Journal of leisure Heseanh Copyright 20062006, Vol. 38, No. 2, pp. 1H7-2O7 National Rerreaiion and Park Assorialion

Benefits and Challenges of Computer SimulationModeling of Backcountry Recreation Use in the

Desolation Lake Area of the John Muir Wilderness

Steven R. LawsonVirginia Polytechnic Institulp and State University

Robert M. ItamiGeoDimensions Pty., Ltd.

H. Randy GimbleltUniversity of ArizonaRobert E. ManningUniversity of Vermont

This paper descrihes the development and application of a computer-basedsimulation model of recreational use in the John Muir Wilderness Area in theSierra Nevada Mountains of California, USA. The results of the study demon-strate, conceptually, how sitnulation modelitig can be used as a tool for under-standing existing visitor use patterns within the John Muir Wilderness Area,estimating the impact of increasing visitor use levels on management objectives,and evaluating the effects of alternative policy decisions on visitor flows andvisitor use conditions. This study also identifies and discusses potential chal-lenges of applying computer simulation to backcountry recreation managementand provides recommendations for further research to address these issues.

KEYWORDS: Visitor flows, monitoring, wilderness managpmeid. Limits of AcceptableChange, crowding.

Introduction

In the United States, legislation dictates that Wilderness areas should bemanaged to, among other things, provide recreational visitors with "oppor-tunities for solitude or a primitive and unconfined type of recreation" (Hen-dee & Dawson, 2002). However, the growing popttlarity of outdoor recrea-tion in backcountry settings threatens the ability of wilderness managers toachieve these objectives. For example, increasing recreational use of wilder-ness areas can result in perceived crowding and increasing conflict amongdifferent types of users (e.g., hikers and packstock; Manning, 1999). These

Address correspondence to: Steve Lawson, Department of Eorestry. 307A Cheathani Hall, Vir-ginia Tech, Blacksburg, VA 24061, Phone: (540) 231-8303, Email: lawsom&uUdu.

Autlioi nottr: The authors would like to thank the following people for thfir scholarly reviewand assistance with this study: David Colt-, Aldo Leopold Wilderness Research Institute, GaryOye, USDA Eorest Service, Marilyn Hof, Kerri C:ahill, and Jan van Wa^tendonk, U.S. NationalPark .Semce, Mark Vande Kamp, University of Washington, and Aurora Moldovanyi, VirginiaPolytechnic Institute antl State University.

The research resulting in this publication was funded jointly hy ihe USDA Forest Sei-vice,U.S. National Park Service, and the Aldo Leopold Institute.

187

188 lAWSON, 1TA.MI. GIMBl.KTT AND MANMNX.

problems are exacerbated by tbe fact tbat backcountry recreation use tendsto be concentrated both spatially and temporally (Hendec & Dawson, 2002;Lucas, 1980). For example, most wilderness areas are used most heavily dur-ing tbe summer, atid witbin the summer montbs, use can be bcavier on ibeweekends tban during weekdays. Similarly, recreational use tends to concen-trate geograpbically along establi.sbed biking trails/roules, along tbe peripb-ei7 ratber tban witbin tbe interior of an area, and close to desirable naturalfeatures {e.g., water bodies, scenic views).

Rules and regulations designed to manage recreation-related impactssucb as crowding, conflict, and damage to natural resoutces can diminisbvisitors' sense of spontaneity and freedom, tbus eroding tbe primitive andunconfined nature of tbe wilderness experience (Cole, Peterson. 8c Lucas,1987). Managers are faced witb tbe challenge of preventing and mitigatingrecreation-related impacts to wilderness witb tbe tnost unobtrusive, indirect,light-banded means possible (Hendee 8c Dawson, 2002). Tbat is, tnanagersare expected to identify tbe "minimutn tool" required to achieve desiredconditions within wilderness. Conseqtiently, decisions about how to managerecreational use of wilderness are cotnplex.

Recent researcb suggests that computer-based simulation modeling is aneffective tool for belping lo address ibe cballenges as.sociated with managitigvisitor use in backcountry and wilderness settings (Daniel & Gimblett, 2000;Gimblett. Ricbards, 8c Itatni, 2000; Lawson 8c Manning, 2003a, 200: b; Law-son, Manning, Valliete, Sc Wang, 2008; Lawson, Mayo-Kiely, Sc Manning, 2003;van Wagtendonk, 2003; Wang 8c Manning, 1999). For example, simulationtnodeling can be used to describe and understand existing visitor use con-ditions tbat are inherently difBcutt to obser\e. That is, given current man-agement practices and existing levels of visitor use, where and when is visitoruse occurring? By providing managers with detailed information about howvisitors are currently using tbe area, tbis baseline information can assist man-agers in identifying "trotible spots" or "bottlenecks," as well as areas tbatmay be capable of accommodating additional use. Siinuiation tnodeling canalso be used to monitor the condition of "hard to measure" indicator vari-ables (Lawson, Manning, Valliere, 8c Watig, 2003; Wang Sc Manning, 1999).For example, how many encounters do backpacking visitors have with othergroups per day while hiking? How many nights do visitors camp witbin sightof other groups? In addition, simulation modeling can be used to test tbepotential effectiveness of alternative management practices in a manner thatis more comprehensive, less costly, and less politically risky tban on-tbe-ground trial and error (Lawson 8c Manning, 2003a, 2603b). For example,wbat effect does a permit quota bave on tbe number of encounters visitorsbave with other groups while hiking? How would tbe number of biking en-counters cbange as a result of redistributing use from heavily used trailheadsto less commonly used entry points? These capabilities make computer-basedsimulation modeling a useful tool for assisting matiagers in identifying po-tential recreation-related pioblems and evaluating the effectiveness and coststo visitors of potential solutions to these problems.

BF.NEFITS AND C:HALLt:NGt;S OF f:OMPl!TF.R SIML lAllON 189

This paper describes the development and application of a coniputer-hased simulation model of recreational use in the John Muir Wilderness Areain the Sierra Nevada Mountaiins of California, USA. The results of the studydemonstrate, conceptually, how simulation modeling can be used as a toolfor understanding existing visitor use patterns within the John Muir Wilder-ness Area and estimating the effects of visitor use policy decisions on thecondition of crowd ing-related indicators of quality. Furthermore, this studyillustrates how Geographic hifonnation Systems (GIS) can be used to com-municate simtilation modeling results regarding the spatial distribution ofvisitor use of a backcountry recreation area in a manner that may be morenseful for managers and easier for interested publics to understand (I^n-drcs, Spildie, Sc Q^ ieen, 2001), Lastly, this study identifies and describes sev-eral potential challenges associated with applying computer simulation tobackcountr)' recreation management. The lessons learned from this studyprovide new insight regarding how to efficiently and accurately apply com-puter simulation modeling to recreational use management and have impli-cations for future research.

Methods

Description of Study Area

In this study, a computer-based simulation model of recreation use wasdeveloped for a portion of the Humphrey's Basin area of the John MuirWilderness Area. The )ohn Muir Wilderness covers . >H4,()00 acres in the Si-erra and Inyo National Forests, in the Siena Nevada Mountains of California.The area is characterized by snow-capped mountains with hundreds of lakesand streams and lush meadows. Lower elevation slopes are covered withstands of Jeffrey pine, incense cedar, white and red fir and lodgepole pine.The higher elevations are barren granite with many glacially carved lakes.

Data Collection

This study evolved from a larger visitor use study conducted in the InyoNational Forest as part of the Enviroinnental hnpact Assessment process forthe John Muir, Ansel Adams, and Dinkey Lakes Wilderness areas (USFS,2001). In the study reported in this paper, only trip itinerai7 data for theHumphrey's Basin area of the Inyo National Forest were used. All data re-quired for the current simulation stndy, which are described in detail below,were derived from the larger University of Arizona study.

Visitor Characteristics

During the 1999 \isitor use season, diary questionnaires were distributedto hackcotintry visitors in the John Muir Wilderness. Questionnaires weredistributed at trailhead self-registration stations and at ranger stations whenvisitors picked up their agency-issued permit. Randomly selected self-

190 LAWSON, ITAMI, C.IMBI.FXr AND MANNING

registration stations were periodically attended by data collectors who dis-trihuted diaries to visitor groups and collected completed questionnairesfrom groups as they finished their trips. In addition, questionnaires weredi.stributed by commercial packstock outfitters, following instructions givenby the research team.

The diary questionnaire included a series of questions concerning groupand trip characteristics and a map of trails and natural features. Respondentswere instructed to record their route of travel during their visit, includingthe trailhead(s) where they started and ended their trip, and their campinglocation on each night of their trip. Respondents were also asked to reportthe duration of their visit, the number of people in their party, and theirmode of travel. The response rate for the Humphrey's Basin area of the JohnMuir Wilderness was 32.2%, resulting in a total of 324 completed diaries.

Site Characteristics—Trail Network

Data concerning the trail network within the study area were providedhy the USFS Inyo National Forest as a GIS shapefile. The trail network shape-file was created hy heads-up digitizing trails and known informal hikingroutes from 1994 USGS 7.5 minute quadrangles, 1:24,000 scale. These datawere supplemented with information from a campsite inventory completedin the summers of 1999 and 2000. The data included all trail segments andintersections within the stndy area.

Site Characteristics—Campsite Clusters

"Campsite clusters" were created from the visitor surveys by groupingvisitor reported camping locations based on proximity and common access.A single campsite cluster was comprised of al! reported camping locationsthat were within a (subjectively determined) reasonable distance of eachother. The campsite clusters were used to determine camping encounterswithin the travel simulation model. Specifically, groups camping at locationswithin the same campsite cluster were considered to be within close enoughproximity to he within sight and/or sound of each other.

Travel Simulation Model Design and Analysis

The data described in the previous section of this paper were used asinputs in the construction of a dynamic travel simnlation model. The travelsimulation model was developed using Extend software, and a duplicatemodel was developed using RBSim2 software (see Lawson, Manning, Valliere,&: Wang, 2003 and Itami et al., 2004 for detailed descriptions of Extend andRBSim2, respectively). The scope of this paper is limited to disctissing theresulLs of the Extend travel simulation model. Further research is being con-ducted hy the authors of this paper to compare the outputs of the Extendand RBSim2 travel simulation models of the study area.

The Humphrey's Basin travel simulation model is a prohahilistic, steadystate simulation (Law He Kelton, 2000). Prohahilistic simulation models are

BENEFITS .\NU CHALLENGES OF COMPUTER SIMUIATION KU

based on probability theoiy In these models, a simple random sample istaken of the population (in this case visitor trip itineraries). The variationof trips is then modeled based on the probabihty of a visitor selecting a singletrip itinerary out of the entire sample, or alternatively, the probability ofselecting the next destination based on the probability distribution of alldestinations originating from the current destination. Prohability models arethe standard method for modeling baseline conditions. Probability distri-butions for either trip itineraries or origin-destination pairs are a convenientway to "ramp up" numbers of visitors in a simulation since the standardassumption is that as the number of visitors increase, the distribution of tripitineraries will remain the same. Steady state simulations are designed tomodel a system during the period when it reaches its full operating level(e.g., during the peak period of the visitor use season). Conseqnently, steadystate simulations require a "warm up" period to reach the target steady stateoperating level. Furthermore, steady state simulations require substantial rep-lication (e.g., simulated visitor use days) in order to obtain reliable outputsthat are not biased by short-term effects of the probabilistic componentswithin the model or auto-correlation. In the case of steady-state simulation,replications can be obtained by making one single, long run, rather thanmultiple shorter runs. The advantage of making one single, long run is thatit is only nece.ssary to simulate the warm-up period a single time, rather thanfor each shorter simulation. In all of tbe simulations conducted in this study,the model was run for a total of 2,000 simulated visitor days. The first 500days of each simulation were dropped from the study analyses in order toavoid potential start-up effects within the simulation. The outpuLs from theremaining 1,500 days were used to generate the data reported in this study.

As noted earlier, the response rate to the diary questionnaire was 32.2%,which raises the question of whether the study data constitute a represen-tative sample of trip itineraries in Humphrey's Basin or if they are biased asa result of a relatively high non-response rate. Furthermore, while visitorswere randomly sampled at trailheads, there was less control over commercialpackstock outfitters' sampling procedures and therefore, less confidence thatthe sample is completely random. Consequently, the primary purpose of thisstudy was to demonstrate conceptually the potential utility of computer sim-ulation modeling as a tool to help managers address the challenges of mon-itoring and managing recreation use in backcountry and wilderness settings.That is, while the computer simulation model was used to generate quanti-tative results, these data may be biased due to inadequate sampling of camp-ing trip itineraries and should not be used to inform management decisionsin the study area. Therefore, some simplifying asstimptions were constructedinto the model. First, data concerning packstock trips and day trips wereexcluded from the simulation. Second, the travel simulation model was de-signed to simulate backpacking use within a sub-section of the larger Hum-phrey's Basin study area. Third, the campsite clusters were subjectively de-fined, based on proximity of campsites. Wliile it would be possible to useGIS to define campsite intervisibility more realistically, subjectively definedcampsite clusters were considered sufficient, given the conceptual nature of

192 1J\WSON. ITAMI. GIMBl.ETT AND MANNIN(;

the study. The section of the .siudy area for which the model was developedis lelerred lo as the Desolation Lake Locale. Figure 1 presents a map ofHumphrey's Basin with the Desolation Lake Locale highlighted by a box.Locations of the three access points into the Desolation Lake Locale arelaheicd on the map by ihe letters "TH" and a rode number, which will bereferred to later in the paper.

The majority of visitor vise occurs within the Humphrey's Basin areaduring the summer months, with considerably less recreation use during theearlier and later months of the season. Therefore, the computer simulationmodel was designed to focus on the "peak" period of the visitor use seasonby excluding data concerning trips starting before July 1, 1999 and afterSeptember 30, 1999. The simplifying assumptions described in the pre\'iousparagraph, coupled with ihe decision to focus on the summer months of thevisitor use season, resulted in a total of 190 useable trip itineraries includedas inputs into the travel simulation model.

Tlie travel simulation model is designed to allow the user to manipulateseveral management-related parameters within the model. This feature ofthe model allows the user to estimate the effect of alternative managementpractices and visitor use scenarios on \isitor use densities and hiking andcamping encounters within the study area. For example, the modelis de-signed to allow the user to control the number and timing of trips stiirtingeach day from each of the three entiy points into the study area. This ca-pability allows the user to design simulations that test the potential effect ofincreasing total use levels, trailhead quotas, temporal and spatial redistribu-tion of visitor use, and other management practices on crowding-related in-dicators of quality within the study area.

Model Outptits

A series of simulations were conducted to generate a common set ofoutpuLs concerning visitor use densities and hiking and camping encounters.The common data generated within this series of simtilations included:

{1) Avmige biking me per day, by trail seginent. Average hiking use perday is calculated for each tiail segment by summing the number of groupsthat pass through each trail segment during the cotirse of the simulationand dividing by the total number of days simulated.

(2) Average hiking encounters per group per day, by trail segment. Hikingencounters are calculated for each trail segment on each day that at leastone group passes along the trail segment. Two types of hiking encounterswere calculated within the simulation model. "Overtaking encounters" aredefined as one group passing another group while traveilinir in the samedirection along the trail. "Meeting encounters" are defined as two groupspassing each other while travelling along the trail in opposite directions(Schechter & Lucas, 1978). Ihe average number of hiking enconnters pergroup per day is calculated for each trail segment by summing the totalnumber of hiking encounters along the trail segment throughout the sim-

BENEHTS AND CHALLENGES OE COMPUTER SIMl'IATION 193

194 LAWSON, ITAMI, GIMBLETT AND MANNING

ulation and dividing by the total number of groups that hiked the trail seg-ment during the simulation.

(J) Average camping use per night, by campsite cluster. Average campinguse per night is calculated for each campsite cluster by summing the numberof groups that camp at the campsite cluster during the course of the simu-lation and dividing by the total number of nights simulated.

(4) Axierage camping encounters per group per night, l)y campsite cluster.Camping encounters per group per night are calculated for each night thata campsite cluster is occupied by one or more parties. The number of camp-ing encounters a group has on a given night is equal to the number of othergroups camping in the same campsite cluster on the same simulated night.The average number of camping encounters per group per night is calcu-lated for each campsite cluster by summing the total number of campsiteencounters throughout the simtilation and dividing by the total ntimber ofgroups that camped at the campsite chister during the simulation.

Baseline Simulation

The first simulation conducted with the travel simulation model wasdesigned to generate the outputs described above for baseline visitor uselevels and existing management practices, where the baseline level of visitoruse is assumed to be equal to the number of groups that completed the diaryquestionnaire during the summer months of the sampling period (July, Au-gust and September, 1999). This simulation is referred to as the IX simula-tion throughout the remainder of this paper. In addition to generating andreporting tabular results of the IX simulation, the simulation outputs wereimported into a GIS and a map was created to illustrate the baseline spatialpattern of visitor use in the study area.

Increasing Visitor Use Simulation

A second simulation was conducted to estimate the potential effect ofincreased visitor use of the study area on visitor use densities and encountersalong trail segments and within campsite clusters. Within this simulation, theaverage number of trip starts per day was increased from baseline levels by400% at each of the three trailheads in the study area. The outputs describedabove were generated for this scenario. This simulation run is referred tothroughout the remainder of this paper as the 4X simulation.

Maximum Allowable Use Simulation

A series of simtilations were conducted to demonstrate the capability oftravel simulation modeling to assist managers in estimating the total dailyu.se that can be accommodated within an area without violating crowding-related standards of quality (Manning, 1999). Specifically, this series of sim-ulations was designed to estimate the maximum level of use that could beaccommodated in the study area without the number of groups in a selected

BENEEITS AND CH.'XLt.ENGES OE COMPUTER SIMl iATION 195

campsite exceeding five for more than 5% of nights (an arbitrarily selectedpotential standard of quality for camping use density). This was done byincrementally increasing or decreasing the simulated use levels evenly acrossthe three entry points until the result "converged" on the desired level ofcampsite cluster use (Lawson, Manning, Valliere, &: Wang, 2003). This anal-ysis illustrates how simulation modeling can be used to establish trailheadquotas to achieve desired social conditions within a wilderness area, and isreferred to as the maximum allowable use simulation throughout the re-mainder of the paper.

Model Verification and Validation

Before using a computer simulation model to support decision-making(e.g., establishing trailhead quotas, redistributing recreation use within abackcountry area, etc.), it is necessary to examine whether the model andits outputs are "correct." The process for doing so involves model verificationand validation (Sargent, 1998). Model verification is concerned with assuringthat the computer model has been implemented correctly and has beendescribed as dealing with btiilding the model right (Banks, Carson, Nelson,&r Nicol, 2001). Model validation tests whether model results accurately rep-resent the "real-world" system the model is intended to replicate and hasbeen described as dealing with building the right model.

Several model verificati(jn techniques described in Law and Kelton(2000) were conducted to verify that the simulation model of the DesolationLake Locale was implemented correctly. First, individual components of thesimulation model were developed and debugged to ensure that they func-tioned properly before integrating them into the full model. For example,modules for the campsite clusters, trail segments, trail junctions, and trail-heads were developed and tested individually before being combined in thefull model. Second, the authors conducted a structured "walk-through" ofthe simulation model to review the logic and operation of the simulationmodel. Third, Extend's built-in debugging functions were used to detect andcorrect data coding errors in the full model that resulted in incorrect routingof simulated trips. Fourth, Extend's animation function was used to visuallyverify that the model did not contain errors. Fifth, the trip itinerary datareported by respondents to the diary questionnaire were used as the basisfor a quantitative verification technique. Specifically, the proportion of camp-ing use for each campsite cluster was calculated for the diary trip itinerarydata and compared to the proportion of camping use for each campsitecluster, based on the results of the IX simulation. This technique provideda method for comparing the output of the model with known analyticalresults (Schecter Sc Lucas, 1978).

In general, the most powerful techniques for validating computer sim-ulation models are those that compare model output data to data from theactual system the model is designed to replicate (Law 8c Kelton, 2000). Dueto a tack of data from the actual system (e.g., hiking encounters per group

196 lAWSON. ITAMl. GIMBLETT AND MANNINC.

and hiking use per irailhead in the Humphrey's Basin area), it is not feasibleto conduct quantitative techniques such as results validation to assess thevalidity of the computer simulation model. However, non-statistical tech-niques were used to assess the content and face validity of tlie .simulationmodel by examining the rea.son able ness of the inputs and outputs of thesimulation model. Furthermore, sensitivity analyses were conducted to testwhether the model outputs change in the expected direction in response tochanges in selected input parameters (e.g., total use).

Results

Simulated Use Lniels: IX and 4X Simulations

Table 1 reports the mean nmnber of simulated trip starts per day bytrailhead for the IX and 4X simulations. The traillieads are differentiatedwith a code number tbat was assigned to them during the data collectionprocess. As the data in Table I suggest, the baseline level of visitor use inthe study area is relatively low, witb an average of less than two trip startsper day from tbe most heavily used of the three trailheads (Trailhead 93).Even with a 400% increase in visitor use, two of the three trailbeads wouldhave less than one trip start per day into the Desolation Lake Locale.

Camping Use and Encounters, by Campsite Cluster: IX and 4X Simulations

Table 2 reports average camping use per night and average campingencoimters per group per night, by campsite cluster for the IX and 4Xsimulations. Results ofthe IX simnlation suggest that under existing condi-tions, camping densities are low throughout the entire study area. In ali ofthe campsite clusters within tbe study area, there is an average of less thanone camping group per night. Similarly, the data suggest that under existingconditions, very few visitors encounter other groups while camping.

The 4X simulation results suggest that if use were to increase by 400%at each of tbe three trailheads in the study area, visitors who camp withincampsite clusters 7 and 37 would encounter an average of three other groupsper night. Furtbermore, visitor use densities and camping encounters wouldbe moderately bigb in several other campsite clusters, inchiding dusters 42,44, 46, and 47. However, throtighout tbe remainder of tbe study area, camp-ing densities and encounters would remain relatively low.

TABLE 1

Simulated Mean Numhn- of liackpackiri^ Trip Starts per Day, by Trailhead

\\\ ;i;t TH 94 TH 999

IX Simulation 1.89 O.Ot 0.144X Siiniilalinii 7.61 0.04 0.56

BENKFITS AND CHALLENGES OF (X)MPL]TER SIMULATION 197

TAHLF 2Ax.'n'age Camping Use and Encounters, hy Campsite

Cluster—-IX and 4X Simulations

ClusierID

7363738394041424445

46

4748

495051

5253565780

81

Mean Use'IX Simulation

(1.8fi0.120.740.05

o.ir.0,050.260,32(t,440,1 :i0,480..S10.140.040.120.070.020.04O-IO0-140.1!0.07

Me-an Kncoimters''IX Simiilaiion

(t,900,140,750,060,120,030,220,330,430.120,510.250.140.000.130.040.000-10O.OM

0-1 .^

0.090,02

Mean Use"4X Simulation

3.430.473.040,22O,.52

0,21

0,9. 11,44

L84

o,r.61,901,210,.560,130,460,2.-.{),08

0,18

0,-33

0,H0

0.42

0,2,-1

Mean Encounters''4X Simutation

3.400.443.010.280.510.190.901.411.930.651.891.120.590.140-470.260.100.140,390,610.460,23

•*M{'an number ol camping groiijDs per riiglii-''Mean number oi'camping encoLiniers per git.up per

The jnap in Figure 2 portrays the spatial distribution of camping usewithin the study area for the IX simulation. Wliile the data in Table 2 suggestthat use throughout the sUidy area is low, (he map .show.s ihe rehitiiie densityof camping use. Specifically, larger circles on the map correspond to higheruse camping locations, while smaller circles correspond to lower nse loca-tions.

Hiking Use and Encounters, by Trail Segment: IX and 4X Simulations

Table 3 reports average hiking nse pei" day and average hiking encoun-ters per group per day, by trail segment for the IX and 4X sinuilations.Result.s of tlie IX ,siinulati(jn suggest that, under exi.sting conditions, hikingdensitie.s are low throughout most of the study area, with moderate levels ofvisitor use along several trail segments (e.g., trail segments 2, 4, 5. 9, 10, 11).In addition, there are very few hiking encoimters among groups under ex-isting conditions.

198 l j \WS()N. ITAMI. GIMBI-KIT AND MANNING

Camping use categories(average number of groups per night)

• Low (0.(10-0.15)

O Moderate (0.16-0.47)

O High (0.48-0.86)

Figure 2. Spatial Distribution of Camping Use—IX Simulation

Results of the 4X simulation suggest that while hiking densities wouldincrease along several trail segments in the study area if use were to increase4-fold at each of the trailheads, hiking encounters would remain lowthroughout the trail network. In fact, the model estimates that hikers alongonly one trail segment (segment 5) would have an average of more than 1encounter per group per day.

The map in Figure 3 portrays the spatial distribution of hiking use withinthe study area for the IX simulation. Again, overall recreational use of thestudy area is quite low, however, the map illustrates the relative density ofhiking use throughotit the trail network. In particular, thicker lines on themap correspond to higher use trail segments, while thinner lines correspondto lower use segments.

Maximum Allowable Use Simulation

As stated earlier, simulation modeling can he used to help managersestimate the impact of alternative policy decisions related to visitor use andvisitor flows within a recreation area. Table 4 reports the results of a seriesof simulations designed to estimate the maximum amount of use that could

BKNKF1T.S AND CHAI.I .KNt.KS O F C O M P U IKK S I M U I A I I O N 1 9 9

TABIJi: 3Average Hiking Use and Encounters, hy Trail Segment—IX and 4X Simulations

Trail ID

23456789

10U121314151617181920212223242526272829303132333435363738394041

132

Mean Use''IX Simiiiation

3.510.083.513.430.580.140.043.353.283.200.120.200.802.95l.IO2.472.410.150.990.900.770.090.132.310.151.080.150.451.290.680.630.041.870.071.430.290.881.290.221.250.06

Mean KiuuiiiUcrs''IX Simiilaiioii

0.200.000.110.^40.030.030.000.110.100.050.000.010.040.200.020.110.050.010.010.030.060.000.010.070.020.060.010-020.010.030.050.000.090.00O.OH0.020.060.210.010.070.00

Mean Use"4X Simula lion

14.020.35

14.0213.75

2.350.550.18

13.4113.1712.830.420.863.31

11.724.569.779.610.624.133.703.210.430.499.160.504.470.681.935..M2.772.610.167.220.265.541.133.665.030.844.870.26

Mean Encoiin lei's''4X Simulation

0.750.000.421.480.110.060.010.400.400.170.020.040.160.800.060.420.190.050.070.100.200.030.050.270.040.220.080.080.020.120.180.020.S70.030.300.070.350.760.040.350.02

•'Mean number of hiking groups per day.''Mean number of hiking cncoiuilcrs per group per use day.

200 L.WSON, IIAMI, (.lMiil.i-iT AMi MANNINC,

Hiking use calegories(average number of groups per day)

Low (0.04-0.62)

— Moderate (0.63-2.30)

^ ^ High (2.31-3.51)

I) 1)5 \ 2 Miles

Figure 3. Spatial Distribution of Hiking Use—IX Simulation

be accommodated in the study area without the number of groups campingwithin a selected campsite cluster exceeding 5 more than 5% of nights. Theresults of this simulation suggest that use could be dramatically increasedfrom existing levels without exceeding this standard. While the standard andcampsite cluster selected for this analysis are hypothetical, the analysis dem-onstrates the potential capability of computer-based simuiation modeling toassist managers in evaluating the implications of policy decisions.

TABU: 4Maximum Allowable Use, by Trailhead, jar Hypothetical Camping Use

Density Standard

TH TH 94 [ I i 9'

[10.80, 11.10]'0.06"

[0.05, 0.081" [0.74, 0.82]'

^Simulated iiu an niimhcr of trip starts per day,'"95% confidence interval for simulated mean number of trip stalls per day.

BENEFITS AND (:HA1.LEN(;ES OF COMPUIER SIMULATION 201

Model Verification and Validation

The results reported in Table 5 suggest that the computer simulationmodel ha.s been constructed correctly and that operating errors have beeneliminated through the debugging and verification techniques descrihcd ear-lier in this paper. Specifically, as indicated in Table 5, there is no substantivedifference between the distribution of campsite cluster use reported in thediary survey and the IX simulated trips. Furthermore, a paired t confidenceinterval for the difference between the diaiy survey data and the simulationmodel outputs suggests that there is no statistically significant difference be-tween the survey data and simulation outputs. These verification results sug-gest that the model has heen constructed correctly.

Due to the low response rate to the questionnaire and a lack of data totest for .systematic differences among respondents and non-respondents, it isdifficult to assess the content validity of the model, which is concerned withthe reasonableness of the model inputs. Furthermore, statistical comparisons

TABIJ-: 5Travel Simulation Model Verification Results

Clu.stcr ID

7363738394041424445464748495051525356578081

Survey Data''

0.180.020.160.010.030.010.050.070.090.040.100.060.030.010.030.010.000.010.02o.os0.020.01

Modfl Output''

0.180.020.160.010.030.010.050.080.100.030.100.060.030.010.020.010.000.010.020.030.020.01

Difference'

0.000.000.000.000.000.000.00

-0.01-0.01

0.010.000.000.000.000.010.000.000.000.000.000.000.00

•Pioportion of total caniping use based on sui vey data itineraries.''Proportion of total camping use based on simulation model output.'Survey data proportion minus simulation model proportion.

2 0 2 LAWSON. ITAMI. CilMBLETT .AND MANNING

of model outputs to actual system data are not possible due to a lack of"ground-truthing" data. However, the results of the IX and 4X simulationssupport the internal and face validity of the model. That is, as the totalsimulated use of the Desolation Lake Locale is "ramped up" in the 4X sim-ulation, estimates of camping use and encounters increase for all campsiteclusters (Table 2). Furthermore, estimates of hiking use and encounters in-creased from the IX to the 4X simulation for all but one trail .segment (therewas no change in encounters for trail segment #3).

Conclusions

The Benefits of Computer Simulation for Backcountry Recreation Management

The study described in this paper illustrates the potential usefulness ofcomputer-based simulation modeling in monitoring and managing recrea-tional use in backcounti"y and wilderness landscapes. Dispersed recreationin such areas is inherently difficult to observe directly. However, the studyfindings suggest that by collecting representative data on recreational uselevels and patterns by means of trailhcad counts and a diary survey of asample of visitor groups, it is possible to develop a computer simulationmodel to estimate detailed levels and patterns of visitor use. The model de-veloped for the Desolation Lake Locale illustrates how computer simulationcan inform managers about levels of use and resulting encounters at all trailsegments and campsite clusters within a backcountry recreation area, andthis information can be used for several purposes, including identifying po-tential bottlenecks or congested sites, scheduling maintenance and patrolactivities, and educating visitors about the conditions they are likely to ex-perience.

This study also demonstrates how computer simulation can he used formonitoring purposes. Monitoring is becoming increasingly important in parkand wilderne.ss planning and management and plays a \ital role in applica-tion of the Limits of Acceptable Change (LAC; Stankey et al., 1985) andVisitor Experience and Resoiuce Protection (MilRP; Manning, 2001; NationalPark Service, 1997} frameworks developed and used hy the U.S. Forest Ser-vice and U.S. National Park Service, respectively. These frameworks requireformulation of indicators and standards of quality for resource and experi-ential conditions in parks and wilderness. Indicator variables must be mon-itored to help ensure that standards of quality are maintained. Often timesin monitoring, relatively easy to measure variables, such as parking lotcounts, will be observed as a means to estimate the condition of harder tomeasure variables, such as use densities and encounters within a recreationsystem (Hollenhorst, Whisman, &: Ewert, 1992). Similarly, simulation modelslike the one developed in this study can be used to monitor crowding-relatedindicator variables such as trail and campsite encounters. Trailhead counts(gathered on a periodic basis by means of automatic trail counters, self-registration stations, or permit data) can be used to run the model and

BKNEFITS AND <:HAl.l.KNGES OF COMPtJTER .SIMUIATION 203

estimate trail and campsite encounters. Moreover, as the "Maximum Allow-able Use" simulation conducted in this study demonstrates, computer sim-ulation models can be used in a more "proactive" way by estimating the totaldaily use that can be accommodated without violating crowding-related stan-dards of quality. In this way, a traiihead quota or permit system couid bedesigned to ensure that crowding-related standards of quality are main-tained.

Travel simulation modeling can be used to test the potential effective-ness of management practices, such as those designed to reduce trail andcampsite encounters. For example, travel simulation modeling provides man-agers with a tool to estimate the potential efTect of redistributing use amongentry points to a wilderness area, or altering the temporal distribution of useon visitor flows and visitor use-related conditions. Wliile the level of visitoruse in the Desolation Lake Locale is too low to demonstrate this capabilityof travel simulation modeling, several other studies have illustrated this(Manning Sc Potter, 1984; McCooI, Lime, &: Anderson, 1977; Potter 8c Man-ning, 1984; Smith 8c Krutilla, 1976; Underhill, Xaba, 8c Borkan, 1986; VanWagtendonk Sc Coho, 1986; Wang Sc Manning, 1999). For example, in a studyat Isle Royale National Park, a travel simulation model was developed to testthe effectiveness of a range of management practices designed to reducecrowding within the Park's backcountry campgrounds (Lawson 8c Manning,2003a, 2003b). Travel simulation results from the study suggested that im-plementing a permit quota, a fixed itinerary system, and/or campsite devel-opment would reduce crowding in backcountry campgrounds, while redis-tributing use among the entry points to the Park's backcountry would notbe an effective strategy. These findings assisted managers in identifying man-agement practices that would potentially reduce campground crowding,while avoiding the costs associated with instituting potentially ineffectivemanagement policies. Findings from a travel simulation model of visitor usealong the Appalachian Trail suggested that the number of hiking encountersalong the Trail could be reduced by altering the number and timing ofarrivals at various trailheads (Manning Sc Potter, 1984; Potter 8c Manning,1984). In fact, spatial and temporal redistributions of use along a section ofthe trail were found to be more effective at reducing the number of hikingand camping encounters than across-the-board use limits. In such cases, sim-ulation modeling may he a useful tool for optimizing the design of trailheadquota systems and/or information and education programs that redistributeuse across starting locations and starting times.

Finally, the results of this study demonstrate that computer simulationprovides outputs that integrate well with other resource data and assess-ments. For example, by exporting computer simulation results to a OIS da-tabase, it is possible to conduct overlay analyses with resource data to ex-amine relationships between natural resource characteristics (e.g., resourcefragility, resource impacts, etc.) and existing visitor use patterns. Further-more, the integration of GIS and computer simulation technologies provides

204 LAWSON, ITAMI, GIMRLFTT AND MANNINC

m a n a g e r s vviili a lool lo illiislrate wilh m a p s ilif pcHfiiiial e f i t i u of a l l f i i ia -

tivc visitor use policy decisions and management practices on visitor use pat-terns and natural resources within a dispersed, backcountry setting.

The Challenges of Computer Simulation for Backcountry Recreation Management

In addition to demonstrating the potential value of computer simulationfor backcountry recreation management, this study bas identified several an-alytical issues that need to be addressed tbrougb further research. Tbe firstcballenge tbat emerged in tbis study bas lo do witb collecting reliable, rep-resentative data with which to construct computer simulation models. Back-coimtry recreation areas, by definition, are dispersed in nature, often havingmany remote enti"y points. Consequently, it is often difficult to capture travelitineraries and otber data needed to construct a computer simulation modelusing on-site survey tecbniques. The response rate to the diary questionnairein tbis study (32.2%) bears evidence to tbis cballenge and calls into questiontbe representativeness of the camping itinerary data used to construct thesimulation model of the Desolation Lake Locale. Other studies, however,have been more successful in collecting data for computer simulation mod-els. For example, in the study at Isle Royale discussed earlier in this paper,the computer simulation model was constructed from a census of all permitsissued during the 2001 visitor use season. In addition, computer simulationsof visitor use on the Carriage Roads and the loop road on Scboodic Penin-sula of Acadia National Park were constructed from data collected throiighvisitor sui"\eys that bad response rates of 83.3% and 95.4%, respectively. Fur-tber researcb and coordination witb land managers is needed tcj identifymethods for collecting trip and visitor characteristics data in a reliable andconsistent manner. For example, alternative data collection methods, includ-ing glolial positioning system, aiUomatic timing systems (i.e., race technol-ogy), mechanical counters, and related technologies sbould be comparedwith on-site survey methods (and eacb otber) to assess tbeir effectiveness forcollecting travel ititierary data. Tbe relative strengths and weaknesses of al-ternative methods for collecting travel itinerary data should be documentedand publisbed. Fiutbermore, iu areas wbere peruiits are issued, lesearcbersshould work witb managers to develop standardized permit applications witbquestions designed to collect tbe information needed to construct valid com-puter simulation models.

Secondly, in tbis study, validation of tbe simulation model was limitedto techniques tbat rely on intuitive judgments. For example, sensitivity anal-yses conducted witb tbe simulation model suggest tbat wben total simulateduse was increased, estimates of biking and camping use and encounters in-creased. While this analysis supports the internal and face validity of thesimulation uiodel. it is less useful for making conthisious concerning bowwell tbe sitnulatiou model outputs represeut correspondiug dat;i for tbe ac-tual Desolation Lake Locale trail and campsite system. Due to a lack of suf-ficient data couceruing actual biking aud camping use and encounters.

AND (::n.u.i,i:N(;Ks OF COMPL'TKR SIMLIAIION

quantitative validation terhnitjiies were not tised in thi.s study. Fin thermore,existing applications of computer simulation to recreation management andplanning, inclnding the example presented in this paper, have generallylacked or been limited in the use of quantitative validation techniques. Thisiniderscores again the need for ftnther re.search desigued to improve re-searchers' and managers' ability to collect data thai tan be nscd not only asinputs to a computer .simulation model, but as the basis for quantitative,results validation of simulation models. Furthermore, wbile there is a rela-tively exteTisive body of literature describing validation techniqties for sim-ulation models of manufacturing systems, there is a lack of recent researchconcerning the appropriateness of alternative statistical techniqnes for vali-dation of computer simtilation models of parks, wilderness and related out-door recreation systems. Consequently, existing applications of comptitersimulaiion, including tlie example presented in this sitidy, may overstate theutilit)' ot computer sinuilation. More research is warranted to develop stan-dardized methods and procedures to assess the validity of computer siniti-tation models designed for outdoor recreation management.

Third, as Tioted earlier in this pajjer, the compnter sinnihuion model ofthe Desolation lake Locale is a steady state simulatit)n model. One of themost difficult challenges of steady state simtilation is determining the appro-priate rtintime length needed to generate statistically valid model outputs(i.e., outputs that are not biased by start-up effects or autocorrelation; Cen-teno & Reyes, 1998). Through previous researcli ou simulation, technic|ueshave been developed to establisb steady state runtime lengths needed toachieve specified levels of accuracy for model outputs (Centeno 8c Reyes,1998). However, like the quantitative validation methods referred to in theprevious paragraph, these methods were designed primarily for manufactur-ing applications and were of limited utility in this study. Conseqtientiy, tberimtime lengtli for the Desolation Lake Locale simulations was selected ar-bitrarily. Uliile intnitively one would expect the 2.000 day runtime lengthused in this study to be adequate, the melbods developed in the simulationliterature could not be used to test ottr intuition. Before the full potentialof comptuer simulation can be realized for outdoor recreation management,n)ore research is needed to develop and standardize methods to establisbsteady slate simulation riuUinie lengtbs.

Fourtb, an underlying assumption of the 4X and ihe Maxinumi Allow-able Use Simulations is that the relative frequency of the hiking and campingitineraries that visitors follow will remain the same, even as use levels rise.Since people's choice of hiking and camping itineraries are often driven bythe amount of time they have and the location of key attractions {e.g., waterbodies, vistas, mountain tops, etc.), this assumption seems reasonable. How-ever, ibis assumption tnay become more problematic for dramatic increasesin overall use levels in which case use densities may override the attractive-ness of specific landstapc features and, to a lesser extent, time constraintsin determining people's hiking and camping routes. In cases where the sce-narios to be simulated deviate dramatically Irom existing conditions, it may

20fi lAWSON. ITAMI, <;iMBI,KTT .AND MANNIN<;

be more appropriate to use an agent-based .simulation approach (Daniel &Gimblett, 2000). With this approach, simulated visitor behavior is driven bya set of rules that take into account such factors as the number of othervisitors in selected locations, time constraints, and the attractiveness of land-scape features. A primary challenge of agent based simulation lies in definingrules that accurately represent human behavior. If rules of human behaviorare not specified correctly, results of agent based simulations will not be valid.Consequendy, the choice of simulation approach must be based on the ob-jectives of the simulation project and more research is needed to developguidelines to assisi computer simulation users in making the best choiceamong simulation options.

In conclusion, this study demonstrates that computer simulation is a toolwith great potential for assisting managers as they attempt to addre.ss thecomplexities of managing recreational use in dispersed, backcountry settings.In addition, this paper provides new insights about the limitations and chal-lenges of computer simulation for backcountry recreation management. Tohelp ensure that managers can confidently use siiiuilation as a standard toolin recreation planning and management, future research should focus onaddressing the challenges and limitations described in this paper.

References

Banks,J., Carson,J.. Nelson, B., & Nicoi, D. (2001). Discrete-fvent xyHem simulation. Uppt-r SaddleRiver, NJ: Prentice Hall.

Centeno. M. & Reyes, M. (1998). Su you have yotir niockl: What to do nexi? A tutorial onsimulation oiitpui analysis. I'wreedings of Ihr 1998 Winter Simulation OmfereTice, pp. 23-29.

Cole, D., Peterson, M., & Luta.s. R. 1*>87. Man<i^ng wildtniess rerrmiion use: common pmhlenvi andpotential solutions (Research Paper IN 1-2. 0): USDA Forest Service, tntennountain ResearchStation.

Daniel, X, & GimbleU, R. 2000. Autonomous ageiiLs in the park: An introdtiction to the GrandCanyon river trip simulation model. Inlemalionnl journal of Wildtrmess, 6(3), 39-43.

Gimblett, R., Richards. M., & ItamI, R. 2000. RBSim: Geographic simulation ol wilderne.ss rec-reation he\iA\'Mr. JouTJial of Forestry, 99{4), .^642.

Hendee, J. & Dawson, (". 2002. WiUiei-ness management: Sleiuard.ship and protection of resources andvalues (3"' ed.). Golden, CO: Fulcrum Publishing.

Hotlenhorst, S., Whisman, S., & Kwcri, A. (1992). Monitoring visitor use in barkcountry and wild^r-ness: A rmiew of methods ({w^neral Technical Report PSW-GTR-134): USDA Forest Service,Pacific Southwest Research Station.

Itami, R., Raulings, R., MacLarcn, G., Hirst, K., Gimblelt, R., Zanon, D., & Chladek, P. (2004).RBSim 2: Simulating the complex interactions between human movement and the outdoorrecreation enviromiiem. journal of Nature (lonsiTiialion, 11(4). 278-286.

Landres, P., Spildie, D., &: Queen, L. (2001). GIS applications to wilderness management: Potentialuses and limitations ((ieneral Technical Report RMRS-GTR-80): USDA Forest Semce. RockyMountain Research SLation.

Law, A. M.. & Kelton. W. D. (2000). Modeling simulation and analysis (3"' ed.). Boston: McGrawHill.

Lawson, S. X: Manning, R. (2003a). Research lo inform management of wilderness camping atLsie Royale National Park: Pail I—df.scriptive ves,enrc\\. Journal of Park and Recreation Admin-istration, 27(3), 22

BENEFITS AND C:HALLKN(.;KS OK f:OMPUTER SIMUIATION 207

Lawson, S. Sc Manning, R. (2003b). Research to infonn management of wilderness camping atIsle Royalc National Park; Pan II—presrriptive rcsearth. Joumat of I'ark and Recreation Ad-ministration, 2l{?i), 43-5(i.

Lawson, S., Manning, R., Valliere, W.. & Wang. B. (2003). Proathvt- moniioring antl adaptivemanagement of social carrying capacity in Archts National Park: An application of com-puter simulation modeling, yowr/i/i/ of Envimnmentat Mnnagemenl, 68, 305-313.

Lawson. S., Mayo Kiely, A., & Manning, R. (2003). Integrating social science into park andwilderness management at Isle Royale National Park, (korge Wright Forum, 20(3), 72-82.

Liica.s, R. (1980). Use patterns and visitnr rharncteristics, attitudes, a»d preferences in nine wildernessand other roadtfsi areas (C^neral Technical Report INT-253): USDA Eorest Service, Inler-moiintain Research Station.

Manning, R. (2001). Visiloi (.-xpt'i icnte and resource protection: A frame-work for managing thec a r r y i n g c a p a c i t y o f n a t i o n a l p a r k s , foumat of Park and Recreation Administration, I9(\), 9 3 -

1 0 8 .

Manning, R. (1999). Studies in (lutdooi recreation: Search and research fvr satisfaction. Cor\"allis, OR:Oregon State University Press.

Manning, R., & Poller. F. (1984). Computer sinmlation as a Lool in teaching park and wildernfssmanagement. y<»wrj)«/ of Environ mtmtiil Education, /5. ! -9.

McCool, S., Lime, D., & Anderson, D. (1977). Sinuilation inodcling as a tool for managing riverrecreation. RiTer mrealioti manngemeiil and research syinposinm procpedings ((icncral TechnicalReport NC-28), USDA Forest Service, North Cenlral Research Station.

Natitjnal Park Service (1997). VEEP: The visitor experience and resource protection (VERf) framework:A handhook for planners and managers. USDI National Park Service Technical Report.

Potter, E, Sc Manning, R. (1984). Application ofthe wilderness travel simulaiion model to theAppalachian Trail in Vermont. Environmental Management. 8, 543-530.

Sargent, R. (1998). Verification and validation of simulation models. Proceeding ofthe 1998 WinletSimulation Conference, 121-130.

Shechter, M. & I,iiLa.s, R. (1978). Simutation of recreational use for park and wddemess management.Washington, DX'..: Resource.s for the Future.

Smith, K., & Krutilla, J. (1976). Structure and properties of a witdemess travel simulator. Baltimore,MD: Johns Hopkin^i University Press.

Stankey, G.. Cole, D., Lucas, R., Peterson, G., Erisst-ll, S. & Wa-shhume, R. (1985). The limits ofacceptable change (/-AQ .system far witdemess planning (General lechnical Report INT-176):USDA Eoresl Service, IniermotniLain Research Station.

Underhill, A., Xaba, A.. & Borkan, R. (1986). The wilderness use simulation inudel applied toColorado River boating in tlraiul Canyon National Park, USA. Environmenlcd Management,10. 367-374.

United States Forest Service. (20(M). Wildenics.s recreation study: Appendix I and Appendix N.John Muir/Amet Adams Cjf Dinkey Lakes Wilderness Management Ptan Final Environment ImpactStatement.

Van Wagtendonk, |. (2003). The wilderness simulation model: A historical perspective. Interna-tionat Joumat of Wilderness, 9(2), '.l-l.'i.

Van Wagtendonk, ]. 8c C^oho, P. (1986). Trailhead <|uotas: Rationing tise lo keep wilderness wild.Joumat iif Forestry, S4, 22-24.

Wang, B., &c Manning, R. (1!)9*)). Computer simtilation modeling for recreation management:A study on carriage road use in Acadia National Park, Maine, USA. Envinmrtumtat Manage-ment, 23, 193-203.