@prefix vivo: . @prefix edm: . @prefix ns0: . @prefix dcterms: . @prefix dc: . @prefix skos: . vivo:departmentOrSchool "Arts, Faculty of"@en, "Geography, Department of"@en ; edm:dataProvider "DSpace"@en ; ns0:degreeCampus "UBCV"@en ; dcterms:creator "Grímsdóttir, Harpa"@en ; dcterms:issued "2009-11-24T21:45:00Z"@en, "2004"@en ; vivo:relatedDegree "Master of Science - MSc"@en ; ns0:degreeGrantor "University of British Columbia"@en ; dcterms:description """Over the last 34 years Canada has had an average of 11 avalanche fatalities per year and during the past five years this average has increased to 16 fatalities per year. Today, avalanche accidents happen primarily to people during recreational pursuits, and about half of the victims over the last 20 years were backcountry skiers. Backcountry skiing operations in Canada are making a constant effort to improve their avalanche safety. This study is based on data from a large heli-skiing operator in Canada; Canadian Mountain Holidays (CMH). The first objective of this study was risk analysis based on data from CMH's database, Snowbase. Skier triggered avalanches were analysed in term of factors such as elevation level, aspect, stability rating and the time of the year. When looking at human triggered avalanches, it is not possible to analyse the risk associated with these factors based on avalanche data alone; it is essential to have an idea about where and when people are skiing. Fortunately, Snowbase also contains information about the usage of defined ski runs within the operation areas, and therefore it is possible, perhaps for the first time, to extract some ideas about the relative risk associated with the different factors. The study shows that the historical risk of accidentally triggering an avalanche greater than size 1 depends highly on the stability rating, with the highest risk under "poor" stability. The risk is greater in higher elevation levels than lower down, and it is lower during late season than earlier on. The risk does not depend as much on aspect as may be indicated from avalanche data alone. However, it is relatively high in the N-NE-E sector. These factors are not independent of each other so analyses of combined factors were also performed. The second objective was to extract knowledge on avalanche risk management from professional mountain guides. Questionnaire and interviews with professional mountain guides were used as the tools for that, as well as observation of guiding in action and analysis of remarks in avalanche reports. The focus was on terrain selection and group management in terms of avalanche risk. My study indicates that when selecting terrain, guides first look at the overall shape and size of the terrain, but avalanche history of terrain and inclination are also important factors. Group management is an important part of avalanche risk management and the most important tools used by the guides are instructions to guests and the selection of regrouping spots. Experience is a significant factor in both terrain selection and group management. The third objective was to look at the possibility of using rule based decision methods for professional mountain guides in Canada. The result was that strictly rule based methods are not appropriate, because the guides are most likely able to make better decisions, and take more factors into account, based on their experience. However, some structure is desirable for the decision making process, in order to minimize the risk of human error, and among guides, some structure already exists. Based on the results from the first two objectives of the study, reccommendations are given in the form of suggestions rather than rules."""@en ; edm:aggregatedCHO "https://circle.library.ubc.ca/rest/handle/2429/15692?expand=metadata"@en ; dcterms:extent "16649731 bytes"@en ; dc:format "application/pdf"@en ; skos:note "A V A L A N C H E R I S K M A N A G E M E N T IN B A C K C O U N T R Y SK I ING O P E R A T I O N S B y H A R P A G R J M S D O T T I R B . S c , The University of Iceland, 1998 A THESIS S U B M I T T E D IN P A R T I A L F U L F I L M E N T OF T H E R E Q U I R E M E N T S F O R T H E D E G R E E OF M A S T E R S OF S C I E N C E in T H E F A C U L T Y OF G R A D U A T E STUDIES (Department of Geography) We accept this thesis as conforming to the required standard T H E U N I V E R S I T Y OF BRIT ISH C O L U M B I A June 2004 © Harpa Grimsdottir, 2004 Library Authorization In presenting this thesis in partial fulfillment of the requirements for an advanced degree at the University of British Columbia, I agree that the Library shall make it freely available for reference and study. I further agree that permission for extensive copying of this thesis for scholarly purposes may be granted by the head of my department or by his or her representatives. It is understood that copying or publication of this thesis for financial gain shall not be allowed without my written permission. Name of Author (please print) Date (dd/mm/yyyy) Title of Thesis: / f t / ^ / W ^ c t f/jt MfitftiOWWU't D e 9 r e e : Milkf nP JctMCP, _ Y e a r : 7J)0U( Department of / r c t ^ / T f y Qo\\ The University of British C lumbia Vancouver, B C Canada Abstract Over the last 34 years Canada has had an average of 11 avalanche fatalities per year and during the past five years this average has increased to 16 fatalities per year. Today, avalanche accidents happen primarily to people during recreational pursuits, and about half o f the victims over the last 20 years were backcountry skiers. Backcountry skiing operations in Canada are making a constant effort to improve their avalanche safety. This study is based on data from a large heli-skiing operator in Canada; Canadian Mountain Holidays (CMH) . The first objective of this study was risk analysis based on data from C M H ' s database, Snowbase. Skier triggered avalanches were analysed in term of factors such as elevation level, aspect, stability rating and the time of the year. When looking at human triggered avalanches, it is not possible to analyse the risk associated with these factors based on avalanche data alone; it is essential to have an idea about where and when people are skiing. Fortunately, Snowbase also contains information about the usage of defined ski runs within the operation areas, and therefore it is possible, perhaps for the first time, to extract some ideas about the relative risk associated with the different factors. The study shows that the historical risk of accidentally triggering an avalanche greater than size 1 depends highly on the stability rating, with the highest risk under \"poor\" stability. The risk is greater in higher elevation levels than lower down, and it is lower during late season than earlier on. The risk does not depend as much on aspect as may be indicated from avalanche data alone. However, it is relatively high in the N - N E - E sector. These factors are not independent of each other so analyses of combined factors were also performed. The second objective was to extract knowledge on avalanche risk management from professional mountain guides. Questionnaire and interviews with professional mountain guides were used as the tools for that, as wel l as observation of guiding in action and analysis of remarks in avalanche reports. The focus was on terrain selection and group management in terms of avalanche risk. M y study indicates that when selecting terrain, guides first look at the overall shape and size of the terrain, but avalanche history of terrain and inclination are also important factors. Group management is an important part of avalanche risk management and the most important tools used by the guides are instructions to guests and the selection of regrouping spots. Experience is a significant factor in both terrain selection and group management. The third objective was to look at the possibility of using rule based decision methods for professional mountain guides in Canada. The result was that strictly rule based methods are not appropriate, because the guides are most l ikely able to make better decisions, and take more factors into account, based on their experience. However, some structure is desirable for the decision making process, in order to minimize the risk of human error, and among guides, some structure already exists. Based on the results from the first two objectives of the study, reccommendations are given in the form of suggestions rather than rules. Table of contents Abstract ii Table of contents iv List of figures vi List of tables viii Acknowledgements x Chapter 1. Introduction 1 1.1 Report structure 2 1.2 The source of data and the area of study 3 1.3 The decision making process of guides in backcountry skiing operations 4 1.4 Research goal and objectives 7 1.5 Research rationale 7 Chapter 2. Literature review 9 2.1 Avalanche risk and backcountry travel in Canada 9 2.2 About risk 10 2.3 Slab avalanches 13 2.4 Spatial variations in the snow cover 14 2.5 Decision making and avalanche risk management 14 2.6 Decision making of experts in natural settings 15 2.6.1 Rational analysis 20 2.7 Human factors in avalanche forecasting 22 2.8 Decision making of mountain guides 23 2.9 Rule based approaches to decision making 24 2.9.1 The Swiss method of Werner Munter 24 2.9.2 The French Nivo Test 30 Chapter 3. Statistical risk analysis of skier triggered avalanches 32 3.1 Introduction 32 3.2 Data 32 3.2.1 Snowbase 32 3.2.2 Digital maps 37 3.3 Methods and results of single factor risk analysis 38 3.3.1 Elevation levels 39 3.3.2 Stability ratings 46 3.3.3 Aspect 50 3.3.4 Time of the year 55 3.3.5 Inclination 58 IV 3.4 Results from risk analysis of combined factors 60 3.4.1 Aspect and elevation levels 61 3.4.2 Aspect and time of the year 64 3.4.3 Elevation levels and time of the year 67 3.4.4 Aspect and stability 68 3.4.5 Stability ratings and elevation levels 70 3.4.6 Stability ratings and the time of the year 72 3.4.7 Stability, aspect and elevation 75 3.5 Remarks 76 3.6 Summary 78 3.7 Datasets for comparison 82 3.7.1 Avalanches where someone was caught 83 3.7.2 Skier accidentals, not including remotely and sympathetically triggered avalanches 88 3.8 Summary and conclusions 96 Chapter 4. The decision making process of heli-skiing guides in terms of avalanche risk 98 4.1 Introduction 98 4.2 Data and Methods 98 4.2.1 Questionnaire 98 4.2.2 Interviews 99 4.2.3 Observation 101 4.3 Results and conclusions from Questionnaire 102 4.3.1 Personal information 102 4.3.2 Terrain selection 102 4.3.3 Group Management 104 4.4 Analysis of remarks in reports of avalanches where someone was caught (Dataset 2) 106 4.4.1 Remarks on terrain 107 4.4.2 Human factors 108 4.5 Results and conclusions from interviews and observation 112 4.5.1 Characteristics of heli-skiing guides at CMH 112 4.5.2 Terrain selection in terms of avalanche risk 112 4.5.3 Summary on terrain selection 120 4.5.4 Group management in terms of avalanche risk 121 4.6 Factors potentially affecting decisions in a heli-skiing operation 127 4.7 The role of experience in decision making 130 4.7.1 Experience in avalanche risk management 130 4.7.2 Experience in the terrain 134 4.8 Rule based decision methods for professional mountain guides? 135 Chapter 5. Conclusions 137 5.1 Rule based approach 137 v 5.2 Recommendations, based on pattern from risk calculations 139 5.3 Further recommendations and contemplations 143 5.4 Suggestions for future work 145 References 147 Appendix A. The 3x3 Formula 150 Appendix B. The Canadian size classification system for avalanches 151 Appendix C. Copy of questionnaire. 152 Appendix D. Results from questionnaire 159 List of f igures Figure 1. Location of CMH's heli-skiing operation areas (CMH, 2004) 4 Figure 2. Canadian avalanche fatalities by activity for 230fatalities between October 1984 and September 2003 (CAA, 2004). 10 Figure 3. Risk - Decision matrix for backcountry skiing and helicopter skiing (McClung, 2002) 12 Figure 4. Comparison of the elevation level of avalanches [P(Et/AJJ with 95% confidence interval and usage of elevation levels [P(Ei)J. 41 Figure 5. The relative risk in the different elevation levels: P(ET /AJ / P(EJ. 43 Figure 6. The relative risk in the different elevation levels when avalanches with starting area in 2200 m a.s.l. are classified as \"alpine \". 44 Figure 7. The fraction of avalanche sizes in the different elevation levels: P(SZt /E/AAJ. 45 Figure 8. Comparison of the stability ratings under which avalanches fell[P(St/AJJ and usage of stability ratings [P(Sj)]. 48 Figure 9. The relative risk under different stability ratings: P(Si/As) / P(Sj). 49 Figure 10. The probability of sizes of avalanches under the different stability ratings: P(SZ, Figure 11. Comparison of aspect of avalanches [P(ASi /AJJ and usage of aspects [P(ASj)]. 51 Figure 12. Comparison of aspect of avalanches [P(ASj /AJJ and usage of aspects [PfAStf. 52 Figure 13. The relative risk in different aspects: P(ASj /AJ/PfASJ. _53 Figure 14. A radar plot of the relative risk in different aspects: P(ASi/As)/P(ASl). : 54 Figure 15. The relative risk in three different classes of aspect: P(ASi/AJ/P(ASi). 55 Figure 16. Comparison of the time periods when avalanches fell [P(Ti IAJ] and the usage of time periods [P(T)]. 57 Figure 17. The relative risk during different time periods: P(Tt P(Td- 57 v i Figure 18. The fraction of avalanche sizes during different time periods: PiSZjTMs)- 58 Figure 19. Inclination of the starting areas of avalanches: P(f /AJ._60 Figure 20. Aspect of avalanches at different elevation levels: PfASjEjAAJ. 61 Figure 21. Usage of aspects at different elevation levels: P(ASj /EJ). _61 Figure 22. Alpine - Comparison of aspect of avalanches [P(ASj /E^AA^)] and the usage of aspects [P(ASt /E)]. 62 Figure 23. Treeline -Comparison of aspect of avalanches [P(ASi /EJAAS)] and the usage of aspects [P(ASj I Ed]. 63 Figure 24. Sub-treeline -Comparison ofprobability of aspect of avalanches [P(ASt /EjAAJ] and the probability of usage of aspects [P(AS, IE^]. 63 Figure 25. Relative risk in the different elevation levels and aspects: [PfAStAEjAJ]/ [P(AS,AEM. . 64 Figure 26. Aspect of avalanches in different time periods: P(ASi /TIAAJ. 65 Figure 27. Usage of aspects in different time periods: P(ASt /TJ. 65 Figure 28. Relative risk in the different aspects at different time periods: [PCAS^TjAJJ/fPfASiATJJ. 66 Figure 29. Elevation levels of avalanches during different time periods: P(E, iTMs)- 67 Figure 30. Usage of elevation levels during different time periods: P(E,/Ti). 67 Figure 31. Relative risk in the different elevation levels at different time periods: [P(EIATI /AJ]/ [P(EIATJ)]. 68 Figure 32. Aspect of avalanches under different stability ratings: P(ASi/SjAAJ. 69 Figure 33. Usage of aspects under different stability ratings: P(ASt 69 Figure 34. Relative risk in the different aspects under different stability ratings: [P(ASiASi /AJJ/ [PfASjASJ]. 70 Figure 35. . Stability ratings under which avalanches fell in different elevation levels: P(Si /EJAAJ. 71 Figure 36. The usage of stability ratings in different elevation levels : PfSjEJ. 71 Figure 37. Relative risk in the different elevation levels under different stability ratings: [P(SIAEJ /AJJ/fPfoAEJJ. 72 Figure 38. Stability ratings under which avalanches fell during different time periods: P(Sj /TJAAJ. 73 Figure 39. The usage of stability ratings during different time periods: P(S,/Ti). 73 Figure 40. Relative risk during the different time periods under different • stability ratings: [P(SiATt /AJ]/ [P(TJAEJ]. 74 Figure 41. Comparison of the elevation levels of avalanches where someone was caught fP(Ej /AJ] and the elevation levels of skier triggered avalanches greater than size l[P(EtUj]. 85 Figure 42. Comparison of aspect of avalanches where someone was caught [P(ASj /AJ] and the aspect of skier triggered avalanches greater than size 1 P(ASt /AJJ. 86 Figure 43. Comparison of the time periods when avalanches, where someone was caught, fell [P(Tt /AJ] and the time periods when skier triggered avalanches, greater than size IJell [P(TjAJ]. 87 vi i Figure 44. Comparison of inclination of the starting area of avalanches in Dataset 2 [P(Tt /AJJ and inclination of the starting area of avalanches in Dataset 1 [P(Tt /AJJ. 88 Figure 45. Comparison of the elevation levels of avalanches in Dataset 3 [P(Ei /AJJ and the elevation levels of avalanches in Dataset 1 [PfEjAJ]. 89 Figure 46. Comparison between Dataset 3 and Dataset 1 for relative risk in the different elevation levels. 90 Figure 47. Comparison of aspects of avalanches in Dataset 3 [P(AS/AJJ and Dataset 1 [P(AS/AJ]. 91 Figure 48. Comparison between Dataset 3 and Dataset 1 on relative risk in the different aspects. 91 Figure 49. Comparison of the time periods of avalanches in Dataset 3 [P(T/AJJ and avalanches in Dataset 1 [P(T/AJJ 92 Figure 50. Comparison between Dataset 3 and Dataset 1 on relative risk during different time periods. 93 Figure 51. Comparison of stability ratings under which avalanches in Dataset 3 [P(Si I A,)] and Dataset 1 [P(S, I A J] fell. 94 Figure 52. Comparison between Dataset 3 and Dataset 1 on relative risk under different stability ratings. 94 Figure 53. Comparison of inclination of starting areas for avalanches in Dataset 3 [P(It /AJJ and Dataset 1 [P(U /AJJ. 95 List of tables Table I. Hazard potential in Munter's Reduction Method 27 Table 2. Reduction factors (RF) in Munter 's Reduction Method 28 Table 3. [P(EASAAS /AJJ / [P(EASAAS)] 75 Table 4. Relative risk for single risk factors 78 Table 5. Relative \"risk numbers \"for different stability ratings during different time periods 80 Table 6. Combined risk factors, associated with the highest historical risk 81 Table 7. Combined risk factors, excluding VP/P stability ratings, associated with the highest historical risk 81 Table 8. Questionnaire: Gender 159 Table 9. Questionnaire: Age 159 Table 10. Questionnaire: Guiding education 160 Table 11. Questionnaire: How inclination is taken into account 160 Table 12. Questionnaire: The importance of inclination under \"good\" stability rating 161 Table 13. Questionnaire: The importance of inclination under \"fair\" stability rating 161 Table 14. Questionnaire: The importance of inclination under \"poor\" s ta b ility ra ting 161 Table 15. Questionnaire: The importance of large scale aspect under \"good\" stability rating 162 Table 16. Questionnaire: The importance of large scale aspect under \"fair \" stability rating 162 Table 17. Questionnaire: The importance of large scale aspect under \"poor \" stability rating 162 Table 18. Questionnaire: The importance of small scale aspect under \"good\" stability rating 163 vi i i Table 19. Questionnaire: The importance of small scale aspect under \"fair \" stability rating 163 Table 20. Questionnaire: The importance of small scale aspect under \"poor\" stability rating 163 Table 21. Questionnaire: Hazard rating of small scale terrain features under \"good\" stability rating 164 Table 22. Questionnaire: Hazard rating of small scale terrain features under \"fair\" stability rating 164 Table 23. Questionnaire: Hazard rating of small scale terrain features under \"poor\" stability rating 165 Table 24. Questionnaire: Hazard scores of small scale terrain features 165 Table 25. Questionnaire: Hazard rating of terrain factors in morning meetings under \"good\" stability rating 166 Table 26. Questionnaire: Hazard rating of terrain factors during skiing under \"good\" stability rating 166 Table 27. Questionnaire: Hazard rating of terrain factors in morning meetings under \"fair\" stability rating 167 Table 28. Questionnaire: Hazard rating of terrain factors during skiing under \"fair \" stability rating 167 Table 29. Questionnaire: Hazard rating of terrain factors in morning meetings under \"poor\" stability rating 167 Table 30. Questionnaire: Hazard rating of terrain factors during skiing under \"poor\" stability rating 167 Table 31. Questionnaire: Hazard scores of terrain factors in morning meetings 168 Table 32. Questionnaire: Hazard scores of terrain factors during skiing 168 Table 33. Questionnaire: Clients following instructions 169 Table 34. Questionnaire: The effect of clients not following instructions 169 Table 35. Questionnaire: The selection of regrouping spots 170 Table 36. Questionnaire: The importance of regrouping spots 770 Table 37. Questionnaire: Spacing between skiers when skiing downhill 171 Table 38. Questionnaire: Spacing between skiers when traversing steep slopes 171 Table 39. Questionnaire: The importance of spacing between skiers 172 Table 40. Questionnaire: Strictness of group management / 73 ix Acknowledgements This study was supported by U B C Graduate Fellowship, Wendy Fan Memoria l Scholarship, Josephina T Berthier Fellowship as well as the Natural Sciences and Engineering Research Counci l of Canada and Canadian Mountain Holidays Chair in Snow and Avalanche Science at the University of Brit ish Columbia, Canada. This research was possible only because o f the support o f the people behind Canadian Mountain Holidays (CMH) . I must thank the administration team at C M H for providing data from their excellent Snowbase database, as wel l as additional data in form of avalanche paper reports. The database was originally created by Roger Atkins, a guide at C M H . I also thank the administration for giving me the oportunity to visit the heli-skiing operation areas of C M H . I would especially like to thank all the guides at C M H , who welcomed me to their operations, helped me any way they could, and gave me insight into their world. In particular, I am grateful to all the guides who answered my questionnaire and/or participated in the interviews. Gratitude is extended to Dr. Dave McClung for supervising this thesis and providing support for perhaps an unusual Physical Geography thesis. Pascal Hageli is thanked for his help and translation work. On a personal level my thanks go to my husband and two sons and the rest of my family for mental support throughout this study. Chapter 1. Introduction Backcountry skiing and backcountry travelling in mountains during winter has become increasingly popular in Canada over the last decades. The most significant risk associated with this kind of travel is due to avalanches, and a number of skiers and travelers in Canada loose their lives to avalanches every year. Many backcountry skiing companies operate in Western Canada, and their goal is to provide people with opportunities to ski in the backcountry under the supervision of educated and experienced ski-guides. Clients are either brought to the hills on foot, by a snowcat or snowmobile, or with helicopters. The guides in these companies make many decisions every day regarding the safety of their clients, most importantly in terms of avalanche risk. The decisions are mostly based on the experience and knowledge of individual guides. Quite a few attempts have been made to create rule based decision methods to help winter backcountry travellers estimate avalanche risk and select routes accordingly. In Europe, different methods have been developed in different countries, but in North-America such methods have not been in common use. Rule based methods are in most cases created with recreational travelers in mind however, professional guides have also been using them as an assistant tool. Rule based methods for decision making are at least partly based on statistics on avalanche accidents. The problem with defining risk factors from such statistics is that human triggered avalanches are usually encountered so human factors come into play. The natural release of avalanches happens only a very small percentage of the time, typically for a total of a few hours each winter (McClung, 2001). The probability of being caught by a 1 naturally released avalanche while traveling in the backcountry is therefore very low, and it is estimated that for about 90 percent of backcountry avalanche accidents, the avalanche is triggered by the victim or someone in the victim's party (e.g. Jamieson and Geldsetzer, 1996; Logan and Atkins, 1996; Tremper, 2001; McCammon, 2000). Therefore, it is a combination of human and physical factors that lead to an avalanche. Due to human factors, it is not sufficient to look at data on avalanches in order to define risk or risk factors for human triggered avalanches; it is crucial to look at data and information on the behaviour of travelers as well . No avalanche occurrences in a certain area during a given time period might simply reflect a low number of travelers rather than low risk. Information on where and when people travel in the backcountry is usually not readily available, and that is one of the big problems when analysing risk or defining risk factors associated with human triggered avalanches. 1.1 Report structure This report is divided into five chapters. In Chapter 1, a brief background of the study is presented, the company providing data for the study is introduced and the decision making process o f guides in backcountry skiing operations is described. The research goals and objectives are presented as wel l as the research rationale. Chapter 2 includes a review of the literature on avalanche accidents in Canada, risk studies, slab-avalanches, snow structure studies and avalanche risk management. A review of the literature on ideas and theories about how experts and professional guides make decisions is also included and examples of rule based decisions methods are presented. Chapter 3 contains results of risk analysis on skier triggered avalanches, as well as other statistical analysis based on C M H ' s database. In Chapter 4, results from questionnaires and interviews with professional guides about 2 avalanche risk management, are presented. In Chapter 4, the main focus is on terrain selection and group management. The results from observation during field trips and analysis of remarks in avalanche reports are also included in the results. In Chapter 5, conclusions are drawn from the research and recommendations based on the study are discussed. 1.2 The source of data and the area of study This research is based on data from Canadian Mountain Holidays (CMH), a heli-skiing company which operates in the Interior Ranges of British Columbia. C M H is the largest heli-skiing operator in the world, operating on more than 20,000 km of land in the Columbia Mountains of British Columbia, specifically in the Cariboo, Monashee, Purcell and Selkirk mountain ranges. Elevation ranges from 400 m in valley bottoms to approximately 3500 m above sea level. The snow climate is transitional with a strong maritime component (Hageli and McClung, 2003). Studies indicate that the natural avalanche activity on persistent weak layers is approximately 20% of the overall natural activity. The average value can vary between 0% during a maritime winter to about 40% for a winter with a strong continental influence (Hageli and McClung, 2003). C M H has 12 different operation areas, and employs more than 100 professional guides. Each operation area has between 200 and 300 defined skiruns, and the skiing activity is restricted to those runs. 3 16 T O E D M O N T O N 80 km/50 mites J f t t Vafemount ' ^ ALBERTA ^ — - V a f e m o u n t 4^ 1 1 J \" u ' * / 93 \"?s Silvertip ^ E ^ K Monashees H^^^Adamaf t t s / BANFF B R I T I S H 5C o t h J ! p r C O L U M B I A ( \\ Revebtoke 0 ,£^bte 6 TKam]oops S i c a m o u » ^ J ^ ^ ^ ^ g ^ ^ ^ \\ ^ Bugaboos AREASl K e i o w n a , ( Kootenay Galena Figure 1. Location of CMH's heli-skiing operation areas (CMH, 2004) C M H keeps a digital database on avalanche occurrences and the usage of skiruns, among other factors, and part of this study is based on analysis of those data. The information on the usage of skiruns is important, because an idea can be derived about the usage of different terrain factors and conditions, as well as when the usage occurs. Thus, from the data on skier triggered avalanches along with the data on usage it is possible, perhaps for the first time, to form some ideas about risk factors, in terms of terrain, stability conditions and time of the year. 1.3 The decision making process of guides in backcountry skiing operations The daily decision making process of guides in Canadian Mountain Holidays ( C M H ) is described below. The description is based on observations and interviews which were done for this study (refer to section 4.2 for a description of methods). It is included in this Chapter 4 because terms described here are used throughout the thesis. The decision process is similar in many other backcountry skiing companies in Canada. There are two steps in the daily decision making process of guides in C M H : 1) Morning meetings hi the mornings, before the skiing starts, the guides in each operation area have a morning meeting with two main tasks: a) Stability ratings. The stability ratings of the day for different elevation levels are decided. The stability ratings are: very poor, poor, fair, good and very good. The rating is given for: alpine, treeline and sub-treeline. Refer to page 36 for definition of the stability ratings. b) Runlist. The guides within an operation area make collective decisions regarding which runs can be skied during the day, in terms of avalanche risk. Those runs are assigned the color green. However, even though a run is made green, it does not imply that all parts of it are considered safe to ski, and it may be discussed during the meeting which areas of a green run should be avoided. Runs that are not considered safe to ski during that day are made red. If one of the guides thinks a run should be red for the day, it will be red, regardless of the other guides' opinions. Sometimes a run is assigned the color yellow, which means that it can only be skied after some conditions have been fulfilled (e.g. stability tests performed). Red runs cannot be \"greened\" during the day. 5 In the meeting, the first ideas about which areas to ski during the day are formed and a rough plan is made. The decisions during the morning meetings are based on factors such as the accumulated knowledge of how the snow-cover has been evolving, data from the I N F O - E X (a database consisting of snow and avalanche data from many different operations), the general weather forecast, and in some cases on information from remote weather stations. In general, during morning meetings, decisions on where to ski are made on the large scale, based on general data and information and large scale terrain characteristics. 2) Decision mak ing dur ing ski ing. One of the guides has the role of lead guide for the day, and he or she chooses the runs (from the \"green\" runlist) and skis the first line with his group. However, every guide is responsible for his or her own line within the runs. During skiing, every guide is constantly evaluating the avalanche risk, based on information that is directly related to the problem. They feel the snow while skiing and they may do stability tests in different locations. They talk to other guides and might observe some avalanche activity. During skiing they are constantly choosing terrain to ski, and the terrain factors of interest are often of a smaller scale than those the runlist is based on, such as small convex rolls, windloaded pockets or chutes in the trees. One of the guides is the snow safety guide for the day, and he or she focuses primarily on looking at the structure of the snow and how it interacts with terrain. He or she usually travels alone with a small helicopter and digs snow pits, does stability tests and feels the snow in different runs. The result of this work is a direct input to the decision making process, both in meetings (morning and evening) and during skiing. 6 Thus, it can be stated that the decision making process begins in the morning meetings as a general, large scale thinking, based on general information and data, and then during skiing, it narrows down to decisions on small scale terrain, based on information directly about the situation at hand (as described by McClung (2000)). 1.4 Research goal and objectives The objectives of this thesis are: 1 . Analysis of avalanche risk factors in terms of terrain: The analyses are based on a database that C M H maintains on skier triggered avalanches and usage of ski-runs. 2. Extraction of knowledge from professional guides, regarding avalanche risk management during backcountry skiing: The focus is particularly on terrain selection and group management. 3. A discussion of the possibility of rule based decision methods in terms of avalanche risk, for professional guides in Canada. 1.5 Research rationale A l l attempts to analyse avalanche risk factors for backcountry travelers have suffered from the lack of knowledge of where and when people are travelling. It is not possible to form an idea about risk, associated with human triggered avalanches, without having at least some clues about how and when people are using the terrain. The database analysed in this study contains relatively good data on skier triggered avalanches and recording of the usage of ski-runs. Thus, for the first time, it has been possible to take usage into account and to form ideas on relative risk associated with different factors. 7 Professional mountain guides hold a great amount of knowledge on avalanche risk management during backcountry traveling. A big part of this knowledge only exists in the guides' heads and has never been written down. In this thesis, an attempt is made to extract some of this knowledge and set it up in a formalized way. A big part of the daily decision making process of professional guides, in terms of avalanche risk, is based on the experience and knowledge of individual guides. In this thesis, the possibility of formalising the process, or using rule based methods for decision making in order to minimize errors, is discussed. Risk of avalanches wi l l always be associated with winter backcountry travelling in mountainous areas. However, it is important for the backcountry skiing companies in Canada to keep looking for ways to improve avalanche risk management in their operations, and the main purpose of this research is to contribute to that. 8 Chapter 2. Literature review 2.1 Avalanche risk and backcountry travel in Canada In the period from 1970 until the summer of 2003, 336 people in Canada have been ki l led by avalanches, which is an average of 11 avalanche fatalities per year. During the past five years this average has increased to 16 fatalities per year. The number of people venturing into the mountains for winter recreation has increased, and, therefore, an increase in the number of fatalities can be expected. (Jamieson and Geldsetzer, 1996; C A A , 2004). In the first half of the 20 t h century, avalanche accidents happened primarily to people working in, or driving through, avalanche terrain. Today, avalanche accidents happen primarily to people during recreational pursuits. As seen on Figure 2, most of those ki l led by avalanches in Canada are backcountry skiers, followed by snowmobilers (Jamieson and Geldsetzer, 1996; C A A , 2004). On a worldwide basis, more than 80% of fatal avalanche accidents now occur during backcountry travel (McClung and Schearer, 1993). Furthermore, the greatest risk for winter backcountry travelers is due to snow avalanches (Daffern, 1999). Between October 1984 and September 2003, 16% of the recreational accidents occurred in the Coast Mountains of B C , 47% occurred in the Interior Ranges of B C and 34% occurred in the Rocky Mountains. Winter recreation is more intense in the Interior Ranges than in the Rockies, which might explain the higher number of accidents there (Jamieson and Geldsetzer, 1996; C A A , 2004). 9 Figure 2. Canadian avalanche fatalities by activity for 230 fatalities between October 1984 and September 2003 (CAA, 2004). 2.2 About risk Risk has been defined in many ways. A broad definition by McClung (2002) is: \"risk is the probability or chance of death or losses\". A \"chance of death or losses\" refers to a general thinking about risk, not in a numerical way, while \"probability\" indicates a statistical matter. Narrower definition is needed, for each study involving risk analysis. For my study, risk contains the following elements: • Probabi l i ty or chance of avalanche triggering during a given period. • Vulnerabi l i ty is the probability of death, injuries or losses related to avalanche size. 10 • Exposure is the fraction of time or space exposed. A similar form of this nomenclature is established in several disciplines (e.g. Keylock, 1997; National Research Counci l , 1991; Fel l , 1994). In this thesis, I often refer to Historical risk, which is defined by Finlay and Fel l (1997) as a factual record, which may be used as a guide as to what is acceptable. Acceptable r isk Fel l 's and Hartford's (1997) definition of acceptable risk is: \"A risk for which, for the purposes of life or work, we are prepared to take pretty well as it is with no regard to its management\". It is the level of risk that people accept under given circumstances. If a level of risk associated with a place or an activity is higher than acceptable, some measures may be taken in order to reach acceptable levels. The level of acceptable risk in a society can be estimated indirectly by reviewing the decisions that have been taken historically in order to reduce individual risks (Wilhelm, 1997). The level of acceptability depends among else on cultural values in the society and the existing levels of risk from different sources. It also depends on whether or not the risk is considered voluntary and whether or not it is considered controllable (Fell, 1994). In general, the society may accept levels of voluntary risk as much as 1000 times higher than for involuntary risk (Starr, 1969). People undertake voluntary risk i f they believe that the benefits from the activity justify the cost. Driving is for example considered mostly voluntary risk, and people might accept it because the benefit of getting between places in a relatively short time is considered high enough to justify the risk. According to Wi lhelm (1997) it can be assumed that risk due to avalanches is mostly voluntary during activities such as backcountry skiing or ice cl imbing, and it is mostly involuntary for residential areas. 11 The term \"acceptable r isk\" implies that people are wi l l ing to take a certain amount of risk. L ike stated above, they consider the benefits of the activity high enough to justify the risk. Wi lde (2001) argues, on the basis of traffic accident data, that people constantly try to achieve an optimal value of risk called \"Target R isk \" which optimizes the difference between potential gains and losses during human activity. McC lung (2002) constructed a Risk-Decision Matrix following Adams (1995). Correct decisions are displayed to fall within two limits which define the lower and upper limits of acceptable risk for an individual. The area of correct decisions between these two limits is referred to as the Operational Risk Band (ORB) (Figure 3). The upper limit of the O R B is near the maximum risk people are wi l l ing to accept, while the lower limit represents the lower bound of acceptable risk for an individual associated with excessive conservatism. Going below the lower limit often represents lack of action where opportunities are missed such as exhilarating skiing. Wi lde's (2001) Target Risk is near the upper limit of the O R B . McClung (2002) argues that the goal of a helicopter skiing operation (from the avalanche perspective) is to maintain risk in the O R B : to provide more exciting skiing than a fixed lift ski area but to keep risk below one which provides excessive danger to clients. Propensity to take Risks Human Perception Data Decision-Analysis Making Accidents: Type I Error t Operational Risk Band Excessive Conservatism: Type II Error Figure 3. Risk - Decision matrix for backcountry skiing and helicopter skiing (McClung, 2002) 12 According to these suggestions, people are not constantly trying to minimize their risk, but rather keep it close to an \"optimal degree\" or within an \"Operational Risk Band\" by evaluating the costs and benefits of attending a \"r isky\" activity. How people perceive risk is not necessarily in correlation with the calculated value of risk. Risk perception depends on the individual's judgement and involves considerable subjectivity, governed by psychological factors (Finlay and Fel l , 1997). The \"Target R isk \" or \" O R B \" of an individual is based on risk perception rather than calculated probabilities. Refer to section 2.7 for further discussion on perception. 2.3 Slab avalanches The most common and serious form of avalanches involved in accidents are dry slab avalanches. Dry slab avalanches account for nearly all the avalanche deaths in North America (Tremper, 2001). A snow slab is a cohesive layer of snow with a thinner, weaker failure layer beneath it, and it becomes a slab avalanche once it is cut out around all boundaries by fractures (McClung and Schaerer, 1993). Slabs form because weather deposits snow in layers, and each different kind of weather affects the snow in a different way (Tremper, 2001). According to McClung and Schaerer (1993), two requirements must be met before a propagating shear fracture takes place within the slab: 1) The shear stress must approach the shear strength in the weak layer and 2) the rate o f deformation in the weak layer must be fast enough to provoke fracture. Instability develops when the first condition is met, but dry slab avalanches can start only when both requirements are fulfil led. It is convenient to distinguish between natural slab avalanches (not caused by humans) and artificial releases (caused by humans), including releases by explosives, 13 skiing, or other influences. The principal difference is the rate of application of the energy that starts the failure process. For skier triggered avalanches, the application of energy is due to dynamic forces from the weight of the skier, in addition to the weight of the slab. The most common trigger for natural dry slab avalanches is addition of weight by new snowfall, blowing snow, or rain (McClung and Schaerer, 1993). 2.4 Spatial variations in the snow cover Resent studies show that the spatial variability of snow stability can be great, even within relatively uniform slopes (e.g. Kronholm et al., 2002; Birkeland and Landry, 2002). Due to this variability, it cannot be assumed that a stability test is representative for larger area. h i order to have skier-triggered avalanches, some instability or imperfections have to be occurring in the snow-cover. The principal problem of forecasting natural avalanches is due to temporal and spatial variations of these instabilities (McClung, 2002), but human triggered avalanches have other influences including perception. A s the hazard can change on a micro-scale in the backcountry it is almost impossible to know where the imperfections are and thus, the avalanche forecasting process is a kind of a probabilistic analysis and decision-making is the key element for determining whether hazard exists. 2.5 Decision making and avalanche risk management Avalanche forecasting is the prediction of current and future snow instability in space and time, relative to a given triggering level (McClung, 2002). The goal of avalanche forecasting is to minimize uncertainty about instability introduced by three principal sources of uncertainty: 1) the temporal and spatial variability of the snow cover (including terrain 14 influences); 2) any incremental changes from snow and weather conditions; and 3) any human factors including variations in human perception and estimation (McClung, 2002). Due to the spatial and temporal variability in snowpack and weather, it is not possible to predict where and when every avalanche w i l l happen, and therefore a residual uncertainty exists in any avalanche forecast. For backcountry travelers, the residual uncertainty results in risk o f triggering avalanches; a risk that may be managed in terms o f probabilities with decision making, but it cannot be eliminated. The term \"avalanche risk management\" in this study refers to this. 2.6 Decision making of experts in natural settings. Features that help define a natural decision-making setting are: time pressure, high stakes, experienced decision makers, inadequate information (information that is missing, ambiguous, or erroneous), ill-defined goals, poorly defined procedures, cue learning, context (e.g., higher-level goals, stress), dynamic conditions, and team coordination (Orasanu and Connolly, 1993). Gary K le in (1998) and his team studied how people, e.g. fireground commanders make decisions in natural settings. One of the main results was that usually the commanders could reliably identify good options without comparing it to any others. On contrary, the classical theories o f decision making assumed formal analysis and comparison o f choices. Other studies of experts in complex real-world situation also suggest that the experts do not use heuristic or analytical strategies (e.g. Dreyfus and Dreyfus, 1986). K le in (1998) developed the Recognition-Primed Decision Model (RPD) in order to explain the decision making process of experts. The keypoints of the model is that experienced decision makers access a situation and judge it familiar (situation awareness), 15 and then courses of action can be quickly evaluated by imagining how they w i l l be carried out. They look for the first workable option they can find, not the best option. However, i f too many expectancies are violated in the process of action, the decision might be reconsidered. The emphasis is on being poised to act rather than being paralyzed until all the evaluations have been completed. The major factor that distinguishes experienced from less experienced decision makers is their situation assessment ability, not their reasoning processes per se according to K le in (1989), and Orasanu (1990). The situation assessment often happens intuitively, and according to K le in (1998), intuition depends on the use of experience to recognize key patterns that indicate the dynamics of the situation. Because patterns can be subtle, people often cannot describe what they noticed, or how they judged a situation as typical or atypical. Mental simulation plays an important role in the R P D model (Klein, 1998). It is the ability to imagine people and objects consciously and to transform those people and objects through several transitions, finally picturing them in a different way than at the start. It takes a fair amount of experience to construct a useful mental simulation. Mental simulation lets us explain how events have moved from the past into the present, and it also lets us project how the present wi l l move into the future. In that process we look at each step to see i f there could be a problem, trying to find pitfalls in advance. Then the plan might be modified to overcome the pitfalls, or it may be rejected. The more experienced the decision makers are, the more clear-cut are the expectancies. B y checking whether the expectancies are satisfied, the decision maker can judge the adequacy of the mental simulation. 16 Due to memory limitations people usually construct mental simulations using around three variables and around six transitions. K le in (1998) identifies some shortcomings of mental simulations: 1. The biggest danger of using mental simulation is that you can imagine any contradictory evidence away. If you are determined enough, you might never give up an explanation you have established. Once we have built a mental simulation, we tend to fall in love with it. 2. When you are pressed for time, you may not do as careful a job in building or inspecting the mental simulations you have constructed. 3. We have trouble constructing mental simulations when the pieces of the puzzle get too complicated - there are too many parts, and these parts interact with each other. According to K le in (1998), stories of incidents are used by experts to understand and they are useful as a form of vicarious experience for people who did not witness the incident. They also help to preserve values, by showing newcomers the kind of environment they are entering. K le in (1998) states that for researches, these kinds of stories also help understand situations and relationships and stories are a powerful method for elicting knowledge. If experts are asked what makes them so good, they are l ikely to give general answers that do not reveal much. But i f the researcher can get them to tell about tough cases, nonroutine events where their skills made the difference, then the researcher have a pathway into their perspective of the world. Recognitional decision making is defined as a singular evaluation. According to K le in (1998) people are more likely to use singular strategies when the time pressure is 17 greater, when people are more experienced in the domain, when the conditions are more dynamic and when the goals are i l l defined. In contrast people seem more l ikely to use comparative evaluation when they have to justify their choice, when the situation is computationally complex and when people are faced with unfamiliar situations. Thus, novices are on average more l ikely than experts to use comparative evaluation according to K le in (1998). Dreyfus and Dreyfus (1986) state in a similar way that novices tend to make decisions in an analytical way, based on context free rules, while experts make decisions in an intuitive way. According to Baron (1988), people in unfamiliar situations readily adopt simple guidelines and recipes for action, and w i l l typically not abandon them until they clearly fail. K le in (1998) states that there are many things experts can see that are invisible to everyone else. Below are a few examples: • Patterns that novices do not notice • Anomalies - violations of expectancies. Novices are confused by much that happens to them because they have so much trouble forming expectancies. As they begin to form expectancies, they also begin to get surprised each time an expectancy is violated. • The big picture (situation awareness). Experts have an ability to judge proto-typicality. Whereas novices may be confused by all the data elements, experts see the big picture, and they appear to be less l ikely to fall vict im to information overload. • Opportunities and improvisations. One aspect of being able to improvise is the ability of experts to generate counterfactuals: explanations and 18 predictions that are inconsistent with the data. Perhaps they have this ability because they have learned not to place too heavy a reliance on data. Novices, in contrast, have difficulty imagining a world different from the one they are seeing. • Differences that are too small for novices to detect. • Their own limitations. h i general, many studies of experts in naturalistic settings suggest that it is not feasible to apply classical decision research analyses to many real-life situations, and professional decision makers seldom rely upon classical theory to make decisions (Orasanu and Connolly, 1993 and Beach and Lipshitz, 1993, K le in , 1998). \"The standard advice for making better decisions is to identify all the relevant options, define all the important evaluation criteria, weight the importance of each evaluation criterion, evaluate each option on each criterion, tabulate the results, and select the winner. In one form or another, this paradigm finds its way into training programs the world over. Again and again, the message is repeated: careful analysis is good, incomplete analysis is bad. And again and again, the message is ignored; trainees listen dutifully, then go out of the classes and act on the first option they think of. The reasons are clear. First, the rigorous, analytical approach cannot be used in most natural settings. Second, the recognitional strategies that take advantage of experience are generally successful, not as a substitute for the analytical methods, but as an improvement on them\" (Klein, 1998). Klein 's studies involved mostly firefighter commanders and military personnel, but also airline pilots, nurses and other people making decisions in natural settings. In this paper it is suggested that part of the decision making process of mountain guides may follow a procedure similar to K le in 's R P D . There are, however, some limitations to that, since the decisions of mountain guides are, among else, based on knowledge of the snowpack that has accumulated through the season, rather than solely an evaluation of a single event. A lso, the high number of pieces in the puzzle that interact in a complicated way, along with a lack of 19 feedback to decisions, may cause problems for mountain guides using \"recognitional\" strategies. 2.6.1 Rat ional analys is It has been suggested above that experts do in general not rely on rational analysis when making decisions in naturalistic settings. However, rational analysis has its place in many forms of decision making. According to K le in (1998) rational analysis is a specialized and powerful source of power, however, the role it plays in decision making varies and depends on the task. Analysis and calculations let us find trends in noisy data. Rational analysis reduces the chance that an important option w i l l be overlooked. Decision trees and cost-benefit analysis can help us make sense out of complicated choices. For rational thinking we need to (Klein, 1998): • Decompose. We have to analyze a task - break the task, idea, or argument into small units, basic elements, so we can perform different calculations on them. • Decontextualize. Since context adds ambiguity, we must try to find units that are independent of context. We try to find a formal way to represent the world, to treat it as a representation, a picture, a model. We try to build theories and maps to substitute for having a sense of the task or the equipment. • Calculate. We apply a range of formal procedures on the elements, such as deductive rules of logic and statistical analyses. • Describe. A l l the analyses and representations should be open to public scrutiny. 20 Some of the limits of rational thinking according to K le in (1998) are: • Lack of Basic Elements. There are not any \"primit ives\" that naturally exist. The components defined are arbitrary and depend on individual goals and methods of calculation. There is no right way to break down a task. Different people find different schemes. • Ambiguous rules. Rules and procedures take some sort of the if-then form. We rarely try to plan out every contingency. Instead, we try to make it easy to understand the intent behind the rule or order. • Difficulty of setting up the calculations. Even when we know which rules apply and which to perform, we still have to initialize the equations or arguments. When the calculations require people to estimate probabilities or utilities, to estimate their values or to make other unnatural judgments, we are going to have trouble. • Combinatorial explosions. Formal methods of rational analysis can run into difficulties when they consider a large set of factors (as is found in a natural setting) and try to work out the implications of all the different permutations. In our everyday lives, we do not face combinatorial explosions because we are not relying on calculations. We use experiential sources of power to frame situations and arrive at manageable representations. Then, i f necessary, we bring in the analytical methods to add precision. The analytical methods run into limits when we try to use them without recourse to the experiential sources of power. 21 2.7 Human factors in avalanche forecasting Since most deaths in western Europe and North America are caused by people triggering the avalanches themselves (McClung and Schaerer, 1993), the root cause of most these accidents is failure in human perception (McClung, 2002). Perception is equivalent to one or more people's picture of reality based on information processing derived from the senses. It is important at several scales including individuals, groups, and levels of government (McClung, 2002). McClung (2002), defines common elements affecting human perception about instability in avalanche forecasting. Targeted education and experience are defined as positive factors affecting perception. McClung (2002) argues that experience combined with targeted education and objective reasoning is very effective for influencing correct decisions in avalanche forecasting and improving human perception. McC lung (2002) also lists examples of biases as negative elements affecting the human perception. According to McClung (2000) avalanche forecasting is not an event, it is an evolutionary process arising from information about the state of instability in the snow cover assembled cumulatively in time. Reasoning in avalanche forecasting involves both inductive and deductive processes (McClung 2002). LaChapelle (1980) argues that the fundamental reasoning process in avalanche forecasting is a dynamic, (mostly) inductive integrative process which is probabilistic in character with an intuitive component which is very difficult to reduce. McC lung (2002) states that even though avalanche forecasting is mostly an inductive process, proper forecasting should include results from deductive reasoning. Results from deductive reasoning come from targeted education including models. 22 McClung (2000) defines two general types of data used to forecast avalanches: 1. singular or case data about the specific situation at hand and 2. distributional data and information about similar situations in the past. Distributional data are linked to experience with avalanches and terrain, and inexperienced people may have extremely limited view of distributional information related to terrain at location (McClung, 2002). McClung(2002) argues that it is essential to combine both data types in avalanche forecasting. 2.8 Decision making of mountain guides Walter Bruns, a mountain guide and the president of C M H , describes the thought process of mountain guides in his paper \"Snow Science and Safety for the Mountain Guide (1997) \". The following is based on Bruns' paper. Due to the complicated environment, guides look for common themes and think in terms of patterns, relating these patterns more and more subconsciously as the complexity increases. Experience enters by providing a greater number of patterns to draw on and intuition enters by virtue of subconscious processes. Guides tend to think in terms of scenarios, and therefore powers of observation and a memory that captures images are crucial attributes of good guides. They take the patterns from hazard assessment, link them as still frames of a moving picture on the terrain and then run that movie forward in time to anticipate consequences. Simplification is important: each situation is reduced to its most basic component situations and dealt with in causal sequence. Input factors are screened so that only the most relevant need to be considered. Uncertainties are translated into levels of non-confidence; those for which there is lowest confidence become prime subjects for more extensive discussion. The process is iterative to identify and reject redundancies and inconsistencies. 23 Left-brain guides who are linear, analytic thinkers are quickly mired in uncertainty, and frozen into inaction, unless they make increasingly radical assumptions. Right-brain, kaleidoscopic thinking lends itself more readily to the challenge. Guides contribute to stability, hazard or risk assessments by telling stories of their experiences, relating a series of images, patterns or scenarios to the less experienced. 2.9 Rule based approaches to decision making A few different rule based approaches to decision making for travellers in avalanche terrain, have developed in Europe. This section contains a description of the method widely used in Switzerland, and for comparison, a short description of the French method is included as well. 2.9.1 The Swiss method of Werner Munter The following is based on: Munter (2003): 3X3 Lawinen: Risikomanagement im Winter sport. According to Munter, at the beginning of the 80's the current school of thought was dominated by the idea that the snow cover is more or less homogenous. It was believed that it is possible to analyse the snowpack in an entire area with a representative snow profile, and that the stability should be very similar across an individual slope. The new way of thinking according to Munter is: there is no release of slab avalanches without material flaws in the snow cover. A small super weak zone (material flaw) is needed for the initial fracture and a widespread weak layer is necessary for the fracture propagation. If there is an additional load, such as the weight of a skier, the minimal size of the deficit zone decreases. 24 Instead of looking at a snow profile, Munter attempted to determine the hazard rating by analysing the spatial distribution of stability. He found that each snowpack is a patchwork of areas of weak, intermediate, and strong stability, and under all conditions there are weak spots in the snow cover. It is the number, size, and distribution that are different depending on the hazard rating. Munter carried out a research project called M I S T A , in which he related results o f a large number o f \"Rutschblock\" stability tests (about 600 in Switzerland) to the official hazard bulletin. He found that the proportion of the stability class 'weak' (results from Rutschblock tests) doubles from one hazard rating to the next (exponential growth). The rule o f thumb is: 5% \"weak\" = L O W 10% \"weak\" = M O D E R A T E 20% \"weak\" = C O N S I D E R A B L E 40% \"weak\" = H I G H Based on his findings, Munter determined that a Rutschblock score is a probabilistic variable of a normal distributed population. Instead of identifying individual weak spots, it is only possible to estimate the hazard potential of an entire region (few km 2 ) according to Munter. Observations, such as alarm signs or critical new snow depth and measurements such as snow profiles, are qualitative point measurements and cannot be used to make conclusions about the stability across entire slopes or regions. Instead, they give us valuable information about the general conditions of a region. 25 The 3x3 formula The 3x3 formula is a tool Munter designed as a framework for a comprehensive analysis of avalanche hazard. It takes into consideration three different factors for the assessment o f potential dangers in a given slope: 1. Snow and weather conditions 2. Terrain 3. Human factors A l l these factors are considered at three geographical scales: 1. Regional 2. Local 3. Zonal It is compared to three sequential filters, which go from coarse to intermediate to fine (Refer to Appendix A for a more detailed description on the 3x3 formula). The idea is that on each scale it should be evaluated whether the snow and weather conditions, the terrain and the group is suitable for the trip. The answer to all three questions should be yes at all three scales. The method is a formalised way of applying the classical decision method at three different scales. A n example of evaluating terrain characteristics according to the 3x3 method would be to start the evaluation before the trip by looking at maps (coarse filter), then the knowledge is verified at the beginning of the trip (intermediate filter) and finally on the zonal scale, the individual slope is analysed (fine filter). 26 The overall safety rate of the 3X3 formula is considered too low, and therefore it is combined with a method called: \"The Reduction Method\" at each step. The Reduct ion Method Munter's motive for developing the Reduction Method was that especially for inexperienced people, he found there where not enough concrete rules, which left too much space for error in their decisions. His aim was to create a method with precise decision tools, without assuming special knowledge of snow science methods from the users. Input parameters should be limited to topographical parameters, and evaluation should only include simple considerations and combinations, and it should only involve thinking and not digging a snow pit or new expensive equipment. The goal of the method was to reduce the avalanche fatalities in Switzerland by half without an unacceptable decrease of the action range. Hazard potential The M I S T A project forms the basis for the definition of the hazard potential for Munter's Reduction Method. It showed that the hazard potential increases by a factor of 2 between different hazard levels. Therefore the numbers representing hazard potential are linked to the hazard rating of the avalanche bulletin as follows: Table 1. Hazard potential in Munter's Reduction Method Hazard rating in avalanche bulletin Hazard potential in Reduction Method Low 2 Moderate 4 Considerable 8 27 Reduction factors (RF) Reduction factors represent a quantitative measure of how certain behaviour reduce the hazard potential. They are based on statistics on avalanche accidents in the Swiss Alps. Table 2. Reduction factors (RF) in Munter's Reduction Method First class RF Reduction factor Steepest section 35-39' (below 4 2200 m a.s.l. Treeline >1800 and < 2200 m a.s.l. Sub-treeline: <1800 m a.s.l. In Snowbase, the maximum elevation in m a.s.l. is given for the starting areas of avalanches. Keep in mind that the elevation is not an exact measurement and it is usually recorded with 100 m intervals. Usage of elevation levels: The number people that used a specific run on a given day, is recorded in Snowbase. In order to estimate the usage of elevation levels for this research, a digital elevation model (DEM) was used to define the elevation levels of the runs and that way, the elevation levels of each run was recorded. If all or a part of a run lies above 2200 m a.s.l. it is assigned a 39 checkmark for \"alpine\". For every day this run has been used, the value 1 is added to the usage of \"alpine\". One run can have a checkmark for one, two or all three elevation levels. This is obviously not an accurate measurement of usage of elevation levels, since the usage of a big run that lies completely in the alpine adds the same value to the usage of \"alpine\" as the usage of a run with only a small part in the alpine. I am, however, looking at thousands of runs in total, and tens of thousands of user days and therefore, this error should not affect one elevation level more than others, and it can be assumed that the errors mostly cancel out. Results: Let: Ej = Elevation level (i = alpine, treeline, sub-treeline) A s = Accidentally skier triggered avalanche greater than size 1 Then: P ( E ; | A s) is the probability of being in a given elevation level, given a skier triggered avalanche greater than size 1. P ( E i ) is the fraction of time spent in a elevation level, during skiing. In the chart below, P(Ej | A s ) is compared to P(Ej). In other words the elevation levels of avalanches are compared to the usage of elevation levels. 40 ap(Ei\\As) mP(Ei) Alpine Treeline Sub-treeline Figure 4. Comparison of the elevation level of avalanches [P(Ej | As)] with 95% confidence interval and usage of elevation levels [P(Ej)]. The error bars for the avalanche data in the charts in this section represent 95% confidence assuming a large (n<100) random sample from a population with multinomial distribution. In Barber (1988) it is shown that for binomial random variable representing the number of successes in n trials where the probability of success in each trial is n, the confidence interval for n at a level of confidence 1 - a is: P - Z a/2>/(p(l-p)/n) < 71 < p + Z a/2V(p(l-p)/n) or p±z a / 2 V(p(l-p)/n) (3.1) where p is the sample proportion of successes in n trials and z is the value of the standard normal distribution with an upper-tail probability of a/2. In Olkin et al. (1980) it is shown that for trinomial distribution: o 2 x l = n p i ( l - p O , i = 1,2,3. (3.2) Thus, the the confidence level for trinomial distribution is: P i ± z o/2V(pi(l-pi)/n), i = 1,2,3. (3.3) 41 The confidence intervals give some ideas about what the sample of avalanches tells about the population it is sampled from. The sample, in this case, contains all recorded accidentally skier triggered avalanches greater than size one, over a seven years period. The population could be defined as all avalanches with the same criteria at the C M H operation areas, including the future ones and, thus, the analysis might have a predictive value. The question is whether seven years of recorded data can be defined as a random sample of the population. In reality, the weather and snow conditions probably show some variations over large time periods, adding to the uncertainty. The most important source of uncertainty is, however, the variations in human behaviour, which is an inseparable part of human triggered avalanches. There are no confidence intervals displayed for the usage data, since the sample is about 45,000 user days/run and thus, the interval would become very small and negligible. Figure 4 shows that more avalanches fell in the treeline than in other elevation levels, or almost 45% of the avalanches. About 30% fell in the alpine and 25% sub-treeline. However, the alpine part of runs were used only about 21% of the time, while treeline was used about 42% of the time and sub-treeline about 37% of the time. This indicates that the historical risk is higher in the alpine than in the treeline, and higher in the treeline than sub-treeline. Now: P ( A A E ) = P(A)P(E I A ) and P (EAA) = P(E)P(A | E) 42 and P(AAE) = P(EAA) Thus: P(E)P(A | E) = P(A)P(E | A) P(A | E ) = P(A)P(E | A)/P(E) (3.4) Thus, the historical risk of triggering an avalanche greater than size one in a given elevation level is calculated as: P(A S I E ) = [P(AS) P(E, | A S )] / [ P ( E 0 ] The calculations are based on historical records. For the purpose of this research, the interest is not in the actual probability, but rather in the relative probability between different elevation levels. Thus, we can remove P(A S) from the equation and work with risk proportionality: P(AS | EO oc P(E, | A S ) / P(E0 (3.5) 1 I_ • • 11 Alpine Treeline Sub-treeline Figure 5. The relative risk in the different elevation levels: P(Ej | A s ) / P(Ei). The error bars in Figure 5 are derived from the confidence levels of P ( E A S ) . 43 Figure 5 shows that the historical risk of triggering an avalanche greater than size 1 is more than two times higher in the alpine than sub-treeline, and the risk in the treeline is approximately in between. Within the 95% confidence level for the avalanche data, the difference is statistically significant. Thus, the historical risk increases on average with higher elevation levels. A closer look at the data shows that a large number of avalanches have starting areas recorded at exactly 2200 m a.s.l. According to the classification used here, these avalanches fell in the treeline. If we add them to the alpine avalanches the chart showing relative risk in the different elevation levels would look like this: 2.50 -> 2.00 — 1.50 1.00— 0.50 — 0.00 -I ' 1 - i ' * 1 '• ' ' ' ' -Alpine Treeline Sub-treeline Figure 6. The relative risk in the different elevation levels when avalanches with starting area in 2200 m a.s.l. are classified as \"alpine\". Figure 6 indicates that the historical risk in the elevation area that includes the alpine and the uppermost part of the treeline, is considerably higher than in lower elevation levels. A l l skier triggered avalanches greater than size 1 are treated the same in the risk calculations above, and as stated before, this reflects a kind of acceptability in an operation. 44 However, this definition of risk includes only the probability of avalanches plus exposure. The probability of injuries or death (vulnerability or consequences) most likely rises with increasing size of avalanches and, therefore, it is of interest to look at the sizes of avalanches in the different elevation levels. Let: SZi = Size of an avalanche (i = 1.5, 2, 2.5, 3, 3.5) A s = Accidentally skier triggered avalanche greater than size 1 E i = Elevation level In the chart below, the fraction of avalanche sizes are compared between the different elevation levels. Then: P(SZj EJAA s ) is the fraction of avalanche sizes in elevation levels. 0.6 0.5 0.4 0.3 • 3.5 • 3 • 2.5 • 2 • 1.5 0.2 0.0 0.1 Alpine Treeline Sub-treeline Figure 7. The fraction of avalanche sizes in the different elevation levels: P(SZi | EJAAS). 45 Figure 7 shows that there is a lower fraction of size 1.5 avalanches and a higher fraction of bigger avalanches in the alpine than in the lower elevation levels. The sizes of avalanches in the treeline and sub-treeline are similar. 3.3.2 Stability ratings Methods: Stability ratings when avalanches fell The stability rating is given for the three different elevation levels: alpine, treeline and sub-treeline. Thus, in order to define under which stability rating an avalanches fell, the elevation of the starting areas has to be taken into account. For example, an avalanche with the starting area in the alpine (> 2200 m a.s.l.) fell under the stability rating given in the morning meeting for the alpine that day. It should also be noted that each operation area has its own stability rating for the day. Refer to page 36 for a detailed description of the stability ratings. Here, the stability ratings \"good\" and \"very good\" are combined into one group: G / V G . The same applies to the stability ratings \"poor\" and \"very poor\" (VP/P). The stability rating \"fair\" (F) is treated as it is. The reason is that the stability ratings \"very good\" and \"very poor\" are used relatively seldom (less than 2% of the time), and the combination of groups simplifies the analysis and results. Usage of stability ratings. The stability rating is given for the three different elevation levels daily. For example, i f a stability rating is \"fair\" in the alpine on a given day, then for each run used that day, which 46 is partly or fully in the alpine, the count 1 (1 user day) is added to the usage of the stability rating \"fair\". The accuracy of the measurements is affected by the same errors as for the analysis of the usage of elevation levels. Results: Let: Si = Stability rating (i = V P / P , F, G / V G ) A s = Accidentally skier triggered avalanche greater than size 1 Then: P(S j | A s ) is the probability of a stability rating, given a skier triggered avalanche greater than size 1 . P(Si) is the fraction of time a stability rating is used. On the chart below, P(Sj | A s ) is compared to P(Sj). That is, the stability ratings under which avalanches fell are compared to the usage of stability ratings. The error bars represent 95% confidence as before. 47 0,7 0,6 0,5 0,4 0,3 0,2 0,1 0 VP/P F G/VG Figure 8. Comparison of the stability ratings under which avalanches fell [P(S, | As)] and usage of stability ratings [P(Sj)]. Figure 8 shows that more than half of the avalanches fell under the stability rating \"fair\" and almost 30% under \"very poor\" or \"poor\" (VP/P) stability ratings. However, the stability rating \"fair\" (F) was used almost half of the time, while V P / P were used only 8% of the time. Thus, the historical risk under V P / P stability ratings is higher than under \"fair\" stability rating. The historical risk under \"good\" and \"very good\" (G /VG) stability rating is lower than for the others, since less than 20% of the avalanches fell under those stability ratings while they were used almost 45% of the time. The relative risk is calculated as (refer to equation 3.4 and expression 3.5 on page 43): P ( A s | S i ) o c P ( S i | A s ) / P ( S i ) 48 VP/P F G/VG Figure 9. The relative risk under different stability ratings: P(Sf IAJ / P(SO. Figure 9 shows that the relative risk of triggering an avalanche greater than size 1, under the stability rating \"poor\" or \"very poor\" is about 3.5 times higher than under \"fair\" stability rating. The risk under \"fair\" is however more than two times higher than under \"good\" or \"very good\" stability rating. The difference is statistically significant. As before (refer to page 45), it may be interesting to look at the fraction of avalanche sizes under different stability ratings (Figure 10). 49 0.6 0.5 0.3 0.2 0.0 • 1.5 • 2 • 2.5 • 3 • 3.5 VP/P G/VG Figure 10. The probability of sizes of avalanches under the different stability ratings: P(SZ; | SJAAS). Figure 10 shows that under G / V G stability ratings the fraction of the smallest (size 1.5) avalanches is higher than for the other stability ratings. The stability rating \"fair\" has the highest fraction of size 2 avalanches. Under V P / P stability ratings the fraction of size 1.5 avalanches is slightly higher than under \"fair\" stability ratings however, the fraction of size 3 avalanches is also higher. The fraction of size 2.5 avalanches is quite similar under all stability ratings. 3.3.3 Aspect Methods: Aspect of avalanches The starting areas are recorded with eight different aspects: N , N E , E, S E , S, SW, W and N W . Note that the recorded aspect is often the estimation of the guide. For example, an aspect recorded as N could easily be N E or N W . Usage of aspects 50 With the help of D E M s and GIS, each run is assigned a main aspect. If a run with the main aspect \"North\" is used for one day, the count 1(1 user day) is added the usage of \"North\". This is not an accurate measurement of the usage of aspects since within each run, starting areas with variable aspects may be found. However, as before, the errors should mostly cancel out due to the high number of runs and user days. Results: Let: A S , = Aspect of starting areas (i = N, N E , E, SE, S, SW, W, NW) A s = Accidentally skier triggered avalanche greater than size 1 Then: P(ASj | A s) is the probability of an aspect (of starting area), given a skier triggered avalanche greater than size 1. P(AS0 is the fraction of time each aspect is used. N 0,3 j 0,25 1 Figure 11. Comparison of aspect of avalanches [P(ASj | As)] and usage of aspects [P(ASi)]. 51 0,35 0,3 0,25^ • ?(.45 1.4.) mp(AS') N NE E SE S SW W NW Figure 12. Comparison of aspect of avalanches [P(ASj | As)] and usage of aspects [P(ASj)]. Figure 11 and Figure 12 show that most avalanches fell in northern, northeastern and eastern aspects. These, along with northwest, are also the aspects that are most frequendy used. However, the ratio between avalanche occurrences and usage is greatest in north aspects, since 25% of the avalanches fell there, but it was only used about 17% of the time. The opposite is true for the aspect northwest, since it is used 16% of the time and only 10% of the avalanches fell there. The relative risk is calculated as (refer to equation 3.4 and expression 3.5 on page 43): P (A S | A S O oc P ( A S i | A s ) / P(ASd 52 2.00 1.80 1.60 1.40 1.20 1.00 0.80 0.60 0.40 0.20 INI NE E SE S SW W NW Figure 13. The relative risk in different aspects: P(ASj | As) / P(ASi). The error bars are as before derived from the 95% confidence level for the avalanche data. The relatively large error bars in Figure 13 reflect that the uncertainty increases when we are looking at eight classes instead of just three before. The large error bar for south aspects shows that there are not as many avalanches recorded there as in e.g. north aspects. The risk is on average higher in the main aspects; N, S, E and W than other aspects. This is most l ikely due to inaccurate recording of aspects and in reality the risk is probably more rounded out between aspects. 53 Figure 14. A radar plot of the relative risk in different aspects: P(AS( I As) / P(ASi). Figure 14 shows that i f looked at in terms of 180 degrees sectors, the risk is generally slightly higher on east aspects than west aspects, but there is little difference between north and south aspects. It would be of interest to simplify the aspect data and use fewer classes. It is however not simple to create new classes. If I would e.g. choose to use only the main aspects N, E, S and W, where would avalanches recorded in other aspects belong? Would it be reasonable to take avalanches recorded with N E aspects and divide them equally between N and E aspects? There are some indications from other datasets (see section 3.5) that the aspects N, N E and E belong together in a class rather than e.g. N W , N and N E . I chose not to split up the defined classes in Snowbase but combined them in three different classes (Figure 15). 54 Let: AS i = Aspect of starting areas (ASi = N or N E or E, A S 2 = S E or S or SW, A S 3 = W or N W ) 1.40 1.20 1.00 0.80 0.60 0.40 0.20 N-NE-E SE-S-SW W-NW Figure 15. The relative risk in three different classes of aspect: P(ASil A J / P C A S i ) . Figure 13 and Figure 15 indicate that the risk of triggering avalanches greater than size 1 is not as dependent on aspects as suggested by data on avalanche occurrences alone. On average, the risk is lowest in the W - N W sector and relatively high in the N - N E - E sector. However, the results depend on how the classes are defined. 3.3.4 Time of the year Methods: One of the factors recorded in Snowbase, that may affect avalanche risk, is the time of the year. The characteristic of the snowcover changes significantly from the early winter, to mid winter and towards the spring. In co-operation with C M H guides (Bezzola, 2003: personal communication), three time periods within the skiing season were defined: 55 1) Ea r l y winter: December 1 - January 31. The time when solar radiation does not have much affect. Aspect is not one of the most important things when it comes to choosing terrain. 2) M i d winter: February 1 - M a r c h 15. Solar radiation becomes increasingly important. The snowpack gets thicker and much more complex, with large variations in structure. 3) Late season: M a r c h 15 - A p r i l 31. Warmer temperatures and solar radiation affect the terrain selection a lot. The snow gets wet and heavy in lower elevations. Poor skiing conditions, not just in the afternoon, except i f there is corn snow in the afternoon of sunny days. Results: Let: Tj = Time of the year (T, - Dec 1 - Jan 31, T 2 = Feb 1 - Mar 15 ,T 3 = Mar 1 6 - A p r 31) Then: P(Ti | A s ) is the probability of a time period, given a skier triggered avalanche great than size 1. P(Tj) is the fraction of time that a time period is used. 56 Dec 1 -Jan 31 Feb 1 - Mar 15 Mar 16-Ap 30 Figure 16. Comparison of the time periods when avalanches fell [P(Ti | As)] and the usage of time periods [PCT,)]. The relative risk is calculated as (refer to equation 3.4 and expression 3.5 on page 43): P ( A s | T i ) « P ( T i | A s ) / P ( T i ) The results are shown in Figure 17 below. Dec 1 -Jan 31 Feb 1 - Mar 15 Mar 16-Ap 30 Figure 17. The relative risk during different time periods: P(Tj | As) / P(T(). 57 Figure 17 shows that the historical risk during late season is only half of what it is during mid winter. There are at least two possible reasons for that: 1) The snowpack is probably on average more stable during late season than earlier on. 2) During late season the instability is often due to solar radiation, which is a relatively manageable factor for professional guides. Note that the 95% confidence intervals overlap for early winter and mid winter. The fraction of avalanche sizes during the different time periods are displayed in Figure 18. • 1,5 • 2 • 2,5 • 3 • 3,5 — — L . it Dec 1-Jan 31 Feb 1 - Mar 15 Mar 16-Ap 30 Figure 18. The fraction of avalanche sizes during different time periods: P C S Z J T A A S ) . Figure 18 suggests that avalanches are on average slightly larger during early winter than later on. Note that there are fewer avalanches during late season to base the analysis on, and, therefore, the statistical uncertainty is higher than for early and mid winter. 3.3 .5 Inclination Methods: Inclination of starting areas: 58 The inclination of the starting areas of avalanches is in most cases recorded. Often, it is not a result of a measurement but an estimation of the guide and the inclination tend to be recorded with 5 degrees intervals. Usage of inclination: I did not do any analysis on the usage of inclination because it is not possible to estimate it in a sufficient way. The only information on usage of terrain in \"Snowbase\" is the usage of whole runs. Within a run it is possible to choose between many different lines and I do not have information on whether the steepest parts of the runs were skied or not. Furthermore, inclination is important on the small scale; only a small piece of terrain of the critical inclination is needed for a slope to avalanche. The resolution of the D E M s used here is too coarse to identify all potential starting areas in terms of inclination. With no usage data it was not possible to do risk analysis on the inclination factor. However; results of inclination analysis for the avalanche data are presented below. Results: Let: Ii = Inclination of the starting area of an avalanche (Ii - 15° -27° , h = 2 8 ° - 3 2 ° , I3 = 3 3 ° - 3 7 ° , Lt = 3 8 ° - 4 2 ° , I 5 = 4 3 ° - 4 7 ° , I 6 > 4 7 ° ) Then: P(Ii I A s ) is the probability of an inclination, given a skier triggered avalanche greater than size 1. 59 0,5 0,45 0,4 0,35 0,3 0,25 0,2 0,15 0,1 0,05 0 15-27 28-32 33-37 38-42 43-47 >47 Figure 19. Inclination of the starting areas of avalanches: P(Ij | As). The median inclination is 35 degrees, which is lower than for datasets in many other studies on human triggered avalanches (e.g. Schweizer and Jamieson, 2000 and Schweizer and Lutschg, 2000). That might reflect the terrain selection of guides. 3.4 Results from risk analysis of combined factors The factors analyzed in section 3.3, cannot be assumed to be independent from each others. Therefore, it is necessary to analyze any data on combined conditions directly from the database. The results in this section are not presented with confidence intervals, and in fact the statistical uncertainty is high because the data is divided in many classes. However, the analyses represent the historical risk in C M H from the period the data were collected. 60 3.4.1 Aspect and elevation levels The aspect of starting areas is dependent on elevation as can be seen in Figure 20, which shows the aspect of starting areas of skier triggered avalanches in the different elevation levels: Let P(ASj | E | A A S ) be the probability of an aspect, given a skier triggered avalanche greater than size 1 and an elevation level. Let P(ASj | Ei) be the fraction of time each aspect is used, given an elevation level. Note that the probabilities of aspects sum up to 1 for each elevation level for both the avalanches and the usage. This makes it easy to compare the probability of aspects in the different elevation levels (see Figure 20 and Figure 21). Avalanches P(ASi | EJAA S ) Usage P(ASJ|EJ) Figure 20. Aspect of avalanches at different elevation Figure 21. Usage of aspects at different levels: P(AS; | E : A A s). elevation levels: P(ASil E F). 61 From Figure 20 it may be concluded that in the treeline and sub-treeline a higher ratio of skier triggered avalanches released in northern aspects than for avalanches in the alpine. Figure 21 shows that this is not explained by higher usage of runs with northern aspects in the treeline and sub-treeline. On the contrary, the usage of aspects is not highly dependent on elevation levels. The probability of aspects at each elevation level is shown in below. Alpine NW N 0,25 j 0,2- NE W i 1 v SW SE H—' E — P{AS,\\E,r\\A.) — P(AS,\\E,) s Figure 22. Alpine - Comparison of aspect of avalanches [P(ASj | EJAAS)] and the usage of aspects [P(ASj | Ej)]. 62 Treeline Sub-treeline N N 0.4 T W I 1 1 H SW NW SE NE -^i E P ( A S , \\ E i ( ^ A * ) — P(AS,\\E>) SW SE S s Figure 23. Treeline -Comparison of aspect of avalanches [P(ASj | E ;AA s)] and the usage of aspects [P(AS; | E)]. Figure 24. Sub-treeline -Comparison of probability of aspect of avalanches [P(ASj | EJAAS)] and the probability of usage of aspects [P(ASj E)]. Figure 22, Figure 23 and Figure 24 indicate that in the alpine, the risk is highest in east aspects as well as southeast and south, while the risk is by far highest in north aspects in the treeline and sub-treeline. The relative risk in the different elevation levels and aspects is calculated as (refer to equation 3.4 and expression 3.5 on page 43): P(AS | ASjAEi) oc [P(ASiAEj As)]/ [P(ASiAEj)] 63 It is not clear why the risk in the alpine is highest in E-SE-S aspects, while it is highest in northern aspects lower down. Perhaps, wind loading has a greater effect in the alpine, resulting in increased risk in eastern aspects. Solar radiation might also pose more risk in the alpine than lower down, due to the open terrain. Lower elevations are probably more prone to surface hoar formation, which might also be more persistent than in the alpine, due to lesser wind effect. That might result in greater risk in northern aspects at lower elevations. However, these are only speculations based on physical reasoning since the data are not available to prove them. 3.4.2 Aspect and time of the year It is l ikely that the aspect of avalanches, as well as the usage of aspects, depends on the time of the year. 64 Let P (ASi | TJAA s ) be the probability of an aspect, given a skier triggered avalanche greater than size 1 and a time period. Let P (ASj | TO be the fraction of time each aspect is used, given a time period. The probabilities of aspects sum up to 1 for each time period, for both the avalanches and the usage, on the charts below. Avalanches P (ASi | T ; AA s ) Usage P (ASi | Tj) Figure 26. Aspect of avalanches in different time Figure 27. Usage of aspects in different time periods: P (ASi | TJAA,). periods: P(AS; IT). Figure 26 shows that in the period from March 16 through the end of the season, more than 30% of the avalanches were recorded in north aspects. The ratio of north aspect avalanches is slightly lower for the earlier time periods. The main difference in usage of aspects (Figure 27) between the different time periods is that in late season, northern and northwestern 65 aspects are used more frequently than earlier, on the cost of southerly and easterly aspects. Southern and southwestern aspects are used more frequently during early winter than later on. The relative risk in the different aspects at different time of the year is calculated as (refer to equation 3.4 and expression 3.5 on page 43): P ( A S | ASJATJ) OC [P(ASJATJ I A S ) ] / [P(ASIATJ)] • Dec 1-Jan 31 mFeb 1 - Mar 15 • Mar 16-Ap 30 N NE E SE S SW W NW Figure 28. Relative risk in the different aspects at different time periods: [P(AS;AT; | AS)]/ [P(ASJATJ)]. It is interesting that both the highest and one of the lowest risk numbers are found in the south aspect, the highest during mid season and the lowest during late season. The low risk in south aspects during late season is probably because the risk due to solar radiation in southern aspects is relatively obvious and manageable. This risk is maybe more difficult to manage in mid-winter, when solar radiation is starting to affect the snow stability. These results are, however, affected by statistical uncertainty, which is especially high in southern aspects where relatively few avalanches are recorded. 66 In early winter, the risk is on average higher in N W , N, N E and E aspects than other aspects. During late season, the risk is relatively higher in N, N E , E and SE aspects than other aspects. In mid winter, the risk numbers spike in northern and southern aspects and are lowest in N E and N W aspects. Note that inaccurate recording affect these results, indicating higher risk in main aspects than others. 3.4.3 Elevation levels and time of the year The elevation levels of avalanches and the usage of elevation levels are affected by the time of the year. Figure 29 suggests that a higher ratio of alpine avalanches and a lower ratio of sub-treeline avalanches are found during late season than earlier in the winter. The ratio of usage of the alpine increases as the season goes on. Avalanches P(E, | T . A A , ) Usage P(Ej | T , ) ~ y , . 1 Dec1-Jan31 Feb1-Mar15 Mar16-Ap30 Dec 1-Jan 31 Feb 1-Mar 15 Mar16-Ap30 Figure 29. Elevation levels of avalanches during F i g u r e 3 0 U s a g e o f elevation levels during different time periods: P(ES | TyvA,). different time periods: P(E I Ts). 67 The relative risk in different elevation levels and at different time of the year is calculated as (refer to equation 3.4 and expression 3.5 on page 43): P ( A S | EIATI) - [P(EJATJ I A,)] / [P(EiATi)] 3 j - —| 2,5 1 2 • Alpine a Treeline • Sub-treeline Dec 1-Jan 31 Feb 1-Mar 15 Mar16-Ap30 Figure 31. Relative risk in the different elevation levels at different time periods: [PCEJAT; | A J ] / [P(E;ATi)]. Figure 31 indicates that during the whole year, the risk is highest in the alpine and lowest at sub-treeline. The risk at treeline is higher during mid winter than early winter, while the risk at the other elevation levels is very similar between these two periods. Like already noted, the risk is in general much lower during late season than earlier in the winter. 3.4.4 Aspect and stability The aspect of avalanches is not highly dependent on stability ratings. It is e.g. less dependent on the stability ratings than on elevation levels (See Figure 32 and Figure 33). 68 Avalanches P(ASi | (S,AA S) Usage P ( A S i l S i ) Figure 32. Aspect of avalanches under different F i 8 u r e 3 3 - U s a g e of aspects under different stability ratings: P(AS; | S,AAs). stability ratings: P(AS; | Ej). The aspect pattern for both avalanches and usage under V P / P stability ratings differ slightly from \"fair\" and G / V G , which are more or less similar. The relative risk in the different aspects under different stability ratings is calculated as (refer to equation 3.4 and expression 3.5 on page 43): P ( A S | ASIASI) oc [P(ASjASi | A S ) ] / [P(ASiAS0] 69 • VP/P • F • GA/G 1 1 i 11 -1 I I i In f l N NE E SE S SW W NW Figure 34. Relative risk in the different aspects under different stability ratings: [P(AS,ASJ | A S)]/ [P(AS;AS;)]. Figure 34 shows that in all aspects the relative risk is highest under \"very poor\" and \"poor\" stability ratings and lowest under \"good\" and \"very good\" stability ratings. When the stability rating is VP/P the risk is highest in northeast aspects, while it is highest in north aspects when the stability rating is \"fair\", and south aspects under \"good\" and \"very good\" stability ratings. When the stability rating is F and G/VG the risk is relatively higher in southern aspects than when the stability rating is P/VP. 3.4.5 Stability ratings and elevation levels Figure 35 and Figure 36 indicate that stability ratings and elevation levels are interdependent for both avalanches and usage. 70 Avalanches P(Sj EJAA S ) Usage P(Si | Ej) Alpine Treeline Sub-treeline A | P i n e Treeline Sub-treeline Figure 35. . Stability ratings under which avalanches F igur e 36. The usage of stability ratings in different fell in different elevation levels: P(Si | E;AA s ) . elevation levels: P(Sj | E ;). Figure 35 shows that avalanches most often fell under \"fair\" stability rating in all elevation levels. However, the ratio of avalanches falling in \"fair\" stability is highest in the alpine and diminishes with lower elevation levels. The opposite is true for \"very poor\" and \"poor\" stability while there is not a great difference between elevation levels under \"good\" and \"very good\" stability ratings. In the alpine and treeline, \"fair\" is the most commonly used stability rating. Sub-treeline, the stability ratings G/VG are the ones most often skied. The relative risk in the different elevation levels under different stability ratings is calculated as (refer to equation 3.4 and expression 3.5 on page 43): P ( A S | EIASO cc [P(EiASi | A S )] / [P(EiASi)] 7 1 4,5 Alpine Treeline Sub-treeline Figure 37. Relative risk in the different elevation levels under different stability ratings: [P(SJAE; | A,)]/ [PCSJAEJ)]. Figure 37 shows that the relative risk is by far highest under \"very poor\" and \"poor\" stability ratings in all elevation levels, and it is always higher than risk under \"fair\" stability rating. It is interesting when looking at both elevation levels and stability ratings, that the relative risk is highest in sub-treeline and V P / P stability, even though the risk is in general lower sub-treeline than at higher elevation levels. This is due to how rarely V P / P stability ratings are used sub-treeline, while about 35% of the avalanches recorded there fell under those stability ratings. 3.4.6 Stability ratings and the time of the year Figure 38 shows that the stability rating, under which avalanches fell, is highly dependent on the time of the year. In early winter, more than 40% of the avalanches fell under V P / P stability ratings while the ratio is less than 5% for late season. The ratio of avalanches falling under G / V G conditions is lowest in early winter (8%) and highest during the late 72 season (40%). Figure 39 shows that the fraction of time V P / P stability ratings are used goes from 10% in early winter down to less than 5% in late season. \"Fai r \" is the most used stability rating during early winter and mid winter, while G / V G is the stability rating most often used during late season. Avalanches P (S i |T iAA s ) Usage P(S, TO Dec 1-Jan 31 Feb1-Mar15 Mar 16-Ap 30 Dec 1-Jan 31 Feb 1-Mar 15 Mar 16-Ap 30 Figure 38. Stability ratings under which avalanches fell during different time periods: P(Si | TJAA S). Figure 39. The usage of stability ratings during different time periods: P(Sj I Tj). The relative risk during different time periods under different stability ratings is calculated as (refer to equation 3.4 and expression 3.5 on page 43): P(AS | SiATi) oc [P(SiATi | A s ) ] / [P(SiATi)] The results are shown in Figure 40 below: 73 • I I • VP/P • F • GA/G I — i _ n ^ • Dec 1 -Jan 31 Feb 1 -Mar 15 Mar 16-Ap 30 Figure 40. Relative risk during the different time periods under different stability ratings: [P(S;ATJ | AS)]/ [P(TJAEJ)]. The relative historical risk is by far highest under VP/P stability ratings in early winter and mid winter. The difference in risk between the stability ratings is even greater during early season than mid-winter, and the lowest risk overall (when looking at stability ratings and time of the year) is during early winter under G/VG stability ratings. During late season the risk is highest under \"fair\" stability rating and very low under P/VP stability ratings compared to earlier in the season, which is very interesting. The number of avalanches the analysis is based on, is lower late season than earlier and therefore the statistical uncertainty is higher. It should be noted that by definition, P/VP stability ratings are associated with natural avalanche activity. Those kinds of conditions are most often found during and after major snowfalls, which usually happen during early winter or mid-winter, rather than late in the season. 74 c g TO > Q) CD \"D C ro -t—' o CD CL CO ro -O ro CO r-ro — o O C/3 rt :Cfl o > C c • —• > _CJ o o o c_ co rt co 5 o ,rt —i ~ CJ c o o

U O X> co o c i—i rt rt o cfl O > Is CJ cj o > rt CJ o C cc — rt cc rt — U O T 3 CS 1 2 o rt E cc O O u .rt U cfl rt rt CS ~ rt -o •— o E o CO cn CJ > \"5) JCJ rt CJ > CJ o CC C \"3 o o CD cc O CJ 13 o cc rt — co CJ \"CJ _r t rt > rt oo < < oo < oo < < < 3 c • co CN LO T— o o CM CD CO CM LO CM o £ > d d d d d d d d S O 3 CO V CO CO T f CO CO CN CD o o o CO CD £ LL 3 CO c\\i d d d d d d d 0) CD CM CO CO \"t o o o o o LO r--E co T T Tf ' CM d od CN •c CL £ CL s> 3 CO CO CD T r co CO p CO o o co CM O d d d T— d d d d Treelin GA/G LO o o CN CO CO LO CO T f CO oo LO T h o co ine CM T— T ~ d d d d d O LL 0) i-1-OT CO LO CN CO LO CM CM CD p o o o N; CD T f 0) •E CL d LO Csi T~ CM co d CM Treel o p T— co xr OT O CM co CO CM LO Alpine G/VG d d d d d CM CN o p OT CN O OT L^\" OT CM T— OT r— Aspect Z LU Z 1X1 LU CO CO CO 5 Z iJ-M 3.5 Remarks The factors analysed in section 3.3 are not independent of each other, as shown in section 3.4. However, the dataset is too small for it to be feasible to analyse the combination of more than two factors at the time. As described in this Chapter there are some sources of uncertainty associated with the recording of data, and the methods used to estimate the usage of terrain. The uncertainty should always be kept in mind while looking at the results of the analysis. When analysing one factor at the time in section 3.3 the statistical 95% confidence level is shown. The assumption is that the avalanches analysed are a random sample from a population. The population may be defined as future avalanches in C M H , and thus, the analysis might have predictive function. In reality, however, the assumption doesn't hold, since the various factors affecting the avalanche risk are not constant over time. The snow and weather patterns may change over the years, and most importantly, a part of the whole picture is human behaviour, which is probably very dynamic. The confidence interval should, however, still give some ideas about statistical uncertainty in the calculations. For the analysis on combined factors in section 3.4 the confidence interval is not shown. Therefore, those results only describe historical risk for the time period of the data. The aspect data in the combined analysis are used as they come from the database, and not merged into fewer classes, even though that would simplify the calculations and decrease the statistical uncertainty. The reason for this is that by looking at the data, different classification seems appropriate for the different combinations. For example, the aspect E may belong with SE and S aspects when looking at the relative risk in the 76 Alpine, while it belongs with N and N E aspects when looking just at the aspect, and not elevation. The usage data reflect the decisions of heli-skiing guides. It is not clear how representative the data are for the usage of other groups, such as recreational skiers. The heli-skiing guides in C M H are professional guides, and most of them have long experience in managing avalanche risk in the backcountry. Another factor affecting the data is that helicopters facilitate the movement over land in heli-skiing operations, and it is e.g. sometimes possible to go directly from one low incline run to another, or from one north facing run to another. The data might, however, give some general indications; such as that all aspects are probably not used equally by backcountry skiers. Northern aspects and lee aspects may be preferred by skiers in the northern hemisphere. However, the risk may be less dependent on aspects in the Columbia Mountains in B C than some other places. Some guides of the C M H state that they are often dealing with persistent weak layers appearing on every aspect, which they believe is characteristic for this area. The avalanche data are also affected by the risk management of guides from the time of the data. For example, the relatively low inclination of starting areas might reflect that guides often choose to ski in mellow slopes, especially when they are concerned about avalanche triggering. Therefore, the results of the risk analysis reflect the residual, historical risk after decisions had been made by guides. 77 3.6 Summary Below, the single factors from the risk analysis are listed with the factor associated with the highest relative risk first. The numbers for relative risk are obtained directly from the analysis in section 3.3. Table 4. Relative risk for single risk factors Risk factor Relative r isk V P / P 3.7 Alpine 1.4 M i d winter 1.2 N - N E - E 1.2 F 1.1 Treeline 1.1 Early winter 1.0 S E - S - S W 0.9 Sub-treeline 0.7 W - N W 0.7 Late season 0.6 G / V G c 0.4 Table 4 shows that the historical risk depends most on stability rating of the factors analysed in this research. The highest risk overall was under V P / P stability ratings and the lowest risk under G / V G stability ratings. The difference between late season and the earlier time periods is also significant, while the difference between mid winter and early winter is not. Thus, early winter and mid winter can be combined to one class which has significantly higher risk than late season. The variation in risk with elevation levels is 78 also significant. It is difficult to estimate the variation of risk with aspects. When looked at in terms of 180 degrees sectors there is not a significant difference between northern and southern aspects, but the risk is somewhat higher in eastern aspect than western. However, the risk in north, northeast, and east aspects is higher than in other aspects. The significance of the analyzed factors, in terms of risk, ranks as follows: 1. Stability ratings 2. Elevation levels 3. Time of the year (early and mid winter vs. late season) 4. Aspect When looking at combined factors, the historical risk is always highest under \"poor\" and \"very poor\" stability ratings, independent of other factors, with only two exceptions: The first one is that the relative risk is slightly higher under \"fair\" stability rating in N aspect than P /VP stability rating in S W aspect. The other exception is more remarkable: The historical risk under \"poor\" and \"very poor\" stability ratings is relatively low during late season, and considerably lower than the risk under \"fair\" stability rating any time during the season. One could assign relative risk numbers, to the different stability ratings during different time periods, based on the idea of Munter's (2003) hazard potential. If we select the number 1 for the risk during late season under \"good\" and \"very good\" stability ratings, the relative risk numbers for stability ratings and time periods look like this: 79 Table 5 . Relative \"risk numbers\" for different stability ratings during different time periods Very poor/poor Fa i r Good/very good Dec 1 - Jan 31 10 2 0.3 Feb 1 - M a r 15 8 2 1 M a r 1 6 - A p r 31 1 1.5 1 Table 5 reflects that the relative risk under \"poor\" and \"very poor\" stability ratings were very high during early winter and mid season, but not during late season. Below, the combined factors (analysed in section 3.4.) that are associated with relatively high historical risk, are listed. The numbers for relative risk are derived directly from the analysis in section 3.4. The ten combined factors associated with the overall highest relative risk are listed in Table 6. Not surprisingly, all of them include the stability rating \"very poor\" or \"poor\". In Table 7, the combined factors associated with the highest relative risk, under other conditions than V P / P stability ratings, are listed. Keep in mind that when splitting the data up like this, the uncertainty is high. 80 Table 6. Combined risk factors, associated with the highest historical risk Risk factors Relative r isk V P / P and N E 6.2 V P / P and N 5.0 V P / P and Dec 1 - Jan 31 4.8 V P / P and Sub-treeline 4.3 V P / P and W 4.2 V P / P and Feb 1 - Mar 15 3.9 V P / P and Treeline 3.8 V P / P and E 3.7 V P / P and SE 3.5 V P / P and Alpine 3.1 Table 7. Combined risk factors, excluding VP/P stability ratings, associated with the highest historical risk Risk factors Relative r isk Alpine and S 2.2 Alpine and E 2.1 Feb 1 - Mar 15 and S 2.1 Alpine and SE 2.0 Feb 1 - Mar 15 a n d N 1.9 Fair and N 1.9 Alpine and S W 1.8 Treeline and N 1.8 Dec 1 - Jan 31 and N 1.7 Alpine and Feb 1 - Mar 15 1.7 Alpine and Dec 1 - Jan 31 1.6 Fair and Alpine 1.5 Dec 1 - Jan 31 and N E 1.5 Note that combined factors including aspect may in some cases show unreasonable high risk, especially in the main aspects like N or S, due to inaccurate recording of aspects. The relative risk in aspects with few avalanches recorded, such as south aspects, is associated with high statistical uncertainty. 3.7 Datasets for comparison The criterion chosen for the avalanche dataset used for the risk analysis in this study (sections 3.3 and 3.4) was: All skier triggered avalanches greater than size 1, with the exclusion of skier controlled avalanches. It has been argued that this definition reflects some kind of acceptability, and it gives a relatively large dataset with relatively comprehensive data on avalanches. However, many of the avalanches in the dataset may not have been a big threat to skiers. Some of the smaller avalanches may have posed no risk to skiers, and some of the remotely triggered avalanches may have been in a safe distance, thanks to good terrain selection by guides. Therefore, it is interesting to compare the statistics of that dataset to other datasets extracted from Snowbase by using different criterion. 82 In order to define whether or not a difference in proportions between two datasets is statistically significant, I use a two-sample difference-of-proportions test for 711 - 7 1 2 (Barber, 1988). There are two independent random samples of size n\\ and nj from which two values of the sample proportion statistic P are determined. These two values are denoted pi and Pi-Hypotheses: Ho: Tti -%2 = Do H A : Tti - 7t2 D0 (two-tailed) Here, Do is taken to be zero, and thus, the test is used to examine the hyothesis that TCI -7T.2 = 0, or equivalently %\\ - 712 Test statistic (Barber, 1988): Pi-P2-D° (3.6) lP\\{\\-P\\)lm+P2(\\-P2)lm] Decision rule I reject H 0 i f z < -z^ or i f z > z^; otherwise I accept H 0 . When looking at 95% confidence, Zgj2 = 1.96. The datasets I am looking at are not totally independent and the samples cannot be described as random in reality, but this method still gives some ideas about statistical significance in the differences between the datasets. 3.7.1 Ava lanches where someone was caught The dataset is extracted from the table S Y S A D M A V N O T A B L E in Snowbase with the addition of paper records that date back to 1982. It contains information on avalanches 83 where one or more people were caught according to records. Being caught in an avalanche is not clearly defined, and it can mean everything from being buried by debris up to knees or taking a short ride on the surface of an avalanche, to being fully buried deep under avalanche debris. The total number of recorded avalanches where someone was caught, is 189 avalanches in the time period from 1982-2002 (with the addition of two fatal avalanches from before 1982). Below, the elevation level, aspect, time of the year and inclination of avalanches where someone was caught, are compared to accidentally skier triggered avalanches, greater than size one. The stability rating under which avalanches fell is not included since I do not have information on stability ratings from the time before the digital database. Note that some overlapping exists between the two databases. The error bars represent 95% confidence levels as in section 3.3. Elevat ion level: Let: A c = Avalanche where someone was caught (Dataset 2) A s = Accidentally skier triggered avalanche, greater than size 1 (Dataset 1) Ej = Elevation level (i = alpine, treeline, sub-treeline) Then: P(Ej | A c ) is the probability of elevation level, given an avalanche where someone was caught. 84 P(Ei | A s ) is the probability of elevation level, given an accidentally skier triggered avalanche greater than size 1. mP(Ei\\Ac) aP(E,\\A.) Alpine Treeline Sub-Treeline Figure 41. Comparison of the elevation levels of avalanches where someone was caught [P(Ej I Ac)] and the elevation levels of skier triggered avalanches greater than size 1 [P(Ej | As)]. Figure 41 shows that when comparing dataset 1 and 2, relatively fewer avalanches in dataset 2 fell in the alpine, and relatively more avalanches fell at sub-treeline elevations. However, the difference is not statistically significant for any of the elevation levels. Aspect: P(ASi | A c ) is the probability of an aspect (of starting area), given an avalanche where someone was caught (Dataset 2). P(ASj | A s ) is the probability of an aspect (of starting area), given an accidentally skier triggered avalanche greater than size 1 (Dataset 1). 85 Figure 42. Comparison of aspect of avalanches where someone was caught [P(AS; f Ac)] and the aspect of skier triggered avalanches greater than size 1 P (ASj I As)]. Figure 42 shows that the aspects of avalanches are very similar for the two datasets. It is likely that the recording of aspects, and perhaps other factors, is done more carefully for notable avalanches than for other avalanches. Avalanches in Dataset 2 are not as likely to be recorded in one of the main aspects (N,E,S, and W) on the cost of other aspects, as avalanches in Dataset 1. Most avalanches in Dataset 2 fell in the sector N - N E - E , with relatively higher number in N E than for Dataset 1. This might indicate that a part of the high number of avalanches recorded with N and E aspects for Dataset 1 actually were more to the N E . Also, a part of the avalanches recorded with S aspects in Dataset 1 may have been more to the SW. 86 Time of the year : Refer to page 55 for definition of time periods. P(Tj | A c ) is the probability of a time period, given an avalanche where someone was caught (Dataset 2). P(Tj | A s ) is the probability of a time period, given an accidentally skier triggered avalanche greater than size 1 (Dataset 1). \\P(Ti\\Ac) \\P{Ti\\A.) Dec 1 -Jan 31 Feb 1 - Mar 15 Mar 16-Ap 30 Figure 43. Comparison of the time periods when avalanches, where someone was caught, fell [P(Tj | Ac)] and the time periods when skier triggered avalanches, greater than size 1, fell [PfTj | As)]. Figure 43 shows that when comparing dataset 2 to dataset 1, relatively fewer avalanches in dataset 2 fell during early winter and relatively more avalanches fell late season. The difference is, however, not statistically significant for any of the time periods. Incl inat ion: P(Ii | A c ) is the probability of an inclination, given an avalanche where someone was caught (Dataset 2). 87 P(Ii I A s ) is the probability of an inclination, given an accidentally skier triggered avalanche greater than size 1 (Dataset 1). 0,5 -I 1 0,45 jM 15-27 28 -32 33 -37 38 -42 43 -47 >47 Figure 44. Comparison of inclination of the starting area of avalanches in Dataset 2 [P(Tj | Ac)] and inclination of the starting area of avalanches in Dataset 1 [P(Ti | A*)]. Figure 44 indicates that the inclination of starting areas in Dataset 2 is on average lower than for Dataset 1. The number of avalanches with inclination of starting areas around 30° is significandy higher for Dataset 2. The mean is 34.5° in Dataset 2, while it is 35.5° in Dataset 1. However, the median is the same for the two datasets or 35°. 3.7.2 Skier accidentals, not including remotely and sympathetically triggered avalanches A l l avalanches that are recorded as skier accidentals in Snowbase are included in this dataset (dataset 3). The difference from dataset 1 is that avalanches that are recorded as remotely or sympathetically triggered are not included, which leaves 150 avalanches in the dataset. 88 Elevation: Let: A a = Accidentally skier triggered avalanche, greater than size 1, excluding remotely and sympathetically triggered avalanches (Dataset 3) A s = Accidentally skier triggered avalanche, greater than size 1 (Dataset 1) Ei = Elevation level (i = alpine, treeline, sub-treeline) Then: P(Ei | A a ) is the probability of elevation level, given an avalanche in Dataset 3. P(Ei | A s) is the probability of elevation level, given an avalanche in Dataset 1. mP(Ei\\A.) BP(Ei\\A,) Alpine Treeline Sub-treeline Figure 45. Comparison of the elevation levels of avalanches in Dataset 3 [P(Ej | AJ] and the elevation levels of avalanches in Dataset 1 [P(Ej | AJ]. The difference between the two datasets is statistically significant with 95% confidence in the alpine and sub-treeline. 89 The relative risk is calculated as shown in section 3.4 (refer to equation 3.4 and expression 3.5 on page 43): P ( A a | E i ) « P ( E i | A a ) / P ( E i ) UP{Ei\\Aa)IP(Ei) nP(E,\\AAI P(E>) Alpine Treeline Sub-treeline Figure 46. Comparison between Dataset 3 and Dataset 1 for relative risk in the different elevation levels. Figure 46 shows that when remotely triggered avalanches are excluded, the risk for the different elevation levels is more or less similar, implying that remotely triggered avalanches are more common in the alpine than lower down. Aspect: P(ASi | A a ) is the probability of an aspect (of starting area), given an avalanche in Dataset 3. P(ASi | A s ) is the probability of an aspect (of starting area), given an avalanche in Dataset 1. 90 Figure 47. Comparison of aspects of avalanches in Dataset 3 [P(AS | AJ] and Dataset 1 [P(AS | A,)]. The relative risk is (refer to equation 3.4 and expression 3.5 on page 43): P (A a | ASi) « P(ASi | A , ) / P(AS0 • P(ASi\\Aa)/P(ASi) M P(ASi\\A.)lP(ASi) N-NE-E SE-S-SW W-NW Figure 48. Comparison between Dataset 3 and Dataset 1 on relative risk in the different aspects. 91 Figure 48 indicates that when excluding remotely and sympathetically triggered avalanches, the risk is lower in southern aspects than for Dataset 1, and slightly higher in other aspects. However, the difference is not statistically significant. T ime of the year : Refer to page 55 for definition of time periods. P(T i | A a ) is the probability of a time period, given an avalanche in Dataset 3. P(Ti | A s ) is the probability of a time period, given an avalanche in Dataset 1. *P(Ti\\A.) MP(Ti\\As) Dec 1 -Jan 31 Feb 1 - Mar 15 Mar 16-Ap 30 Figure 49. Comparison of the time periods of avalanches in Dataset 3 [P(T f A„)] and avalanches in Dataset 1 [P(T IAJ] The difference between Dataset 1 and 3 is statistically significant for early winter, but not the other periods. The relative risk in Figure 49 is defined as (refer to equation 3.4 and expression 3.5 on page 43): P f A a l T O o c P C T i l A a H P C T i ) 92 UP(Ti\\Aa)l P(Ti) UP(Ti\\As)/P(Ti) Dec1-Jan31 Feb 1-Mar 15 Mar16-Ap30 Figure 50. Comparison between Dataset 3 and Dataset 1 on relative risk during different time periods. Figure 49 shows that remotely triggered avalanches are relatively more common early winter than later on. Figure 50 shows that when remotely triggered avalanches are excluded the relative risk of triggering an avalanche greater than size one is similar for early winter and late season, and almost two times higher during mid winter. Stabi l i ty rat ings: Let P(Si | A„) be the probability of a stability rating, given an avalanche in Dataset 3. Let P(Sj | A s ) be the probability of a stability rating, given an avalanche in Dataset 1. 93 \\P(S,\\Aa) lap(S,\\As) VP/P G/VG Figure 51. Comparison of stability ratings under which avalanches in Dataset 3 [P(Sj I Aa)] and Dataset 1 [P(Sj I As)] fell. The relative risk in Figure 51 is defined as (refer to equation 3.4 and expression 3.5 on page 43): P(A a |s i)«P(S i |A a)/P(Si) 5 4,5 4 3,5 3 2,5 2 1,5 1 0,5 0 • P{Si\\Aa)IP(Si) • P(Si\\A.)IP(Si) VP/P G/VG Figure 52. Comparison between Dataset 3 and Dataset 1 on relative risk under different stability ratings. 94 Relatively higher number of avalanches in Dataset 3 than Dataset 1 fell under G / V G stability ratings and lower number under other stability ratings (Figure 51). It shows that remotely skier triggered avalanches are relatively more common under VP/P stability ratings and less common under G / V G stability ratings than other skier triggered avalanches, which is not surprising. Still, the difference in relative risk between stability ratings is high for Dataset 3, with the highest risk being under VP/P stability ratings and lowest under G/VG stability ratings (Figure 52), even though the difference is slightly smaller than for Dataset 1. The differences between the datasets are not statistically significant. Inclination: Let P(Ii | A a ) be the probability of an inclination, given an avalanche in Dataset 3. Let P(Ii | A s ) be the probability of an inclination, given an avalanche in Dataset 1. 0,6 - i 0,5 0,3 0,4 0,2 -• P(Ii\\As) 0,1 -0 -15-27 28-32 33-37 38-42 43-47 >47 Figure 53. Comparison of inclination of starting areas for avalanches in Dataset 3 [P(Ij I Aa)] and Dataset 1 [P(Ii | As)]. 95 Figure 53 indicates that the inclination of starting areas is, on average, slightly lower when remotely and sympathetically triggered avalanches are excluded. This is expected since remotely triggered avalanches often propagate to steeper areas than the skiers would be skiing on. The difference between the two datasets is, however, not statistically significant, and it is lower than the difference between Dataset 1 and Dataset 2. The median is the same for all datasets; 35 degrees. 3.8 Summary and conclusions The components of risk defined in section 2.2 are: probability, vulnerability and exposure. By focusing on all accidentally skier triggered avalanches greater than size one we are only looking at probability and minimum magnitude. It is not clear how this relates to risk of being buried in an avalanche, risk of injuries or risk of death. The vulnerability of people rises on average with bigger avalanches, and the exposure is probably on average lower for remotely triggered avalanches than skier triggered avalanches that are not remote. Therefore, it was of interest to compare datasets 2 and 3 defined in section 3.5 to the main dataset. The analyses show that there is less difference in historical risk between elevation levels in Datasets 2 and 3 than in Dataset 1. Perhaps it indicates that the historical risk, when taking all the components of it into account, is not as dependent on elevation levels as suggested by Dataset 1 alone. However, part of the risk in the upper elevation levels in Dataset 1 is due to remotely triggered avalanches. Some of them might have posed no risk to the skiers, but many of them probably did. The average exposure for remotely triggered avalanches is probably lower than for other skier triggered avalanches, however that does not necessarily indicate that the risk of injury of death is on average lower for 96 remotely triggered avalanches. Some of the most dangerous avalanches a group of skiers can encounter are large, alpine, remotely triggered avalanches that start above the group, and have gained speed and momentum before hitting the people The aspects of avalanches in Dataset 2 and Dataset 3 are very similar to the aspects in Dataset 1, with the majority of avalanches in the N - N E - E sector. Avalanches in datasets 2 and 3 are slightly more l ikely to happen during late season and less l ikely to happen in early winter than avalanches in Dataset 1. Dataset 2 and 3 include higher ratio of avalanches with starting area in 30 degrees and lower ratio of 40 degrees than Dataset 1. This difference is greater between Datasets 1 and 2 than Datasets 1 and 3. However the median inclination is 35 degrees for all three datasets. 97 Chapter 4. The decis ion making process of heli-skiing guides in terms of avalanche risk 4.1 Introduction The community o f professional heli-skiing guides holds a great amount o f experience and knowledge in managing backcountry avalanche risk. Many of the guides spend a few months in the field every year, and during their day of work, they cover a large area of terrain. The first objective of this chapter is to extract some of the knowledge held by professional heli-skiing guides and write it down in a structured way. The main focus is on terrain selection and group management. The second objective is to discuss the possibility of a rule base approach to decision making of guides. In order to do that I need to formulate an idea about how guides make decisions, and how experience is used in the process. 4.2 Data and Methods During February, March and Apr i l , 2003,1 spent a total of 17 days visiting 7 of the Canadian Mountain Holidays ( C M H ) heli-skiing operation areas in the Columbia Mountains in British Columbia, Canada. In August, 2003,1 visited one more operation area. During these fieldtrips, I asked the guides in the areas to complete a questionnaire, I did in-depth interviews with 1 or 2 experienced guides in each area, and I observed the working day of the guides. 4.2.1 Questionnaire Guides from 9 different operation areas completed the questionnaire, which was answered by 40 guides in total. I handed the questionnaire to the guides, and answered any questions they had 98 regarding it, in eight operation areas, and my supervisor D. McClung handed the questionnaire out in one area. The questionnaire had two goals: 1) To examine which factors in terms of terrain selection and group management, are most important to guides, and 2) under which stability rating do different factors start to become significant. The questionnaire is found in Appendix C and the raw results from it are displayed in Appendix D. The advantage of questionnaires is that the answers are given in a structured way, and they are relatively easy to analyse. The analyses do not involve much subjectivity. A disadvantage of the questionnaire in this case, is that factors are taken out of context of the situation at hand, and many guides complained about that. In reality, it is likely that the importance of e.g. slope inclination in the decision making process, depends on the situation that day, in terms of weather, snow structure and human factors as well as the stability rating. The environment, in which the heli-skiing guides make their decisions, can be defined as natural settings (see page 15). Some researches have suggested (e.g. Klein, 1998) that experts under natural settings do not make decisions in an analytical way. Therefore, it may be difficult for guides to think in an analytical way, when answering a questionnaire. 4.2.2 Interviews In order to include information that cannot be contained in a structured questionnaire, in-depth interviews were conducted with a total of 10 experienced guides. The assumption was that guides may not be able to explain in an analytical way, how they make decisions, and which are the most important factors influencing their decisions. Therefore, I decided to put the guides in a situation where they imagine a realistic setting, and base their answers on that. The method I used was to interview the guide in front of the computer of the lodge, while looking at run 99 pictures. Every lodge has digital run photos of many of their runs, which are often used as a base of discussion during guides' meetings. Sometimes we would look at runs that we had skied that very same day and, in other cases, I asked the guide to pick some runs that are rated \"green\" but are not straightforward in terms of terrain selection. For example, we looked at runs where it is possible to make some mistakes and get into trouble, even when rated \"green\". I asked the guides to explain how they would travel from the top to the bottom of the run, and what they would be thinking on the way, in terms of terrain selection and group management. I also asked the guide to explain the runlist of the day; what makes runs red and what makes them green. Furthermore, I asked all the guides the same three questions: 1 . How long have you been a guide? 2. Why did you decide to become a ski-guide? 3. Do you think Munter's rule based decision making methods could be applied to professional heli-skiing guides in Canada? The main advantage of this interview method is that, opposed to the questionnaire, things are kept in context. Also, the guides are used to discussing terrain selection while looking at run pictures in the computer of the lodge. The disadvantages however, are that it is difficult to interpret and analyze the results of such loosely structured interviews, and the researcher will unavoidably influence the results. No two people analyzing the data might get exactly the same results. Another drawback is that the interviews are not independent of the conditions at the time they were taken. The season of 2003 was characterized by a shallow snowpack, many persistent weak layers, deep instability and, in general, very low confidence in the snow stability. The interviews were tape recorded with the permission of each guide interviewed, and then typed on computer. After typing the interviews, I divided them into different subjects for 100 analysis. In this thesis, I have sometimes taken the words of individual guides and repeated them verbally. In those cases, it is important to remember that spoken word often differs from that printed. A l l interviews are kept anonymous. 4.2.3 Observat ion During the fieldtrips, I was present at morning- and evening meetings of the guides every day. Whenever there was a space in the helicopter, I went out skiing with a guide and his or her group. The main focus was on terrain selection and group management. When skiing I tried to have as little effect on the group and the guide as possible, and simply observe the guiding. However, I was one of the group, and the guide and the guests knew that I was doing an observation. Therefore the observation lies somewhere between a non-participant observation and participant observation as defined by Sarantakos (1998). During the visits I kept a diary of my observations. It was necessary for me, as an outsider researching decisions in terms of avalanche risk management, to observe guiding in action. It gave me knowledge and overview necessary to work on this project. One of the advantages of observations according to Mahr (1995) is that it approaches reality in its natural structure, and studies events as they evolve. It also offers first-hand information without relying on the reports of others. However, like with the interviews, the results of such observation will always be coloured by the researcher. It is exposed to the observer's bias, selective perception and selective memory (Sarantakos, 1998). Examples of further limitations according to Sarantakos (1998) are that observation cannot provide information about past, future or unpredictable events, and it cannot offer data related to frequency of behaviour. In participant observation the observer is a part of the situation that is 101 being observed, and it offers no control measures regarding bias, attitudes and opinions of the observer. Furthermore, observation cannot offer quantitative generalisations on the results. As stated before, avalanche hazard was a major concern for the guides during my fieldtrips, which made the observations especially interesting. 4.3 Results and conclusions from Questionnaire The raw results from the questionnaire are displayed in Appendix D 4.3.1 Persona l information O f the guides who answered the questionnaire, 90% are males, and most were between 35 and 50 years of age, and have substantial amount o f experience. The average number o f years the guides in the questionnaire had been working as guides is 12. A l l the guides at C M H have the minimum education of assistant guide and 80% of the guides in the questionnaire were either ski guides or mountain guides. The answers in the questionnaire did not vary much with the age and experience of guides. 4.3.2 Terrain select ion The guides were asked to rate the importance of four different terrain factors for the decision making process with relation to the stability rating, and whether the decisions are made during morning meetings or skiing. The four factors are: 1) general shape of terrain, 2) inclination, 3) aspect and 4) elevation level. Furthermore, a few more questions about these factors were included. According to the questiomiaire, the general shape of the terrain is more important in the process of terrain selection than any of the other factors. This applies under all stability ratings and both during morning meetings and skiing. The guides were also asked to rate the importance 102 of five different small scale terrain features: 1. convex rolls 2. windloaded depressions 3. gullies 4. small bowls and 5. the area around cliffs, under different stability ratings. Refer to Appendix D on methods to estimate the overall ranking of importance. Convex rolls were considered the most hazardous feature on average under all stability ratings, and windloaded depressions were rated number two. Gullies were rated number three on average under all stability ratings, however, their importance increases with poorer stability ratings. Areas around cliffs were rated number 4 under \"good\" stability ratings, but number 5 under \"fair\" and \"poor\" stability ratings. The rating of small bowls is the other way around. The importance of small bowls increases with poorer stability ratings. Inclination was rated the second most important terrain factor both during morning meetings and skiing. It was rated important under all stability ratings, however, the level of importance rises significantly between the stability rating \"good\" and \"fair\", while there is not a big difference in the importance of inclination between \"fair\" and \"poor\". O f 40 guides, 22 said that they are constantly aware of the inclination of the terrain they are skiing, and 32 guides said that they first and foremost think about the shape of the terrain, rather than inclination directly. Thus, it may be concluded that the general shape of the terrain is more important than any single terrain factor, however, guides often keep a constant idea about the inclination of the terrain they are skiing. 103 The importance aspect was on average rated number three out of the four terrain factors. In question two the aspect was divided into: large scale aspect, that is the main aspect of a ski-run as a whole, and small scale aspect, that is the aspect of individual slopes within a ski-run. The importance of large scale aspect was rated on average higher than the importance of small scale aspect. Unl ike inclination, the importance of aspect increased gradually between \"good\", \"fair\" and \"poor\" stability ratings. On average, the importance of aspect was rated slightly higher for the decision process during morning meetings, than during skiing. The importance of elevation levels was on average ranked no. 4 out of the 4 factors. Elevation levels were however rated more important than aspect in \"fair\" and \"poor\" stability ratings during skiing. In summary, the general shape of the terrain was considered the most important terrain factor and convex rolls the most hazardous small scale terrain features. Inclination was the second most important terrain factor, and aspect and elevation levels number three and four. 4.3.3 Group Management The guides were asked a few questions on instructions to guests, regrouping spots, spacing between skiers and whether or not stricter group management would reduce the avalanche risk in the operation. The majority of the guides, or 82%, checked the statement: \"Most clients follow my instructions but once in a while I have problems with individual guests in this regard\". The majority, or 56% said that they think that clients not following orders have very much effect on the avalanche risk in their operation as a whole, while 26% thought it has some effect, and 18% didn't think it has much effect. Thus, most guides believe that clients not fol lowing instructions have at least some effect on the avalanche risk in the company as a whole. 104 Of the respondents 38% checked the statement: \"I always choose the regrouping spots very carefully. I am constantly trying to find a spot that can not be triggered and is not threatened by avalanches from above, no matter how the snow stability is that day\". A higher number, or 62%, checked the statement \"It depends on the stability how carefully I choose the regrouping spots\". Out of those, 16 guides, or 70%, said that they started worrying significantly about regrouping spots under \"fair\" stability ratings, while 6 guides, or 26%, said that \"poor\" stability rating was the threshold. A majority, or 64% of the respondents said they believed regrouping spots are very important when thinking about the overall avalanche risk in their operations, and 33% said that it was somewhat important. Thus, the importance of regrouping spots was on average rated higher than the importance of clients not fol lowing instructions. O f 37 guides, 8 (22%) checked the statement: \"I always tell my skiers to keep proper distance from each other (at least 2 or 3 turns), no matter how the snow conditions are\" (that is when skiing downhill). O f 39 guides, 10 (26%) checked the statement: \"When traversing steep slopes I only let one skier traverse at a time, no matter how the stability rating is\" . The rest of the guides checked: \"It depends on the stability whether I control the distance between my skiers\". L i both cases, the majority of guides or 80%, checked \"fair\" when asked under which conditions they would use spacing. When asked how important they believed spacing between skiers is, when thinking about the overall avalanche risk in their operation, 31% checked \"very important\", 53% checked \"somewhat important\" and 16% checked \"not very important\". Thus, the importance of spacing between skiers was on average rated lower than for regrouping spots, or clients not following instructions. 105 The result from the last question in the questionnaire (see Appendix D) indicates that the guides have very different opinions on whether or not stricter group management would lower the avalanche risk in their company. In summary, the respondents of the questionnaire believe that out of the three factors, the selection of regrouping spots has the most effect on avalanche risk in the company, clients not following instruction the second most effect and the spacing between skiers was number three. The guides were, however, not asked to rate those factors relative to each other. It depends on the conditions how important regrouping spots are in the decision making, and whether or not guides think about spacing their skiers apart. According to the questionnaire, \" fair\" stability rating is in general the threshold when guides start worrying significantly about the selection of regrouping spots and spacing. 4.4 Analysis of remarks in reports of avalanches where someone was caught (Dataset 2) If an avalanche is recorded as \"notable\" in Snowbase, the guide involved makes a special report. The reports often contain a short description of circumstances, e.g. on the terrain where the avalanche fell and human factors leading to the avalanche. These remarks contain valuable information that is not found for other avalanches in the database. In this study, I looked at all \"notable\" avalanches for which it was recorded that someone was caught. Refer to page 35 and 83 for further description of the dataset (Dataset 2). The total number of recorded avalanches where someone was caught, is 189 avalanches in the time period from 1982-2002. I read the remarks in the avalanche reports to identify the most common factors mentioned in the remarks, in terms of terrain and human factors. The results are presented in this section. 106 4.4.1 Remarks on terrain Of 189 reports in total, 87 reports (46%) contain no information at all on terrain. The amount of information in the remaining reports varies. The most frequently mentioned terrain features are listed below, with the number of times the feature is mentioned listed in parentheses: • Convex rolls (32) are by far the most common terrain feature mentioned in the remarks. • Steep slopes (18). O f those, 8 were recorded as short, steep slopes, and 4 recorded as steep, windloaded slopes. • Windloaded pockets (6). • Trees were mentioned in one way or another in 28 reports. In 17 of those, the debris piled up against trees. Six times the victims held on to trees in order to save themselves. In four cases it was reported that avalanche victims were injured by hitting trees. Other terrain features were mentioned less often. Remarks Since the analysis is based on text in the reports, the words chosen to describe terrain depend on the guide who writes the report. In some cases a drawing is attached to the report. I did not attempt a deeper interpretation of terrain description in text or analysis of the drawings. Convex rolls were simply by far the most mentioned terrain feature in the reports, indicating that those are terrain features that are easily defined and often cause problems. Small scale terrain features are mentioned more often than large scale terrain features. 107 4.4.2 Human factors Human factors include information on decisions or behaviour that led to the notable avalanche. Some human factors from the reports are listed here, and the number of reports mentioning each factor is listed in parentheses. In addition, some results of statistical analysis of information about triggers and victims are presented. Decisions and behaviour: • Run had been skied before (41). Out of those, it was stated in 21 reports that the run had been skied the same day, and in nine reports that the run had been skied in the days before. For the reminder, it was not clear when the run had been skied. • Run or line had been skicut or checked by a guide (22). • The slope had already slid (7). • Clients did not do what they were expected to do (33). In six reports, the guide stated that the client was intentionally not fol lowing instructions. In five reports, the reason given was that a client could not follow the instruction, because he or she lost control of the skiing. In four reports, the reason given was that a client did not respond, in the intended way, to a warning from the guide. • Avalanche incident in relation to regrouping spots (16). In these cases, either the avalanche was triggered from a regrouping spot, or it caught people in regrouping spots. L i ten cases the regrouping spot was selected by the guide, but in other six cases, the clients regrouped in places not selected by the guide. • Skiers fell or took off skis (8). Either the fall o f the skier or a skier who had taken off his or her skis after a fall triggered the avalanche. 108 • Skiers were probably trying to find untracked snow (13). It may be assumed from the reports that the avalanche triggering were the results of skiers trying to find fresh snow. However, this is not always explicitly stated and therefore this is my interpretation of the text. • The avalanche was triggered while traversing a slope (13). Remarks: Some factors are more likely than others to be mentioned in reports. Therefore, such a summary is not necessarily a true representation of the most common factors; however it still gives some valuable information. The guides often mentioned that the run had been skied before, checked by a guide, or had already slid before. These are probably factors that are likely to be mentioned in the remarks when they happen, because the guides are justifying their decisions, and expressing surprise over the incident. The high number of these cases may also indicate too high confidence in runs that have already been skied or checked by guides. When avalanches are triggered by clients, who are not where the guide intended them to be, it is probably also likely to be mentioned in the guide's report. The guide could be suggesting that it was not only a wrong decision by himself or herself that lead to the incident. The frequency of such incidents, emphasizes how important group management is for the avalanche risk management. Avalanches triggered from, or affecting regrouping spots may not be mentioned in order to justify guide's decisions, especially not when the spots had been selected by the guide. 109 However, regrouping spots are a clearly definable factor in guiding, and they may therefore be likely to be mentioned. The impact of a falling skier can trigger an avalanche, where normal skiing may not. Such an incident is probably l ikely to be reported. When people go heli-skiing, they are expecting to ski untracked snow most of the time. It is also often easier than skiing in tracks in the deep snow. Therefore, there is a tendency for skiers to hunt the new snow, which might increase the avalanche risk in two ways: 1) it might lead people to more dangerous places, after the best line has been tracked up, and 2) since it causes people to spread out over terrain, it is more l ikely that someone hits a sweet spot or imperfection (see e.g. sections 2.3 and 2.8). For analysis, the information has to be broken into basic elements or factors. Some basic elements have developed among guides, as reflected in the report text. Examples include \"regrouping spots\" or \"traversing a slope\". The elements that guides most often talk about are most l ikely to be mentioned in reports, which probably also reflects their importance. However, when text is broken down into factors, the result is always affected by the interpreter, and his or her motives. Other people might perceive the text in the reports differently with changes in ranking the factors. Statistics on triggers and vict ims: The ranking of the trigger in the group In 89 reports it is mentioned where the trigger was in the group. Out o f those: • 20% of the avalanches were triggered by a guide. • 8% were triggered by first skier other than guide. 110 • 42% were triggered by 2nd to 4th skier. • 30% were triggered by one or more of the skiers behind the 4 t h skier. The avalanches triggered by first skier other than a guide, were triggered by skiers who either had passed the guide, or were not skiing the guide's line. The total group size is reported in about 90% of the reports. The most common group size is 12 people, while the average is 10 people, including guide. These historical statistics indicate that the probability of a guide triggering an avalanche is higher than for any other person in the group. One of the first four skiers in the group is more likely to trigger an avalanche than the skiers behind them, or on average more than two times more likely. However, about 1/3 of the avalanches were triggered by someone who came after the first four skiers, showing that the risk is not solely associated with the first skiers. More than one person caught In 46% of the avalanches where someone was caught, it was more than one person who got caught. One or more people triggering For 125 reports, indications on whether one or more skiers triggered the avalanche are available. Out of those, 82 (66%) avalanches were probably triggered by only one skier, while 43 (34%) were probably triggered by more than one skier. These numbers involve some subjectivity in the interpretation of text in reports. I l l Trigger caught For 137 of the reports, indications are given about whether the person who triggered the avalanche got caught, and in 87 (63%) of those avalanches the trigger was caught. 4.5 Results and conclusions from interviews and observation This section contains the combined results and conclusions from the in depth interviews and observation. The results are also compared to results from questionnaire and analyses of avalanche reports. The focus is on terrain selection and group management. 4.5.1 Characteristics of heli-skiing guides at CMH When asked in the interviews about the reasons for becoming ski-guides, the first answer was usually the love for the mountains and the great outdoors. The second and third reasons were the enjoyment of being with people and sharing the experience of the outdoors with them, and the enjoyment of skiing. As one of the guides said: \"A bad day of skiing is better than a good day at the office \" Heli-ski guiding is not only hard work, it is a lifestyle, and it is not enough to find skiing fun to be a good guide; love for the mountains is an essential part. 4.5.2 Terrain selection in terms of avalanche risk Size and shape of the terrain Interviews and observations gave similar indications to the questionnaire; the shape of the terrain as well as the size of it, are the factors first considered by guides in any terrain selection. The terrain factors that most often made runs \" red\" (on the day of the interview), according to my interviewees, were large, alpine terrain features. For example, i f a part of a run 112 was a big, open alpine bowl, the run would most likely be red. Also, any unavoidable exposure to big terrain is likely to result in a red run. In general, guides first think about potential consequences of something going wrong in an area. Therefore, the first concern is usually whether the terrain (size and shape) is capable of producing large avalanches. In such terrain, the decisions of guides might be different and more conservative than in terrain that is not considered capable of producing very large avalanches, such as broken or forested terrain. This applies, even when other conditions, such as inclination, aspect and snowpack characteristics, are equal. A large avalanche can put people at greater risk of death than a small one, and it also puts more people at risk. Therefore, this dependence in the decision making on the size and shape of the terrain is an important part of avalanche risk management for backcountry skiing operations. Terrain can be described as fractal geometry, which implies it is not only inclination and aspect that defines the capability of terrain to produce avalanches. An experienced guide develops a feeling for the type of terrain, in terms of the shape along with other factors, which is most likely to produce avalanches, either natural or human triggered. Terrain features For convenience and simplification, terrain is often classified into different terrain features. Some terrain features are often mentioned among guides: • Convex rolls were the single terrain feature most often mentioned as a matter of concern in the interviews, which is in lines with both the questionnaire and avalanche reports. Convex rolls are small scale terrain features, which are relatively easily triggered due to gradient in slope angle with ease in producing shear fractures. 113 Windloaded depressions are of importance according to questionnaire and avalanche reports, however, these terrain features were not discussed a lot in the interviews. Windloaded depressions are features where drifting snow has accumulated. If there is a weak layer under the drifted snow, these pockets can be touchy. Moraine features are generally avoided, especially when the stability is a concern. According to interviews and observation, guides working in glaciated areas are always concerned with moraines and they select their routes accordingly. Moraine features often have thin snowpacks which makes them prone to skier triggering. They are known for their odd behaviour, and are often triggered remotely. Cutblocks are always of concern for guides working in logged areas, according to interviews and observation. Cutblocks often grow surface hoar wel l , because of the relatively low elevations, and lack of trees and wind. A lso, they lack the anchoring of the surrounding forest and, in general, the snow conditions in a cutblock may be very different from the surrounding areas. Gullies are not one of the most easily triggered features, according to some of the guides I interviewed, because they have support from below. However, i f the stability is poor enough for them to be triggered, the consequences may be serious, since gullies often turn out to be a \"terrain trap\" when the snow piles up in the bottom. This is in lines with the results from the questionnaire, were gullies were rated more important with poorer stability ratings. 114 Glades and chutes in the trees are features of interest to guides according to interviews and observation, especially in steep terrain. They are similar to cutblocks but often smaller in scale. Shallow spots in the snowcover are potential triggering spots. A n important part of avalanche risk management is to try to recognize where the snowpack might be shallow, especially i f there are deep, persistent weak layers around. Examples of potential shallow spots are on convex rolls and moraines, and rocks sticking out o f the snow might indicate such areas. However, it is sometimes difficult, even for experienced guides, to estimate how the thickness of the snowcover varies with the terrain. Trees and forested areas are treated differently from anything else. A t C M H , a lot of the skiing takes place in the trees, and in Brit ish Columbia there is perhaps more tree-skiing going on than in many other places in the world, e.g. in Europe. Trees serve as anchors for snow slabs, and, where the forest is thick enough, the risk of triggering an avalanche is usually considered minimal. Furthermore, avalanches triggered in the forest are, in general, smaller than avalanches in open areas. For these reasons, the forest is often considered a shelter from avalanche risk. However, should an avalanche release in or above forested areas, the trees may cause injuries to people caught in even small slides. A large percentage of avalanche deaths are caused by trauma due to people striking objects. 115 History of the terrain Within individual heli-skiing operations, accumulative knowledge exists on the avalanche history of individual runs. Guides in the area have the knowledge in their minds, but some of it is kept in Snowbase. Here, short term history of terrain is defined as the history of avalanche activity after the most recent snowfall, or significant weather changes. Experienced guides within an operation area can estimate which slopes are likely to slide during natural avalanche cycles. After a slope has slid naturally it may be considered safe to ski, even i f it is relatively steep. The runlists in the mornings were often affected by these factors during my visits, and a steep run in big terrain might even be green on an otherwise conservative runlist, i f the guides are confident that all of the critical slope(s) already slid. The same applies to smaller scale terrain features considered during skiing; an otherwise hazardous feature may be considered safe for skiing i f it already slid. The guides, however, often discuss whether everything has slid, or i f there are still some dangerous pockets hanging around. A lso, there might be a discussion on whether or not the new snow on the bedsurface of an avalanche could slide and cause any harm. During periods of natural avalanche activity and stability concerns, the short term history is an important factor in the decision making process. Long term history of terrain is defined as the avalanche history of individual slopes over the years of operation. Some runs contain terrain features that are known \"repeaters\", which means that they are known to avalanche often. It is not always known why certain places tend to slide so often, only that they do. This knowledge of avalanche history plays a very important role in the decision making process of guides, and I believe that, on average, it is the second most important factor in the terrain selection, after the shape and size of the terrain. 116 Slope incl inat ion The factor most often mentioned in the interviews, as characterising green runs was low inclination. I found that the guides I interviewed always had an idea about the slope angle when they went down the runs in their mind, and all o f them talked a lot about inclination. I did however not verify whether or not their ideas about inclination were correct. Some guides would talk about slope angles, usually with 5 degrees intervals such as 30, 35, 40. Some guides only used terms such as \"mel low\", \"gentle\", \"steep\", and \"very steep\". When talking about degrees, everything below 30 degrees was generally considered mellow, inclination between 30 and 35 degrees was starting to be of concern, and anything above 35 degrees was considered steep. These definitions may reflect the situation at the time of the interviews. I believe that inclination is one of the most important factors in the decision making process of guides, but it is generally considered in context of the overall shape of the terrain. If the terrain is of very low inclination (the limit may be around 30 degrees) it may often be assumed that the risk of triggering an avalanche is negligible and no further steps are needed to reduce avalanche risk. However, as soon as the inclination approaches levels of concern, a combination of other factors is considered. Aspect Aspect was not often mentioned in the interviews, which might reflect that aspect was not a very important factor for decision making at the time the interviews were conducted. In general, aspect becomes increasingly important late in the season when solar radiation starts to 117 have effect. A lso , aspect was mentioned in relation to places where surface hoar tends to grow and persist. Elevation levels Elevation levels were not mentioned often directly during the interviews, unless when talking about the quality of skiing rather than avalanche risk. However, the variation of the forest-cover with elevation seems to be very important in the decision process, as is the relation between surface hoar growth and persistence, and elevation levels. R u n has been skied before How often and when a run has been skied before affects the terrain selection. One of the most common characteristics of \"green\" runs at the time, according to my interviewees, was that it had been heavily skied throughout the season. These heavily used runs are more l ikely to be used on a given day, than other runs. There are a few possible reasons for this: • The more the guides ski a certain run, the more relevant information they have on the run. Thus, the level of uncertainty is lower on a heavily skied run, than on a run that has not been skied a lot throughout the winter. Guides often talk about the confidence they have in ski-runs and the snowpack. Low confidence often indicates high uncertainty and that affects decisions very much. • Some of my interviewees mentioned that skier compaction can be a factor. A few runs are skied so intensively that weak layers are believed to be destroyed as soon as they form. Sti l l , many guides are hesitant about relying on the effect of skier compaction, except under special circumstances. One guide said: \"The different guides have different 118 philosophy about it [skier compaction]... For me skier compaction meaning that there is a weak layer form at the surface, and we ski in the weak layer and then it snows on top, I treat a slope like that as if it has not been skied. Numerous times slopes release that I know had been skied with a weak layer. The second where I do feel comfortable with skier compaction is on a steep slope, where the weak layer was formed, and then we got some amount of snow, 10 or 20 cm of new snow on it. And then we might ski the line and it feels like as you put turns in the slope you are affecting quite a wide area because the snow sluffs down under you, and you are pushing the snow and you are really disturbing the weak layer. ... The groups would start from one side and ski it completely. \" Another guide talked about being unable to use skier compaction in the 2002-2003 season to stabilize runs, because the snowpack was very facetted, and \"tracks in recrystallized snow don't make much of a difference \". Overall, guides have different theories about skier compaction, and many of them are very cautious about putting trust on the stabilizing effects of it. • Perhaps a psychological factor is also a part of the picture. Guides may feel better on runs simply because they have been heavily skied. A related question is whether the evidence that a run has been skied before the same day or the days before, affects terrain selection. This is a factor often mentioned in remarks with reports for notable avalanches (refer to page 108), possibly because it surprised the guide that there was an avalanche incident in a slope that had been skied shortly before. Do guides sometimes use previous skiing as an indicator of stability? One guide said: \"..and those tracks, they start to make you think that it is stable because the whole group went down there, it must be alright. 119 That, in my mind, is stability evaluation by group loading, you now, you put a group on it and it didn 't go, so it is okay -1 don't think so. Let's go back to the pit data here. What is the snowpack structure saying\"? 4.5.3 Summary on terrain selection Both during morning meetings and skiing, the decision process of guides in terms of terrain selection is focused first on the size and the shape of the terrain. During morning meetings the emphasis is on large scale terrain features, while during skiing, every small terrain feature may get attention. In general, (at least when the stability is a major concern), the first concern is whether the terrain is capable of producing large avalanches, and whether and, for how long, the skiers are exposed to such terrain. For example, big, open terrain in the alpine lead to more conservative decisions than broken or forested terrain, everything else being equal. Some defined terrain features are often mentioned when guides are discussing terrain. The small scale terrain features, which get the most attention, are convex rolls. Examples of other small scale features often mentioned are: moraines, cutblocks, gullies, chutes and glades, windloaded depressions and shallow spots. The avalanche history of the terrain is also a very important factor in the terrain selection process o f heli-skiing guides, both the history o f avalanche releases after the latest storm, as wel l as the avalanche history of the slopes over the years. Most guides probably have a constant idea about the inclination of the slopes they are skiing. M y assumption is that during the time of the interviews and observation of this study, the shape and size of the terrain, the avalanche history and inclination were by far the most important factors in the terrain selection of guides in terms of avalanche risk management. 120 Furthermore, these factors are always in context with each others. Other factors such as aspect and elevation were not as important at the time, except when evaluating skiing quality. The most common factors characterising the \"green\" runs of the day, according to interviews, were: Low inclination, heavy usage through the season, the run already slid, and some guides talked about safety margins within the runs. The most common characteristics of red runs were big terrain and unavoidable exposure to hazardous terrain. 4.5.4 Group management in terms of ava lanche risk Instructions from guide to guests It is quite a common statement in the remarks with notable avalanches, that the client who triggered the avalanche was not located where the guide intended him or her to be (see page 108). Perhaps guides are l ikely to make that remark every time they have a reason to, since it provides an explanation for the incident and takes some of the guilt away from the guide. The question is whether clients not following instructions of a guide, add significantly to the avalanche risk in backcountry skiing operations like C M H . If there is a significant amount of incidents that are at least partly due to a client who is somehow not located were the guide intended him to be, it may not be sufficient to focus on avalanche forecasting and terrain selection in order to lower avalanche risk; the attention has to be on group management as well . Many of the guides I interviewed emphasised the importance of instructions to the guests. The guides agreed on that instructions need to be simple, and the guide has to face the guests and be loud and clear and make sure everybody understands the most important factors. If instructions are too detailed or complicated, the guide risks losing the attention of the guests or that the guests won't understand the most important factors. One guide said it was important always to explain why the instructions are given, because i f people don't understand the purpose 121 of the instructions, they might ignore them. However, not all guides emphasize this and I observed a considerable difference between individual guides in the nature and amount of instructions given to clients. I l iked to have more rather than less instructions, and I believe that inexperienced heli-skiers prefer quite detailed instructions, while the more experienced guests only want the most important instructions. One of the interviewees said: \"There are two ways of guiding: It is guiding with instructions and guiding by leading\". This is one way to look at the different guiding styles, and most guides use a combination of both ways. When a guide uses speed to increase spacing between skiers, he or she is guiding by leading. It may be assumed from the interviews that guiding by leading is in general used more when the skiers in the group have greater experience. In general, the approaches of guides to giving instructions are as different as the guides and different strategies work for different personalities. Regrouping spots According to interviews and observation, the selection of regrouping spots is an important and distinct factor in the group management process. Regrouping seems to be a relatively simple process in the mind of experienced ski-guides such as my interviewees. People tend to ski the fall line, and the key is to regroup every time there is a change in the direction o f the route relative to the fall line, and also every time before a hazardous feature, such as a crevasse or a moraine, is passed. In addition to the terrain factors, the behaviour and the skiing ability of the group also determines the regrouping spots. When the group is strong and following the guide well , it is not necessary to regroup as often as with a group of weak skiers or skiers that are not behaving well . Experience is the key to selecting regrouping spots, and that includes both 122 experience in terrain evaluation, and experience in group management. The guide typically uses the same basic regrouping spots every time he takes a group down a certain line in a run, and those are determined by the terrain. Then he or she might want to regroup more often based on the skiing ability and behaviour of the group. Two \"types\" of avalanche risk are managed with the selection of regrouping spots: 1. Avalanche risk at the regrouping spot, which is both due to the possibility of the skiers triggering an avalanche from the regrouping spot, and the risk of an avalanche triggered somewhere else reaching the regrouping spot with the potential of catching many skiers. 2. Risk of guests ending up in places where they can trigger an avalanche, i f the guide does not regroup at the right time and locations. The remarks with notable avalanche reports (see page 108) indicate that the former factor might at least occasionally be a problem. However, the guides I interviewed seemed to be mainly concerned with the latter factor. Spacing between skiers Some guides say that spacing between skiers, meaning skiing one at a time, is not a tool that is used in heli-skiing. It simply takes too much time, and the cost is therefore too high. One guide said: \"Ifyou find yourself in a situation where you want to let skiers go one at the time, you should not have been there in the first place\". However, some pf the guides I interviewed, stated that they sometimes let skiers go one at the time, and a few guides always ask their skiers to keep proper distance from each others, also due to the risk of collisions. Some guides said that they sometimes control the distance between skiers, not with instruction, but with the pace of skiing. If they wanted skiers to be spaced widely apart, they would simply ski fast themselves, and as a result, the group would stretch out. Within C M H some operations have smaller private groups of 123 maximum four guests. Spacing seems to be used more in the group management of those small groups than with the bigger groups of 11 guests, because the guides find it easier to have a stricter control over a small group, and they are able to \"sneak\" down lines where they would not take the bigger groups. As with regrouping spots, there are two \"types\" o f avalanche risk that may be affected by controlling spacing between skiers: 1) the risk of triggering an avalanche may increase due to the load of many skiers in a small area of terrain (however, people don't agree upon whether this is the case), and 2) i f skiers are close together, there is an increased probability of more than one skier getting caught. The notable avalanche reports show that in over 40% of the cases when someone was caught in an avalanche, more than one person was caught. Group size Most of the guides I interviewed mentioned the group size in one way or another. A l l seemed to think that a group of 11 skiers is a large group, and talked about how much easier it is to manage smaller groups. Large groups use a lot o f space in the terrain and therefore require big margin o f safety. This means that the route has to contain a broad area of relatively safe terrain, because it is not possible to sneak past hazardous areas. Instructions are also more l ikely to get lost in a large group than a small one, and it is more difficult for the guide to make sure everyone understands what he is saying. 124 Group management and experience Some of the guides I interviewed said that group management becomes easier with increased experience. They have learned to give instructions efficiently, what to expect from guests and where guests might be lost. Confidence increases with experience, and it is a key factor in group management. Some guides said that in their first years of guiding, they felt like they had to provide excellent skiing for the guests all the time. With time they realised that this is not the case, and in the end, it is only about skiing in a general sense. The guests wi l l generally be happy with a few highlights, even i f the skiing is not excellent every run, every day. Therefore, the pressure of providing excellent skiing all the time reduces with increased experience. Group dynamics At C M H , the majority of the guests come again and again. Therefore, many guests have substantial experience in heli-skiing. The experience of the guests and their personalities affect the group dynamics. It is necessary to explain a lot to new people, but they are usually conservative and stay with the group, and are therefore not l ikely to cause problems. The majority of the experienced guests are very aware of the avalanche risk and they respect the guides' decisions. However, the interviews indicated that, on average, more experienced guests are more likely to cause problems than new guests, perhaps because they become overconfident or comfortable in the terrain and settings. One guide pointed out that when guides have stronger groups they ski faster and thus, process information faster, increasing the probability of an error. Furthermore, sometimes guides 125 may become distracted by the energy of the group, adjusting the pace for them and perhaps get a little carried away. Some guides said that they often form ideas about what kind of skiers they have in their groups by watching their behaviour before they start skiing. Noticing how they act in the avalanche transceiver training, and when they put their skis on. The guide can have a big influence on the group behaviour. For example, I observed that a small compliment from the guide on a group's behaviour can make the skiers happy and determined in keep on doing their best to follow instructions. One of the guides I interviewed was very concerned with how guides treated their guests. He said \"I would like to run the way I would like to be treated if I'm on a vacation. Because these people are on vacation and they pay a large amount of money for the vacation. I don't want to treat them like I am the king or I am the boss \". He said that the key is to build a good relationship with your group and that some guides have problem with groups because they treat the guests poorly. He emphasized that guides have to show leadership and help the guests to have a good time, even when conditions are poor. They should not say negative things about the guests to other guides, because that can have a snowballing effect and tension in the guest-guide relationship forms even before a guide gets the chance to know the particular guest. On the other hand, a good, positive group management can improve the behaviour of a group and, thus, decrease the overall risk of someone in the group triggering an avalanche. One guide suggested that too much technology can be dangerous to guests. He said that people show up now with goggles with a little fan and a big helmet. They have got a radio and an avalanche transceiver, an Avalung, and then they have a little video camera and some cameras. \"At the time to get out to the field they are so covered in technology they can't really ski. And 126 they are totally isolated from their environment\". I often noticed that the skiers with helmets did not hear what the guide said at the regrouping spot. Lack of avalanche risk awareness among guests is sometimes a problem according to some of the guides as well as my own observation. Perhaps people, who are not very knowledgeable about the nature of avalanche risk, don't perceive it. Risk perception, when rushing on skis through trees in steep slopes, is something that comes naturally, while risk perception in a 33 degree, open slope during a period of poor stability may be far from the actual risk. Thus, when clients in a backcountry operation are seeking their \"Target R isk \" or \"Optimal Risk Band\" (refer to section 2.2) they may underestimate the avalanche risk drastically. Therefore, I believe, it is important for guides to help the clients perceive the avalanche risk, in order to make them understand the rationale behind the instruction, increasing the probability that it wi l l be followed. However, some of my interviewees talked about how important it is that the guests feel safe in the skiing program, and that is probably the reason why many guides do not talk a lot about avalanche risk to their clients, even when the stability is a concern. One guide pointed out that the client needs to trust the guide, otherwise he or she w i l l start make decisions of his or her own, which wi l l not be as good as the guide's and is unacceptable in a guided party. 4.6 Factors potentially affecting decisions in a heli-skiing operation In every backcountry skiing operation a number of factors, in addition to avalanche risk evaluations, affect decisions on where and how to ski. However, every factor influencing terrain selection may affect the overall avalanche risk and therefore, it is important for people making the decisions to be aware of all those factors. Below is a list of factors that have the potential to sometimes affect decisions at heli-skiing operations. The list is a result of my observations during the fieldtrips to Canadian 127 Mountain Holidays, and it should be viewed only as a few examples of factors with the potential to affect decisions. • The weather often is a limiting factor when it comes to terrain selection. Poor weather may restrict the program, even though the avalanche risk is not considered high. • The visibility affects the skiing in many ways. The terrain selection becomes more difficult for guides when visibil i ty is poor, e.g. when the light is flat or the visibil i ty limited by fog or snow. A l l guides, but especially those not very familiar with the terrain, are in greater risk of ending up in places they didn't mean to ski , when the visibil i ty is poor. Some runs may not be desirable to ski with a group of people in poor visibil i ty due to risk of loosing track of clients, or clients ending up in hazardous places. • Helicopter flying is a big factor in all decision making in heli-skiing operations. The flying ability of a helicopter is closely related to the weather and visibility. Some days, the runlist might have 100 green runs, but only 10 of them may be reached by the helicopter due to flying conditions. Another effect is that the helicopter flight has to be effective. The guides try to get as many vertical meters skiing as possible out of as few flying hours as possible. The program should run as smoothly as possible and the helicopter should not have to wait too long between runs. Finally, the guides use the flying time to look at the runs from the air and think about terrain selection. • . Guides are constantly trying to provide the best quality skiing possible. If some elevation levels, aspects or areas have snow that is not considered good for skiing (e.g. breakable crust) those places are avoided as much as possible. If the quality of skiing for the previous days has been poor, guides may experience pressure, either from themselves or 128 their clients, to find better skiing. On the other hand, after a few days of top quality skiing, the atmosphere is more relaxed. • The behaviour of the group affects decisions on where and how to ski. Guides tend to ski differently when they are confident that the group is following them quite wel l , than with a group of \"misbehaviours\". If people in the group are not fol lowing instructions wel l , the margin of safety needs to be larger, and the guide keeps them far away from \"the edge\". • The guides keep track on the number of vertical meters the groups are skiing. C M H guarantees their clients a certain amount of vertical meters skied in their program. If they do not reach the minimum amount, the company refunds some money. A lso, a record is kept of the number of vertical feet each client has skied in total with the company. When they reach one mil l ion feet, they get a powder suit with \"mi l l ion feet\" label on the arm. A l l this results in the number of vertical meters skied being a potential pressure factor on decisions. If one group gets more vertical meters than another group, it may create tension and pressure. • Client pressure in general is a factor that has the potential to affect decisions. The clients at C M H pay a lot of money to spend a week heli-skiing. In many cases they are wealthy people who may own or direct big companies, and they are more used to giving instructions than following them. Some of the guests are very experienced backcountry-or heli-skiers and have their own opinions about where and how to ski. Guides, however, seem to be very aware of this potential pressure factors and take care that their decisions are not affected. A lso, the majority o f clients respect the guides' decisions and do no try to affect them. 129 • The run usage since last snowfall is also a potential pressure factor for decisions. Hel i -skiing clients expect to ski in untracked powder snow most o f the time. When there has been a while since last snowfall and the amount of skiable runs is limited, the runs get tracked up. That puts pressure on guides to venture further for untracked snow. 4.7 The role of experience in decision making A.l A Exper ience in ava lanche risk management Avalanche decision making can be defined as a type of \"natural decision making\" (Atkins, 2000) and mountain guides in heli-skiing face most of the conditions that characterise natural decision-making settings as defined by Orasanu and Conol ly (1993) (see page 15): They make decisions under time pressure since they have to make relatively fast decisions on terrain selection while skiing in the field (to ski over a roll or around it, go on the left or right side of a cutblock), even though the decisions are based on accumulative knowledge. The time pressure is, however, not of the same level as for fire commanders or airline pilots. High stakes are involved, since avoiding avalanche incidents, may be the matter of life or death. The guides are very experienced, since they usually spend a lot of time in the field, making decisions all the time. In their decision making, guides face uncertainties due to spatial and temporal variations in the condition of the snow cover. According to Walter Bruns (1997), the goal of the mountain-guide is to deliver safety and enjoyment to the clients. Safety and enjoyment often pul l in opposite direction, and the result is an unclear goal. Guides need to perceive patterns (cue learning) due to the complexity of the information they process, and they face dynamic conditions, because of the variability in e.g. the weather, snowcover, and terrain. Guides meetings are the field for team work, and guides within an operation area make many decisions collectively, even though guides also have to make many decisions on their own during skiing. 130 According to the Recognition-Primed Decision Model (Klein, 1998) (refer to page 15), experienced decision makers under natural settings, access a situation and judge it familiar, and then courses of action can be quickly evaluated by imagining how they w i l l be carried out. Hence, they do not make decisions in a strictly analytical way. Walter Bruns (1997) suggests that analytical methods are minimal in the decision-making process of mountain guides. The interviews and observations in this study, also indicated that analytical methods are not dominant in the decision process of experienced guides, especially not in the decisions made during skiing. (By \"analytical methods\" in this context, I mean methods to make decisions, e.g. some sort of if-then rules. The decisions may still be based on the laws of physics.) I believe that he Recognition-Primed Decision Model (Kle in 1998) fits wel l with some parts of the decision making process of professional ski-guides, especially decisions made in the field during skiing. Experienced heli-skiing guides describe that when they f ly over a run, they can identify the safest line from the top of the run to the bottom, right away. They do not have to analyse every step o f the way. This does , however, not mean that the decisions are not based on solid data. There is quite a strict framework for what data (both distributional and singular data as defined by McClung (2002)) are collected and used to make decisions. It is the process of the data ending up as decisions in the field that is not very transparent and happens mostly inside the head of the guides. I believe that, due to their experience, the guides are able to see the different data as a pattern and get an idea about the situation. They rarely use simple, context free rules and analytical methods while selecting routes and managing groups in the field. However, when they have to explain how they select routes and manage their group, they can be very analytical, and as was the case in the interviews. When explaining decisions, the interviewees started to analyse the situation and the decision making process in details. This in accordance with Kle in 's 131 (1998) theory, that people are more l ikely to use comparative evaluation when they have to justify their choices. It may also show that guides do recognise the great number of factors their decisions are based on, even though they don't go through them consciously while making the decisions. M y interviewees talked a lot about experience and how important it is in the decision making. Gut feeling or listening to the stomach are terms that guides often mention. If the \"feel ing\" is not right they don't like to go, even though they can't explain why. Bruns (1997) says that guides look for common themes and begin to think in terms of patterns, relating these patterns more and more subconsciously as the complexity increases. Experience enters by providing a greater number of patterns to draw on, and intuition enters by virtue of subconscious processes. This corresponds with Kle in 's (1998) ideas about experience and intuition and their role in the situation awareness of people (section 2.6): Experience allows people to be able to access a situation and judge it familiar and intuition depends on the use of experience to recognize key patterns that indicate the dynamics of the situation. McClung (2002) finds that both experience and targeted education are needed for successful decision making in avalanche forecasting. In reality, it may often be difficult to distinguish between the effects of targeted education and experience on decisions. Targeted education is a part of the experience of all professional guides. The situation awareness of guides in terms of avalanche risk evolves the whole winter and each day it depends on how it was the day before. Then fresh data during morning meetings, as well as discussion with the other guides results in a general idea about the conditions that day. Individual guides then pick up additional data and information during skiing, and they need to form a situation awareness for every slope skied on a given day. Therefore, I believe that the 132 situation awareness is based on a large amount of data and information, but it often forms subconsciously and is sometimes referred to as intuition or gut feeling. When the guides have made an assessment of the current conditions (situation awareness), they have to decide how they are going to act. According to Bruns (1997), guides think in terms of scenarios. They take the patterns from hazard assessment, l ink them as still frames of a moving picture on the terrain and then run that movie forward in time to anticipate consequences. This is what K le in (1998) refers to as Mental simulation, which is the ability to imagine people and objects consciously and to transform those people and objects through several transitions, finally picturing them in a different way than at the start. He states that it takes a fair amount of experience to construct a useful mental simulation. During my interviews the guides would often start to describe in details what might happen i f a certain decision was made, or a certain route selected. Some talked about that they imagine their way down the terrain, like a f i lm they have in their mind. This indicates that experienced guides often use mental simulations when deciding how to act based on their situation assessment. Because of memory limitations, people usually construct mental simulations using around three variables and around six transitions (Klein, 1998). Similarly, Bruns (1997) states that the trick throughout is to simplify: each situation is reduced to its most basic component situations and dealt with in causal sequence. Input factors are screened so that only the most relevant need to be considered. K le in (1998) identifies some possible problems of using mental simulation (Refer to section 2.6). One of the shortcomings is that we have trouble constructing mental simulations when the pieces of the puzzle get too complicated - there are too many parts, and these parts 133 interact with each other. Clearly, there is a certain risk of that in the case of mountain guides, since there are a large number of factors, which might affect avalanche risk. According to K le in (1998) one of the ways experts learn, is that they obtain feedback that is accurate, diagnostic, and reasonably timely. In domains that are not marked by opportunities for effective feedback, mere accumulation of experience does not appear to result in growth of decision expertise. The lack of feedback might be one of the problems for building expertise in the field of guiding. Most of the time, even bad decisions do not result in avalanche triggering. However, every guide recognizes that it might be impossible to know how close they were to triggering an avalanche and thus, the evaluation o f the quality o f decisions is difficult. One o f my interviewees said that every guide should ask himself or herself: How many times have I been somewhere Ishouldn't have been, but I didn't get the evidence to tell me Ishouldn't have been there? A.1.2 Exper ience in the terrain It is not only the general experience as a guide that influences the decisions of ski-guides. The experience in the terrain where they are skiing also affects the decision making process. In general, my interviewees believed that experience in terrain was a very good thing to have, since it makes it easier to select lines down the runs and avoid hazardous features. Some areas are by experience known to be hazardous, even though nobody really knows why. Therefore, guides, who are experienced in the terrain, have ideas about which runs and which lines are appropriate to ski under the different conditions. This knowledge is not structured and guides I talked to did not think it would be possible to write it down, because it is based on a complicated situation awareness, they cannot explain. Thus, for experienced guides, quick decisions on terrain selection in the field are fewer than for guides not experienced in the terrain. 134 It is also easier to give instructions to the group when the guide is familiar with the run. When the guide is not familiar, he or she might have to stop more often and give more instructions. However, some guides pointed out that i f they stay for a long time at the same place, it might lead to overconfidence. When guides go to new areas they usually have to be extra alert all the time, which might be good for avalanche risk management. They have to be very careful in their terrain selection just to find their way down to the pickup, and therefore keep their eyes open all the time. 4.8 Rule based decision methods for professional mountain guides? When asked about the possibility of adapting Munter's (2003) decision methods to professional mountain guides, my interviewees agreed upon that the Reduction Method (refer to page 27) was a good tool for recreational skiers, who don't know a lot about snow science. However, most of them did not believe that the method, or a similar method, should be used by professional guides. One guide said that i f they were brought up with using this method, it might be one of the tools for them to use, but experience would still be the most important part of decision making, and it would always shine over everything else. This was the main argument against using a simple rule based decision making method like Munter's: the guides believe they can make better decisions based on their experience, and take more factors into account. The Reduction Method is very different from their decision methods, even though they are looking at the same factors as the Reduction Method is based on. According to K le in (1998) and Dreyfus and Dreyfus (1986) inexperienced people are in greater need for analytical methods to make decisions, than experts. A lso , whereas novices may be confused by all the data elements, experts see the big picture, and they appear to be less likely to fall vict im to information overload. According to this, the Reduction Method may be good for 135 recreational skiers, while professional guides maybe better off using recognitional strategies based on experience. Rule based decision methods are examples of rational analysis, which reduces the chance that an important option wi l l be overlooked. Thus, it reduces the probability of errors in the decision making, and that is the main advantage of rule based methods. However, there are some limitations to rational analysis as described by K le in (1998) in section 2.6. Munter (2003) breaks the decision making into the basic elements of hazard rating, inclination, aspect and group factors in his Reduction Method. B y doing that, those factors are taken out of context of the situation at hand, and guides are often reluctant in doing that. On a given day, they might estimate that the avalanche risk is high in southern aspects due to solar radiation however, one of the southern facing runs has been in shadow from another mountain-face the whole day, and may therefore be skied even though most other southern facing runs are avoided. On another day, northern aspects may be avoided in lower elevations, because of a persistent surface hoar layer, while there is no need to avoid northern aspects in the alpine, because the surface hoar never formed in elevations above the valley fog. 136 Chapters. Conclusions 5.1 Rule based approach In my opinion, a strict, rule based approach is not desirable for the decision making process of professional guides, in terms of terrain selection or group management. Most of their decisions in the field are not made in a pure analytical way, but follow a proceedure partially contained in the descriptive \"Recognition-Primed Decision Mode l \" (Klein, 1998). However, the data and information gathering are largely formalized processes. Due to their experience, perhaps along with targeted education, professional ski-guides are able to evaluate a greater number of factors in a more complicated way than would be desirable in a strictly rule based decision making method. A lso, rule based methods similar to Munter's Reduction Method (Munter, 2003) require breaking the task into basic elements, such as inclination, aspect, elevation levels, which means taking these elements out of context of the big picture, while guides might be able to look at the big picture in context. For example, guides look at the size and the shape of the terrain as a whole, and the shape of the terrain is not easily broken into \"basic elements\"; it is more than inclination, elevation and aspect. Furthermore, analyses in section 3.4 in this thesis indicate that factors, such as elevation levels and aspects of avalanches, as well as the stability rating and time of the year when the avalanches fell, are interrelated. Even though the historical risk as defined in this study is in general highest under \"poor\" and \"very poor\" stability ratings that is not the case during late season. And even though the historical risk is in general highest in the N - N E - E sector, that is not the case in the alpine. When adding other factors, such as weather, snow and human factors, the picture gets even more complicated. 137 Methods similar to the French NivoTest (refer to section 2.9.2) attempt to keep some context and look at more factors than is done in the Reduction Method, such as snow factors. However, I beleive such methods risk becoming too complicated, and they don't result in a yes or no decisions anyway. The conclusion is that while simple rules based on a few single elements of avalanche risk (e.g. Munter's Reduction Method) may be good enough for recreational ski tourers, and potentially lower their avalanche risk, a method like that does not reflect the big picture well enough for professional mountain guides. Finally, one correct way of breaking avalanche risk into basic elements does not exists, and different researchers find different ways. There are, however, some shortcomings in relying too much on experience based decisions. Due to the many factors that affect avalanche risk, guides may sometimes get into trouble using their traditional methods for decision making. Furthermore, a lack of feedback may result in difficulties building up expertise in the field. Some structure and rational analysis analysis in the decision making system, might eliminate a number of errors caused by human misperception and other human factors. The decision making process, especially the data and information collecting process, does have some structure at C M H and most other backcountry skiing operations in Brit ish Columbia. The stability rating and runlist, made during morning meetings, are examples of such structure, which most l ikely eliminates quite a few errors from the decision making process. Those are, however, only tools to add structure to the decision making process, not tools to make the actual decisions. The runlist is a kind of a safety valve, assuring that guides don't end up in a ski-run without a formal group decision that the run is safe to ski that day. The decision process during the skiing day in the field is less staictured. 138 Risk calculations, based on avalanche statistics and usage data from the past in C M H , might serve as an advising tool in the decision making process, without letting strict rules replace experience. In fact, it should be desirable for guides, and the company as a whole, to analyse historical data in order to find patterns that cannot be seen without data analysis and calculations. 5.2 Recommendations, based on pattern from risk calculations Munter's (2003) idea about defining a hazard potential (in his method it is based on the public hazard rating) and then reducing the hazard with reduction factors, is a good example of how it is possible to think about avalanche risk management in a structured way. At Canadian Mountain Holidays ( C M H ) , the guides make their own stability rating, which is more detailed and at a smaller spatial scale than the public avalanche bulletin. In fact, the public avalanche bulletin for an area, is partly based on the stability ratings and information from backcountry skiing operations, such as C M H . The stability rating is therefore the best snowpack evaluation for the area, but it is made by the same people who are making the decisions in the terrain. Therefore, it is not an independent rating in the same way as the public bulletin in Munter's hazard potential (Munter, 2003). A lso, the stability rating is not a hazard rating, and the definition of the ratings (page 36) includes the presence or absence of naturally triggered.avalanches. Due to that, the stability ratings do not and should not directly reflect avalanche hazard for skiers. Sometimes the hazard under \"fair\" stability rating maybe extremely high, because there is a high risk of triggering large avalanches, but there are not many natural avalanches occuring, and therefore the stability is not rated \"poor\". A n d sometimes the hazard under \"poor\" stability rating may not be very high, because even though avalanches are easily released, the slab over the critical weak layer may be very thin and thus, only small avalanches are expected. 139 According to the calculations in Chapter 3, there is a big difference in historical risk between the stability ratings, when risk is defined as the probability of accidentally triggering an avalanche, greater than size 1. This is interesting because since the guides make the stability ratings themselves they are very much aware of the stability problems under \"poor\" stability ratings. The historical risk is nine times higher under \"poor\" or \"very poor\" stability ratings than under \"good\" or \"very good\". The historical risk is greater in higher elevation levels than lower down, and it is lower during late season than earlier on. The results for the relative risk calculated for those three factors; stability rating, elevation levels and time of the year, are quite clear and the statistical uncertainty is not too high since there are only three classes. The historical risk varies least with aspect o f the factors analysed in this study. The risk is, however, on average higher in the N - N E - E sector than in other aspects. It is also on average higher in the eastern 180° than the western 180°, while the difference between the northern and southern 180° is not great. The risk is not as dependent on aspects as indicated by avalanche data alone, since the aspects with the highest number o f avalanches, N , N E and E, are also the aspects most used. The question is whether it is the risk management of guides that reduces the difference in risk between aspects so drastically. I do not think this is the case, since the questionnaires and interviews indicated that aspect is in general not one of the most important factors in terrain selections. Therefore, one of the most important results o f this thesis is that aspect is a weak predictor of risk. Even though it is not clear how the usage data in this study relates to the usage of other groups, it show that it cannot simply be assumed that all aspects are used equally. Therefore, it cannot be assumed either, when some aspects have three times more avalanches than other aspects, that the risk is also three times higher there. Methods like Munter's Reduction Method 140 (Munter, 2003), who are based on avalanche data without usage data, might thus overestimate the dependence of risk on aspects. Furthermore, it can certainly not be assumed that i f 60% of avalanches happen in the N - N E - E sector, the avoidance o f this sector would decrease avalanche incidents by 60%. It should be kept in mind, that the analysis in this study do not reflect the unbiased risk potential of the situation since the data include the decisions of guides on whether, where and how to ski under the different stability ratings. The results rather reflect the residual, historical risk, after decisions had been made by guides. However, this w i l l always be the case with data on human behaviour. It might seem tempting to define risk numbers for different stability ratings and use as a hazard potential in a risk reduction formula following Munter (2003), and then define reduction factors based on analysis of other factors. If we would e.g. assign the number 2 to the risk under 'good' and 'very good' stability ratings, the number would be 6 for 'fair ' stability rating and 18 for 'poor' and 'very poor' stability ratings. We could imagine that staying below treeline would give some reduction factors, while skiing in the alpine would not. Ski ing outside o f the N - N E - E sector might even give a small reduction factor and so could skiing during late season. The problem with these calculations is that the factors are not independent of each others. It would be ideal to create a matrix with all the different factors, and calculate the historical risk under all possible combinations. Unfortunately, the database is too small to tolerate such detailed calculations. Therefore, I have looked at the combination of two different factors at the time,which is quite speculative. 141 Results of the risk analysis (refer to section 3.3 and 3.4) indicate that the probability of triggering an avalanche greater than size one could be lowered by taking more precautions than before under following conditions (the factors associated with highest historical risk are listed first): 1. The historical r isk is very high in \"very poor \" and \" p o o r \" stabil i ty ratings in al l aspects and elevation levels. 1.1. The risk under V P / P stability ratings is high during early and mid winter, while it is relatively low during late season 1.2. V P / P stability ratings and the N-NE-E sector is especially associated with high historical risk 2. The alpine 2.1. The sector E-SE-S-SWin the alpine has high risk 2.2. The risk is relatively high during early winter and mid winter in the alpine 3. Ea r l y and mid winter 3.1. As stated before the risk during early and mid winter is very high under VP/P stability ratings, and it is high in the alpine 3.2. Southern aspects are associated with high risk during mid winter (Feb 1 - Mar 15), while the sector N-NE has relatively high risk during early winter (Dec 1 - Jan 31) 4. N - N E - E 4.1. In addition to the factors mentioned above ( \" V P / P \" stability ratings and \"early winter\"), the risk in the treeline is especially high in the N - N E - E sector 4.2. The risk during \"Fair\" stability ratings is also especially high in the N - N E - E sector 142 It depends on the situation at hand, which actions would be the most effective in lowering the risk. The results of the questionnaires and interviews included in this study, may indicate a few relatively general factors that could be used to reduce the risk of triggering an avalanche: • Avoiding terrain, capable of producing big avalanches • Staying on slopes below 35 degrees or even staying below 30 degrees i f there is a special concern. • Few people in a group (e.g. 4 or less) may reduce the risk potential • Selecting regrouping spots with extra care, and regroup more often • Increased spacing between skiers • Skiing one at the time while skiing down or traversing steep slopes • Taking extra care avoiding convex rolls • Taking extra care when skiing around other hazardous small scale features, such as moraines, windloaded pockets, cutblocks, chutes and glades. 5.3 Further recommendations and contemplations Heli-ski ing guides make high stakes decisions every day in their work. Therefore, it is important for guides to constantly be aware of what is affecting their decisions at the time. There are so many factors that guides have to take into account during their entire decision making process during the day, that there is always a potential of factors like time pressure, flying, visibil ity or client or peer pressure affecting the decisions on the cost of avalanche safety. According to the questionnaire, the main difference in the decision making process, in terms of both terrain selection and group management, occurs between the stability ratings \"good\" and \"fair\" rather than between \"fair\" and \"poor\". The reason is probably that the stability rating \"fair\" is very comprehensive in terms of actual avalanche hazard for skiers, which can 143 sometimes be very high. Historical risk analyses in this study show that the difference in the historical risk of accidentally triggering an avalanche greater than size 1, is great between the stability ratings \"fair\" and \"poor\". Even when remotely triggered avalanches are excluded, the risk is still by far the highest under P / V P stability ratings. A lso, the difference in ratio of sizes 1.5 avalanches is not great: about 48% of avalanches under V P / P stability ratings were size 1.5 while the ratio is close to 43% for avalanches under \"fair\" stability ratings. Perhaps guides underestimate the risk under \"poor\" stability rating on average as compared to \"fair\" stability rating. The intensity of skiing in the runs, affects the terrain selection o f guides. There are many good reasons for this. However, it may be important to bear in mind that intensive skiing alone is not always a reliable indicator of a stable run. The same applies to runs that have been used before the same day or the days before. Guides have different theories on skier compaction, and many experienced guides are very reluctant to completely rely on it. Usually terrain selection based on skier compaction is made as a group decision during morning meetings, which should increase the safety margins of those decisions. However, the guides in each operation probably have more homogeneous views than guides between different operations, because they affect the views of each other. Instructions from guides to guests are an important part of avalanche risk management and should be constantly evaluated and discussed among guides. Simple, clear instructions are important, but guides have to be careful not assuming that the guests know more than they do about e.g. stability or avalanche risk. It is important for guests to be told exactly what is expected from them. Many guests at C M H are experienced heli-skiers, but that should not affect the 144 amount of instructions given to less experienced guests. A compliment by a guide might sometimes be more powerful than instructions to ensure good behaviour o f a group. Guides have different opinions and use different methods on controlling spacing between skiers. There are some indications that increased average spacing between skiers, and letting skiers more often ski one at a time, may decrease the risk of skiers triggering, or getting caught in avalanches. The cost of increased spacing would however have to be evaluated, and that brings us to the concept of acceptable risk, or the Operational Risk Band (refer to section 2.2). Avalanche risk management in a backcountry skiing operation as with other fields, is a kind of a cost and benefit analysis. It is not desirable to be too conservative, since people would not come heli-skiing i f they were only offered to ski slopes under 25 degrees. Therefore, it is important for backcountry skiing operators to acknowledge that there is some cost associated with being more conservative. However, increased knowledge on avalanches, and avalanche risk management could add to the safety, without adding too much to the operational cost. 5.4 Suggestions for future work Professional heli-skiing guides are experts in managing avalanche risk during backcountry travelling and they do make excellent decisions based on their experience and targeted education. However, it is important to keep looking for ways to eliminate errors in the decision making process. As C M H ' s database grows, more statistical analyses can be performed, and other databases might be included in such analyses. It is especially important to continue to form ideas about how and when backcountry travellers use terrain. It would be interesting to analyse the terrain usage of other groups, such as recreational travelers, but that is associated with some difficulties in data collection. 145 Experienced heli-skiing guides hold a lot of knowledge about avalanche risk management - a knowledge that can, and should be explored much deeper than done in this study. 146 References Adams, J., 1995: Risk, University College London. London, U K . Atkins, D., 2000: Human Factors in Avalanche Accidents, Proceedings of the International Snow Science Workshop, B ig Sky, Montana, U S A , 46-50. Barber, G. M . , 1988: Elementary Statistics for Geographers, The Gui l ford Press, New York, U S A . Baron, J. 1988: Thinking and Deciding. Cambridge University Press, Cambridge, U K . Beach, L.R. and Lipshitz, R., 1993: Why classical decision theory is an inappropriate standard for evaluating and aiding most human decision making, in Decision Making in Action: Models and Methods, G. K le in et al. (eds.), Ablex Publishing, Norwood, N J , U S A , 21-35. Bezzola, C , 2003: Personal communication, E-mail and conversations spring and fall of 2003. Birkeland, K .W. and Landry, C C , 2002: Changes in spatial patterns of snow stability through time, Proceedings of the International Snow Science Workshop, Penticton, B C , Canada. Bolognesi, R., 2004: Evaluer le risque avalanche avec le NivoTest (Resume). Webpage sited May 25, 2004, http://www.meteorisk.com Bruns, W., 1997: Snow science and Safety for The Mountain Guide. International Snow Science Workshop 1997, Canadian Avalanche Association, Revelstoke, B C , Canada, 203-206 C A A , 2004: Canadian Avalanche Association, Trends and Patterns in Avalanche Accidents. From avalanche Accidents in Canada Volume 4: 1984 - 1996. Supplemented September 2003 with data from the Canadian Avalanche Association, webpage sited February 26, 2004, http://www.avalanche.ca/accident/index.html C M H , 2001: CMH Snow Safety Guidelines, Rev. Nov. 2001. C M H , 2004: Canadian Mountain Holidays, webpage sited February 15, 2004, http://www.cmhski.com Daffem, T., 1999: Avalanche Safety, Rocky Mountain Books, Calgary, Alberta, Canada. Dreyfus, H.L. and Dreyfus, S.E., 1986: Mind Over Machine: The Power of Human Intuition and Expertise in the Era of the Computer, Free Press, New York, U S A . Fel l , R., 1994: Landslide risk assessment and acceptable risk, Canadian Geotechnical Journal, 31, 261-272. Fel l , R. and D. Hartford, 1997: Landslide risk management. Landslide Risk Assessment, Cruden & Fell (eds), 51-109. 147 Finlay, P.J. , and Fel l , R., 1997: Landslides: risk perception and acceptance, Canadian Geotechnical Journal, 34, 169-188 Hageli, P. and McClung , D .M . , 2003: Avalanche characteristics of a transitional snow climate -Columbia Mountains, Brit ish Columbia, Canada, Cold Regions Science and Technology, 37, 255-276. Jamieson, B. and Geldsetzer, T., 1996: Avalanche Accidents in Canada Volume 4: 1984-1996, Canadian Avalanche Association, Revelstoke B C , Canada. Keylock, C , 1997: Avalanche Risk in Iceland, M.Sc. thesis U B C , Vancouver, Canada. K le in , G. , 1998: Sources of Power. How People Make Decisions, Massachusetts Institute of Technology, London, England. Kronl iolm, K., Schweizer, J . and Schneebeli, M . , 2002: Spatial variability o f snow stability on small slopes, Proceedings of the International Snow Science Workshop, Penticton, B .C , Canada. Kurzeder, T. and Feist, H., 2003: Powder guide. Managing avalanche risk, Mountain sports press, C O , U S A . LaChapelle, E.R., 1980. The fundamental processes in avalanche forecasting. Journal of Glaciology, 26, 75-84. Logan, N. and D. Atkins, 1996: The Snowy Torrents: Avalanche Accidents in the United States, 1980-1986, Colorado Geological Survey Special Publication 39, Colorado Geological Survey, Department of Natural Resources, Denver, Colorado, U S A . McCammon, I. 2000: The Role of Training in Recreational Avalanche Accidents in the United States, Proceedings of the International Snow Science Workshop, B i g Sky, Montana, U S A , 37-45. McClung, D., 2000: Predictions in avalanche forecasting, Annals of Glaciology, 31. McClung, D., 2002: The Elements of Appl ied Avalanche Forecasting. Part I: The Human Issues, Natural Hazards, 25, 111-129. McClung , D. and P. Schaerer, 1993: The Avalanche Handbook, The Mountaineers, U S A . Ministry of Sustainable Resource Management, 1996: Gridded D E M Specification. Brit ish Columbia Specifications and Guidelines for Geomatics. Webpage sited Apr i l 15, 2003, http://home.gdbc.gov.bc.ca/products/griddem.htm Munter, W., 2003: Risikomanagement im wintersport.. Agentur Pohl & Schellhammer, Gannisch-Partenkirchen, Germany. National Research Counci l , 1991: Uses of Risk analysis To Achieve Balanced Safety In Building Design and Operations, National Academy Press, Washington, D . C , U S A . 148 Olkin, I., Gleser, L.J . and Derman, C , 1980: Probability Models and Applications, Macmil lan Publishing, New York, U S A . Orasanu, J . and Conolly, T., 1993: The reinvention of decision making. In K le in , G. et al. (eds.), Decision making in action: Models and methods, Ablex Publishing Norwood, N J , U S A , 3-20. Sarantakos, S., 1998: Social Research - second edition, Macmi l lan press L T D , London, U K . Sivardiere, F., 2003: Presentation in C A A Avalanche Workshop, December 2003, Vancouver. Starr, C , 1969: Social benefit versus technological risk, Science, 165, 1232-1238. Schweizer, J . and Jamieson, J.B. , 2000: Field Observations of Skier-triggered Avalanches, Proceedings of the International Snow Science Workshop, B i g Sky, Montana, U S A , 192-199. Schweizer, J . and Liitschg, M . , 2000: Measurements of Human-triggered Avalanches from the Swiss Alps, Proceedings of the International Snow Science Workshop, B i g Sky, Montana, U S A , 200-207. Tremper, B., 2001: Staying Alive in Avalanche Terrain. The Mountaineers, U S A . Wilde, G.J.S., 2001: Target Risk 2 -A new psychology of safety and health. What works? What doesn't? and why, P D E Publications, Ontario, Canada. Wi lhelm, C , 1997: Wirtschaftlichkeit im Lavinenschutz - Methodik und Erhebungen zur Beurteilung von Schutzmassnahmen mittels quantitativer Risikoanalyse und okonomischer Bewertung. Mitteilungen 54. Eidgenossisches Institut fur Schnee- und Lavinenforschung. Davos, Switzerland. 149 Appendix A. The 3x3 Formula The following table is based on the translation of Munter's (2003) 3x3 method in Kurzeder and Feist (2004) 1. Regional filter Pre-departure preparation, including alternatives 1. Snow and Weather Conditions 2. Terrain 3. Human Factor • Avalanche bulletin • Weather foreast • Expert opinions and information • A map with the largest scale available • Guide-books, photos, air photos • Personal knowledge of the terrain and area you will be traveling in • Who is coming? • Level of experience and skill • Equipment • Physical fitness, mental strength • Is there a designated leader? 2. Local filter Within range of vision Selection of route and alternatives 1. Snow and Weather Conditions 2. Terrain 3. Human Factor • Snow: wind loading, signs of danger, critical amount of new snow, wind: ripples, dunes or sastrugi, is the AB correct? - make changes in case of doubt, identify dangerous areas • Weather: visibility, clouds, wind speed, dangerous wind loading?, precipitation (amount of new snow), temperature changes • Terrain features • Slope angle • Aspect(s) • Are the tracks left by previous riders or skiers adapted to current terrain and current danger factors? • Who are my companions? • Check emergency equipment (tranceivers, etc.) • Are other groups traveling on the same route? (Discuss and coordinate plans with them, if helpful.) • Time factor (remaining time): Are you on time? Or is it getting too late? 3. Zonal filter Check the slope in front of you 1. Snow and Weather Conditions 2. Terrain 3. Human Factor • How much new snow has fallen on the slope you are going to ride? • Recent wind deposition? • Influence of solar radiation? • Where could a slab release? • How big would the slab be? • Situation (terrain and people) above and below • Steepest section: how steep? • Aspect? • Typical avalanche terrain: steep, rocky slope; steep leeward slope; steep shaded forest glade? • Slope shape • Elevation • Frequent traffic (regularly tracked-out slope) • Ability, discipline, fatigue of group members • Precautions: safety distance, go one at a time, follow same track, identify \"islands of safety\" as meeting points • Avoid dangerous areas • Avoid too risky routes >choose alternative routes 150 Appendix B. The Canadian size classification system for avalanches Size Descript ion Typ ica l Mass Typ ica l Path Length Typ ica l Impact Pressures 1 Relatively harmless to people <10t 10m 1 kPa 2 Could bury, injure, or kill a person 10 2 t 100 m l O k P a 3 Could bury a car, destroy a small building, or break a few trees 10 J t 1000 m 100 kPa 4 Could destroy a railway car, large truck, several buildings, or a forest with an area up to 4 hectares 10 4 t 2000 m 500 kPa 5 Largest snow avalanches known; could destroy a village or a forest of 40 hectares 10 5 t 3000 m 1000 kPa (McClung and Schaerer, 1993) 151 Appendix C. Copy of questionnaire Dear guide This questionnaire is a part of my Master's thesis at U B C . I am doing a research on the avalanche risk management at backcountry skiing operations. A part of the research is to capture the knowledge and experience of professional backcountry skiing guides, and that is the purpose of this questionnaire. The questionnaire focuses on two subjects: firstly on the use of terrain in avalanche risk management and secondly on group management and the connection to avalanche risk. Personal Information Gender: • Male • Female Age: • Under 25 • 25-29 • 30-34 • 35-39 • 40-49 • 50-59 • 60 or older Guiding experience in years: Guiding education: Mountain guide Ski guide Assistant guide Other, what? Terrain 1.1. How do you take inclination into account when you are guiding a group of skiers? Check the statement that is closest to reflect your guiding style. You may check more than one statement. I am constantly aware of the inclination of the terrain I am skiing. I only think in terms of inclination when I believe it is necessary due to the conditions. First and foremost I think about the shape of the terrain (convex rolls, gullies, bowls, depressions etc.), rather than inclination directly. I usually do not think in terms of inclination. Other, what? 152 1.2. How important is inclination to you when guiding under different conditions. If the level of importance does not change with the stability, please check only one box under \"good stability\". Good stability • Not important • Somewhat important • Very important Fair stability • Not important • Somewhat important • Very important Poor stability • Not important • Somewhat important • Very important 2. How important is aspect of different scales to you when guiding under different conditions. If the level of importance does not change with the stability, please check only one box under \"good stability\". i) Large scale aspect. That is the main aspect of a ski-run as a whole Good stability • Not important • Somewhat important • Very important Fair stability • Not important • Somewhat important • Very important Poor stability • Not important • Somewhat important • Very important ii) Small scale aspect. That is the aspect of individual slopes within a ski-run. Good stability • Not important • Somewhat important • Very important Fair stability • Not important • Somewhat important • Very important Poor stability • Not important • Somewhat important • Very important 153 3. Which small scale terrain features do you regard as the most hazardous in general under different conditions? Please put the number 1 in front of the most hazardous terrain feature, 2 in front of the second most hazardous etc, for each stability rating. If this does not depend on the stability in your opinion, please rank only the features under \"good stability\". Good stability Convex rolls Windloaded depressions Gullies Small bowls The area around cliffs Other terrain features, which? Fair stability Convex rolls Windloaded depressions Gullies Small bowls The area around cliffs Other terrain features, which? Poor stability Convex rolls Windloaded depressions Gull ies Small bowls The area around cliffs Other terrain features, which? 154 4. Please rank the following terrain-factors depending on their relative importance in your opinion in terms of avalanche risk management. Put number 1 in front of the most important factor, 2 in front of the second most important factor, etc. for each stability rating both for the morning meetings and during skiing. Good stability: When the runlist is made during morning meetings Elevation level Inclination Aspect General shape of terrain During skiing Elevation level Inclination Aspect General shape of terrain Fair stability: When the runlist is made during morning meetings Elevation level Inclination Aspect General shape of terrain During skiing Elevation level Inclination Aspect General shape of terrain Poor stability: When the runlist is made during morning meetings Elevation level Inclination Aspect General shape of terrain During skiing Elevation level Inclination Aspect General shape of terrain 155 Group Management la. How well do the clients follow your instructions? Please check only one statement that is closest to reflecting your experience. Very well, I have rarely any problems regarding that. Most clients follow my orders but once in a while I have problems with individual guests in this regard. The majority of clients follow my orders but it is quite common to have incidents where a guest or guests are not located where I intended them to be. Not well. It is a serious problem and too many guests are not doing what they are told. Other, what? lb. How much effect do you think that clients not following instructions have on the avalanche risk in your operation as a whole? Very much • Some • Not much • None 2. How do you choose the regrouping spots in terms of avalanche risk? Please check only one statement that is closest to reflecting your guiding style. a) I always choose the regrouping spots very carefully. I am constantly trying to find a spot that can not be triggered and is not threatened by avalanches from above, no matter how the snow stability is that day. b) It depends on the stability how carefully I choose the regrouping spots. If option b) is chosen please answer the following question: Under which conditions do you worry significantly about regrouping spots? • Good stability • Fair stability • Poor stability Other, what? 2b. How important do you believe regrouping spots are when thinking about the overall avalanche risk in your operation? • Very important • Somewhat important • Not very important • Not important 156 3. How do you control the spacing between skiers in your group? Please check the statements that are closest to reflect your guiding style. I) When skiing downhill a) I always tell my skiers to keep proper distance from each other (at least 2 or 3 turns), no matter how the snow conditions are. I make sure that they follow my orders. b) It depends on the stability whether I control the distance between my skiers. If option b) is chosen please answer the following question: Under which conditions do you use spacing when skiing downhill? • Good stability • Fair stability • Poor stability II) When traversing steep slopes a) When traversing steep slopes I only let one skier traverse at a time no matter how the stability rating is. b) It depends on the stability whether I control the distance between my skiers If option b) is chosen please answer the following question: Under which conditions do you use spacing when traversing steep slopes? • Good stability • Fair stability • Poor stability Other, what? 3b. How important do you believe spacing between skiers is, when thinking about the overall avalanche risk in your operation? • Very important • Somewhat important • Not very important • Not important 4. Below are some statements on the strictness of group management and freedom of movement. The statements apply to your operation as a whole rather than your personal guiding style. Please check only one statement that is closest to reflect your views. I don't think a stricter group management would lower the risk of triggering avalanches in my operation. I believe a stricter group management would somewhat lower the risk of triggering avalanches in my operation, but the cost of restricted freedom of movement would be too high to justify it. I believe that we can use stricter management than we do now, in order to lower the risk of triggering an avalanche, without restricting the freedom of movement too much. Other, what? 157 Appendix D. Results from questionnaire Below, the raw results from the questionnaire are presented. The questionnaire was answered by 40 guides. Personal information Gender: Table 8. Questionnaire: Gender Male 36 (90%) Female 4(10%) Table 9. Questionnaire: Age <25 1 25-29 5 30-34 3 35-39 13 40-49 16 50-59 2 159 Guiding experience in years: Range: 1-34 years Mean: 12 years Guiding education: Table 10. Questionnaire: Guiding education Assistant guide 8 (20%) Ski guide 12 (30%) Mountain guide 20 (50%) Terrain selection 1.1. How do you take inclination into account when you are guiding a group of skiers? Check the statement that is closest to reflect your guiding style. You may check more than one statement. Table 11. Questionnaire: How inclination is taken into account 1. I am constantly aware of the inclination of the terrain I am skiing. 6 2. I only think in terms of inclination when I believe it is necessary due to the conditions 1 3. First and foremost I think about the shape of the terrain (convex rolls, gullies, bowls, depressions etc.), rather than inclination directly 16 4. I usually do not think in terms of inclination. 0 Statement 1 and 3 16 Statements 1, 2 and 3 1 1.2. How important is inclination to you when guiding under different conditions: 160 Good stability: Table 12. Questionnaire: The importance of inclination under \"good\" stability rating Not important 1 Somewhat important 25 Very important 13 Fair stability: Table 13. Questionnaire: The importance of inclination under \"fair\" stability rating Not important 0 Somewhat important 2 Very important 36 Poor stability: Table 14. Questionnaire: The importance of inclination under \"poor\" stability rating Not important 0 Somewhat important 1 Very important 37 2. How important is aspect of different scales to you when guiding under different conditions: i) Large scale aspect, that is the main aspect of a ski-run as a whole Good stability: Table 15. Questionnaire: The importance of large scale aspect under \"good\" stability rating Not important 3 Somewhat important 24 Very important 9 Fair stability: Table 16. Questionnaire: The importance of large scale aspect under \"fair\" stability rating Not important 0 Somewhat important 15 Very important 17 Poor stability: Table 17. Questionnaire: The importance of large scale aspect under \"poor\" stability rating Not important 1 Somewhat important 8 Very important 23 162 ii) Small scale aspect, that is the aspect of individual slopes within a ski-run. Good stability: Table 18. Questionnaire: The importance of small scale aspect under \"good\" stability rating Not important 3 Somewhat important 26 Very important 6 Fair stability: Table 19. Questionnaire: The importance of small scale aspect under \"fair\" stability rating Not important 0 Somewhat important 20 Very important 13 Poor stability: Table 20. Questionnaire: The importance of small scale aspect under \"poor\" stability rating Not important 0 Somewhat important 16 Very important 17 3. Which small scale terrain features do you regard as the most hazardous under different conditions? Please put the number 1 in front of the most hazardous terrain feature, 2 in front of the second most hazardous etc, for each stability rating. 163 The tables below show the number of times each terrain feature was assigned a certain rating; e.g., under good stability rating, convex rolls were rated number one a total of 20 times. Good stability: Table 21. Questionnaire: Hazard rating of small scale terrain features under \"good\" stability rating Terrain feature 1 2 3 4 5 6 Convex rolls 20 6 2 2 0 0 Windloaded depressions 11 6 4 2 1 0 Gullies 2 9 3 8 2 0 Small bowls 0 2 3 5 7 2 The area around cliffs 3 5 7 0 6 1 Fair: Table 22. Questionnaire: Hazard rating of small scale terrain features under \"fair\" stability rating Terrain feature 1 2 3 4 5 6 Convex rolls 23 6 1 0 0 0 Windloaded 10 9 3 3 0 0 depressions Gullies 5 6 7 6 1 0 Small bowls 2 5 4 4 7 1 The area around 4 4 3 3 6 0 cliffs 164 Poor : Table 23. Questionnaire: Hazard rating of small scale terrain features under \"poor\" stability rating Ter ra in feature 1 2 3 4 5 6 Convex rolls 25 6 1 0 0 0 Windloaded 11 8 4 1 1 0 depressions Gullies 8 6 7 5 0 0 Small bowls 4 4 4 7 3 1 The area around 5 2 1 2 11 1 cliffs One way to estimate the overall ranking of importance is to assign a terrain feature 6 points each time it is ranked number 1, and 5 points each time they are ranked number 2. With this method, the total points under the different stability ratings are like following: Table 24. Questionnaire: Hazard scores of small scale terrain features Ter ra in feature Good Fa i r Poor Convex rolls 164 172 184 Windloaded depressions 120 126 127 Gullies 97 108 121 The area around cliffs 84 77 73 Small bowls 53 80 88 165 4. Please rank the following terrain-factors depending on their relative importance in your opinion in terms of avalanche risk management. Put number 1 in front of the most important factor, 2 in front of the second most important factor, etc. for each stability rating both for the morning meetings and during skiing. The tables below show the number of times each terrain factor was assigned a certain rating; e.g., under good stability rating and during morning meetings, the general shape of terrain was rated number one a total of 25 times Good stability: When the runlist is made during During skiing: morning meetings: Table 25. Questionnaire: Hazard rating of terrain Table 26. Questionnaire: Hazard rating of terrain factors in morning meetings under \"good\" factors during skiing under \"good\" stability stability rating rating Terrain factor 1 2 3 4 Terrain factor 1 2 3 4 Elevation level 2 9 5 13 Elevation level 3 9 4 10 Inclination 5 14 5 6 Inclination 8 14 5 3 Aspect 4 11 10 5 Aspect 1 9 11 5 General shape 25 3 4 2 General shape 27 2 3 3 of terrain of terrain 166 Fair stability: When the runlist is made during morning meetings: Table 27. Questionnaire: Hazard rating of terrain factors in morning meetings under \"fair\" stability rating Terrain factor 1 2 3 4 Elevation level 8 6 4 11 Inclination 8 16 6 2 Aspect 6 7 10 5 General shape of terrain 23 4 3 4 Poor stability: When the runlist is made during morning meetings: Table 29. Questionnaire: Hazard rating of terrain factors in morning meetings under \"poor\" stability rating Terrain factor 1 2 3 4 Elevation level 9 5 4 11 Inclination 9 18 4 1 Aspect 7 4 13 4 General shape of terrain 24 5 0 4 During skiing: Table 28. Questionnaire: Hazard rating of terrain factors during skiing under \"fair\" stability rating Terrain factor 1 2 3 4 Elevation level 7 8 5 10 Inclination 9 17 3 3 Aspect 4 6 11 6 General shape of terrain 27 4 3 2 During skiing: Table 30. Questionnaire: Hazard rating of terrain factors during skiing under \"poor\" stability rating Terrain factor 1 2 3 4 Elevation level 8 6 5 10 Inclination 9 16 4 1 Aspect 5 6 11 5 General shape of terrain 26 6 0 3 167 In order to estimate the overall ranking of importance a similar way was used as before (question 3). The terrain features are given 4 points each time they are ranked number 1 and 3 points each time they are ranked number 2. With this method, the total points under the different stability ratings are as follows: When the runlist is made dur ing morning meetings: Table 31. Questionnaire: Hazard scores of terrain factors in morning meetings Ter ra in factor Good Fa i r Poor General shape of terrain 119 114 115 Inclination 78 94 99 Aspect 74 70 70 Elevation level 58 69 70 Dur ing sk i ing: Table 32. Questionnaire: Hazard scores of terrain factors during skiing Ter ra in factor Good Fa i r Poor General shape of terrain 123 128 125 Inclination 87 96 93 Aspect 58 62 65 Elevation level 57 72 70 168 4.3.3 Group Management l a . How well do the clients follow your instructions? Please check only one statement that is closest to reflecting your experience. Table 33. Questionnaire: Clients following instructions Very wel l , I have rarely any problems regarding that. 4 Most clients follow my instructions but once in a while I have problems with individual guests in this regard. 31 The majority of clients follow my instructions but it is quite common to have incidents where a guest or guests are not located where I intended them to be. 3 Not wel l . It is a serious problem and too many guests are not doing what they are told. 0 l b . How much effect do you think that clients not following orders have on the avalanche risk in your operation as a whole? Table 34. Questionnaire: The effect of clients not following instructions Very much 22 Some 10 Not much 7 None 0 2. How do you choose the regrouping spots in terms of avalanche risk? Please check only one statement that is closest to reflecting your guiding style. Table 35. Questionnaire: The selection of regrouping spots a) I always choose the regrouping spots very carefully. I am constantly trying to find a spot that can not be triggered and is not threatened by avalanches from above, no matter how the snow stability is that day 14 b) It depends on the stability how carefully I choose the regrouping spots. 23 If option b) is chosen please answer the following question: Under which conditions do you worry significantly about regrouping spots? Good 1 Fair 16 Poor 6 2b. How important do you believe regrouping spots are when thinking about the overall avalanche risk in your operation? Table 36. Questionnaire: The importance of regrouping spots Very important 25 Somewhat important 13 Not very important 1 Not important 0 170 3. How do you control the spacing between skiers in your group? Please check the statements that are closest to reflecting your guiding style. I) When skiing downhill: Table 37. Questionnaire: Spacing between skiers when skiing downhill a) I always tell my skiers to keep proper distance from each other (at least 2 or 3 turns), no matter how the snow conditions are. I make sure that they follow my orders. 8 b) It depends on the stability whether I control the distance between my skiers. 29 If option b) is chosen please answer the following question: Under which conditions do you use spacing when skiing downhill? Good 2 Fair 21 Poor 3 II) When traversing steep slopes: Table 38. Questionnaire: Spacing between skiers when traversing steep slopes a) When traversing steep slopes I only let one skier traverse at a time no matter how the stability rating is. 10 b) It depends on the stability whether I control the distance between my skiers 29 If option b) is chosen please answer the following question: Under which conditions do you use spacing when traversing steep slopes? Good 2 Fair 21 Poor 3 3b. How important do you believe spacing between skiers is, when thinking about the overall avalanche risk in your operation? Table 39. Questionnaire: The importance of spacing between skiers Very important 11 Somewhat important 19 Not very important 6 Not important 0 172 4. Below are some statements on the strictness of group management and freedom of movement. The statements apply to your operation as a whole rather than your personal guiding style. Please check only one statement that is closest to reflect your views. Table 40. Questionnaire: Strictness of group management I don't think a stricter group management would lower the risk of triggering avalanches in my operation. 15 I believe a stricter group management would somewhat lower the risk of triggering avalanches in my operation, but the cost of restricted freedom of movement would be too high to justify it. 6 I believe that we can use stricter management than we do now, in order to lower the risk of triggering an avalanche, without restricting the freedom of movement too much. 13 173 "@en ; edm:hasType "Thesis/Dissertation"@en ; vivo:dateIssued "2004-11"@en ; edm:isShownAt "10.14288/1.0099777"@en ; dcterms:language "eng"@en ; ns0:degreeDiscipline "Geography"@en ; edm:provider "Vancouver : University of British Columbia Library"@en ; dcterms:publisher "University of British Columbia"@en ; dcterms:rights "For non-commercial purposes only, such as research, private study and education. Additional conditions apply, see Terms of Use https://open.library.ubc.ca/terms_of_use."@en ; ns0:scholarLevel "Graduate"@en ; dcterms:title "Avalanche risk management in backcountry skiing operations"@en ; dcterms:type "Text"@en ; ns0:identifierURI "http://hdl.handle.net/2429/15692"@en .