Data collection at gaging stations near Mount St. Helens often was motivated by rainstorms and impending storm flows. Efforts to measure sediment discharge during storm flow can be justified by reference to earlier studies and statistical analyses of daily sediment discharges. Leopold and others (1964, p. 72) studied sediment-discharge records for streams throughout the United States and found that a large part of the annual sediment discharge occurred during moderate floods with frequent recurrence. An analysis of daily sediment discharges from streams near Mount St. Helens shows that emphasis on data collection during storm flows was appropriate.
Examples of percentage distribution of sediment discharge with time are given for available water years in table 3 for six gaging stations. Five gaging stations had 9 years of daily sediment discharge record, and the Clearwater Creek gaging station had 7 years of record. For each station, the daily sediment discharges were ordered by magnitude. Of the cumulative sediment discharge at those stations, more than 60 percent of the sediment was transported on 5 percent of the days, over the long term. From 33 to 37 percent of the sediment was transported on the highest 1 percent of days in the North Fork and the mainstem Toutle Rivers. In contrast, 58 to 60 percent of the sediment was transported on the highest 1 percent of the days from the airfall-affected streams, the Green River and Clearwater Creek. Because a large percentage of sediment discharge occurs in a small percentage of time, sediment transport during storm flows was measured whenever possible.
Sediment discharge is measured during storm flow using standard components of flow measurements; at multiple points, one measures the flow depth, the stream velocity, and the sediment concentration. At moderate discharges, these components can be measured directly with fair-to-excellent accuracy. During storm flows when sediment discharges are most extreme and measurements are most critical, basic data become difficult to obtain, and accuracy may be poor. After the 1980 eruption devastated the forested lands, high stream velocities and hazards from floating woody debris hampered measurements of stream discharge so that accuracy was often poor.
Forests to the north around Mount St. Helens were largely destroyed by the eruption blast (fig. 10). As the major lahars coursed through the river valleys, they distributed the uprooted trees into the path of future storm flows, along flood plains and in stream channels. Cut logs, uprooted trees with large root balls and branches, tree stumps, broken branches, and fine roots were all transported during storm flows (fig. 11) . In most streams of the forested Cascades Range, storm flows transport woody debris that usually decreases in quantity after peak river stage. The abundant woody debris in the Toutle and the upper Lewis River streams, however, endangered submerged equipment long after the peak stage occurred.
Equipment that was ordinarily robust enough for flood measurements was relatively unstable in the high flow velocities (greater than 10 ft/s) of the steep channels (16 to 24 ft/mi). The brass sediment samplers and the lead weights used for discharge measurements were dragged tens of feet downstream after immersion and would spin and swing wildly on their suspension cables after removal from the flow. Larger samplers and sounding weights increased somewhat the stability of suspended equipment. Still, nozzles on sediment samplers were often bent or broken by collisions with debris. Metal cups of the Price AA velocity meter might be distorted or crushed repeatedly during a discharge measurement. To avoid damage to cable-suspended equipment, conventional measurements of discharge and sediment concentration were postponed for short periods at high flow. Methods to reduce damage to equipment were adopted, which included:
These are accepted flood-measurement techniques (Buchanan and Somers, 1969), although the extent of their application in streams near Mount St. Helens was unusual. Conventional measurement techniques were resumed after stream conditions became more favorable during storm-flow recession. Also, the amount of debris transported during storm flows decreased to nuisance levels a few years after the 1980 eruption. Reduced hazards to equipment eventually allowed standard discharge measurements to be obtained at higher flows.
Rainstorms on tributary drainage basins of the Toutle and the upper Lewis Rivers produced storm runoff rapidly, and the rainfall was generally widespread. To gain synoptic information about sediment discharge from affected drainage basins, repetitive measurements were made at several gaging stations simultaneously during storm flows over 2 to 3 days. Fully outfitted vehicles, two-way radios, satellite telemetry, portable generators, and high quality rain gear were essential to obtain acceptable measurements of sediment discharge during storm flow. Satellite telemetry that relayed gage height to a central location helped direct attention to sudden changes in the hydrologic situation.
Most measurements of storm flow were accompanied by constant rain, and night operations were common during storm flows lasting several days. High intensity lights powered by gasoline generators were arranged at bridges and cableways where measurements were made. Hazards to equipment were increased by low visibility of floating debris. Suspension cables occasionally were snapped by unseen debris, and expensive samplers and sounding weights were lost. Although night operations are not routine in stream monitoring, they were essential to maintaining time resolution of unpredictable variations in sediment discharge.
Examples of Transport Variability During Storm Flow
Detailed measurements of stream velocity, cross-sectional area, and sediment concentration revealed rapid changes in those components during storm flow at gaging stations. Data from those measurements illustrate the variability of flow hydraulics and sediment concentration that would affect sediment-discharge computations.
Sandy streambeds and high sediment concentrations caused the post-eruption stream flow to be hydraulically smooth and to build flood waves rapidly, as described by Dunne and Leopold (1981, p. 10). During water year 1981, times of travel for flood waves between the North Fork Toutle River at Kid Valley and the Toutle River at Highway 99 ranged from 2 to 2.3 hours, at stream discharges greater than 10,000 ft3/s. Mean velocity of flood waves through the 23.1-mi reach ranged from 14.5 to 16.9 ft/s (Dinehart, 1982). Times of travel between the North Fork Toutle River at Kid Valley and the Toutle River at Tower Road were similar for flood waves originating from rapid spillage of small ponds on the debris-avalanche deposit. The velocity of flood waves in 1987 and 1990 was measured at 13 ft/s between Kid Valley and Tower Road.
Flood hydrographs recorded on the sediment-affected streams were steeper than those recorded prior to the eruption. Pre-eruption hydrographs of selected flood peaks at the gaging station Toutle River near Silver Lake showed rates of stage rise that ranged from 0.3 to 1.1 ft/hr when measured along the steepest segment of the hydrograph. A rate of stage rise of 2.8 ft/hr was measured at the same station in a post-eruption hydrograph of storm flow. At the Toutle River at Highway 99, a stage rise of 6 ft/hr was measured during storm flow on February 19, 1981.
If flood waves and sediment concentration peaks coincide, the instantaneous rates of sediment discharge can be increased greatly. For example, the peak sediment-discharge rate was about 19 million tons per day for more than an hour on February 19, 1981 at the Toutle River at Highway 99. Peak discharge and peak sediment concentration differed in time by 2 hours; if they had coincided, peak sediment discharge could have been as high as 29 million tons per day. Flood-wave and sediment-concentration peaks were closer in time nearer the debris-avalanche deposit, which would have contributed to higher sediment-discharge rates at the upstream stations.
A representative list of the storm flows monitored during the study period 1980-90 is given in table 6 . Stream- and sediment-discharge records of the North Fork Toutle River at Kid Valley were used for illustration. In the table, mean stream discharge and sediment discharge are each totaled for 4 days of storm flow, which includes the duration of typical storm flows in this region. To maintain consistency among storm flows, the day of peak flow was usually chosen as the second of the four days. This list can be used to locate days of high sediment discharge in published data (U.S. Geological Survey, 1980-90; Dinehart and others, 1981; Dinehart, 1986, 1992b).
Peaks and Lags of Sediment Concentration
Storm flows were measured and sampled repetitively in November and December 1980 to evaluate the first widespread erosion of eruption deposits. At the gaging station at the North Fork Toutle River at Kid Valley, stream discharge and sediment concentration peaked within about 0.5 hour of each other. Several miles downstream at the Toutle River at Highway 99, peak sediment concentration lagged peak stream discharge by nearly 2 hours on November 7 and 8, 1980 ( fig. 20 ).
Lagging of sediment peaks has been noted in other river systems where the primary sediment source is many miles upstream from the sampling station (Heidel, 1956). Flood-wave celerity is greater than flow velocity; sediment that is transported near stream velocity can arrive at a gaging station long after the flood-wave peak has passed. As sediment from the debris-avalanche deposit in the upper North Fork Toutle River valley was transported in the Toutle River, the difference between flow and flood-wave velocity increased the time lag between flood-wave peaks and the arrival of suspended sediment downstream.
In complex hydrographs that include runoff from a sequence of rainstorms, the lag of suspended sediment is less identifiable. Many storm flows produced broad, fluctuating peaks of sediment concentration not closely associated with peak discharge. The relation between sediment concentration and peak discharge was examined by dividing the sediment concentration into sand and fine concentrations. Although the sediment in most samples from the November and December 1980 storm flows were not divided into sand and fine concentrations, repetitive samples from subsequent storm flows were routinely divided.
Some features of stream discharge can be inferred from sand and fine concentration curves that were plotted for eight separate storm flows (figs. 21-28). At gaging stations distant from the dominant sediment source, the peak value of fine concentration lagged peak discharge by 1 or 2 hours. The trace of fine concentration resembled the shape of stream-discharge hydrographs. Changes in fine concentration between successive samples were gradual.
Sand concentration varied more erratically. Because local suspension of sand increases with stream velocity, sand concentration will increase with rising stream discharge. Sand concentration also increases as sand is transported from distant, upstream sources. Successive sand concentrations, separated by only minutes, often differed by more than a factor of two. The maximum sand concentrations can be attributed to sampling at the high transport regions of bedforms, to "boils" of sand suspended by turbulence, or to accidental contact of the sampler nozzle with the sandy streambed. Sand concentration often increased several hours after fine concentration had receded. Erratic variations in sand concentration were typical of flow recession and may have indicated bedform migration during sampling.
The storm flow of February 19, 1981 was sampled as often as every 5 to 15 minutes ( fig. 21 ). Fine concentration reached a maximum value of 219,000 mg/L about 1.1 hours after peak stage. Sand concentration was 158,000 mg/L about 1.2 hours after peak stage. Sand concentrations of two later samples (248,000 and 166,000 mg/L) indicated a subsequent increase and decline about 2.2 hours after peak stage. During recession of fine concentrations, each value was lower than the preceding one, whereas sand concentrations of the same samples varied erratically with time.
The time lag between peak discharge and peak fine concentration changed in later storm flows. Smaller storm flows were sampled repetitively at the Toutle River at Highway 99 on October 6-7 and December 2, 1981 (figs. 22 and 23 ). Samples spaced as closely as 5 minutes apart revealed a gradual rise to peak fine concentration that lagged peak stage by less than 1 hour. The sequence of sand concentrations showed increases near peak stage, and sand concentration did not recede as smoothly as fine concentration.
The storm flow of December 5-6, 1981, was sampled simultaneously at the gaging stations at the Toutle River at Highway 99 and the Cowlitz River at Castle Rock, which are separated by 3.7 river miles ( fig. 24A, B ). The shape of the fine-concentration curve was similar between the two stations, although the sand-concentration curves showed little resemblance to each other. The lag of fine concentration that was observed during the December 2, 1981, storm flow at the Toutle River at Highway 99 was not apparent in the curves of fine concentration for December 5-6, 1981. Instead, fine concentration reached a maximum nearly 2 hours before peak stage at both the Toutle River at Highway 99 and the Cowlitz River at Castle Rock.
When the influence of a distant sediment source decreases, local sediment sources can produce a sediment concentration peak nearer to the peak discharge in time. In a storm flow derived from snowmelt in the lower elevations of the Toutle River basin, peak fine concentration coincided with peak discharge on January 23, 1982, at the Toutle River at Highway 99 ( fig. 25 ). In the following two water years, peak fine concentration preceded peak discharge by several hours on December 3, 1982 ( fig. 26 ), and on November 3, 1983 ( fig. 27 ), at the Toutle River at Tower Road. The decrease in sediment lag can be interpreted as a result of decreasing dominance of sediment discharge from the debris-avalanche deposit. Additional factors that influence sediment lag have been described by Williams (1989).
A storm flow on February 20, 1982, breached the embankment impounding Jackson Lake, which had been formed at Jackson Creek along the southern margin of the debris-avalanche deposit. Outflow from the lake created a flood wave that was sampled at the Toutle River at Highway 99 ( fig. 28 ). Later that day, the north embankment on the temporary sediment-retention dam N1 was breached when the North Fork Toutle River overflowed its existing channel ( fig. 29 ). Sediment from the retention dam was eroded and transported through the breach. At the North Fork Toutle River at Kid Valley and the Toutle River at Tower Road, a small peak of fine concentration lagged the associated increase of discharge by more than 1 hour. As sediment from the N1 breach was transported downstream, the associated sediment peak was diffused and diluted, as shown in samples collected at the gaging stations at the North Fork Toutle River at Kid Valley, the Toutle River at Tower Road, the Toutle River at Highway 99, and the Cowlitz River at Castle Rock ( fig. 30 ). The small sediment wave was superimposed on the flood recession.
An eruption at Mount St. Helens on May 14, 1984, generated a lahar that entered the North Fork Toutle River (Pringle and Cameron, 1986). Sediment samples of the diluted flow provided another example of lag effects from a distant sediment source ( fig. 31 ). Repetitive samples, collected at the Toutle River at Tower Road for 4 hours after peak discharge, traced the fine concentration only to the beginning of concentration recession. The curve of fine concentration was smooth, reaching peak concentration 3.5 hours after peak discharge. Sand concentrations near peak discharge were not significantly different from concentrations prior to arrival of the diluted phase of the lahar, with the exception of one sand concentration at peak discharge that was the highest of the sequence on May 14, 1984. Sand concentrations again increased near the end of the sampling episode.
Sand and fine concentration curves are available for storm flows only through water year 1984, before the gradual replacement of box samples by automatic pumping samples. After water year 1984, repetitive samples were collected at a single vertical only when automatic sampling was not available or was unreliable. As noted earlier, the proportion of fine and sand concentrations in a pumped sample does not represent the flow, and the sand concentration may be under- represented.
The preceding examples of sediment lag confirm that, in the Toutle and the Cowlitz Rivers, sediment concentration during storm flows was not a simple function of stream discharge. Previous studies have noted that sediment-transport curves derived from instantaneous samples in lagging flows will have significant errors (Guy, 1964; Marcus, 1989). Direct computation of sediment discharge from coincident time plots of sediment concentration and stream discharge is the most accurate method available for these lagging flows. The extensive collection of sediment samples provided more reliable daily sediment discharges than could be obtained from sediment-transport relations.
