hazards. The following guidelines are suggested instead: 



1. Limit size of clearcut units; 



2. Stagger location of clearcut units or blocks both in space 

 and in time; 



3. Leave buffer zone of trees above and below haul roads; 



4. Leave buffer zone of undisturbed vegetation along all 

 streams. 



SELECTION LOGGING 

 VERSUS CLEARCUTTING 



The analyses and findings reported here clearly recommend 

 leaving as much residual timber stand as possible from the point 

 of view of preventing surf icial and mass erosion. The greater the 

 amount of standing timber, the smaller the amount of soil root 

 cohesion loss, the smaller the rise in piezometric levels in a 

 slope, and the greater the amount of effective buttressing and 

 soil arching action by residual vegetation. All these beneficial 

 influences are favored by a selection logging system as 

 opposed to clearcutting. 



SITE PREPARATION AND 

 ABANDONMENT PROCEDURES 



There are a number of measures that are routinely employed 

 in conjunction with timber harvest operations to minimize slope 

 stability problems. These measures are usually specified in 

 various State and Federal forest practice rules. They include 

 such procedures as the seeding and scarifying of roadbeds, 

 removal of temporary road fills, construction of waterbreaks, 

 disposal of slash, and establishment of "vegetation leave 

 areas." 



The concept of vegetation leave areas is of particular concern 

 and interest in view of the findings reported here. Trees and 

 woody vegetation should be left undisturbed in critical areas 

 such as steep, slide-prone slopes. Vegetation should also be 

 left intact as much as possible along the margins of haul roads 

 and streams. Gonsior and Gardner's (1971) recommendation 

 bears repeating in this regard, namely, that barriers of live trees 

 should remain undisturbed immediately below the toe of fill 

 slopes and above cut slopes. This recommendation should be 

 weighed, however, against the likelihood of trees falling across 

 roads, owing to possibility of greater vulnerability to root dam- 

 age and windthrow after right-of-way-clearing. 



Hydraulic structures should be constructed with regard to 

 residual areas of slope vegetation. Crossroad drains and water- 

 bars should drain water onto undisturbed vegetation, not over a 

 fill slope or into another road or skid trail. Undisturbed vegeta- 

 tion should be left to provide water spreading areas large 

 enough to accommodate all water draining from roads, skid 

 trails, and similar locations. Particular care should be taken to 

 avoid "stream piracy" during water spreading operations. This 

 can easily happen when water is intercepted by the readout in 

 one or more microwatersheds and is carried downslope along 

 the road in the road drainage system and allowed to spread in 

 an adjacent microwatershed. 



General Slope Hazard Rating Scheme 



The recommendation to leave vegetation intact and in-place 

 in critical areas during timber harvest operations requires that 

 some procedure be employed to identify slopes prone to high 

 mass erosion hazard. Several schemes have been devised for 

 identifying hazardous slopes (Radbruch-Hall 1976; Ward 1976; 

 and Simons and others^). Most of these methods consist of 



mapping information on slope gradient, soil type, geology, hy- 

 drology, and past landslide occurrence. This information is 

 integrated by linear combination or factor overlay techniques 

 (Hopkins 1977) to produce a composite map of relative slope 

 hazard. 



An alternative approach is to base slope hazard ratings on a 

 geotechnical model employing principles of limiting equilibrium. 

 Geotechnical models such as the infinite slope analysis used 

 here (equation 2) explicitly account for the primary factors in 

 landslide occurrence such as soil strength, ground water influ- 

 ences, vegetative effects, and slope inclination. Geotechnical 

 models represent actual field conditions; hence, they can be 

 used to analyze the response of a hill slope to temporally and 

 spatially varying factors. The geotechnical models or slope 

 stability analyses are routinely used by engineers to evaluate 

 the stability of a particular hillslope, determine the influence of a 

 particular slope modification, and to assess the effectiveness of 

 a particular slope protection measure. 



One of the main difficulties with geotechnical models for slope 

 hazard analysis is that they are deterministic. As such they do 

 not satisfactorily take into account uncertainity and variability in 

 the input parameters. A way around this dilemma has been 

 developed by Wu (1 976), Ward (1 976), and Simons and others^ 

 by casting the stability equation or factor of safety equation in a 

 probabilistic framework. Instead of computing a single valued 

 safety factor for a slope, one computes a probability of failure. 

 Calculated probabilities can then be grouped into three hazard 

 classes as suggested by Simons and others^, namely; 



1. High probability when P[F«1] > 60 percent; 



2. Medium probability when 30 «P [F^l] « 60 

 percent; 



3. Low probability when P [F«1] < 30 percent. 



where P [F « 1] is the cumulative probability that the safety 

 factor (F) is less than or equal to one. 



Computation of the probability of failure requires knowledge 

 of the mean and variance of input variables in the safety factor 

 equation. This type of information is seldom available without 

 extensive testing. This difficulty can be overcome by assuming 

 that the input variables are uniformly distributed, random vari- 

 ables. With this assumption the mean of a random variable is 

 simply found as 



X — Xg + Xb 



2 



and the variance as 



Var [X] = (Xb - Xa) ^ 

 12 



where Xg and Xt, are the lower and upper limits on the variable X. 

 Thus, probability of failure can be estimated solely from knowl- 

 edge of the range in each variable, information which is readily 

 available. Simons and others^ show that the assumption of a 

 uniform distribution provides a conservative estimate of prob- 

 ability of failure. The authors also provide a well-documented 

 example or application of their method for identifying potential 

 landslide areas in terms of their probability of failure. 



21 



