Each group interview on lightly used trails cost approximately $100, compared to 

 a cost of only $5 per interview on heavily used trails, because few interviews were 

 obtained per sampling unit. For example, stratum 4 (containing many lightly used 

 trails) yielded only 12 interviews during 36 days of interviewing; stratum 7 yielded 

 only 7 interviews during 22 days of interviewing. The least-used third of all trails 

 yielded less than 5 percent of all registrations. Deletion of these trails from the 

 sampling frame would have resulted in only a slight loss in the precision of estimates 

 and would have cut sampling costs almost in half. Use of lightly used trails could 

 have been adequately estimated by means of registration stations only, without employing 

 uptrail interviewing. 



Elimination of all lightly used trails from the sampling frame would have meant 

 that the large, highly stratified sample would not have been necessary. Relationships 

 between registration data and actual use were fairly similar from trail to trail; 

 variation was small enough that strata representing different trails, seasons, and time 

 of week could be pooled without significant loss of precision of the estimates. Con- 

 siderably less than the 110 sampling units used in the study would have been expected 

 to yield use estimates of acceptable precision. 



A presampling method is needed to determine optimum stratification and to identify 

 lightly used trails. Installing registration stations on all trails 1 year prior to 

 field sampling would likely provide a satisfactory screening method. Although all 

 entering groups do not register, the completed registration forms would indicate the 

 relative magnitude of use on each trail. 



Best Estimation Procedures 



Both linear regression and ratio estimate (correction factor) procedures yielded 

 acceptably accurate estimates. However, the ratio estimates were somewhat more precise 

 (had smaller error terms), and would generally be easier for administrators to apply. 



Extension of Results to Other Wildernesses 



The prediction equations or correction factors generated in this study are not 

 applicable to other unsampled wildernesses. Yet, our results were quite similar to 

 those obtained by Wenger and Gregersen on the Three Sisters and Mountain Lakes 

 Wildernesses in Oregon. The similarity in registration-use relationships for these 

 three areas suggests that more economical ways of estimating use and establishing re- 

 lationships between use and registration might be found. It is possible that particular 

 types of wildernesses might have similar registration rates and use patterns. Other 

 day-use hiking areas might also prove to be fairly similar. If further studies confirm 

 this hypothesis, it might be possible to sample registration responses at one wilderness 

 area only enough to confirm an expectation based on the t\^e of wilderness involved. 



Furthermore, it might also turn out that relationships change little over time. 

 The general similarity in use patterns on the Mission Mountains Primitive Area, which 

 was sampled during 1968, and the Oregon wildernesses, which were sampled 6 and 7 years 

 earlier, indicates relative stability over time. If so, estimating equations might 

 be used for rather long periods of time, perhaps 5 years or more, and only require 

 limited periodic checking to detect any shifts with time. 



Personal Interviews and Automatic Counters 



The largest cost item in this and earlier studies was personal interviewing, which 

 accounted for approximately 75 percent of the total sampling cost. Therefore, we 

 should seek to eliminate or reduce the need for interviewing by devising some substitute 

 methods, such as mechanical or electrical counters or automatic cameras. Although 



36 



