a sgfi , A a“ . .. u‘imrfigw ‘. h...» x _ 42..» a . d ,wrm mums | 3:05 I LIBRARY Michigan State University This is to certify that the dissertation entitled CHARACTERIZING AND MODELING THE RECREATION USE OF DISTANCE SEGMENTED USDA FOREST SERVICE VISITORS presented by Eric M. White has been accepted towards fulfillment of the requirements for the Doctoral degree in Forestry flaw/U- MajGEB‘ofessor’s Signature 4i»?1 03+ 17,, 2005’ Date MSU is an Affirmative Action/Equal Opportunity Institution PLACE IN RETURN BOX to remove this checkout from your record. TO AVOID FINES return on or before date due. MAY BE RECALLED with earlier due date if requested. DATE DUE DATE DUE DATE DUE MAY 2 7 2009 120908 m CHARACTERIZING AND MODELING THE RECREATION USE OF DISTANCE SEGMENTED USDA FOREST SERVICE VISITORS By Eric M. White A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Forestry 2005 ABSTRACT CHARACTERIZIN G AND MODELING THE RECREATION USE OF DISTANCE SEGMENTED USDA FOREST SERVICE VISITORS By Eric M. White The recreation behavior, consumption patterns, and activity participation of visitors to lands managed by the USDA Forest Service (USDA FS) is highly variable. To adequately manage and plan for recreation at the local level, USDA FS natural resource managers must identify the types and quantity of recreation use at individual national forests. This study presents an approach to segmenting and modeling the recreation use of national forest visitors that informs recreation management and planning decisions. Under the adopted segmentation framework, national forest visitors are classified into distance-based visitor segments based upon the proximity of their home to national forest visited. Three distance segments are recognized: Local, Mid-distance, and Long-distance. Local visitors live very close to the national forest, Mid-distance visitors live within a moderate drive of the forest resource, and Long-distance visitors live in the “rest of the world”. Using visitor survey data obtained for USDA FS Regions 2 and 9 via the National Visitor Use Monitoring (NVUM) project, visitors in the three segments are characterized in terms of their recreation behavior, consumption patterns, and activity participation. Statistical tests are completed to determine differences in visitor characteristics both between study regions and between the visitor segments themselves. Few statistical differences are found between study region after accounting for differences due to visitor segmentation and trip type. Capitalizing on the segmentation framework, recreation use models are developed to predict the forest-level recreation use of Local and Mid-distance recreation segments. Models of Local segment recreation use predict visitation based upon local population counts, participation rates and annual visit frequencies. The recreation use of Mid-distance visitors is modeled via multi-site zonal travel cost models. Separate zonal travel cost models were estimated for Mid-distance day trips and Mid-distance overnight trips. While the parameters and coefficients of the constructed models were consistent with theory, evaluation of model prediction proved inconclusive. ACKNOWLEDGEMENTS This dissertation could not have been completed without the support of many individuals. First and foremost I would like to thank Larry Leefers for his support and guidance as my committee chair. I have greatly enjoyed having him as a mentor and friend. I thank him for his dedication and hard work in assisting me through the PhD process. I would also like to thank the members of my committee, Dan Stynes, Steve Friedman, Charles Nelson, and Don English, for their assistance. I owe a special debt of gratitude to Daniel Stynes for hiring me on as a graduate assistant—I very much enjoyed working with you. My graduate assistantship was funded with support from the USDA Forest Service. I thank Ross Arnold, USDA FS Washington Office, for his continuous efforts to keep the money flowing. I also thank Mike Vasievich for providing the initial link between the NVUM project and MSU. The MSU Department of Forestry and the MSU Graduate School also provided funding at points throughout my program. I am also grateful for the assistance provided by Don English, Stan Zamoch and Sue Kocis as I waded my way through the N VUM data and process. A special thanks to Greg Alward and Susan Winter for their assistance throughout my time here at MSU. I thank my family for their support and love, in this and in all things. Last, but certainly not least, I thank Anita Morzillo. I look forward with great anticipation to our adventures and our long life together. iv TABLE OF CONTENTS List of Tables ..................................................................................... viii List of Figures ..................................................................................... xii Chapter 1. Introduction ............................................................................ 1 Research Problem ......................................................................... 3 Research Objectives ...................................................................... 9 Organization of the Dissertation ....................................................... 10 Chapter 2. Literature Review .................................................................... 12 Demand and Use Estimation in Recreation Planning ............................... 13 Methods to Estimate Recreation Demand and Use...... ....15 Structural Models of Recreation Demand and Use .................................. 17 Population-Specific Models ............................................................ 19 Travel Cost Models ...................................................................... 20 Early Travel Cost Applications ......................................................... 22 Contemporary Zonal Travel Cost Models ............................................ 24 Contemporary Recreation Models Incorporating GIS .............................. 26 Recreation Segmentation. .' ............................................................. 29 Relationship to the Current Study ...................................................... 33 Conclusion ................................................................................ 37 Chapter 3. Methods .................................................................................. 38 Conceptual Discussion of Distance-based Visitor Segmentation ................... 38 National Visitor Use Monitoring Project ............................................. 46 Cl. -lL NVUM Visitor Survey ......................................................... 48 Data Weighting Schemes ...................................................... 49 Study Area ................................................................................ 51 NVUM Survey Data ..................................................................... 53 Distance Segmentation of NVUM Survey Respondents ........................... 56 Characterizing Distance-based Segments ............................................... 59 Modeling Segment Visitation .......................................................... 61 General Description ............................................................ 62 Local Visitor Recreation Use Model ......................................... 68 Mid-distance Visitor Recreation Use Models ............................... 74 Evaluation of Model Predictive Ability ............................................... 80 Chapter 4. Results and Discussion ............................................................... 88 Introduction ............................................................................... 88 Characteristics of Distance Segmented Visitors ...................................... 88 Segmentation of NVUM Survey Respondents .............................. 88 Regional Comparisons within Segments ..................................... 90 Distance Segment Comparisons .............................................. 98 Summary of Distance Segment Characteristics ........................... 111 Recreation Use Models ................................................................ 112 Local Permanent Resident Model ........................................... 113 Local Seasonal Resident Model ............................................. 123 Mid—distance Recreation Use Models ....................................... 127 Evaluation of Combined Model Predictions ............................... 136 vi Model Application to Out-of—Sample National Forests ................... 139 Discussion of Model Results .................................................. 141 Chapter 5. Summary, Conclusions, and Policy Implications ............................... 146 Introduction .............................................................................. 146 Summary ................................................................................. 146 Conclusions .............................................................................. 152 Policy Implications ..................................................................... 154 Limitations and Recommendations for Future Research .......................... 157 Appendix A ....................................................................................... 162 Introduction .............................................................................. 163 Proxy Sites ............................................................................... 163 Sample Day Selection .................................................................. 164 Survey Revisions ....................................................................... 165 NVUM Visit Estimation ............................................................... 167 Appendix B ....................................................................................... 195 Appendix C ....................................................................................... 207 Literature Cited ................................................................................... 167 vii LIST OF TABLES Table 1. Postulated Characteristics of Local, Mid-distance, and Long-distance Visitors .................................................................... 42 Table 2. Administrative National Forests (NF) within USDA FS Regions 2 and 9 ...... 51 Table 3. NVUM Survey Variables ............................................................... 55 Table 4. Distance bands used in Local and Mid-distance recreation use models ......... 64 Table 5. Forest aggregation for estimating AVij .............................................................. 71 Table 6. Spatial data sources for publicly-owned land ........................................ 79 Table 7. NVUM Estimates of Local Visitor Recreation Use for USDA FS Regions and National Forests (NF) in the Study Area .......................................... 83 Table 8. NVUM Estimates of Mid-distance Visitor Use for Regions and National Forests (NF) in the Study Area ............................................................................ 85 Table 9. NVUM Estimates of Summed Local and Mid-distance Recreation Use .......... 86 Table 10. NVUM Estimates of the Percentage of Local Visitor use as a Function of Summed Local and Mid-distance Use ................................................... 87 Table 11. Number of NVUM Survey Respondents by Distance Segment and USDA FS Study Region ................................................................. 89 Table 12. Percent of Total Recreation Use by Distance Segment and USDA FS Study Region ................................................................. 90 Table 13. Trip-type Segment Shares by Distance Segment and USDA FS Study Region .................................................................. 92 Table 14. Mean Values for Visitor Characteristics of Interest by distance Segment, Trip-type, and USDA FS Region ........................................... 94 Table 15. Day of National Forest Arrival by Distance Segment, Trip-type, and USDA FS Region .................................................................... 95 Table 16. Primary Activity by Distance Segment and USDA FS Region .................. 98 viii Table 17. Mean Values for Visitor Characteristics of Interest by Distance Segment, and USDA FS Region ........................................................ 99 Table 18. Day of National Forest Arrival by Distance Segment and USDA FS Region ........................................................................ 107 Table 19. Primary Activity by Distance Segment and USDA FS Region ................. 108 Table 20. Trip-type Segment Shares by Distance Segment and USDA FS Study Region .............................................................................. 109 Table 21. Mean Values for Visitor Characteristics of Interest by Distance Segment, Trip-type, and USDA FS Region ............................... 110 Table 22. Regional Participation Rates of Populations Residing Within 30 Miles of National Forests in the USDA FS Study Regions .................... 114 Table 23. Forest-level Distance Band Populations for NVUM forests sampled in FY2001, FY2002, FY2003 ............................................... 115 Table 24. Distance Band Annual Visit Frequencies of Local Visitors by Region and National Forest Aggregation Group .................................... 116 Table 25. Model and NVUM Estimates of Permanent Resident Local Visitor Recreation Use .................................................................. 1 18 Table 26. Local Visitor Model Errors ......................................................... 119 Table 27. Local visitor Segment Shares Estimated Under Two Alternative Weighting Schemes ....................................................... 123 Table 28. Seasonal Homes Proximate to NVUM forests sampled in FY 2001, FY2002, and FY2003 ....................................................... 124 Table 29. Model Estimates of Seasonal Local Resident Recreation Use. . . . . . 126 Table 30. Model and NV UM Estimates of Seasonal Resident Local Visitor Recreation Use for FY 2003 NVUM National Forests ............................. 127 Table 31. Day and Overnight Mid-distance Segment Recreation Use Models, USDA FS Region 2 ........................................................... 130 Table 32. Day and Overnight Mid-distance Segment Recreation Use Models, USDA FS Region 9 ........................................................... 132 ix Table 33. Model Predictions and NVUM Estimates of Mid-distance Segment Recreation Use, USDA FS Region 2 ...................................... 133 Table 34. Model Predictions and NVUM Estimates of Mid-distance Segment Recreation Use, USDA FS Region 9 ...................................... 134 Table 35. Mid-distance Visitor Model Errors ................................................ 135 Table 36. Model Predictions and NVUM estimates of Local and Mid-distance Recreation Use ......................................................... 137 Table 37. Local Percentage of Local and Mid-distance Segment Recreation Use based upon Model Prediction and NVUM Estimates ......................... 138 Table 38. Model and NVUM Estimates of Permanent Resident Local Visitor Recreation Use for Out-of-Sample Forests ................................. 139 Table 39. Model Predictions and NVUM Estimates of Mid-distance Segment Recreation Use for Out-of—Sample Forests ............................... 140 Table 40. Local Percentage of Local and Mid-distance Segment Recreation Use based upon Model Prediction and NV UM Estimates, Out-of-Sample Forests ................................................................. 141 Table B-1. NVUM Estimates Recreation Use by Forest and Distance Bands from 30 — 200 miles ................................................................... 174 Table B-2. NVUM Estimates of the Percentages of Day and Overnight Trips by 30 — 200 Mile Distance Band for National Forests in Region 2. . . . . . . . .....175 Table B-3. NVUM Estimates of the Percentages of Day and Overnight Trips by 30 - 200 Mile Distance Band for National Forests in Region 9 ............ 176 Table B-4. NVUM Estimates of Day and Overnight Recreation Use by Distance Band for National Forests in Region 2 ................................. 177 Table B-5. NVUM Estimates of Day and Overnight Recreation Use by Distance Band for National Forests in Region 9 ................................. 178 Table B-6. Modeled Recreation Use Estimates and Confidence Intervals for National Forests in the Study Area Sampled in FY 2001, FY2002, FY2003 ......................................................... 179 Table B-7. Local Use Percentages of Modeled Recreation Use for National Forests (NF) in the Study Area .......................................... 180 Table B-8. Mid-distance Use Percentages and Primary Purpose Percentages of Modeled Recreation Use for National Forests (NF) in the Study Area. 181 Table C-l. Standard Errors of Annual Visit Frequency Estimates by Region and National Forest Aggregation Group ............................... 183 Table C-2. Multi-site Zonal Travel Cost Model Inputs, USDA FS Region 2 ...... 184 Table C-3. Multi-site Zonal Travel Cost Model Inputs, USDA FS Region 9 ...... 193 xi LIST OF FIGURES Figure 1. USDA FS Regions 2 and 9 ....................................................... Figure 2. Distance Segmentation of USDA FS Visitors .................................. Figure 3. Cumulative Percent of Visitation Across Visitor Origin Distance Bands...... Figure 4. River and lake barriers to travel in the study area .............................. Figure 5. Euclidean and Barrier-adjusted Distance Estimation .......................... Figure 6. Conceptualization of quantifying a forest’s local population ................ Figure 7. Percent of housing units classified as seasonal homes in Census 2000 for census block groups within 30 miles of national forests in USDA FS regions 2 and 9 ........................................................................ Figure 8. Distance Segment Visit Frequency, USDA FS Region 2 (Truncated at 12 Annual Visits, Exposure Weighted) ............................. Figure 9. Distance Segment Visit Frequency, USDA FS Region 9 (Truncated at 12 Annual Visits, Exposure Weighted) ........................... Figure 10. Duration of National Forest Visits by Distance Segments, USDA FS Region 2 (Rounded to Nearest Hour, Truncated at 24 hours, ExposureWtd.) ............................................. Figure 11. Duration of National Forest Visits by Distance Segments, USDA FS Region 9 (Rounded to Nearest Hour, Truncated at 24 hours, ExposureWtd.) ............................................. Figure 12. Number of sites (and GFA days) by Distance Visitor Segment, USDA FS Region 2 (Figure Truncated at 5 Sites, Exposure Weighted) ...... Figure 13. Number of sites (and GFA days) by Distance Visitor Segment, USDA FS Region 9 (Figure Truncated at 5 Sites, Exposure Weighted). . . . Figure 14. Number of Activities by Distance Segmented, USDA FS Region 2 (Figure Truncated at 10 Activities, Exposure Weighted) ............ Figure 15. Number of Activities by Distance Segmented, USDA FS Region 9 (Figure Truncated at 10 Activities, Exposure Weighted). . . . . . . . . xii ..... 9 ....39 .58 ....66 ....67 ....69 ....72 ..100 ...100 ..101 ..102 ...103 103 ..104 105 Figure 16. People per Vehicle by Distance Segment, USDA FS Region 2 (Exposure Weighted) ........................................................ 106 Figure 17. People per Vehicle by Distance Segment, USDA FS Region 2 (Exposure Weighted) ........................................................ 106 xiii CHAPTER 1 INTRODUCTION Recreation, encompassing a wide array of activities, is a fundamental component of American life. While many recreation pursuits take place indoors, recreation in the outdoors is prominent. It is estimated that 98% of people in the US, 16 years of age or older, participate in some form of outdoor recreation activity in a given year (Interagency Survey Consortium, 2002). The rate of participation in outdoor recreation activity has remained steady to slightly increasing since the early 1960’s (Outdoor Recreation Resources Review Commission, 1962; National Park Service, 1986; Interagency Survey Consortium, 1995; Interagency Survey Consortium, 2002). Individuals in the US. are most likely to participate in relatively passive outdoor recreation activities such as walking (82%) or taking part in family gatherings (74%). More active outdoor recreation pursuits such as developed camping (26%), mountain biking (21%), or hunting ( 11%) are undertaken with less propensity (Interagency Survey Consortium, 2002). While privately-owned lands also provide recreation opportunities, publicly- owned lands have long served a central role in the provision of places for people to recreate. The basis for public entry to lands for the purpose of recreation was established by the “fishing and fowling” laws enacted in the colonial period (Douglass, 1999). The establishment of public recreation areas began in 1710 with the city forest of Philadelphia and continued in the mid 1800’s with the work of Frederick Law Olmstead and the advent of city parks (Douglass, 1999). Public recreation on federally-owned lands in the US. began in the latter half of the 19‘h century and expanded rapidly in the early 20th century; concurrent with the expansion of the forest reserves managed by the then US. Forest Service and with the establishment of the National Park Service. In the early 21St century, the role of publicly—owned lands as the primary provider of outdoor recreation opportunities is firmly cemented. Wellman and Propst (2004) state: “. . .these (lands) have allowed many American to participate in forms of recreation that honor their pioneering heritage, respond to their desire to take risks, allow them temporarily to move from their complex everyday environments into quieter and greener places, and offer special opportunities for them to discover things about themselves and their surroundings”. Recreation opportunities are provided on lands owned by all levels of government. City- and county-owned lands may support a limited number of outdoor recreation activities (e.g. walking, picnicking, etc.), while state- and federally-owned land areas may provide opportunities for a greater variety of pursuits (e. g. picnicking, camping, rock climbing, hunting, etc.). Federally, the USDA Forest Service (USDA PS), the National Park Service (NPS), the Bureau of Land Management (BLM) and the US. Army Corps of Engineers (ACOE), among others, manage land for public recreation. In 2003, NPS sites across the nation received an estimated 266 million visits (NPS, Public Use Statistics Office, 2004). The ACOE receives approximately 385 million visits annually at its managed recreation areas (U .S. ACOE, 2004). Annual visitation to USDA Forest Service lands is estimated to be 205 million visits (English, Pers. Comm.) while the BLM reports approximately 61 million annual visits (BLM, 1999). The NPS manages 95 million acres of land and the ACOE manages 12 million acres of land and water. Both agencies manage lands located throughout the nation. The USDA FS and BLM manage much larger areas than the other two agencies, 191 million acres and 262 million acres, respectively. Lands managed by the BLM are located almost exclusively in the western contiguous US. and Alaska while USDA FS land is concentrated in the contiguous west, but located throughout the nation. With extensive land holdings, accessibility to much of the US. population, a variety of natural features, and high rates of visitation, the USDA FS is an indispensable provider of recreation opportunity in the US. and is the focus of the research presented here. Research Problem Characterizing recreation visitors and quantifying recreation use are central to USDA FS planning activities (Dana, 1957; Propst, 1985; Alig and Voss, 1995). Characterizing recreation visitors in regard to their consumption patterns and preferences assists in identifying the types of recreation opportunities that may best meet demand, determining what recreation facilities to develop and where, and selecting specific “on the ground” management actions. Reliable estimates of visitation are useful for forest plan revision, for completing economic impact analysis, in developing estimates of social benefits, and for completing benefit-cost analyses. Visitor characterization and estimation of recreation use are often completed concurrently and can be achieved via three approaches: conducting a complete census of visitors, implementing a visitor sampling program, or developing and applying recreation demand models. A census of all visitors to USDA FS lands is generally accepted as impractical and unnecessary. Attempts to determine recreation use via a census in the fledgling years of the USDA FS proved mostly futile (Waugh, 1918). The dispersed nature of recreation and the expansive land area, accessible at a multitude of points, precludes the counting of every USDA FS recreation visitor. Furthermore, the reliability of visitor sampling approaches and recreation modeling makes a census unnecessary. The USDA FS has implemented a number of projects aimed at estimating recreation use and visitors characteristics via visitor sampling. Recent projects of note include PARVS, CUSTOMER, and National Visitor Use Monitoring (NVUM). PARVS and CUSTOMER were initiated in the latter half of the 1980’s while NVUM began in the year 2000 and continues today. While PARVS and CUSTOMER surveys provided useful information, the sample sizes were small, implementation was not standardized, and the visitor samples were not representative of all USDA FS visitors (Alward et al., 1998). In response to these problems, USDA FS scientists and analysts developed a national-level, standardized method for quantifying recreation use and visitor characteristics called National Visitor Use Monitoring (English et al., 2002). Through NVUM the USDA FS develops national, regional, and forest-level estimates of visitor use and characteristics. Models of recreation use can be used to characterize and quantify recreation use under both current as well as alternative conditions and can provide analysts with a greater understanding of the processes behind recreation use patterns and behavior. Grant and others (1997) identified five general applications of scientific models: 1) providing a sound conceptual framework for future research, 2) evaluating alternative hypotheses about system structure or function, 3) describing system behavior under normal conditions, 4) predicting system response to specific management schemes or environmental situations, and 5) heuristically exploring the dynamics of a system of interest. A limited number of recreation models have been completed specifically for resources managed by the USDA FS (e.g. English and Bowker, 1996; English and Home, 1996; Alig and Voss, 1997; Loomis et al., 2001; Betz et al., 2003). Of these, only models developed by English and Home (1996) were used to estimate recreation use for multiple national forests. The remaining models were aimed at quantifying recreation use for individual sites or for specific activities occurring on individual national forests. For the purpose of planning and management, USDA FS recreation use has typically been characterized (or segmented) and quantified based upon the primary activities of recreation visitors (USDA FS recreation is also commonly characterized and quantified by Recreation Opportunity Spectrum, ROS, class). This primary activity approach was used in the first census of visitors in 1916 as well as the PARVS surveys of the 1980’s (Waugh, 1918; Alward et al., 1998). While this approach quantifies the types of activities users are engaged in it may not provide particularly informative or reliable information concerning behavior and/or patterns of recreation use. Specifically, the behavior and use patterns of visitors within activity segments may be just as variable as those of the visitor population at large. An alternative approach to segmenting recreation visitors is one based upon the interceding distance between visitor residence and the forest resource. Specifically, does the visitor live very close to the forest (a local visitor), within a moderate drive from the forest (a mid-distance visitor), or several hundred miles or more from the forest (a long- distance visitor)? A limited number of recreation studies have shown that the intervening distance from the visitor’s residence to the recreation destination may influence recreation visitor consumption pattern, activity participation, recreation behavior, and trip motivation. Strauss and others (1993) found the proximity of hunters to the hunting destination influenced their pattern of recreation consumption. Hunters at the Delaware Water Gap who lived outside the local area were found to hunt almost exclusively on Saturdays, while those from the local area hunted on days throughout the week. Though such a pattern is intuitive, the consumption patterns of hunters are frequently characterized without regard to hunter residence. Recreation participation by certain demographic groups may be more likely given their proximity to the resource. Faunce and others (1979), in their study of Maine hunters, found that participation of women hunters was more common among resident Maine hunters than among non-resident Maine hunters. This difference likely relates to differences in hunter motivation between those living in the State and those living out-of-state. In regards to recreation behavior, Stynes and White (2003) have shown that the expenditures of USDA FS recreation visitors who live in the local forest area are almost always less than that of non-local users within trip-type segments, regardless of recreation activity. Finally, Etzel and Woodside (1983) found that visitors who recreated near home were motivated more by opportunity for relaxation and recuperation while visitors recreating farther from home expected more stimulation and entertainment in their recreation experience. Additionally, Faunce and others (1979) found that non- resident Maine hunters were more frequently motivated by the social interaction component of hunting than resident hunters. Segmenting recreation users based upon the distance from their home to the recreation resource may provide for a greater understanding of consumption patterns, activity participation, and recreation behavior of USDA FS visitors. This may lead to more informed decisions by resource managers regarding the provision and management of recreation opportunities. Assuming that visitors within distinct distance segments have unique recreation behaviors, activity participation, and consumption patterns, quantifying forest-level recreation use of the segments would allow individual forests to identify their “recreation markets”. Recreation models can be used to determine the expected forest-level recreation use of the distance-based segments under current and alternate conditions. In addition to predicting use, recreation models developed for individual distance segments can provide insight into the underlying mechanisms influencing recreation use of visitors within those distance segments. An understanding of the mechanisms influencing recreation use of distance segment visitors can aid resource planners and managers in identifying factors that influence the level of recreation use at an individual national forest. The USDA FS has traditionally taken a regional approach to planning for and managing actual recreation use (e. g. 1990 RPA assessment regional recreation use values). The assumption of regional recreation differences may result, in part, from the perception that regional variation in the types of natural features managed yields differences in recreation behavior and participation. Given this, testing for regional differences in consumption patterns, activity participation, and recreation behavior of distance segmented recreation visitors between distinct USDA FS regions is one component of this study. USDA FS regions 2 and 9 are particularly appropriate for comparative analysis (Figure 1). National forests within these regions are spatially disjoint from one another and the natural features managed are, in general, quite different. . l uswa Region 2] E] National Forests Figure 1. USDA FS Regions 2 and EIRoaionaI Boundariu Research Objectives Given the above, the general objective of this study is to identify the characteristics and model the forest-level recreation use of distance-segmented USDA FS recreation visitors. The four specific objectives of the study are to: 1) identify the recreation behavior, activity participation, and consumption patterns of Local, Mid-distance, and Long-distance USDA FS visitors in USDA FS Regions 2 and 9, 2) statistically compare the recreation behavior, activity participation, and consumption patterns of visitors to USDA FS Regions 2 and 9 within the distance-based segments, 3) statistically compare the recreation behavior, activity participation, and consumption patterns of Local, Mid-distance, and Long-distance visitors within USDA FS Regions 2 and 9, and 4) model forest-level recreation use of Local and Mid-distance recreation visitors for national forests located in USDA FS Regions 2 and 9. Organization of the Dissertation Chapter 2, Literature Review, is an examination of I) demand and use estimation in forest recreation planning, 2) broad approaches employed to estimate recreation demand and use 3) use of structural recreation demand and use models and 4) the role of recreation segmentation in recreation planning. Also included are discussions of the theoretical basis and evolution of travel cost modeling, descriptions of contemporary applications of recreation modeling utilizing spatial analytical techniques, and identification of commonly employed recreation segmentation approaches. Chapter 3, Methods, provides a discussion of the data and analytical techniques used in achieving the research objectives identified above. The chapter begins with a conceptual discussion of the distance-based segments and the relationship of the segmentation to the recreation literature. A short description of the NVUM project, survey data and recreation use estimates follows. Further detail on the NVUM sampling scheme and the NVUM procedures for estimating recreation use can be found in Appendix A. The remainder of Chapter 3 is devoted to describing the specific procedures 10 used in analyses. This section begins with delineation of the boundaries separating the distance segments from one another. The approaches to characterizing visitors within the distance segments and the statistical tests employed to identify differences between the study regions and between the distance segments are then identified. The final portion of the Chapter relates to the development and evaluation of the Local and Mid-distance recreation use models. Chapter 4, Results and Discussion, is divided into two halves; the first relates to objectives 1 — 3 while the second relates to Objective 4. The first half of the Chapter identifies the characteristics of the distance-based visitor segments as well as results of statistical comparisons both between study regions and between distance segments. In the second half of the Chapter, the constructed recreation use models are presented, and the parameters and coefficients are discussed. The predictive ability of the recreation use models is evaluated against NVUM recreation use estimates. In Chapter 5, Summary, Conclusion, and Policy Recommendations, a summary of the research undertaken is presented, and conclusions related to the research objectives are identified. Policy implications, in regard to forest planning activities, are presented and discussed. The limitations of the research and future research needs are outlined. 11 CHAPTER 2 LITERATURE REVIEW The impetus for this research is the need to quantify the types and numbers of national forest recreation users for forest-level planning. The role of recreation use estimation in recreation planning activities is identified in the first section of this chapter. The second section includes discussion of how recreation demand is incorporated in the planning process and the approaches to developing recreation use estimates. The use of recreation models, particularly the travel cost model, is described in greater detail. The third section contains a description of traditional USDA FS approaches to segmenting recreation visitors as well as some alternative approaches to visitor segmentation. The chapter closes with a summary of how the literature relates to this study. An influential report that deserves a special introductory note is “Assessing Demand for Outdoor Recreation” published in 1975 by the National Academy of Sciences. This report was written by the Committee on Assessment of Demand for Outdoor Recreation Resources established by the now dissolved Bureau of Outdoor Recreation. The committee was tasked with investigating four specific components of recreation demand analysis: identifying the objectives of demand analysis, reviewing systematic models of demand estimation, specifying the parameters of a recommended demand estimation approach, and identifying the steps of to effectively estimate demand. Two appendices were also included in that report. The first (written by Bev Driver and Perry Brown) examined recreation demand from a social-psychological perspective while 12 cc to 1 mt. Sus incr the second (written by V. Kerry Smith) provided a comprehensive analysis of recreation demand models. The report (with its appendices) is a preeminent piece of recreation demand literature and is cited frequently within this chapter. Demand and Use Estimation in Recreation Planning Public recreation resources are limited. Like all scarce resources, public resources must be allocated between competing demands. Whereas allocation of traded resources can be determined by markets, public recreation resources generally provide non-market public goods and services and, as such, their allocation is not as easily determined. In the course of natural resource planning, the “demand” for recreation resources or activities as well as the benefits or values that those resource or activities provide are frequently quantified. The Renewable Resource Planning Act of 1974 and the National Forest Management Act of 1976 require the USDA PS to complete national-level natural resource assessments every 10 years and to develop individual forest-level plans every 10 to 15 years. Both the national assessment and individual forest plans must incorporate multiple uses (including recreation) of natural resources as mandated by the Multiple Use Sustained Yield Act of 1960. At the state-level, efforts to quantify recreation demand increased dramatically in response to the 1962 report of the Outdoor Recreation Resources Review Commission (ORRRC) and the associated enactment of the Land and Water Conservation Fund. 13 The specific role of quantifying recreation demand for the purpose of planning is in “illuminating and measuring the implications of alternative planning, provision, and management decisions” (NAS, 1975). In particular, quantifying recreation demand provides information for three kinds of planning decisions: policy decisions, allocation decisions, and site-specific resource provision and management. These kinds of planning decisions correspond to three aspects of recreation demand: demand for recreation in the context of broad social and economic policy, demand for alternative types of recreation, and the demand for site-specific recreation resources (NAS, 1975). The first aspect of recreation demand focuses on the importance of outdoor recreation opportunities as a component of overall social structure. Decisions that affect recreation opportunities have an impact on “. . .a range of diverse elements, including economic, industrial, and population growth. ...” (NAS, 1975). The second aspect of demand encompasses those decisions relating to determining the type, quantity, and location of recreation opportunities, and well as the strategy and timing for their implementation (NAS, 1975). Assessment of this aspect of demand requires that the planning agency recognizes what recreation opportunities are desired by users and what management actions are needed to fulfill those desires. The third aspect of demand focuses on choosing the site and selecting the kind and quantity of recreation resources, facilities, and programs provided (N AS, 1975). This can be achieved by identifying potential users, recognizing the alternative sites available to them, and determining the relationship between site characteristics and the user group 14 (NAS, 1975). In all cases, recreation decisions are facilitated by quantifying the expected mix of user segments and the activities and site characteristics that these segments desire (Hendee, 1967; NAS, 1975; Dwyer et al., 1977; Bojanic and Wamick, 1994). While it is mandated that recreation demand be incorporated in planning activities, doing so is made difficult, in part, by disparities in identifying what “recreation demand” represents. In the Preface of the 1975 report of The Committee on Assessment of Demand for Outdoor Recreation Resources (The Committee), the Associate Director of the Bureau of Recreation described the concept of demand as “[o]ne of the most often used, least understood, and most significant concepts in outdoor recreation planning” (NAS, 1975). In the vernacular sense, recreation demand may be conceptualized as the total number of visits to a recreation resource or the expected participation in a recreation activity during a given period. Economists, however, view recreation demand not as one point of consumption but rather as a schedule of expected recreation use given a range of prices or costs (Clawson and Knetsch, 1967). From the perspective of social- psychologists, recreation demand represents the preferences or desires of individuals regardless of whether or not those desires and preferences result in recreation participation (Driver and Brown, 1975). Methods to Estimate Recreation Demand and Use The Committee, adopting a broad definition of demand, outlined four approaches to estimating demand for outdoor recreation: application of standards, projections of use, 15 structural models, and expression of perceived wants (NAS, 1975). Standards, used to determine the “desired” number of recreation opportunities for a given population size or area, have been identified primarily for urban settings. The Committee suggested standards are not particularly effective at indicating recreation demand. In part this is due to the assumptions of applying this technique, namely that social recreation desires, the quality and attributes of recreation sites, and population characteristics are homogenous across areas (NAS, 1975). The second approach, projecting demand, is achieved, generally, by extrapolating historic visitation patterns to some point in the future. This technique intemalizes the relationship between visit volume and price for every point in time. As such, the use of projections “. . .ignores the interaction of social and economic conditions with recreation and the changing determinants of individual recreation decisions, which may, in fact, cause people to behave differently in the future” (NAS, 1975). Structural models (the third approach) estimate demand by parameterization of the relationships between demand/supply of recreation opportunities and the factors that influence demand. This allows the user to explicitly identify the relationships between demand and the factors influencing recreation consumption (N AS, 1975). Structural models allow for demand to be quantified for proposed and existing sites as well as developing demand estimates after proposed quality improvements or degradation at existing sites (NAS, 1975). One of the drawbacks to this method is that while relationships are assumed to be causal they may in fact be spurious. Other limitations to this approach include some difficultly in capturing the motivation of individual 16 recreationists via independent variables and the lack of variation in independent variables—which may lead to poor prediction (NAS, 1975). The final method of estimating demand is accomplished via public input on the number and types of recreation opportunities desired. This can be accomplished through surveys, analysis of public participation in the planning process, and/or examination of operation budgets (NAS, 1975). Structural Models of Recreation Demand and Use Models of recreation demand and use can be developed along a continuum from those that are exceedingly simplistic to those that are very complex. Moeller and Echelberger (1974) stated the selected model “should be dynamic so that changes in management policy, recreation supply, price, etc. can be incorporated into the model to determine their potential impact. . .”. Structural models of recreation may be used to estimate recreation use given supply and demand factors (e. g. Walsh et al., 1992; Hanink and Stutts, 2002), to identify the use value of recreation resources (e. g. Bowker et al., 1996; Fix and Loomis, 1997), or to accomplish both (Boxall et al., 1996; Loomis, 2002). Smith (1975) identified three broad classes of models of recreation demand and use: site- specific user models, population-specific models, and site-specific area models. Site-specific user models depend upon data collected from individual site users. This data may include their motivations, attitudes, or satisfaction levels (Smith, 1975). Models of this type do not assume homogeneity in services consumed from the site to 17 recreation users. On the contrary, “(these) studies attempt to determine what individuals reactions to heterogeneous services will be” (Smith, 1975). Fix and Loomis (1997), Provencher and Bishop, (1997), and Loomis and others (2001) adopted the site-specific user approach to modeling recreation behavior. Population-specific models differ from the other two model types in that information is collected from both participants and non-participants (Smith, 1975). These types of models generally collect information on demand for particular services or activities (Smith, 1975). From the information collected, participation (or participation rate) for specific activities (or in some cases sites) can be determined for groups of users (Smith, 1975). Smith suggests that two problems exist with this approach. First, the observations (survey responses) are “. . .the result of the individual’s own demand and effective supply”. The implication being that the model output incorporates many different individual demand and supply conditions to estimate a single demand relationship. Second, for the purpose of benefit estimation, this procedure may determine average expenditure per day while the demand curve operates on the basis of marginal costs of production. Using this approach results in inappropriate calculations of benefit (Smith, 1975). Population-specific models include those of Cicchetti (1973), Walsh and others (1988), Bowker and others (1999), and Romano and others (2000). Site-specific area models are also developed from information gathered from site users (Smith, 1975). Most important to the use of this approach is identification of the origin of the visitor. Travel cost models, as described in Clawson and Knetsch (1967), 18 fall into this class of demand models. In general, these models relate the number of recreation visits over a range of distances that recreationists travel (representative of cost from an economic standpoint) to develop the recreation demand curve and expected use for the site. Since the data collected in this method is aggregate in nature, the characteristics of individual users are not identified; rather they are assumed to be homogenous within zones. Models by Hellerstein (1991), Loomis and others (1995), and Cho and others (2001) are contemporary examples of this approach. Population-Specific Models Cicchetti (1973) introduced a two-step population-specific model to forecast recreation use for specific recreation activities. Estimation of the model required survey data collected from both participants and non-participants. The first stage of the model predicted the probability that an individual (or individuals) would participate in a given recreation activity while the second stage of the model predicted the expected number recreation occurrences for a participating individual in a given year. The total recreation use in any one activity is obtained by combining the number of participants with the expected number of recreation occurrences per participant. Socio-economic, demographic, and recreation supply variables were included as independent variables in both model stages. Using estimates of future population counts and demographic and socio-economic conditions, Cicchettii (1973) completed several example applications to forecast future recreation use for a variety of water-based and land-based recreation activities. 19 Travel Cost Models In their seminal text Clawson and Knetsch (1967) discussed at length the concept of estimating recreation demand curves, recreation use, and recreation benefit, through the use of travel cost models. Despite the prominence of the text, the authors deferred to Hotelling (1949) and Van Doren (1960) as providing the initial impetus to using travel cost to estimate recreation demand curves. Using distance traveled as a proxy for cost, Clawson and Knetsch (1967) constructed recreation site demand schedules for a variety of recreation areas. In general, origins in close proximity to the recreation site (with associated low travel cost) contribute a greater number of visits to the sites (per capita) than origins at a greater distance from the recreation site (higher travel cost). Total economic benefit to society is calculated as the area under the second stage demand curve (Clawson and Knetsch, 1967). The second stage curve represents what the demand schedule would be under increasing price. Given this, total consumer surplus can be calculated as the entire area under the demand curve. The travel cost approach to demand and benefit estimation, as outlined by Clawson and Knetsch (1967) and further developed by Cicchetti and others (1976) and Cesario and Knetsch (1976), is best applied under the following conditions: 1) travel costs are variable among users, 2) proposed changes are large enough to alter travel costs to users, and 3) travel costs are primarily associated with the recreation site under study (Dwyer et al., 1977). Travel cost modeling is not particularly suited for sites that attract u331‘s primarily from areas in close proximity to the site, for sites that attract many pass- 20 thru or non-primary users, or for sites that are exceptionally large in size with multiple entrances (Dwyer et al., 1977). Several assumptions are required when estimating demand and benefits. Specifically, recreation visits do not include trips to other recreation destinations, the travel portion of the trip does not provide any benefit to the visitor, trips are of uniform duration, and individuals have similar travel patterns and travel means (Loomis and Walsh, 1997). Given the conditions and assumptions, travel cost models are better suited to modeling the recreation use of visitors who visit one recreation resource and who travel a moderate distance to do so. The recreation use of those visitors who live in close proximity to the recreation resource or those who visit the resource as part of a multi-purpose trip is better modeled using a different approach. Travel cost models can be constructed for single sites (simple travel cost models) or for multiple sites within a region (regional multi-site travel cost models) (Dwyer et al., 1977; Loomis and Walsh, 1997). The regional travel cost model includes multiple recreation sites (frequently) of varying quality so the impacts of potential recreation quality changes on an individual site can be identified. Dwyer and others (1977) stated that regional travel cost models can be of two primary forms; specifically, a system of linear demand equations or a single demand equation incorporating a gravity model. Regardless of form the general procedure to construct a travel cost model is as follows (Dwyer et al., 1977): 1) Complete a survey of visitors or households to collect visit, trip, and demographic information, 21 2) Classify sites based upon their amenities and the recreation opportunities provided, 3) Define the origin unit of analysis, 4) Estimate travel cost on a round trip basis, 5) Determine round-trip travel time, 6) Identify substitute recreation destinations, 7) Derive socio-economic variables of interest from necessary data, and 8) Estimate the demand model using the appropriate functional form. Early Travel Cost Applications The initial applications of the method outlined by Clawson and Knetsch (1967) laid the groundwork for the development of the zonal travel cost model that we know today. Three examples of these early applications are Burt and Brewer (1971), Cesario and Knetsch (1976), and Cicchetti and others (1976). Burt and Brewer (1971) presented a conceptual model for deriving the social benefit of additions to an existing set of outdoor recreation sites. This is an extension (though the estimation of benefit is approached from a somewhat different perspective) of Clawson and Knetsch’s (1967) model of a single recreation site. The authors subsequently applied their model to estimate the demand and benefit of proposed ACOE reservoirs in Missouri using a household survey and a system of demand equations where the explanatory variables are travel price from visitor origins to water-resource 22 destinations and income. For the three proposed reservoirs the authors estimated an annual visitation of 1.1 million household visit days with an associated annual benefit of $8.5 million dollars. Cicchetti and others (1976) extended the work of Clawson and Knetsch (1967) to explicitly incorporate the availability of substitute recreation destinations in the calculation of social benefit. The authors applied their model to a set of ski areas in southern California to determine the net social benefit associated with development of a proposed ski area. The authors used a system of demand equations incorporating travel cost to develop an estimate of the social benefit of the proposed area. Comparing the total social benefit of the development with the costs (and using appropriate discounting), the authors suggested the proposed development had a negative net social benefit. In contrast to the previous example Cesario and Knetsch (1976) used a single demand equation to estimate their travel cost model. The traditional travel cost model proposed in Clawson and Knetsch (1967) did not include the use of a travel time variable in predicting demand. This was due in part to the multicollinearity of travel cost and travel time. Cesario and Knetsch (1976) suggested failing to incorporate travel time would lead to overestimation of consumption from users at distant origins. The use of these long-distance visitors would be overestimated as a result of failing to account for the greater cost (value) of the increased time required to reach the recreation site. To correct this overestimation, the authors constructed a composite variable representing 23 both travel costs and time which was then included in the model of demand. The model was found to perform well when applied to a set of Pennsylvania State Parks. Contemporary Zonal Travel Cost Models In their 1986 article Ward and Loomis completed a comprehensive overview of the evolution of travel cost modeling from the late 1970’s to the late 1980’s. The authors described three empirical forms of travel cost modeling that continued to evolve during the period: zonal, individual, and hybrid. The zonal approach to travel cost modeling is described above. To overcome some of the limitations of the zonal technique the individual travel cost model was developed. In this formulation of the travel cost model, individual specific travel times, travel costs, and socioeconomic characteristics are identified. Travel cost models are developed from this information for individuals (rather than zones) where the dependent variable is the individual’s trips (rather than zonal per capita trips). The hybrid travel cost model incorporates portions of both the individual and zonal travel cost models. Hybrid models employ nested decision trees to estimate individual demand models (Ward and Loomis, 1986). These models frequently take the form of multi-nomial logit models (e. g. Provencher and Bishop, 1997). Contemporary examples of zonal travel cost models include Hellerstein (1991), Loomis and others (1995), and Bowker and English (1996). Hellerstein (1991) used permit count data to construct a model of demand for the Boundary Waters Canoe Area (BWCA). The model predicted the per capita visitation from counties surrounding 24 BWCA. To control for the constrained nature of the dependent variable (i.e. visitation to BWCA from countyi) both Poisson and Negative Binomial models were used. For comparison, a semi-log model was specified and fitted using ordinary least squares and a bias correction recommended by Stynes and others (1986). The author found the Poisson model estimated via pseudo-maximum likelihood performed best. Model selection had little influence on the coefficient estimates but had significant influence on the coefficient standard errors. Loomis and others (1995) developed and tested the transferability of demand models between ACOE districts. The authors constructed demand models for day users and campers for three ACOE districts. The transferability of these models between ACOE districts was subsequently tested. Explanatory variables included visitor demographics (aggregated at the county level), the reservoir characteristics of the site(s) of interest, available substitutes, and characteristics of the facilities available at the site(s). The models were fitted using a non-linear least squares model and the Heckman two-stage model (Heckman, 1979). The authors stated that the Heckman model was particularly suited to data with a large number of zeros in the dependent variable. While both models performed well, the Heckman model had all the expected coefficient signs. Using the Chow test the authors found that coefficient estimates were statistically different between all models and could not be transferred across regions. Despite this, the authors state that the models of ACOE districts in the mid-south (two of the considered districts) were very similar to one another and suggested that geography and similarity of demographics may greatly influence the transferability of demand models. 25 English and Bowker (1996) developed zonal travel cost models for day trip visitors to the Chatooga River along the border of Georgia and North Carolina. While the primary research objective was to determine the impact of alternate cost specifications on consumer surplus estimates, their paper also serves as a good example of a contemporary approach to model specification when estimating zonal travel cost models. Models were estimated using both single log ordinary least squares regression and Tobit models estimated via maximum likelihood estimation. The OLS model was estimated with the dependent variable number of trips +1, while the Tobit model was fit with the dependent variables number of trips and trips/capita. A value of 1 was added to the OLS formulation to insure that zones with zero trips were included in the model. Without the adjustment these zones would fall out of the model since the natural log of zero is undefined. Contemporary Recreation Models Incorporating GIS The advent of Geographic Information Systems (GIS) has greatly improved the ability of researchers to analyze recreation data in a spatial context. Several recent recreation demand studies have explicitly incorporated the use of a GIS in the modeling process (Alig and Voss, 1995; Lovett et al., 1997; Brainard et al., 2001; Hanink and Stutts, 2002). Alig and Voss (1995) attempted to model current camper use, and to predict future camper use for the Chequamegon and Nicolet national forests in Wisconsin using 26 age-specific participation rates and demographic data. Counties were the unit of analysis for visitor origins. The proximity and demographics of populations living around the national forests were determined via a GIS incorporating spatial databases of county boundaries, national forest boundaries, and US. census bureau data. Camper recreation use was modeled as a function of the population and participation rates of those living within 125 miles (the estimated maximum travel distance of campers) of the national forests. The model predictions of recreation use were significantly greater than the observed campground visitation. The authors cited several reasons for failing to validate the model, including the failure to incorporate substitute recreation sites and a non- uniform pattern of population distribution. Furthermore, the authors stated that the estimates of visitation were highly dependent upon the determination of the market area, for which no reliable data existed. Lovett and others (1997) were more successful in developing a recreation demand model for woodlands in England. First, using a geographic information system (GIS) the authors determined network distance and travel time for origin-destination pairs identified from a sample of woodland visitors. Second, the authors developed a substitute grid surface from a land use classification system to determine a substitute index for each origin, explicitly capturing the availability of substitute sites. Finally, socioeconomic and demographic variables were associated with each origin in the study. Incorporating population, travel time, the interaction between travel time and population, and unemployment rate, the authors used Poisson regression techniques to specify the recreation use model. The model successfully predicted total site visitation but performed 27 poorly at estimating visitation by origin. The authors suggested that identifying different sub-groups within the population and specifying multiple demand functions to represent these groups might improve model performance. Brainard and others (2001) improved the model constructed by Lovett and others (1997). The primary improvements to the model were the calculation of a market access variable and the inclusion of site amenity data. Market access for each site in the study was quantified by using a distance decay function developed from visitation count data and an interpolated population surface estimated from the US. Census. The use of a market access variable greatly improved the performance of the model. The authors suggested using larger sample sizes and developing a set of user-specific models would increase the transferability of the model. Most recently, Hanink and Stutts (2002) developed demand models for National Battlefield Parks. The model estimated expected visitation based upon market access, the distance to substitute battlefields, and battlefield amenities. Market access in this case was determined based upon distance to metropolitan statistical areas rather than the population of distributed areas. The specified models performed very well at predicting total battlefield visitation. Independent variables explaining a large proportion of variation in visitation were “market potential” and “number of historic battlefield casualties”. 28 Recreation Segmentation To allocate recreation resources between competing recreation uses, planners must identify the types, quantities, and locations of the recreation opportunities demanded. In turn, decisions related to site-specific management require an understanding of the kind and quality of recreation resources, programs, and facilities demanded by users (N AS, 1975). Models estimating total use or demand in aggregate provide little information in the context of these decisions. To provide more information, expected use levels or demand is frequently estimated within user groups (user segments). These user segments should be meaningful in that they are applicable to planning and management decisions while at the same time explaining recreation behavior within groups. Segmentation is defined as the “process of dividing a large heterogeneous population into smaller homogenous subsets” (Bojanic and Wamick, 1994). Reid (1989), as cited in Bojanic and Wamick (1994), identified five broad approaches to segmentation: geographic, demographic, psychographic, behavioral, and benefit. Geographic segmentation classifies visitors based upon the location of their residence. The demographic approach classifies visitors based upon demographic and socioeconomic variables such as age, gender, income, race, family status, etc. Psychographic segmentation classifies users based upon the motivations of the user while behavioral segmentation segments visitors based upon their actions or activities. A 29 benefits-based segmentation classifies visitors based upon the personal benefits or utility received from recreating. The USDA FS has a lengthy history of classifying users based upon their primary recreation activity (one aspect of visitor recreation behavior). The first attempt to census visitors to USDA FS lands was mandated to be made within specific recreation activities (Waugh, 1918). More recent survey efforts, such as the CUSTOMER survey in the 1980’s, have also attempted to estimate visitor use within specific activity groups of interest (Alward et al., 1998). The USDA FS also uses a benefits-based segmentation approach when classifying recreation use by Recreation Opportunity Spectrum (ROS) classes. One assumption of the ROS approach is that users of the different ROS class areas have differing expectations and receive differing benefits; that is, visitors who recreate in primitive non-motorized areas likely receive different benefits than visitors recreating in rural areas. This approach to segmentation is complicated by visitors who recreate within multiple ROS classes and uncertainty as to whether many visitors can readily identify and make recreation decisions incorporating the ROS classification. Activity segmentation is attractive in that the relationships between activities and management actions appear straightforward. However, users within activity groups can still remain very heterogeneous in their recreation characteristics and behavior. Average trip spending for USDA FS fishing parties can vary from $46 to $222 depending upon Whether the party is on a day trip from the local area or whether they live outside the local area and are spending the night on the forest (Stynes et al., 2003). Similarly, non- 30 local snowmobile parties spend nearly 50% more in total than local parties; spending $20 more in restaurants and $15 more on gas and oil. Faunce and others (1979) have found differences in spending, motivation, and participation when comparing resident and non-resident Maine hunters. N on-resident hunters spent nearly twice their resident counterparts and were more frequently motivated by the social interaction aspect of hunting while resident hunters were more motivated by sustenance. The participation rate of women among resident hunters was much greater than non-resident hunters. Etzel and Woodside (1982) found differences in expenditures, frequency of participation, family status, and expectations when comparing “near home” and “distant” general vacation travelers. Travelers staying near home spent less on their trip, had previously visited the area more times, had larger travel groups, and expected rest and relaxation rather than excitement, personal growth, or intellectual stimulation compared to distant travelers. Some researchers have employed recreation segmentation approaches based on factors other than primary activity. In his dissertation aimed at profiling recreation users, Hendee (1967) segmented visitors into five recreation groups based upon the recreationists’ stated recreation resource preference: national forest wilderness users (dispersed areas), national park wildemess users (dispersed areas), national forest car campground users, national park car campground users, and state park users. User groups differed along socioeconomic variables including age, income, family status, and in their attitudes toward management activities and use of natural resources. 31 Cordell (2003) has segmented the US population based upon recreation behavior and participation patterns using data from the National Survey on Recreation and the Environment. Eight recreationist segments are identified: inactives, passives, nature loving drivers, nature and family, activity samplers, motor consumptives, skiers, and enthusiasts. Inactives represent the largest segment and are least likely to participate in outdoor recreation while enthusiasts represent the second smallest segment but have the highest rates of outdoor recreation participation. Demographic and socioeconomic patterns differ between groups as do attitudes toward natural resource use and management. Identifying meaningful differences in terms of demographic characteristics and attitudes can assist natural resource managers and policy makers in decision-making. Stynes (1999), Stynes and others (2003), and Stynes and White (2003), for the purpose of explaining spending, have adopted a user segmentation approach based upon the type of trip completed and whether the visitor lives in proximity to the recreation resource. The trip type includes day trips and overnight trips with overnight trips further differentiated by the type of lodging used. Trip type segmentation performs better at explaining individual spending than activity segmentation—the former explaining 18% of variation while the latter explains only 2% (Stynes and White, 2003). In addition, party characteristics, such as trip length, number of people, and number of children, were also found to differ by trip type segment (Stynes and White, 2003; Stynes et al., 2003). 32 Relationship to the Current Study Assuming there are differences between distance segments in terms of recreation behavior, activity preference, and participation, models estimating recreation use within segments will assist in long—term, forest-level allocation, and site-specific recreation planning. To be most useful, the model output should produce information that can be used to identify the level of participation in different types of activities, the types of recreation trips undertaken, the characteristics of visitors and visitor parties, and temporal patterns in recreation use. If possible, it would also be useful to identify what spatial patterns of recreation use are expected. The expected recreation use of distance segments accompanied with a characterization of the distance segments is expected to provide such information. Frequently, recreation use models have estimated the recreation consumption (or demand) for a specific recreation resource conjointly for all potential recreation visitors (or for all visitors participating in an activity of interest). However, it seems logical that visitors located at a range of distances from the resource respond differently to the factors that influence recreation behavior. For example, visitors in close proximity to resources may incorporate travel cost into their recreation behavior differently than those located farther away. In this case, using a single model relating travel cost to recreation use for all visitors does not appear to be appropriate. Inability to replicate visitation by origin while correctly predicting total visitation (e.g. in Lovett et al., 1997) may be a manifestation of this functional change in relationship between travel distance and visitation. 33 The failure to incorporate substitute recreation sites in the study conducted by Alig and Voss (1995) resulted in poor model performance, specifically overestimation of recreation use. Selecting an approach to quantifying the availability of substitute recreation sites for recreation use models is challenging. In part, this is due to the lack of a clear theoretical basis for identifying substitutes and the small number of studies investigating how recreationists View substitutes and how they decide between alternate recreation areas. Three frequently adopted approaches to quantifying substitute availability include 1) distance (or travel cost) to a single (or a limited number of) alternative recreation site(s), 2) development of a substitute index incorporating distances to multiple sites and the quality (or attractiveness) of those sites, or 3) creation of a spatial substitute layer using a GIS. Hellerstein (1991) and Boxall and others (1996) adopted the first approach in their travel cost studies of the Boundary Waters Canoe Area in Minnesota and Rocky- Clearwater forest in Alberta, respectively. This approach to quantifying substitutes seems to be more frequently adopted when there is only one recreation resource of interest and a nearby recreation area is clearly identifiable as a substitute. In the former, the substitute site was Algonquin Provincial Park in Ontario (offering similar canoeing and backcountry experiences) while in the latter study the substitute was Bow-Crow Forest (a nearby Alberta Forest Service public land area). In both studies the coefficient on the substitute variable was statistically significant and had the expected positive sign. In a Variation of this approach some researchers have identified a number of potential 34 substitute sites (e. g. all state parks) and identified the minimum distance (or travel cost) to the single nearest substitute site for each visitor origin (e. g. Fix and Loomis, 1997). Substitute indexes were adopted in Loomis and others (1995) and Hanink and Stutts (2002). In the former, the substitute index calculated for each visitor origin was the ratio of lake acres (the selected measure of recreation site quality) to travel distance summed across all identified substitute sites. In the latter, the authors used two substitute indices, one incorporating competing site quality (the number of historic battlefield casualties) and one not incorporating competing site quality. In both of these studies, recreation use models were estimated for multiple recreation areas and there was no clearly identifiable single recreation site that could serve as a substitute. In both studies the statistical significance of substitute index variables were mixed depending upon the chosen functional form of the model. Brainard and others (2001) opted to develop a “recreation potential surface” for use as a substitute variable in their model of demand for recreation at forestry commission recreation sites in England. The authors first placed a grid of sample points at 5km spacing throughout the study area. A measure of access to publicly-accessible forestland was then computed for each sample point. Interpolation was then used to develop values in the intervening distance between samples, yielding a continuous surface throughout the study area. One problem in the adopted approach is that the modeled recreation sites are included in the calculation of the recreation potential surface—so it is not a strict substitute surface. When incorporated in the models of 35 recreation use the recreation potential variables were statistically significant, though the direction of the relationship was inconsistent. The use of a single site substitute variable seems inappropriate for the current study since there is likely more than one substitute for any individual national forest. However, one option is to identify a spatial database of potential substitute sites (e. g. national parks, national forests, BLM land, etc.) throughout the study area and identify the nearest substitute destination of each ownership type. The development of a recreation potential surface is probably not feasible given that a substitute-only surface would have to be computed for every origin-forest pair. The traditional activity-based USDA FS approach to segmentation yields information related to the number of individuals participating in a specific recreation activity. However, within activity groups recreation behavior and participation are highly variable. In addition, classifying visitors by primary activity can be imprecise since many recreationists participate in multiple activities on a given recreation visit. For example, what is the proper activity classification for an angler who also spends two nights camping on the national forest? A further discussion of the limits to segmenting visitors in terms of primary activity can be found in Stynes and others (2003). Alternative approaches to segmentation such as using distance and trip type segmentation and/or visitor motivation have been successful at explaining variation in recreation behavior (e.g. Faunce and others, 1979; Etzel and Woodside, 1982; Cordell, 2003; and Stynes and White. 2003). The adopted distance segmentation (and associated recreation use models) 36 is expected to provide information for forest planning that may not be captured via activity segmentation alone. Conclusion This chapter examined three areas of recreation literature relating to the objectives of this study. The initial section of this chapter included a discussion of the role of demand analysis in resource planning. This discussion was centered primarily on the 1975 NAS document “Assessing Demand for Outdoor Recreation” and the types of planning questions addressed through demand analysis. The second area of focus in this chapter was on methods commonly used to estimate recreation demand and use. Specific emphasis was given to the travel cost model. A discussion of some contemporary examples of recreation demand and use models was also included. The final area of literature examined in this chapter was approaches to segmenting recreation visitors. This section described the traditional approach to recreation visitor segmentation, some of the inconsistencies in recreation behavior that result from this approach, and some alternative segmentation approaches. 37 CHAPTER 3 METHODS The primary focus of this Chapter is a description of the methods used to achieve the dissertation objectives identified in Chapter 1. This Chapter begins with a conceptual discussion of the distance-based segmentation and the relationships between the segmentation and the recreation literature. Given its importance to this study, a brief description of the NVUM project, the sampling design, and the visitor survey procedures follows. The remainder of the chapter is devoted to description of the specific analytical approaches in this study. The study area is described and the NVUM survey data used in the study are identified. Specific approaches used to delineate the distance-based segments and the statistical tests used to compare segment visitors between USDA FS regions and between distance segments (objectives 1 — 3) are discussed. The closing section details construction and evaluation of the models used to predict the recreation use of Local and Mid-distance visitors for individual national forests in the study regions, Objective 4. Conceptual Discussion of Distance-Based Visitor Segmentation The distance-based approach to visitor segmentation adopted in this study classifies recreation users (and potential users) based upon the proximity of their residence to the recreation resource. Three distance-based visitor segments are 38 recognized: Local visitors, Mid-distance visitors, and Long-distance visitors. Within segments, visitors may be further disaggregated by recreation trip-type (e. g. day trip or overnight trip). Visitors are classified based upon their proximity to the recreation resource under consideration. As such, individuals in a single location may be classified as Local visitors (or potential local visitors) to one forest and Mid-distance visitors (or potential Mid-distance visitors) to another forest, etc. Spatially, Local visitors originate from areas in close proximity to the forest boundary, Mid-distance visitors originate from areas farther from the forest boundary, and Long-distance visitors originate from the “rest of the world” (Figure 2). Conceptually, visitors within each of the distance segments share similar recreation behaviors, recreation participation patterns, and have similar responses to the factors that influence recreation (distance, amenities, substitutes, and socioeconomic conditions, etc.) as other visitors within that segment. National Forest Local Mid-distance Long—distance Figure 2. Distance Segmentation of USDA FS Visitors. 39 The distance-based segmentation can be grounded, to some extent, in the classification of outdoor recreational uses and resources as described in Clawson and others (1960) and in Clawson and Knetsch (1967). Three classes of recreation uses and resources were identified by the authors: user-oriented, intermediate, and resource-based. Under the Clawson and Knetsch classification scheme, individual recreation resources are classified into one of the categories based primarily upon the proximity of the resource to major populations and the natural features of the resource. However, it seems reasonable that a single national forest could represent different resource types (and provide for multiple recreation uses) to multiple individuals. That is, visitors originating from varying distances may recreate at a national forest as if it were a user-oriented resource, an intermediate resource, or a resource-based recreation resource. User-oriented areas, as outlined in Clawson and Knetsch (1967), are identified by “their ready accessibility to users”. This proximity allows users to recreate at the resource with high frequency and during a “variety of leisure times” (daily leisure times and/or weekend leisure times). Visitors to these areas frequently participate in “general” recreation activities such as swimming, picnicking, or walking. Intermediate areas are those that are located at moderate distance from visitors. Given this distance, recreationists are generally constrained to recreating at intermediate areas during long day trips and weekend visits (Clawson and Knetsch, 1967). Recreation activities at intermediate areas more frequently include camping, hunting, and fishing than visitors to user-oriented resources. 40 IS Iimit Resource-based areas, under the Clawson and Knetsch (1967) classification, are farthest from populations and have “outstanding physical resources”. Recreation at these resources generally occurs during a vacation and may occur coincident with other activities (the resource may not be the primary purpose of the trip to the area). Individual annual visitation rates are very low. Visitor recreation activities include those at intermediate areas as well as more general and “interpretive” activities such as sightseeing, nature study, and visiting historical places. Using Clawson and Knetsch’s classification scheme as a framework, postulated characteristics of distance segment visitors were identified. Local visitors have either permanent or seasonal homes in close proximity to the national forest under consideration. Conceptually, these visitors take frequent trips with each trip typically being of short duration—day trips or shorter (Table 1). Longer overnight trips are taken with less frequency and may be associated with special recreation activities (e. g. hunting). Local visitors recreate at a limited number of sites and participate in a limited number of activities on any given trip. Expenditures on individual trips are relatively small—though in aggregate cumulative spending by these visitors may be high. Conceptually, Local visitors are more likely to be involved in forest planning activities and to comment on recreation management decisions. Due to their increased visit frequency and proximity to the forest, these visitors are most affected by natural resource management decisions on an individual forest. The forest-level visitation of this segment is limited mostly by the local population. For many forests these visitors will represent the majority of forest visitation. 41 \ dist Terr Table 1. Postulated Characteristics of Local, Mid-distance, and Long-distance Visitors. Local Mid-distance Long-distance Recreation Characteristic Visitors Visitors Visitors .-Anaaal-IaP_Ete999991/.-----------hi_gh ............ {gadgets $9.199! ....... Yea/Jaw .............. generally moderate .-IFIP.PB¥?P§911--------------.----_--_.Sh91it ............ Eaters--.------_----'.--§h9tt-t.9_!99€19rat.e.-_- pass through day . generally day and overnight or overnight off- ..ItipIyps---.-------------..----.----s!ay._tr.irz§ ....... trips ...................... ferasttrin ............ Primary Purpose of the Trip is nearly .Fatima!Eaterifiaqtsatiea-runaways ......... typically. ................ irrfrsqaentlr .......... single or ...S.i.t.e.§.Y.i§.i.t.e.€l-Rer.tIip-_-_------------.f9.W. ............. 2991.621? ................. $1951.99! fart ......... single or .-Agtixitissrsrtfirz---------__------__.f.e.V.V. ............. militias ................. answerer ......... PartySlzesmall ........... medatatsjalatga ..... variable. ............... may occur Weekend/weekday Visitation throughout throughout the Pattemweek ........... primarily3r9§isen€i.---wssls ................... Trip Spending (Attributable to Forest)10w ............. its!) ...................... metals ............... peak and shoulder .-.S.§§§999I-B§¥I9FflBEYEIEPIQP---.X‘?§:€9§P§l----§9§§9’3§ .................. peak. ................... ..IIIISIESEEP.Xi.S.i.t9¥-1.’.r.98r.?m..S.------IQYY ............. high ...................... high .................... Likelihood of Involvement in unlikely to be .-E9I‘E~°:t..Plflliiii.'}g..491i.‘.’.iii?§--------X??X.1.iIS?I)/.----9?9€19!?§?.1)f.1.ik?lx ..... 11.1%.)!st .............. proximity of forest to tourist Primary Limiting Factor proximity of forest destinations, cities, Affecting Total Group local to population thoroughfares, Visitation population centers airports Mid-distance visitors originate from beyond the local area but within a moderate distance of the forest under consideration. Mid-distance visitors complete up to several recreation visits to the national forest in a given year with annual visitation rates decreasing as the distance to the forest increases (Table 1). These visitors generally have 42 for; CEO." ltt‘ft teen inilt SL105 son W01 of l the rela tip 103 longer trip durations, a greater propensity to visit multiple sites within the national forest, and greater likelihood to participate in multiple activities per trip than Local or Long- distance visitors. Mid-distance visitors are more likely than Local visitors to be interested in interpretive programs and visitor centers. Visitors from this area are more likely to include an overnight stay in their trip than visitors from Local or Lon g-distance areas. Visits to the forest are more likely to be secondary to some other trip purpose for these visitors than for Local visitors. Recreation spending by Mid-distance visitors initiates the majority of local economic activity directly attributable to recreation on the national forest. As such, the recreation use of these visitors can be very important to economics dependent upon recreation spending. Conceptually, the forest-level recreation use of these visitors is influenced by the distance from the forest to population centers, the existence of s ubstitute recreation areas, the characteristics of the recreation resource, and the socioeconomic characteristics of potential visitors. Visitors originating from beyond the Mid-distance area (i.e. the “rest of the world”) are categorized as Long-distance visitors. This group is comprised of two types of visitors: 1) those in the forest area primarily for some purpose other than recreating on the forest (i.e., business, visiting a nearby tourist destination, visiting friends and relatives, traveling through the area, etc.) and 2) visitors who have traveled to the area expressly to visit the forest. Long-distance visitors may confine their recreation activity to a select number of particularly attractive or well-known recreation sites on the forest 43 he: por- OPPI' reprc PIES: sub: bid four Lil: fore like PTOx litre (Table 1). Regardless of trip-type, visitors in this group have annual visitation rates that are very low. Long-distance visitor typically complete either day trips or overnight off- forest trips. When on day trips, the visitor is frequently passing through the area. A limited number of Long-distance visitors spend the night on the national forest. Since the national forest may not be the primary reason for the trip away from home, much of the spending of by these visitors may not be attributable to the national forest. Conceptually, the number of Long-distance visitors will be influenced by the proximity of the forest to population centers, tourist destinations, lodging, major travel routes, special recreation opportunities, and airports. For most forests, Long-distance segment visitation will represent a small percentage of total forest recreation use. Some literature lends support to the link between the distance segmentation in the present study and the resource classes of Clawson and Knetsch (1967). Dwyer and others (1977) stated “different values may be placed on recreation participation by different subsets of the population”. In particular the authors stated these subsets may be defined by distance from the facility (among other possibilities). Strauss and others (1993) have found distance effects in the participation of hunters in the Delaware Water Gap. Likewise, Nelson and Lynch (1995) found that visitors to undeveloped portions of the forest whose primary residence was within the forest proclamation boundary were more likely to participate in hiking/walking and less likely to participate in off-road vehicle (ORV) use. Spending patterns of recreation visitors have been shown to be related to the proximity of the visitor’s residence to the forest (Stynes and White, 2004). Nationally, recreation parties originating from greater than 30 straight-line miles away from the 10:: int 115:: pnr. outs: chr forest spend 50% more money in the local forest area per trip than parties originating from within the local area (Stynes and White, 2004). Finally, there is some evidence that distance to the recreation site influences visitor decision-making. Reiling and others (1993) found that campers were indifferent between comparable camping sites when travel times were less than two hours. However, travel time to the sites was highly important in campground selection when traveling greater than two hours from home. Regarding participation in specific recreation activities, Bristow and others (1993) have found that boaters in Massachusetts depended primarily upon local supply of boating opportunities while campers generally traveled outside the local area for camping opportunities. Distance-based segmentation will provide an alternate way of quantifying and characterizing USDA FS recreation use. Whereas activity-based segmentation identifies the activity patterns of visitors, distance-based segmentation may explain the patterns of recreation participation and consumption as well as the recreation behaviors of visitors. Models predicting the expected recreation use of the distance-based segments will identify the types of recreation visitors expected and may identify the “recreation markets” of individual national forests (assuming the relationships between recreation use and influencing factors remain the same). 45 n, Sim. 1156 mm lode almo; use. Q ICCIell manor airs COUlll National Visitor Use Monitoring Project Estimates of USDA FS visitation developed via the USDA FS National Visitor Use Monitoring (NVUM) project and the visitor survey data collected in that process are used in this research. NVUM began as a pilot project in 1998 and was implemented throughout the National Forest System (NFS) in 2000. The goal of NV UM is to develop statistically reliable national, regional, and forest-level estimates of USDA FS recreation use (English et al., 2002). Unlike previous USDA FS visitor survey projects (e. g. PARVS and CUSTOMER), NVUM uses a consistent visitor survey and sampling scheme implemented on all units within the NPS. The data used in this research are drawn from the first four-year NVUM cycle which occurred from calendar year 2000 through fiscal year 2003 (ending in September 2004). In the first NVUM cycle, approximately M: of the administrative forests in the NPS were sampled every year. Concomitant with the implementation of NV UM were changes in the unit of measure for USDA FS recreation use and the adoption of a more conservative approach to defining what “counts” as USDA FS recreation. Previously, the USDA FS relied almost exclusively on the Recreation Visitor Day (RVD) as the unit to measure recreation use. One RVD equals one person recreating for one 12-hour period. Under NVUM, recreation is quantified on a “visit” basis. A visit is “one person entering and exiting a national forest. . .for the purpose of recreation” (English et al., 2002). A visit may last just a few minutes or several days. An individual camping for a week on a national forest is counted as one visit while an individual staying overnight in a hotel off of national forest 46 Olffi area land and visiting the forest on three days is counted as three visits. Previously, USDA FS visitation estimates included individuals traveling through national forests for purposes other than recreation as well as individuals viewing national forest scenery from an off- forest roadways, airplanes, ships, etc. Under NVUM, the definition of a recreation visit has changed to exclude these cases from visit counts. NVUM employs a double sampling approach, completing visitor counts and administering visitor surveys at selected locations on selected days (sample days) within individual forests. Estimates of recreation use are constructed by combining traffic count data with information obtained from visitor surveys.I Sample days are selected via a stratified random sample from a population of site days identified for each administrative forest by personnel from that forest. Site days are stratified by the type of recreation location (site type) and the expected level of exiting recreation traffic (site strata). Site types incorporated in NVUM include day-use developed sites GDUDS), ovemight-use developed sites (OUDS), Wilderness areas (WILD), and the general forest area (GFA).2 DUDS sites have received “moderate to heavy” modifications for the purpose of visitor convenience, education, and comfort as defined in the USDA FS Infrastructure Access (INFRA) database (English et al., 2002). These sites include picnic areas, fishing sites, interpretive sites, visitor centers, etc. OUDS sites are overnight sites that have received “moderate to heavy modification” as defined by INFRA and are generally developed campgrounds, cabins, hotels, resorts, etc. (English et al., 2002). ' Appendix A includes detailed descriptions of the NVUM protocol and use estimation procedure. 2 A fifth site type, viewing corridors, exists within the N VUM sampling process. However, recreation use on this site type does not contribute to the overall visit estimate and is not considered in this study. 47 W17. .7. IL all} 3116. 03‘“ [he ;' WILD areas are congressionally designated Wilderness areas. Sampling locations for WILD areas are located at trailheads and access points (English et al., 2002). The GFA encompasses the remainder of the forest area not elsewhere classified. Sample locations for the GFA are generally at trailheads, parking lots, and NFS roads exiting the national forest (English et al., 2002). Site strata are based upon the “expected level of exiting visitor traffic relative to all site days in that site type” (English et al., 2002). Strata include high, moderate, low and closed/no expected exiting recreation use. Individual forest managers classify site days into the site strata based upon their own judgment and the recreation use patterns on the individual forest. NVUM Visitor Survev Visitors (or parties) who stop at a voluntary check point established on the sample day are questioned to determine if they are candidates to complete the NVUM visitor survey. Visitors qualify to complete the survey if they are 1) recreating on the forest and 2) exiting the recreation site (DUDS, OUDS, WILD) or forest (GFA) for the last time that day. If a party of visitors, rather than a single individual visitor, stops at the check point the individual 16 years of age or older with the most recent birthday is asked to complete the survey. The survey consists of a general questionnaire completed by all respondents and two supplemental questionnaires (a satisfaction supplement and an economic supplement) that are completed by subsets of survey respondents. 48 l’lSiI ‘ list. 6116.“. llktlz.‘ “he: “Big I TIMI: Crimp}: The NVUM general survey is designed to gather information relating to the duration of time spent recreating on the forest, the number of USDA FS recreation sites visited, participation in recreation activities, past visit frequency to the forest, the home ZIP code of the respondent, the purpose of the trip, party characteristics, and demographic information. The economic supplement to the general survey gathers information on the duration of the trip away from home, annual recreation expenditures, and trip-related expenditures made in the local area. All portions of the survey are designed to be read to survey respondents by trained interviewers and all parts, including the economic addition, are completed on-site during the interview.3 Data Weighting Schemes Three weighting schemes are available for use with the NVUM survey data: exposure weights (Exth), national forest visit expansion weights (NVEXPAND), and visit weights (VisWt). Exposure weights adjust the sample for overrepresentation of those visitors who visit multiple sites and/or spend multiple days in the GFA. Due to their extensive movement and extended time in the national forest, these visitors have a greater likelihood of being sampled. Formally, the weight is computed as 1 Ex Wt-=—, p 1 NSj where N 81- is the number of sites (and days in the GFA) visited by individual j. Exposure weights for the study area visitor sample range from 0.03 to 1. 3 There is anecdotal evidence that the economic supplement was sometimes handed to the respondent for completion rather than read aloud. 49 is: sit: (Er RIC 0003 16V CI I bet) a NV’T The NVEXPAND weights expand the visitor sample to conform to the NVUM use estimate. The weight for an individual respondent j is the product of the number of site visits the respondent represents (SVE,-) and the exposure weight for individual j (Exthj): NVEXPANDj = SVE, * Exthj. The NVEXPAND weights for data included in this study range from 1.6 to 51,450. Visit weights correct the sample for overrepresentation of those visitors that visit an individual national forest many times in a given year. Similar to exposure weights, those visiting an individual national forest many times annually have a greater likelihood of being sampled. Weights for individual respondents are based upon the product of Exth,- and the inverse of the reported number of annual visits for individual j (nfv12moj+1): VisWt - = Exth - * ‘ l * J J (nfv12moj +1) A limited number of individuals reported annual visitation rates greater than 365. In these cases, the reported annual visits were truncated to 365. This research follows the weighting approach adopted in Stynes and others (2003). Exposure weights (Exth) are used when there is no expectation that site type or level of site use (the factors in calculating NVEXPAND) would influence visitor behavior. The characteristics of visits are primarily estimated using Exth’s. NVEXPAND weights are always used when estimating total visitation or the visitation of 50 ad? Reg Met" Nth Pike Rio l Rot. San , ShnE Whit PM? a subset of visitors. The VisWt’s are used only when estimating the mean number of annual visits. Study Area The study area for this research is USDA FS Regions 2 and 9 (Figure 1, Chapter 1). Region 2 is geographically located in the Rocky Mountain region of the western US. while Region 9 encompasses the northern portion of the eastern United States. Twelve administrative forests are located in Region 2 and 14 administrative forests are located in Region 9 (Table 2). Both regions include a mix of “general” and “tourist destination” forests. Table 2. Administrative National Forests (NF) within USDA FS Regions 2 and 9. Region 2 Region 9 Arapaho and Roosevelt NF Allegheny NF Bighorn NF Chequamegon / Nicolet NF Black Hills NF Chippewa NF Grand Mesa, Uncompahgre and Gunnison NF Green Mountain And Finger Lakes NF Medicine Bow NF Hiawatha NF Nebraska NF Hoosier NF Pike and San Isabel NF Huron-Manistee NF Rio Grande NF Mark Twain NF Routt NF Monongahela NF San Juan NF Ottawa NF Shoshone NF Shawnee NF White River NF Superior NF Wayne NF White Mountain NF In addition to national forests, Region 9 includes the Midewin National Tallgrass Prairie (MNTP)——the first federally-desi gnated tall grass prairie in the Nation. MNTP was 51 €51. ICC fag: Cf.“ Uillt nan, “tie B)" :5]. We125: admin: 333353;: Finis Louie: I. established in 1996 on a former ammunition plant and depot near the city of Joliet, Illinois. To insure visitor safety, the site has, for the most part, been closed to public recreation while the remnants of the previous ammunition production and storage facilities are removed. Recently, a limited number of trails, recreation sites, and a visitor center have been established. Given its limited access for recreation activity, its uniqueness as a USDA FS public land area, and limited NVUM sample size (80 respondents), MNT P survey respondents and visit estimates have not been included in this analysis. In Region 2, four of the 12 administrative national forests also manage nearby national grasslands. NVUM recreation use estimates for the forest/ grassland aggregates are reported as a single recreation use estimate for the administrative national forest. Due to the difficulties that may arise from modeling recreation use to national grasslands and national forests in aggregate, the estimates of grassland recreation use on three of the four administrative national forests were subtracted from the respective total administrative national forest recreation use estimates.4 Correspondingly, surveys completed by respondents sampled on the excluded national grasslands were removed from the survey database. Recreation visitors sampled on national grasslands managed by the Nebraska NF were included in the NVUM visitor sample and the recreation use of the Nebraska NF By isolating those NVUM respondents sampled on national grasslands and using the N VEXPAND 'eights a rough approximation the percentage of total administrative national forest use associated with the Iministered national grassland can be made (English, Pers. Comm). A more comprehensive estimation of assland recreation use would require identification of the number of identified sitedays within stratum on isslands relative to the number on the national forest itself. Having done this the NVEXPAND weights uld be adjusted to identify national grassland recreation use. 52 tot N: in no: in .- .\'\' and its managed grasslands was modeled in aggregate. This approach was adopted for three reasons: 1) use on Nebraska NF grasslands accounted for a significant percentage of total Nebraska NF recreation use, 2) a greater percentage of respondent sampled on the Nebraska NF originated from outside the “local area”, and 3) differences between the natural features of the national forests and national grasslands within this aggregate are not as substantial. Survey Data In the first cycle of NVUM, 16,991 visitor surveys were completed by visitors to administrative forests in USDA FS regions 2 and 9. All respondents completed the general survey and ‘A (4,479) completed both the general survey and the economic addition. Stynes and others (2003) found no statistical differences between the economic sub-sample and the general survey respondents. Variables applicable to this study obtained directly from the survey instrument and variables computed from survey data are shown in Table 5. All NVUM survey data were obtained from NVUM administrators in ASCII flat file format. The data had undergone a preliminary “cleaning” by USDA FS personnel to remove inconsistencies found during the visit estimation process. Additional cleaning activities to facilitate economic analysis were completed by MSU personnel and are outlined in Stynes and White (2002), and Stynes and others (2003). Several variables used in this study were computed from information provided by JVUM survey respondents (Table 3). The duration of the forest visit (visdur) was 53 estimated as the difference between the reported ending of the forest visit (forend) and the reported beginning of the forest visit (forarrive). This calculation was performed using date and time functions in Microsoft Excel®. The number of sites visited (expose) was computed as the sum of the number of developed recreation sites visited (numsites) and the number of days spent recreating in the GFA (GFAdays). In addition to identifying their primary recreation activity, NVUM respondents were also asked to report all the recreation activities in which they participated. The number of activities (numact) is the count of these reported participation activities. The study variables fordist and triptype were constructed as part of the economic analysis of the NV UM survey data as reported in Stynes and others (2003) and Stynes and White (2005a). Fordist is the straight-line distance between the centroid of the reported home ZIP code of the NVUM respondent and the nearest boundary of the visited national forest. Approximately 15% (2,770) of non-foreign NVUM respondents in the study area either failed to report a ZIP code or provided a ZIP code that did not conespond to the ZIP code databases used to identify respondent origins. In these cases fordist values are classified as “missing”. Additional discussion related to the calculation of forest distance calculation can be found in Stynes and others (2003). 54 Table 3. NVUM Survey Variables. Survey Computed Variable Variable Description Variable Variable regionUSDAFSRegwnl ......................... .-191991.-...--..N.21i.99a1-19.1:9§t 31911951 ........................................ 1 ......................... ..iattiat.....-.M911th./.c19x11i.r.r1_e. 9.1 merrier ................................ 1 ......................... .-faranrs-.....Manthtdexltime 91.41am! at target. ........................ 1 ......................... .-19end.........Marthtdexltima 9.1 lasting. the. fares! ...................... 1 ......................... nflemo Number of times visited this NF during the ............................ Past. earl .-1191”..§11€§-.----N}111112?£91.?91?.$:.QIJP§:-W1EP§11?§1119211991 .......... 1 ......................... ..Qfééaxs.....flumber 91 daysiaihegfié ................................. 1 ......................... .1129949....-...1191116_.Z.11?.9.9§19-9.f_amends!!! .............................. 1 ......................... .£91918?"..--.-3911191199111-1§§-19§.i§l?111.91.?.191§1g!1-99911111/.----_-----.1 ......................... peopveh Number of people in the respondent's ............................. Y ehlclel madman/equity ................................................ 1 ......................... .-Etitantl. anarr 291129.926. -91. 111.9 11113-19. 111.9 area .................... 1 ......................... nfprime Is the national forest your primary ............................. <1 estln3t10n°1 lotyPeTr/reoflodgmsusedl ......................... .-xisdyr---------9.11111.th91.t11e.19ts§.t-xi§i.t ................................................... 1 ........ .-E’SRQ§9--------18.911199191.1?9£§?11911-§11?§ rated. .................................................. .-namaet. - - - - - _ - 15191111291 91.a.c.t.ixi.t.i9§n.ar.1i.c.i.pat§si. in. ................................................ fordist Distance from respondent's ZIP code to the ............................. f. orestboundaryl triptype Respondent’s trip type (i.e. day trip, ............................. 9 xemight.trimetufiraantrip)---__--.-----------_-_------_------1-------- .-avatrzaad. - - - - 1512111931?! 191%! 31919219399911 was!!! .................................... 1 ........ tamer! .......... 1. terms? ......................................................................... 1 ........ l viswt expwt*(1/nfvl2mo+1) a Year 4 data only b Years 1 - 3 data only In the course of development of recreation spending profiles for USDA FS visitors, NVIJM survey respondents were classified into one of four trip types: day trips, vernight on forest trips, overnight off forest trips, and not primary trips. Day trip visitors e those who reported they were not spending any nights away from home as part of :ir national forest visit. Overnight on forest visitors were those who answered yes to the 55 question “did you spend last night on the forest?”. Those visitors who reported they would be spending at least one night away from home, not on the national forest, were classified as overnight off-forest visitors.5 For the study presented here, the overnight segments have been combined into a single overnight group (OVN/OVNNF). Stynes and White (2005b) found that NVUM respondents likely had difficulty determining whether they indeed “spent last night on the forest”. Not primary visitors are those who reported that the national forest was not their primary recreation destination. Due to how the “not primary” question was worded in the first three N VUM years the number of not primary visitors was likely underestimated (see Appendix A and Stynes and White 2005b). Distance Segmentation of NVUM Survey Respondents NVUM survey respondents were assigned to one of the three distance segments based upon the distance between their reported home ZIP code and the forest boundary (fordist).6 To do so, the distance boundaries (i.e. travel distances) separating Local visitors from lVIid-distance visitors and Mid-distance visitors from Long-distance visitors had to be identified. This was achieved via the approach described in the following paragraphs, incorporating the NVUM data and NVUM regional visit estimates. Regional-level visitation estimates for five-mile travel distance bands surrounding atr'onal forests were computed using the NVUM survey data and the NVEXPAND E n the first three years of NVUM, respondents were only asked to report the number of nights away from me rather than the number of nights in the local area. As such, some visitors who were passing through local area and but spending the night away from home outside the local area are classified as overnight forest visitors. his approach is consistent with the NVUM economic reports: Stynes and White (2002) and Stynes and ers (2003). 56 weights. Those respondents whose primary activity was downhill skiing were excluded from computation of these visit estimates.7 The cumulative percentage of the total regional-level visitation was computed for each five-mile travel distance band and depicted graphically (Figure 3).8 The percentage of total visitation associated with each distance bands and the marked changes in the relationships between travel distance and visitation rate changes were used to inform decisions regarding segment separation distances. Two important characteristics of the Local visitor segment were included in the conceptual discussion: 1) Local visitors are the majority of total forest visitation (particularly at the regional- and national-level) and 2) travel distance has only a minor influence on the rate of visitation. In their economic analysis of the NVUM data, Stynes and others (2003) utilize a 30-mile boundary to identify local national forest visitors. In both study regions, visitation by those living within 30 miles of forest boundaries constitute more than 50% of total regional-level visitation (Figure 3). Prior to the marked change in slope located at approximately 30 miles, marginal visitation levels of this group of visitors changes very slowly with increasing travel distance. Given the consistency with the postulated Local visitor characteristics and for compatibility with NVUM 7 Since most downhill ski areas were NVUM proxy sites, visitor surveys were completed by downhill skiers at a lower intensity than visitors to non-proxy sample sites (see Appendix A). Due to the high use at ski areas and the low intensity visitor surveying, the average NVEXPAND weight for NVUM skier survey respondents is more than 3 times greater than the average NVEXPAND weight for other survey respondents. Given the potential for individual skier respondents to dominate any analysis of recreation use when considered by distance band (using NVEUAND); downhill skier visits were excluded from segment delineation and from development or evaluation of the recreation use models. Skier survey respondents were included in analysis of visitor segment characteristic—where most analyses are based upon Exth. 8 Cumulative percentages estimated for individual forests in the study area are available from the author. 57 definition of “local visitors”, the travel distance separating Local from Mid-distance visitors was established at 30 miles from origin to forest boundary. 100 90 80 70 4O 3O Cumulative Percentage 20 10 60~ 50— i é N 8 I I I I I (a) :5 U" 0) ‘1 8 8 8 8 8 Visitor Origin Distance (Miles) 8 0 § S Region 2 - - - - Region 9 30 Miles 200 Miles Figure 3. Cumulative Percent of Visitation Across Visitor Origin Distance Bands. Conceptually, the visitation of the Mid-distance segment should be influenced significantly by travel distance and the total visitation of this segment should comprise the majority of non-local visitation. Between the 30-mile and 200-mile distance bands the influence of travel distance on marginal visitation is apparent (Figure 3). At distances of greater than 200 miles, the cumulative visitation levels of the distance bands increase at a relatively uniformly rate toward 100%—indicating no obvious changes in the relationship between travel distance and marginal visitation. At the ZOO-mile band, 80% of the 58 visitation in Region 2 and 90% of the visitation in Region 9 has been accounted for. Given these factors, 200 miles was selected as the origin to forest boundary travel distance separating Mid-distance from Long-distance visitors. Characterizing Distance-based Segments Visitors within the three distance segments were characterized in regard to the following recreation behavior, activity participation, and consumption pattern variables: annual visit frequency, visit duration, number of sites visited per visit, number of activities per visit, visitor primary activity, party size, visit trip type, and weekday/weekend visit patterns (Objective 1). Statistical tests were conducted to determine if the characteristics of distance segment visitors differed between the study regions (Objective 2). Subsequent to these comparisons, a second group of statistical tests were completed to identify differences between the distance segments themselves (Objective 3). The postulated characteristics of distance segment visitors, as well as the postulated differences between the distance segments, are listed in Table 1 (within the conceptual discussion). All statistical tests were completed using SPSS 12®. Selection of the appropriate tests was based upon Howell (1997). Statistical tests were not completed if any one group under consideration had less than 30 observations. Chi-square analysis was used in all tests where the variable of interest was nominal in nature (e. g. trip-type). The Mann- Whitney U test was employed when variables that were ratio in nature (e. g. annual visits) 59 were compared between two groups (e. g. study regions). When completing comparisons of a ratio variable between more than two groups a two-step procedure was completed. Kruskal-Wallis ANOVA (K-W ANOVA) was used to determine if there was a statistical differences between all groups under consideration (e. g. Local, Mid—distance, and Long- distance segments). Subsequent to a finding of statistical difference between all groups, a series of Mann-Whitney U tests were completed between pairs of groups (e. g. Local and Mid-distance segments) to determine specific pairwise statistical differences. The non-parametric tests Mann-Whitney U and the K-W ANOVA were chosen for use in this study over comparable parametric tests (ANOVA and student’s t-test) for several reasons. First, the population distributions of the recreation characteristics under consideration do not meet the assumption of normality required for parametric tests.9 Second, Mann-Whitney U (and the related K-W ANOVA) tests the broad hypothesis that samples were drawn from identical populations—consistent with the objectives of this study (Howell, 1997). Comparatively, the hypothesis of the student’s t-test is that samples were drawn from populations with the same mean. Finally, the results of non- parametric tests, as compared to parametric tests, are not overly influenced by the inclusion of extreme values (e. g. annual visit rates of 365) whereas parametric tests can be (Howell, 1997). Analyses were completed using the following weighting schemes, outlier removal, and data cleaning rules. With two exceptions Exth’s were used for analyses 9 Many argue that t-tests are robust to violation of the normality assumption if the sample size is sufficiently large to yield a normal sampling distribution of the mean under the central limit theorem. 60 relating to visitor characteristics; NVEXPAND weights were employed when estimating the frequency of primary activities while VisWt’s were used for analysis of annual visit frequency. When completing analysis of party size, respondents reporting party sizes of greater than 7 individuals were excluded from analysis. Similarly, respondents reporting a trip of greater than 30 days away from home were excluded from estimation of visit duration. Removal of these outliers is consistent with the approach adopted for economic analysis of the NVUM data (Stynes et al., 2003). Lastly, in several cases, individual respondents reported annual visit frequencies (nfv12mo+1) of greater than 365 visits. In these cases, the reported number of annual visits was reduced to 365 visits.10 Modeling Segment Visitation Objective 4 is to model the forest-level recreation use of Local and Mid-distance visitor segments to national forests in USDA FS Regions 2 and 9. The remainder of this chapter includes a general description of the modeling approach, a specific discussion of the procedures used to construct the recreation models for each distance segment, and lastly, explanation of the approach used to verify and validate the predictive ability of the constructed models. to The average numbers of visits annually is not appreciably influenced by reducing the maximum number 0f annual visit to 100. 61 General Description The models constructed here are meant to predict only the recreation use of Local and Mid-distance visitors recreating on national forest land engaged in activities other than downhill skiing. The recreation use of downhill skiers and visitors to national grasslands is not modeled in this research. For both of these recreation use types, the factors influencing recreation use are likely different from those influencing visitors engaged in traditional outdoor activities on national forest land. Moreover, downhill skier use was not modeled as these visitors were sampled at a very low intensity, leading to very high individual NVEXPAND weights that may adversely influence derivation of band-level recreation use (see footnote 7, this Chapter). Approximately 15% of NVUM survey respondents did not provide a valid ZIP code. As the origin of these visitors could not be determined, the recreation use associated with these visitors was also excluded when constructing and evaluating the recreation use models.11 The forest-level recreation use of Local visitors was modeled using a simple approach incorporating counts of the population living at several distance zones around national forests, estimates of the percent of those individuals participating in national forest recreation, and estimates of the number of visits completed by participating individuals. Mid-distance recreation use was modeled via more complex multi-site zonal travel cost models. While the recreation models are formulated differently, the basic " Alternatively, it could have been assumed that the origin distribution of those providing a ZIP code was representative of those failing to provide a valid ZIP and this use could have been included in modeled recreation use. However, there is no basis for this assumption and it seems reasonable to expect that some respondents may be more likely to not report a ZIP code. 62 premise of each is to model forest-level recreation use based upon information about the populations living in zones around national forests and the patterns of national forest recreation use by individuals in those zones. NVUM survey data and recreation use estimates for national forests sampled in NVUM years 2, 3, and 4 were used to construct and verify the performance of the models. Recreation use estimates for forests sampled in NVUM year 1 were retained for model validation (ascertaining the out-of-sample predictive ability). Retaining a portion of the data to test the out-of-sample predictive ability of models is an accepted form of model validation (Grant et al., 1997; Haefner, 1997). When using a zonal approach to modeling recreation use, the zonal aggregation of the visitor origins must be defined. Typically, zone definition is relatively arbitrary (English and Bowker, 1996). In this study, twenty zones (or distance bands), spanning travel distances from zero miles to 200 miles were delineated (Table 4). The 20 zones are comprised of groups of origins that share a similar distance from origin to forest boundary.12 Units of analysis for NV UM survey respondents and for population socio- demographic characteristics represent two datasets. In the case of the former, visitor origins are assumed to be the reported ZIP codes of the NVUM respondents. For the latter, census block groups were selected as the origin unit of analysis. ZIP codes could not be used in this study as the origin unit of analysis for the population data for several '2 Due to large percentage errors in distance estimation, the 0 — 5 mile band was joined with the 5 — 10 mile band to form one 0 — 10 mile distance band. 63 Table 4. Distance bands used in Local and Mid-distance recreation use models. Distance from origin to forest boundary Band Label Local Segment Bands Origin within forest boundary 0 0 S 10 miles 10 ' 10 S 15 miles 15 15 S 20 miles 20 20 S 25 miles 25 25 S 30 miles 30 30 S 35 miles 35 Mid-distance Segment Bands 35 S 40 miles 40 40 S 45 miles 45 45 S 50 miles 50 50 S 65 miles 65 65 S 80 miles 80 80 S 95 miles 95 95 S 110 miles 110 110 S125 miles 125 125 S 140 miles 140 140 S 155 miles 155 155 S 170 miles 170 170 S 185 miles 185 185 S 200 miles 200 reasons: 1) comprehensive spatial databases of ZIP codes are expensive to obtain at this scale of analysis, 2) US. Census data are not reported by postal ZIP code (Census data are reported by ZIP code tabulation areas rather than postal ZIP codes), and 3) ZIP codes may apply only to businesses or PO. boxes. Census block groups were selected as the unit of analysis for population data because the 2000 US. Census data are reported by census block group and these data are in the public domain. 64 The impact of using two datasets is limited as both NV UM origins and population origins are ultimately aggregated by distance bands measured from the respective origin centroid to the forest boundary. Additionally, the recreation use models for Local and Mid-distance visitor segments are estimated incorporating observations from all forests in the respective study regions—further lessening the impact of any errors at the origin level. Despite these controlling factors, this difference remains a potential measurement error in constructing the recreation use models. The distances from origins to destinations for the NVUM survey respondents were calculated from the NVUM variable “fordist”. This straight-line distance was calculated in ArcView3.2a using the “distance by ID” extension (Jenness, 2004) and spatial databases of the geographic centroids of ZIP codes reported by NVUM respondents and the national forest boundaries. The projection for this calculation was Albers Equal Area Conic, distance units equal to miles. “Fordist” was constructed to classify visitors into local and non-local groups for NVUM economic analysis and has been used frequently in analysis of the NVUM data. For a detailed discussion of the fordist variable and its calculation see Stynes and others (2003). The distances from census block group origins to forest destinations were estimated using a straight-line calculation that incorporated an adjustment for barriers to travel (i.e. large lakes, major rivers). This procedure reduced the potential for assigning to the wrong zones the band population estimates of individual forests. First, straight-line distances were computed for all origin/destination pairs (census block group centroid to 65 national forest boundary) with an intervening distance of less than 200 miles. Second, straight-line routes crossing major lakes and rivers within the study area (Figure 4) were identified. Third, the barrier-adjusted Euclidean distance (BED) was calculated for each of these origin/destination pairs using the procedure below. As with NV UM distance, the analysis projection was Albers Equal Area Conic, distance units equal to miles. '4 # A la... M ‘" 1""ng Figure 4. River and lake barriers to travel in the study area. (Source: ESRI, Redlands, CA) 3 crossing the The latitudes and longitudes of all road bridges and car ferriesl barriers shown in Figure 4 were identified using Delorrne Street Atlas 2004 were imported to ArcView 3.2a and converted to the analysis projection. Points allowing '3 The car ferries crossing the Great Lakes were not included in the constructed bridge and ferry database since they charge significantly greater fees than the other general river and/or publicly subsidized ferries. 66 travelers to circumvent the Great Lakes (e. g. a point south of the southern tip of Lake Michigan) were digitized directly in ArcView 3.2a. Barrier-adjusted Euclidean distance was computed by first identifying the barrier crossing nearest the intersection of the Euclidean route and the barrier. Next, the Euclidean distances from the origin to the crossing identified in the previous step and from the crossing to the nearest boundary on the destination forest were determined. Lastly, these two distances were combined to form BED (Figure 5). Origin O 1% National Forest A/ Transportation Barrier II Bridge \ Euclidean Distance '\ Barrier-adj usted Distance Figure 5. Euclidean and Barrier-adjusted Distance Estimation. Mountainous areas are often perceived as barriers to travel. However, in the context of this analysis, mountainous regions do generally not meet the definition of barriers, namely “completely inhibiting travel”. While often circuitous, travel through mountainous regions is generally possible. Individual mountains may serve as barriers to 67 travel; however, identifying individual barrier mountains located within the study area as well as the appropriate bypass points is beyond the scope of this analysis. Locgl Visitor Recreation Use Model The Local population is comprised of those individuals living within 30 miles of the forest boundary. Total Local segment visitation includes the recreation use of permanent as well as seasonal Local residents. The recreation use of these two groups was modeled separately due primarily to differences in how the US. Census Bureau quantifies permanent and seasonal residents. Permanent residents are quantified on a person basis while the seasonal population is quantified on a household basis. The permanent resident Local model is presented first followed by the seasonal resident Local model. Conceptually, the visitation of the permanent Local visitor segment is a function of the local population, the rate of national forest recreation participation, and the annual visit frequencies of Local visitors. The number of Local segment visits to an individual national forest (Tqu) is computed as I TVLj = 201019,-j * Pan,- * AVij), i =1 where Popij is the local population within each distance band i of forest j, Part, is the percentage of the population in distance band i that participates in national forest recreation estimated at the regional-level, and AVij is the number of annual visits of 68 participating recreation users in distance band i for forest j. The populations located in five distance bands around each national forest in the study regions (Popij) were estimated from the 2000 US. Census data, spatial databases of census block group geographic centroids (U .S. Census, 2004b) and national forest boundaries (USDA FS, 2000) (Figure 6). . Census Block Centroids - National Forest [:| Distance Bands Figure 6. Conceptualization of quantifying a forest’s local population. In some portions of the study area (particularly in Region 2) local populations are located in proximity to multiple administrative national forests. Conceptually, recreation use by this segment is a function of convenient access to the natural resource. Given this, it is assumed that visitors located within 30 miles of multiple national forests would 69 choose to recreate at the closest. To that end, census block groups located proximate to more than one national forest were included only in the population counts of the closest national forest. In cases where census block group centroids were located at equal band distance from more than one administrative national forest the block group population was split evenly between forests. In addition to other national forests, some census block groups are proximate to recreation opportunities managed by other public land agencies. In these cases, it is assumed that the availability of substitutes is accounted for in the annual visitation rates of NVUM respondents sampled on that forest. AVij was estimated from the NV UM survey data using the nfv12mo variable, weighting cases with the visit weights (VisWt). Since NVUM respondents were asked to not include the current visit, one visit was added to the reported nfv12mo to compute nfv12mo+1. To preserve local differences, AVij values were obtained by aggregating respondents across forests located in proximity to one another (Table 5). If less than 30 NVUM cases occurred within a band after forest aggregation, the regional AV, was substituted for AVij. The participation rates of Local visitors were determined at the regional-level based upon the NVUM use estimate for Local visitors, the annual visit frequency of Local visitors, and the local permanent population. Formally, the participation rates of local visitors within band i (Parti) were estimated as _ TVi (POPi * AVI) ’ Part,- 70 where Pop, is the study region population of band i, TV, is the NVUM regional use estimate for visitors from distance band i and AV, is the average number of annual visits by visitors from band i estimated at the regional-level using the approach outlined above. Table 5. Forest aggregation for estimating AV”. Reg'gn 2 Region 9 Bighorn National Forest Hoosier National Forest Shoshone National Forest Black Hills National Forest Nebraska National Forest Arapaho and Roosevelt National Forests Medicine Bow National Forest Pike and San Isabel National Forests Grand Mesa, Uncompahgre and Gunnison National Forests Rio Grande National Forest Routt National Forest San Juan National Forest White River National Forest Mark Twain National Forest Shawnee National Forest Allegheny National Forest Monongahela National Forest Wayne National Forest Green Mountain And Finger Lakes National Forests White Mountain National Forest Hiawatha National Forest Huron-Manistee National Forest Ottawa National Forest Chippewa National Forest Chequamegon / Nicolet National Forest Superior National Forest As constructed, a principal assumption of the permanent resident Local model is that individuals within USDA FS regions participate in national forest recreation with the same proclivity as other individuals located at equal distance from forests in that region. It is assumed that much of the forest—level variation in recreation use due to climate, locally-popular recreation pursuits, and natural resource features is captured in the annual visit frequencies. 71 In areas where seasonal homeownership is common, permanent resident population figures underestimate the effective local population. Based on data obtained in the 2000 Census, approximately 570,000 seasonal homes are located within 30 miles of national forests in the study area—80,000 in Region 2 and 490,000 in Region 9. Seasonal homes located near national forest local areas are most common in the Great Lakes and northeastern portions of Region 9 and along the front range of the Rocky Mountains in Region 2 (Figure 7). Given the potential level of recreation use by seasonal homeowners and their family and friends, it is important to include seasonal homes in any model of national forest visitation. »~..L .... Percent Seasonal Homes - 67 - 100 Percent - 34 - 66 Percent iii... 1 - 33 Percent No Seasonal Homes I: Regional Boundaries Figure 7. Percent of housing units classified as seasonal homes in Census 2000 for census block groups within 30 miles of national forests in USDA FS regions 2 and 9. 72 Unfortunately, very little information concerning seasonal homeowners can be gathered from the NV UM survey data. There is no way to identify seasonal homeowners from surveys completed in NVUM years 1 — 3 and only a partial enumeration of seasonal homeowners is feasible using Year 4 NVUM survey data. As such, it was impossible to estimate participation rates and annual visit frequencies by distance band (as in the permanent resident model) for those owning seasonal homes in the local area using the NVUM survey data. Lacking this ability, it was assumed that the participation rates and annual visit frequencies of seasonal local residents were the same as permanent local residents. The US. Census Bureau reports the number of seasonal housing units. Thus, multiplying participation rates by the number of households will underestimate total seasonal homeowner visitation as there are frequently multiple individuals visiting any single seasonal home. To correct for this, the average party size of seasonal homeowner respondents (2.6 individuals) was identified from partial sample of seasonal homeowners available in the Year 4 NVUM data.14 The resulting model to predict the total visitation of seasonal homeowners in the local area of forest j (T VLSj) is I TVLSJ- = 25H,-j * Pan,- *AVij * PS, i=1 where SH,, is the number of seasonal homes in band i of forest j, Part, is as defined above, AV,, is the annual visit frequency of local permanent residents in band i of forest j, and PS is the average party size of seasonal homeowner parties. The numbers of seasonal 14 Due to small sample size, mean party size was estimated from all seasonal homeowners sampled nationally using Exth. 73 homes located in the five distance bands around each national forest in the study regions (SI-Iij) were estimated from the 2000 US. Census data following the same approach used for permanent local residents. As in the permanent resident model, census block groups proximate to multiple national forests contributed to the local seasonal population of only the closest national forest. Mid-distance Visitor Recreiion Use Models Of the three distance segments, the recreation use of Mid-distance segment visitors best matches the assumptions and conditions for successful application of a travel cost model. Visits completed by individuals within this segment are likely to be single destination trips, it is likely that these visitors receive no benefits during travel to the recreation location, and sufficient variation in travel cost exists within the segment (Loomis and Walsh, 1997). It is difficult to apply a travel cost model to model the recreation use of the Local segment since distance band travel distances are very similar within that segment (due to their proximity to the forest and the selection of a 30—mile boundary). Use of a travel cost model for Long-distance recreation use is probably inappropriate given that these visitors likely receive some benefit from traveling to the area and trips completed by this segment are more likely to include multiple destinations. Mid-distance visitors may complete day trips to the national forest or trips to the national forest that include overnight stays either on or off the forest. Additionally, these visitors may complete national forest visits that are secondary to some other trip purpose. 74 A confounding issue in developing travel cost models is variation in time spent at the recreation destination. Variation in on-site time may change the relationships between travel cost and recreation use (Loomis and Walsh, 1997). Specifically, visitors who spend a greater amount of time at the recreation area may be willing to incur greater costs than those spending a shorter time. In addition to travel cost, the relationships between recreation participation and the other factors influencing recreation (e. g. resources characteristics, substitutes, etc.) may vary depending upon the type of trip and time spent on site. With these considerations in mind separate travel cost models were developed for Mid-distance segment day trips and Mid-distance segment overnight trips within each study region. Consistent with theory, “not primary” Mid-distance recreation use was excluded from the constructed zonal travel cost models. Functional forms commonly chosen for use in zonal travel cost models include single—log ordinary least squares (OLS), the Tobit model, and the Heckman two-stage model (e.g. Loomis et al., 1995). The single log OLS model is attractive because 1) interpretation of the independent variable coefficients estimated via OLS is straightforward and 2) use of a single-log dependent variable for visits yields non- negative model predictions of recreation use. A drawback to the single log OLS is that the model cannot be estimated with zero-value observations as the natural log of zero is undefined. The typical solution to this problem is to add one visit (or a small value) to all observations to insure computability (e. g. English and Bowker, 1996). In the Tobit model and two-stage Heckman model, recreation use is modeled as both a function of the likelihood of participation as well as the number of visits of participating populations. 75 The Tobit model incorporates both the likelihood of participation and level of recreation use in one model while participation and recreation use level is modeled separately in the Heckman two-stage model. Both model formulations explicitly incorporate zero-value observations. The multi-site travel cost models were estimated for each trip-type/region combination following the general approach to empirical specification adopted by English and Bowker (1996). Models for day trips and overnight trips were estimated using both single log ordinary least squares (OLS) and Tobit formulations. Models were estimated with several different formulations of the dependent variable. Dependent variables used in the OLS formulation included ln(distance band visits,,) and ln(distance band visits/population (1,000’s)). Dependent variables in the Tobit model formulation were distance band VISItSij and distance band visits,,/capita. In all cases, “distance band visits” is the number of day or overnight trips originating from distance band i of forest j. The selected dependent variables were chosen for consistency with theory and for model computability. All models were estimated using Eviews 4.1® (Quantitative Micro Software LLC, Irvine, CA). When using a natural logarithm dependent variable, English and Bowker (1996) added one visit to each observation to insure model computability. Attempts to make a similar adjustment in this study resulted in poorly performing models that yielded predictions of forest-level recreation use that were unrealistically low. Given this result, this adjustment was not made and zero visit observations were dropped from the OLS formulations. The impact of excluding zero visit observations from the OLS models 76 should be minimal as the range of observed number of visits is large and there are a number of observations with low recreation use (e. g. < 250 visits). Zero visit observations are included in the Tobit model formulation and comparison of the predictive ability of between the two study models should help to identify the impact of dropping zero visit observations in the single-log OLS model. Dependent variables for the Mid-distance segment models were constructed using the NVUM estimates of the number of forest-level day and overnight visits originating from the Mid-distance segment distance bands identified previously. In the first cycle of NVUM only respondents to the economic survey can be classified by trip-type. As such, a three-step process was required to estimate band-level recreation use for the trip-types under consideration. First, the forest—level recreation use of all trip types (day, overnight, not-primary trips) in each Mid-distance segment distance band was estimated for all forests in the study area (Appendix Table B-l). Second, the percentage of day, overnight, and not primary recreation visits for each forest—level distance band was determined from the NVUM economic subsample and the NVEXPAND weights (Appendix Tables B-2 and B-3). In several cases no economic subsample observations occurred in a given forest/distance band combination. In these cases, the regional trip-type percentages for the distance band were substituted. Third, forest-level visit estimates by trip-type and band were computed as the product of the forest-level visit estimate obtained in step one and the trip-type percentages obtained in step two (Appendix Tables B-4 and B-5). 77 The forest-level recreation use of Mid-distance day trip visitors (NDDJ) and Mid- distance overnight trip visitors (MDO,) in Region 2 was modeled as: MDDj =F(P0p,-j,D-- I SNF-- SBLM SNPS w a" 11* .7. RUCU’AJ’UJ') ijs MDOJ- = F(Pop,j,D 1,-J-,SNF~ SBLM ,1, ,j,SNPS,-j,RUC-- Aj,UJ-), i!" U’ respectively. Likewise, the forest-level recreation use of Mid-distance day trip visitors (MDD,) and Mid-distance overnight trip visitors (MDO,) in Region 9 was modeled as: I SNF- MDDJ- =F(P0p,-j,D ,1, ,j, ,j, SNPSij,RUC,-J-,Aj,UJ-), SNF-- MDOj=F(P0p,J-,Dij,l ,J, SNPSU- ,RUC-- will». ij . respectively. Independent variables include the population in band i of forest j (Pop,,), the maximum distance from band i to forest j (D,,), the median income of band i of forest j (I,,), the distance to the nearest other national forest land of band i of forest j (SNF,,), the straight-line distance to the nearest BLM land of band i of forest j (SBLM,,), the straight- line distance to the nearest NPS land of band i of forest j (SBLM,,), the Rural Urban Continuum (RUC) code of band i of forest j (RUC,,), the acreage of forest j (Aj), and the number of major separate units on forest j (U j). The independent variables for each distance band are computed by aggregating the respective values of all census block groups located within each distance band. Additionally, band values for variables I,,, SNF,,, SBLM,,, SNPS,,, and RUC,, were computed by weighting each census block group observation by the census block group population—forming population-weighted band values for these variables. The sources of data for the independent variables are as follows. Census block group population estimates and median income figures were obtained from the US. 78 Census Bureau (2004a). Distances from each census block group centroid to the nearest other national forest land15 , the nearest BLM land, and the nearest NPS land were determined using ArcView 3.2a based upon the spatial datasets obtained from the sources identified in Table 6. The RUC codes for each census block group were determined from the database of county-level RUC codes developed by Beale (2004). Each census block group is located within a single county. RUC codes quantify the rurality of individual counties based upon: 1) whether the county is classified as a metropolitan area by the Office of Management and Budget, 2) whether it is adjacent to a metropolitan area if a non-metropolitan area, and 3) the county’s urban population. Increasing RUC values indicate increasing rurality.16 The acreages of individual national forests were obtained from the USDA FS 2004 Land Areas Report (USDA FS, 2004). The number of separate units was determined based upon a subjective count of the number of major spatial units managed by each forest. For example, the number of separate units managed by the Hiawatha NF is two while the number of separate units managed by the Arapaho- Roosevelt NF is one. Table 6. Spatial data sources for publicly-owned land. Land Agency Data Source USDA Forest Service USDA FS, 2000 National Park Service NPS, 2002 Bureau of Land Management National Atlas of the US, 2005 -_ '5 National forests located in regions other than the study region were included in this computation. RUC code was included in the models as a continuous variable. Inclusion of RUC as a dummy variable (1 = metropolitan area) yielded equivalent results. 79 Evaluation of Model Predictive Ability The predictive ability of Local and Mid-distance recreation use models were evaluated in three tiers. First, the model predictions of forest-level recreation use for each segment (i.e. Local and Mid-distance segments) were compared to the NVUM estimates of forest-level recreation use for those segments. Second, the summed model predictions of forest-level Local and Mid-distance recreation use were compared to the NVUM estimates of Local and Mid-distance recreation use. Finally, the forest-level percentages of Local recreation use (as a function of predicted Local and Mid-distance recreation use) were compared to the forest-level NVUM estimates of the percentage of Local recreation use (as a function of the NVUM estimates of Local and Mid-distance recreation use). NVUM offers a clear advantage over previous USDA visitor sampling programs in that it is implemented on all forests in the NPS via a consistent approach. Additionally, the visitor survey data collected via the NVUM process represent the most comprehensive USDA FS visitor survey dataset available to date. However, the dispersed nature of USDA FS recreation use, the large land areas managed, and the number of access points to individual national forests make USDA FS visitor sampling challenging. Ultimately, the NVUM recreation use figures are estimates of actual visitation, the accuracy of which is dependent upon the appropriateness of the sampling design and application of the visitor sampling protocols. 80 The accuracy of NV UM forest-level recreation use estimates for particular groups of recreationists (e. g. Locals, snowmobilers, visitors from a given distance band, etc) are particularly dependent upon the ability of NVUM to obtain surveys from visitors in those groups and the appropriateness of weights to expand those samples of visitors to total recreation use. Failure to obtain surveys from visitors in certain visitor groups may lead to recreation use estimates that appear unreasonably low. For example, since no snowmobilers were surveyed on the Huron-Manistee NF (located in the northern lower peninsula of Michigan) the NVUM recreation use estimate for that national forest indicates there was no snowmobile recreation use during the NVUM sample year. Conversely, obtaining a large number of surveys from individuals in a given visitor group may lead to recreation use estimates that seem questionably high. Biases in the obtaining representative visitor samples could influence the estimates of distance-based segment use for individual national forests. Stynes and others (2003) have raised concerns as to whether the NVEXPAND weights (employed to estimate total recreation use) are appropriate when applied to visitor characteristics (including group membership). Of particular concern is the potential for individual respondents with high NVEXPAND weights to exert undue influence over results of analyses of visitor characteristics (including the recreation use of visitor groups) (Stynes et al., 2003). The influence that individual observations with large NVEXPAND weights have over the results increases as sample sizes decrease. The potential impacts from observations with large NVEXPAND weights (if they occur) will 81 primarily influence estimation of the recreation use models—when a small number of cases are used to identify recreation use by distance band. Despite the potential limitations, N VUM recreation use estimates are the only comprehensive set of USDA FS recreation use figures obtained via a consistent national- level program. As such, the NVUM use figures are deemed the best figures to evaluate the ability of the constructed models to predict distance-segment recreation use. In all evaluations, statistical uncertainty in the NVUM use figures is reflected in the 80% confidence intervals constructed around the estimates. NVUM estimates of forest-level Local recreation use range from 57,000 to 2 million visits in Region 2 and from 79,000 to 727,000 visits in Region 9 (Table 7). The forest-level estimates of Local visitor recreation use were obtained via a multi-step process. This is required because the published NVUM estimates of forest-level recreation use differ from recreation use estimates obtained by using the N VUM survey data and NVEXPAND weights alone. First the forest-level recreation use associated with visitors of known origin not engaged in downhill skiing and recreating on national forest land (termed “modeled recreation use”) were estimated for each study area forest (Appendix Table B-6). These figures were estimated by multiplying the published NVUM estimates of forest-level recreation use by the NVEXPAND weighted percentage of NVUM survey respondents meeting the modeled recreation use criteria on each forest. Next, the forest-level modeled recreation use figures were multiplied by the 82 NVEXPAND weighted percentage of NVUM survey respondents classified in the Local visitor segment on each forest (Appendix Table B-7). Table 7. NVUM Estimates of Local Visitor Recreation Use for USDA FS Regions and National Forests (NF) in the Study Area. NVUM Estimate Lower Estimate Upper Estimate Bighorn NF 273,995 198,388 355,876 Black Hills NF 506,796 398,248 622,623 Grand Mesa, U.,G. NF 1,388,635 953,697 1,848,310 Medicine Bow NF 288,978 236,296 345,489 Nebraska NF 56,678 40,789 74,836 Pike San Isabel NF 2,045,153 1,714,877 2,389,610 Routt NF 587,221 501,804 679,250 Shoshone NF 258,705 214,246 306,760 White River NF 1,279,373 1,171,184 1,392,239 Region 2 6,821,822 6,333,542 7,321,118 Allegheny NF 555,824 381,840 744,479 Chequamegon-Nicolet NF 595,152 389,868 822,667 Chippewa NF 726,966 592,578 872,045 Hoosier NF 365,668 295,388 440,773 Huron-Manistee NF 212,354 129,357 307,506 Mark Twain NF 412,738 350,419 477,656 Monongahela NF 262,522 214,339 314,594 Ottawa NF 78,848 61,991 97,583 Shawnee NF 316,315 270,318 365,126 Wayne NF 248,914 200,208 301,341 Region 9 4,102,540 3,734,356 4,480,997 In addition to statistical uncertainty in the forest-level estimates of total recreation use, there is also statistical uncertainty in the forest-level estimates of Local recreation use. The error in this use estimate is a combination of error in estimating total recreation use (estimated by NVUM personnel) and error in estimating the percentage of forest- level recreation use associated with Local visitors. Confidence intervals (at the 80% 83 level) around the percentages of forest-level Local visitor recreation use were computed as pil.28l7*0'p, where p = the sample estimate of the percentage and 6,, is the standard error of the Local _ [pa-p) O'p— —N . The upper estimates of forest-level Local recreation use (Table 7) were obtained by percentage: multiplying the upper estimates of total recreation use (Appendix Table B-6) by the upper estimates of the forest-level Local percentage (Appendix Table B-7). Likewise, the lower estimates of forest-level Local recreation use (Table 7) were obtained by multiplying the lower estimates of modeled recreation use (Appendix Table B-6) by the lower estimates of the forest-level Local percentage (Appendix Table B-7). Forest-level estimates of Mid-distance recreation use were computed following the same approach as above with one addition. The models developed to predict Mid- distance recreation use exclude “not primary” recreation use. To be consistent, estimates of forest-level Mid-distance recreation use also exclude not-primary recreation use (Table 8). Specifically, forest-level estimates of modeled recreation use (Appendix Table B-6) were multiplied by forest-level estimates of the percentage of Mid-distance use and by the forest-level estimates of Mid-distance “primary purpose” recreation use (Appendix Table B-8). Upper and lower percentage estimates were not constructed around the percentages of “primary purpose” recreation use. 84 Table 8. NVUM Estimates of Mid-distance Visitor Use for Regions and National Forests (NF) in the Study Area.ll NVUM Estimate Lower Estimate Upper Estimate Bighorn NF 99,692 68,193 135,990 Black Hills NF 27,680 15,875 42,481 Grand Mesa, U.,G. NF 312,161 199,280 442,999 Medicine Bow NF 177,220 142,030 215,708 Nebraska NF 32,674 22,516 44,637 Pike San Isabel NF 179,644 130,947 235,966 Routt NF 392,194 328,547 461,412 Shoshone NF 53,634 39,561 69,781 White River NF 702,753 631,665 777,533 Region 2 1,969,848 1,790,973 2,156,547 Allegheny NF 340,138 227,368 466,716 Chequamegon-Nicolet NF 833,607 556,978 1,131,889 Chippewa NF 442,223 349,658 544,799 Hoosier NF 126,631 94,776 162,927 Huron-Manistee NF 456,977 292,176 634,072 Mark Twain NF 55,123 41,207 71,043 Monongahela NF 246,185 203,838 291,264 Ottawa NF 32,836 25,618 40,898 Shawnee NF 133,763 109,090 160,889 Wayne NF 130,716 100,655 164,355 lLegion 9 3,000,003 2,720,308 3,288,796 a Day and overnight primary purpose visits only The NVUM estimates of forest-level total Local and Mid-distance recreation use (Table 9) were calculated as the sum of the NVUM estimates of Local and Mid-distance recreation use (the “NVUM estimate” columns only) in Tables 9 and 10. The upper and lower estimates around these values figures were calculated using the forest-level 80% error rates reported by N VUM personnel. Finally, the NVUM estimates of the forest- level percentages of Local use (Table 10) were obtained by dividing the NVUM estimates of Local use (Table 7) by the NVUM estimates of summed Local and Mid- 85 estimates. distance recreation use (Table 9). No confidence intervals were constructed around these Table 9. NVUM Estimates of Summed Local and Mid-distance Recreation Use.“ NVUM Estimate Lower Estimate Upper Estimate Bighorn NF 99,692 68,193 135,990 Black Hills NF 27,680 15,875 42,481 Grand Mesa, U.,G. NF 312,161 199,280 442,999 Medicine Bow NF 177,220 142,030 215,708 Nebraska NF 32,674 22,516 44,637 Pike San Isabel NF 179,644 130,947 235,966 Routt NF 392,194 328,547 461,412 Shoshone NF 53,634 39,561 69,781 White River NF 702,753 631,665 777,533 Region 2 1,969,848 1,790,973 2,156,547 Allegheny NF 340,138 227,368 466,716 Chequamegon-Nicolet NF 833,607 556,978 1,131,889 Chippewa NF 442,223 349,658 544,799 Hoosier NF 126,631 94,776 162,927 Huron-Manistee NF 456,977 292,176 634,072 Mark Twain NF 55,123 41,207 71,043 Monongahela NF 246,185 203,838 291,264 Ottawa NF 32,836 25,618 40,898 Shawnee NF 133,763 109,090 160,889 Wayne NF 130,716 100,655 164,355 Region 9 3,000,003 2,720,308 3,288,796 a Mid-distance recreation use includes only primary purpose visits. 86 Table 10. NVUM Estimates of the Percentage of Local Visitor use as a Function of Summed Local and Mid-distance Use. NVUM Bighorn NF 73% Black Hills NF 95% Grand Mesa, U.,G. NF 82% Medicine Bow NF 62% Nebraska NF 63% Pike San Isabel NF 92% Routt NF 60% Shoshone NF 83% White River NF 65% Region 2 78% Allegheny NF 62% Chequamegon-Nicolet NF 42% Chippewa NF 62% Hoosier NF 74% Huron-Manistee NF 32% Mark Twain NF 88% Monongahela NF 52% Ottawa NF 71% Shawnee NF 70% Wayne NF 66% Region 9 58% 87 CHAPTER 4 RESULTS AND DISCUSSION Introduction Results of analyses undertaken to achieve the objectives identified in Chapter 1 are detailed in this chapter. The first half is devoted to characterizing the recreation characteristics of Local, Mid-distance, and Long-distance visitors (Objective 1). Included in this portion of the chapter are the results from statistical comparisons between study regions and between the distance segments (Objectives 2 and 3). The second half of the chapter focuses on the models developed to predict the recreation use of Local and Mid- distance visitors in USDA FS Regions 2 and 9 (Objective 4). The parameters and performance of the constructed models are described in detail. Characteristics of Distance Segmented Visitors Segmentgtion of NVUM Survev Respondents Of the 19,146 respondents sampled under NVUM in USDA FS regions 2 and 9, 39% (7,378) were classified as Local visitors, 27% (5,210) as Mid-distance visitors, and 20% (3,788) as Long-distance visitors (Table 11). Greater percentages of survey respondents in Region 2 are classified as Local or Long-distance visitors compared to 88 Region 9. Correspondingly, a greater percentage of survey respondents are Mid-distance visitors in Region 9. Nationwide, approximately 14% of NVUM survey respondents failed to report the ZIP code of their permanent residence or reported an invalid ZIP code. Consistent with this, 14% of non-foreign NVUM survey respondents (2,770) sampled in the study area failed to report a valid ZIP code (or reported an invalid ZIP code). These respondents are categorized as “missing” and excluded from further analysis. NVUM survey respondents originating from foreign countries are classified as Long-distance visitors. Table 11. Number of NVUM Survey Respondents by Distance Segment and USDA FS Study Region. Region 2 Region 9 Total Distance Full Economic Full Economic Full Economic Segment Sample Sample Sample Sample Sample Sample Missing 1,304 79 1,466 62 2,770 141 Local 4,246 1,151 3,132 789 7,378 1,940 Mid-dist. 1,938 530 3,272 878 5,210 1,408 Long-dist. 2,705 691 1,083 Zfl 3,788 220 Total 10,193 2,451 8,953 2,028 19,146 4,479 Objectives 1 through 3 relate to characterizing visitors within the distance segments and completing statistical comparisons. Statistical comparisons are completed between study regions, within segments, and between the distance segments themselves within study regions. Regional comparisons of distance segments are presented first followed by comparisons between segments. 89 Regional Comparisons within Segments Local visitors comprise the largest share of total recreation use in both regions— 53% in Region 2 and 49% in Region 9 (Table 12).1 In Region 9, 42% of total recreation use is associated with Mid-distance visitors, substantially greater than the 21% in Region 2. Conversely, Long-distance visitors account for 27% of all use in Region 2 and only 9% of visitation to Region 9 national forests. Table 12. Percent of Total Recreation Use by Distance Segment and USDA FS Study Regjon.a Region 2 Region 9 Local 53% 49% Mid-distance 21% 42% Lon g-distance 27% 9_% Total 100% 100% ' Estimated based upon the full NVUM survey sample and NVEXPAND weights. Differences between regions in the percentages of Mid-distance and Long- distance visitors may be explained, in part, by three factors. First, the populations surrounding national forests in Region 9 are greater and more uniformly spatially distributed than populations around Region 2 national forests—thereby creating a greater number of potential Mid-distance visitors. Second, the unique natural features located in and around Region 2 national forests, particularly those located in Colorado, attract a greater number recreation users traveling more than 200 miles from home. In particular, opportunities for downhill skiing attract a significant number of these visitors. Downhill ’ Removing skiers, Local visitation represents 63% of use in Region 2 and 50% of use in Region 9. These non-skier shares are consistent with Figure 3 in Chapter 3. 90 skiers account for approximately 10.7 million visits to Region 2 national forests with more than 40% of these visitors traveling greater than 200 miles to reach the forest. In Region 9, only 320,000 visits are made primarily for downhill skiing; of those visits, only 4% resulted in travel of more than 200 miles from home. Third, Region 2 forests likely receive more visitors traveling through the area on their way elsewhere (e. g. traveling to the western US.) than Region 9 forests. Three types of recreation trips are recognized in this study: day trips (Day), trips involving an overnight stay either on the national forest or off the national forest (OVN/OVNNF), and trips where the primary purpose was something other than recreating on the national forest (Not Primary). Aggregating across distance segments, a significant difference exists between study regions in frequency of trip-type (Table 13).2 Within distance segments, a statistical difference betweens study regions in trip type frequency exists only for the Mid-distance segment. In both regions, 72% of Local visits are day trips. Overnight trips represent about 25% of visits while “not primary” trips comprise only a very small percent of all visits. Trip-type propensity is nearly identical between study regions. Mid-distance recreation use most frequently involves an overnight stay away from home (Table 13). However, approximately 20% of Mid—distance visits in each study region are day trips. Compared to Local visits, Mid-distance visits are more frequently “not primary” trips. 2 Detailed statistical tables for all statistical tests in this Chapter are available from the author. 91 Table 13. Trip-type Segment Shares by Distance Segment and USDA FS Study Region.“ Day OVN/OVNNF" Not Primary‘ Total P-valued Local Region 2 72% 24% 3% 100% 0.77 Region 9 72% 25% 4% 100% Mid-dist. Region 2 24% 66% 11% 100% 0.01 Region 9 17% 70% 14% 100% Long-dist. Region 2 8% 59% 34% 100% 0.60 Region 9 6% 60% 34% 100% Total Region 2 46% 42% 12% 100% 0.00 Reg’on 9 40% 48% 12% 100% 3‘ Estimated based upon the economic NVUM economic sample and Exth. b Recreation trips involving an overnight stay in the local forest area or on the national forest. c Trips not made primarily to recreate on the national forest. d Statistical comparisons between regions within distance segments were completed using contingency table analysis. The majority of Long-distance visits in both regions are classified as Overnight trips. In both study regions, “Not Primary” trips comprise 34% of Long-distance visitor recreation use. In NVUM year 4, survey respondents were asked to identify the primary reason of their trip away from home.3 In Region 2, 74% of “Not Primary” Long-distance visitors reported “visiting other recreation areas” as the primary purpose of being away from home. “Not Primary” Long-distance visitors in Region 9 more commonly cited “business and family” (35%) as their primary trip purpose than the same visitors in Region 2. Given differences (both statistical and practical) between study regions in the distribution of trip-types within the three distance segments, statistical tests of regional 3 NVUM survey forms are available from Susan Kocis, Field Coordinator National Visitor Use Monitoring Program, USDA Forest Service. 92 differences in segment characteristics were completed incorporating trip-types using only the NVUM economic sub-sample (Tables 14 and 15). For the most part, after controlling for differences in distance segment and trip type, visitors to national forests in USDA FS Regions 2 and 9 have recreation behaviors that are not statistically different. The greatest number of statistical differences between study regions occurs in the Local visitor segment. Local visitors in the two regions differ in the number of annual visits, the duration of the national forest visit, the number of people per vehicle and national forest arrival day. Across all trip-types, Local visitors in Region 2 complete fewer annual visits than those in Region 9 (Table 14).4 Region 2 Local visitors on day trips spend a longer period on the national forest and recreate in smaller parties. On overnight trips, Local visitors in Region 2 spend a shorter period of time on-forest than their Region 9 counterparts.5 lastly, Local day trip visitors in Region 2 tend to visit more frequently throughout the week than their Region 9 counterparts (Table 15). The difference in annual visitation rates between study regions is slight. This difference may be largely driven by the spatial arrangement of populations around national forests in the study regions. In Region 9, a slightly greater percentage of total Local segment use is associated with visitors living within 10 miles of the national forest boundary than in Region 2 (76% compared to 68%).6 Assuming that those living farther ’ Annual visit median values for Local visitors in Regions 2 and 9 are 2.0 and 3.0, respectively 5 On-forest duration statistics are likely influenced by the ability of NV UM survey respondents to identify and differentiate lands managed by the USDA FS. 6 Based upon NVUM sample of Local visitors and NVEXPAND weights. 93 Table 14. Mean Values for Visitor Characteristics of Interest by distance Segment, Trip-type, and USDA FS RegLon. Visit Distance Study Annual duration Number Number of People! _§e_gment Trip type Region visits“ (hours)" of Sites Activities Vehicle Local Day 2 3.0“ 1. ° 32° 21° Day 9 2.6“ 1.1“ 31° 23° OVN/OVNNF 2 240°l 1.3° 4.6“ 25° OVN/OVNNF 9 40.2d 13° 4.6“ 23° Not primary 2 26f 1.3f 4.2f 22° Not primary 9 3.2“ 1.1f 34" 22° All trip-types 2 6.1‘1 All trip-types 9 6.0d Mid-dist. Day 2 4.3“ 1.1“ 2.9“ 2.4“ Day 9 4.2“ 1.2c 3.1“ 2.2“ OVN/OVNNF 2 222° 13° 42° 25° OVN/OVNNF 9 280° 1.4“ 4.4° 25° Not primary 2 3.0“ 1.3“ 4.2“ 2.6“ Not primary 9 3.6“ 1.3“ 3.6“ 2.8“ All trip-types 2 2.6“ All trip-types 9 1.6“ Long-dist. Day 2 20" 1.3r 3.4f 2.4f Day 9 1.9f 1.3f 3.8f 1.8f OVN/OVNNF 2 55° 1.5“ 39° 2.8d OVN/OVNNF 9 8.0“ 1.o° 42° 2.5‘1 Not primary 2 30° 15° 4.8“ 25° Not primary 9 2.2“ 1.4“ 3.6“ 2.4“ All trip-types 2 1.6“ All trip-types 9 1.8“ Note: Statistical tests completed using Mann-Whitney U, Exth. ' Not computed within trip-types since variable corresponds to previous trips of unknown type. VisExpwt used in statistical analyses, “ Median values reported, “ P-value > 0.05, no statistical difference between regions, “ P—value < 0.05, statistical tests between regions, “ P-value < 0.01, statistical tests between regions, f Number of cases in at least one group is less than 30, statistical test not completed. 94 Table 15. Day of National Forest Arrival by Distance Segment, Trip-type, and USDA FS Region.“ Distance P- iegment Trip-type Sun. Mon. Tues. Wed. Thurs. Fri. Sat. value Local Day2 29% 13% 7% 8% 10% 8% 24% 0.001 Day“ 34% 10% 5% 6% 6% 8% 33% OVN/ OVNNF“ 16% 7% 7% 6% 6% 25% 32% 0.265 OVN/ OVNNF" 13% 5% 5% 9% 12% 30% 26% Mid-dist. Day“ 40% 8% 3% 6% 10% 5% 27% 0.553 Day9 37% 10% 4% 4% 4% 7% 35% OVN/ ovr~n~n=2 17% 4% 7% 5% 10% 27% 30% 0.402 OVN/ OVNNF’ 12% 5% 4% 6% 11% 30% 31% Not primary“ 9% 5% 7% 14% 7% 26% 33% 0.064 Not primary“ 28% 8% 8% 3% 8% 14% 31% Long-dist. OVN/ OVNNF“ 15% 13% 15% 14% 8% 15% 20% 0.051 OVN/ OVNNF“ 14% 10% 4% 16% 14% 19% 23% Not primary“ 8% 11% 20% 20% 14% 16% 11% 0.170 Not primary“ 10% 10% 13% 11% 13% 18% 25% “ Segment, trip-type combinations with less than 30 cases in at least one region are excluded. Statistical analysis completed using contingency table analysis and Exth. Comparisons between regions using exposure weights assumes there is no bias in site stratification or the propensity for visitor sampling to occur on a given day(s) between regions. 2 Region 2. 9 Region 9. 95 from national forests visit with less frequency, it is reasonable that Region 2 Local visitors, overall, would have lower mean annual visitation rates. Mid-distance visitors in regions 2 and 9 are not statistically different (within trip- types) from one another in terms of annual visit frequency, on-forest duration, number of sites visited, number of recreation activities participated in during recreation visits, or party size (Table 14).7 Likewise, there are no statistical differences in arrival patterns (Table 15). Statistical similarity between study regions (within this distance segment) contrasts to the commonly held perception that recreation visitor behavior (in terms of these characteristics) differs by USDA FS region. This finding is particularly important, as these visitors (non-locals, whose primary trip purpose is generally visiting the national forest) are frequently the focus of economic impact and resource valuation studies as well as analyses of management alternatives and strategies. Across all trip-types, Long-distance segment visitors in Region 2 complete fewer national forest visits annually than comparable visitors in Region 9 (Table 14). A greater percent of Region 2 Long-distance visitors (68%) complete just one annual visit compared to Region 9 Long-distance visitors (58%).8 It is unclear how many of the Long-distance visitors, in either region, are completing their first trip to the national forest. 7 Comparisons not incorporating trip-type also resulted in no statistical differences in Activity visitor characteristics between study regions. 8 Based upon the NVUM survey sample, weighted by VisExpwt. 96 The prevalence of large parties among Long-distance overnight visitor groups in Region 2 leads to the statistical difference in party size between regions (Table 14). In Region 9, only 1% of recreation use in this segment/trip-type combination is associated with parties of 5 — 7 people. In Region 2, large parties represent 14% of use in this segment. Visitors in large parties in Region 2 primarily participate in general tourism- type activities such as “driving for pleasure” (52%) “nature study” (17%) or “viewing nature” (14%).9 These large parties likely typify general tourist visitors drawn to Region 2 national forests by unique natural features as well as their proximity to urban centers and large national parks. Within distance segments, the primary recreation activities of recreation users are largely similar between the study regions (Table 16). Across all distance segments, a substantially greater percentage of recreation use in Region 2 is associated with downhill skiing. This is expected given the renown and sheer number of ski resorts located in the national forests of Region 2. Within distance segments, Local visitors in Region 2 are more commonly engaged in downhill skiing or biking and less commonly in hunting or fishing than their Region 9 counterparts. Likewise 45% of Mid—distance visitors in Region 2 are downhill skiers compared to only 13% in Region 9. Additionally, Mid- distance visitors in Region 2 are more frequently biking and less frequently fishing, hiking, or cross-country skiing than the same visitors in Region 9. Similarly, a greater percentage of Region 2 Long-distance visitors are engaged in downhill skiing and fewer are hiking or fishing than in Region 9. 9 Based upon NVUM survey respondents reporting between 5 - 7 people per vehicle, weighted by NVEXPAND 97 Table 16. Primary Activity by Distance Seggm and USDA FS Region.“ Local Mid-distance Long-distance Region Region Region Region Region Region 2 9 2 9 2 9 Developed Camping 2% 2% 3% 5% 2% 3% Prim. Camping & Backpacking 2% 1% 1% 4% 0% 3% Resort 0% 1% 0% 1% 0% 4% Picnic 2% 3% 2% 0% 1% 1% Nature-related 8% 7% 5% 7% 10% 13% General/Relaxing 5% 5% 4% 6% , 4% 7% Fishing 7% 13% 3% 9% 3% 7% Hunting 5% 13% 6% 9% 3% 5% OHV use 3% 3% 4% 5% 1% 2% Driving 3% 6% 2% 2% 4% 1% Snowmobile 4% 3% 2% 3% 1% 5% Boating 2% 1% 1% 1% 1% 1% Hiking 19% 16% 7% 13% 11% 25% Biking 7% 2% 6% 1% 6% 0% Downhill skiing 14% 7% 45% 13% 45% 4% Cross-country skiing 5% 7% 1% 8% 0% 3% Other non-motorized 4% 3% 2% 2% 1% 2% Other 3% 4% 1% 1% 1% 4% No primary Activity 4% 1% 2% 3% 2% 3% Multiple primary activities 2% 2% 3% 7% 2% 5% Total 100% 100% 100% 100 % 100% 100% “ Estimated from full NVUM sample, weighted by NVEXPAND. DisLance Segment Comparisons Distance segment visitors were characterized in terms of annual visitation, visit duration, number of sites visited, number of activities, and party size (Table 17). Aggregating across trip-types, Local visitors complete a large number of national forest visits with most visits lasting less than four hours. These visitors recreate at a limited number of sites, generally participating in a moderate number of recreation activities in small recreation parties. Mid-distance visitors generally visit a given national forest up to 98 several times a year, generally spending more than six hours on the national forest. Mid- distance visitors recreate at a greater number of sites, participate in more activities, and have larger recreation parties than Local visitors. Visitors traveling greater than 200 miles to the forest have very low annual visitation rates with on-forest durations of most visits being less than 5 hours. These visitors recreate at the greatest number of sites, participate in the greatest number of activities, and recreate in the largest visitor parties. Table 17. Mean Values for Visitor Characteristics of Interest by Distance Segment, and USDA FS Region. Visit Study Distance Annual duration Number Number of People! Region Sggment visits“ (hours)" of Sites Activities Vehicle 2 Local 6.1I 3.8‘ 1.1I 3.1I 2.11 Mid-dist. 2.61 6.62 1.32 2.82 2.42 Long-dist. 1.62 4.23 1.43 3.03 2.93 9 Local 6.01 3.41 1.21 2.91 1.91 Mid-dist. 1.61 7.32 1.22 3.42 2.42 Long-dist. 1.82 4.01 1.57- 3.92 2.52 Note: K-W ANOVA was used to identify statistical differences between all distance segments within regions. Mann-Whitney U was used to identify specific differences between segments. Analyses were completed using the exposure weights. “ Weighted by VisExpwt, “ Median values reported. 1' 7" 3 Statistically different subgroups within regions. There is a clear difference between distance segments in the number of annual visits reported (Figures 8 and 9). More than half of the Long-distance segment visitors sampled complete only one annual visit to the national forest—approximately 95% complete no more than three annual visits. Approximately 40% and 30% of Mid-distance visitors and Local visitors report only one annual visit, respectively. While both of these segments visit more often than Long-distance visitors, Local visitors more frequently complete a greater number of annual visits than Mid-distance visitors. The 95th percentile 99 :3 l l ' n ' —o—Local —l— Mid-Dist. - -A- - Long-Dist ii N ix? Percent of Respondents s I l I —L 53 0% 5 6 7 8 9 10 11 12 AnnualVlslts Figure 8. Distance Segment Visit Frequency, USDA FS Region 2 (Truncated at 12 Annual Visits, Exposure Weighted). 70% 60% T A. r. 50% __._'._% e— ' s “g ' s C 400k q : I ___E44fi—e—- fl!---—-* * ”v— +Local 3. I _._ Mid-Dist. . .‘p -Long-Dist. 0123456789101112 AnnualVlsits Figure 9. Distance Segment Visit Frequency, USDA FS Region 9 (Truncated at 12 Annual Visits, Exposure Weighted). 100 for Mid-distance visitors is seven annual visits. Comparatively, the 95‘h percentile for Local recreation visitors is 20 annual visits. Within distance segments, substantial variability exists in the duration of on-forest visits. At most, the modes of visit duration for each distance segment are reported by only 19% of respondents (Figures 10 and 11). Though variability is wide, the majority of respondents report visits of 10 hours or less. Local visitors are most likely to have on- forest durations of 1 to 5 hours. Forest visits of greater than 5 hours are reported more frequently by Mid-distance and Long-distance visitors. In both regions, Long-distance visitors are more likely than visitors from the other two segments to have national forest visit durations of one hour or less. 18% 16% 14% 12% 10% 8% 6% Percent of Respondents 4% 2% 0% . _, _2 _.—Local _I— Mid-Dist. ‘ - * -Long-Dist. I ‘. 0 5 10 15 20 Visit Duration Figure 10. Duration of National Forest Visits by Distance Segments, USDA FS Region 2 (Rounded to Nearest Hour, Truncated at 24 hours, ExposureWtd.). 101 —o—Local —.— Mid-Dist. - g- - Long-Dist. Percent of Respondents 0 5 10 15 20 Visit Duration Figure 11. Duration of National Forest Visits by Distance Segments, USDA FS Region 9 (Rounded to Nearest Hour, Truncated at 24 hours, ExposureWtd.). Compared to the divergent patterns between segments in annual visits and visit duration, the numbers of sites visited by visitors within each distance segment are quite similar (Figures 12 and 13). Approximately 90% of Local visitors in both regions visit only one recreation site or recreate only one day in the GFA. The percentages of single site visitors in the Mid-distance and Long-distance visitor segments are slightly lower. In Region 2, Long-distance visitors are more likely to visit a greater number of sites than Mid-distance visitors. In Region 9, propensity for visiting a large number of sites is similar between the two segments. 102 1 00% 90% 80% - 70% 60% 50% 40% 30% Perwnt of Respondents 20% 1 0% 0% —0— Local —.— Mid-Dist. - -A- - Long-Dist. I I I I I I “HAM-=4 0 3 4 5 Number of Sites Figure 12. Number of sites (and GFA days) by Distance Visitor Segment, USDA FS Region 2 (Figure Truncated at 5 Sites, Exposure Weighted). 1 00% 90% 80%- 70% 60% 50% 40% 30% Percent of Respondents 20% 1 0% 0% —0— Local —I— Mid-Dist. //” - 1- - Long-Dist. i 1L Number of Sites Figure 13. Number of sites (and GFA days) by Distance Visitor Segment, USDA FS Region 9 (Figure Truncated at 5 Sites, Exposure Weighted). 103 Mid-distance activity participation is somewhat bi-modal (Figures 14 and 15). Visitors are most likely to participate either in a single recreation activity or in 3 to 5 activities. More than four recreation activities are completed with decreased propensity— though such a pattern is most common among Long-distance visitors in Region 2 and among Mid-distance visitors in Region 9. Local visitors less frequently participate in multiple activities than visitors in other visitor segments. Local visitors may be characterized as “focused visitors”, coming to the forest and participating only in one or two specific activities, while Mid-distance and Long-distance visitors more frequently participate in multiple recreation activities during a single visit. 25% 20% . i l: 15% Local 3 —-— Mid-Dist. E 10% - *- - Long-Dist. s 5% - 0% Number of Activities Figure 14. Number of Activities by Distance Segmented, USDA FS Region 2 (Figure Truncated at 10 Activities, Exposure Weighted). 104 25% 8 +Loca| _.— Mid-Dist. - -A- - Long-Dist. 10% Percent of Respondents 9 5% 0% Number of Activities Figure 15. Number of Activities by Distance Segmented, USDA FS Region 9 (Figure Truncated at 10 Activities, Exposure Weighted). The patterns of party size are moderately bi-modal (Figures 16 and 17). In both regions, Long-distance and Mid-distance visitors are most likely to recreate in either two or four person parties. Three person parties or parties with more than four persons are less common among these visitor segments. Local visitors are most likely to recreate in two person parties and nearly equally likely to recreate in either three or four person parties. Party sizes of more than four people are less common among this segment. Large parties of recreation users are most common in the Lon g-distance visitor segment. 105 Percent of Respondents —o—Local —-— Mid-Dist. - ‘- - Long-Dist. 3 4 5 People per Vehicle Figure 16. People per Vehicle by Distance Segment, USDA FS Region 2 (Exposure Weighted). 60% Si 5 5 Percent of Respondents 5 —L 23 0% —o—Local —|— Mid-Dist. - *- - Long-Dist. People per Vehicle Figure 17 . People per Vehicle by Distance Segment, USDA FS Region 2 (Exposure Weighted). 106 Across all distance segments, the greatest numbers of recreation visitors first arrive at the national forest on Friday, Saturday, or Sunday (Table 18). Arrivals on Monday, Tuesday, Wednesday, or Thursday are more common by Long-distance visitors than other distance segments. This pattern is consistent with the assumption that Long- distance recreation visitors are either on vacation or in the area for business, both likely to include weekday travel. The arrival patterns of Local and Mid-distance visitors are comparable—though Local visitors are slightly more likely to arrive at the national forest during the workweek. This similarity is contrary to the postulated characteristics identified in Chapter 3. Table 18. Day of National Forest Arrival by Distance Segment and USDA FS Region.“ Study Distance P- lggion Segment Sun. Mon. Tues. Wed. Thurs. Fri. Sat. Total value Region 2 Local 25% 11% 7% 8% 9% 15% 25% 100% 0.000 Mid-dist. 21% 7% 7% 6% 10% 23% 26% 100% Long- dist. 15% 12% 16% 14% 11% 15% 17% 100% Region 9 Local 27% 8% 5% 7% 8% 15% 31% 100% 0.000 Mid-dist. 19% 6% 5% 6% 9% 24% 30% 100% Long- dist. 15% 10% 11% 11% 11% 20% 23% 100% “ Statistical analysis completed using contingency table analysis and Exth’s. Comparisons between regions using exposure weights assumes there is no bias in site stratification or the propensity for visitor sampling to occur on a given day(s) between regions. In Region 2, greater percentages of Local visitors engage in hunting, fishing, and hiking than visitors within other Region 2 distance segments (Table 19). More of these visitors recreate on the national forest with no particular primary activity. Local visitors 107 in Region 9 are more commonly hunting, fishing, or “driving for pleasure” than other distance segments in the region. Mid-distance visitors in both regions are more likely to be camping, operating OHV’s, and downhill skiing than others. In both regions, Long- distance visitors are more commonly completing “nature-related” activities. In Region 2, these visitors more frequently are “driving for pleasure” than other segments while in Region 9 they are more commonly hiking, snowmobiling, or using resorts than other segments. Table 19. Primary Activity by Distance Segment and USDA FS Region.“ Region 2 Region 9 Mid- Long- Mid- Long- Local dist. dist. Local dist. dist. Developed Camping 2% 3% 2% 2% 5% 3% Primitive Camping & Backpacking 2% 1% 0% 1% 4% 3% Resort 0% 0% 0% 1% 1% 4% Picnic 2% 2% 1% 3% 0% 1% Nature-related 8% 5% 10% 7% 7% 13% General/Relaxing 5% 4% 4% 5% 6% 7% Fishing 7% 3% 3% 13% 9% 7% Hunting 5% 6% 3% 13% 9% 5% OHV use 3% 4% 1% 3% 5% 2% Driving 3% 2% 4% 6% 2% 1% Snowmobile 4% 2% 1% 3% 3% 5% Boating 2% 1% 1% 1% 1% 1% Hiking 19% 7% 11% 16% 13% 25% Biking 7% 6% 6% 2% 1% 0% Downhill skiing 14% 45% 45% 7% 13% 4% Cross-country skiing 5% 1% 0% 7% 8% 3% Other non-motorized 4% 2% 1% 3% 2% 2% Other 3% 1% 1% 4% 1% 4% No primary activity 4% 2% 2% 1% 3% 3% Multiple primary activities 2% 3% 2% 2% 7% 5% MI 100% 100% 100% 100% 100% 100% floured from full NVUM sample, weighted by NVEXPAND. 108 The most marked differences between distance segments occur in trip-type frequency (Table 20). Local visitors generally recreate on day trips and less commonly on overnight trips. “Not primary” trips are rare for this visitor segment. Mid-distance visitors most commonly complete overnight trips, recreating less commonly on day trips and infrequently on “Not primary” trips. Very few Long-distance visitors complete primary purpose day trips, instead taking a large number of “Not primary”. These differences are marked and consistent with the expectations identified in Chapter 3. Table 20. Trip-type Segment Shares by Distance Segment and USDA FS Study _Rr_:_glon.d Day OVN/OVNNF“ Not Primary“ Total Region2 Local“ 72% 24% 3% 100% Mid-dist.“ 24% 66% 11% 100% Long-dist.c 8% 59% 33% 100% Region9 Local“ 72% 25% 4% 100% Mid-dist.“ 17% 70% 14% 100% Long-dist.“ 6% 62% 33% 22% “‘ b' “ Statistically different subgroups within regions. “Estimated based upon the NVUM economic survey subsample. Exth’s applied. Statistical comparisons between regions within distance segments were completed using contingency table analysis. “ Recreation trips involving an overnight stay in the local forest area or on the national forest. Trips not made primarily to recreate on the national forest. Within trip-types, the characteristics of visitors in distance segments are less frequently statistically different (Table 21). Statistical differences between distance segments remain for visit duration and the number of people per vehicle when accounting for trip-type. For the most part, patterns of site use and the number of activities are statistically similar across segments within trip-types. 109 Table 21. Mean Values for Visitor Characteristics of Interest by Distance Segment, Trip-type, and USDA FS Region. Visit Number Distance duration Number of People! Trip Type Reg‘Lon Segment (hours)“ of Sites Activities Vehicle Day Region 2 Local 30‘ 1.1‘ 3.2‘ 21‘ Mid-dist. 4.32 1.1‘ 29‘ 2.42 Long-dist. 20° 1.3‘ 3.4‘ 24"2 Region 9 Local 2.6‘ 1.1‘ 3.1‘ 2.3‘ Mid-dist. 4.22 1.2‘ 3.1‘ 22' Long-dist. 1.9“l 1.3al 3.8a 1.8“ OVN/OVNNF Region 2 Local 24.0‘ 1.3‘ 4.6‘ 2.5‘ Mid-dist. 22.2‘ 1.3‘ 4.22 2.5‘ Long-dist. 5.52 1.5‘ 3.92 2.82 Region 9 Local 40.2‘ 1.3‘ 4.6‘ 2.3‘ Mid-dist. 28.02 1.4‘ 4.4‘ 2.52 Long-dist. 8.03 1.6‘ 4.2‘ 2.5‘2 Not primary Region 2 Local 26° 13‘ 4.2‘ 2.2‘ Mid-dist. 3.0‘ 1.3‘ 4.2‘ 26‘ Long-dist. 3.0‘ 1.5‘ 4.8‘ 25‘ Region 9 Local 3.2“ 1.1“ 3.4“ 2.2“ Mid-dist. 3.6‘ 1.3‘ 3.6‘ 28‘ Long-dist. 2.22 1.4‘ 3.6‘ 2.4‘ Note: Statistical tests completed using Mann-Whitney U incorporating the exposure weights. “ Median values “ Not computed within trip-types since variable corresponds to previous trips of unknown type. VisExpwt used in statistical analyses, :“ 3 Statistically different subgroups within regions. In both study regions, Local visitors on day trips spend a shorter period of time on-forest than Mid-distance visitors on day trips. This pattern is consistent with the postulated characteristics identified in Chapter 3. Considering overnight visits, the on- forest durations of Local, Mid-distance, and Long-distance visitors follow a continuum from longest forest duration to shortest forest duration, respectively. In Region 2, only the visit duration of Long-distance OVN/OVNNF visitors is statistically different from the 110 other segments while all OVN/OVNNF segments are statistically different from one another in Region 9. Average party sizes range from 2.1 to 2.8 individuals for all segment trip-type combinations (excluding those with less than 30 cases). The party sizes of Local day trip groups in Region 2 are statistically different from that of parties in the Mid-distance segment in that Region. Similarly, Local visitor groups on overnight trips are smaller than Long-distance segment visitor parties. In Region 9, the size of overnight Local visitor parties is statistically different from that of Mid-distance visitor parties on ovemight trips. Summgy of Distance Segment Characteristics The characteristics of visitors within the three distance visitor segments are largely consistent with those postulated (Chapter 3, Table 1). Local visitors complete multiple trips annually, primarily on short-duration day trips. Local visitors typically visit a small number of sites and participate in few activities on any single visit. When on overnight trips, Local visitors spend a substantial time on-forest and participate in a larger number of activities. Contrary to expectations outlined in Chapter 3, Local visitors primarily first arrive at the national forest during the weekend rather than the weekday— though Monday arrivals are relatively common among Local visitors. Mid-distance visitors recreate on a given national forest less frequently than Local visitors (though the difference is not statistically significant) and more frequently than 111 Long-distance visitors. When completing a national forest recreation visit, Mid-distance visitors typically spend the night away from home. Among distance segments, these visitors are least likely to spend only a short period of time (less than 5 hours) on-forest during their recreation visit. In aggregate, these visitors recreate on more sites, participate in more activities, and are in larger recreation parties than Local visitors. The characteristics unique to Long-distance visitors are low annual visit frequencies and a propensity for national forest visits of very short duration. More than 50% of Long-distance visitors in both regions complete only one annual national forest visit and more than 95% complete less than four visits annually. Approximately 50% all Long-distance visitors spend less than four hours on the national forest. In Region 9, nearly 20% stay for less than one hour. By far, this distance segment has the greatest percentage of visitors classified as “Not Primary”, more than 20%. Among distance segments, those traveling greater than 200 miles visit the greatest number of sites, participate in the greatest number of activities, and recreate in the largest parties. Compared to Local and Mid-distance visitors, Long-distance visitors are most likely to arrive at the national forest during the workweek. Recreation Use Models Objective 3 of this study is to develop models to predict the annual forest-level recreation use of Local and Mid-distance segment visitors. Recreation use estimates for national forests sampled under NVUM in FY2001, FY2002, and FY2003 were used to 112 construct recreation use models and to verify mode] operation. Recreation use estimates from forests sampled in calendar year 2000 are used to validate model performance (ascertain the model’s out-of—sample predictive ability). In this section, the Local recreation use models are presented first followed by models constructed to predict the recreation use of Mid-distance visitors. Local Permanent Resident Model Construction of the models predicting the recreation use associated with permanent Local area residents required estimation of regional-level estimates of the percent of the population participating in national forest recreation, identification of forest-level Local populations, and estimation of annual visitation rates of those participating in national forest recreation, all by distance band. Evaluation of the Local model predictions required development of the NVUM estimates of forest-level Local segment recreation USC. Regional-level participation rates of Local populations living in distance bands surrounding national forests were estimated using the formula developed in Chapter 3. This formula incorporated band-level populations, NVUM use estimates, and annual visitation rates, aggregated within region (Table 22). For the most part, the empirically estimated regional participation rates decrease with increasing distance, as expected. One notable exception to this pattern is a substantial increase in participation in the 25-30 mile 113 distance band in Region 2. This increase in percent participating is a manifestation of the difference in spatial units of analysis between the NVUM data and the Census data.'0 Table 22. Regional Participation Rates of Populations Residing Within 30 Miles of National Forests in the USDA FS Study Regions. Distance Band 0 0-10 10-15 15-20 20-25 25-30 Region NVUM Use 2 Estimate 1,723,5 16 3,476,959 585,723 647,619 181,577 206,428 Population 134,035 1,076,205 467,167 680,461 5 1 1,090 140,893 Annual Visits fi._8 M 4._8_ 4; 3+6 3g Percent Participating 87% 44% 26% 21% 10% 41% Region NVUM Use 9 Estimate 1,166,158 2,259,467 270,956 90,953 15 1,668 164,1 13 Population 385,018 2,235,003 1,282,808 1,280,169 1,197,016 1,587,431 Annual Visits L8 7_.0_ $9 2L6 43 3,4 Percent g Partflating 39% 15% 5% 2% 3% 3% Across all bands, the participation rates of those living in Region 2 are greater than those of the Region 9 population. Within bands, the participation rates of those living within national forest boundaries (0 Band) are substantially greater than pOpulations living outside national forest boundaries. The occurrence of very high —~ 10 Aiarge number of ZIP code origins in the 25 - 30 mile distance band (also the 20 - 25 mile band) in Region 2 are located in the Denver metropolitan area. However, the census block groups in the 25-30 mile band are located, largely, on the outskirts of the Denver metropolitan area (near the Denver airport) in an area With a small population. In the future, it may be advantageous to aggregate the 20-25 mile distance band and the 25-30 mile distance band to smooth this estimate. 114 participation rates among those living within national forest boundaries is logical as the national forest is “at the doorstep” of this population. Within regions, Local populations are highly variable (Table 23). In Region 2, only 30,000 people live within 30 miles of the Routt NF while 2.1 million people live near the Pike San Isabel NF. The Local population of the Hiawatha NF is only 93,000 residents while the Local population of the Mark Twain NF is 1.8 Million people. Local populations around Region 9 national forests are typically greater than those of Region 2. Table 23. Forest-level Distance Band Populations for NVUM forests sampled in FY2001, FY2002, FY2003.“ Distance Band 0 0-10 10-15 15-20 20-25 25-30 _R_ggion 2 Mom NF 1,322 11,684 23,162 4,963 4,301 2,490 Black Hills NF 24,371 114,182 8,413 5,013 1,987 0 Grand Mesa, U.,G. NF’s 17,937 56,119 81,250 38,399 5,994 625 Medicine Bow NF 0 34,325 15,070 60,605 41,851 39,216 Nebraska NF 4,507 34,338 13,114 10,936 13,147 10,845 Pike San Isabel NF 33,071 755,161 322,666 531,226 408,905 55,697 Routt NF 4,621 16,262 3,370 6,140 0 0 Shoshone NF 3,011 11,345 17,917 5,476 8,398 23,219 White River NF 46,517 55,709 1,833 5,110 1,150 625 Region 9 Allegheny NF 20,332 113,825 120,692 108,484 74,660 79,758 Chequamegon-Nicolet NF 24,468 94,671 64,2 85 96,731 72,5 86 65,320 Chippewa NF 14,866 50,161 30,452 15,733 17,668 11,460 Hoosier NF 26,027 264,409 163,230 122,646 141,626 269,580 Huron-Manistee NF 88,017 380,337 105,201 164,850 180,481 279,015 Mark Twain NF 86,030 559,874 269,382 382,272 290,601 244,049 Monongahela NF 31,414 104,822 65,629 67,451 96,558 248,230 Ottawa NF 10,431 37,426 6,552 18,601 13,230 6,734 Shawnee NF 26,819 249,541 11 1,037 96,633 67,285 84,795 Wayne NF 56.614 379,938 349,274 208,574 238,802 297,367 “ Estimated using Barrier-adjusted Band distance. Census block groups applied only to the nearest forest. 115 The mean numbers of annual visits completed by visitors living within Local distance bands (Table 24) were estimated for the national forest aggregates. The number of annual visits declines with increasing distance from forest up to 15 miles from the forest boundary. From 15 to 30 miles the annual visit frequencies are relatively constant. The annual visit estimates of Region 2 Local visitors are, for the most part, greater than those reported by Region 9 Local visitors. Table 24. Distance Band Annual Visit Frequencies of Local Visitors by Region and National Forest Aggflrgation Group.“ Distance Band Aggregation Group“ 0 0-10 10-15 15-20 20-25 25-30 Region 2 1 14.1 10.7 4.8“ 6.3 3.6 “ 3.5 “ 2 10.7 6.7 6.3 4.5 “ 3.6 “ 3.5 “ 3 11.0 5.7 3.3 2.7 3.0 3.3 4 16.7 11.7 6.4 5.9 3.6 “ 3.6 Regional Avg. 14.8 7.3 4.8 4.5 3.6 3.5 Region 9 l 6.8 6.6 3.1 3.5 4.7 4.2 2 9.3 7.1 4.9 3.6 c 4.8 3.2 3 6.7 6.4 4.3 3.6 “ 2.9 3.4 “ 4 5.6 4.4 4.0 “ 3.6 “ 4.9 “ 2.6 5 12.3 9.2 4.2 3.6 4.9 3.4 Regional Avg. 7.8 7.0 4.0 3.6 4.9 3.4 “ Estimated from the NVUM survey and weighted by VisExpwt. Standard errors available in Table 01 “ Aggregation Groups defined below. Region 2: l Bighorn NF, Shoshone NF; 2 Black Hills NF, Nebraska NF; 3 Arapaho and Roosevelt National Forests, Medicine Bow NF, Pike and San Isabel National Forests; 4 Grand Mesa, Uncompahgre, and Gunnison National Forests, Rio Grande NF, Routt NF, San Juan NF, White River NF. Region 9: l Hoosier NF, Mark Twain NF, Shawnee NF; 2 Alleghany NF, Monongahela NF, Wayne NF; 3 Green Mountain and Finger Lakes National Forests, White Mountain NF; 4 Hiawatha NF, Huron-Manistee NF, Ottawa NF; 5 Chippewa NF, Chequamegon/Nicolet NF, Superior NF. “ Number of cases is less than 30. Region-level average substituted. 116 Population counts of those living in distance bands around national forests were combined with regional-level participation rates and estimates of annual visit frequencies to predict forest-level recreation use by those living within 30 miles of the national forest boundary (I able 25). Model predictions of forest-level Local visitor recreation use range from 112,000 visits to 3 million visits for national forests in Region 2 and from 51,500 visits to 903,000 visits for national forests in Region 9. In Region 2, only one forest prediction of Local visitor use (Black Hills NF) is consistent with the NVUM Local visit estimate. The regional summation of forest-level predictions in Region 2 falls below the Region 2 NVUM Local visitor estimate. Of the eight Region 2 national forests where model predictions are outside the NVUM Use estimates, the model underestimates Local use for six national forests and over-estimates Local use for two national forests. Model predictions for three of the six under-predicted national forests were within 90,000 visits of lower NVUM Local visit estimate. The two forests over-predictions were well above the NVUM upper estimates of Local segment visits. The Local visitor model performs slightly better in Region 9 than in Region 2. In three cases, the predicted Local visitor use falls within the confidence intervals of the NVUM Local visit estimates. Although not within the confidence intervals, the predicted Local visitor use for Ottawa NF is very close to lower NVUM estimate. The regional summation of predicted forest-level Local visitor use falls within the confidence interval . of the NVUM Region 9 Local use estimate. Of the seven Region 9 national forests where 117 model predictions are outside the 80% NV UM confidence interval, the recreation use of Local visitors is under-predicted on four national forests and over- predicted on three national forests. Table 25. Model and NVUM Estimates of Permanent Resident Local Visitor Recreation Use. Local Visitor Lower Upper Use NVUM Use NVUM NVUM Prediction Estimate Estimate Estimate Bighorn NF 112,122 273,995 198.388 355.876 Black Hills NF 583,526“ 506,796 398.248 622.623 Grand Mesa, U.,G. NF 736,311 1,388,635 953,697 1,848,310 Medicine Bow NF 200,592 288,978 236,296 345,489 Nebraska NF 195,758 56,678 40,789 74,836 Pike San Isabel NF 3,004,975 2,045,153 1,714,877 2,389,610 Routt NF 164,189 587,221 501,804 679,250 Shoshone NF 157,285 258,705 214,246 306,760 White River NF 973,054 1,279,373 1,171,184 1,392,239 Total 6,127,811 6,821,822 6,333,542 7,321,118 Alfgheny NF 246,456 555,824 381,840 744,479 Chequamegon-Nicolet NF 280,043 595,152 389,868 822,667 Chippewa NF 149,101 726,966 592,578 872,045 Hoosier NF 410,796“ 365,668 295,388 440,773 Huron-Manistee NF 512,529 212,354 129,357 307,506 Mark Twain NF 903,695 412,738 350,419 477.656 Monogahela NF 279,677“ 262,522 214,339 314,594 Ottawa NF 51,438 78,848 61,991 97.583 Shawnee NF 354,919‘ 316,315 270.318 365.126 Wame NF 759,006 248,914 200,208 301,341 Total 3,947,661‘ 4,102,540 3i34,356 4,480,997 * Model prediction falls within the 80% confidence interval of the NVUM Local visitor use estimate. 118 The absolute values of raw errors in model predictions range from 76,000 visits to 960,000 visits in Region 2 and from 17,000 visits to 510,000 visits in Region 9 (Table 26). Six of the nine national forests in Region 2 have percentage errors (percentage error = absolute raw error/NV UM Local visit estimate) of less than 50%. Two forest-level percentage errors are below 25%. The absolute value of the raw error between the summation of the Region 2 Local model predictions and the NVUM regional estimate was approximately 700,000 visits, corresponding to a percent error of 10%. Table 26. Local Visitor Model Errors. Absolute Errora Percentage Error” Bighorn NF 161,873 59% Black Hills NF 76,731 15% Grand Mesa, U.,G. NF’s 652,324 47% Medicine Bow NF 88,386 31% Nebraska NF 139,080 245% Pike San Isabel NF 959,821 47% Routt NF 423,032 72% Shoshone NF 101,420 39% White River NF 306,319 24% Region 2 694,011 10% Allegheny NF 309,368 56% Chequamegon-Nicolet NF 315,109 53% Chippewa NF 577,864 79% Hoosier NF 45,128 12% Huron-Manistee NF 300,175 141% Mark Twain NF 490,958 119% Monongahela NF 17,156 7% Ottawa NF 27,410 35% Shawnee NF 38,604 12% Wayne NF 510,092 205% Region 9 154,879 4% a abs(Model prediction — NVUM Local visit estimate) b (absolute error/NVUM Local visit estimate)*100 119 In Region 2, the predicted Local visitor use for the Nebraska NF (and administered grasslands) was vastly different from the NVUM estimate of Local use. In fact, the predicted use of Local visitors was greater than the NVUM estimate of all recreation use on the Nebraska NF (127,000 total visits). Such a large overestimation is indicative that either the regional participation rates or aggregate annual visit frequencies are inappropriate for the Nebraska NF. The Region 2 participation rates are largely influenced by national forests located in the Rocky Mountain region of Colorado and Wyoming. The outdoor recreation resources and services available on these mountain national forests likely differ from those of the Nebraska NF. In addition, the recreation behavior and participation patterns of those living around the Nebraska NF may be more similar to populations proximate to Region 9 national forests given the Nebraska NF’s distance from the mountain national forests. Substituting the Region 9 participation rates for Region 2 participation rates in the Nebraska NF Local visitor model yields a prediction of 59,881—nearly identical and within the confidence intervals to the NVUM estimate. Given this result, Region 9 participation rates appear more appropriate for the Nebraska NF than Region 2 participation rates. Errors are generally greater in Region 9 than in Region 2. Only four of ten national forests in Region 9 have percent errors of less than 50%. Percent errors for the Huron-Manistee NF, the Mark Twain NF, and the Wayne NF are all above 100%. On those Region 9 national forests where the model predictions are within the NVUM Local 120 visitor use estimates (Shawnee NF, Hoosier NF, Monongahela NF), both absolute and percent errors are quite small. Summing all Region 9 national forest model predictions, the regional prediction of Local use has an absolute raw error of 155,000 visits, a percent error of only 4%. The national forests in the study area with the greatest population sizes are the Pike San Isabel, the Mark Twain NF, and the Wayne NF, respectively. The predicted Local use levels for each of these forests are significantly greater than the NVUM estimate. The percent errors for the Wayne NF and Mark Twain NF are among the highest in Region 9. In all three cases, the prediction from the Local visitor model is greater than the NVUM estimate of total recreation use for the national forest. Commonly, the participation rates of large population areas are less than those where population numbers are small. Overestimation for these national forests is indicative of the need to develop adjusted participation rates applicable to forests with large population areas. Such participation rates could be developed by pooling visit estimates among forests with large populations and re-estimating participation rates. When there are statistical differences between the two, Local model predictions tend to be lower than the NVUM Local visit estimates. Of the 15 cases where statistical differences exist, the model predictions are lower than the NVUM estimates in 10 cases. Non-random error in specifying the model inputs or over-estimation of Local visitor use from NVUM may cause the consistent underestimation of Local segment visits. Concerning the former, the model assumes that all residents within a given distance band 121 participate with the same proclivity. If a disproportionate share of the population within the distance band is located closer to the forest than the Mid-distance, visitation may be underestimated. This may be particularly applicable to the zero — 10-mile band where a large percentage of residents may reside directly adjacent to the forest boundary. Based upon their proximity, these residents may have much higher participation rates than other individuals in the distance band living farther away from the national forest. Consistent underestimation may also be indicative of errors in the NV UM Local use estimate. In this study, the NVUM estimate of Local use is developed from the visitor survey data and the NVEXPAND weights. Stynes and others (2003) have questioned whether the N VEXPAND weights are appropriate for estimating visitor characteristics and trip types from the NVUM survey data. Use of the NVEXPAND weights assumes that the NVUM approach to stratifying recreation sites captures variability in visitor origins and/or visitor characteristics. The exposure weights, which assume the NVUM stratification does not explain variability in visitor characteristics, have been adopted by Stynes and others (2003) when estimating visitor characteristics and, in some instances, when estimating visitor shares. For national forests in the study area, use of exposure weights yield lower Local visitor shares in many cases (Table 27). In several cases the differences between forest-level estimates of Local segment shares under the alternate weighting schemes are greater than 10 percentage points—particularly in Region 2. Consistently high NVUM estimates of Local use could lead to the pattern of under- estimation found in the Local visitor model. 122 Table 27. Local visitor Segment Shares Estimated Under Two Alternative WeigflLSchemes. NVEXPAND Exth Bighorn NF 52% 31% Black Hills NF 69% 59% Grand Mesa, U.,G. NPS 67% 55% Medicine Bow NF 51% 51% Nebraska NF 45% 30% Pike San Isabel NF 73% 75% Routt NF 39% 38% Shoshone NF ' 53% 49% White River NF 44% 44% Region 2 58% 51 % Allegheny NF 57% 54% Chequamegon-Nicolet NF 38% 38% Chippewa NF 55% 46% Hoosier NF 71% 67% Huron-Manistee NF 30% 38% Mark Twain NF 84% 70% Monongahela NF 38% 35% Ottawa NF 34% 26% Shawnee NF 62% 61% Wayne NF 64% 69% Region 9 51 % 49% Local Seasonal Resident Model The number of seasonal homes located within the Local areas of national forests in the study was identified from US. Census data (Table 28). In Region 2, Pike San Isabel NF followed by the White River NF and Grand Mesa, Uncompahgre, and Gunnison NF have the greatest number of seasonal homes close to the national forest. The Nebraska NF has the lowest number of seasonal homes in the local area of the National Forest. The Huron-Manistee NF, in Region 9, has more than 100,000 seasonal homes located in its local area—by far the largest number of seasonal homes among 123 national forests in this study. Elsewhere in Region 9, the Chequamegon—Nicolet NF, the Alleghany NF, and the Mark Twain NF all have 30,000 or more seasonal homes located within 30 miles of the forest. The Shawnee NF has the fewest number of seasonal homes proximate to the national forest. Table 28. Seasonal Homes Proximate to NVUM forests sampled in FY2001, FY 2002, and FY 2003.a Distance Band 0 0-10 10-15 15-20 20-25 25-30 _R__e_gion 2 Bighorn NF 313 598 442 24 271 57 Black Hills NF 1,523 838 33 112 30 0 Grand Mesa, U.,G. NF’s 4,014 2,004 145 66 17 2 Medicine Bow NF 0 1,450 327 636 121 535 Nebraska NF 99 586 173 103 210 115 Pike San Isabel NF 5,429 7,964 920 1,690 259 138 Routt NF 1,174 976 391 33 0 0 Shoshone NF 506 513 868 973 785 564 White River NF 7,632 1,847 36 60 25 2 Region 9 Allegheny NF 9,982 8,212 2,099 4,472 4,976 6,192 Chequamegon-Nicolet NF 12,674 22,831 1 1,998 10,662 5,170 5,697 Chippewa NF 4,948 9,541 3,250 3,467 3,635 2,597 Hoosier NF 804 2,366 1,431 1,805 2,342 1,022 Huron-Manistee NF 25,106 34,947 12,328 14,021 10,335 11,473 Mark Twain NF 5,256 9,585 2,893 5,355 2,873 8,256 Monongahela NF 5,624 4,004 2,179 5,062 4,448 3,049 Ottawa NF 3,688 8,654 2,241 1,089 1,632 183 Shawnee NF 649 1,375 432 853 485 772 Wayne NF 746 3,853 2,995 2,931 2,338 1,964 ‘ Estimated using Barrier-adjusted Band distance. The number of seasonal households within census block groups counted only for the nearest forest. Due to data limitations, it was assumed that seasonal homeowners in the Local area visit with the same propensity and with the same annual frequency as permanent local residents. An estimate of the average party size of seasonal homeowner parties originating from seasonal homes was identified using the NVUM FY 2003 national-level 124 survey data. Nationally, seasonal homeowner parties average 2.6 individuals per party. For the Local visitor seasonal model it is assumed that recreation parties associated with seasonal households in all Local distance bands are comprised of 2.6 individuals. The numbers of visits associated with seasonal homes in the local areas of national forests were estimated from the counts of seasonal homes, participation rates, annual visit frequencies, and party sizes (Table 29). The predicted number of national forest visits associated with seasonal homes is highly variable. In Region 2, only 8,500 visits to the Nebraska NF are associated with seasonal homes while more than 310,000 visits are predicted for the White River NF. The model predicts only 8,600 seasonal home visits to the Shawnee NF and nearly 250,000 seasonal home-related visits to the Chequamegon-Nicolet NF. Although there are many more seasonal homes in Region 9, visitation associated with seasonal homes is nearly equal between the two regions—due to the greater assumed participation rates and visit frequencies of Region 2 residents. It is difficult to verify the performance of the seasonal homeowner Local visitor model. The NVUM visitor surveys used in the first three years of NVUM did not clearly identify seasonal homeowners, precluding direct estimation of seasonal homeowner recreation use for forests sampled during these years. In the Year 4 survey, respondents were asked to report the type of local lodging used, including owned-homes, if they were “staying away from home on this trip”. The percent of survey respondents selecting “owned home” can be used to develop the NVUM estimate of seasonal homeowner use. However, this is likely not a complete enumeration of seasonal homeowner use as it is 125 reasonable to expect that many seasonal homeowners may not have answered the question since they were indeed staying at a home they owned (i.e. “not staying away from home”). Table 29. Model Estimates of Seasonal Local Resident Recreation Use. Forest Model Estimate Bighorn NF 19,347 Black Hills NF 43,890 Grand Mesa, U.,G. NF’s 179,194 Medicine Bow NF 13,190 Nebraska NF 8,517 Pike San Isabel NF 192,660 Routt NF 59,184 Shoshone NF 31,566 White River NF 313,134 Total 860,682 Allegheny NF 121,326 Chequamegon-Nicolet NF 248,499 Chippewa NF 98,865 Hoosier NF 13,466 Huron-Manistee NF 214,88 1 Mark Twain NF 65,970 Monongahela NF 68,254 Ottawa NF 37,187 Shawnee NF 8,648 Wayge NF 21,1 18 Total 829,960 Three of the seven forest-level model predictions (all in Region 9) for Year 4 national forests fall within the confidence intervals of NVUM seasonal homeowner recreation use estimates (Table 30). Seasonal resident model predictions for Region 2 were all greater than the NVUM estimates. However, it is unlikely that the NVUM estimate of no seasonal homeowner visits to the Shoshone NF is correct given the sheer 126 number of seasonal homes proximate to the NF. For this NF (at least), the NVUM estimate likely underestimates true seasonal homeowner recreation use. Table 30. Model and NVUM Estimates of Seasonal Resident Local Visitor Recreation Use for FY 2003 NVUM National Forests. Count of Local NVUM Lower Upper Seasonal Visitor Use Use NVUM NVUM Homes Prediction Estimate Estimate Estimate Black Hills NF 2,536 43,890 3,930 0 13,175 Grand Mesa, U.,G. NF’s 6,248 179,194 17,474 5,227 54,177 Shoshone NF 16,399 31,566 0 0 0 Hoosier NF 9,770 13,466‘ 6,742 217 15,299 Monongahela NF 24,366 68,254‘ 49,532 31,063 71,708 Ottawa NF 17,487 37,137" 47,669 32,694 65,358 Wayne NF 14,827 21,118 3,394 0 8,642 1' Model prediction falls within the 80% confidence interval of the NVUM Local visitor use estimate. Mid-distance Recreation Use Models The recreation use of day trip and overnight trip Mid-distance visitors was modeled separately. Day trips are those where the recreation visitor does not spend a night away from their home. Overnight trips include those visits involving an overnight stay on the NF as well as those including a stay overnight off-forest.ll The models developed for Mid-distance visitation predict only the recreation use where the primary trip purpose is recreation at the NF. “Non-primary” trips are excluded, as these trips are likely a function of factors other than those typically included in travel cost models. n In the first three years of NVUM survey, respondents were asked to report the length of their trip away from home rather than the number of nights in the local forest area. Given this, some modeled overnight trips will include nights spent away from home but not in the local forest area. 127 The multi-site zonal travel cost models of Mid-distance segment recreation use were estimated using both single-log models estimated via ordinary least squares (OLS) and Tobit models estimated via maximum likelihood estimation (MLE) for each study region. The single-log OLS models were estimated with the dependent variables day and overnight visits and day and overnight visits per thousand population. Tobit models were estimated with the dependent variables day and overnight visits and day and overnight visits per capita. Eviews 4.1 (Quantitative Micro Software LLC, Irvine, CA) was used to estimate all models as well as to forecast forest-level Mid-distance recreation use based upon the fitted models. When forecasting, Eviews 4.1 software corrects for the bias resulting from the retransforrnation of log estimates. Stynes and others (1986) identified bias in coefficients and use estimates when using log formulations of travel cost models. White’s heteroskedasity test was completed on all OLS models. In all cases, the null hypothesis of homogeneity could not be rejected. The “best performing” single-log OLS and Tobit models in each region were selected to predict forest-level day and overnight Mid—distance segment recreation use. Models were selected based upon 1) consistency of the coefficient estimates with theory and 2) the smallest mean absolute percentage errors between model prediction and observed values of band-level recreation use. The number of observations used for estimating the day and overnight single-log OLS differ from the number of observations used to estimate the Tobit models since zero-visit observations are excluded from the single-log formulations. The greatest number of zero-visit observations occurs for day 128 trip recreation use and, correspondingly, these models have the smallest number of observations. Preliminary models constructed using alternate dependent variable formulations were compared to one another, in part, based upon their mean absolute percent errors (MAPE). Despite the selection of final models that had the smallest MAPE values (in addition to coefficient estimates consistent with theory) the final model MAPE values are still moderately large for Region 2 models and very large for Region 9 models. When interpreting these MAPE values it is important to recognize that the model MAPE’s depict the differences between observed dependent variable (e. g. day trips/capita) and the predicted dependent variable—rather than observed total recreation use versus predicted total recreation use. Given that the dependent variables used in estimating the models are generally small values (e. g. visits per capita), moderate divergences from these values can result in large MAPE values. In Region 2, day trips are largely dependent upon travel distance. In both model formulations, day trips decrease with increasing distance, all else being equal, as expected (Table 31). In the OLS formulation, the coefficient on RUC is also statistically significant. Increases in RUC code (increasingly rural areas) leads to fewer day trip visits originating from that distance band, all else remaining the same. The coefficients on median income, distance to substitutes, and recreation site acreage and number of units were not statistically different from zero. 129 Table 31. Day and Overnight Mid-distance Segment Recreation Use Models, USDA FS Region 2.a OLS Model Dep. Variable: ln(Distance Band Visits/ Population(l,000s)), Variable Day T-Value OVN T-Value Constant 12.546 353"" 6.464 3.72‘" Population (1,000’s) 0.000 -0.40 0.000 -1.12 Distance -0.054 4.19“" -0.015 4.95"" Med. Income (1,000’s) -0.046 -1.09 0.009 0.39 Nearest Other NF Land 0025 -0.92 -0.010 -0.98 Nearest BLM Land 0.011 0.36 0.012 1.29 Nearest NPS Land 0.007 0.44 ~0.001 -0.14 RUC Code -0.514 -1.80* -0.122 086 Forest Acreage (1,000’s) 0.000 -0.68 0.000 -0.95 ‘ Number of Separate Forest Units -0510 -129 -0312 -203” N 33 95 Adjusted R-squared 0.35 0.31 F-Value 2.91" 562“" Mean Absolute Percentage Error 38% 121% Tobit Model Dep. Variable: Distance Band Visits/Capita Variable Day Z-Value OVN Z-Value Constant 0.704 1.19 0. 842 2.19" Population (1,000’s) 0.000 0.72 0. 000 -0.71 Distance -0005 -205" -0. 003 395“" Med. Income (1,000’s) -0.007 -1.01 -0.004 -0.80 Nearest Other NF Land -0004 -073 -0.005 -186" Nearest BLM Land -0.002 -0.46 0.003 1.43 Nearest NPS Land 0.001 0.43 0. 000 -0.03 RUC Code 0047 -1.06 -0. 053 -1.74" Forest Acreage (1,000’s) 0.000 0.74 0. 000 -0.39 Number of Separate Forest Units -0.031 -0.50 -0. 055 -1.44 N 124 124 Tobit o 0.17 0.22 Log likelihood —10.92 -26.64 Mean Absolute Percentage Error 7% 17% a Model dependent variables and inputs shown 111 Table C- 2 Lp-value < O. 10 M*p-value < 0. 05 p-value < 0. 01 130 Considering overnight recreation use, coefficients for travel distance and the number of separate forest units were statistically different from zero in the OLS overnight model while coefficients on travel distance, distance to nearest substitute national forest, and RUC code were statistically different from zero in the Tobit formulation. In both models, as expected, increasing distance leads to decreases in overnight recreation use, all else remaining equal. In the OLS formulation, a greater number of separate forest units yields lower overnight recreation use. In the Tobit model, origins more rural in character (increasing RUC code) are associated with lower overnight recreation use. The coefficient on distance to nearest substitute USDA FS land is contrary to expectations. However, given the density of USDA FS land in and around Region 2, a complementary rather than substitute relationship between national forests is conceivable. In Region 9, travel distance is the only statistically significant predictor of day trip recreation use (Table 32). Increasing travel distance leads to significant decreases in day trip recreation use, all else being equal. In the overnight recreation use models, the coefficients on population, travel distance, distance to nearest substitute national forest, and forest acreage and number of separate units are all statistically different from zero in the OLS formulation. The signs of all statistically significant coefficients are consistent with expectations. In the Tobit model, travel distance, RUC code, and forest acreage and number of units are all statistically significant predictors of overnight recreation use. The signs of these statistically significant coefficients are consistent with expectations. 131 Table 32. Day and Overnight Mid-distance Segment Recreation Use Models, USDA FS Region 9.a OLS Model Dep. Variable: ln(Distance Band Visits/Population(l,000’s)) Variable DaL T-Value OVN T-Value Constant 1.174 0.52 3.680 2.34”— Population (1,000’s) 0.000 -1.18 0.000 -170" Distance 0,024 4.77"" -0020 639*" Med. Income (1,000’s) 0.059 1.29 0.004 0.15 Nearest Other NF Land 0.004 0.88 0.006 2.13" Nearest NPS Land -0.009 -0.94 -0.005 -0.85 RUC Code 0.336 1.42 0.058 0.34 Forest Acreage (1,000’s) 0.000 0.03 0.002 3.87"“ Number of separate Forest Units -0.084 -0.74 -0.405 -5.23m N 69 131 Adjusted R-squared 0.45 0.56 F-Value 808"" 21.31"" Mean Absolute Percentage Error 110% 252% Tobit Model Dep. Variable: Distance Band Visits Variable Day Z-Value OVN Z-Value Constant -4,877 .3 -0.24 30,5360 1.31 Population (1,000’s) 0.3 0.15 -2.1 -0.89 Distance -1432 340"” -112.8 -2.51" Med. Income (1,000’8) 349.9 0.91 234.0 0.55 Nearest Other NF Land -7 .8 -0. 19 68.3 1.53 Nearest NPS Land 4.6 0.06 -3.0 -0.04 RUC Code -866.5 -041 -7,620.8 3.18"" Forest Acreage (1,000’s) 5.7 1.08 18.1 3.12"“ Number of separate Forest Units 40.4 0.04 -3,113.1 276'" N 140 140 Tobit a 17,096 20,842 Log likelihood -874.41 -1,506 Mean Absolute Percentage Error 713% 1,182% a Model dependent variables and inputs shown in Table C-3 ‘ p-vaiue < 0.10 " p-value < 0.05 "I p-value < 0.01 132 In Region 2, model predictions from either the OLS or Tobit formulations fall within the NVUM estimates for eight of the nine national forests used in model development (Table 33). In four cases predictions from both formulations fall within the confidence intervals. The summation of predicted Region 2 forest-level Mid-distance recreation use also falls within the NVUM regional Mid-distance segment use estimate. Use predictions are generally consistent between the two model formulations, though large differences do exist for the Nebraska NF and the White River NF. In all cases (except the Rout NF) where model predictions fall outside the NVUM confidence intervals, model predictions fall below the NVUM estimates. Table 33. Model Predictions and NVUM Estimates of Mid-distance Segment Recreation Use, USDA FS Region 2.‘I Lower Upper NVUM NVUM NVUM OLS Models Tobit Models Estimate Estimate Estimate Bighorn NF 43,177 38,775 99,692 68,193 135,990 Black Hills NF 25,121" 19,713" 27,680 15,875 42,481 Grand Mesa, U.,G. NF’S 138,352 290,152' 312,161 199,280 442,999 Medicine Bow NF 167,741" 208,947" 177,220 142,030 215,708 Nebraska NF 39,783" 5,552 32,674 22,516 44,637 Pike San Isabel NF 156,519“ 199,490‘ 179,644 130,947 235,966 Routt NF 419,313‘ 539,219 392,194 328,547 461,412 Shoshone NF 57,379" 64,966‘ 53,634 39,561 69,781 White River NF 370,492 642,862‘ 702,753 631,665 777,533 Rggm 2 1,417,876 2,009,675" 1,969,848 1,790,973 2,156,547 ‘ Combined recreation use of Day and Overnight trips. NVUM estimates exclude recreation use where the NF is not the primary trip purpose. * Model prediction falls within the 80% confidence interval of the NVUM Mid-Dist. visitor use estimate. There was less consistency between model predictions and NVUM estimates of NIid-distance recreation use in Region 9 (Table 34). In only four of ten cases were forest- level predictions from either model formulation within the NVUM confidence intervals. 133 Though not within the confidence interval, the OLS prediction for the Wayne NF was within approximately 500 visits of the lower NVUM estimate. In only one case, the Huron-Manistee NF, did model predictions from both the OLS and Tobit models fall within the NVUM estimate. The regional summation of predicted forest-level Mid- distance recreation was outside the NVUM confidence interval. In Region 9, model predictions under the two formulations were frequently disparate. Predictions under the OLS formulation typically were below the NVUM estimates while Tobit model predictions tended to be greater than the NVUM estimates. In particular, Mid-distance segment use predictions for the Ottawa NF and the Mark Twain NF were drastically greater than the NVUM estimates. Table 34. Model Predictions and NVUM Estimates of Mid-distance Segment Recreation Use, USDA FS Region 9.a Lower Upper Tobit NVUM NVUM NVUM OLS Models Models Estimate Estimate Estimate Allegheny NF 175,380 388,120' 340,138 227,368 466,716 Chequamegon-Nicolet NF 252,227 535,414 833,607 556,978 1,131,889 Chippewa NF 80,425 303,915 442,223 349,658 544,799 Hoosier NF 129,161" 288,652 126,631 94,776 162,927 Huron-Manistee NF 350,460‘ 517,792‘ 456,977 292,176 634,072 Mark Twain NF 50,163~~ 278,779 55,123 41,207 71,043 Monongahela NF 304,444 394,705 246,185 203,838 291,264 Ottawa NF 85,188 303,324 32,836 25,618 40,898 Shawnee NF 73,230 207,301 133,763 109,090 160,889 Wayne NF 100,167 229,221 130,716 100,655 164,355 Region 9 1,600,844 3,447,222 3,000,003 2,720,308 3,288,796 ' Combined recreation use of Day and Overnight trips. NVUM estimates exclude recreation use where the NF is not the primary trip purpose. * Model prediction falls within the 80% confidence interval of the NVUM Local visitor use estimate. Absolute errors between the Mid-distance segment model predictions and NVUM estimates range from 2,500 visits to 580,000 visits under the OLS formulation and from 134 8,000 to 450,000 visits under the Tobit formulation (Table 35). Most percent errors in the Region 2 models are less than 30%—all are less than 100%. Percent errors of Region 9 estimates are generally greater, but most are still less than 50%. The percent errors for the Mark Twain NF and Ottawa NF under the Tobit formulation are substantial. Table 35. Mid-distance Visitor Model Errors. Absolute Absolute Percentage Percentage Error (OLS)' Error (Tobit) ' Error (OLS) b Error (Tobit) b Bighorn NF 56,515 60,917 57% 61% Black Hills NF 2,559 7,967 9% 29% Grand Mesa, U.,G. NF’s 173,809 22,009 56% 7% Medicine Bow NF 9,479 31,727 5% 18% Nebraska NF 7,109 27,122 22% 83% Pike San Isabel NF 23,125 19,846 13% 11% Routt NF 27,1 19 147,025 7% 37% Shoshone NF 3,745 1 1,332 7% 21% White River NF 332,261 59,891 47% 9% Total 551,972 39,827 28% 2% Allegheny NF 164,758 47,982 48% 14% Chequamegon- Nicolet NF 581,380 298,193 70% 36% Chippewa NF 361,798 138,308 82% 31% Hoosier NF 2,530 162,021 2% 128% Huron-Manistee NF 106,517 60,815 23% 13% Mark Twain NF 4,960 223,656 9% 406% Monongahela NF 58,259 148,520 24% 60% Ottawa NF 52,352 270,488 159% 824% Shawnee NF 60,533 73,538 45% 55% Wayne NF 30,549 98,505 23% 75% Total 1,399,159 447,219 47% 15% ‘ abs(Model prediction - NVUM Mid-distance visit estimate) b (absolute error/NVUM Mid-distance visit estimate)*100 135 The statistical relationships and predictive abilities of the recreation use models estimated in this study using OLS and Tobit model formulations are very similar. Lacking a clear difference in the performance of the two models, the single-log OLS model is more attractive given its common usage and the ease of interpreting variable coefficients and model summary statistics. Independent variable coefficients estimated using the Tobit model are more difficult to interpret as they reflect both the likelihood of participation as well as the level of recreation use of those participating. Evaluation of Combined Model Predictions To this point, Local and Mid-distance model predictions have been evaluated individually. The second and third evaluations of model performance are based upon the combined model predictions. The summed model predictions for Local (permanent residents only) and Mid-distance segments were compared with the NVUM estimates of summed Local and Mid-distance recreation use (Table 36). Combined, Local and Mid- distance model predictions for Region 2 generally fall below the NVUM estimates of use. Only the summed model predictions for the Black Hills NF fall within the NVUM estimates. This pattern of underestimation can be traced to the consistent underestimation Of Local recreation use predictions. In Region 9, summed model predictions for the Hoosier NF, Huron-Manistee NF, and Shawnee NF fall within the NVUM confidence intervals. The Local and Mid-distance model predictions (individually) were within the NVUM confidence intervals for the Hoosier NF while only the model predictions for Local and IVIid-distance use (individually) were statistically consistent with NVUM data 136 for the Shawnee and Huron-Manistee NF, respectively. No clear pattern of error for Region 9 combined predictions. exists Table 36. Model Predictions and NVUM estimates of Local and Mid-distance Recreation Use. Local+OLS Local+Tobit NVUM Mid-distance Mid-distance Estimate Lower Upper Bighorn NF 155,299 150,897 373,687 284,152 463,222 Black Hills NF 608,647 ' 603,239,. 534,476 437,362 631,590 Grand Mesa, U.,G. NF ’5 874,663 1,026,463 1,700,796 1,205,014 2,196,578 Medicine Bow NF 368,333 409,539‘ 466,198 399,951 532,445 Nebraska NF 235,541 201,310 89,352 71,491 107,213 Pike San Isabel NF 3,161,494 3,204,465 2,224,797 1,913,103 2,536,491 Routt NF 583,502 703,408 979,415 899,887 1,058,943 Shoshone NF 214,664 222,251 312,339 274,515 350,163 White River NF 1,343,546 1,615,916 1,982,126 1,889,759 2,074,493 Region 2 7,545,687 8,137,486 8,791,670 8,278,236 9,305,104 Allegheny NF 421,836 634,576 895,962 646,078 1,145,846 Chequamegon- N icolet NF 532,270 815,457 1,428,759 997,702 1,859,816 Chippewa NF 229,526 453,016 1,169,189 1,006,087 1,332,291 Hoosier NF 539,957' 699,448 492,299 415,156 569,442 Huron-Manistee NF 862,989. 1,030,321 669,331 445,975 892,687 Mark Twain NF 953,858 1,182,474 467,861 407,086 528,636 Monongahela NF 584,121 674,382 508,707 439,116 578,298 Ottawa NF 136,626 354,762 1 1 1,684 95,870 127,498 Shawnee NF 428,149. 562,220 450,078 400,974 499,182 Wayne NF 859,173 988,227 379,630 320,825 438,435 Region 9 5,548,505 7 ,394,883' 7,102,543 6,576,245 7,628,841 * Model prediction falls within the 80% confidence interval of the NVUM Local visitor use estimate. Thus far, the performance of recreation use models has been evaluated based upon ability to predict visit estimates consistent with the NVUM estimates. An 137 alternative evaluation is the ability of the models to predict the percent of recreation use associated with the Local and Mid—distance visitor segments consistent with the percentage estimates of Local and Mid-distance recreation use based upon NVUM. Given the differences in recreation behavior and participation, identifying the percentage of use by alternative visitor segments can enhance recreation managers’ decision-making—even if the total number of visits is unknown. For most forests, the percentages of predicted Local visitor use (as a function of summed Local permanent resident and Mid-distance use) are very similar to percentages derived from the NVUM estimates (Table 37). Table 37. Local Percentage of Local and Mid-distance Segment Recreation Use based gon Model Prediction and NVUM Estimates. Local Percentage, Local Percentage, Local+OLS Mid- Local+Tobit Mid- distance“ distanceb NVUM Bighorn NF 72% 74% 73% Black Hills NF 96% 97% 95% Grand Mesa, U.,G. NF 84% 72% 82% Medicine Bow NF 54% 49% 62% Nebraska NF 83% 97% 63% Pike San Isabel NF 95% 94% 92% Routt NF 28% 23% 60% Shoshone NF 73% 71% 83% White River NF 72% 60% 65% Region 2 81 % 75 % 78 % Allegheny NF 58% 39% 62% Chequamegon-Nicolet NF 53% 34% 42% Chippewa NF 65% 33% 62% Hoosier NF 76% 59% 74% Huron-Manistee NF 59% 50% 32% Mark Twain NF 95% 76% 88% Monongahela NF 48% 41 % 52% Ottawa NF 38% 14% 71% Shawnee NF 83% 63% 70% Wayne NF 88% 77% 66% Region 9 71 % 53 % 58% ' (Local model prediction/ (Local prediction + OLS Mid-distance prediction))* 100 b (Local model prediction/ (Local prediction + Tobit Mid-distance prediction))*100 Note: alternatively, the percent of primary purpose Mid-distance visitor use is (1- Local percent) 138 Model Application to Out-of-Sample National Forests The constructed Local and Mid-distance segment models presented in the previous sections were applied to Year 1 NVUM forests to assess their out-of—sample predictive ability. Predictions from the Local permanent resident model fall below the NVUM estimates for all Year 1 forests (Table 38). This pattern of disparity is consistent with the results from model verification. Table 38. Model and NVUM Estimates of Permanent Resident Local Visitor Recreation Use for Out-of-Sample Forests. Model NVUM Estimate Estimate Lower Upper Region 2 Arapaho- Roosevelt NF 2,325,266 4,003,679 3,180,636 4,858,144 Rio Grande NF 183,341 753,544 463,998 1,070,810 San Juan NF 553,017 951,972 703,868 1,222,135 Region 9 Green Mountain NF 568,255 1,459,691 1,181,227 1,756,025 Hiawatha NF 92,644 279,962 230,528 333,743 Superior NF 168,510 2,006,922 1,538,942 2,506,133 White Mountain NF 191,044 471,622 296,058 671,081 Mid-distance model predictions fall within the NVUM estimates for two of three of the Year 1 Region 2 national forests (Table 39). Mid-distance model estimates for Region 9 national forests are all below the NVUM estimates. The Region 9 national forests used in model validation include some of the premier recreation forests in that Rfigion. In particular, the Green Mountain NF and the White Mountain NF located in the 139 New England portion of the Region offer excellent recreation opportunities. Assuming NVUM estimates are correct, the constructed Mid-distance models likely do not capture the “attractiveness” of these premier Region 9 national forests. Better measures of forest attractiveness may improve the ability of the Mid-distance models to predict recreation use at national forests such as the White Mountain NF. Table 39. Model Predictions and NVUM Estimates of Mid-distance Segment Recreation Use for Out-of-Sangrle Forests.‘I OLS Tobit NVUM Models Models Estimate Lower Upper Region 2 Arapaho-Roosevelt NF 378,554" 302,062“ 309,323 295,844 415,587 Rio Grande NF 277,749 461,068 131,098 90,865 271,829 San Juan NF 173,253 245,964* 273,200 234,152 462,819 Region 9 Green Mountain NF 178,750 358,706 690,772 537,698 860,306 Hiawatha NF 47,410 199,278 135,298 106,615 167,362 Superior NF 259,214 619,555 1,047,040 780,706 1,340,882 White Mountain NF 130,156 398,000 1,321,450 872,923 1,796,634 ' Combined recreation use of Day and Overnight trips. NVUM estimates exclude recreation use where the NF is not the primary trip purpose. * Model prediction falls within the 80% confidence interval of the NVUM Local visitor use estimate. The Percentages of forest-level Local visitor use, as estimated from the constructed Local and Mid-distance models, are consistent with the NVUM estimates for four of the seven out-of—sample national forests (Table 40). Model-based Local percentages are below NVUM estimates for the Rio Grande NF and the Superior NF and higher than NVUM estimates for the White Mountain NF. These validation results are Similar to those found under verification. 140 Table 40. Local Percentage of Local and Mid-distance Segment Recreation Use based upon Model Prediction and NVUM Estimates, Out-of-Sample Forests. Local Percentage, Local Percentage, Local+OLS Mid- Local+Tobit Mid- distance”I distanceb NVUM Region 2 Arapaho-Roosevelt NF 86% 89% 93% Rio Grande NF 40% 28% 85% San Juan NF 76% 69% 78% Region 9 Green Mountain NF 76% 61% 68% Hiawatha NF 66% 32% 67% Superior NF 39% 21% 66% White Mountain NF 59% 32% 26% ' Local model prediction/ (Local prediction + OLS Mid-distance prediction) b Local model prediction/ (Local prediction + Tobit Mid-distance prediction) Note: alternatively, the percent of primary purpose Mid-distance visitor use is (1- Local percentage) Discussion of Model Results A Clear difficulty in evaluating the performance of the Local and Mid-distance recreation use models is that both the model predictions and the comparison values (i.e. the NVUM figures) are estimates of the actual recreation use. Given disparity in the estimates, it is impossible to definitively determine which (the model predictions, the NVUM figures, or both) is “incorrect”. Even consistency between the two estimates does not definitively confirm that recreation use has been correctly quantified—though it lends support to both estimates. At the finest level of evaluation (i.e. number of visits), model predictions were frequently inconsistent with the NVUM estimates. The promising exception to this was 141 the performance of the Mid-distance segment models estimated for Region 2 national forests. Summed output from Local and Mid-distance models (the second level of evaluation) also was generally inconsistent with the NVUM estimates. Use was typically underestimated in Region 2 while there was no discemable pattern in Region 9. At the most aggregate evaluation level (percentage of Local recreation use) model outputs actually compare well with NVUM estimates. In particular, Region 2 predictions are almost entirely consistent with the NVUM estimates. In Region 9, considering only Local percentages from the “Local + OLS Mid-distance” column, model predictions are largely “in the ballpark” with the NVUM estimates. Assurrring NVUM is “correct”, the models appear to predict reliable estimates of the relative percentages of visitors from the two distance segments considered. The ability to estimate relative percentages of use by distance segments may be beneficial for resource planners when identifying national forest “recreation markets”. One confounding issue (that remains unresolved) in comparing model predictions with NVUM use estimates is seasonal homeowner recreation use. Seasonal homeowner recreation use is clearly identified in the model predictions but is not clearly identified in the NV UM use estimates. There is no single question in the NVUM survey that can be used to identify national forest users recreating on the national forest coincident with use of their seasonal home. In NVUM Year 4, seasonal homeowners can be partially identified via their responses to two questions; however, correct identification depends largely on how respondents interpret the questions. Regardless, if a seasonal homeowner 142 were surveyed it is not clear what ZIP code (permanent or seasonal residence) the respondent would provide as the ZIP code question stated only “what is your home ZIP code?” rather than “what is the ZIP code of your permanent home?”. Therefore, it is unclear in what distance segment (i.e. Local, Mid-distance, and Long-distance) that seasonal homeowner recreation use should be counted. Given this uncertainty, predicted seasonal homeowner use cannot be Simply added into one distance segment or another and thus remains largely excluded from comparisons with NVUM estimates. This imprecision will more greatly influence the comparisons between model predictions and NVUM estimates for forests with extensive seasonal homeownership and corresponding use. In future years, it would be beneficial to include distinct questions to identify seasonal homeowners and the locations of their seasonal and permanent residences. Irrespective of the prediction comparisons, the parameters and coefficients of the constructed models are largely consistent with theory. Participation rates and annual visit frequencies estimated for the Local models are downward sloping (for the most part) and appear reasonable. For the Mid-distance models, the coefficients on travel distance are negative and significantly different from zero, as expected. Significant coefficients on other variables in the Mid-distance models appear reasonable and are nearly always consistent with expectations. The approach to modeling recreation use undertaken in this research differed from Other studies in that the recreation use of those living at different proximities to the national forest and completing different trip types (in the case of the Mid-distance 143 segment) were modeled separately rather than in one single model. The aim of this approach was to capitalize on differences in how visitors respond to the factors influencing recreation use (e. g. travel distance, resource characteristics, socioeconomic conditions, etc). Two common Challenges to travel cost modeling are 1) limited variability in travel distance (or cost) and 2) disparities in on-site time between user groups. Separation of models for Local and Mid-distance recreation use likely controlled for both of these problems. Travel distance is relatively similar for those living in the Local area (due largely to the small range in distance) and travel distance appears to have limited impact on marginal visitation rates of those living in this area (Chapter 3, Figure 3). Conversely, the Mid-distance segment can be characterized by highly variable travel distance and decreasing marginal visitation corresponding to increases in travel distance. Concerning on-site time, comparisons of visit durations for Local and Mid-distance visitors revealed statistically significant differences within trip type between the two segments. Use of separate models for these two groups controlled fOr this statistical difference in visit length. The development of separate zonal travel cost models for day and overnight trips highlighted the differences in relationships between recreation use and factors influencing recreation between the two trip types that might not be captured in a single zonal travel cost model. In three of four day trip models, travel distance was the sole independent variable with coefficients statistically different from zero. Comparatively, at least two 144 independent variables (more frequently three or more) were statistically significant predictors of overnight recreation use. Separation of Mid-distance visitors into day and overnight trips also controlled for problems related to differences in on-Site time that may have otherwise influenced the performance of the travel cost models. 145 CHAPTER 5 SUMMARY, CONCLUSIONS, AND POLICY IMPLICATIONS Introduction This closing chapter serves as a final discussion of the research. In the first section, the research objectives, methods, and results of the research are briefly restated. Based upon the results, the study conclusions and implications for policy are identified in the next two sections. The Chapter closes with a discussion of the study limitations and an outline of recommendations for future research. Summary The general objective of this research was to characterize and model the recreation use of USDA FS visitors classified into three distance-based visitor segments. The four Specific objectives of the study were to: 1) identify the recreation behavior, activity participation, and consumption patterns of Local, Mid-distance, and Long-distance USDA FS visitors in USDA FS regions 2 and 9, 2) statistically compare the recreation behavior, activity participation, and consumption patterns of visitors to USDA FS Regions 2 and 9 within the distance-based segments, 146 3) statistically compare the recreation behavior, activity participation, and consumption patterns of Local, Mid-distance, and Lon g-distance visitors within USDA FS regions 2 and 9, and 4) model forest-level recreation use of Local and Mid-distance recreation visitors for national forests located in USDA FS regions 2 and 9. Much of the data used in this research were obtained from the USDA FS NVUM project. NVUM visitor survey data were used to identify and complete statistical analyses of the distance segment visitor characteristics, and the NVUM recreation use estimates were employed to estimate and evaluate the predictive ability of the recreation use models. Prior to addressing the specific objectives of the study, the boundaries separating the distance-based segments from one another were identified. Based upon previous research and an examination of the regional patterns of recreation use, Local visitors were defined to originate from within 30-miles of the national forest, Mid-distance visitors from between 30 and 200-miles of the national forest, and Long-distance visitors from the “rest of the world”. Local visitors typically complete a large number of visits annually, with each visit generally being of short duration. On any one visit, these visitors typically participate in a limited number of activities, visit a limited number of sites, and recreate with few other individuals. In contrast, Long-distance visitors complete very few visits annually, recreate in large parties, visit multiple recreation sites, and participate in a number of recreation activities. For the most part, the characteristics of Mid-distance 147 visitors generally fall between the extremes exhibited by the other two segments. One characteristic of note, however, is that Mid-distance visitors exhibit the longest visit durations of the three distance-based segments. Within trip-type and distance segment, few statistically significant differences in recreation behavior exist between visitors to national forests located in the two study regions. Worth special mention are the findings of no statistical differences in the recreation characteristics of Mid-distance visitors between USDA FS regions 2 and 9. Statistical differences that were found include differences in annual visits, the duration of visits, and the party sizes of Local and Long-distance visitors on some trip-types. Across all distance segments, a greater percent of recreation use in Region 9 relative to Region 2 is associated with hunting and fishing. Conversely, downhill skiing and mountain biking are more common for all distance segments in Region 2. In addition to the above differences, Mid-distance visitors in Region 9 are more frequently engaged in cross-country skiing and hiking than comparable visitors in Region 2. Similarly, hiking is much more common among Long-distance visitors in Region 9 than in Region 2. Comparing the distance segments to one another, statistically significant differences between two or more segments were found for all the visitor characteristics under consideration. The majority of the differences are consistent with the postulated characteristics of distance segment visitors identified in Chapter 3. There were a greater number of statistical differences between segments among Region 2 visitors than among 148 Region 9 visitors. In Region 2, the recreation Characteristics of all three distance segments were frequently statistically different from one another. In Region 9, the recreation characteristics of Local visitors were typically unique from those exhibited by Mid-distance and Long-distance visitors. Many of the statistical differences between the distance segments can be traced to differences in trip-type. Local visitors are most likely to complete day trips, Midsdistance visitors are most likely to complete overnight visits, and Long-distance visitors take more “not primary” trips than any of the other two segments. When accounting for trip-type, the recreation behavior of visitors in the three distance segments is more similar. However, statistical differences do remain for visit duration and party size variables. Differences in the number of visits annually cannot be tested within trip-types. Considering recreation activities, Local visitors are more likely to fish than visitors in the other two segments. In Region 2, Local visitors more frequently engage in hiking than visitors in the other two segments, while in Region 9 Local visitors more frequently hunt and drive for pleasure than other visitors. Long-distance visitors more frequently visit the national forest to complete nature-related activities (i.e. viewing scenery, viewing wildlife, nature study, or visiting a nature center) than visitors in the other two segments. Long-distance visitor are also more frequently engaged in hiking than Mid-distance visitors. The most frequently reported primary activities for Mid- distance visitors are downhill skiing and hiking. Additionally, these visitors are more 149 likely than the other two visitor groups to be using OHV’s or to be camping in developed portions of the forest. Separate models were developed for the Local and Mid-distance visitor segments. The recreation use of Long-distance visitors was not modeled as the factors influencing the recreation are not clearly understood. Local recreation use was modeled using population figures obtained from the 2000 Census and participation rates and annual visit frequencies estimated from NVUM survey data and recreation use figures. Mid-distance recreation use was modeled via multi-site zonal travel cost models incorporating variables related to travel distance, income, the degree of rurality, the availability of substitutes, forest acreage, and number of forest units within an administrative forest. For the most part, the parameters and coefficients of the models are consistent with theoretical expectations. In the case of the Local models, the estimated participation rates generally decrease with increasing distance to the national forest. The estimated annual visit frequencies of Local visitors are reasonable and consistent with expectations. Considering Mid-distance models, the coefficients on travel distance are negative and significant in every case. When Significant, the coefficients on other variables in the Mid- distance models are reasonable and largely consistent with expectations. Forest-level predictions of recreation use were generally not consistent with the NVUM estimates of recreation use. In Region 2, only one forest-level prediction of Local Visitor use was within the confidence interval of the NVUM estimate. In Region 9, three predictions were within the NVUM confidence intervals. Region 2 model predictions of 150 Local use typically fell below the NVUM estimates while there was no particular pattern of discrepancy for the Region 9 Local model predictions. Predictions of Mid-distance recreation use were more frequently consistent with the NVUM estimates. Consistency with the NVUM estimates was more frequent in Region 2 than in Region 9. Percent errors for all models were generally below 50%, though percent errors for some forest- level predictions were quite high. At the most aggregate comparison level (percent of Local visitor use), model predictions and the NVUM estimates were largely consistent. This was particularly true in Region 2. In the modeling approach adopted here, Local visitors were modeled separately from Mid-distance visitors and the recreation use of Mid-distance day trip visitors was modeled separately from recreation use of Mid-distance overnight visitors. In a conventional modeling approach, the recreation use of all these groups would likely have been modeled in aggregate. Considering just Mid-distance visitors, the variables found to be statistically significant predictors of recreation use were quite different. In all but one case, the only statistically significant coefficient in models of day trip recreation use was travel distance. In contrast, in models of overnight Mid-distance recreation use, at least two variable coefficients (and more commonly four) were Statistically different from zero. Forest acreage, number of forest units, RUC code of the origin, and travel distance commonly influenced overnight Mid-distance visitor recreation use. 151 Conclusions Segrnenting recreation users based upon their proximity to the recreation resource yields visitors groups with distinct recreation consumption patterns, recreation behavior, and primary activity preferences. Notable and defining differences occur in the number of visits completed annually, the duration of those visits, party size, and trip-type. Worthy of particular note are the marked differences in trip type frequency. Nearly 75% of Local visitors complete day trips, more than 60% of Mid-distance visitors complete overnight trips, and Long-distance visitors can be typified by the percent of “not primary” visits undertaken (33%). Even when accounting for trip type differences, many of the distinctions and patterns in recreation characteristics remain evident—though not necessarily statistically significant. Within distance segments and trip-types, the recreation characteristics of national forest visitors in Regions 2 and 9 are generally analogous. Statistical differences do exist in some cases for Local and Long-distance visitors. The most notable difference between the two regions, within segments, is a dramatic difference in the percentages of visitors engaged in downhill skiing. In particular, Mid-distance and Long-distance visitors in Region 2 are much more likely to be downhill skiing than their Region 9 counterparts. Even with a clear difference in recreation activity, there are no statistically significant differences in the recreation characteristics of Mid-distance visitors between the two study regions. 152 Using NVUM recreation use estimates, N VUM visitor survey data, and Census 2000 population estimates, the participation rates of Local segment visitors were determined as the initial step in development of the Local recreation use model. In general, as expected, the participation rates of Local populations decrease with increasing distance to the national forest. An estimated 87 % and 39% of those living within the proclamation boundaries of Region 2 and Region 9 national forests participate in national forest recreation, respectively. At the farthest distances, the participation rates of Local visitors in Regions 2 and 9 fall to approximately 10% and 3%, respectively (excluding the 30-mile band in Region 2). The greatest numbers of visits annually are completed by those living in the two nearest distance bands. The number of visits completed annually is relatively constant for those living beyond 15 miles from the forest boundary. Multi-site zonal travel cost models of day and overnight Mid-distance recreation use were constructed for USDA FS regions 2 and 9. Mid-distance segment day trip recreation use is related primarily to the travel distance between visitor origin and the national forest. Conversely, the recreation use of overnight Mid-distance visitors is influenced by a number of factors, including travel distance, substitutes, and forest characteristics. Both model formulations adopted in this study appear appropriate for use in estimating the recreation use of USDA FS visitors at this scale of analysis. It is quite difficult to evaluate the predictive ability of the constructed models. The central limiting factor in an evaluation of the models is that the “true” forest-level recreation use of Local and Mid-distance visitor segments is not known. Consistency 153 between the model predictions and the N VUM estimates (when it occurred) only implies that the two are consistent. Some additional complicating factors in using the NVUM estimates to evaluate model predictions is the inability to quantify recreation use associated with seasonal homes in the NVUM estimates and the significant number of NVUM respondents (and associated recreation use) with an unknown origin. While there is no evidence to suggest that the current models are sufficient, the forest-level recreation use of Local and Mid-distance visitor segments can be modeled using the basic approach adopted in this study. That is, predicting forest-level recreation use based upon estimates of the populations living in proximity to national forests and knowledge of the recreation consumption patterns of those populations is possible given some model refinements and a more definitive evaluation. A clear benefit of modeling recreation use via the approach adopted here is the ability to predict future recreation use levels given estimates of future conditions under the assumption that all relationships between recreation use and population characteristics remain the same. Policy Implications The distance-based approach to segmenting recreation visitors is grounded, somewhat, in the classification of outdoor recreational uses and resources developed by Clawson and Knetsch (1967). The approach presented here classifies visitors based upon their proximity to the resource while Clawson and Knetsch (1967) Classify recreation resources based, largely, upon proximity to visitor populations. The observed recreation 154 behavior of Local, Mid-distance, and Long-distance visitors largely correspond to the postulated characteristics of recreation visitors to Clawson and Knetsch’s user-oriented, intermediate, and resource-based recreation resources, respectively. Given their distinct recreation Characteristics, distance-based segmentation can provide a framework for classifying USDA FS visitors. Ultimately, distance-based segmentation represents another way for USDA FS planners and managers to consider recreation use. Incorporating the distance-based segmentation adopted in this study with the traditional activity-based segmentation may increase the information provided by both. Currently, many national forests are undertaking “recreation niche” or “recreation market” analysis. The goal of these analyses is to identify the “role” or “Special opportunities” that the individual forests play in providing recreation opportunity to the public. Distance-based segmentation seems to offer an excellent framework for individual national forests to identify, at least in part, their role in the provision of recreation opportunities both now and under future conditions (using predictions of future visitor segment use). For example, a national forest that determines its recreation use is predominantly associated with Local visitors could identify their recreation market as serving frequent visitors who are primarily on day trips, recreating in small visitor parties, and spending only a short period of time on the forest during any one visit. Given the observed pattern of recreation activities presented here, popular recreation activities on this national forest would likely include hunting, fishing, and hiking. Similarly, a 155 national forest attracting a significant number of Long-distance visitors could expect visitors that visit very infrequently, perhaps for the first time, recreate in large parties, and visit a number of recreation sites. A forest attracting a number of Long-distance visitors may do well to offer a variety of interpretive recreation activities since Long- distance visitors frequently engage in passive nature-related pursuits. Further understanding of the motivations of distance-segmented visitors would add the understanding of a forest’s market. Comparisons between visitors in Region 2 and Region 9 revealed few statistical differences after accounting for distance segment and trip-type. Similarly, Stynes and White (2003) have found that USDA FS regions explain very little of the variation in visitor spending. Both of these findings are counter to the notion that USDA FS regions can be used to explain or delineate visitor characteristics. Based upon the results of this study, visitor proximity to the resource and trip-type probably capture more variation in recreation behavior than USDA FS region. Region probably does explain some variation in recreation activity propensities given that some USDA FS regions do have unique natural features conducive to some specific activities. In this study, zonal travel cost models were employed only for a specific subset of recreation use. In a conventional application, a single model may have been estimated for all recreation use or for all recreation use within a certain trip-type (e. g. day trips). At a minimum, it appears that separate travel cost models should be estimated for visitors engaged in different types of trips. Based upon the results of this study, the factors 156 commonly accepted as influencing recreation use (e.g. distance, substitutes, amenities) likely influence the recreation use of day and overnight visitors differently. Given variation in the recreation characteristics of Local, Mid-distance, and Long-distance visitors, it also seems likely that distance, substitutes, and Site amenities influence the recreation use of these groups in different fashions. Models that combine visitors originating from these different distance segments likely fail to capture these different functional relationships. However, a comprehensive evaluation of alternative model formulations is required to determine this for certain. Limitations and Recommendations for Future Research This study is based upon the revealed behavior of recreation visitors (as quantified via the NVUM survey data). The recreation characteristics identified in this study do not provide information related to the desires of the distance-segment populations, but rather their consumptive patterns. Additional surveys are required to identify the desires of the distance-based segment visitors and those of the general population. Similarly, the motivations for recreation of national forest visitors were not quantified in this study. It is likely that the motivations of visitors within the three distance segments differ, but additional study is needed to quantify these differences. A greater understanding of visitor motivation in the context of proximity to the recreation resource would likely be very informative. 157 The models in this study were constructed using data from 19 administrative national forests located in two USDA FS regions. The other 100 administrative national forests located in seven other regions offer a variety of recreation Opportunities and serve diverse recreation visitors. The models presented here apply only to visitors to national forests located in USDA FS Regions 2 and 9. There is no evidence to suggest that the parameters and coefficients of the estimated models are transferable to national forests located in other areas. Likewise, there is no support for applying the parameters and coefficients estimated in this study for recreation use on lands managed by other government agencies. Two measurement problems frequently arise in estimating zonal travel cost models: quantifying the site characteristics that influence the behavior of recreation visitors and quantifying the availability of substitute recreation Sites. Coarse measures of both were employed in this study. Different approaches to quantifying the attractiveness of national forests and national forest substitutes may identify relationships between these factors and recreation use not found in this Study. Additionally, the incorporation of substitute recreation opportunities managed by state and/or local agencies in the models of recreation use may result in different relationships. In the course of this study, many opportunities for future research have been identified. 0 Seasonal homeowners likely represent a Significant component of national forest recreation use, particularly in some portions of the country. Currently it is difficult to 158 identify their recreation use, and little is known about seasonal homeowner visitors. Future studies should be directed at identifying the national forest participation rates and recreation characteristics of these users. With the recent expansion in the second- home market, the extent of seasonal homeowner recreation use will likely increase. Local visitor recreation use constitutes more than 50% of the recreation use in regions 2 and 9. Given the importance of their recreation use, a household survey directed specifically at populations living around national forests would be beneficial. Of particular interest is greater information concerning the consumption patterns, motivations, and desires of these individuals. This research assumes that Local segment visitors recreate at the nearest national forest when more than one forest is located within 30 miles of the visitor’s origin. The validity of this assumption can be examined using the NVUM survey respondents and spatial databases of visitor origins and national forest boundaries. Such an analysis would yield a greater understanding of local visitor consumption patterns. The boundaries between the three distance segments were identified via an aggregate approach. Additional research aimed at better identifying the boundaries between Local, Mid-distance, and Long-distance visitors would be beneficial. Further refined boundaries could be identified via analysis of changes in recreation behaviors or based upon the motivations of users. An examination of patterns of day trip use and overnight trip use may be particularly useful in future delineations of the boundaries between distance segments. The Local visitor models show promise in the ability to estimate forest—level recreation use of local populations. Future development of these models is 159 appropriate. Specific future tasks may be to develop participation rates applicable to populations within certain age classes, better understanding of how local users choose between federal, state, and locally—managed recreation opportunities within the local area, and identification of ways to incorporate more local variation into the models. Distance is a key variable influencing recreation use. In this study, a Euclidean-based approach incorporating an adjustment for barrier crossings was adopted for calculating distances from populations to forest resources. Alternate approaches to estimating distance post-hoe include using simple Euclidean distance, network distance, or the development of a cost surface using a GIS. Further research is needed to compare these alternate post-hoc approaches to distance calculation and the potential impacts of their use on recreation research. In a recent revision of the NVUM survey instrument, survey respondents are asked to report their travel distance. This offers an excellent opportunity to identify the patterns in how visitors report travel distance (e. g. rounding) and to evaluate the consistency of post-hoe travel distance calculations with reported travel distances. While the Mid-distance models perform fairly well, some further refinements may lead to more informative models. In particular, development of an alternate approach to aggregating visitor origins into zones may lead to better identification of the factors influencing recreation use by these visitors. Additionally, as stated, refinements in quantifying both the national forest attractiveness and substitute availability may be beneficial. Currently only federal lands are considered as substitutes. Future studies may choose to include recreation areas managed by state and local governments as potential substitutes. Additional information from recreation users regarding the site 160 Characteristics influencing their recreation decision and the other sites they considered would be very useful. The recreation use of Long-distance visitors was not modeled—largely due to uncertainty into what factors influence the recreation use of these individuals. Development of models of Long-distance recreation use, coincident with refinement of the other two models, would create a package of models that could estimate the expected recreation use from each distance segment; thereby providing estimates of total recreation use. 161 APPENDIX A 162 Introduction This Appendix includes additional detail on the NVUM process. Included are detailed descriptions of how proxy sites are incorporated in the NVUM process, the process for identifying sample days within national forests, Changes to the visitor questionnaire that have occurred during the first NVUM cycle, and of the mathematical procedures used to estimate forest-level recreation use based upon NVUM traffic counts and visitor surveys. Proxy Sites At some sites within national forests visitors are required to pay a user fee or obtain a recreation permitl. The types of sites that frequently require permits or fees include ski areas, developed campgrounds, and some Wilderness areas. These sites differ from others because reliable estimates of visitation, the permits or fees (the proxy), already exist. As such, visitation at “proxy Sites” can be estimated by identifying a conversion factor between the proxy and the number of site visits. The appropriate conversion is identified via additional questions asked of survey respondents on proxy site sample days. On proxy sites the strata is based upon the proxy site type and the type of proxy used (e. g. OUDS with fee envelopes) rather than expected level of exiting recreation traffic. Because visitation estimates could be developed based primarily on the proxy, fewer visitor surveys are administered on sample days at proxy Sites. ’ Some sites require the user to purchase a Recreation Fee Demo sticker or another broadly applicable use permit. Broadly applicable fees and permits are not considered here. 163 Sample Day Selection Approximately 200 sample days are allocated to each administrative forest. These 200 sample days represent, on average, 64,000 Site days identified for each administrative forests (English et al., 2002). On an individual sample day a traffic counter is installed at the site for a 24-hour period and visitors interviews are conducted during a six-hour interview period (from either 8 am to 2 pm or 2 pm to 8 pm, altered as needed to constrain visitor surveying to daylight hours only). The number of sample days to occur in each Stratum within each administrative forest is determined at the USDA FS regional level. The three-step allocation process is described in detail in English and others (2002) and outlined here. Each region is allocated 200 sample days per adrrrinistrative forest sampled in a given NVUM year. First, each administrative forest is allocated eight sample days for viewing corridor sites. Next, each administrative forest is allocated, via a stratified approach, up to 50 sample days for the proxy strata. Finally, eight sample days are allocated to each non-proxy strata within each administrative forest (each administrative forest may have up to 12 non-proxy strata). Any remaining unallocated Sample days (at the regional level) are allocated across strata across administrative forests based upon the product of 1) the standard error of the strata estimated from previous NVUM years2 and 2) a weight of 20 for high use sites, 10 for medium use sites, and one for low use sites. The effect of this final step is to allocate the remaining sample days to those strata with perceived high levels of recreation use and large variation in observed recreation use. 2 No standard errors for strata existed in the first two years so this allocation was based solely upon the strata weights. 164 The selection of individual sample days (dates and locations for sampling) within strata within individual administrative forests is completed by NVUM national-level staff using a simple random sample with the following minor adjustment. Logistical limitations constrain sampling to three or fewer Sites on a single calendar day on a single administrative forest. For those calendar days on individual forests where more than three sample days are selected, NVUM personnel retain two of those sample days, place the other(s) back in the population of site days and then resarnple the necessary number of sample days from the population. In the first cycle of NVUM no attempts were made to insure that a temporally or spatially representative group of sample days within strata within an administrative forest was obtained. Survey Revisions During the first NVUM cycle the survey form has undergone several revisions. Copies of the survey forms (including the economic addition) used in the first cycle are available from USDA FS NVUM personnel. Most of the revisions during the first NVUM cycle were made to correct spelling and/or grammatical errors and to change the survey form layout. However, in the final year of the cycle a significantly revised economic supplement was introduced. Changes introduced in this final form (Revision 5) came about, in large part, due to the analyses of survey results collected in the initial years of NVUM. AS result of these changes there are some differences between the year 4 data and the data collected in the previous three years. 165 Three primary changes occurred in Revision 5: 1) substantive changes to the “primary purpose” question, 2) changes in questions relating to the visitor’s trip length and stay in the local area, and 3) removal of the “Sharing question” and explicitly requesting trip expenditures per party rather than per person. The “primary purpose” question is used to identify visitors whose trip to the forest is secondary to some other trip purpose. Identifying “non-primary” visitors is integral to correctly estimating economic impact and estimating use values, and also provides some information relating to the motivation of the visitor. In the first three years, trip purpose was determined via a two part question (questions 3 and 4 economic addition, forms 1 — 4). In the course of analyzing the economic survey data it was determined that these questions did not definitively determine whether the forest was the primary reason the visitor was away from home and that the answers provided were frequently inconsistent. In Revision 5 these questions were removed and replaced with a single question on the general survey (Question 11 general survey, Revision 5) that specifically detemrines whether recreation on the national forest is the primary reason the visitor is away from home. Use of the revised question yields a greater percentage of non-primary visitors (Stynes and White, 2005b). Visitor trip length is primarily used to determine whether visitors are on day or overnight recreation trips. It is also used to place Spending on a per-night basis and can be used in analysis of visitor recreation behavior. Revisions l — 4 determined trip length, in terms of days and hours away from home, via a single question (Question 2 economics addition, revisions 1 — 4). While the original question determines total trip length what is 166 more informative for economic impact analysis and planning applications is the length of Stay in the local area (preferably in terms of nights rather than days). In Revision 5 respondents report both the number of nights away from home (if any) as well as the number of nights in the local area (questions 1 — 3 economic addition, Revision 5). In addition, the respondent is asked to identify the types of overnight lodging used in the local area (Question 4 economic addition, Revision 5). Using the previous surveys, visitors passing through the local forest area and staying overnight away from home were classified as overnight visitors when they Should have, in fact, been classified as day trip visitors (since they were not staying overnight locally). The changes to the trip length question alleviates this problem and results in fewer overnight visitors in the Year 4 sample. NVUM Visit Estimation The NVUM traffic counters and visitor surveys are used together to develop estimates of total visitation on individual sample days for non-proxy sites. Estimates of sample day visitation for non-proxy sites are computed as follows (all formulae in this section are adapted from English et al., 2004): SVhi = Chi * Ph *Vh where SVh, is the estimate of site visits for given stratum h on sample day i, Chi is the number of cars obtained on the traffic counter adjusted for number of axles and two way traffic, Pb is the proportion of exiting vehicles that are last exiting recreationists (estimated from the visitor survey) averaged across all sample days in the strata, and V1, is the number of persons in last exiting recreation vehicles (estimated from the survey) 167 averaged across all sample days in the strata. The average number of site visits across all sample days in given stratum h are computed as: i=1 "h where m, is the number of sample days in stratum h. To estimate the number of total Site visits for stratum h, the average site visit estimate for stratum h is multiplied times the number of site days (Nb) in the population of stratum h: SVh = N, * S—Vh, Site visit estimates for proxy sites are developed from the following: 1) the annual total proxy count for a given site k within a given proxy stratum h (PM), 2) the proxy compliance rate for a given site k within a given proxy stratum h (Cth), and 3) a conversion factor (21.) to facilitate transfer from the proxy count for stratum h to a visit count for stratum h. The first two components are combined to develop the effective “compliance based proxy count” for each site day (PChk): P pchk =__hk_ Cth . The site day proxy count is combined to form a mean daily proxy count for the stratum h: K ZPChk Z k=1 9 N hk where Nhk is the number of Site days for site k stratum h. 168 The conversion factor Z1. is constructed from visitor survey responses completed on the proxy sample day, combining 1) the reported number of proxies completed per visitor or group (e. g. number of fee envelopes used to cover an individual or group while camping) and 2) the reported number of people each proxy covers (e. g. number of people covered by a single fee envelope). Combined they form an aggregate conversion factor for stratum h: "h 250,,- 2h : i=1 "h 25R,"- i=1 9 where SGh, is the sum of group size for stratum h on given sample day i and SRM is the sum of number of proxies for stratum h on a given sample day i. From the compliance based proxy count and the proxy conversion factor the average site visit estimate for proxy stratum h can be computed: "517;. = Z}. *F'C'h . The total number of Site visits for proxy stratum h is computed by multiplying the average site visits for stratum h by the number of site days in stratum h: S Vh = S—Vh * N h The average number of Site visits for all sample days, combining proxy and non-proxy sites, on a given national forest is computed as _. H _ SV = 2W}, *SVh , h=1 169 where u@— Nh _ _H . :ZNh h=1 The total number of site visits for a national forest is SV=SV*N, where N is the total number of site days on the national forest. While the number of site visits is of interest for some applications, the number of visits to the national forest as a whole is generally more useful. Since many visitors will recreate at multiple locations within an individual forest the number of site visits overestimates the total number of national forest visits. To convert site visits to national forest visits survey respondents are asked to report the number of Sites visited and the number of days spent in the GFA during the current recreation visit (NShjj). With this information the number of national forest visits on a given sample day (NFV,) can be computed as: NFVh, = Chi * P], * CHARM. CBAR, is computed as n V .. CBARhi = 1 2 hi} LEvh, FINShij ’ where LEV, is the number of last exiting vehicles in Stratum h on sample day i, thj is the number of visitors in vehicle j in stratum h on sample day i, and NShjj is the number of 170 sites visited by vehicle j in stratum h on sample day i. The average number of forest visits across all sample days in given stratum h are computed as —— "h NFV . NFVh = Z——’"—. i=1 "h The number of national forest visits for stratum h is then computed as ~th = Nth *Nh. To estimate the number of national forest visits from the proxy sites the conversion factor ANFa is developed: nh ZSCh, ANFh =t_=l_ nh 9 25R,“- i=1 where SC” is defined SC - it ’"7 ) hr - . j=l NShij The mean number of national forest visits to proxy Sites is then computed as NFVh = ANFh *P—Ch. From this the total number of national forest visits for stratum h is ~th = NFVh * Nh . The average number of national forest visits for all sample days on a given national forest is computed, combining proxy and non-proxy sites, as 171 __ H _ NFV= zwhi‘NFVh, h=1 where N h W}, — H 2”}: h=1 The total number of national forest visits for a given national forest is then NFV=NFV*N, where N is the number of site days for the national forest. 172 APPENDIX B 173 J I awhm ohm; awhm Edam SNJ. “mmfi mom.: cavdw EMS 35.3 mmad 3%? Cam; Rad m2 0533 comm cam 08% coca www.N_ gwd mmofi vwad mmnd Nomfim min: $w.- «3.2 #9.. _ "HZ 00.565 94.2 amuse. mm}. wotm Rog 24.2 93. 81m 83. «2. 8m «dam SQN mm»; m2 .2585 $3.2 omofi whufi 3&3 03.9 2 m.mm w. m. _ m www.mv 08.34 ovficv 89mm hmbfi ‘43.: www.mv m2 Eosewcocovz mvm; Ewe.— mmm moo; $34.9 «3.: owwfi m2; 81m hmwfi 3m.m 08.0 Emfi _v~.N ..wZ £95. :32 02 Rm enhm ~23. 3 go macaw 02.3. 80.3 ommdm «3.2. mwwdm 8o.m 3.3 95.5 ”2 gmmcaztcoam mafia mm a .N New; om ~ mum van at; won and 2&3 Eton momxam Swag Soda "—2 c2803 wows. Sow 08.3 owwéz $2 2 Snow Sham 45.2 34.8 Seam «made. one; www «8 "2 «32520 mood 086 96.2 0864 Sofia Swdm S wém 332 83% 33.8 98.9 wmhde «3.8— 8%: ".2 .282 -cowofigconu 0m fix of 90m wmo cum wwwfi Stem 51¢ 03.? am 3% woofim 03.2 ”48.3.. www.mm "HZ Cosmo—2 a eons— hmofi o o o own; 8%: $3.: 2.0.3 85an wmndmm Sham wnmém can mowd "2 .33 BE? woo.m mmwd «mm; m: 0 Eva: Nag. 03; mm: .N 3%.: N86 #3 9% wow; "—2 0:27.25 mvw o o Gfim Rafi ~26 flux—m mondm 94$: 28.2. www.cm wmfimm www.mv ownm ”—2 2.5% mmafi m2 3&2 NNQm o Siva Smd EVO— mbaé 855 ~33” 9A.: cmmé Safe "—2 755$ new 3E ~36 :wé _No.m enhm Sad 43d 2 flu 23. 8: mafia ewe mm~.m o o m2 8.3562 ave; w _ h; mom; Ewe new and 31mm one; _ Rwdm mmofim w ~ 9m are women 36.3 ml 30m 05232 036 08.x. 0 ~34 mwmé 20.2 9%.? nmmdm— www.mm 39mm Rafi 03.6 new. H Sod m2 6:: .362 285 o o mmv mohm $3“ vwmd amwa mwe emm 30.2 o E.— o o 2:3 xofim was. 84.: So; 0 o $3 4% «3.2 mag omwem 83 cm 34.2 o m2 Sofia N :emwom 8N mm“ :3 mm a: ma— 3 w ma cw me am me cw mm 38.—er.— .2232 A355 65m 8:3me 69:5 :3 1 en 98.... 355 3535 can teach .3 8: :euaouuom 8353mm 55; .—.m 035—. 174 .33 ...CNEE: 87:: at: Emuto>0 60:: Nan: ”womb at: 3:5 .:0 2:0 9:: 30:80.0 203 £965 . $0 $0 $0 $0 $00: $00: $00: $3‘ $50 $m0 $2 $00: $05 $:m Z>O ":7: 5.5: 80:? $0 $0 $0 $0 $0 $0 $0 $:: $::. $0m $0 $0 $8 $00 .30: ":2 :25: 2:55 $0:: $00 $00: $m0 $0 $9: $9: $2: $0 $:0 $:0 $~0 $05 $0 Z>O ":7: 2808.5 $0 $0 $0 $0 $0 $0 $0 $m $0 $8 $0 $0 $m~ $00: .95 ":7: 280800 $0 $00 $0 $~0 $00: $00: $00: $00: $00 $3: $::. $00: $00: $50 Z>O ":7: 0:0: $0 $0 $0 $0 $0 $0 $0 $0 $00 $0 $00 $0 $0 $mm b5 ":2 25: $00: $00 $00: $m0 $0 $00: $m> $00: $:0 $0 $0 $00: $05 $:0 750 ”:7: :30: cam 8?: $0 $0 $0 $0 $0 $0 $0 $0 $0m $00: $0:V $0 $m~ $0: 9am: ":7: :33: cam 8?: $00: $00: $00: $00 $00: $00: $00: $00: $00: $00: $00 $00: $0 $0 z>o ”:7: 33.507: $0 $0 $0 $0 $0 $0 $0 $0 $0 $0 $9: $0 $0 $0 9.5 ":7: 38507: $00: $00 $0 $00: $00: $00 $0:. $00: $m0 $00 $0 $00 $00 $3 750 ":7: 33: 25:00:): $0 $0 $0 $0 $0 $0 $:: $0 $0 $0: $0 $0 $:: $0m 9am: ":7: 30m 05060:): $00 $0 $0 $00 $00: $00: $00 $00 $0:. $00 $0 $0~ $~m $0m Z>O ":2 6:3 .32): 055 $0 $0 $0 $0 $0 $0 $0 $: $0 $00 $0 $0 $0 $:0 man: ":7: 6:3 .303: 055 $0 $0 $00: $N0 $00: $00: $0:. $5 $:0 $00: $0 $~0 $0 $0 750 ”:7: $::: U6.0:: $0 $0 $0 $0 $0 $0 $0 $m $:m $0 $0 $0 $0 $0 bun: ":7: 2:5 V:35 $vw $00: $00: $0 $0 $00: $00: $wb $v0 $0h $0 $N0 $0 $0 Z>O ":7: 52:05 $0 $0 $0 $0 $0 $0 $0 $~N $0 $v~ $0 $0 $v0 $0 .05 ":2 Eosmx: 8a m»: 3: mm: 2.: mm: a: ma 8 m. cm mu 8 mm 2:. .33.: .2282 .05 02:5 05:: 8.835 Mu =0:we~: :: 38.5,: 3.8507: .8: 055 3530:: 0:: 00a I 00 .3 02.5. 335930 :05: man :0 mow—3:33.: 2:: 00 8:08:03: 552/: .NA: 030,—. 175 .35 .50th 82: 02.5 Emeo>O .09: :00 “womb QB 35 mo 25 BE 3.20220 203 20:05 . $00 $2 $00: $00: $00: $00: $00: $00 $2 $00 $0 $0 $00: $0 z>o "220203 $00 $2 $0 $0 $0 $0 $0 $3 $2 $2. $00: $00 $0 $8 28 222%? $00: $00: $00: $02 $2 $00 $00: $: $00 $2 $00 $5 $00 $2 z>o "2280320 $0 $0 $0 $0 $00 $2 0 $0 $00 $00 $2 $0 $2 $00 $00 25 "Hz 8:320 $00 $00 $00 $00 $0 $3. $00 $2. $0 $00: $0 $00 $0 $0 z>o "220335 $0 $0 $0 $2 $0 $$ $0 $0: $0 $0 $0 $2 $0 $00: 28 m2 $55 $0 $$ $00 $2 $00 $3 $2 $2 $2 $8 $2 $00 $00 $2 2.5 ":2 20500802 $0 $0 $0 $0 $3“ $0 $0 $0 $2 $2 $2 $2 $8 $2 25 "Hz 22.000802 $3 $00: $:0 $00: $0 $00 $2 $20 $2 $3 $0 $00 $2 $00: 2.5 "22033.33: $0 $0 $0 $0 $8 $0 $0 $3 $2 0 $$ $2 $0 $00 $0 28 ":2 503,0 03: $3 $2 $00: $2: $00 $00 $00 $00 $00 $00 $00 $00: $00 $2 750 "228252-85: $0 $2 $0 $0 $0 $0 $2 $0 $0 $2 $00 $0 $2 $00 25 ":2 8252-005: $3. $2 $00: $00: $00: $0 $00 $00 $20 $20 $3 $2 $00: $3 z>o "2.20.8: $0 $2 $0 $0 $0 $0 $0 $0 $2 $2 $00 $00 $0 $:0 28 m2 .280: $00 $2 $00: $00 $00: $00: $00: $0» $00: $00: $0$ $00: $00 $20 750 "22030026 $0 $2 $0 $0 $0 $0 $0 $0 $0 $0 $00 $0 $2 $2. 25 .22 0300020 $0 $2 $00 $00 $2 $00 $00 $00: $20 $3 $2. $00 $00 $00 z>o mz 3.022-800.0336 $0 $00 $0 $0 $00 $0 $0 $0 $0 $00 $0 $2 $2 $2 28 "Hz s_oo_z-0omoe§§u $00: $2 $:0 $00 $00 $00: $00 $00 $2 $00 $00: $00 $00: $00: z>o "2202022 $0 $2 $0 $2 $3 $0 $0 $2 $0 $0 $0 $0 $0 $0 23 "Hz 202.022 Sn 2: 02 mm“ 0: man 0: m0 0» m0 0m 2. 3. mm 2i. 08.0,,— 3:02.02 -05 1:028:52: ...a :cmwom 5 38.8,.— ..Eéaz .80 0:0: 8:305 0:2 000 u 00 3 are 32505 0:0 :5 00 80559.0.— 2: 00 83.5.02 5:52 .02 2.: 176 o o o 9 ohm; 3%: wmv.: www.mm £0.02 528m 202.: 32% mam 81m 750 22 002m 0:55 o o o o o c o 03.2 25.: Saw: owfim o mg 8:0 .25 "2 53m 22>? vain womd SN; 2: 0 Eva: SS. «:2 o 8%: wcofi Sm vmo o Z>O 22 25:85 o o o o o o c on mmm; maqm o o no: we: .25 "Hz 25:85 o o o 30.: unqm N26 mafia 328 80.3 Smdm «3.3 wmfimm www.mv Stu Z>O "Hz 030% o o o o o o o o 80.2. o 33: c 0 m2; 2.5 m2 3:3— mmoa mm: 32m _ m _ in o mvvém mm: 822 wQO o o 2%.: 32% 22% Z>O "2 .32 50 9:0 o o o o o o o o 95.— Nonfim com; o emo; 20.: 095 "Hz 32 50 8:0 ~85 :3. :Nod 926 Sufi vmo.m EQN m3;V men; mowd woo mm~.m o o 750 "Hz 87.0502 9 o o o o o o o o o 2 m o o o .95 EA 8.00502 0.0.: «me; o Gwd Sm 2.06 w: .2 08.: m: w.m_ camdm o mmofi 32% awdm 750 m2 225 050602 c o o o o 9:.” Sod o o mafia o 0 001m 0min .95 m2 38 8.2002 wand o o moo; owmé 28.2 03.3 52:2 80.2 Sad: o omo wow amoN 750 m2 Nam: 9:20 o o c o o o o mam; ommJ 05.2 o o o Nam 209 "Hz 0.0 .0002 985 o o mmv we?“ mwo.m vwmd comm mom 3N god: c on _ o o 750 "HZ 0:5 0.02m 0 o o o o o o «N 8: o o o o o .95 m2 0:5 0.005 3.3. 8%: ~84 o o 02.: van v8.8 325 30.2 o «N ova o Z>O mz 0.855 o o o o o o o 32m .34 can o o 322 o .95 22 Eonwfi 000 m2 0: m2 0: m2 0: m0 00 2. 0m 2. 0.. mm 22. $20.2 .05 .2232 team 3:505 .N new. 3— E 88.8,.— _a:o=az you .055 35505 .3 on: 55.3.83— %mfigc 6:: man— u: mfiafiumm E2 .94— 039—. 177 000.0 000 000.0 30.0 20.0 000.0 000.: .000 000.2 000.0 0 3.0.. 000; :0 z>o 028003 000; 000 o 0 0 0 0 00.2 000.0 000.0 000.0 000.2 0 000.0 05 022.003 000.0 0.00 000.0 20.0 000.0 08.0 000.0 :0; 02.0 5.00 000.0 06.0 000.0 000.0 z>o 0230320 0 o o 0 03.0 30... 0 000.0 000.0 000.0 000.. 000.0. 000.0 000.0 08 02803000 000.0. 000.0 000; 0:.0 0 000.0 000.... 000.0 3.0 000 o 83 o o z>o 02033.0 0 o o 00 0 003 o 000 000 o o 00.0 0 000.. 03 "02 0355 0 000.0 000.0 000.2 000.0 03.00 20.0 000.00 000.00 000.00 000.00 000.0. _00.0 0010 z>o mz Eocawcocoz o o o 0 000.0 0 0 000.0 000 000.0 50.0 08.0 000.: 000.00 00a 02 Bonawcocoz 3.0 000.0 0: 08.. 30 003. 000; 000 000 000.0 0: 000.0 000 30.0 z>o 02.003. 0.02 00 o 0 0 000.2 000 0 20 30; 000 0:... 000 000... o 05 025030 0002 00 000 000.0 000.0 000.00 000.00. 000.9. 000.00 000.00 000.0. 000.2 08.0 30.0 000.00 z>o 02 300§02é052 0_ 02 o 0 000.0 0 000.0 0 83 00040 000.2 0 000 000.0 08 02 300602-055: 000.0 30.. 000; 00_ 000 o 80; 0.0 08.0 000.0 30.2 00_ 000.00 000.00 z>o 02.208: 000 0: o 0 o o o o 00.0 000.2 03.0. 002.0 0 000.0 05 ”02020800 000.0 000.0 000.: 000.00 000.5 000.00 000.00 000.0 000.00 000.00 20.00 000.0 000 000 z>o 020300020 000 000 0 08.0 0 o o o o 0 000.2 0 02 000 0.5 ”020300020 000 000.0 5.: 000.00 000.00 000.00 30.00 000.00 000.00. 000.00 000.00 000.2 000.00 02.00 z>o 02.282 -cowofigconu 0 20.0 o 0 000.00 0 o o 00.0. 000...: 0 30.00 000.2 .000 03 020282 -cowofigconu 000.0 00_ 000 000 000 000.0 9.0.: 000.0 30.00 000.00 000.00 000.0 0.00.0.0 000.00 z>o 02050022 0 00 0 E 00 o o 003 20.0 000.0 0 o o 0 08 02050022 08 000 a: 000 03 0S 0: 00 an 00 cm 00. 3 0n 2:0 08.8..— -05. .2502 mug—am— 09:53a— d :cWB— E 88.09.— 30.332 .08 enam— 35505 .3 003 553.53— .:dmfioZV can .005 .«c gamumm 55>Z .m.m 035,—. 178 Table B-6. Modeled Recreation Use Estimates and Confidence Intervals for National Forests in the Study Area Sampled in FY2001, FY2002, FY2003. Modeled Recreation CI Lower Use Upper Use 80% Estimate Estimate Region 2 Bighorn NF 529,242 24% 402,436 656,048 Black Hills NF 734,802 18% 601,289 868,316 Grand Mesa, U.,G. NFs 2,067,412 29% 1,464,761 2,670,063 Medicine Bow NF 562,138 14% 482,258 642,018 Nebraska NF 126,775 20% 101,433 152,117 Pike San Isabel NF 2,811,811 14% 2,417,876 3,205,746 Routt NF 1,494,708 8% 1,373,338 1,616,079 Shoshone NF 492,390 12% 432,762 552,019 White River NF 2,894,585 5% 2,759,697 3,029,473 Region 9 Allegheny NF 970,968 28% 700,165 1,241,771 Chequamegon-Nicolet NF 1,550,322 30% 1,082,590 2,018,054 Chippewa NF 1,319,701 14% 1,135,603 1,503,799 Hoosier NF 515,539 16% 434,754 596,324 Huron-Manistee NF 700,099 33% 466,476 933,722 Mark Twain NF 491,409 13% 427,575 555,243 Monongahela NF 694,512 14% 599,503 789,521 Ottawa NF 230,575 14% 197,925 263,224 Shawnee NF 509,560 1 1% 453,967 565,153 Wayne NF 387,176 15% 327,203 447,150 179 Table B-7. Local Use Percentages of Modeled Recreation Use for National Forests (NF) in the Study Area. Lower Percentage Upper Percentage Percentage Estimate Region 2 Bighorn NF 51.8% 49.3% 54.2% Black Hills NF 69.0% 66.2% 71.7% Grand Mesa, U.,G. NFs 67.2% 65.1% 69.2% Medicine Bow NF 51.4% 49.0% 53.8% Nebraska NF 44.7% 40.2% 49.2% Pike San Isabel NF 72.7% 70.9% 74.5% Routt NF 39.3% 36.5% 42.0% Shoshone NF 52.5% 49.5% 55.6% White River NF 44.2% 42.4% 46.0% Region 9 Allegheny NF 57.2% 54.5% 60.0% Chequamegon- Nicolet NF 38.4% 36.0% 40.8% Chippewa NF 55.1% 52.2% 58.0% Hoosier NF 70.9% 67.9% 73.9% Huron-Manistee NF 30.3% 27.7% 32.9% Mark Twain NP 84.0% 82.0% 86.0% Monongahela NF 37.8% 35.8% 39.8% Ottawa NF 34.2% 31.3% 37.1% Shawnee NF 62.1% 59.5% 64.6% Wayne NF 64.3% 61.2% 67.4% 180 Table B-8. Mid-distance Use Percentages and Primary Purpose Percentages of Modeled Recreation Use for National Forests (NF) in the Study Area. Lower Upper Primary Mean Percentage Percentage Purpose Percentage Estimate Estimate Percentagg_ Region 2 Bighorn NF 19.6% 17.6% 21.5% 96.3% Black Hills NF 3.8% 2.6% 4.9% 100.0% Grand Mesa, U.,G. NP 16.4% 14.8% 18.0% 92.1% Medicine Bow NF 34.9% 32.6% 37.2% 90.3% Nebraska NF 29.8% 25.7% 33.9% 86.5% Pike San Isabel NF 6.6% 5.6% 7.7% 96.2% Routt NF 28.9% 26.3% 31.4% 90.8% Shoshone NF 12.5% 10.5% 14.5% 87.1% White River NF 27.7% 26.1% 29.3% 87.6% Region 9 Allegheny NF 36.0% 33.4% 38.6% 97.3% Chequamegon-Nicolet NF 56.2% 53.7% 58.6% 95.7% Chippewa NF 34.1% 31.3% 36.9% 98.2% Hoosier NF 25.5% 22.6% 28.3% 96.4% Huron-Manistee NF 66.2% 63.5% 68.9% 98.6% Mark Twain NF 13.4% 11.6% 15.3% 83.4% Monongahela NF 51.7% 49.6% 53.8% 68.6% Ottawa NF 30.7% 27.9% 33.5% 46.4% Shawnee NF 27.5% 25.2% 29.9% 95.3% Wayne NF 34.7% 31.6% 37.8% 97.3% 181 APPENDIX C 182 Table C-l. Standard Errors of Annual Visit Frequency Estimates by Region and National Forest Aggeggtion Group.”I Distance Band Aggregation Group” 0 0.10 10-15 15.20 20-25 25-30 Region 2 1 12.2 5.4 2.3 2 6.3 3.7 5.7 3 15.1 1.2 1.6 0.8 1.1 0.9 4 7.7 4.6 3.4 2.4 2.9 Regional Avg. 5.3 1.3 1.5 1.0 1.2 0.9 Region 9 1 3.5 2.0 1.3 1.8 2.9 2.8 2 6.8 1.7 1.7 1.7 1.3 3 4.9 3.2 2.8 2.8 4 5.4 2.5 1.1 5 7.7 3.8 4.6 3.9 1.3 4.6 Regional Avg. 2.3 1.1 1.0 1.0 1.3 0.7 ‘ Estimated from the NVUM survey and weighted by VisExpwt. b Aggregation Groups defined below. Region 2: 1 Bighorn NF, Shoshone NF; 2 Black Hills NF, Nebraska NF; 3 Arapaho and Roosevelt National Forests, Medicine Bow NF, Pike and San Isabel National Forests; 4 Grand Mesa, Uncompahgre, and Gunnison National Forests, Rio Grande NF, Routt NF, San Juan NF, White River NF. Region 9: 1 Hoosier NF, Mark Twain NF, Shawnee NF; 2 Alleghany NF, Monongahela NF, Wayne NF; 3 Green Mountain and Finger Lakes National Forests, White Mountain NF; 4 Hiawatha NF, Huron-Manistee NF, Ottawa NF; 5 Chippewa NF, Chequamegon/Nicolet NF, Superior NF. c Number of cases is less than 30. Region-level average substituted. 183 SN; hm mm mm oowfim E fin o cod o cod mm m2 2:2 xoflm we ._ we 3 mm padm Roam av 3.0 o cod 8N m2 Eonwa we: on 0m 2 cowdm vwméfi ma aod o cod m2 m2 50.35 mg.— wm mm 2 Eafim $2.62 m 36 o cod o: m2 .555 we: on vm m— 08.3 Ema... o 86 o cod m2 m2 Eozwmm we: 3. cm mm cmodm mmaJm o cod o cod o3 m2 :5sz mo: mm M: E wtfim movfim Nv cod c cod m2 "2 5235 mo: 3 2 mm 3&2.“ mhvdm cm 86 o 86 9 fl m2 Eonwa we _ 4 mm. 3 Va 26.? wot; 09 d mo 3 mod 3 m2 Eonwa we; 9. m: 3 319. 2.19; 53 3.0 2 5.0 ow m2 82me we ._ av 2 cm gown 08.3 SN mad ca 8.0 no ”—2 E235 no: em 9 on wadm v23 o 86 o 8.0 ow "HZ Eonwa we 4 I” R an Ommém wm~.m h Ed D cod 9 m2 :5sz m3 4 an S on 3min and 2: S .o 3.2 no; ow m2 fienwmm we; mm a co 8m.mm 35.5 o 8.0 o cod mm m2 .555 r E a m e ..m D... . 0.. r ) e o o a h a d mm Mm mm mm WWW Wm m mm mm M Wm m. Wm N ..m w m B N o .m U U U U B m F .N :emwom mm «RED £23:— Eve—Z «menu .96.; EEVN 36%—=2 .N-U 039—. 184 185 SQN :V 00 0 «3.3. m0w.mm 50 00.0 3.0 mm ”2 .0 0-D can”: 280 SN; 0: v: 50 00¢.va 09.00 0 00.0 00.0 00m "2 £5 0.85 Sum; 03 000 #0 304m 000.50 0 00.0 00.0 mm- "HZ $:: fism Sum.— N0 mm mm 001nm 30.00— m 00.0 00.0 0: "2 £5 x85 Sum.— wa 00 3. www.mn 53.00 mm 00.0 00.0 03 ”Hz 3.5 fism Sum.— mfi mm 0m wmwfim «05.2. me. #00 00.0 03 m2 $:: 8.5 SN; 0: mm mm 50.3 «3.05 S. 00.0 00.0 mm_ ”2 .25.— x85 Sum; 00— mm 0m 000.2 03.00 N... 3.0 00.0 0: "HZ v.25 8.5 gm; no 0.0 mm mmvém manfiv 2 00.0 00.0 no m2 2:3 32.5 new; mm 0v Vm 000.mm 03.2 m— «0.0 _0.0 00 m2 £5 #85 SN; vm mm mm mmofim $12 000 00.0 00.0 00 m2 2:: 8.5 SN; 9. 00 mm M: 0.2.. $0.0 0 00.0 00.0 00 "HZ $:: 32.5 Sum; 0v 0N mm 03.9. 9 m.mm h .00 00.0 0..» "Hz 2.5 .35 SN; 9. 3. 0m $0.0m made 0 00.0 00.0 0v "HZ $:: seem r. .6 w e . 0.. .m 0.. d ...m ..3 Em M m .. . N . N. y . y. .. .m ( D .m P P D P D D o m .m w R c m B N o M .m 0 U ow U w B m B .3%—SUV .N-U visa. >00.N >00.N h00.~ >00.N Sod >00.N >00.N 500d >00.N Sod NV 00 00 mv 00. um 00 0v 9. _m 3. 00 00 mm 0— Z 2 2 0m mm a a 0— 0— 0— 030m 30.9 03.9. 05.0 001% 005.0 0 0N0.0v 3.0.0»V 000.0»q 30.0”V 000.03 N010?” v2.03 05.3.04 02.400.— 000.03 $40.0: Nwwdm N0m.Nm www.mv N— : ms. VN N0 #0 of 0N 00.0 5.0 00.0 N00 00.0 3.0 8.0 5.0 85 m2 m2 .0 0.: 0392 2.5 86 9: m2 .0 0.: ram”: 230 86 we .2 .o too '30: 250 So o: ...z .o 0.: ram”: 220 86 3 m2 .0 to: 9302 280 8.o ow uz .e 0.: 0%”: Eco w; no .2 .o 00.0 00 m2 .0 00.0 00. m2 0 Number of Units NF Acres (Th.) RUC Code DistNPS Dist. BLM Nearest OthNF Med. Income Pop. OVN Use/Th. Pop. OVN Use/Cap. Day Use/Th. Pop. y... .m mm mm am :5:er .N..U asap. 186 000.0 N0 mm mm 000.0»V wwN._N_ 0 ~00 00.0 03 m2 30m 2.282 000; _0 00 MN 0— 0.00 00— .0; E _0.0 00.0 03 m2 30m 8.232 000.0 _ 0 0m 0m N— 1N0 50.0— 0; N. ~00 00.0 0_ _ "HZ 30m 05232 000; N0 mm mm 0100 NNBJ 00 3 N00 00.0 00 ”HZ 30m 05282 000.0 00 cm 0— 000.00 000.5% N0 00.0 00.0 0w "HZ 30m 05282 000.0 00 00 NN 000.00V 0—0Kmm 00 00.0 00.0 00 "*2 30m 05282 000.0 hm mm 0— 000.00 00 0.0% 0 00.0 00.0 00 m2 30m 2.582 000.0 0m 00 S 50.3‘ 000.000 00 00.0 00.0 01 "HZ 30m 05282 000; 0N ow N_ 000.3» 20.02 00 00.0 00.0 00 m2 225 E282 0004 0v mm 0_ 00nd»4 N0m.m0 #5. 34.0 00.0 0v mZ 30m 05282 000.0 Ow 0— #0 05.0.54 :0.0_ 0000 M04 :4 00 m2 30m 05282 000.N 00 0m 3 0mw.0v Nag—mm v 00.0 00.0 00m ”—2 .0 to: 9am”: Eco 000.N 00 am 2 000.00 20000 0 00.0 00.0 000 .02 .0 Dip 0am”: 0:30 h00.N we mm mm m— 0.00 §.00N 0 00.0 0 00.0 02 m2 .0 Rob ram”: 2.20 r m m e n... u m. m .m .m c ) C e m M W W m . . N v. . V. d .m .M m m .m n. W P u a M P a «u m o m U A a U L o 0 V o .0 D D .m P P D P D D o N .m m a c m n N 0 mm o u ow n W m m F 3.2800 .N0 0.000. 187‘ 000.. 00 000 000 00. .00 000.00. 0. 00.0 00.0 00. ".2 8.3502 000.. 0.. 00. 00. 000.00 000.00. 0. 00.0 00.0 00. 2 8.3502 000.. ... 00. 00. 00..00 000.000 . . .00 00.0 00. ".2 8.3502 000.. 00. 0.0 0.0 0.0.00 000.00. 0. .00 00.0 0.. ”.2 5.3502 000.. 00 00. 00. 000.00 000.00 00 000 00.0 00 ".2 $8502 000.. 0.. 00. .0. ..0.00 000.00 00 00.0 00.0 00 "H2 8.3502 000.. 00. 00. 0. . 0.00 000.00 00 00.0 00.0 00 ”.2 8.3502 000.. 00. 00. 00. 000.00 000..0 0 .00 00.0 00 ".2 5.3502 000.. .0 00 .0. 000..0 000.00 00. 0.0 00.0 00 ".2 8.3502 000.. 00 00. 00. 000.00 000. .0 0 00.0 00.0 00 "02 53502 000.. 00 00 00 0.0.00 000.00 0 00.0 00.0 00 ”.2 8.3502 000.. 00 00 00 000.00 000.00. 0. 00.0 00.0 000 ".2 Ram BEBE 000.. .0 00 .0 000.00 00..00. .. .00 00.0 00. m2 30m. 2.6.82 000.. 00 00 0. 000.00 000.000 0 00.0 00.0 00. 22 30m. 2.6.32 000.. 00 00 0. 000.00 000.000 .0 00.0 00.0 00. ”.2 30m. 2:282 r m m e .m 0.. .m 0.. m m m c ) c e m M w. m . m 9.. N n.. a v. m. y a .a m U Am U0 mL m... .00 o VMo W aMo aC mu m... m... m R C m B D N o M m 0 U ow U w B m m F €0.80. .00 0.5a... 188 m 0N. .. N .4 mm 8 one? 8...... ... mm 8... o 8... m... ...z .33. m 0N... N om ..N N. M38 3.08. NVN mN... 8. o. .o on 02 98... m N... N ..N ..N o. ..NnNn 83% 8. 2... o 8... a. ..z .53. m 0N... m mm NN N. 83.... 80...... NaN ..N... c 8... 8 ..z .33. m 0N... m 8 v. o 8N; m SNNM mo 8... Nm 8... mm ..2 9.5.. m NNN.N .. a. a. w. .3... 398m 2. 8... o o N. ..z .33.. 5m 9.... NNN.N ..N m. ..N .8.Nm 08...... o. 8... o o o: 02 .38. 5% 9.... NNN.N on mm aN Nam... 02.8. N... 0.... o o 8 ..z .88. gm 9.... NNN.N mm 9. .N 0%.? NNn.mNm w 5... m S... 8 ..z .3... 5% 9.... NNN.N 3 NM N. 2an 9.3: o o S. 0.... S ".2 .89. 5m 9.... NNN.N mm NN N. NRA... 80.8 o o N. No... on ”.2 .33.. 5m 9.... NNN.N a. 9. ..N ..Nmfi .20? 08 m... o o a. ..z .39.. 5% 9.... NNN.N N... N a. $4.3 30.8 mm 3... .. a... o... ..z .3... Em 9.... NNN.N .m N N. 8.8 838 w... a... a. 3... mm .2 .3... 5m 9.... 3.... No. mNN R. 9.3.. 3.0% N S... o 8... SN .2 8.3502 3.... a. m... 0.. 8...? 3...: a .o... o 8... m... ..z 9.3502 3.... ..m ..N. NM. 8...? N233 .. 8... o 8... o: ".2 8.3502 3.... a. NNN 8N wN..Nm 3.3... N. 8... o 8... mm. ..2 9.35.2 r. 8 m e .m 0.. h 0.. M m m c ). c e m M mm a m m. N m. N a y n... y a m U Ah U0 m... m m m o VMo VC ..Mo aC mm m P P D P D o m... “a .2. m B D... M... on ow u w D m... 6.2800 .00 0.50... 189 hmvN Vw MN . . NNm.wm 00W6MN 0 00.0 00.0 mm. ”.7. 80:09.0 hmvN mm . . N. .m..0m NVMfM. 0 00.0 00.0 0.... "H7. 2.9.8.5 hmvN . . . N. a. 3...?” 0.5.x... 0.. 50.0 00.0 mN. W7. 2.9.8.5 hmvN mo N. .N 0.1...0 .m0.v.. 00 50.0 00.0 0. . ”.7. ozonmonm hmvN mm .. .N w0..wm EQWa m. .0.0 00.0 m0 m7. 25:00.5 hmvN 0% w. MN wimom 03.02 0 00.0 .0.0 00 ”.7. 2.2.02.0 hmvN 9. ON wN .ow.0m vwoNv. mm 00.0 m0.0 we ”.7. 2.9.8.5 hmvN mm mm .N 2de wmmf wwm and 00.0 0m ".7. ecosmosm hmvN m? N. mN oma..v 0mm.m. hm 00.0 00.0 mv m7. 2.2.095 hMYN mm m. .N 8N6v mwodN 0m m0.0 . 0.0 0v. ”.7. 2.9.02.0 hmvN MN. 0N wN 0w..wm 000... 0 00.0 m. .0 mm ".7. ozonmonm m ON... 5 No hm 0m mooKm .meN. 0 00.0 0 00.0 00N ".7. 9.0% m 0N... N. we mm NM N.V.VM VNW..w 0 00.0 0 00.0 mw. "HZ anew. m ON... 0 mm mm mm 05...; mohmw 0 00.0 0 00.0 0h. "H7. .501 m 0N... v R. an .m cmo.Vm vm..m0N o .0.0 0 00.0 mm. "HZ fined m 0N... m mm VN 0m n. wdm .de. v. .0.0 0 00.0 0... "H7. .50“ m 0N... N mm b. v. .w..mv .0N.mcv m. .0.0 0 00.0 mN. ”.7. 5.5”. m 0N... v mm mm m. mmflfin .VNfimN 3 mod 0 00.0 0.. m2 6:3. m 0N... N om Nm a. N. Ndm ..Nvch m. . . . .0 0 00.0 ma "HZ 3.5% m 0N... . NW 3. ON 0.04% .aodNN. 00 w0.0 mm 3.0 0w ".7. 630% r N. m S e . m. h D.. I. mm c) Ce m M an .m r... Nb... Na v. n.. ya d m... U A“ UM ML uh .mm o VMO VC 8M0 aC Mm m... m R D e t P P D P D D o N M W C m B N 0 M ..m 0 U 0 W U W B M F €3.50. .N.U 035% 190 chad mm mm on 5min h. m.m.. o oo... o cod mm. ”.7. .03... 2...? cnmfl on cm on cmmdm 02.9.. .. ...d o co... 9.. L7. .25. 2...? chad mm vm mm mad.“ wmmdcm me 3... o co... mm. ”.2 .o>.~. 2...? 03d mm 3 on wOmdm nmmdv. vm. m. .o o oo... o.. ".7. .03”. 2...? 08d .m cm .0. 36.2. ..madow cm 3.0 m. 8... mo ”.2 .95. 2...? 0N~.N .m cm v. 35...... nmofiow cm 3... m. 8... ma ”.2 .95. 2...? chad 0.. am om vwmdm 8.98%. .m. m . .0 ea c. .o ow ".7. .03”. 2...? chad om mm c. a. ..mm mwmdwm. om. m. .o mm ac... no ...7. .o>.~. 2...? chad on m. N. 8&3 watmm .5... .N... .v 3... om ".7. .o>..~. 2...? cNN.N mm mm c. 89.9 $5.? 80. .0.. o co... 9. ...Z .95. 2...? chum ..m N. a 2%.: 8th .N no... 0 .od ow ...Z .39. 2...? Eda mm .. m. 8...: $5.? av mod 2. 2... mm ”.2 .25. 2...? SYN mo. vm .. N013 w.m.voo v 8... o co... com ".7. 2.9.85 5min no mm v. 8de www.mon a. No... o 8... mm. ".2 2.9.0....m Rim am 3 v. ENS. $5.3. .. .0... o 8... o... ".2 29.8% r. E m e .m 0.. .m 0.. m .m .m c J C e m t M L m .M . m 0.. m. N a v. n.. v. a d m U A“ UM EL .mah .mo 0 WM0 VC 3M0 aC Mm m... m ( D D e t c P P D P D D o N .m M R C m B N O M .m 0 U 0 W U W B m F 05.80. .N-U o1...... 191 .02.... n 2...: 02.2.... ..< .207. N chN.N v 0.. VN .N NNN...q 05.69» 0 00.0 0 00.0 00N hwznuwzm N ebN.N 0 mm 0.. m... wbmdm 35.00. 0 00.0 0 00.0 mm. KZMomfl N chN.N m 0.. wN ..v 000.9.” 0. ..0m. 0 00.0 0 00.0 0... ".7. ~me 2...? m. m) c. m ..m . mm. m y... y... .2 . mm mm. mm . mm mm. m... m mum. mm mm... mm mm mm 05:00. N... 2...... 192 _ 8. m 8 o. .3... .0.... .m 88.. o o a. ".2 .3895 _ 8. m 8. n. Sn... .8... m. on. m .2 8 ".2 .3255 _ 8. o 8 e. 2.... an... m. 8. 3 ..m m. ...z .3395 v ..3 m e. E. 8...: 88.... o .8 o o 8. ".2 3.8.2.82886 v ..3 v v. 9.. 8.... 8...... m .9... v .2... 3. m2§82é8§§85 . ..2 v o. 8. 5.8 8...... m. 3.: o o o: ".2 Eogzémoegaofi v ..2 . 8 8. 8.8 2.8.... a. 8...... o o .2 m2 3.8.2-8.3386 . ..3 . E x: «.2. 8...... o. .8... m. .8... .2 m2 3622-886385 e ..2 . .. .3 8.... 51.8.. m. .3... o o 8. ”.2 3322-885386 v ..2 m 8 E .8... 8...... me .3... o o o: ”.z §8_z-=o.oe§§u . ..2 v E 5 5.... 83.8 m. 2.... o c 8 ".2 58.28.08.385 . ..2 v 8 o: 8...... 9.53. m. . .8... 2 :3. o. ”.2 .28_z-=o.oa§§o v ..n. v 8 8 8m... 2...... .3 8.8 at 8...: _ we ".2 5322-888385 v ..3 m 8 .2 .8... .36.: o... .8... o o 8 ”.2 530.28.053.06 v ..2 m 8 o: .8... .8... m. 8.... 8. .8... a. .2 6.8.2-888286 w ..3 v 8. .2 8.... 3...: mm. 35.8 B as... 8 ”.2 3_8_zeo.o§=§o v ..2 m 8 m. 8.... 8...... . .m .2... m. a... m. .2 8.828.388... . .E . 8 8 :3. «8.8... w 81. o o 8. .2 £2.22 . an N v. 8. 5.... .88... o .2 o .. a. ”.2 22.22 . ..m m 8 8. .8... 89.8.. o a... o o o: ".2 €2.22 . ..m . .. m. .8... «3...... o 8. o 3 m2 ".2 .5822 . .3 m c. 8 .8... 88.8... o R. o R .3 ”2.5822 . 2. . mm 8 8.... 2...... w m... o o n... ".2 €2.22 _ 2. m .. .. .8... 82.8... . 8.... o o .2 ”2.53.2? . 2. . .2 .. 81:. 2...... m R... . .2; 8 ".z 22.22 d 2. . .: 8 81:. 818... 5 3.... . 2... o. ".2 €2.22 . .3 . m2 .2 8m... 2...... .m E... m .m... we ".2 22.22 . .3 w m: 8.. a... .8... mm. .8... o o 8 ".2 .58»? . 2 m m .2 .2 8.... 8...... m. 2... o o 9. ".2 32.22 _ 2 m m 8. E 8.... $5: 9: 35... o o 8 ...z 32.22 _ 2. w 5 .2 8o... ......_ 8. 8.... o o m. ".2 22.22 r m m . e .m a .m u m .M .m c \l C e m M m . . ..U V. . U .0 m u A u w . m w w w a w y m m m 0 MM mm RC m NM MID. P. 0MP. m DMD- m DB MF .8 5%.. m... ..an .35... .252 .80 .38... ...ch 88.232 .90 93.... 193 . ..8 . .2 ... .8... ....2. .. 88.. o . .. ...z 8......283: . .8 v .2 ... 8...... 8...... S 3.... . .... 2. ".2 8252-85: . ..8 m 3. 2.. ...... 8...... m. 8.... .. .... .. ”.2 8...:«2.83.. . 8. . o. 2.. 8.... ......... o .8. o 8. 8. ".2 .28... . 8. . .. ... 8...... 8.8.... o .... o ... ... ”.2 .28... . 8. m .2 .2 3.... ....88. . .8. o . ... ".2 .28... . 8. .. ... 8. ....2. ......2... o ... o o ... ”.2 .28... . 8. . ... .. ...... ....83. o ... o o 2.. ".z .28... . 8. . ... .. .8... 2.8... o o o o ... ...z .28... . 8. . ... .. ...... .88... . 8... o o 2. ".2 .28... . 8. . .... 3. 8.... 3...... o ... o o 8 ..z 3.8... . 8. . .... 2. 2...... 8...... m 88.. o .... o. ".2 .28... . 8. . ... .... ...... 8...... . ...... .. ...... .. ".2 .28... . 8. . ... ... ....2. 2...... o. .3... . 2.... o. ”.2 .28... . 8. . 3. ... ...... ....o.. o ... ... ...... .. ".z .28... . 8. . 8 ... .8... ...... .m .....8. o . o 2. ..z .28... . 8. . 8 ... 3o... 3...... .v ...... a. ...... .. ..z ......8: . ... . .. ... ....2. .8... . .8. . ... 8. ".z A.32....6 . ... m .. .2 .8... .8... .. ...... . ... ... ".2 .3256 . ... m 8. .... .8... ...... o. .8... o o ... ".z .3256 . ... . 8. .... 3.... 3...... m. ....8 m 2... ... ".2 «325.6 . ... . .. ... 3.... 8...... ... ...... o o 2.. ".2 d.322...6 . ... . .. 8. 8.... 8...... 8 ...... o o ... .2 «32...... . 8. . ... o. 8.... 3...... .. ...... o o 2. ..z .3256 . ... .. .. .. 3.... ...... ... .88 o o 8 ...z «3256 . ... .. ... .. 8.8. 8...... ... 3...... o o o. ...2 «325.6 . 8. m .2 .. 28... 8...... 8. .3... o o .. ...z .3256 . .8 v 8 .. ...... 3o... 2.... ...... .... ...... o. ”.z .3286 . .8 m 8 .. .3... ...... w. 88.. o o ... ”.2 «328.6 . ... m .2 .. ...... 82.... a. ... . ... 2. ".z ‘.3286 . ... m 8 .. ...... ...... .. .... ... ... .. ".2 «3256 r. ...u m ..m w m M m m .. ) c .. m . m m . . u y . u m U A U .m m .... m w. W w. N w. v. mm m o £5.80. .n.D 035. 194 . .8 . .. .. ...... ...... 8 ...... . .... .. .2 2.5.852 . .8 . .. .. ...... 8.8. o. .... 8 ...... 8 .2 2.5.8.2 . .8 . .. .. ...... ...... 8 8... 8. ...... .. ”.2 2.5.882 . 8... . ... ... ...... 8...... o ... o 8 8. .z .53... .52 . 8... . ... ... ...... .. ...... . .... o . ... ...z .53. .52 . 8... . ... .2 ...... 8...... o ... o . ... ...z .53... .52 . 8.. . ... ... ...... .8..... F .8. o . ... .2 .53... .52 . 8... N ... ... ...... .... .... o 3. . ...... 8. .z .53.. .52 . 8... . ... .. .. .... 8...... . ...... o ... ... .z .53.. .52 . 8.. . ... .. 8.... ...... . 8... o o 2. .z .53. .52 . 8... . ... .. ...... ...... . ... . .. . .. .z .53... .52 . 8... . ... .. ...... .....8. o ... . .... o. .z .53... .52 . 8... . ... .. ...... ....8... . .... o 8. .. .z .53. .52 . 8... N .2 .. ...... .8... o ... . .... o. .z ..«3. .52 . 8.. . 8 .. ...... ...... .F .... _ ... .. .z .53.. .52 . 8.. . .. .. ...... ...... . ... t .... 8 .z .53... .52 . 8... . .. .. ...... ...... . .... o o .. .z .53.. .52 . ... . .. ... ...... ...... . o .. o .. 8. .2 8252-85: . ... . .. ... ...... ...... . ... o ... ... .2 8252.25. . ... . .2 ... ...... ...... . ...... o . ... .2 8252.2... . .8 F .2 ... ...... .3...... N ...... o . ... .2 8.252-852 . ... . ... ... ...... .88... N. ...... F .... 8. .2 8.2.2.8... . .8 . .2 ... ...... 8...... N. ...... o o ... ”.7. 8252.2... . ... . ... ... ...... ...... . 8 8.8 . .... 2. .2 8252.2... . ... . ... .... 8.... ...... ... ...... o o 8 .2 8252.2... . ... N ... ... ...... ...... 8 ...... N .8. o. ".2 8252-52... . ... . ... ... 3.... ...... 8 ...... .. .... . .. .z 82.52-85: . ... . 8 ... ...... ....8. .8 .. ..2 8. ...... .. .2 82.52825. . .8 m .2 ... ...... ....2. .. .8. o o .. .2 8252.25. . .8 . .2 ... ...... 8.... F F 3... . ... 8 .2 8252-852 . ... . 3. 8. ...... 8...... .. ...... w. .... .. .2 8252.85. ... m m F . e .. u. h. u. m .m .m c \l C e m m N m . . U V. . U w. u A u . m . .. w w a .. mm o .. m... mm RC m DNM mm P 0MP m 0MP m. on MR 8.280. .80 0...... 195 . ... N ... .. ...... 8...... N .... m 8... o. .2 8.3... . ... m .2 .. ...... 8...... 8 ...... m ...... .. .2 8.3... . ... m .2 .. ...... ...... m. .... . .... o. .2 8.3... . ... m 8 .. ... ... ...... a. 2... 8 .... .. .2 8:3... . ... . 8. .. ...... ...... m. .... .. ...... 8 .2 8.3... . ... m .. 8 8.... ...... .N 8... .. ...... .. .2 8.3... . 8. m .. ... ...... 8...... . .... o o 8. .2 .32... . 8. N .. ... ...... 8...... . .... o o ... .2 .326 . 8. N .. .2 ...... ......o. . .... o . .... .2 .32... . 8. m .. .. ...... ...... m .. ... o .. ... .2 .326 . 8. . .. .. ...... 8.... o o o o 8. .2 .325 . 8. m .2 .. ...... ...... o. .8. m .... ... .2 .32... . 8. . 8 .. ...... ...... m. .... o o 2. .2 .32... . 8. . ... .. .88 ...... .... .... N ... 8 .2 .326 . 8. m 8 .. ...... 8...... m ... . ... o. .2 .32... . 8. . 8 .. ...... ...... o. ... o o .. .2 .326 . 8. . .. .. ...... ...... o o o o o. .2 .32... . 8. . .. .. ...... ...... m. .2. cm. .... .. .2 .32... . 8. m .. .. ...... ...... o o o o 8 .2 .325 . 8. m .. .. ...... 8.8 o o .. .... .. .2 .32... . .8 N .2 8 ...... 8...... o o o o 8. .2 2.5.8.52 . .8 m .2 .. ...... 8...... . .... c o ... .2 2.5.882 . .8 m .2 .. ...... 8.8... F 8... o o 8. .2 2.5.8.52 . .8 N 8. .. ...... 8...... m ....2 o o ... .2 2.5.882 . .8 N .. .. ...... 2...... m ...... ... 8... 8. .2 2.5.8.52 . .8 u .. .. ...... ......... m 2.... o o ... .2 2.5.882 . .8 m .. .. ...... 2...... N ...... o o 2. .2 2.5.882 . .8 m .. .. o..... .....8.. 8 ...... . .... 8 .2 2.58282 . .8 N .. .. 8.8 .88... 8 ...... o ... o. .2 2.5888. . .8 m o. .. ...... ....8... mm ...... m 8... .. .2 2.5.882 . .8 m .. .. ...... ...... .. ...... o. .... ... .2 2.5.882 r ......) c. m .. .. .. . .. . ...... n. .. M. m U A m m m m a .m M m no. 0 w W m m no. y D m ...... o N .m w D N o m U 0 U m m F €2.80. .m.0 .35. 196 . ... . .2 .. ...... .85.... o .... o .... 8. .2 ...... . ... . .. .. ...... 8...... o ... o ... ... .28.... . ... m .. .. ....8 .28... _ .... o o 2.. .2 ......» . ... . .. S ...... 5...... . .... o . ... .2 ...... . ... . .. .. ...... 8...... . ..... o o 8. .2 ...... . ... . .. .2 ...... ......... . ..... o . ... .28.... . ... N .2 .. ...... 8...... . ...... o o 2. .28.... . ... N .2 .. ...... ......... . .... . ....2 8 .28.... . ... . .. .. ...... 3...... . ...... . .... o. .2 ...... . ... . .. o. 8... .85.... . ...... . .... .. .28.... . ... m .. 8 ...... ...... o o t .8. o. .2 ......» . ... . .2 8. ...... ...... F ...... .. ...... .. .28.... . ... m .2 .2 8.... ...... . .... o o 8 .2 ...... . ... N .2 .2 ...... ...... F ... .. .... .. .2 ...... . ... . ... .. ...... 8...... N 8... o o 8. .28.3... . ... . ... .. ...... ....8... o ... o . ... .2 8.3... . ... . ... .. ...... 8...... . 8... o o 8. .2 8.3... . ... . ... .. ...... 2...... . ...... o . ... .28.3... . ... N .2 .. ...... ......... N 8... . ...... 8. .28.3... . ... m 8 .. ...... 8...... . ...... . ...... ... .2 8.3... . ... . 8 .. .2... ...... . .... o o 2. .28.3... . ... . .2 .. ...... ...... F .... . .... 8 .2 8.3... r m w t e . u . u w um um Ce m mm ...m .. mu... U y... U ma .... U A T U .m . a m o o a o v. m m m mom W( RC W DM Min. P DMD. m DMP m DB MF 3...... ..-. ...... 197 LITERATURE CITED Alig, J.T. and RR. Voss. 1995. National Forests of Wisconsin Demographic and Recreation Participation. USDA FS Research Paper FPL—RP-542 Alward, G.A., J .R. Arnold, D.B.K. English, and D.W. McCollum. 1998. Developing expenditure profiles for Forest Service recreation visitors. Unpublished draft of a general technical report Beale, C. 2004. 2003 Rural-urban continuum codes. Available on-line: http://www.ers.usda.gov/briefing/rurality/RuralUrbCon/ Betz, C.J., J .C. Bergstrom, M.J. Bowker. 2003. A contingent trip model for estimating rail-trail demand. Journal of Environmental Planning and Management 46(1): 79 — 96 Bojanic, DC. and RB. Wamick. 1995. Regional trade market analysis: resort marketing approaches. IN Proceedings of the 1994 Northeastern Recreation Research Symposium. April 10 — 12, 1994 Saratoga Springs, NY. USDA FS GTR-NE-l98 Bowker, M.J, D.B.K. English, and J .A. Donovan. 1996. Toward a value for guided rafting on southern rivers. Journal of Agricultural and Applied Economics 28(2): 423 — 432 Bowker, J.M., D.B.K. English, and H.K. Cordell. 1999. Projections of outdoor recreation participation to 2050. IN Outdoor recreation in American life: a national assessment of demand and supply trends. H.K. Cordell, Principal Investigator. Sagamore Publishing, Champaign, IL Boxall, P.C., B.L. McFarlane, and M. Gartrell. 1996. An aggregate travel cost approach to valuing forest recreation at managed sites. The Forestry Chronicle 72(6): 615 - 621 Brainard, J., I. Bateman, and A. Lovett. 2001. Modelling demand for recreation in English woodlands. Forestry 74(5): 423 — 438 Bristow, R.S., S.P. Caron, and K.H. Green. Investigating spatial structure influence on recreation activity packages. [N Proceedings of the 1993 Northeastern Recreation Research Symposium, April 18 - 20, 1993, Saratoga Springs, NY. USDA FS GTR-NE-185 Bureau of Land Management. 1999. Public Land Statistics, 1998. Available on-line: http://www.blm.gov/natacq/pls98/ 198 Burt, OR. and D. Brewer. 1971. Estimation of net social benefits from outdoor recreation. Econometrica 39(5): 813 — 827 Cesario, E]. and J .L. Knetsch. 1976. A recreation site demand and benefit estimation model. Regional Studies 10: 97 - 104 Cho, K-I, I.Lee, and T. Var. 2001. Application of travel cost model to measure economic value of recreation and tourism resources. Tourism Analysis 6: 17 — 27 Cicchetti, C]. 1973. Forecasting recreation in the United States: an economic review of methods and applications to plan for environmental resources. Lexington Books, Lexington, KY Cicchetti, C.J., A.C. Fisher, and V.K. Smith. 1976. An econometric evaluation of a generalized consumer surplus measure: the Mineral King controversy. Econometrica 44(6): 1259 - 1276. Clawson, M. and J .L. Knetsch. 1967. Economics of outdoor recreation. John Hopkins Press Baltimore, Maryland Clawson, M., RR Held, and CH. Stoddard. 1960. Land for the Future. John Hopkins Press, Baltimore, MD Cordell, H.K. 2003. Outdoor recreation: segmenting americans and markets to better see the trees. Available on-line: http://www.srs.fs.usda.gov/trends/OIAAZ.pdf Dana, ST. 1957. Problem analysis: research in forest recreation. USDA Forest Service publication. Washington, DC. Douglass, R.W. 1999. History of outdoor recreation and nature-based tourism in the United States. IN Outdoor recreation in American life: a national assessment of demand and supply trends. H.K. Cordell, Principal Investigator. Sagamore Publishing, Champaign, IL Driver, BL. and P.J. Brown. 1975. A social-psychological definition of recreation demand, with implications for recreation resource planning. IN Assessing Demand for Outdoor Recreation. National Academy of Sciences. US. Government Printing Office. Washington, DC. Dwyer, J .F., J.R. Kelly, and MD. Bowes. 1977. Improved procedures for valuation of the contribution of recreation to national economic development. Final Report to the Office of Water Research and Technology, US. Department of Interior. English, D.B.K. National Visitor Use Monitoring Project Manager. USDA FS Washington Office. 199 English, D.B.K., and J .M. Bowker. 1996. Sensitivity of whitewater rafting consumer surplus to pecuniary travel cost specifications. Journal of Environmental Management 47: 79 - 91 English, D.B.K. and A. Home. 1996. Estimating recreation visitation response to forest management alternatives in the Columbia river basin. Journal of Applied Recreation Research 21(4):313 - 334 English, D.B.K., S.M. Kocis, S.J. Zamoch, and J .R. Arnold. 2002. Forest Service national visitor use monitoring process: research method documentation. USDA FS General Technical Report SRS-57 14 p. Etzel, MJ. and AG. Woodside. 1983. Segmenting vacation markets: the case of the distant and near-home travelers. Journal of Travel Research Spring: 10 - 14 Faunce, R.F., A.S. Kezis, and GK. White. 1979. Characteristics of Maine’s resident and non-residents hunters. Bulletin 760, Life Sciences and Agriculture Experiment Station, University of Maine. Fix, P. and IE. Loomis. 1997. The economic benefits of mountain biking at one of its meccas: an application of the travel cost method to mountain biking in Moab, Utah. Journal of Leisure Research 29(3): 342 - 352 Grant, W.E., E.K. Pedersen, and S.L. Marin. 1997. Ecology and natural resource management: systems analysis and simulation. John Wiley & Sons. New York, NY Haefner, J .W. 1997. Modeling biological systems: principles and applications. Chapman and Hall. New York, NY Hanink, D.M. and M. Stutts, 2002. Spatial demand for National Battlefield Parks. Annals of tourism research 29(3): 707 — 719 Heckman, J. 1979. Sample selection bias as a specification error. Econometrica 47: 153 - 162 Hellerstein, D.M. 1991. Using count data models in travel cost analysis with aggregate data. American agricultural economics association 73(3): 860 - 866 Hendee, J .C. 1967. Recreation clientele-the attributes of recreationists prefen'ing different management agencies, car campgrounds or wilderness in the pacific northwest. Doctoral Dissertation. University of Washington. Hotelling, H. 1949. The economics of public recreation. IN The Prewitt Report. National Park Service. 200 Howell, DC. 1997. Statistical methods for psychology. Wadsworth Publishing Company, Belmont, CA Interagency Survey Consortium. 1995. National Survey on Recreation and the Environment 1994-1995. USDA FS Recreation, Wilderness, and Demographics Trends Research Group, Coordinator. Available on-line: http://www.srs.fs.usda.gov/trends/Nsre/nsre2.html Interagency Survey Consortium. 2002. National Survey on Recreation and the Environment 2000-2002. USDA FS Recreation, Wilderness, and Demographics Trends Research Group and Human Dimensions Research Laboratory University of Tennessee, Coordinators. Available on-line: http://www.srs.fs.usda.gov/trends/Nsre/nsrerep.html Jenness, J. 2004. Nearest features (nearfeat.avx) extension for ArcView 3.x, version 3.8a. Jenness Enterprises. Available on-line: http://www.jennessent.com/arcview/nearest_features.htm Loomis, J .B. 2002. Quantifying recreation use values from removing dams and restoring free-flowing rivers: a contingent behavior travel cost demand model for the Lower Snake River. Water Resources Research 38(6): 2-1 — 2-7 Loomis, J .B. and R.G. Walsh. 1997. Recreation Economic Decisions: comparing benefits and costs, 2“‘1 Edition. Venture Publishing, State College, PA Loomis, J ., A. Gonzalez-Caban, and J. Englin. 2001. Testing for differential effects of forest fires on hiking and mountain biking demand and benefits. Journal of Agriculture and Resource Economics 26(2):508 - 522. Loomis, J., B. Roach, F. Ward, and R. Ready. 1995. Testing transferability of recreation demand models across regions: a study of Corps of Engineers reservoirs. Water Resources Research 31(3): 721 — 730 Lovett, A.A., J.S. Brainard, and U. Bateman. 1997. Improving benefit transfer demand functions: a GIS approach. Journal of environmental management 51: 373 - 389 Moeller, G.H. and HE. Echelberger. 1974. Approaches to forecasting recreation consumption. IN Outdoor Recreation Research: Applying the Results. North Central Forest Experiment Station GTR-NC-9 National Academy of Sciences. 1975. Assessing demand for outdoor recreation. US. Government Printing Office, Washington DC. National Atlas of the United States. 2005. Federal Lands of the United States. Available on-line: http://www.nationalatlas.gov/atlasftp.html?openChapters=chpbound% 201 23chpbound National Park Service. 1986. 1982-1983 Nationwide Recreation Survey. US. Government Printing Office, Washington DC. Available on-line: http://www.srs.fs.usda.gov/trends/Nsre/nsre8283.htrnl National Park Service. 2002. National Park Service—park boundaries. Available on-line: http://www.nps.gov/gis/data_info/ National Park Service, Public Use Statistics Office. 2004. Statistical Abstract 2004. Available on-line: http://www2.nature.nps.gov/stats/abst2004.pdf Nelson, CM. and J.A. Lynch. 1995. Estimating dispersed recreational use on Michigan’s state and national forests. IN Proceedings of the 1992 Northeast Recreation Research Symposium. April 10 — 12, Saratoga Springs, NY. USDA FS GTR-NE- 176 Outdoor Recreation Resources Review Commission. 1962. Outdoor Recreation for America. US. Government Printing Office. Washington DC. Propst, D.B. (compiler) 1985. Assessing the economic impacts of recreation and tourism. Proceedings of the conference and workshop. Michigan State University, East Lansing, Michigan. May 14 - 16 1984 Provencher, B. and RC. Bishop. 1997. An estimable dynamic model of recreation behavior with an application to Great Lakes angling. Journal of Environmental Economics and Management 33: 107 - 127 Reid, RD. 1989. Hospitality marketing management, 2“‘1 Edition. Van Nostrand Reinhold, New York, NY Reiling, S.D., C. Trott, H.T. Cheng. 1993. Users’ preferences for selected campground attributes. IN Proceedings of the 1992 Northeast Recreation Research Symposium. April 10 — 12, Saratoga Springs, NY. USDA FS GTR-NE-176 Romano, D., R. Scarpa, F. Spalatro, and L. Vigano. 2000. Modelling determinants of participation, number of trips and site choice for outdoor recreation in protected areas. Journal of Agricultural Economics 51(2): 224 - 238 Smith, V.K. 1975. The estimation and use of models of the demand for outdoor recreation. IN Assessing Demand for Outdoor Recreation. National Academy of Sciences. US. Government Printing Office. Washington, DC. Strauss, C.H., T.W. Moran, G.L. Storm, and RH. Yahner. 1993. Economics and demographics of northeastern sport hunters. IN Proceedings of the 1992 Northeast 202 Recreation Research Symposium. April 10 — 12, Saratoga Springs, NY. USDA FS GTR-NE-l76 Stynes, DJ. 1999. Guidelines for measuring visitor spending. Available on-line: httpzllwww.prr.msu.edu/mgm2/econ/ Stynes, DJ. and EM. White. 2002. Spending profiles of national forest visitors year 1 forests. Available on-line: httpzllwww.prr.msu.edu/stynes/nvum/NVUMZOOOReport.doc Stynes, DJ. and EM. White. 2003. Spending profiles of national forest visitors. Available on-line: httpzllwww.prr.msu.edu/stynes/nvum/SpendingProfilesShort.pdf Stynes, DJ. and EM. White. 2004. Spending profiles of national forest visitors, 2002 update. Available on-line: httpzllwww.prr.msu.edu/stynes/nvum/NV3YearFinal.pdf Stynes, DJ. and EM. White. 2005a. Spending profiles of national forest visitors, NVUM four year report. Available on-line: httpzllwww.prr.msu.edu/stynes/nvum/NV4Year.pdf Stynes, DJ and EM. White. 2005b. Effects of changes in the FY 2003 NVUM instrument and development of national forest visitor spending profiles for lodging-based segments. Available on-line: http://www.prr.msu.edu/stynes/nvum/NVUMFY2003report.pdf Sytnes, D.J., E.M. White, and LA. Leefers. 2003. Spending profiles of national forest visitors Years 2000 and 2001. Available on-line: httpzllwww.prr.msu.edu/stynes/nvum/NVUM2Year.pdf Stynes, D.J., G.L. Peterson, and DH. Rosenthal. 1986. Log transformation bias in estimating travel cost models. Land Economics 62(1): 94 - 103 US. Army Corps of Engineers. 2004. Recreation at a glance. Available on-line: http://corpslakes.usace.army.mil/employees/recreation/glance.html US. Census Bureau. 2004a. Census 2000, Summary File 1. Available on-line: http://factfinder.census.gov/home/saff/main.html?_lang=en US. Census Bureau. 2004b. Census block groups cartographic boundary files descriptions and metadata. Available on-line: http:l/www.census.gov/geo/www/cob/bg_metadata.html USDA Forest Service. 2000. National Forest Boundaries (Lower 48).Available on-line: http://roadless.fs.fed.us/documents/feis/data/gis/coverages/ 203 USDA Forest Service. 2004. Table 3, Areas by Region. IN Land Areas Report, September 30,2004. Available on-line: http://www.fs.fed.us/1and/staff/lar/LAR04/table3.htm Van Doren, CS. 1960. Recreational usage and visitors expenditure, Gavins Point dam and reservoir, summer 1959. Business research bureau, State University of South Dakota Bulletin Number 65 Walsh, R.G., J .R. McKean, and J .G. Hoff. 1992. Effects of price on forecasts of participation in fish and wildlife recreation: an aggregate demand model. Journal of Leisure Research 24(2): 140 - 156 Walsh, R.G., D.A. Harpman, J .G. Hoff, K.H. John, and J .R. McKean. 1988. Long-run forecasts of participation in fishing, hunting, and non-consumptive wildlife recreation. IN Outdoor Recreation Benchmark 1988: Proceedings of the National Outdoor Recreation Forum. January 13 - 14, Tampa Florida. USDA FS GTR-SE- 52. Ward, FA. and J.B. Loomis. 1986. The travel cost demand model as an environmental policy assessment tool: a review of literature. Western Journal of Agricultural Economics 11(2): 164 — 178 Waugh, FA. 1918. Recreation uses on the national forests. USDA Forest Service publication. Washington, DC. Wellman, J .D. and DB. Propst. 2004. Wildland Recreation Policy: An Introduction. Krieger Publishing. Melbourne, FL 204 llHfllllljlllflllHHHINllilllllllllllllllllflll 3 02736 5