Statistical analysis of logging effects on flooding in the Snoqualmie River basin, Washington

Stephen D. Rosencrantz
Pascal Storck
Dennis P. Lettenmaier

Department of Civil Engineering Box 352700
University of Washington, Seattle, WA 98195


Abstract

Field studies have shown that substantial changes in snowmelt during rain-on-snow events can occur due to removal of forest cover. Such effects are of particular concern in the coastal mountain ranges of northwestern North America, where major vegetation changes have occurred in the last 50 years due to logging, predominantly in the so-called transient snow zone. Among the hydrologic changes associated with logging that have been observed in field studies are differences in snow accumulation as a result of canopy interception changes, and differences in latent and sensible heat associated with increased wind at the snow surface. However, the field studies are of necessity essentially point observations; at the watershed scale, the effects of vegetation changes on any particular flood are complicated by variations in antecedent snow accumulation, spatial differences in temperature and precipitation during the storm, and the area- elevation distribution of the watershed. A detailed analysis of flood peak changes was conducted for the Snoqualmie River, Washington, a 1560 square km west Cascades watershed, and four of its major tributaries, which have areas ranging from 14 to 400 square km. Annual flood peaks and peaks-over-threshold (POT) series were analyzed for trends, using the Mann-Kendall test. The results show that statistically significant increases have occurred in the annual maxima series for all of the headwaters catchments, but not for the main stem. With one exception, however, there were no statistically significant trends in the POT series, nor were there statistically significant changes in partitions of the POT series for large versus small flood magnitudes. Analyses of partitions of the POT series according to the relative contribution of snowmelt showed that for the main stem and one of the headwaters catchments, statistically significant increases occurred for those floods with the smallest snowmelt signature. This result is somewhat counterintuitive, however it is consistent with findings of a previous study by Storck et al using a model residuals approach.


Introduction

There is a perception that the frequency and severity of flooding has increased in the watersheds located on the western slopes of the Washington Cascades, especially during the rain on snow (ROS) events which occur when warm fronts follow periods of cool, wet weather. This perception has been reinforced by major floods occurring during November, 1986, 1990, and 1995, all of which caused major damage. Point scale field studies have shown that removal of the forest canopy can cause substantial increases in both snow accumulation before ROS events and snowmelt during these events (Berris and Harr, 1987; Coffin and Harr, 1992). This suggests that forest harvesting could be an important contributor to recent floods, because many western Cascades watersheds have been heavily logged during the last 40 years, and much of the logging has occurred in the so-called transient snow zone. A recent investigation of small to intermediate scale catchments in western Oregon has shown a significant increase in smaller rain-on-snow events (Jones and Grant, 1995).

This research investigates the historical streamflow record of the Snoqualmie River watershed, Washington and several of its smaller tributaries for evidence of changes in flood magnitudes over the period of record. The Snoqualmie River basin is of particular interest because of its proximity to the Seattle metropolitan area, and because of flood damage that has occurred in the floods of 1986, 1990, and 1995 (three of the five largest floods in the 67 year period of record). In addition, trend analysis of the climatic forcings, as represented by simulated records of the water available for runoff (WAR, also sometimes termed rain-plus-melt) were conducted.


Method

We examined both the historical Annual Maximum Series (AMS) and the Peaks over Threshold (POT) series for USGS gauges on the main stem of the Snoqualmie River Watershed and four of its subcatchments, the North Fork, Middle Fork, South Fork Snoqualmie Rivers and the North Fork Tolt River. (See Figure 1.) This work focuses on the longest concurrent period of record for all five gauges, which extends from 1961 to 1993. Figures 2a-d show the harvest history for each of the subcatchments ( N. Fork Snoq., Mid. Fork Snoq., S. Fork Snoq., N. Fork Tolt. ) expressed as the percent of each of the subcatchments with trees less than 5 and 20 years old. Harvest histories were reconstructed from data provided by Weyerhauser Co., the Washington Department of Natural Resources, and the U.S. Forest Service. The character of the harvest pattern is somewhat similar for the four subcatchments: the transient snow zone (300- 900 m) tended to be harvested earlier than the higher elevations. The timing of the harvesting is somewhat similar for the North Fork Tolt and North Fork Snoqualmie. The peak harvest occurred earlier for the Middle Fork, and later for the South Fork. The total fraction of area harvested is significantly higher in the North Fork Tolt than in the other three subcatchments. For the main stem (not shown), the fraction of young trees tends to be smaller, and with an earlier peak, than for the headwaters catchments.



The AMS were taken directly from USGS. However, the POT were constructed from mean daily discharge, because the USGS has changed the threshold for their POT records, which are based on instantaneous discharges. We set our thresholds so that each record contained an average of three events per year. The POT series was further partitioned into "large" and "small" events such that large events occurred on average once per year. To assure that the POT series for the subcatchments contained the same events, we used the POT series for the main stem as an index and extracted the maximum mean daily flow which occurred on that or the previous day for each of the four subcatchments.

To simulate WAR we used the Distributed Hydrology Soil Vegetation Model (DHSVM) of Wigmosta et al., (1994) as extended for use in forested mountain catchments by Storck, et al (1995). DHSVM requires observations of temperature, dewpoint temperature, windspeed, incoming solar radiation, incoming longwave radiation, cloud cover and precipitation. Of these data, all but the radiation data were available from the Stampede Pass weather station which is located at the crest of the Cascade Range just outside the Snoqualmie Watershed. Incoming solar radiation was estimated from calculated extraterrestrial radiation and adjusted for cloud cover. Incoming long wave radiation was calculated assuming an emissivity of one with a cloud temperature equal to the dewpoint temperature and a clear sky temperature equal to local air temperature. Since DHSVM was only used to estimate the precipitation and snowmelt series, the model was run for 16 points which represented the area-elevation curve. The snowmelt for these points was then integrated to produce time series of basin-total snowmelt. Precipitation at each elevation was adjusted from the Stampede Pass observations according to precipitation-elevation curves constructed for each basin, which were constrained to reproduce the mean annual water balance.

For these simulations, the vegetation cover for each point was set to old-growth Douglas Fir for the entire simulation. DHSVM was run at a three hour time step and the total precipitation, snow water equivalent and snowmelt during each time step was recorded. The model output was then used to construct a ROS index for each event in the POT record by calculating the three day (two days prior to and one day during each event) catchment-total snowmelt divided by the total precipitation for the same period. We segregated each POT series into three categories using the ROS index: events for which snowmelt was less than 5% of precipitation, 5-10%, and over 10%, respectively. Each of the WAR series were tested for trends using the Mann-Kendall test, as were each of the AMS and POT flood series. For all analyses, a significance level of 0.05 was used.


Results

WAR Series:

None of the WAR series had statistically significant trends, suggesting that over the period of analysis, the climatic forcings have not changed.

Annual Maximum Series:

Figure 3 shows each of the five AMS from 1961 to 1993. ( Main Stem, N. Fork Snoq., Mid. Fork Snoq., S. Fork Snoq., N. Fork Tolt. ) For all four of the subcatchments, there were significant uptrends. However, there was no significant trend at the main stem station.

POT series:

Figure 4 shows the POT series for all events. ( Main Stem, N. Fork Snoq., Mid. Fork Snoq., S. Fork Snoq., N. Fork Tolt. ) Only for the North Fork Tolt, the most heavily harvested of the watersheds, was there a significant increasing trend. There were no statistically significant trends for partitions of the POT record into either small or large events.

Segregation of POT series events by ROS index:

For three of the basins (North Fork, Middle Fork, and South Fork of the Snoqualmie River) there were no statistically significant trends in any of the partitions of the POT series by ROS index. However, for the North Fork Tolt and main stem Snoqualmie, there was a statistically significant increasing trend for those events during which WAR was less than 5% of the total storm precipitation. For these two watersheds, the trends disappear as the ROS index is increased beyond 5%. Figure 5 shows the partitioned POT series for these two watersheds.
N. Fork Tolt (<5%).
N. Fork Tolt (5% - 10%)
N. Fork Tolt (>10%)
Main Stem Snoq. (<5%)
Main Stem Snoq. (5% - 10%)
Main Stem Snoq. (>10%)

POT difference series, North and Middle Fork

Figure 2 shows that from 1961 to present, the area of the North Fork with young trees increased while the Middle Fork decreased. These contrasting trends in harvest history do not translate into trends in the POT difference series for the two watersheds. We examined three difference measures, a raw difference of the daily discharge from each basin, a normalized difference using daily discharges normalized by the average basin discharge, and a logarithmic difference.


Synopsis

o No statistically significant trends were found for any of the WAR (climatic forcing) time series;

o For all of the headwaters streams, there are statistically significant increasing trends in the AMS, but only for the North Fork Tolt, the most heavily harvested subcatchment, were there statistically significant increases in POT series;

o There were no statistically significant increases in the POT series partitioned into large or small events for any of the subcatchments or the main stem;

o Only for the North Fork Tolt and main stem Snoqualmie, were there statistically significant (increasing) trends for partitions of the POT series by ROS index; for these two gauges, statistically significant increases occurred for those events with the smallest snowmelt component of total storm WAR.

o No statistically significant trends were found for a series of POT differences for the two subcatchments (North and Middle Fork) with the greatest contrast in harvest histories. The result was the same for three normalization schemes, and for partitions of the series by event magnitude and ROS index.

o These results are consistent in some respects, and contrast in others, with a related study of the Snoqualmie River main stem by Storck et al. (1995) which examined residuals of DHSVM simulations of AFS and POT series from observed peaks. Unlike this study, Storck et al found statistically significant increases in normalized residuals of the POT series for the smallest events, but like this study, they found those increases only in events with small ROS indices.


References

Berris, S. N., and R. D. Harr, Comparative snow accumulation and melt during rainfall in forested and clear-cut plots in the western Cascades of Oregon, Water Resources Research, 23(1), 135-142, 1987

Coffin, B. A., and R. D. Harr, Effects of forest cover on volume of water delivery to soil during rain-on-snow, Final Report, Project SH-1, Sediment, Hydrology, and Mass Wasting Steering Committee, Timber/Fish/Wildlife program, Olympia, Washington, 1992.

Jones, J. A. and G. E. Grant, Peak flow responses to clearcutting and roads, Western Cascades, Oregon, in press, Water Resources Research, 1995

Storck, P. A., D. P. Lettenmaier, B. A. Connelly, T. W. Cundy, Implications of forest practices on downstream flooding: Phase II Final Report, Washington Forest Protection Association, 1995.

Wigmosta, M. S., D. P. Lettenmaier, and L. W. Vail, A distributed hydrology-vegetation model for complex terrain, Water Resources Research, 30(6), 1665-1679, 1994.


Hydrology Homepage / University of Washington / hydro@hydro.washington.edu