Chapter 3:Paleohydrology

Rafted Pumice

In the Soloman Butte area, the rounded pumice on the plateau (west of the Williamson River, Figure 3) and south of the plateau was likely deposited as the waters overtopped the high point along the plateau. The pumice that is found in the scour channels was then deposited as the water receded and flowed down the canyon.

Isopleths of the climactic pumice fall indicate that clasts of pumice the size of those that are found rafted in the Wocus Bay area fell out of the plume within approximately 20 km of the vent. The inundation area for the highest stand of the marsh overlapped this isopleth by 2 km, thus offering a possible source for the pumice of the size rafted into the Wocus Bay area. However, pumice clasts were also transported into the basin by the pyroclastic flows, which contain up to 10% pumice clasts in the deposits west of Little Wocus Butte and approximately 50% pumice clasts in the Military Crossing quadrangle to the north. The pyroclastic flows are the likely source for the rafted pumice because they would have transported the clasts well within reach of the backflooding on the western edge of Klamath Marsh. The clasts were then eroded from the pyroclastic flow and rafted throughout the basin. Some distinction can be made between the clasts of pumice that have been water transported verses those that have simply weathered from the pyroclastic flow. The clasts that are found within the pyroclastic-flow section and weathering from it are salmon pink, while those that have been rafted by water or are pyroclastic fall are buff. Roundness alone cannot be used to distinguish mode of transport because the clasts within the pyroclastic flows are also rounded. Where the large rounded clasts were found in linear deposits, in abundances greater than those in the pyroclastic-flow deposits, or in areas with no pyroclastic flow they were classified as rafted.

Reworked pyroclastic-fall deposits

Between the 1397 m elevation of the middle Holocene terrace and the modern marsh level, pumice-rich deposits lack the size distribution and sorting characteristics of the undisturbed pyroclastic-fall deposits. Samples of pyroclastic-fall deposits from above and below 1397 m were collected from pits excavated along the survey transects Wocus Bay East and Hog Creek (Figure 11). The grain size of the pyroclastic-fall deposits decreases down slope from the tread of the middle Holocene terrace (Figure 14). The thickness of the pyroclastic-fall deposits also decreases from the tread. The decreases in grain size and thickness are a result of the backflooding eroding and floating the pyroclastic-fall deposits. The larger clasts of pumice float easier (Whitham and Sparks, 1986) and are therefore less prevalent on the middle Holocene terrace scarp than in the undisturbed section.

Figure 14: Grain size analysis of samples collected from the middle Holocene terrace scarp at Hog Creek and Wocus Bay east.

Figure 15: Thickness of pyroclastic-fall deposits on the middle Holocene terrace scarp at Hog Creek. Five times vertical exaggeration.

Figure 16: Thickness of pyroclastic-fall deposits on the middle Holocene terrace scarp at Wocus Bay East. Five times vertical exaggeration.

Several linear channels have been eroded into the pyroclastic-flow deposits in the Wocus Bay and Soloman Butte quadrangles. The channels in the Wocus Bay quadrangle are cut into a gently sloping plain underlain by pyroclastic-fall deposits (4.5 m) that dips toward the marsh on the west flank of Little Wocus Butte (Figure 11). These channels are about 1.5 to 2 m deep and 5 m wide. Rafted large clasts of pumice are found on the pyroclastic-flow deposits that these channels are incised into and within the channels. The channels in the Soloman Butte quadrangle are carved into a sloping plain west of the Williamson River canyon (Figure 17). Here, they are also eroded into pyroclastic-flow deposits and are up to 5 m deep and 10 m wide.

Figure 17: Scour channels and rafted pumice on the plateau west of the Williamson River canyon and location of the pyroclastic-flow blockage in the Soloman Butte quadrangle. Coordinates are UTM Zone 10, 6 m contour interval.

The thickness of the pyroclastic flows in this area is greater than 15 m (this minimum thickness was determined by hand augering within a 10 m deep scour channel in the pyroclastic flow to the 5 m limit of the auger). Rounded, large clasts of pumice are also found in these channels and on the pyroclastic flows that the channels are incised into. Channels in both locations are straight with no meanders. None of the channels shows evidence of recent stream occupation and all are forested with stands of Lodge Pole Pine.

The eastern slope at the mouth of the Williamson River canyon is devoid of pyroclastic-fall deposits to an elevation of 15 m above the current channel. The pyroclastic-fall deposits in this area normally average 1.5 m in thickness. Well-rounded river cobbles have been deposited on this surface to the upper level of scouring.

The Williamson River canyon opens rapidly to a broad plain comprised of reworked pyroclastic deposits, gravels, and rounded boulders of three distinct lithologies (Figure 18). The lithologies found within the Williamson River canyon are thick palagonitic deposits produced from hydrovolcanic eruptions, distinctly wavy-textured basalt flows, and massive, diktytaxitic basaltic andesite flows (Lee and Cummings, in preparation). The palagonitic deposits crop out at the mouth of the canyon and in several locations south of the canyon mouth (Lee and Cummings, in preparation). The wavy-textured basalt flows crop out within the canyon approximately 6 km from the boulder deposit (Figure 18) and are also found along the canyon rim to the south.

Figure 18: Soloman Butte quadrangle with location of boulder deposit, wavy outcrop, and location of pyroclastic-flow blockage. Coordinates are UTM Zone 10, 6 m contour interval.

Figure 19: Rounded boulder of the wavy textured basalt at the mouth of the Williamson River canyon. Boulder is 6.7 m in its longest dimension.

The columnar jointed, diktytaxitic basaltic andesite crops out along the edge of the Chemult Plateau to the west.

The deposit contains boulders up to 6.7 m in the longest dimension (Figure 19) and grain size decreases away from the mouth of the canyon (Figure 20). The boulder deposit is bordered to the east by the current Williamson River channel and to the west by the Chemult Plateau (Figure 20). To the south, the boulder deposit grades to reworked pyroclastic-fall and pyroclastic-flow deposits. These deposits contain rounded clasts of pumice up to 7 cm in diameter with lenses of lithic rich sand (Appendix E).

Figure 20: Locations of rafted pumice (*), auger holes (+), gravel pit, and boulder deposit (stippled pattern) with average intermediate axis diameter of erratics. Coordinates are UTM Zone 10, 6 m contour interval.

Portions of the deposit have been removed by quarrying, but the largest boulders were left behind. This has allowed for accurate measuring of the boulder diameters as well as the deposit thickness in these areas (Table 8).

The late Holocene terrace is cut into an organic-rich mixture of pumice and ash deposits at approximately 4.5 m above the current marsh. This terrace is only found in three of the locations surveyed (Figure 12) and has a tread that varies from 24 to 40 m in width.

The most recent fluctuations in the extent of Klamath Marsh are likely a combination of the redistribution of water for agricultural purposes, possible anthropogenic modification of a basalt flow, and climate change. Since the U.S. Bureau of Reclamation initiated the Klamath Reclamation Project in 1905 over 80% of the Klamath Basin’s 1400 km² of wetlands have been converted to farm and cattle land . Agricultural needs have lead to the diversion of the Williamson River before it reaches the marsh, which has likely reduced inflow affecting the surface elevation. The level of the marsh is controlled by input to the marsh and the base level is controlled by the Kirk Sill, a lava flow at the southern end of the marsh that water must first overtop before entering the Williamson River canyon. Anecdotal evidence reports that the Kirk Sill was blasted with dynamite at the turn of the century to lower the level of Klamath Marsh.

Early descriptions of Klamath Marsh by trappers and explorers provide evidence of the most recent fluctuations. Peter Skene Ogden was the first to describe Klamath Marsh when he visited it in November, 1826 (Davies, 1961; Dicken, 1980). His party visited a native village in the middle of Klamath Marsh that was surrounded by water too deep to be approached by foot or horseback. J.C. Frémont visited this same village in the summer of 1843 and found no difficulty in riding to the village .

Table 8: Location (UTM Zone 10), size, lithology, and velocity required for transport of measured boulders at the mouth of the Williamson River canyon.

The fluctuations described by early visitors to the marsh were likely seasonally or climatically induced and would have formed the late Holocene terrace. The lowering of the Kirk Sill coupled with the diversion of the Williamson River have likely reduced the level of the marsh from the late Holocene terrace surface to its current elevation.

The formation of the terraces is separated into three stages (Figure 21):

Stage 1: During the Pleistocene, Pleistocene lake Chemult occupied the area now known as Klamath Marsh. Seasonal and more long-term climatic changes during the Pleistocene induced fluctuations in the extent of Lake Chemult. These fluctuations planed the Pleistocene terrace into bedrock and colluvium.

Stage 2a: Middle Holocene eruptions from Mount Mazama deposit an average of 3.5 m of pyroclastic fall on the Wocus Bay quadrangle.

Stage 2b: Pyroclastic flows from the eruption form a blockage in the Williamson River canyon. Backflooding inundates the marsh, and forms a lake with a shoreline at 1397 m rafting the ash and pumice from the recently deposited pyroclastic fall and flow.

Stage 2c: Catastrophic breaching of the blockage results in the rapid draining of the marsh, stranding water rafted clasts of pumice on the middle terrace.

Stage 3: Climatic fluctuations plane the late Holocene terrace and agricultural modifications lower the marsh to its present elevation of 1376 m.

Figure 21: Three-stage development of terraces found along the eastern edge of southern Klamath Marsh. Figure is not to scale.

To backflood the Klamath Marsh to the 1397 m elevation of the middle Holocene terrace a blockage of the basin’s only outlet is required. The logical place for this blockage is in the canyon west of Soloman Butte (Figure 22). The river is confined to a slightly sinuous bedrock canyon that is only 75 to 100 m wide. Following the eruptions of Mount Mazama the basin was inundated with pyroclastic fall and pyroclastic flows. The pyroclastic flows were confined to drainages as they traveled off the eastern flanks of Mount Mazama en route to the Klamath Marsh where they were directed by the topography towards the Williamson River canyon. Within the canyon the flows likely became constricted and formed a blockage. The blockage is believed to have occurred at the point where the canyon begins to narrow and meanders slightly (Figure 17 and 18). The size of a blockage in this location was estimated from the topographic map to be 38 m high. Water entering the basin then backed up behind this constriction to the elevation of the middle Holocene terrace slope break at approximately 1397 m. The upper elevation of the plateau west of the Williamson River canyon coincides with that of the middle Holocene terrace slope break. Thus, the elevation of the plateau was a limiting factor for the extent of the backflooding caused by the pyroclastic flow constriction. This also suggests that the water was overtopping both the constriction and the plateau. Rafted pumice found across the plateau support this interpretation (Figure 17).

Figure 22: Extent of backflooding that resulted from the pyroclastic-flow blockage in the Williamson River canyon.

The rounded clasts of pumice found in the Wocus Bay area were water transported to their current location when the damming of the Williamson River backflooded the Klamath Marsh. The clasts are found at the 1397 m upper limit of the floodwaters in only a few locations. The deposits are also found in constrictions such as Skellock Draw in the Wildhorse Ridge quadrangle, and the constriction between Yoss Creek Meadow and Wocus Bay, where Williams (1942) first described the rounded pumice. In these locations, the pumice-laden floodwaters would have backed up in these hydraulic constrictions as the area was being drained. This resulted in the high concentration of rounded pumice clasts in these areas. The clasts are also commonly found on the broad Pleistocene terrace in the Wocus Bay quadrangle. Deposition on this surface is also thought to have occurred as the water receded stranding the larger clasts on the broad terrace as the impounded water rapidly receded. The linear deposits that are found to the west of Little Wocus Butte are interpreted as shorelines of the temporary lake.

Formation of the scour channels is thought to have occurred when the backflooding rapidly receded. The water saturated ash would have been easily eroded as the level of backflooding dropped. The channels in the Soloman Butte quadrangle are larger than those in the Wocus Bay area. The channel size is likely the result of the steeper gradient of the surface in the Soloman Butte area. Incision of the channels occurred as the impounded water was evacuated down the canyon after failure of the blockage. The presence of these channels in both areas and their linear morphology suggests that the draining of the marsh was rapid (Figures 15 and 16).

The presence of rounded erratic boulders, scoured surfaces, deposition of river cobbles above the current channel, and the scour channels, all in the Soloman Butte quadrangle indicate a catastrophic breaching of the pyroclastic flow constriction within the Williamson River canyon. The coinciding elevation of the middle Holocene terrace and the plateau as well as the presence of rafted pumice on the plateau south of the constriction indicate that the impounded water overtopped the pyroclastic flow constriction. Overtopping leads to rapid dam failure by the headward erosion of channels in the downstream side of the dam.

Boulders that are found on the outwash plain are a conglomeration of material that was in the channel prior to the flood and material that was plucked from the outcrops lining the canyon. The wavy textured basalt boulders crop out in the canyon at the location of the suspected blockage. This lithology is also found south of the blockage along the canyon rim, but these outcrops would have been above the level of scouring in the canyon. Talus from these outcrops likely contributed to the boulders that were entrained by the downstream flooding. The boulders of hydrovolcanic deposits found in the outwash plain were transported less than 1 km, but their presence is significant. This lithology is quite susceptible to physical weathering and would not have withstood typical stream transport to its present position. It can then be inferred that these boulders were plucked from the hydrovolcanic outcrop near the mouth of the canyon and deposited in their present position in a single event.

The velocity of the flood and size of the material transported rapidly decreased as the flood left the canyon. Beyond the delineated boulder deposit are deposits of rounded pumice, sand, and ash. Hand auger SB4 (Figure 20, Appendix E) encountered 74 cm of reworked pyroclastic deposits that were underlain by cobbles. These cobbles could be river deposits or flood deposits, but since undisturbed pyroclastic fall was not encountered in this section, the underlying cobbles are likely part of the flood deposits. This stratigraphy indicates that the flood velocities were decreasing and deposition of finer grained material was taking place. The lenses of sand intermixed with rounded clasts of pumice in auger SB5 suggest either pulses in the waning flood discharge or primitive channel migration.

The current Williamson River channel flows along the eastern edge of the boulder deposit and then crosses the deposit about 1 km from the mouth of the canyon (Figure 20). The concentration of the boulder deposit to the west of the current channel is the product of hydraulic routing of the outburst flood. Two factors likely influenced this routing. The paleochannel of the Williamson River could have occupied a channel west of its current position. The outburst flood upon exiting the canyon would have been directed by the channel to the west. Another possibility is that as the floodwaters left the canyon, they would have been deflected to the west by a westward bend in the canyon at its mouth. This is the more likely scenario given that the water was scouring surfaces approximately 15 m above the current channel at the mouth of the canyon. The paleochannel of the river likely had little influence on a flood of this magnitude.

The paleohydrologic reconstruction of the outburst flood is based on upstream and downstream geomorphic evidence and hydraulic principles. Estimates of the peak-discharge of the outburst flood were determined with three methods:

1. Flow-competence
2. Empirically based dam-break model
3. Physically based dam-break model

Flow competence equations are based upon empirical correlations and theoretical formulas relating the largest clasts transported by a flood to velocity, shear stress, and stream power ; . For this study the regression relation for velocity developed by O’Connor (1993) in a study of the Bonneville flood is used. O’Connor (1993) used geologic evidence coupled with the step-backwater method to develop regressions for velocity, stream power, and shear stress. This regression is particularly useful for this study where the deposit has been quarried because it is based upon the largest clasts transported rather than a median grain size. The following equation (O’Connor, 1993) was used to estimate velocity:

v = 0.29D_i^0.60

where D_i is the intermediate axis diameter (cm), and v is the flow velocity (m s^-1). The results of the flow-competence calculations and their averages at each site are given in Table 8. The discharge (Q) was then estimated by the relation Q=wdv, where w is the flow width and d is the flow depth. The flow width was estimated from the width of the boulder deposit and the depth was estimated from the elevation of the scour surface at the mouth of the canyon and the depth of the boulder deposit elsewhere (Table 8). The maximum discharge estimate with this method is 1.5 ´ 10⁴ m³s^-1. An average discharge of 1.3 ´ 10⁴ m³s^-1 was calculated for two sites (Table 8).

Costa and Schuster (1988) used a regression analysis with potential energy as the independent variable to derive equations for landslide, glacier, moraine, and earth- and rock-fill dam-break discharges. The data used to construct the equations are gathered from historic dam-failures where good estimates of potential energy of the back-flooded area and resultant discharges were available. For this study, the equation for a landslide dam-break is used. This equation was chosen because landslide dams are typically wide at their base, comprised of large volumes of material, and are usually easily eroded . These characteristics make them most analogous to a pyroclastic flow dam in the Williamson River canyon. The empirical relation for peak discharge (Q_p) from the failure of a landslide dam (Costa and Schuster, 1988) is:

where the potential energy of the lake (E_p) =r_wgdV₀. For the lake formed by the pyroclastic flow constriction, d = 17 m (drop in water from elevation of middle Holocene terrace at 1397 m to the elevation of the late Holocene terrace at1380 m), and V₀ = 5.7 ´ 10⁹ m³ (volume drained from the lake). Using r_w = 10³ kg m^-3 (density of water) and g = 9.8 m s^-2 (acceleration due to gravity), this equation yields a peak discharge of 2.2 ´ 10⁴ m³s^-1.

Walder and O’Connor (1997) evaluated regression relations that are used for predicting peak discharge and determined that factors not used in these relations actually provided the greatest influence on the discharge. These factors are the erosion rate of the dam (k), and the morphology of the dam and basin.

Walder and O’Connor (1997) outline "benchmark" predictive relations that are based on mean values from typical natural dams and can be used for hazard-assessment purposes. Fundamental to these relations is the dimensionless parameter h, that is defined as

where k is the erosion rate of the dam (m s^-1) , V₀ is the volume of water drained by the breach (m³) , g is the acceleration due to gravity (m s^-2) , and d is the drop in water level during down stream flooding (m). If this parameter is significantly greater than 1, then the following equation is justified for determining peak discharge

where Q_p is peak discharge, and D_c is height of dam crest relative to dam base (Table 9). The value 1.94 is a multiplier derived from the morphology of the breach. For the pyroclastic flow dam on the Williamson River the erosion rate (k) is unknown. Data compiled by Froehlich (1987) show that k for constructed earthen dams are between 1 and 10³ m hr^-1. The compaction of a pyroclastic flow dam is likely less than that of a constructed earthen dam, but values from 1 to 10⁴ m hr^-1 were used to determine h and the results are all significantly greater than 1 (Table 9). The erosion rate is actually not significant for lakes that have large volumes for a given depth because erosion of the dam will be complete before any significant decrease in volume has occurred. If the dimensionless parameter V*₀ (V*₀=V₀ / d³) is greater than 10⁴, erosion rate has little affect on peak discharge . For the flooded area behind this blockage that was drained V*₀ equals 1.2 ´ 10⁶.

The peak discharge that was determined with the physically based dam-break model of Walder and O’Connor (1997) is 1.3 ´ 10⁴ m³s^-1. A summary of estimated discharges and their associated variables is found in Table 9.

Table 9: Summary of discharge estimates and variables.

The highest discharges determined using the flow-competence equation and the channel dimensions (Table 8) were for boulders in the gravel pit located in the middle of the deposit. The intermediate axis of these boulders was smaller than the measured intermediate axis of the boulders at the mouth of the canyon that were not excavated. This discrepancy in discharge is the result of overestimation of the channel width and/or depth at the gravel pit. The channel width was determined from the width of the boulder deposit and assumes that the deposit was emplaced in one event. The deposit is likely more of a fan that had channels migrating across its width as the flood stage decreased. The estimation of channel width and depth at the mouth of the canyon is thought to be more accurate because it is constrained by the river to the east and the plateau to the west. The maximum estimated discharge for this location is 1.07 x 10⁴ m³s^-1. However, the velocity estimates for these boulders is likely conservative because they were not excavated and their complete dimensions could be concealed by overburden.

The flow-competence equation is also likely a conservative estimate of the peak discharge because it is based on boulders found ~6 km from the suspected dam location. The discharge likely decreased over this distance, but the decrease would have been small because the flow was confined to the canyon across this distance. Material that is currently found in the canyon was not used in the flow competence equation because it is difficult to distinguish the material associated with the dam-break from colluvium. Another factor that leads to the underestimation of discharge with this method is the size of the material that was available for transport. Columnar jointed basalt would produce boulders with a size limitation dependent on the jointing pattern. The wavy textured basalt is a massive unit and the material derived from its outcrops would not have these limitations.

The empirically based dam-break model is limited in its estimate of the discharge because it only uses the volume of water impounded by the blockage and the drop in the level of this water. This method would provide the same discharge estimation for a dam that was several kilometers in width as it would for a dam that was only a few meters in width, provided that the volume and drop in water level of the flooded areas was equal. This equation must also be used with caution because it is derived from data supplied by a variety of sources at different locations. This method is based on discharges gathered downstream of the breach with no consideration of energy losses or translation of the flood wave and is therefore an underestimation of discharge in most situations . The most significant limit to this method is that most discharges are controlled by factors other than the volume and drop in lake level (Walder and O’Connor, 1997). The factors that are usually the primary influences on peak discharge are erosion rate of the impoundment and to a lesser extent the geometry of the breach and basin (Walder and O’Connor, 1997).

The physically based model is considered to be the most accurate estimate of discharge at the breach. Uncertainties in this method lie with the calculation of shape multiplier. This factor is based on average shape parameters (channel size and morphology) from historic dam-breaks and these parameters are not exactly known for the constriction on the Williamson River. Variances in these parameters would not however greatly affect the estimated peak discharge.

Rough estimates for the duration of the backflooding from the pyroclastic flow dam can be made indirectly from comparison to landslide dams, geologic evidence, and hydrologic characteristics of the basin.

The three factors that control the stability of natural dams are the rate of inflow to the impoundment, the size and shape of the dam, and the geotechnical characteristics of the dam (Costa and Schuster, 1988). Rate of inflow to the impoundment is difficult to estimate, but the existence of landslide dams is short when a small slide blocks a stream with a large drainage area . Since the dam and the plateau were overtopped by floodwaters, it can be concluded that the rate of inflow was greater than the rates of seepage through the impoundment and evaporation. The relatively uniform grain size of this impoundment would have prevented extensive piping through the dam. The dam was likely thicker than typical constructed earthen dams, because the pyroclastic flows were utilizing the channel as well as flowing into it from the west creating a long constriction in the canyon. This thickness would have contributed to the stability of the impoundment. Landslide dams that are composed of large or cohesive materials are more stable than those composed of soils or loose rock (Costa and Schuster, 1988). Ash, although fine grained, is quite cohesive when moist. The factors controlling the failure of the dam after its overtopping would have been the erosion rate of the ash on the downstream side and the length of the impoundment within the canyon. Failure likely occurred shortly after the overtopping of the dam, so the longevity of the pyroclastic flow impoundment was primarily controlled by the rate of inflow to the impoundment.

To fill the basin with water to an elevation of 1397 m from the late Holocene elevation of Klamath Marsh requires the addition of 5.7 ´ 10⁹ m³ of water. To achieve this with the highest recorded discharge from the gauging station along the Williamson River at the Kirk Sill would require 4.75 years. This is a likely overestimate for several reasons. At the time of the cataclysmic eruption, glaciers perhaps 60 m thick occupied the Sun and Kerr Valleys . The eruption would have caused extensive melting of these glaciers and the resultant meltwater from Kerr Valley would have contributed to the flooding of Klamath Marsh. Water that was present in the marsh at the time of the eruption would have been displaced by the sinking of pyroclastic fall and the emplacement of pyroclastic flows. The depth of these deposits in Klamath Marsh is over 20 m in some locations.

The rafted clasts of pumice found in Wocus Bay can be loosely used to infer the longevity of the lake. A simple experiment consisting of a bucket of water and rafted clasts of pumice retrieved from Wocus Bay revealed that these clasts would currently float no longer than 5 days. These samples weathered since they were erupted and certainly were able to float for a longer period prior to this weathering. Rafted pumice for the 1952 eruption of Volcan Barcena on Isla San Benedicto, Mexico produced pumice rafts that were sighted floating in the Pacific Ocean for a minimum of three years following the eruption . Eruptions from the Rabaul Caldera, Papua New Guinea in 1878 and 1937 produced pumice rafts several meters thick that hampered shipping for several months after the eruptions .

An estimate of the time to drain the lake can be made by assuming that a hydrograph of the discharge is triangular and dividing twice the volume by the peak discharge. Using this method, it would have taken 2.5 days to drain the flooded area impounded by the dam.

The extensive Pleistocene terrace that has been planed into bedrock and colluvium is the product of climate fluctuations during the Pleistocene. Pleistocene lake Chemult, which covered an area of 390 km², likely occupied this basin between about 32 ka and 11 ka. The increased inflow to the basin lead to the development of the Williamson River canyon. Desiccation of this lake exposed fine grained lake sediments that were then easily eroded and transported by winds forming the parabolic dunes found in the northeast corner of the basin

The cataclysmic eruption of Mount Mazama (7627± 150 cal yr B.P.; and resultant pyroclastic flows were responsible for backflooding that rose 22 m above the current level of Klamath Marsh to an elevation of approximately 1400 m. This large-scale backflooding was the result of a pyroclastic flow constriction along the Williamson River in the Soloman Butte quadrangle. Failure of the pyroclastic flow constriction was catastrophic and resulted from overtopping by the impounded water. Discharges generated by the dam failure are estimated to have been 1.3 x 10⁴ m³s^-1. The resultant downstream flooding transported boulders up to 6.7 m in length approximately 6 km. The boulder deposit rapidly grades to clasts 1 m in diameter in 1 km. The flood deposits have confined the current Williamson River channel to the east at the mouth of the canyon. The upstream affects of the dam include terrace formation, channels scoured into pyroclastic-flow deposits, reworking of pyroclastic-fall and pyroclastic-flow deposits, and the inundation of 5.6 x 10⁸ m².

The most recent fluctuation in the extent of Klamath Marsh is attributed to the agricultural redistribution of water resources in the Williamson River basin. The diversion of inflow to the basin and the purported lowering of the Kirk Sill are the likely mechanisms behind the lowering of Klamath Marsh to its current level.

[Next Section] [Previous Section] [Table of Contents]