SNAPS. Northern Periphery programme. SNAPS Work Package 2: Weather and avalanches. Prepared by. Issue / Revision: 1 / 0

Transcription

1 SNAPS Northern Periphery programme SNAPS Work Package 2: Weather and avalanches Prepared by Magni Hreinn Jonsson and Harpa Grimsdottir, IMO, Iceland Hedda Breien and Krister Kristensen, NGI, Norway Amin Zeinali, LTU, Sweden Issue / Revision: 1 / 0

2 Document controlled by: Harpa Grímsdóttir Page i

3 SNAPS Report GRANT AGREEMENT NR: SUBJECT: Weather and avalanches PROJECT COORDINATOR: IMO ISSUE / REVISON: 1/0 CONTRACTOR S REF: Abstract: The objective of this part of the SNAPS project was to study the correlation between weather and avalanches. Weather observations and avalanche records were compared. A modified version of the nearest neighbour method was used to analyse the Icelandic data but simpler methods for the other countries, were the amount of data was not sufficient for detailed analysis. Critical weather patterns were identified. It was shown, as expected, that precipitation is the factor that plays biggest role in avalanche activity. The avalanche hazard increases with increased wind speed but the effect both of wind speed and precipitation varies greatly with wind direction. This project was funded by EU Interreg AUTHORS: Jonsson, M., Grimsdottir, H., Breien, H., Kristensen, K., Zeinali, A. Page ii

4 Table of Contents 1 Introduction The target areas Iceland... 3 Norway RV 15 Strynefjellet FV 91 through Breivikeidet... 5 FV 63 Geiranger Eidsdal Sweden Avalanche and weather data Iceland Avalanche data Weather data Norway RV 15 Strynefjellet FV 91 through Breivikeidet FV 63 Geiranger Eidsdal Swedish target area Correlation between weather and avalanches - methods Wind direction Results for Iceland Opinion of local experts Statistical analyses Results from Súðavíkurhlíð Wind direction Wind speed Precipitation Wind speed and direction Wind direction and precipitation Wind direction, speed and precipitation Temperature Results from Kirkjubólshlíð Wind direction Page iii

5 5.4.2 Precipitation Wind direction and wind speed Wind direction and precipitation Wind direction, wind speed and precipitation Examples Conclusions Results for Norway RV 15 Strynefjellet Precipitation Wind Conclusions FV 91 through Breivikeidet January 1983, 12:30 Graselva th of March 1997, 01:00 Graselva april 1997, 11:50 Graselva Conclusions FV 63 Geiranger Eidsdal Conclusions Results for Sweden Discussion Conclusions References Page iv

6 1 Introduction This report is a part of the SNAPS project, which stands for Snow, Ice and Avalanche Applications The main goal of the project is to improve the efficiency and safety of transport routes in the Northern Periphery where snow and avalanches are a problem. SNAPS is funded by EU s Northern Periphery Programme (NPP). Main partners in the project are: Finnish Meteorological Institute (FMI), Luleå Technical University (LTU) in Sweden, Norwegian Geotechnical Institute (NGI), Northern Research Institute (Norut) and the Icelandic Meteorological Office (IMO) which is the lead partner. Transport authorities in all partner countries are associated partners, as well as local authorities and a few other organizations. The report describes results on analyses of weather and snow avalanches for target roads in Iceland, Norway and Sweden, and is a part of work package 2. The purpose is to create a tool that is useful for avalanche forecasting for these roads and the methodology may be used for other areas as well. The analyses differ somewhat between the three countries. Better weather and avalanche data exist for the Icelandic target road than the other areas and, therefore, a more detailed statistical analysis was possible. The Icelandic target road has more than 40 defined avalanche paths that pose danger to travellers on the road with frequent avalanches. In Norway, three target roads are included but the number of avalanche paths is lower than in the Icelandic target area. In Sweden, the target road runs to an area with cottages and a hydropower plant. The problem consists of wet avalanches and slush flows overrunning the road, and the dataset is quite small. Page 2

7 Issue / Revision 1/0 2 The target areas 2.1 Iceland The road between the towns Ísafjörður and Súðavík is the target road in Iceland for the avalanche forecasting work in SNAPS. It runs beneath two mountain sides: Súðavíkurhlíð and Kirkjubólshlíð. Each of them has many avalanche paths and avalanches are an essential problem on this route, they close the road and pose danger to travellers. The road is hit by an average of around avalanches each winter. The road is a part of the route from Ísafjörður to Reykjavík and is the only way for transport of goods to and from Ísafjörður over the winter time. Furthermore, it connects towns that are a part of the same service and employment area. Like in many other places, the nature of the traffic has changed over the years. Earlier, individuals living in those towns did not have to travel the road frequently, but now many people drive it every day for work, recreation and services. Figure 1. Number of avalanches from each avalanche path on Súðavíkurhlíð Avalanches hitting the road are most often a direct effect of weather, especially precipitation and wind. In order to predict those avalanches it is, therefore, important to analyse the weather leading to avalanches. However, in some cases the stability of the snow might be an essential factor. Page 3

8 2.2 Norway In Norway the SNAPS project focuses on 3 target roads; a part of RV15 Strynefjellet, FV63 Geiranger- Eidsdal and FV 91 Breivikeidet (Figure 2). Figure 2. The three target roads marked with black rings RV 15 Strynefjellet This road is one of the main transportation routes between west and east. The avalanche prone area is on the western side of the mountain pass, and there are starting zones in several different expositions. Especially the area below Sætreskarsfjellet (Grasdalen) and outside the tunnel in Skjerdingsdalen have throughout the years been hit by avalanches and resulted in closures of the road. Page 4

9 Figure 3. The RV15 Strynefjellet is avalanche prone in several different areas. Weather station is marked with a star FV 91 through Breivikeidet This is known to be an avalanche prone road, but only 5 avalanche occurrences are recorded with dates in the directory of the Road authorities. Due to this, no statistical analysis on weather and avalanche occurrence can be made, but in the paper we take a closer look at the weather prior to some of the avalanches. All these registered avalanches hit the road approximately where the river Graselva crosses the road. The avalanche starting zone is a bowl and catches snow with wind from W-N-NE directions. The target road is marked on the map below (Figure 4), and all the other starting zones in the area have similar exposition. Page 5

10 Figure 4. Breivikeidet. Page 6

11 FV 63 Geiranger Eidsdal This is the only winter access road to the small community of Geiranger at the end of the Geiranger fjord (Figure 5). Alternative access is by an extraordinary ferry service in case of prolonged closures. Figure 5. The Geiranger-Eidsdal area. 2.3 Sweden The target road, Ritsem Road, is located in the north-western part of Sweden. It is a 52 km private road and most of it is within the Stora Sjöfallets National Park. This road is the access road to the Ritsem area as well as Suova hydropower dam and Stora Sjöfallet National Park. Dry avalanches have fallen on the road but they have not been recorded. The main problem for this road is slush avalanches. More than four hundred meters of the road have been exposed to two huge slush avalanches in three years. The avalanches caused a major problem to access the Ritsem area since the road is a deadlock, but fortunately nobody was injured by the avalanches. Page 7

12 3 Avalanche and weather data 3.1 Iceland Avalanche data The Icelandic Road and Coastal Administration has recorded avalanches that hit the roads on Súðavíkurhlíð and Kirkjubólshlíð since the 70 s. In this study we use avalanche records from January 1996 until spring After 1996 the avalanches were recorded in a more systematic way than before and, furthermore, precipitation records used in the statistical analysis are only available from Therefore, earlier records are excluded. The number of recorded avalanches on Súðavíkurhlíð is 452 and 63 in Kirkjubólshlíð. Only avalanches that actually hit the roads are recorded, not those that came to a stop above the roads. The timing of the avalanches is not exact in the records, but the interval in which avalanches could have occurred is recorded. It is very uncommon that an avalanche is observed as it runs. While the road is open road workers or users may observe avalanches. But while the road is closed due to bad weather, snow on the road or avalanche hazard no avalanches are recorded. It is after such periods, while the road is being cleared, that avalanches are recorded. In these cases the avalanches are often recorded with fairly large time interval which spans the closure time of the road. For statistical work, avalanche cycles rather than individual avalanches were used. Each cycle consists of avalanches that are recorded in the same time interval or in overlapping intervals. On Súðavíkurhlíð there are 112 recorded cycles and 23 on Kirkjubólshlíð. It is clear from these numbers that the avalanche activity on Súðavíkurhlíð is higher than on Kirkjubólshlíð. On average, 26 avalanches hit Súðavíkurhlíð road in more than six cycles each winter. In Kirkjubólshlíð 3.6 avalanches hit the road on average in just over one cycle every winter. Out of 23 cycles on Kirkjubólshlíð there were six with no avalanches on Súðavíkurhlíð. In four cases the first avalanches fell on Kirkjubólshlíð followed by avalanches on Súðavíkurhlíð. In the other 13 cycles the first avalanches fell on Súðavíkurhlíð followed by avalanches on Kirkjubólshlíð. The avalanches are recorded in predefined avalanche paths identified by numbers. Figure 6 shows number of avalanches in each path on Súðavíkurhlíð. The paths are 22 in total, and out of them six have more than 20 recorded avalanches and 17 paths have 10 or more recorded avalanches. This indicates that the difference between paths is not great and it is difficult to predict which paths will avalanche in a given weather situation. Page 8

13 Number of avalanches Issue / Revision 1/ Avalanche path Figure 6. The number of recorded avalanches in different avalanche paths on Súðavíkurhlíð Weather data Data from three weather stations are used for the analysis. The mountain weather station Thverfjall is used for temperature, wind speed and direction. It benefits of being on a mountain top, at 753 m.a.s.l., and therefore the recorded wind is not as affected by landscape as weather stations down in the fjords. Precipitation is only recorded in the lowland. When studying avalanches on Súðavíkurhlíð precipitation measurements from Súðavík village were used. The precipitation gauge is located about one kilometre from the closest avalanche path on Súðavíkurhlíð. For studies on Kirkjubólshlíð precipitation from the town Ísafjörður is used. The gauge is about 1.5 km from the closest path. The weather data has one hour resolution. It means that one hour average is calculated for temperature, wind speed and direction, but the precipitation is recorded cumulative. 3.2 Norway Avalanche occurrence data were collected from the Road Authorities and weather data from nearby weather stations analysed RV 15 Strynefjellet The NGI has been doing avalanche research at Strynefjell since the beginning of the '70s and also operates an automatic weather station in the area, located around 1000 m.a.s.l. During the period 1974 to avalanche cycles were recorded by NGI. A report published in 1984 contains an analysis of the weather leading to these avalanches. In our study we have used the findings from these first years and complemented them with the road authorities' avalanche registrations from the years During these last years there are 15 avalanche registrations. It should be noted that avalanche observations from the road maintenance operation only include avalanches that have reached the road. The exact timing of many avalanches is not known, as in some cases only the dates of the avalanches have been recorded. Page 9

14 3.2.2 FV 91 through Breivikeidet Precipitation data are available from two stations each situated around 25 km W (90450 Tromsø) and 25 km E (91270 Lyngseidet) of the target road. Be aware that the measurements at Lyngseidet are only done once every day (07:00) and reflects the total precipitation during the prior 24 h. This means that precipitation measured on the 23 rd is precipitation fallen between 07:00 on the 22 nd and 07:00 on the 23 rd. In Tromsø the precipitation measurements are done twice a day, but are sometimes missing. The data from Tromsø reflect the total precipitation during the prior 12 h. Table 1 Weather stations, parameters and measurement interval TROMSØ LYNGSEIDET PRECIPITATION 12h (07 and 19) 24h (07) TEMPERATURE 6h (07,13, 19) WIND 6h Note that weather coming in from the ocean most likely will reach Tromsø (where the weather data is measured) before it reaches Breivikeidet. This means that the exact timing of the weather data and the avalanche observation are not directly correlated FV 63 Geiranger Eidsdal There are unfortunately very few weather stations in the vicinity of the target road. The map below shows all weather stations in the area, but some of them are new and thus not usable in statistical analyses, and only the Tafjord station is a complete weather station measuring wind direction, wind speed, temperature and precipitation. Tafjord is however very seldom representative because of particular topographic features. Due to this, only a limited weather-avalanche analysis has been performed for this target road. An SM4 snow height sensor was installed in the area in Page 10

15 Figure 7. Weather stations in the area. Target area marked approximately by red ring. 3.3 Swedish target area Two big slush avalanches have hit the road in the last four years. Before that no records exist, however, it is known that dry avalanches have hit the road but the timing is not known. The two slush avalanches fell on 15 th of May 2010 and 17 th of June 2012 and destroyed about 500 m segment of the road. In both cases the road was closed for over two weeks while blocks were being removed and the road was re-paved, see figure x. The cost of removing blocks and repairing the road after the avalanches in 2010 and 2012 is about 4 million SEK in total. The cost of the asphalt layer in 2012 is not included in costs. Weather observation from Ritsem is used in this study. The data is stored by Swedish Meteorological and Hydrological Institute (SMHI) and is available from Figure 8. Slush avalanche in Page 11

16 4 Correlation between weather and avalanches - methods The best way to find correlation between weather and avalanches for places with only few recorded avalanches is to manually look at the weather leading to each avalanche and look for common patterns. This, along with knowledge from local experts, can be used to distinguish typical weather patterns leading to avalanches in each area. This method was used for areas with insufficient data for statistical analysis. The method was also used as a preliminary study before using statistical methods to help select weather parameters and time frames for the study. This was the case in the Norwegian and Swedish target areas. For the statistical analysis in Iceland a modified version of the nearest neighbour method (NNM) is used. In short, a computer program was created that counts occurrences of predefined weather pattern category or categories and finds which of these occurrences are within the timeframe of avalanches. The results are displayed as the proportion of weather occurrences that are linked to avalanches. The weather categories are defined by restraining one or more weather parameters. For example, one can define a category with one hour average wind speed above 20 m/s and more than 10 mm precipitation for a 6 hour time period. Each weather record is classified with the category it fits into. The weather data we used was sampled with one hour interval. Adjacent records were joined together, in one weather situation, and counted as one if they did fall in the same category but this was constrained to six consecutive records or six hours. If an avalanche period is recorded within six hours from beginning of a weather situation the avalanche period is linked to that weather. The length of the time period of weather situations is limited in order to make the results with different categories better suited for comparison. For example, one could imagine a weather category consisting of wind speed above 12 m/s and another one with one-hour precipitation more than 5 mm. The former happens frequently and can last for hours or days at a mountain weather station while the latter is a rarer event and only lasts for few hours. A weather situation that lasts for a longer time period is by nature more likely to be within a timeframe of an avalanche period than a weather situation that lasts for a shorter period and, therefore, comparing the results from these two would not be correct. Furthermore, all the weather situations are linked to avalanche cycles within six hours from their beginning regardless of the timespan of the weather situation. This method, along with the fact that the exact timing of the avalanches is not known, leads to a possibility of each avalanche cycle being counted more than once both within the same weather category and also in two or more different categories. The impact of this does not seem to be great at least not for the weather categories that are of most interest as they tend to be rare and do not last for a long period of time. In some instances it is of interest to break some of the rules of the method that are stated here above. This is especially true when it comes to precipitation over long time period (more than 6-12 hours). In that case, it is of interest to look not only at avalanche occurrences after the weather record but also during the precipitation. This was done in some cases. For shorter duration of precipitation the influence of this is negligible. Page 12

17 4.1.1 Wind direction One of the most defining features of a given weather category is the wind direction. Special attention was thus given to the way avalanche cycles are linked to wind direction. Three different methods were used to define the wind direction for each avalanche cycle: 1. Sorted manually. The weather leading up to each cycle was studied and one and only one wind direction is assigned to each cycle. In many cases this is not a trivial task as wind directions are frequently changing. The most important wind direction in each case was an expert s opinion. It could be the wind direction with the longest duration, with greatest precipitation, highest wind speed or the wind direction just prior to the start of the cycle. In this case each cycle is only counted once. 2. Sorted cycles. The computer program is used in same manner as explained before except that each avalanche cycle can only be linked to the wind direction category it was assigned to manually. In this case each cycle can be counted more than once but only in its wind direction category. 3. All cycles. The computer program is used without any modification. In this case each cycle can be counted more than once and in more than one category. Page 13

18 5 Results for Iceland 5.1 Opinion of local experts A meeting was held with people from The Icelandic Road and Coastal Administration (ICERA) that have good knowledge on avalanches on the target road. The meeting was attended by: Magni Hreinn Jónsson and Harpa Grímsdóttir from IMO and Geir Sigurðsson, Guðmundur Björgvinsson and Jón Baldvin Jóhannesson from ICERA. Súðavíkurhlíð The people from ICERA feel that the avalanche danger is greatest during precipitation in North- Westerly winds. However, there is also risk of avalanches during precipitation in wind from NE-NW and SW. Usually there is no avalanche danger in calm weather. Furthermore the risk is low when it snows on bare ground, and during such conditions it has to snow for a long time before the danger picks up. Avalanches are in most instances connected to high wind speed and snowdrift. However, the people from ICERA are not convinced that snow drift alone triggers avalanches. Kirkjubólshlíð The greatest avalanche danger on Kirkjubólshlíð is thought to be in connection with winds from NE and during heavy precipitation, danger may occur in winds from N as well. Usually avalanches don t hit the road on Kirkjubólshlíð until after avalanches have already hit the road on Súðavíkurhlíð. The avalanches in Kirkjubólshlíð might need twice the amount of precipitation on average. However, avalanches have hit Kirkjubólshlíð without any avalanches in Súðavíkurhlíð after heavy precipitation in winds from E. In South-Easterly wind directions, a danger may occur on a part of the road but that is uncommon during the winter. In wind directions from SW the precipitation is usually not great in this part of the country. In Kirkjubólshlíð it is not uncommon to have more than one avalanche from the same path and avalanches have fallen there after the weather has calmed down, even more than 12 hours later. 5.2 Statistical analyses In the beginning it should be noted that there are differences in the frequency of different weather categories and the results below are, thus, based on different number of weather occurrences. The results are not displayed on graphs unless they have more than 15 cases of a weather situation behind them. For most of the analyses, the temperature was restrained to below -1 C on Tverfjall. At temperatures below this, precipitation falls as snow in the starting areas of avalanches. Six-hour average temperature was used but it did not seem to make much difference if a shorter or a longer average period was used. 5.3 Results from Súðavíkurhlíð For the analyses of avalanches on Súðavíkurhlíð 112 cycles from 1996 to 2013 were used. Precipitation measurements are not available for the whole period so 81 cycles are used for the analyses of precipitation. Page 14

19 Number of avalanche cycles Issue / Revision 1/ Wind direction Figure x shows the number of avalanche cycles associated with different wind directions. Here, the wind directions are divided into four sectors, 90 each. As the most common wind directions in this area are NE and SW (Figure 9), it was decided that it would be appropriate to group them into NE, SE, SW and NW sectors. Figure 10 shows that the most common wind directions are associated with the greatest number of cycles and NW and SE have 9 cycles each NE SE SW NW Figure 9. Number of avalanche cycles during different wind directions. Page 15

20 Figure 10. Wind rose for Tverfjall when temperature is below 1 C. Figure 11 shows the avalanche proportion for each wind direction counted with the three different methods explained in Chapter Wind from north-west have according to this the highest ratio of avalanches. This is in accordance with the feeling of road administration staff. There is great difference between those three methods with sorted manually showing the lowest proportion and all cycles the highest proportion for all wind directions. This is as expected because the cycles, when they are sorted manually, are only counted once. When the NNM is used to assign those cycles to weather records they can be counted more than once (due to the duration of both weather situation and avalanche cycles). When the NNM is used on unsorted cycles each cycle can be included in more than one wind direction. The effect of this is very noticeable for NW-winds which are not very common. The wind direction only has to turn to NW for a brief moment for avalanche cycles to be defined as a NW cycle and likewise for other wind directions. This scenario is similar to that described by the road administration staff that only short duration of wind from NW can trigger avalanches. To observe this phenomenon the NNM has to be used on unsorted cycles and that is the method that will be used for following analysis. Page 16

21 Avalanche proportion (%) Avalanche proportion (%) Issue / Revision 1/0 5.00% 4.50% 4.00% 3.50% 3.00% 2.50% 2.00% 1.50% 1.00% 0.50% 0.00% NE SE SW NW Sorted manually All cycles Sorted cycles Figure 11. Avalanche proportion for different wind directions. Figure 12 shows the avalanche proportion associated with each sector of wind direction as before, but now only instances when wind is above 5 m/s are included. Obviously, the proportion increases somewhat and reaches 7% for the NW sector. That is interesting given that the only restrains are that the wind is from NW, the speed greater than 5 m/s and the temperature below -1 C. Figure 13 and Figure 14 shows the avalanche proportion when the wind direction has been split into eight sectors. Now the NNW stands out followed by NNE. It is interesting to notice how small the proportion is for ENE, less than one third of that of NNE. WSW and WNW show similar results. 8.00% 7.00% 6.00% 5.00% 4.00% 3.00% Sorted manually All cycles Sorted cycles 2.00% 1.00% 0.00% NE SE SW NW Figure 12. Avalanche proportion for different wind directions when wind speed is above 5 m/s. Page 17

22 Avalanche proportion (%) Avalanche proportion (%) Issue / Revision 1/0 6.00% 5.00% 4.00% 3.00% 2.00% All cycles Sorted cycles 1.00% 0.00% NNE ENE ESE SSE SSW WSW WNW NNW Figure 13. Avalanche proportion for different wind directions % 9.00% 8.00% 7.00% 6.00% 5.00% 4.00% 3.00% 2.00% 1.00% 0.00% NNE ENE ESE SSE SSW WSW WNW NNW All cycles Sorted cycles Figure 14. Avalanche proportion for different wind directions when wind speed is above 5 m/s Wind speed Figure 15 shows how the avalanche proportion increases with increased wind speed. In winds below 10 m/s the proportion does not change much with wind speed. Between 10 m/s and m/s the proportion triples. It is interesting to see how the proportion decreases around 25 m/s. The reason for the decrease after 25 m/s is not known, but one explanation could be found in Figure 16. It shows how the average precipitation in Súðavík changes with wind speed and there is a drop in precipitation above 25 m/s. However, precipitation measurements are in general less accurate in high wind speeds and indicate less precipitation than it actually is. Figure 16 shows both the actual measurements from Súðavík and Ísafjörður and preliminary corrected precipitation measurements were wind speed and the type of precipitation at the observations sites are taken into account. This is done by method described in Crochet (1997). Page 18

23 Avalanche proportion (%) Issue / Revision 1/ One-hour average wind speed Three-hour average wind speed Six-hour average wind speed Wind speed exceeds (m/s) Figure 15. Avalanche proportion as a function of wind speed. Figure 16. Average hourly precipitation as a function of wind speed. This is for instances where temperature is below -1 C similar results were found for different temperature. Precipitation in Súðavík to left and Ísafjörður to right. Page 19

24 Avalanche proportion (%) Issue / Revision 1/ Precipitation Precipitation is the single most important weather factor when it comes to avalanche hazard. In Figure 17 avalanche proportion is shown as a function of six-hour precipitation. This is as before limited to conditions when temperature is below -1 C on Tverfjall and, therefore, the precipitation should fall as snow. The proportion reaches 25% when the precipitation is more than 10 mm in six hours. The undulating right part of the curve in the graph is due to a low number of incidents when precipitation is intense Six-hour percipitation exceeds (mm) Figure 17. Avalanche proportion as a function of six-hour precipitation Wind speed and direction The wind direction is an important factor that defines the character of the weather with regard to avalanche danger. Therefore, the following results will be split up according to wind direction (NE, SE, SW and NW). Figure 18 shows the proportion as a function of winds speed for the four different wind directions. When looking at only wind direction and wind speed the avalanche proportion reaches the highest values in NW- and SW-winds, over 20%. In these wind directions the release areas in the mountain are on the lee side and therefore they are wind loaded. In the case of winds from SW this happens during very high wind speeds. Strong winds from NW are so rare that only results up to 18 m/s can be shown for that direction. In some cases the avalanche proportion drops again in very high wind speeds. The difference between one-, three- and six-hour average wind speed is in general not great. Page 20

25 NW direction NE direction SW direction SE direction Figure 18. Avalanche proportion as a function of wind speed for different wind directions Wind direction and precipitation Figure 19 shows the proportion as a function of three- and 12-hour precipitation. As already shown the proportion increases with increased precipitation. The figure indicates as before that the avalanche danger is in general greatest in NW wind directions, followed by NE-wind directions. The southerly winds are trailing behind with SE being higher in this case. Higher proportion values are reached for the 12-hour precipitation especially for other wind directions than NW. Then the results for 24-hour precipitation (not shown here) are similar to the three-hour but the proportion for 72- hour precipitation is much less. This indicates that precipitation for a couple of hours, up to 24 hours and maybe somewhat longer is of greatest importance when it comes to avalanche danger. Figure 19. Avalanche proportion as a function of three-hour and 24-hour precipitation. Page 21

26 Figure 20 shows the avalanche proportion as a function of precipitation duration for different average intensity of precipitation. It is obvious from the figure and as expected that the proportion increases with increased intensity. But this differs between wind directions. In NE-winds the duration seems to be of more importance than the intensity at least when it is more than 1 mm/hour. The proportion is highest in NW wind directions and the danger is high even after a short period of time with heavy precipitation. The proportion also reaches a high value in NE wind directions. The duration is an important factor, after 8-12 hours the avalanche proportion goes up to 25%. SW and SE has similar proportion with SE being slightly higher. NW direction NE direction SW direction SE direction Figure 20. Avalanche proportion as a function of precipitation duration for different average intensity of precipitation Wind direction, speed and precipitation Two scenarios have been set up in Figure 21 where the avalanche proportion is displayed as a function of wind speed for different precipitation scenarios. Once again, NW-winds have the greatest proportion of avalanches in general. It is around 20% and 30% depending on the precipitation scenarios and independent of wind speed. The proportion for NE-winds is around 10% and 20% from 0-10 m/s. Between 10 and 24 m/s the avalanche proportion increases fast with increasing wind speed, it doubles and triples depending on precipitation scenario, and it exceeds the avalanche proportion for NW-winds. However, no results are shown for wind speed over 14 m/s in NW-winds due to lack of data. During SW-winds the avalanche proportion is 5 and 10% for the different precipitation scenarios from approx m/s. For higher wind speeds the proportion increases Page 22

27 somewhat and is close to 15% where it is highest. In SE wind directions the avalanche proportion is in general between approx. 6 and 12%. It increases somewhat with wind speed and reaches 16% where it is highest. In general Figure 21 indicates that the danger of avalanches is twice as high during NE winds than SW or SE winds. During NW-winds the danger is around times higher than during NE wind directions. The danger is especially dependent on wind speed in NE-winds. NW direction NE direction SW direction SE direction Figure 21. Avalanche proportion as a function of wind speed during different precipitation scenarios. Figure 22 shows the avalanche proportion as a function of six-hour precipitation for different wind strengths. The effect of wind strength is most apparent for NE winds. In other wind directions the difference between 5-13 and is very small. However, it is of importance whether the wind speed is below or above 20 m/s for all wind directions. Page 23

28 NW direction NE direction SW direction SE direction Figure 22. Avalanche proportion as a function of six-hour precipitation for different wind strength Temperature Until now the results have been restricted to six-hour average temperature on Tverfjall being below 1 C. At that temperature precipitation in the starting areas above the target road is expected to be in form of snow. Here, the focus is on the effect of temperature alone on the avalanche proportion. Cold weather can contribute to a weaker snowpack. The bounding and stabilisation is slower and weaker than at higher temperatures and it is more likely that weak layers such as facets will form due to high temperature gradient in the snow cover. A sudden increase in temperature alone can trigger avalanches. Figure 23 shows how the avalanche proportion changes when temperature is below or above a given value both for short- and long-term averages (six-hour and 5-day). The proportion increases with decreasing temperature above 10 C. Below that the proportion seems to drop, possibly due to decrease in precipitation at such low temperatures. Page 24

29 Avalanche proportion (%) Issue / Revision 1/ Temperature ( C) Six-hour average temperature upper limit Six-hour average temperature lower limit Five-day average temperature upper limit Five-day average temperature lower limit Figure 23. Avalanche proportion as a function of six-hour and five-day average temperatures. Both when temperature is constrained from above and below. Figure 24 shows the effect of temperature increase on the avalanche proportion. The trend is not stron. The state of the snowpack is of great importance when it comes to the effect of temperature increase. Better results might be obtained by studying both precipitation and temperature increase. Figure 24. Avalanche proportion as a function of temperature increase. The figure to the right shows results when maximum temperature is within 4 C. In the left figure temperature is not restrained. 5.4 Results from Kirkjubólshlíð In this chapter results for Kirkjubólshlíð are presented with a similar structure as for Súðavíkurhlíð. The frequency of avalanches hitting the road on Kirkjubólshlíð is much lower than for Súðavíkurhlíð and, therefore, the avalanche proportion is lower in general. We use the same weather data as before except that now the precipitation measurements are used from the Ísafjörður automatic station instead of Súðavík Wind direction In Figure 25 the avalanche periods are divided to different wind directions on Kirkjubólshlíð sorted manually (see explanations on sorting methods in Chapter 4.1.1). Winds from NE are associated with the highest number of avalanche cycles; other wind directions are only associated with two Page 25

30 Number of avalanche cycles Issue / Revision 1/0 avalanche cycles each. For the analyses 23 cycles were used but only 21 when precipitation is studied as precipitation measurements are not available for the same period as the avalanche records NE SE SW NW Figure 25. Number of avalanche cycles in different wind directions. Figure 26 shows the results from the three sorting methods. The avalanche proportion is lower than for Súðavíkurhlíð in all cases. The proportion in NE-winds is relatively higher than for Súðavíkurhlíð and the difference between NW and NE is not as great as for Súðavíkurhlíð. And we see that NE has the highest proportion in the sorted cycles method. As for Súðavíkurhlíð there is a considerable difference between the three sorting methods. The reasons for this are explained in Chapter The main reason is that the likelihood of counting avalanche cycles more than once and in more than one wind direction varies between the methods. Figure 27 shows the results when the wind is divided into eight sectors. It is interesting to compare this figure with Figure 14 where the same results from Súðavíkurhlíð are presented. NNW still has the highest proportion of avalanches but is now closely followed by NNE. Other directions are considerable lower in proportion with WNW having no avalanche cycles. This is different from Súðavíkurhlíð where WNW had a high proportion of avalanches, although not the highest. This was expected both due to avalanche observations and the different aspect of the mountains. Page 26

31 Avalanche proportion (%) Avalanche proportion (%) Issue / Revision 1/0 1.40% 1.20% 1.00% 0.80% 0.60% 0.40% Sorted manually All cycles Sorted cycles 0.20% 0.00% NE SE SW NW Figure 26. Avalanche proportion during different wind directions when wind speed is higher than 5 m/s. 1.80% 1.60% 1.40% 1.20% 1.00% 0.80% 0.60% All cycles Sorted cycles 0.40% 0.20% 0.00% NNA ANA ASA SSA SSV VSV VNV NNV Figure 27. Avalanche proportion during different wind directions when wind speed is higher than 5 m/s. The relation between wind speed and avalanche proportion is showed in Figure 28. This is similar results as for Súðavíkurhlíð. The proportion picks up around 10 m/s and triples from m/s but drops after that. This is the same as for Súðavíkurhlíð and possible explanations are less precipitation in winds above 25 m/s and less data. Page 27

32 Avalanche proportion (%) Issue / Revision 1/ One-hour average wind speed Three-hour average wind speed Six-hour average wind speed Wind speed exceeds (m/s) Figure 28. Avalanche proportion as a function of wind speed Precipitation Figure 29 shows avalanche proportion as a function of six-hour and three-day precipitation. The proportion is three times higher for three-day precipitation than six-hour precipitation. In the case of Súðavíkurhlíð this is the other way around and 3 to 24-hour precipitation gives higher proportion than three-day precipitation. This indicates that avalanches on Kirkjubólshlíð are the results of precipitation for a longer time period than on Súðavíkurhlíð. This is not surprising since the avalanche starting areas and paths on Súðavíkurhlíð are very steep all the way down to the sea. Avalanches release in the starting area after a short period of snow accumulation and the likelihood of even small avalanches reaching the road is high. In Kirkjubólshlíð the starting areas can accumulate a bigger amount of snow and the road is a little bit further away from the steep hillside. Figure 29. Avalanche proportion as a function of six-hour and three-day precipitation Wind direction and wind speed Figure 30 shows the avalanche proportion as a function of wind speed for the four wind directions. During NE-winds the avalanche proportion increases with increasing wind speed above 10 m/s which is similar to the results from Súðavíkurhlíð, even though the proportion is much lower. Other wind directions do not show that tendency clearly although the proportion rises with wind speed up to 10- Page 28

33 15 m/s in NW-winds. High wind speed from SW or SE does not seem to be a contributing factor in general for avalanche danger on Kirkjubólshlíð. NW direction NE direction SW direction SE direction Figure 30. Avalanche proportion as a function of wind speed for different wind directions Wind direction and precipitation Figure 31 shows the effect of 6-hour and 24-hour precipitation on avalanche proportion. There is quite a difference between those two. It has been concluded that long-term precipitation has more effect than short term on the proportion on Kirkjubólshlíð. This seems to be true but the effect is most apparent for the NW wind direction. SW direction is not associated with avalanche occurrences in either case. Figure 31. Avalanche proportion as a function of one-hour and three-hour precipitation. In Figure 32 the effect of precipitation duration on the avalanche proportion is shown. This is the average intensity and therefore it is not abnormal that the proportion increases from zero with Page 29

34 increased duration in the southerly winds. This indicates that in those cases avalanches have not occurred during or right after precipitation had stopped. The proportion grows both with increased intensity and longer duration. The exception of this is that no avalanches are associated with intensity being over 1 mm/hour in SW-winds. The avalanches associated with SW-winds did occur during no or little precipitation. NW direction NE direction SW direction SE direction Figure 32. Avalanche proportion as a function of precipitation duration for different average intensity of precipitation Wind direction, wind speed and precipitation Figure 33 shows the avalanche proportion as a function of 24-hour precipitation in different wind speeds and wind directions. The proportion is very low in southerly winds, within 1%. It is higher for NW-winds, around 5% and with an outlier close to 15%. The data are not sufficient to show results for wind above 20 m/s in NW-winds. The proportion increases both with increased wind speed, especially when it is above 20 m/s and precipitation during NE-winds. Page 30

35 NW direction NE direction SW direction SE direction Figure 33. Avalanche proportion as a function of 24-hour precipitation for different wind strength. Figure 34 shows the same as previous figure but now for 72-hour precipitation. The results for SE and NW are similar to previous results. The NE has higher proportion than before, the difference between wind speeds is greater and the proportion is very high in strong winds after heavy precipitation. The results for SW are interesting due to the big difference between 24-hour and 72- hour precipitation and also because 5-13 m/s is associated with the highest proportion by far. The difference between 24-hour and 72-hour precipitation for both SW and NE indicates that the avalanches are not necessarily falling during or right after the precipitation. It should also be noted that the results are based on average wind direction for six hours so the precipitation could have fallen during periods with other wind directions, with NE being as the likeliest. It could be concluded from this that long term precipitation causes avalanches rather than short term and especially if this is during or following strong NE-winds. It is hard to explain why lower wind speeds are associated with higher avalanche proportion during SW-winds. Page 31

36 NW direction NE direction SW direction SE direction Figure 34. Avalanche proportion as a function of 72-hour precipitation for different values of wind speed. 5.5 Examples This section contains four examples of avalanche cycles at the Icelandic target road and a description of the weather during the days before. The first example is a rather typical avalanche cycle on Súðavíkurhlíð with precipitation and wind changing to NW. The next three examples are not typical since the cycles do not start until after the precipitation has stopped. In some of the cases a part of the reason is found in the snow layering. 4 th of November 1999 On 4 th of November 1999 five avalanches fell on the road on Súðavíkurhlíð. The weather the days before is shown in Figure 35. Four days prior to the avalanches the total precipitation was over 20 mm, most of it fell 31 st of October and again 3 rd and 4 th of November. Most of the time the wind was from NE, however, around the time the avalanches released the wind was turning to NW. This is in accordance with the feeling of the road workers; that avalanches often run shortly after the wind turns from NE to NW. It was also observed in Chapter Error! Reference source not found. that short uration of NW wind direction could be enough to cause avalanches. When this cycle was given a wind direction manually it was NW as the wind was from NW at the time the avalanches occurred. It could have been sorted as NE cycle since the wind was coming from NE the days before. The NNM program sorts such cycles both as NE and NW. Page 32

37 Figure 35. Weather during the days prior to five avalanches in November The timing of the avalanches was recorded sometime between the two vertical dotted lines. The precipitation measurement is from Súðavík and other measurements from Tverfjall. The dates on the horizontal axis are on midnight. 28 th of February 2001 Figure 36 shows measurements of weather factors prior to seven avalanches that ran in the evening of 28 th of February The accumulated 72-hours precipitation when the avalanches ran was 13 mm. However, few hours passed after the precipitation stopped and before the avalanches started to run. It is surprising how many avalanches ran during such short time period when nothing special was happening in the weather. According to snow pit data, weak layers were forming in the snow pack during the cold days before the avalanches. A little snow drift in moderate winds could have Page 33

38 been enough to trigger the avalanches and maybe the subtle change in wind direction had some effect on the drift. But it is surprising that no avalanches fell on the road during the precipitation. Figure 37 shows the results from snow pits that were taken 25 th and 30 th March. The one from 25 th shows 20 cm of new snow on top of a melt-freeze crust, and the crust is rated as weak. Under the crust is a layer of rounded grains which is in general a strong crystal form. On the 30 th of March, 10 cm of new snow has been added on the top of the snowpack from the 25 th of March. The crust is still present, a layer with faceted grains has formed within the layer under the crust and the binding is poor between the new and old snow. The formation of facets is a consequence of the low temperatures, and the presence of the crust. This cycle was classified as NE both manually and by the NNM program. Figure 36. Weather during the days before a seven day avalanche cycle in March The timing of the avalanches was recorded sometimes between the two vertical dotted lines, the lines are close together so they might print as one vertical line. The precipitation is from Súðavík station and other measurements from Tverfjall. The dates on the horizontal axis are on midnight. Page 34

39 Figure 37. Snow pit data from Kistufell from 25th and 30th of March th of December 2004 Figure 38 shows the weather leading to five avalanches 25 th of December The precipitation the days before the cycle was slightly less than 15 mm. As in the previous example few hours pass after the precipitation stops until the avalanches starts to run. Snow pit data from 28 th of December show three weak layers with facets. Although it was cold in the days before the avalanches the weak layers are most likely older and in fact snow pit from 14 th December shows lot of layering and two facetted layers. It can be concluded that weak layers contributed to the avalanches but it is not clear why five avalanches were triggered in a relatively short time period, but not during the precipitation and stronger winds from NW. When the avalanches fell wind were picking up from ENE which is in general not a bad wind direction according to our previous findings. Page 35

40 Figure 38. Weather during the days before an avalanche cycle with five avalanches in December The avalanches were recorded sometime between the two vertical dotted lines. The precipitation is from Súðavík and other measurements from Tverfjall. The date ticks on the horizontal axis are at midnight. Page 36

41 Figure 39. Snow pit data from Kistufell from 28th of December. Page 37

42 10 th of January 2005 In the avalanche cycles described so far all of the avalanches occurred on Súðavíkurhlíð. Figure 40 shows the weather leading to an avalanche cycle with two avalanches on Kirkjubólshlíð on January 10 th (none on Súðavíkurhlíð). More than three days passed from the stop of the precipitation until the cycle starts. The precipitation was furthermore only 4 mm on the 7 th. The most likely trigger is the wind that was picking up right at the time the avalanches occurred. Figure 40. Weather in the days before two avalanches in January The avalanches were recorded to have fallen sometime between the two vertical dotted lines. The precipitation is from Ísafjörður and other measurements from Tverfjall. The date ticks on the horizontal axis are at midnight. These examples, although they are not typical cycles, show that avalanches may release after the precipitation has stopped and drifting snow alone can trigger avalanches. Furthermore, snow layering may be an important factor in some cases. Page 38

43 5.6 Conclusions The results for Súðavíkurhlíð show that precipitation is the factor that plays biggest role in avalanches on the road. The avalanche proportion increases with increased wind speed but the effect both of wind speed and precipitation varies greatly with wind direction. The worst wind direction is northwest and especially NNW but NE-winds are the most common and therefore the greatest number of avalanches occur during NE-winds. NW-winds The avalanche proportion is in general highest for NW-winds. Precipitation with high intensity can cause avalanches even though the precipitation does not last long. The proportion increases with increased wind speed as in other wind directions but NW-winds are so rare that data is lacking to study this effect when wind speed is above 20 m/s. And it should be noted that for other wind directions the avalanche proportion increases greatly if the wind speed is above 20 m/s. NE-winds Most avalanches release during or after NE-winds, but that is the most common wind direction in the area. Wind speed plays significant role especially if it is above 10 m/s. The precipitation intensity does not have as much influence as in NW-winds, although the proportion increases with both increased intensity and duration of precipitation. SE-winds SE-winds are both uncommon and have little avalanche proportion. However, some avalanches have run in SE-winds and the proportion reaches 15% when there is heavy precipitation for more than six hours. SW-winds Wind speed seems to be of greater importance than precipitation in SW-winds. The proportion reaches 15-20% when wind speed is around 25 m/s. Westerly winds should be unfavourable as release areas are wind loaded in that direction, so this is not unexpected. This should be kept in mind when there is loose dry snow and strong SW-winds. The avalanche proportion for Kirkjubólshlíð is, as expected, lower than for Súðavíkurhlíð. Precipitation for longer time is needed for avalanche releases and avalanches do not necessarily release during precipitation. Strong winds from NE during or after precipitation is the most favourable avalanche condition and the only condition in which the avalanche proportion for the two road stretches comes close. The factors that have been analysed in this study are wind speed and direction, precipitation and temperature. The state of the snowpack and the amount of snow can have great effect on avalanches. Although not studied here it can be assumed that the proportion and frequency of avalanches is higher when there are weak layers in the snowpack or lot of snow. Stable snowpack and little snow will on another hand reduce the proportion and frequency. Page 39

44 6 Results for Norway In this Chapter the findings for the target roads in Norway are presented. The aim was to use the Nearest Neighbour Model that was developed in the project and used for the Icelandic data. This was not completed due to lack of data and resources. Instead older statistical analyses were used and the weather prior to avalanche cycles was explored manually. 6.1 RV 15 Strynefjellet Precipitation It is clear that for most of the avalanche releases, heavy precipitation is the most important factor, however, the variation in precipitation prior to each avalanche is large. For the avalanche cycles between 1974 and 1984, 3- and 5-day precipitation sums are referred to, together with wind and temperature. For the avalanches after 1985, also 1-hour and 24-hour precipitation data are available. The instrumentation has changed somewhat during the years and periods of missing data exist. In general, all precipitation, including solid precipitation, is referred in mm of water. 24-hour precipitation prior to the avalanches varies from 7 to 69 mm ( , Sætreskarsfjellet). None of the recorded avalanches occurred on the first day of snowfall. For the avalanches from Sætreskarsfjellet ( ), the total 3-day precipitation varies between 26 and 102 mm, and the 5-day precipitation between 36 and 121 mm, whilst other meteorological factors are more or less the same: wind between fresh breeze and gale has come from sectors SW to NW, with temperatures below 0 o C. Avalanches from starting zones facing in other directions have similar variations in precipitation, and there is little common in wind data. 7 out of 10 avalanches occurred on dates when hourly precipitation had been above 2 mm/hour. There is one avalanche registration on a date when more than 7 mm/hour is recorded. As we do not have the exact timing of the avalanches we do not know if the precipitation intensity was the direct releasing factor or the total precipitation in itself, however, it seems that when 3-day or 5-day precipitation is relatively high, intensity of 2mm/hour or above might release the avalanche. Based on the precipitation and the avalanche records from 1974 to 1984 avalanche probabilities have been estimated according to the 3- and 5-day precipitation for the main avalanche path in Sætreskarsfjellet, when a given wind direction and wind speed threshold are fulfilled (Bakkehøi, 1985). Page 40

45 Table 2. Avalanche probability according to precipitation in Sætreskarsfjellet (with given wind direction and wind speed threshold). Probability Precipitation (mm) 3 days 5 days 90 % % % % % % % % Wind There are starting zones at Strynefjellet of several different expositions and they are thus exposed to snow loading with wind from different directions. The starting zones in the area can be grouped by their exposition: Group A: S-W Group B: W-N Group C: E Most of the solid precipitation is coming with winds from the NW sector, but there is also precipitation coming from W-SW. Wind from easterly directions seldom produces much precipitation. However, wind from this direction can load snow into starting zones by snowdrift as this wind direction is often associated with cold weather and drainage winds. Below are some figures (Figs ) showing the main loading areas for wind from sectors S-W and W-N (Kristensen, 2003). Page 41

46 Figure 41. Loading areas for wind from sectors S-W. Figure 42. Loading areas for wind from sectors W-N. The largest avalanche path is from the east flank of Sætreskarsfjellet which as can be seen from the map, is mainly loaded by winds from S-W. This is also confirmed by the avalanche and weather data. Page 42

47 Almost all of the avalanches from the eastern flank of Sætreskarfjellet are released after snowfall with winds from SW. One avalanche was released after only 7 mm 1-day precipitation. The 3-day precipitation was 46 mm and the 5 day precipitation 79 mm. On the date of the avalanche the wind changed to SW direction and increased to 11 m/s. It can be speculated that the wind increase the last day loaded the starting zone additionally with drifting snow and caused the release Conclusions For a few avalanche occurrences, precipitation seems to be the only releasing factor, but for most avalanches a combination of precipitation and wind is important. Most avalanches in Sætreskarsfjellet were released after heavy snowfall combined with winds above 5 m/s from SW direction. This wind direction increases snow loading in the starting zone at Sætreskarsfjellet east flank. However, also winds from NW may load staring zones due to cross loading of minor depressions in the slope. Figures show the areas that are loaded by different wind directions. For the other starting zones the weather data reveal no common wind direction. Continuous snowfall for 3-5 days, a wind increase, or precipitation intensity increase, may release the avalanche if the total amount of new snow is approaching considerable. Detailed weather data from the starting areas are very useful in avalanche forecasting, as well as information on exact time for avalanche release. However, in some cases it is not possible to see any direct connection between short term weather and avalanche release. Most likely these are situations where the snow pack stability was weak (caused by layers of depth hoar, faceted crystals etc), and only little loading due to precipitation, wind or temperature increase might be enough to release the avalanche. Information on snow pack layering and stability is thus very useful in avalanche forecasting. This must be done through systematic field work in the area. Information from the observers in the regional forecasting system should also be used. Page 43

48 Figure 43. Illustration of avalanche probability with increasing precipitation for the avalanche in Sætreskarsfjellet. The probabilities are based on a cumulative distribution of three- and five-day precipitation observations that have coincided with avalanche release in the Sætreskarsfjellet avalanche path (Bakkehøi, 1986). 6.2 FV 91 through Breivikeidet January 1983, 12:30 Graselva During the days before the avalanche, a snowstorm occurred with heavy precipitation and winds from WNW. When comparing data from Tromsø and Lyngseidet it seems that the storm started earlier in Tromsø than in Lyngseidet, which is understandable because Tromsø is closer to the ocean than Lyngseidet is and it will take some time for the weather front to move. In Tromsø, a large value of precipitation was measured at 19:00 on the 21 st, and at 07 on the 22 nd, meaning that the main part of the snowstorm occurred somewhere between 07 on the 21 st and 07 on the 22 nd, but light snowfall continued also on the 23 rd. In Lyngseidet the precipitation event was measured at 07 on the 22 nd and at 07 on the 23 rd, meaning that the snowfall occurred somewhere between 07 on the 21 st and 07 on the 23 rd. Almost no precipitation measured between 07 on the 23 rd and 07 on the 24 th. The avalanche was recorded by the road authorities at 12:30 on the 23 rd. Page 44

49 mm Issue / Revision 1/0 The 24 h precipitation measured at 07:00 on the 23 rd of January was 12.7 mm and 48 h precipitation was 34 mm. Tromsø had 5.8 and 32.5 mm respectively, 35 mm for 72 hours. As there might be a delay in the weather front travelling from Tromsø to Breivikeidet we have also registered the highest 24h precipitation that occurred between the 21 st and the 23 rd, this is 26.7 mm. 25 Precipitation (mm), Lyngseidet (91250) January Figure 44. Precipitation at Lyngseidet prior to the avalanche release. Precipitation and wind m/s mm January Precipitation Wind speed Gust Figure 45. Precipitation and wind gust at the weather station Tromsø prior to the avalanche. Arrow marks time of avalanche. Page 45

50 degrees Issue / Revision 1/0 Table 3. Precipitation prior to the event, measured at weather station Tromsø. 24 h precipitation measured at 07 on the 23rd h precipitation measured at 07 on the 23rd h precipitation measured at 07 on the 23rd 35 Largest 24h-precipitation between 21st and 23rd 26.7 The wind speeds (Tromsø) were between 6 and 11 m/s during the snow fall, with gusts as high as 21 m/s, coming from WNW. The wind data is from Tromsø, and indicates that the wind direction changed to SSW during the night before the 23 rd. No wind data exists for Breivikeidet or Lyngseidet, but most likely the whole storm occurred earlier in Tromsø than in Breivikeidet. The avalanche might have occurred during snow fall and wind from WNW, before the change in wind direction Wind direction, Tromsø January Figure 46. Wind direction. Some days prior to the storm there was a cold period. This might have formed surface hoar, but there is no data on this, making this speculation. The bowl in the starting zone might protect the hoar from being destroyed by wind, so that the surface hoar formed a weak layer for the new snow. The temperature rose to around -2 o C at the weather station when the snowstorm arrived. Page 46

51 Temperature, Tromsø (90450) oc January Figure 47. Air temperature th of March 1997, 01:00 Graselva Another classical example of heavy precipitation and winds, but this time with loading from WSW. The data shows a 24 h precipitation of 23.5 mm between 19:00 on the 8 th of March and 19:00 on the 9 th, and 48 h prec. of 33 mm. There was only negligible precipitation before that. The next measurement after the avalanche is at 07:00 on the 10 th (5.5 mm). The highest 24h precipitation during the period was 23.5 mm. The snowfall was accompanied with strong WSW wind (12-14 m/s, gusts of m/s). The highest gust was measured between 19:00 on the 9 th and 07:00 on the 10 th. However, some delay is expected when the front has to move from Tromsø to Breivikeidet. Most likely, also this avalanche has occurred during heavy snow and wind. Precipitation, wind speed and gust m/s mm March 0 Precipitation Av. wind speed Gust Figure 48. Precipitation, wind speed and gust at Page 47

52 Table 4. Precipitation prior to the event, measured at weather station Tromsø. 24 h precipitation measured at 19 on the 09th h precipitation measured at 19 on the 09th h precipitation measured at 19 on the 09th 33 Largest 24h-precipitation between 9th and 10th Wind direction Figure 49. Wind direction at april 1997, 11:50 Graselva Prior to this avalanche there was a 3 days period of heavy snowfall and strong winds. The avalanche happened on the last day of the snowstorm. Measurements show 24 h precipitation of 14.4 mm, 48 h precipitation of 33.3 mm, 72 h precipitation 57.5 mm. The maximum 24 h precipitation during the period was 24.2 mm, which occurred in beginning of the snowstorm. Wind was coming from NW, with a sudden change to ESE on the 19 th, but as in the other situations we think that there is some delay in the weather coming from NW. A change to ESE should not bring any more new snow into the starting zone, however, the snow can get more packed. Page 48

53 Degrees Issue / Revision 1/0 Wind and precipitation m/s April mm Precipitation Wind speed Gust Figure 50. Wind and precipitation at Table 5. Precipitation prior to the event, measured at h precipitation measured at 19 on the 19th h precipitation measured at 19 on the 19th h precipitation measured at 19 on the 19th 57.5 Largest 24h-precipitation between 16th and 19th Wind direction April Figure 51. Wind direction at Page 49

54 6.2.4 Conclusions All the avalanche occurrences that have been analyzed show that avalanches occur after snow storms with heavy precipitation and relatively strong winds. The precipitation has come with wind from westerly directions. These avalanches have starting zones that generally load by wind from SW- NW-N. To summarize, we can see that for all the avalanches, a 24 h precipitation of more than 23 mm was recorded within the two prior days. As for the wind, an average of 10 m/s was recorded during the snowstorm. Due to the fact that we do not have weather measurements in the near vicinity of Breivikeidet, there might be a delay in the weather situation as the front moves from Tromsø to Breivikeidet. This makes it difficult to accurately correlate the avalanche time with the precipitation or wind intensity. The main finding from the study at Breivikeidet is thus the importance of a nearby weather station in connection to avalanche warning, preferably situated in a release area. 6.3 FV 63 Geiranger Eidsdal Avalanche observation records done by the road maintenance exist from 1995 to In this period, 65 avalanches of different sizes reached the road and caused temporary closures. Of these, most of them occurred in 13 distinct avalanche cycles where two or more avalanches reached the road. The records show that the SW and NW wind directions dominate the weather pattern during the avalanche cycles, with avalanche occurrences on alternating slopes depending on wind loading. Page 50

55 Figure 52. Avalanche zones in the Geiranger-Eidsdal area. A simple susceptibility analysis has been done for the avalanche paths, showing which avalanche paths are likely to be loaded during which wind loading (Figure 53). Page 51

56 Figure 53. Colours show how prone the avalanche paths are to wind loading by different wind directions Conclusions The road is an interesting avalanche site, but the weather data is insufficient for more advanced analyses. The area is dominated by fjords and high mountains resulting in local climates that change very much over short distances, making the need for local weather stations even greater. The nearest, complete station is Tafjord, but Tafjord is very seldom representative for the weather in Geiranger and should thus not be trusted as the only weather station in a statistical analysis. Stations east of the area cannot be used either, because of the weather divide between east (continental climate) and west (maritime climate). As for Breivikeidet, our conclusion is that more detailed and closer situated weather data are needed to look at detailed correlations between weather data and avalanche occurrences. Page 52