Dr. Robert Bolognesi R&D Department, Gester SA, Sion, Switzerland.

Transcription

1 NivoLogTM : AN AVALANCHE FORECASTING SUPPORT SYSTEM Dr. Robert Bolognesi R&D Department, Gester SA, Sion, Switzerland. ABSTRACT : NivoLogis an avalanche forecasting support system designed to be daily used by forecasters or ski patrol men who have to secure mountainous regions, ski resorts, roads or railways. The system consists in several software components dedicated to data management, data analysis and data exchange. NivoLog database contains numerical information about weather, snowpack, topography, geography and observed avalanches. This information is structured according to a relational data model. In supplement to numerical information, NivoLog can process images such as maps or photos. NivoLog also includes an analysis module computing micro-regional avalanche risks (according to the standard European scale) and local avalanche probabilities. Expert knowledge and nearest cases are both used by the model. The forecaster can use actual measurements as well as scenarios to get either estimations or predictions. Finally, NivoLog can edit a short bulletin to be sent on the Web and to pagers. A new snow information service is now available: one can daily receive weather data and local avalanche risks on a watch. At present, developments consist in interfacing NivoLog with automatic sensors. I 11 I KEYWORDS: avalanche forecasting, avalanche protection, avalanche engineering. 1. INTRODUCTION NivoLog is an avalanche forecasting support system designed to help to secure ski resorts or mountain roads. This software is the result of many years of scientific studies which started in 1982 at CEMAGREF with the financial support of two organisms : Alpe d'huez ski resort (France) and French National Agency for Technical Research. This cooperation provided the means to create the expert system AvaLog (Bolognesi, 1993). Then developments have been continued at the Swiss Federal Institute for Snow and Avalanche.* Research in Davos, with the goal to merge the principles of NXD (Buser, 1989) and AvaLog into one single system : NXLog (Bolognesi et ai, 1994, 1995). This software was quite efficient and a lot of European ski resorts used it. NivoLog is the outcome of the last findings of computer science research as well,as practical,-y users wishes. It was created in the R&D department of the engineering company "Gester" (Switzerland). This paper presents an overview of the software. 2. SOFTWARE STRUCTURE The experience shows that each user has his own needs and preferences, and likes to work with such and such familiar software. Consequently, NivoLog is designed as a tool box which can be composed according to the desires of each different forecaster. The main module enables to run independent programs dedicated to data acquisition, representation, analysis and communication (see figure 1). Thus every software which can compute data without modifying files format can be included into NivoLog. Since NivoLog data files are ASCII files, the software is widely adaptable. Corresponding author address: Robert Bolognesi, Gester SA, CH-1950, Sion tel: +41 (0) ; fax: +41 (0) ; ' bolo@gester.ch 412

2 Figure 1. NivoLog front screen. The software enables to run independent programs dedicated to data acquisition, representation, analysis and communication. 3. SOFTWARE MODULES 3.1.NivoLoq Data NivoLog Data is the data manager of NivoLog. Information is structured according to a relational scheme which permits data integrity control and compatibility with usual database engines (Le. Microsoft Access). Required measurements concern geography, topography, meteorology, snowpack, and avalanche activity (detailed or summarized). NivoLog Data is a software module (see figure 2) which can edit ASCII data files. Data transfer functions make it suitable for networking. 2.2.NivoLoq View Pictures also provide interesting analysis material to forecasters. So NivoLog includes a module to display maps 8.nd photos (see figure 3) and is compatible with the GIS "Arc View" (see figure 4). 2.3.NivoLoq Graph A little chart is often more informative than a large page of numbers. NivoLog Graph displays simple graphics in order to show quickly meteorological trends (see figure 5). 413

3 J Figure 2. Screen from NivoLog Data. Figure 3. Screen from NivoLog View. 414

4 ,_4- Figure 4. Compatibility between NivoLog and Arc View. 415

5 ----t--h Iii J i i Figure 5. Screen from NivoLog Graph. 2.4.NivoLoq Analvsis This is the heart of the system NivoLog. The prediction model is an expert system including case based reasoning as well as deductive expert reasoning. It provides two types of outputs: local avalanche probabilities and microregional risk index. Local avalanche probabilities are computed from data describing weather, snowpack, and topography, from history of each gully (nearest neighbors method), and from expert rules. These rules come from theoretical studies as well as local observations and practical knowledge. So, local peculiarities are taken into account by NivoLog. Such an approach was already used before in NxLog (Bolognesi et ai, 1995). The micro-regional risk index refers to the theoretical acceptation of risk which both depends on the probability of occurrence and on the consequences of the avalanche. This index is a function of the number of avalanches that may occur and of their average magnitude (see figure 6) : where R =f(n,m) N = class of expected number of avalanches M = class of expected magnitude of avalanches The function f and the classes Nand Mare defined so that NivoLog delivers a diagnostic according to the European avalanche danger scale (see figure 7). 416

6 Figure 6. Avalanche risk computed by NivoLog from expected avalanche activity and expected magnitude of avalanches. Pour Ie secteur : Crans-Montana I Aminona Le: , oj 10:14 isque iaible indice 1 de I'echelle europeenne) e manteau neigeux est stabilise dans la plupart es pentes. Des declenchements d'avalanches ne t en general possibles que par fortes surcharges groupe de skieurs par exemple) sur de Ires rares..,-' "'.,'. -: "'... tow danger... '.'..' (I~veLlonEuropeanavaJanChe danger scale) The snqwpack> is generally weubondedand. stable. Triggering is..possible. only with high.. additibnall()adsv.e~gro:upof skiers}.ona few"... )I~iysteepextrernes16p~s O~ty~feW~rn~llnatur;:lI avalanches(sluff~) pos$ibl e... Figure 7. Screen from NivoLog Analysis (micro-regional prediction). 417

7 2.4.NivoLoq Bulletin When the forecaster has made his opinion (based on data, model predictions, discussions, intuition, etc.) he can use NivoLog to inform the general public. The module NivoLog Bulletin enables to send messages on Internet and on pagers. Thanks to the cooperation of the companies "Meteotech" and "Swatch", snow and avalanche information can be daily received on a pager included in a watch (see figure 8). Thus, everywhere in Switzerland, one only has to look at the watch to know what is the avalanche danger at such and such mountainous area. Figure 8. Real time avalanche information on a watch (Swatch the Beep Alphanumeric). REFERENCES Bolognesi, 1993, Artificial intelligence and local avalanche forecasting : the system AVALOG, proceedings of the International Emergency Management and Engineering Conference, S.C.S., San Diego CA. Bolognesi R., Buser 0., Good W., 1994, Local avalanche forecasting in Switzerland: strategy and tools - a new approach, proceedings of the International Snow Science Workshop 1994, Snowbird, UT. Bolognesi R., Buser 0., 1995, Merging data analysis and symbolic calculation into a diagnosis system for natural hazards, proceedings of the International Emergency Management and Engineering Conference, Nice Buser 0., 1989, Two years experience of operational avalanche forecasting using the nearest neighbours method, Annals of Glaciology, vo1.13, I.G.S., Cambridge. Chritin V., Adam V., Rossi M., Bolognesi R., 1997, Acoustic detection system for operational avalanche forecasting, proceedings of the International Snow Science Workshop 1996, Canadian Avalanche Association, Revelstoke BC Chritin V., Bolognesi R., Gubler H., 1998, Acoustic sensor to measure snowdrift and wind velocity for avalanche forecasting, proceedings of NGI 2~ anniversary conference, in press. Meteo-France, 1994, Guide d'utilisation de I'echelle europeenne du risque d'avalanches, CEN, Saint-Martin d'heres. CONCLUSION NivoLog is constantly progressing. Once a year, all NivoLog users are invited to a meeting in order to define development priorities. This meeting, named TRACE (Table Ronde Avalanche Control in Europe) is very successful: more than 100 people participated to ''TRACE 98". That shows that users feel strongly concerned. At present time, the current work consists in interfacing NivoLog with automatic sensors such as FlowCapt (Chritin et ai, 1998) and Arfang (Chritin et ai, 1997). This link between hardware and software is a first step towards an automatic prediction system. ACKNOWLEDGEMENT Thanks to A.Cerruti, Y.Vollet, V.Chritin, H.Gubler, Th.Wasmer, JF.Blaser and D.Wannier for their helpful cooperation. NB : NivoLog Data and Nivolog Graph are freeware modules, and are provided with the sensors developed by IAV Engineering (i.e. FlowCapt)