the hydrologic variable are compared with a prediction of the variable based on the calibration relationship to evaluate changes caused by treatment. Three pairs of watersheds were used in this study. Road Construction During the summer of 1970, a system of permanent roads was constructed to provide access for logging (fig. 1). Roads have 5.5-m-wide subgrades and 3.7-m-wide driving surfaces. After subgrade and surface courses were compacted with a vibrating roller, road cutbanks and fill slopes were seeded, mulched, and fertilized in late September (fig. 4). All road construction operations were completed before October 1, 1970. During road construction, 6.4 percent, 7.6 percent, and 1.6 percent of CC-1, CC-2, and CC-3, respectively, were cleared (table 2). All but about 2 percent of CC-1 and CC-2 and 0.3 percent of CC-3 were revegetated a year after seeding. In general, seeding was successful in preventing surface erosion from road cutbanks and fill slopes. Logging Logging began in May 1971 and was completed before September 30, 1971. In CC-1, individual trees making up about 50 percent of total basal area were marked for a preparatory shelterwood cut. Timber was cut and removed after spur roads were constructed to tractor landings scattered throughout the watershed. Cull logs were yarded to landings, but slash was left where it fell. When logging was completed, spur roads and landings were scarified and water-barred. Slightly less than half of CC-1 was disturbed to some degree; 13 percent was compacted (table 3). In addition, some road cuts and ditches have intercepted subsurface flow. Table 2--Summary of treatment areas in Coyote Creek Experimental 1/ A preparatory shelterwood cut was made throughout the watershed. Approximately 50 percent of total basal area was harvested. Table 3- Soil disturbance resulting from yarding and slash disposal. Figures are frequencies of occurrence expressed as percentages of total observations made at 3-m intervals along transects 1/ Slightly disturbed litter removed and mineral soil exposed, litter and mineral soil mixed, or mineral soil deposited on litter and slash; deeply disturbed = surface soil removed and subsoil exposed; compacted = obvious compaction due to passage of machinery or logs. 2/ 3/ "Tractor" and "cable" refer to methods of both yarding and slash removal. Composite figures were obtained by weighting tractor and cable percentages by percent of logged area where logs were yarded by each method. To obtain composite percentages for the entire watershed, multiply by 0.30, the proportion of the watershed that was logged. bare. 4/ Columns do not add to 100 percent because disturbed areas may also be compacted or In CC-2, timber was harvested in 20 small clearcut patches ranging in size from 0.7 to 1.4 ha which collectively make up 30 percent of total watershed area. Half the patches, those on more gentle slopes, were logged by tractor whereas other patches were logged with a mobile, high-lead cable system (fig. 1). In the tractor logged units, slash was piled by tractor. Soil was compacted on 26 percent of the tractor logged area; only 5 percent of the tractor logged area remained undisturbed. In the high-lead logged units, soil disturbance was less severe. Soil in 6 percent of logged area was compacted and 42 percent was undisturbed (table 3). Road cuts and ditches have intercepted subsurface flow in several areas in this watershed also. Timber in CC-3 was completely clearcut after spur roads were constructed across the middle of the watershed and along the central ridge. Several wet areas have been exposed and drained by road cuts and ditches along the uppermost road. Timber on 38 ha (77 percent of the watershed) was cleanlogged, i.e. all material >20 cm in diameter and >2.4 m in length was yarded to a high-lead landing (fig. 5). Soil was compacted on only about 7 percent of the clean-logged area and 44 percent was undisturbed (table 3). The use of tractors for yarding and for windrowing slash in the area between the stream and the northwestern boundary of the watershed and in two other smaller areas compacted soil on 27 percent of the area and left only 11 percent undisturbed (fig. 5). Slash piles were burned during the fall of 1973. Streamflow data were analyzed to determine changes in annual yield, seasonal yields, and instantaneous peak flows after timber harvest. Annual and seasonal yields or flows are total volumes of flow computed for a water year or a season and are expressed as a uniform depth of water over a watershed. Peak flow is the maximum instantaneous streamflow attributable to a particular runoff period and is expressed in liters per second per hectare (liters/sec.ha). We used simple linear regression to obtain prelogging prediction equations and, for some analyses, postlogging prediction equations. Annual and seasonal flows and instantaneous peak flows at CC-1, CC-2, and CC-3 each were regressed against corresponding flows at CC-4, the control watershed. Each regression analysis is based on the assumption that individual values of a hydrologic variable--either annual flow, seasonal flow, or peak flows--are independent and random. We recognize, however, that independence may not always occur in this type of study. In the case of annual and seasonal yields, the significance of the difference between prelogging and postlogging data was determined by computing a prediction limit for prelogging conditions. Because timber harvest was expected to increase water yields, we used a one-tailed test for significance of changes in yield at the 0.05 level of probability. Thus, prediction limit is given by (to. 10) (Sy) where Sy is sample standard deviation of predicted Y, and = and Sy.x sample standard deviation from prelogging regression, n = number of observations in prelogging regression, and x = X-X (Snedecor and Cochran 1956, p. 135-140). The hypothesis was that there is no difference between prelogging yield and postlogging yield. If a particular postlogging annual or seasonal yield exceeded the prediction limit, the hypothesis was rejected in favor of the alternate hypothesis that the yields are significantly greater than prelogging yields. The significance of the difference between prelogging and postlogging peak flow data was determined by the principle of "extra sum of squares" (Draper and Smith 1966, p. 67-69). This principle tested the hypothesis that there is no difference between prelogging and post logging regressions, i.e. a1 = a2 and B1 = 82 where a is the intercept of the regression line, al B is the slope, and subscripts 1 and 2 denote prelogging and postlogging periods, respectively (Harr et al. 1975). Changes in Streamflow ANNUAL YIELD In the analysis of annual yields, the calibration period consists of water years 1964 and 1966-70. Because several stream gages were inoperative during part of the 1965 water year owing to bedload filling weir ponds, 1965 data have been excluded from the analysis of annual yield. In addition, data from 1971 have been omitted from the calibration and post-treatment periods because logging was underway during this year and watersheds were in neither an unlogged nor logged condition the entire year. Prelogging correlation between annual streamflow at the control watershed and annual streamflow at each of the other watersheds used in this study was good. Prediction equations for the 5-year calibration period explain at least 98 percent of the variance in annual yield as indicated by r2 values in table 4. Logging operations significantly increased annual yield in all three logged watersheds (fig. 6). Because hydrologic changes caused by timber |