The influence of sampling interval on the accuracy of trail impact assessment

Trail impact assessment and monitoring (IA&M) programs have been growing in importance and application in recreation resource management at protected areas. This paper addresses a fundamental issue in designing trail IA&M surveys: the choice of sampling interval. Specifically, the influence of sampling interval on the accuracy of estimates for selected trail impact problems was examined using a resampling simulation method. A complete census of four impact-types on 70 backcountry trails in the Great Smoky Mountains National Park was utilized as the base dataset fur the analyses. The census data were resampled at increasing intervals to create a series of simulated point datasets, At each sampling interval level, the accuracy of simulated datasets was evaluated by comparing the estimates of frequency of occurrence and lineal extent for each impact-type with actual census values. Simulation results indicate that increasing sampling intervals are associated with an overall increase in accuracy loss for all four impact-types, The direction of accuracy loss for lineal extent estimates is mixed, while frequency of occurrence estimates are consistently and substantially lower than the actual values. Responses of accuracy loss to increasing sampling intervals vary across impact-types on extent estimates, but are consistent on the frequency estimates. These findings suggest that systematic point sampling can be an appropriate method for estimating lineal extent but not the frequency of trail impacts, Sample intervals of less than 100 m appear to yield an excellent level of estimate accuracy for the four impact-types evaluated, The census-based trail survey and the resampling simulation method developed in this study can be a valuable first step in establishing long-term trail IA&M programs, in which an optimal sampling interval range with acceptable accuracy is determined before investing efforts in data collection, (C) 1999 Elsevier Science B.V. All rights reserved.