieth century delta, as there is evidence of local slumping in this area (Fig. 14). It is interesting to note that Benson et al. (2002) documented ∼40 cm of deposition since 1860 at their core site PLB98–2, ∼25 km north of the delta front (Fig. 14). This historical deposition rate of 0.29 cm/yr is not that much higher than the late Holocene rate of 0.23 cm/yr for Pyramid Lake, calculated from the age model for core PLB98–2 (Benson et al., 2002). It is, however, much higher than the late Pleistocene rate of 0.05 cm/yr calculated from the age model for core PLC92B for the period 17–21 ka, located at the north end of Pyramid Lake (Benson et al., 1997). While it is not likely that all of the sediment deposited at site PLB98–2 since 1860 was derived from erosion of the historical Truckee River trench, this type of transient source must be considered when assessing the long-term accumulation of sediments within Pyramid Lake. Additional sources of sediment may have been Mullen Creek on the west side of Pyramid (Fig. 14) or other tributaries that were also affected by base-level lowering.
The fill-cut Qty2 and Qty3 terraces are inset into a much larger trench than the historical one, which was excavated at the end of the Pleistocene when Lake Lahontan was receding to low levels (Bell et al., 2005a). The total volume of this trench, which extends at least tens of kilometers upstream from Numana Dam (Bell et al., 2005b), was not estimated, but if late Pleistocene base-level lowering caused rapid incision of the trench, then the sedimentary record of Pyramid Lake should reflect this relatively rapid influx of sediment. Considering that many of the rivers and streams flowing into closed basins in the Lahontan system and elsewhere occupy deep trenches excavated through deltaic and lacustrine deposits, the relatively rapid transfer of sediment toward basin centers during lowstands ought to be reflected by increases in sedimentation rates in the deep basin archives. A more detailed examination of this phenomenon at Pyramid Lake and elsewhere, however, will have to await further work.
The detailed photographic and topographic record that has been assembled to examine terrace development along the lower Truckee River is probably relatively rare. This record provides important insights into the way in which base-level lowering affects fluvial systems and changes sediment dynamics, not just at Pyramid Lake over the last 100 yr, but at other closed basins and in other time periods as well. As lake level began to fall at the end of the nineteenth century, exposing the relatively steep front of the Truckee River delta, the river began to incise and scour its bed and banks as headcuts migrated upstream. As lake level approached its historical lowstand in 1967, the planform and long profile of the Truckee River evolved from one that was relatively sinuous and gentle to one that was straighter, shorter, and steeper. This evolution is typically masked when studying prehistorical terraces because it is difficult to document the sinuosity of channels associated with different terraces. As a result, the gradients of terraces plotted using a common valley distance tend to be more parallel than the channels that were associated with the terrace remnants, which belies their evolution.
The increase in slope caused by channel straightening and incision has a dramatic effect on sediment transport rates. This increase in the rate of sediment transport probably reflects the degree to which the fluvial system is out of equilibrium. The gradient of the channel would likely lessen through time if there were a stable base level, thereby decreasing the sediment transport rate as the system approached a new equilibrium. A version of this process may already be occurring below Marble Bluff Dam, where the river has been flattening its gradient since installation of the dam in 1975 (Fig. 10).
The total volume of sediment removed from the historical trench since 1891 is estimated to have been ∼60,000,000 m3, which corresponds to a mass of ∼84,000,000 MT if a density of 1.4 MT/m3 is assumed. Most of this sediment is likely to have been silt, clay, and sand. The total mass of sediment delivered to Pyramid Lake during that time is likely to be higher because the trench estimate does not account for suspended load and bed load sediment delivered to Pyramid from upstream of Numana Dam. If the 60,000,000 m3 volume of sediment was spread across the 250 km2 relatively flat floor of the lake, it would be ∼25 cm thick, which is somewhat similar to the 40 cm of historical deposition at a core site ∼25 km north of the delta (Fig. 14). Although late Holocene sedimentation rates in Pyramid Lake are similar to the historic rate, late Pleistocene rates were much lower when lake levels were much higher. Transient increases in sedimentation rates in Pyramid Lake and in other closed basin lakes may therefore be explained by stream incision and sediment flushing and focusing in response to base-level lowering during relatively dry periods.
This work was funded by the U.S. Army Corps of Engineers as part of the Urban Flood Demonstration Program, which also provided the LiDAR data sets. I would like to thank the Nevada Bureau of Mines and Geology for supplying some of the aerial photographs used in this study. I am also very grateful to Pat Glancy and Steve Born for taking the time to dig through their respective archives to locate and lend the original aerial photographs from 1968 and 1969. Two anonymous reviewers deserve credit for providing thorough and insightful suggestions that greatly improved the quality of this manuscript.