Results


This study had three major objectives: (1) to evaluate the use of turbidity and (or) streamflow as a surrogate for quantifying SSL in Johnson Creek, (2) to compute the mean annual suspended-sediment budget for the watershed, and 
(3) to investigate the timing and spatial distribution of SSL in the watershed. Objective 1 was discussed in section, “Regression Model Evaluation.” The monthly and annual SSL budgets and the annual streamflow were evaluated to accomplish objectives 2 and 3. 


Analysis of the suspended-sediment budget of Johnson Creek was divided into two major components—streamflow and suspended-sediment loads (SSLs). Because the annual SSL budget for a given water year is heavily influenced by the quantity and flashiness of streamflow for that water year, the streamflow of the 4 water years covered in this study (2007– 10) was investigated to determine if the SSL computed should be considered extreme or otherwise unusual.


Because SSL is a product of SSC and streamflow 
(eqs. 3 and 4), if SSC is fixed in time, SSL is directly proportional to streamflow. Evaluating the amount of streamflow for a given water year should be a good indicator of how much SSL to expect. However, SSC also is closely correlated with streamflow. As a result, during a storm event, the rate of increase in SSL is greater than the rate of increase in streamflow. This relation results in especially large quantities of SSL being transported during yearly peaks and floods. The relation between increasing streamflow and SSL hereafter is termed “disproportionality.”


An example of this disproportionality is provided by the storm of late December 2007. The heavy rainfall resulted in the streamflow increasing by a factor of 11, SSC increasing by a factor of 20, and SSL increasing by a factor of 157. The disproportionality between increases in streamflow and SSL highlights the importance of analyzing the high streamflow periods for any given water year. If enough high peaks occur in a given water year, more SSL might be produced than another water year with greater average annual streamflow but fewer peaks. The extent of this disproportionality is analyzed in further detail in section, “Analysis of Suspended-Sediment Budget.”


Analysis of Streamflow


Streamflow is an important component in the production of SSL. Water years with especially high or low levels of streamflow are likely to produce SSLs that are not representative of average conditions. The streamflow records at the Gresham and Milwaukie stations cover only 12 and 21 years of record, respectively. Conversely, streamflow has been continuously recorded at the Sycamore gaging station (14211500) since water year 1941, providing 70 years of data to be compared with the streamflow for the period of study, water years 2007–10.


The Sycamore station is geographically located between the Gresham and Milwaukie stations. For reference, the Sycamore station has a drainage area of 26.8 mi², which is about 1.75 and 0.50 times the drainage areas of Gresham and Milwaukie, respectively. Similarly, the average annual streamflow at the Sycamore station is 53 ft³/s, which is about 1.75 and 0.70 times the average annual streamflow of Gresham and Milwaukie stations, respectively.


In addition to overall annual and monthly streamflow quantities, two specific peak streamflows are covered in more detail in the section “Selected Peak Streamflow Events.” The spatial distribution of streamflow also was investigated by calculating the percentage of streamflow at the Milwaukie station originating upstream of the Gresham station (see the section “Spatial Distribution of Streamflow”). This spatial distribution of streamflow will be compared to the spatial distribution of SSL to evaluate changes in sediment yield in the section “Analysis of Suspended-Sediment Budget.”


Water Years 2007–10


Streamflow for the 4-year study period was compared to the streamflow during the 70-year period of record to assess whether the study period is representative of long-term streamflow conditions. Three metrics were used to evaluate the amount of streamflow at the Sycamore station (table 4). All three metrics were selected to identify and to evaluate aspects of the streamflow that could result in an unusual amount of suspended sediment in Johnson Creek during the study period.


The first metric evaluated was mean annual streamflow, which is a measure of how much streamflow passed the station in a given water year. During the 4-year study period, the mean annual streamflow was 51.6 ft³/s, which is close to the average value for the 70-year period of 53.1 ft³/s. The ranks of each water year ranged from 25 (2007) to 49 (2009). Therefore, 24 of the 70 water years on record produced more streamflow than water year 2007, which has the highest mean annual streamflow of the 4-year study period. Conversely, 
21 of the 70 years on record produced less annual streamflow than water year 2009, which has the lowest annual streamflow of the 4 water years studied. In summary, the mean annual streamflow during the 4-year study period is representative of mean annual streamflow over the period of record.


The second metric studied was the annual peak streamflow. Based on equations 3 and 4, and the positive correlation between SSC and streamflow, high annual peak streamflow in the Johnson Creek basin would be expected to transport a disproportionately large amount of suspended sediment. Thus, a water year with constant streamflow should transport less suspended sediment than a water year with the same total streamflow, but that also included several large peaks.


Of the 4 water years in this study, three of the annual peaks were unexceptional. Water years 2007, 2008, and 2010 produced peaks that ranked between 22 and 55 out of the 70 years of record. However, the peak annual streamflow of water year 2009 was the third highest on record, lower only than the peak annual streamflows of 1965 and 1997.


The third metric studied was the kurtosis of the daily mean values of streamflow. Traditionally, kurtosis is considered a measure of the “peakedness” of a population (a distribution with a high kurtosis has a pronounced peak near the mean), although there is some debate as to whether it is more of a measure of heavy tails (abundance of extreme events) (Kaplansky, 1945; Ali, 1974; Johnson and others, 1980). Higher kurtosis values indicate that the variance of the distribution is the result of infrequent extreme deviations. Conversely, frequent, smaller deviations would result in lower kurtosis values. For this study, kurtosis can be considered a supplemental metric to the peak annual streamflow. High kurtosis values suggest a water year with more “flashiness,” or multiple high peaks, and low kurtosis values suggest smaller and (or) less frequent peaks. It is possible for a given water year to have several moderate peaks, none of which qualify as exceptional. In such an example, the total of many moderate peaks could produce a substantial amount of suspended sediment. However, if these peaks occurred in a water year with relatively little streamflow, and if the peak annual streamflow were near or less than average, the first two metrics used to evaluate streamflow would suggest that the water year might have produced relatively little sediment. In this example, a high kurtosis value resulting from numerous peaks would explain SSLs that were higher than would be expected based on the first two metrics alone.


The kurtosis values of water years 2007 and 2010 were low, with ranks of 49 and 48 out of 70, respectively. The kurtosis value for water year 2008 was moderately high, ranking 8 out of 70. This high kurtosis value likely is in part a result of the high streamflows of December 2007 (water year 2008), which were the largest monthly streamflows during the period of study. The water year 2009 kurtosis value was very high, ranking 2nd highest of 70, and lower only than water year 1994. This likely was a result of the large flood in early January, coupled with a series of minor peaks and a relatively low winter base flow.


Together, the three flow metrics evaluated indicate mostly moderate levels of flow during the period of study. The least amount of overall flow of the 4-water-year study was in water year 2009. However, water year 2009 also produced an exceptionally large peak annual streamflow and high kurtosis value.


Selected Peak Streamflow Events


Two peak streamflow events during the period of study merit additional discussion. The first is the high peak streamflow of January 2, 2009. On January 1, 2009, 2.53 in. of rain fell at the Holgate rain gage (fig. 1; U.S. Geological Survey, 2011d). This was the most precipitation that fell during a 24-hour period for the period of study. The following day, another 0.79 in. of rain fell. Combined with relatively high streamflow from precipitation falling during the previous 8 days (2.92 in. total), the result was the third-largest flood in the 70-year history of the Sycamore station. About 37 percent of the streamflow from water year 2009 occurred during the 20-day period from December 25, 2008 to January 13, 2009.


The other peak streamflow occurred in June 2010. Sustained rains through late May and early June resulted in a rare June peak instantaneous streamflow of 501 ft³/s on June 4. Mean daily streamflows were 260, 167, 360, and 175 ft³/s on June 2, 3, 4, and 5, respectively. The June 4, 2010, mean daily streamflow was the second highest of any June value in 70 years of record, second only to June 21, 1984. The June 2, 3, and 5 mean daily streamflows were the 4th-highest, 10th-highest, and 8th-highest streamflows, respectively. The highest streamflow of any June in history was June 2010 as a whole, at 4,640 acre-ft, surpassing the second-highest June streamflow of 3,780 acre-ft in 1984.


Spatial Distribution of Streamflow


Evaluation of the spatial and temporal distribution of streamflow originating upstream of the Gresham station provides a useful context for comparison with the SSL budget and drainage areas upstream of the Gresham and Milwaukie stations. In a uniform watershed with equal characteristics (such as topography, soils, and precipitation distribution), the proportion of streamflow and SSL originating upstream of Gresham would be identical to the proportion of the drainage area that the Gresham station represents relative to the Milwaukie station.


The proportion of monthly streamflow at the Milwaukie station that originates upstream of the Gresham station is computed in table 5. The proportion has a positive correlation with the total monthly streamflow. During the five wettest months of the year, November through March, about one-half of the streamflow at the Milwaukie station originates upstream of the Gresham station (table 5). For reference, the drainage area upstream of the Gresham station represents 29 percent of the total drainage area at the Milwaukie station. As streamflow decreases, the proportion of streamflow originating upstream of Gresham also decreases and reaches a low of 10 percent in August, the month typically with the least amount of streamflow in the year. The decrease in streamflow upstream from Gresham relative to overall streamflow in Johnson Creek is a result of Crystal Springs Creek, which flows into Johnson Creek between Sycamore and Milwaukie and provides a relatively stable source of base flow throughout the year (Lee and Snyder, 2009). Annually, 45.4 percent of the streamflow at the Milwaukie station from water year 2007 to 2010 originated upstream of the Gresham station.


Analysis of Suspended-Sediment Budget


Annual and monthly SSL were computed for the Gresham and Milwaukie stations for water years 2007–10 (table 6, figs. 6 and 7). The SSL values were evaluated in several different ways. To compare SSL output at the two stations, monthly and annual SSL were divided by their respective drainage areas, which results in the suspended- sediment yield (SSY). SSY is a method of standardizing results, which makes for easier comparison between watersheds of differing sizes. The annual SSL and SSY were evaluated along with the findings from the streamflow analysis. The annual and monthly proportions of SSL at the Milwaukie station originating upstream of the Gresham station were also computed and evaluated in relation to the proportion of streamflow originating upstream of the Gresham station.


Annual Suspended-Sediment Loads and Yields


The computed average annual SSL was 1,890 tons at Gresham and 4,640 tons at Milwaukie. These equate to average annual sediment yields of 123 and 87.2 tons/mi² at Gresham and Milwaukie, respectively (table 7). The amount of SSL for each individual water year can be explained at least in part by the previously evaluated streamflow metrics.


On the basis of total annual streamflow, water year 2007 was the wettest of the 4 years (table 4). The late autumn of 2007 was especially wet; November and December produced the second- and third-highest monthly streamflow during the period of study. At the Gresham station, this translated to the highest computed annual suspended-sediment load of the 4 years, 2,350 tons, and annual yield, 153 tons/mi² (tables 6 and 7). For the Milwaukie station, the 2007 SSL was slightly less than the water year 2008 total. The annual SSL for the Gresham and Milwaukie stations in water year 2007 were 124 and 120 percent of the water year 2007–10 averages, respectively.


Water year 2008 was drier than 2007 on the basis of total annual flow at the Sycamore streamflow-gaging station. However, water year 2008 contained a higher peak annual streamflow than 2007 (1,430 ft³/s compared with 1,030 ft³/s), and the kurtosis of the daily mean streamflow for the water year was the eighth-highest on record (table 4), which suggest more SSL. Additionally, December 2007 (water year 2008) was the wettest month during the period of study. This resulted in total SSL values that were close to 2007 values. The annual SSL for the Gresham and Milwaukie stations in water year 2008 were 112 and 123 percent of the water year 2007–10 averages, respectively.


Water year 2009 was the driest of the 4 water years (table 4). However, it also included the third-largest peak annual streamflow on record and the second-highest kurtosis value, suggesting a higher annual total of SSL than what would be expected based on annual streamflow alone. The flashiness of water year 2009 seemed to have more of an effect at Gresham than at Milwaukie. About 2,090 tons of suspended sediment passed the Gresham station in water year 2009 (table 6), which equates to an annual yield of 136 tons/ mi². The annual total of suspended sediment at Milwaukie was computed as 4,440 tons, which is a yield of 83 tons/mi². These totals represent 110 percent and 96 percent of the 2007–10 averages at the Gresham and Milwaukie stations, respectively.


Of the 4 water years studied, 2010 was closest to the median of the 70 years of average annual streamflows. However, the timing of the streamflow was atypical, with a smaller proportion of streamflow occurring in the winter, and a greater-than-usual proportion occurring in late spring and early summer. The average annual flow of 48.4 ft³/s at Sycamore ranked 39th out of 70 years of record (table 4). Both the peak annual streamflow and kurtosis values were relatively low, ranking 55th and 48th, respectively. Due in part to the lack of significant high peak streamflows, the 2010 annual suspended-sediment loads were the lowest of the 4 years at the Gresham and Milwaukie stations. The computed annual totals of SSL at Gresham and Milwaukie were 1,010 and 2,860  tons, respectively. The annual yields for water year 2010 at the Gresham and Milwaukie stations were 66.0 and 53.8 tons/ mi², respectively. These totals represent 54 and 62 percent of the 2007–10 averages at the Gresham and Milwaukie stations, respectively. 


Seasonal Timing of Suspended-Sediment Loads


Monthly SSL totals were investigated to evaluate the timing of SSL transport in the watershed. At the Gresham station, an average of 73 percent of the annual SSL was transported during the 3 months of November–January. On average, only 8 percent of the SSL was transported during the 7 months of April–October. The Milwaukie station produced similar results (72 and 9 percent, respectively).


Large peak streamflows account for a disproportionate amount of annual SSL. The January 2009 monthly streamflow produced more than 54 percent of the annual suspended-sediment budget for both stations. The highest proportion of annual SSL totals for either station was in January 2009. For the Gresham station, the highest monthly SSL value was in January 2009, even though only the fourth-highest streamflow for the study period was measured there during that month. The second-highest monthly SSL value at the Milwaukie station was in January 2009. The highest monthly streamflow for both stations was in December 2008 and produced the highest monthly SSL during the period of study at the Milwaukie station, which was about 25 percent higher than the January 2009 SSL value. The second-highest SSL for the Gresham station was in December 2008. This discrepancy suggests that the Gresham station is more responsive to large peak streamflow events (such as January 2009) than the Milwaukie station.


The 2010 June peak streamflow was unusual because high flows typically do not occur in the watershed during June. June 2010 was one of the periods without turbidity data, and SSL values for the month are estimated. Although the estimated SSL totals for June 2010 were not large compared to those in winter or early spring, they were 17 and 13 times greater than the SSL totals for any other summer month 
(June–September) during the period of study for the Gresham and Milwaukie stations, respectively.


The SSL totals during the winter storms demonstrate that most SSL was transported during large streamflow events. The extent to which the highest streamflows produce SSL is quantified in table 8. A nonexceedance level represents the percentage of time that a specific streamflow of SSL level is not exceeded. For example, the 0.9-nonexceedance level at Gresham is 92.3 ft³/s. Throughout the 12 years of streamflow data collected at the Gresham gaging station, 10 percent of the time, the streamflow rate exceeded 92.3 ft³/s, and 90 percent of the time, the streamflow rate was less than that value. A nonexceedance level of 0.99 is rarely exceeded (an average of one time for every 100 units of measurement), whereas a nonexceedance level of 0.5 indicates that value is exceeded one-half of the time (median).


In Johnson Creek, the highest 1 percent of streamflow (nonexceedance level of 0.99) carried about one-half the total SSL during the 4 years of study at the Gresham and Milwaukie stations. Similarly, less than 1 percent of SSL is transported at streamflows equal to or less than the median streamflow (nonexceedance of 0.5) during the 4 years of study at both stations. 


Figure 8 expands on table 8 but is not directly comparable. The cumulative amount of SSL equal to or greater than specific nonexceedance values of streamflow is shown in table 8. Alternatively, the nonexceedance values of SSL are compared with cumulative values of SSL in figure 8. This comparison provides a means of evaluating the degree of skewness in the distributions of SSL and streamflow. If the streamflow for Gresham were constant throughout the year, the nonexceedance values on the x-axis would be equal to 100 minus the cumulative values of the y-axis. As the line moves farther to the right, more skew is apparent, indicating that a few high values of streamflow or SSL account for a greater percentage of the cumulative totals. For example, in figure 8A, point A shows that the top 1 percent of streamflows (the total of all streamflows exceeding the 99th-percentile) accounts for 14 percent of all cumulative streamflow over the 4 years of study at Gresham. Similarly, point B shows that the top 1 percent of all SSL accounts for 53 percent of cumulative SSL during the 4 years of study.


The distribution of SSL is more heavily skewed than streamflow at the two stations (fig. 8). That is, SSL plots higher on the graph, indicating that the highest SSL nonexceedance values account for a much greater share of cumulative SSL than streamflow. This indicates that most sediment moves through the watershed during storm events, and especially during the heaviest storm events.


Spatial Distribution of Suspended-Sediment Load


The monthly and annual percentages of SSL originating upstream of the Gresham station were calculated by dividing the monthly and annual SSL totals for the Gresham station by the corresponding monthly and annual SSL totals for the Milwaukie station, and then multiplying each total by 100 (table 9). Some suspended sediment may settle out between the Gresham and Milwaukie stations, especially at low flows with lower velocities. For purposes of this report, it is assumed that most sediment that settles between the two stations is later resuspended by future high streamflows. The possible exception is peak streamflows, which could deposit sediment high on the flood plain where water rarely reaches and vegetative growth could capture sediment.


For the wettest months (November–March), the percentage of SSL originating upstream of the Gresham station ranged from 37.0 percent in December to 43.9 percent in January. With only 4 years of data, this relatively small range could be a result of random variation rather than significant differences between these months.


The percentage of SSL originating upstream of the Gresham station peaked at 48.3 percent in April and reached a low of 9.6 percent in July, the driest month in the region. There is a positive correlation between streamflow at Sycamore and the proportion of Milwaukie SSL originating upstream of the Gresham station (fig. 9). For months when the proportion of SSL that originates upstream of the Gresham station is 25 percent or less, average monthly streamflow at the Sycamore station is always less than 10 ft³/s. Conversely, the proportion of SSL that originates upstream of the Gresham station was at least 30 percent during all months with an average monthly streamflow of 100 ft³/s or more. 


This correlation partially is a result of the tributary Crystal Springs Creek, which enters Johnson Creek between the Sycamore and Milwaukie gaging stations. Streamflow at Crystal Springs Creek is fed predominantly by springs, which, in turn, are closely tied to groundwater levels, resulting in a more constant streamflow than in the rest of the watershed (Lee and Snyder, 2009). As a result of this constant source of streamflow, during low-flow periods, streamflow yield (streamflow divided by drainage area) at the Milwaukie station remains high relative to streamflow yield at the Gresham station. Although both stations typically have low SSCs during low-flow periods (the lowest streamflow months of July and August account for an average of less than 0.2 percent of the overall sediment budget for both stations), the seasonally high proportion of streamflow originating from Crystal Springs Creek results in a lower proportion of overall SSL at the Milwaukie station originating upstream of the Gresham station.


Annual SSL originating upstream of the Gresham station accounted for 40.8 percent of the Milwaukie station totals (table 9). The drainage area upstream of the Gresham station is about 29 percent of the area upstream of the Milwaukie station. These drainage areas are based on topography, and the effective drainage area at the Milwaukie station is smaller than the topographical drainage area of the Milwaukie station because of combined sewer systems, stormwater managed infiltration, storm systems, and other anthropogenic water delivery systems. The annual streamflow at the Gresham station is 45 percent of the annual streamflow at the Milwaukie station for the period of study, or 43 percent if the entire concurrent record of both stations (water years 1999–2010) is considered. These results imply that the amount of SSL originating upstream of the Gresham station is proportional to the amount of streamflow originating upstream of Gresham.


This discrepancy between the percentage of SSL and percentage of drainage area is reflected in the annual yield totals. The average annual SSY of 123 tons/mi² at the Gresham station is 141 percent of the average annual SSY at the Milwaukie station, which is 87.2 tons/mi² (table 7). For comparison, the annual streamflow yield at Gresham is 139 percent of the annual streamflow yield at Milwaukie (1.99 and 1.44 (ft³/s)/mi², respectively). These results suggest that nearly all of the higher sediment yield at the Gresham station can be explained by the higher streamflow yield. This is somewhat surprising because the watershed upstream of Gresham is largely a mix of forest (about one-quarter) and agricultural areas (most of the rest), whereas the area of the watershed between Gresham and Milwaukie is largely urban (79 percent). A comprehensive investigation of the sediment availability of the watershed is beyond the scope of this study, but potential explanations include most suspended sediment originating from near-bank locations, more sediment than expected being produced by urban areas, and (or) the forested area upstream of Gresham capturing more sediment than expected.


Potential Errors and Uncertainties


One of the most readily identifiable potential sources of error is the extrapolation of the turbidity-and-streamflow-to-SSC relation. The method of calculating SSL proposed by Rasmussen and others (2009) and previous investigations assumes a linear relation between logarithmic values of SSC and turbidity and (or) streamflow. SSC values extrapolated beyond the range of those used in the regression model development assume that the linear relation between SSC and turbidity and streamflow extends outside that range. The farther that SSC values deviate from this range the less reliable this assumption and the more uncertain are the SSC values.


For the Gresham station, SSC values used in the model ranged from 39 to 288 mg/L. Eighteen percent of the total SSL computed was derived from SSC values below that range and 20 percent of the total SSL computed was derived from SSC values above that range. For the Milwaukie station, SSC values used for the model ranged from 47 to 378 mg/L. Seventeen percent of the total SSL computed was derived from SSC values below that range, whereas 1 percent of the total SSL was derived from SSC values above that range.


As with most physical science studies, there is no physical rationale for this assumption of linearity. Linear relations are assumed primarily for ease of computation and to allow for the calculation of prediction intervals and other error analyses.


Other potential sources of error include random error in any of the SSC, streamflow, or turbidity measurements, and uncertainties in the estimated time series for turbidity and streamflow (which represent 4 percent and 7 percent of the SSL at the Gresham and Milwaukie stations, respectively). Temporal physical factors also can affect the SSC-turbidity relation, including a change in the shape, size, or color of particles in the water (Anderson, 2005) or the presence of microorganisms such as phytoplankton (Rasmussen and others, 2009). These factors could affect the turbidity without affecting SSC, thus altering the relation between the two parameters.


First posted October 3, 2012