A large scale land use experiment has taken place over the last 40 years in the mountainous areas of the northwestern U.S. through timber harvesting. This land use change effects the hydrology of an area through two mechanisms:

• Clear-cut logging which causes changes in the dynamics of Rain-On-Snow (ROS) events due to changes in the accumulation and ablation of snow caused by vegetation effects on snow interception and melt; and

• Construction and maintenance of forest roads which channel intercepted subsurface flow and infiltration excess runoff to the stream network more quickly.

The relative importance of these two mechanisms is uncertain. Analysis of the effects can be approached in two ways; through retrospective analysis or through prediction.

This research involved a field data collection program in support of modeling to predict the relative effects of ROS and road construction events on streamflow. Observations of culvert discharge were made during the winter of 1995 and 1996 for future comparison to point discharge simulated using the Distributed Hydrology Soil Vegetation Model (DHSVM). The overall connectivity of the road network to the drainage network was assessed based on field observations and GIS.

2.1 Background

The field investigation was conducted in Hard and Ware Creeks, two headwater catchments of the Deschutes River in western Washington Figure 1 . The catchments lie within the Weyerheauser Corporation's Vail Tree Farm where extensive harvesting and road construction have taken place as summarized in Table 1, see Figure 2 and Figure 3.

Table 1 Land Use History
	Ware Creek	Hard Creek
Basin area	284 Ha	231 Ha
Road construction	1974 - 1985	1976 - 1980
Total road length	10.7 km	11.4 km
Harvesting	1979 - present	1984 - present
Harvested area (estimated)	66 %	35 %

Soils in the Hard and Ware Creek basins are shallow and stony (0.6 - 1.0 m in depth), overlying fractured andesite, basalt and breccia bedrock (Sullivan et al. 1987). Roads are cut well into the bedrock, intercepting the entire soil profile in many places.

2.2 Discharge Data Collection

Peak culvert discharge was measured for twelve road segments within the Hard and Ware Creek basins from January 1996 through June 1996. Road segments were selected based on characteristics considered most influential to generating a road segment response, including:

• Hillslope position and altitude;
• Soil depth and bedrock condition; and
• Vegetation treatment above the road segment.

The location and the upslope contributing area for each monitored culvert is shown in Figure 4.

Crest recording gauges, consisting of a 4" diameter stilling well and inlet pipe, were constructed at the entrance to each culvert (Figure 5). Peak stages were measured using floating cork and were recorded approximately once per week. The discharge rate was measured in triplicate for different stages by capturing and timing the discharge from the culvert outfall. Measurements of discharge were used to adjust the theoretical stage discharge curve for flow through corrugated steel pipes.

2.3 Peak Flow Events

A severe rain event between February 5 - 8, 1996 produced record flooding in Oregon and the southern third of Washington State.

• The Deschutes River near Rainier, WA peaked at 218 cms, the third largest flow on record.

• Storm damage in the Deschutes basin included several slope failures along the road network as well as a debris flow which destroyed the road in two locations.

• Two neighboring rivers, the Chehalis River near Doty, WA and the Nisqually River, near National, WA exceeded 57 and 53 year flow records.

The precipitation measured at Ware Creek for this event is shown in Figure 6. A second storm in April 1996 delivered approximately 33% of the total precipitation of the February storm, but at a peak intensity of approximately 75% of the February storm, as shown in Figure 7.

2.4 Basin Response

The recorded discharge for the February and April storms normalized by the contributing area is shown in Figure 8 for nine culvert locations and Hard and Ware Creeks. Several features stand out in this figure:

• Negligible response from culvert H018 for the April event - When this culvert was checked on April 27^th it was no longer reading the correct current stage due to six inches of accumulated sediments which blocked the inlet hose. Therefore, the recorded peak may not be accurate.

• Response for culvert W029 in April in excess of the February storm - Stage is read from the height of a cork line inside the PVC pipe. For the February storm the line was scattered and unclear and may be inaccurate.

• Response of Hard Creek in excess of Ware Creek - For both peak events Hard Creek shows a larger response per unit area than Ware Creek. Although the Ware Creek basin has been clear cut significantly more, the Hard Creek basin has a greater road and drainage density.

3.1 Culvert Classifications

Ditch-relief and stream crossing culverts in the Hard and Ware Creek watersheds were located using a hand-held portable Global Positioning System (GPS). Position data was differentially corrected for an estimated position precision of 2-5 m. Each road segment was classified for its potential connectivity to the drainage network according to the following categories adapted from Wemple (1994):

• Not Connected: road segment delivers water to a soil surface where reinfiltration occurs; or

• Directly Connected:

• Stream Crossing: road segment delivers water directly to a natural stream channel, identified by a pre-existing channel both above and below the culvert; or

• Gully: road segment delivers water to an incised gully and evidence of erosion and a channelized surface flow path exists.

Culverts were located and classified in the summer of 1996 by following the surface flow path from the culvert outlet. Culverts classifications will be verified following rainstorms in the winter of 1996. The preliminary results are summarized in Table 2 and illustrated in Figure 9.

Table 2 Culvert Classification Results
Culvert Class	Ware Creek	Hard Creek
Number of culverts	44	65
Directly Connected Stream Crossing	41 %	28 %
Directly Connected Gullied Flowpath	16 %	23 %
Not Connected	43 %	49 %

3.2 Road Segment Contribution

Road locations were surveyed on foot using GPS to distinguish between insloped, outsloped and crown road portions, defined as follows:

• Insloped - greater than one-half of the road surface drains to the ditch;

• Outsloped - greater than one-half of the road surface drains over the hillside; and

• Crown - approximately one-half of the road surface drains to the ditch and one-half drains over the hillside.

The distribution of in- and out- sloped road segments is shown in Figure 10.

3.3 Stream Drainage Density

Total channel length was calculated using the grid analysis features of Arc/Info and a 30 m resolution depressionless DEM. A map of accumulated pixels draining into in each pixel is then generated based on the pixel flow directions. The extent of the stream network is determined from the flow accumulation map by specifying a minimum number of contributing pixels.

The initial stream locations for Hard and Ware Creeks were determined by specifying a minimum contributing area of 2 hectares to represent the approximate stream length during high flow winter runoff events. The derived stream network was compared with field classifications of culvert types for accuracy.

3.4 Extended Drainage Network

The length of roads contributing stormflow directly to streams was calculated by summing the following:

• Only road segments which drain to directly connected culverts;

• The total length of insloped road portions;

• One-half of the length of crown road portions; and

• For segments connected by a gully, the shortest distance between the culvert outfall and a permanent stream.

The change in drainage density is summarized in Table 3 and illustrated in Figure 11.

Table 3 Increase in Drainage Density due to Roads
	Ware Creek	Hard Creek
Original drainage density	3.5	3.6
Contributing road length	1.97	2.26
Extended drainage density	4.4	4.9
Percent increase	25.9	36.7

4.1 Model Description

DHSVM is a spatially distributed hydrologic model, developed by Wigmosta et al. (1994), that resolves the surface energy and water balance at the DEM scale. A complete description of the model, highlighting recent changes are described by W.A. Perkins et al.

4.2 Road Algorithm

The Pacific Northwest Laboratory, under contract to NCASI, has developed a road and channel networking algorithm to account for surface runoff generation by and over forest roads and stream channels. The height of the road cut and depth of the channel incision are specified for each channel segment to determine the quantity of captured subsurface flow. Intercepted subsurface flow and directly intercepted precipitation are routed through the combined road and channel network using a Muskingum-Cunge routing scheme.

Both creeks have been gauged by Weyerhaeuser Co. and continuous measurements of precipitation and discharge are available since 1985. Air temperature at Hard and Ware Creeks is available since 1989.
5.1 Meteorological Variables

• Air temperatures from 1985 - 1989 were filled in with data from the Olympia Airport (9.5 km north west), adjusted by the mean monthly difference for the period of coincident record.

• Relative humidity was calculated using daily minimum temperature as a substitute for dew point temperature, with a correction depending on the dryness of the region (Kimball et al., 1995).

• Attenuation of daily shortwave radiation is calculated based on the daily temperature range, after Bristow and Campbell (1984).

• Longwave radiation is calculated following Bras (1990).

5.2 Distributed Soil and Vegetation Parameters

• Soil types were obtained from the State Soil Geographic (STATSGO) Data Base produced by the United States Department of Agriculture (USDA) Soil Conservation Service (SCS).

• The Weyerhaeuser Co. will compute vegetation parameters, including: leaf area index (LAI), vegetation height and percent of ground cover, using literature values selected by us for each species scaled by the stand age information recorded in their Forest Inventory and Regeneration System database.

DHSVM was applied to the Hard and Ware Creek watersheds using an imposed stream channel network. Future work will include applications with the imposed road channel network. Preliminary comparisons of predicted versus observed hydrographs for the 1990-1991 calibration period are shown in Figure 12 and discussed below:

• The largest hydrograph peaks tend to be underestimated and the smaller peaks are overestimated.

• Predicted hydrograph timing is slower and longer than the observed hydrographs.

• The height of model peaks seems most sensitive to the lateral hydraulic conductivity.

• Hard Creek response is underestimated more than Ware Creek. Although Hard Creek is smaller the discharge is consistently higher than Ware Creek. This may be attributable to variation in precipitation which is not predicted by the model.

Funding for the first author was provided by Valle Fellowship and the National Council of the Paper Industry for Air and Stream Improvement (NCASI) with support from the Weyerhaeuser Company.

Bras, R. L., Hydrology, An Introduction to Hydrologic Science, Addison-Westley Publishing Company, Reading, 1995.

Bristow K. L. and G. S. Campbell, On the Relationship between Solar Radiation and Daily Maximum and Minimum Temperature. Agricultural and Forest Meteorology, v. 31, p. 159-166, 1984.

Kimball, J., S.W. Running and R. Newman, An Improved Method for Estimating Surface Humidity from Daily Minimum Temperature. Submitted to Agricultural and Forest Meteorology, July 1995.

Perkins, W.A., M. S. Wigmosta and B. Nijssen, Development and Testing of Road and Stream Drainage Network Simulation within a Distributed Hydrologic Model, poster presented at the fall meeting of the American Geophysical Union, December 1996.

Shuttleworth, W.J., Evaporation, in Maidenment, D. R. (ed.), Handbook of Hydrology, McGraw-Hill Inc., New York, 1993.

Sullivan, K., S.H. Duncan, P.A. Bisson, J.T. Heffner, J.W. Ward, R.E. Bilby and J.L.Nielsen, A Summary Report of the Deschutes River Basin, Sediment, Flow, Temperature and Fish Habitat, Weyerhaeuser Company, Technical Report, Paper No.044-5002/87/1.

Wemple, Beverley C., Hydrologic Integration of Forest Roads with Stream Networks in Two Basins, Western Cascades, Oregon, Master's Thesis, Oregon State University, 1994.

Wigmosta, M.S., L.W. Vail and D.P. Lettenmaier, A Distributed Hydrology-Vegetation Model for Complex Terrain. Water Resources Research, Vol. 30, No. 6, pp. 1665-1679, June 1994.