Background

Controlling nonpoint source pollutants, such as sediment from silvicultural activities, depends on using only practices that effectively limit the generation and delivery of such pollutants over large areas. Using specific Best Management Practices (BMPs) for specific classes of activities is the primary control strategy, under the Clean Water Act. for nonpoint source pollution. Yet the effectiveness of accepted practices must be established and continuously demonstrated for the strategy to remain credible. For silvicultural activities, up-slope monitoring of implementation and effectiveness at the site of practices is relatively simple and yields immediate results that can be used to discover implementation deficiencies and improve practices (See Roby et al., WMC Newsletter, Fall 1991) However it is desirable to also investigate the downstream cumulative effects of entire implemented sets of BMPs on water quality and beneficial uses, in-channel.

To meet this need, the Six Rivers National Forest, in cooperation with the Pacific Southwest Research Station, implemented a paired watershed study in 1985 at Hetten and Tompkins creeks in the upper Mad River, to evaluate the effects of modern timber harvest practices (using BMPs) on suspended sediment and turbidity in a mountain stream (USDA,1979; Sullivan et al, 1987; NCRWQCB, 1989). Rainfall, streamflow, suspended sediment concentration (SSC) and turbidity were monitored from water years (wy) 1985-1993. Timber harvest treatments were initiated and completed midway through the study in summer/fall 1989. The goal of this study was to isolate suspended sediment concentration and turbidity impacts in response to BMPs implemented over an entire timber sale area.

Study Area

Hetten and Tompkins creeks are both approximately 1 square mile in area and are located in the upper Mad River basin near Ruth, California. Each watershed drains a montaine conifer forest with large stands of live oak on drier sites. Common Franciscan assemblage basement rock in the area has weathered to form coarse, gravely, and slightly acidic well-drained soils. Precipitation ranges from 34 to 141 inches per year, averaging 71 inches annually since 1973. Elevation ranges from 2800 feet at the monitoring stations to 4720 feet at ridges. Stream gradients range from 4-8% near monitoring stations, 10-18% in mid reaches, and 18-35% in headwater streams.

Methods

Sampling. Stage height, suspended sediment concentration (SSC) and turbidity were monitored at Hetten (control) and Tompkins (treated) Creeks during the fall, winter, and spring of each water year from 1985-1993. Stage height was automatically recorded on a data logger. Rating curves converted sampled stage heights to stream water discharge. Piecewise SALT (Sampling At List Time - R. Thomas, 1985), a variable probability sampling algorithm, together with electronic circuitry and pumping mechanisms controlled the automatic collection of water samples. This design selected samples randomly but with frequency in proportion to rising discharge in runoff events. A technician visited each station every two weeks to download electronic data, collect and analyze water samples, and maintain equipment. Data were processed and analyzed at the Redwood Science Lab in Arcata.

Roadbuilding Treatment. A temporary road was constructed (0.46 miles) in the treatment phase of monitoring bringing total road mileage in the Tompkins watershed to 2.54.

Timber Harvest Treatment. In summer 1989, 109 acres were logged from the Tompkins Creek watershed, totaling 16.2% if its basin area: 99 acres were clearcut and 10 acres were selection cut. Timber removed totalled 4.3 million board feet of ponderosa pine, Jeffrey pine, sugar pine, red fir, white fir, and Douglas fir.

Slash Treatment. Slash was either piled by machine and burned or broadcast burned with a surrounding protective fireline in all units except one selective overstory removal unit. Fire accidentally escaped the fireline in one unit, burning an additional 20 acres. Although the accidental burn was not a BMP-recommended procedure, it did ensure logging procedures reflected real-life operations as accidental occurrences are not uncommon in large operations such as this.

Analyses

Initial Analysis. The log of SSC and the log of turbidity were both plotted and regressed against the log of discharge for pre- and post-treatment conditions. Chow's test, which tests for differences in linear regressions, compared pre- and post-treatment regression equations from the treatment and the control stream. Initial Chow's tests (alpha = 0.05) did indicate a change in the SSC and turbidity to discharge relationship at Tompkins Creek after treatment (increase in slope and intercept) but also indicated change at Hetten Creek (control) in the post-treatment period (increase in slope and decrease in intercept).

Substantial variability, the large number of water samples (over 1000 at each creek), the inherent sensitivity of Chow's test to variation, and simple curiosity caused us to reorganize the data and examine it in different ways in order to reduce variation left unexplained by discharge. Analysis by three methods of discharge class distribution provided inconsistent evidence of treatment effects and were disregarded. The remaining analyses examined SSC and turbidity during periods of storm runoff. A comparison of SSC between rising and falling limbs of storm hydrographs revealed more higher sediment concentrations during rising storm limbs. Analysis did not indicate an increase in SSC or turbidity during either the rising or falling limb after the treatment. The following methods did isolate an increase in suspended sediment during storm runoff conditions in the post-treatment period over that established in background.

Storm Suspended Sediment Yield. Single observations were defined as storm events in which stream discharge reached or exceeded 15 cfs (recurrence interval = 0.2 years), to focus on periods when suspended sediment and turbidity effects would most likely occur. Adequately sampled storms from Hetten and Tompkins creeks were selected for suspended sediment yield analysis: 16 paired storms from pre-treatment years (1985-1989), 8 paired storms from post-treatment years (1990-1993). Suspended storm loads were computed with SEDYIELD, a computer program written by Jack Lewis of the Pacific Southwest Research Station. We used the Hetten Creek (control) sediment load (kg/ha) as an independent variable predicting Tomkins Creek (treated) background sediment transport during the pre-treatment period through regression analysis (R2=0.826). Of the 8 post-treatment storms, 7 plotted above the regression line; 3 of these above the upper limit of a 95% prediction band used to envelop natural variability. The largest post-treatment storm contributed 180 kg/ha of suspended storm sediment at Tompkins Creek while the background (Pre-treatment) level for this storm was estimated at <25 kg/ha. The 4 highest yields from the post-treatment period averaged 500% of predicted (background) loads. Recovery was indicated in that the increase in sediment loads over background attenuated with time (Figure 1).

Cleansing Storms: A "cleansing storm" was defined to identify the first storm in any water year capable of entraining and transporting the dust and fine sediment particles accumulated during the summer. Cleansing storms were recorded in 3 of 9 water years at Hetten Creek and in 5 of 9 water years at Tompkins Creek (2 pre-treatment, 5 post-treatment). Cleansing storms were easily identified as outliers in graphs of SSC/discharge plotted against time. Results from Tompkins Creek cleansing storms in the post-treatment phase consistently exceeded the 20% over background turbidity standard established by the local state water quality basin plan-about 50% over background levels with discharge 1-100 cfs. (NCRWQCB, 1989). (Figure 2) SSCs in cleansing storms were nearly 200% of pre-treatment levels with discharges of 1 cfs and 500% of pre-treatment levels with discharges of 10 cfs (Figure 3). Extrapolating the regression line to discharges over 10 cfs (the limit of pre-treatment discharges) suggested increases of 1000% of background fopllowing treatment. In contrast, Hetten Creek cleansing storms in the post-treatment period produced less SSC and turbidity than did those in the pre-treatment phase.

Biological Assessment: While researchers agree that sensitivity of freshwater organisms to SSC is a combined function of the number of exposures an organism receives, the concentration, frequency, and duration of exposures, I found only Newcombe and MacDonald (1991) reviewed work citing duration of exposure with concentration and responses of freshwater organisms. They prepared a regression equation relating intensity (mg*hrs/l) to an increasing hazard rating (HR) scaled from 1-14.

HR = 0.738 (ln intensity) + 2.170 (R2 = 0.638, n=120)

This work has been criticized for over-simplifying cause and effect yet their methods were well suited to our data and allowed biological interpretation. Hetten and Tompkins SSC (mg/l) and time (hrs, as time elapsed between samples) were multiplied and summed to give an "intensity" for each storm Substituting these into their equation, hazard ratings were calculated as follows:

Hazard ratings at Tompkins Creek were higher than at Hetten Creek both in background and in post-treatment conditions. Surprisingly, the HR average from the control watershed (Hetten) increased more than did the HR average from the treated watershed (Tompkins) after treatment. I concluded that HR did not increase with treatment during storms with recurrence intervals >0.2 years. Hazard ratings could not be calculated for cleansing storms as most cleansing storms had only one representative sample and no duration values existed. One cleansing storm at the treated basin retained 2 sample bottles in the data set and a HR of 4.7 was calculated from this.

The highest hazard rating at Hetten and Tompkins creeks, 8, could have resulted, for example, from an SSC of 171 mg/l over a 24 hour duration. This hazard rating indexes to physiological stress and histological changes according to Newcombe and MacDonald. Although these SSC intensities may have had adverse effects on instream organisms, paired T-tests indicated that the treatment means did not exceed the hazard rating of control means. By this analysis any increases in SSC at Tompkins Creek appear to be within the limits of the basin plan standard that requires that there be no adverse effects on beneficial uses.

Ideas for Future Monitoring

Monitoring turbidity only to limit costs. Equipment required in monitoring for suspended sediments includes controlled/gaged channels, pumping samplers, sampling scheme of some kind, laboratory analysis, substantial electricity requirements, and frequent but unpredictable field visits to change and transport bottles. Monitoring with a continuously recording in-stream turbidimeter would tend to be much less expensive than suspended sediment sampling. Presently, turbidity is a more directly quantifiable variable for comparison to NCRWQB standards than is SSC but SSC is more readily interpreted biologically. The relationship of SSC to turbidity was fairly consistent with predictable variation at Hetten and at Tompkins Creek (R2 = .5-.72). Thus, turbidity might be reasonably usable for some monitoring applications as a surrogate for SSC, with frequent calibration.

Analysis. Time and expense could be saved by concentrating monitoring efforts at the onset of seasonal rains where the summer's dust and fine sediments in the watershed "come out in the wash". The cleansing storm method of analysis required samples taken from only the first storms in the water year and could be a cost-effective method of monitoring the most significant pulses of suspended sediment, where this would indicate effects acceptably. The monitoring design should ensure that initial storms are sampled-even those with relatively small peak discharge. Monitoring cleansing storms will not, however, capture the landsliding effects common in many watersheds because they tend to occur late in the water year as the regolith becomes saturated.

Problems with "background:" There were two fundamental problems with "background" which limit the Hetten and Tompkins findings. Most significantly, recovery from 1982 road construction was taking place between 1984 and 1989 when "background conditions" were established. As such, these background levels were most likely elevated beyond their true levels and SSC increases attributable to treatment were probably understated. Hetten Creek (control) has a road encroaching slightly into the monitored sub-basin along a ridge just upstream of the monitoring station. Once again, "background" is elusive like the Ideal Gas Law, examples are difficult to find in real life but the concept is valuable.

Conclusion

No significant increases attributable to the logging treatment were recognizable in the initial, overall analysis of SSC and turbidity. However, by examining only the storms in the data set we isolated detectable changes. Increases over background SSC and turbidity were found in the examination of initial cleansing storms and in suspended sediment storm yields. Organizing the data by discharge classes and by viewing storms in context of their rising and falling limbs did not show any consistent increases attributable to the treatment.

Turbidity increases in cleansing storms were about 50% over background at Tompkins Creek discharges of 1-100 cfs. SSC in cleansing storms was nearly twice background at discharges of 1 cfs to 5 times background at discharges of 10 cfs. Average increases in suspended sediment loads calculated for the 4 largest storms in the post-treatment periods averaged 5 times background levels.

References

NCRWQCB, 1989. Water Quality Control Plan for the North Coast Region. California State Water Resources Control Board.

Newcombe, C.P., and D.D. MacDonald, 1991. Effects of Suspended Sediments on Aquatic Ecosystems. North American Journal of Fisheries Management 11:72-82, 1991

Sullivan, Reid, Lisle, Dollof, Grant, 1987. Stream Channels: The Link Between Forests and Fishes in Streamside Management. Institute of Forest Resources, Seattle WA.

Thomas, R., 1985. Estimating Total Suspended Sediment Yield with Probability Sampling. Water Resources Research 21:1381-1388

USDA-Forest Service, 1979. Water Quality Management for National Forest System Lands in California (Best Management Practices).

Credits

Ken Wright, Chris Knopp, Rand Eads, Bob Ziemer and Bob Thomas worked together in the selection of the Hetten and Tompkins watershed pair, and in the design and implementation of the study. Ongoing conduct of the project included work by Carol Oskewecz, Leslie Wolff, Michael Furniss, Caroline Houle-Stringall, David Thornton, Dustin Villigran, and Jack Lewis. Special thanks to Jack Lewis for extensive assistance in data analysis.