SIGNIFICANCE OF HISTORICAL RECORDS FOR AVALANCHE HAZARD ZONING IN NORWAY
Krister Kristensen*, Carl Harbitz* Norwegian Geotechnical Institute, Oslo
Alf Harbitz** Norwegian Institute of Fisheries and Aquaculture Ltd, Trorns0
ABSTRACT: In avalanche hazard zoning, it is common practice to investigate the previous avalanche history for the area considered. Historical observations of avalanches serve as an aid in the classification of the terrain, and may also serve as verification of estimates of avalanche runout. Conditions influencing the avalanche occurrences may change significantly over time and it is important to take these changes into account when using historical avalanche observations in hazard ioning today. A number of the most extensive avalanches recorded in Norway, are found during the eigthteenth and nineteenth century. The catastrophes may be linked to weather as well as to socio-econornic conditions, in particular deforestation of mountain slopes. The implications of using or disregarding historical avalanche observation are shown in an example of statistical estimation of avalanche runout.
KEYWORDS: Snow avalanche, run-out distance, -probability, extreme value statistics, Western Norway
1. INTRODUCTION Given enough time, nature will again provide the conditions that caused the avalanche in the past. According to the Norwegian Building Assuming that the avalanches were mainly Regulations, the "safe" areas for habitations in the related to the weather conditions, the avalanche snow-avalanche prone regions of Norway are de probability can be related to the probability of fined as areas where the nominal annual prob such weather conditions. ability of a house bein:p hit by an avalanche One problem with this approach is that it should be less than 10-. This means that not may often not be possible to reconstruct the snow only the obvious paths with recurring avalanches conditions and the weather sequences with have to be considered, but also areas where enough detail. Another problem, which will be avalanches are known to have occurred very discussed in this paper, is that there may be rarely, or not at all. other, non-weather related, factors influencing the Land use planning in avalanche prone ar avalanche activity. This makes it important to eas is usually an interdisciplinary task. It includes have a fairty broad historical perspective when avalanche dynamics calculation, geomorphologic using historical avalanche observations. It may interpretation, meteorology, climatology and his also be that some of these factors are not present torical research. The historical records of ava today, and this may have implications on the lanches often serve as an indication of what may estimates of avalanche probabilities. be potential starting zones and as a verification of estimates of avalanche runout. 2. USE OF HISTORICAL DATA The existence of a historical observation is often seen as proof of an existing hazard today. Historical data of avalanche occurrences in Western Norway vary in quality and availability. In many cases the interpretation of them demand Corresponding author addresses: use of historical research methods, particularty the *Krister Kristensen and Carl Harbitz: procedures of source validation. The sources Norwegian Geotechnical Institute, P.O. Box include farm and family histories, newspaper 3930 Ullevaal Stadion, N-0806 OSLO, reports, official reports of damage appraisals, per Norway sonal letters, etc. Problems with the sources E-mail: [email protected]@ngLno include lack of coherent terminology: e.g. they **Alf Harbitz: Norwegian Institute of Fisheries may not distinguish between snow and slush ava and Aquaculture Ltd., 9291 TROMS0, lanches. It may also be that the messenger has Norway underlying motives, such as pleas for tax reduction E-mail: [email protected] or financial support that could lead to overstate ments of the magnitude of the avalanche.
475 Unless photos or maps exist, reconstruct damage. Both settlements are known to "-e ing the spatial distribution of an avalanche in the been permanently settled before 1600 AD. F past at a given site, will often include rather sub thermore, Gj0rven and Tenden are still 0CCu~ jective interpretation. If it is possible to identify today, more or less at the same sites as in 1868 ruins or obtain old maps of building sites, this can and none have suffered any avalanche dartlage provide data on certain points, e.g. which buildings since then. were destroyed and which were not. But often the It may be of interest to look at the COrtcI determination of the avalanche's area of influence tions that led to this unusual avalanche activity in is left to interpretation and avalanche dynamics 1868. Was it only the unusual combination ~ calculations and this is clearly a source of error. A weather sequences or were other factors also im claim about what a snow avalanche debris looked portant? like 100-200 years ago, is unlikely to be verified nor disproved even by detailed sedimentology or 4. CONDITIONS THAT LED TO AVALANCHES archeological methods. IN 1868 Westem Norway is fortunate to have a good record of avalanche events since the "land 4. 1 Weather conditions rent", a measure of taxability, was assessed for individual farms on the grounds of, among other The disasters of 1868 have been linked to things, their vulnerability to avalanche hazard a unique combination of a cold early winter Wilb (Grove and Battagel, 1983). Claims of tax reduc prevailing storms from the Norwegian Sea foIl_ tion could be made after avalanche damages to a ing in February. The weather data available, farm, and all such claims were assessed by official although limited, support this (Kristensen, 1999). tax commissions and thus recorded for the after It is likely that a widespread instability in tile time. A voluminous and fairly accurate record of mountain snow cover developed during the eatt avalanche damage from around 1650 to 1815 of winter because of a shallow snow cover and krf exists. After 1815 the tax assessment system temperatures. In this period a snow cover was changed, but similar data can be extracted from present also in the lowland. From around JTid. other sources. The records available show sev January, the low-pressure activity over W~ eral major avalanche winters during the latter part Norway increased markedly with Northwest winds of the nineteenth century and large amounts of snow. The extraordil'Vly snowfall is mentioned in many eyewitne$ll 3. THE AVALANCHES IN 1868 accounts from the time. Since local weather data from Stryn lie The most disastrous avalanche winter in not available, it is difficult to estimate the retlJII terms of loss of lives occurred in the winter of period for similar weather situations. Preliminay 1867/68 when a total of 161 people were killed. analyses of data from nearby stations indicate ttd This winter, avalanches started running down into there is hardly any reason to conclude that ti! the inhabited areas of the north region of Westem conditions were sufficiently unique to not hat th Norway (figure 1) on February the 6 , with several occurred several times during the preceding 2lI disasters following the next twenty days until the years with known settlement or during the tid th 26 . The communities that suffered the greatest after 1868 (Kristensen, 1999). In an analysi~' losses were Stryn with 35 and Oppdal with 32 major avalanche winters by Fitzharris aad fatalities. All of the victims were caught inside Bakkeh0i (1986) several winters with similar S'II' houses that were destroyed by the avalanches optic characteristics as the 1867/68 winter (Kristensen, 1998). reported. The avalanches of 1868 in Stryn have been examined closer (Kristensen, 1999). 4.2 Other factors Avalanches struck clusters of farms that both had, and did not have a previous history of avalanche The socio-economic conditions in NolVlIIf damage. One of the settlements affected (13 in the 1860-ies were very different from toda'tl victims) had been struck by an avalanche in 1718, The 1860-ies were generally hard times in 150 years before. However neither of the other West-Norwegian countryside. The size d two settlements, Gj0rven, where four farms were popUlation had reached a level where the destroyed and 11 people lost their lives, nor production could hardly keep up. This led to Tenden, where the destruction of two farms also intensification of the agricultural practices claimed 11 victims, had any history of avalanche involved large-scale use of the mountain
476 f forest cutting and for grazing livestock on tion is if and how this influenced the avalanche ormmer pastures above the timberline. A large activity. su rtion of a farm's production was actu~lIy m~ved m the valley floor up into the mountains. Since r 5000 t~~ Number of cotters. production of the sum":,er dairy farms near the SF, Western Norway. timberline increased, so did the demand for fire 4000 (Timberlid 1981) WOOd for heating and making brown whey cheese. There are several contemporary accounts of the 3000 lack of forest, which sometimes led to abandon 2000 ment of the summer dairy farms.
An increasing number of cotters (Iandren 1000 ters usually farming the outskirt~ of a r~ular farm) sometimes led to settlements In marginal areas, 0 no doubt also leading to increased exposure to 1800 1850 1900 1950 2000 avalanche danger. 250 The forester A. T. Gll2lersen (1868) com Fatalities in buildings affected by avalanehes (10 year sums). ments that although the winter conditions were 200 WeslBrn Norway (SF,MR,HO,RO). special, they were not all that unusual. Simil~r snow and weather conditions had been experi 150 enced before without the same dire conse 100 quences. A new and unique factor must therefore have been added, he argues. 50 Gll2lersen leaves no doubt about what he thinks this is: "I have come to the conclusion that 0 a large number of the avalanches last winter is to 1800 1850 1900 1950 2000 blame on the thoughtless, fast spreading destruc tion of the birch forest in the upper mountain Figure 1. Graph showing the number of cotters in meadows" (translation by the authors). Sogn og Fjordane (SF) and the number of The seriousness of situation was can only avalanche fatalities (in the four West-Norwegian be inferred by indirect data, since quantified forest counties of Rogaland, Hordaland, Sogn og data before 1900 are not available. Photographs Fjordane and Ml2lre og Romsdal). from the late nineteenth century show many slopes almost without trees, that are covered with dense forest today. Forest inventories done in the 4.3 Effect on avalanches by forest twentieth century show a dramatic increase in the growing stock in the four West-Norwegian coun The paths of the 1868 avalanches in Stryn ties of Rogaland, Hordaland, Sogn og Fjordane are now covered by dense forest for up to 2/3 of and Ml2lre og Romsdal during the last century; 3 3 their elevation. It is well known that forest in the from 18 mill. m in 1930 to 81 million m in 1998 starting zone, when dense and strong enough, can (Tomter ed., 2000). In the 1930-ies however, the prevent slab avalanche formation. Gll2lersen was pressure on the countryside had already been of the opinion that lack of forest in the starting greatly reduced compared to sixty years earlier, zones was the main cause of the avalanches in because of a large-scale emigration to the USA 1868. However, the avalanches in Stryn may well and the generally increasing industrialization of have started above the natural timberline and it is Norway. Statistics of grazing domestic animals less clear what the effect is on an avalanche that (cattle, sheep and goats) also point in the same enters a forested slope from above. Two effects direction. In the Stryn parish the number was may be possible. First, there may be a small reduced by almost 50% from 1865 to 1907 braking effect on the flowing part of the avalanche. (Timbertid, 1988). Also the peak in the number of Salm (1979) estimates that there will be an effi cotters around 1870 (Timberlid, 1981) supports cient reduction for high tree densities only, e.g. that the population pressure on the countryside more than 1000 trees per hectare. The avalanche was especially high during this period. forces acting on the stem of a tree depend on the Thus it seems clear that around the 1860 velocity and density of the flowing snow, and the ies there had been an unprecedented exploitation drag coefficient around the tree. According to and deforestation of the mountain slopes in the Salm (1979), the energy loss is relatively small. pOpulated valleys of Western Norway. The ques-
477 Second, a forest will influence the stratifi (Harbitz et. ai, in press) with respect to Pf, vm~c. cation of the snow cover as the canopy of the for denotes the cumulative binomial distribution est intercept the falling snow. The snow in a forest parameters nand pf, and where the argument r. is also less subject to wind transportation and the is the actual number of observed avalanches. TIl microclimate near the snow surface will be differ equation above must be solved numerically. ent from an open slope because of the forest's effect on in- and outgoing radiation. The result is 5.2 The probability of extreme runout an overall rise in the snow stability that may reduce the avalanche's entrainment and volume The annual probability of extreme rull- 478 5.3 Example calculations The path to be analyzed below is from 1200 Gj"rven, Stryn, where observations have been made for 350 years (figure 2). In 1868, with other forest conditions than today, an extreme event with an observed a-angle aobs = 29.4° occurred. Except for this observation, there are only few observations of avalanches with run-out angles higher (Le. shorter runout) than the p -angle of 31.5°. These observations are not considered important, except for the fact that they indicate a o 200 400 600 800 1000 1200 1400 1600 1800 2000 2200 2400 significantly lower probability of extreme runout in m other periods. They should be neglected when Figure 2. Terrain profile with observed and probability of release are discussed in combination estimated runout. with the alI3 -model that is based on a database that excludes avalanches with short run-out dis examples. The choice of an upper limit of a 95% tances. confidence interval as an estimator for Pf is rather It is now of interest to analyze how the strict, and an alternative is to choose e.g. a 50% 1868 event may influence potential hazard zones confidence interval instead. This would be more in in the area. Three different cases can be ana agreement with common hazard zoning practice, lyzed: but may not be considered conservative enough when so few observations are available. In addi 1) the observation is considered relevant also tion, appropriate alternatives to the mean in the under current conditions, Le. one avalanche Gumbel distribution are the less conservative has occurred over the last 350 years (since mode and median. As an example, the mode of 1650); the extreme value distribution is the parameter 2) the observation is not relevant today, i.e. no chosen in the 1OO-year wave concept related to avalanches with runout past the jJ-point ocean wave dynamics. In our case, the choice of have occurred during the last 350 years; mode would give approximately 10 higher a 3) the observational period is reduced to cover values in the above calculations. only the period with today's conditions, Le. It should be noted that in these example no avalanches have occurred over the last applications, the statistical uncertainty associated 100 years. with the statistical/topographical alI3 -model is neglected. This is justified by the relatively large Case 1 probably represents the most amount of observations and the fact that the 13 common way of using a historical observation. value in our calculation example is relatively close For case 1 the p,...estimate from Eq. (1) is Pf.O.95 = to the mean of the observed jJ-values. 1.35%, corresponding to a return period M, = 74 years, Le. considerably less than 350 years. In 6. ACKNOWLEDGEMENTS this case ms =1000' 0.0135 =13.5, from Eq. (3) b = -4.~, and finally from Eq. (2) as1 =0.9613 - 6.1 0 This paper is a contribution to the research project =24.3° is a possible estimate for a "safe area". "SIP6 20001018.820 -Risk analysis", funded by This estimate corresponds to the mean 100o-year the Research Council of Norway, and the EU- pro avalanche. In practical terms, we are "95% cer gram "Catastrophic Avalanches, Defense Struc tain" that the unconditional avalanche frequency is tures and Zoning in Europe", partly funded by the at the most 10-3lyear at a =0.9613 -6.1 0 when one European Union under the contract no. EVG1-CT avalanche is observed in 350 years. 1999-0009. . For cases 2 and 3 (no avalanche observa tions) the Pf estimates become 0.95% and 2.95%, respectively. This corresponds to b = -4.0°, aS2 =25.0° for case 2, and b =-6.1°, aS3=22.9 for case 3, i.e. considerably different from the first case. It should be emphasized that considerable COnservatism is built into the above calculation 479 7. REFERENCES Bakkeh0i, S., Domaas, U., and Lied, K. 1983. Cal Lied, K. and Bakkeh0i, S., 1980. Empirical Calcu. culation of Snow Avalanche Run-out Dis lations of Snow-Avalanche Run-Qut tls. tance. Annals of Glaciology 4, 24-29. tance Based on Topographic Parametres. Fitzharris, B. B., Bakkeh0i, S., 1986. A Synoptic Joumal of Glaciology 26 (94), 165-1n Climatology of Major Avalanche Winters in McClung, D.M., 2000. Extreme avalanche ru~ Norway, Joumal of Climatology, Vol. 6, in space and time. Can. Geotech. J. 37 pp.431-446 161-170. • Gl0ersen, A.T., 1868. Sn0skred og Skov0delaeg Salm, B. 1979. Snow Forces on Forest P1ams. gelse paa Vestlandet, in: Morgenbladet 4. Proceedings. IUFRO Seminar Mountain juni 1868 Forests and Avalanches, Davos 1978. pp. Grove J. M. and Battagel, A., 1983. Tax records 157-181 from Westem Norway, as an index of Little Timberlid, J. A 1981. Innleiing, "For Pladsen skaI Ice Age environmental and economic de han slide.." Husmannsskipnaden i Sogn terioration, Climatic Change, 5, 265 og Fjordane. Tiden Norsk Forlag Harbitz, C., Harbitz, A, Nadim, F, in press. On Timberlid, J. A 1988. Driftsendringar i jordbruket probability analysis in snow avalanche somarsak til forsuring av norske vass hazard zoning. Annals of Glaciology 32. drag? Ein samanliknande studie av lit Kristensen, K. 1998 A Survey of Snow Avalanche markstbruket pa Vest- og S0r1andet i pen. Accidents in Norway. 25 Years of Snow oden 1850-1980. 0koforsk rapport 1988:4 Avalanche Research, Norwegian Tomter, S. M. (ed) 2000. Skog 2000. Statistics cI Geotechnical Institute. pp 155-159 forest conditions and resources in Norway. Kristensen, K. 1999. Skredvinteren 1868 i Stryn. NIJOS H0gskulen i Sogn og Fjordane 480