Statistical Evaluation and Modelling of Landslide Initiation Points in Western Norway

Abstract (Swedish)

Norge drabbas årligen av hundratals jordskred vilket leder till betydande ekonomisk skada och fara
för liv. Därför bedriver norska myndigheter och forskningsinstitutioner omfattande forskning om
jordskred och vidtar åtgärder för att övervaka, modellera och slutligen mildra konsekvenserna av och
förebygga dessa. Som en del av sitt arbete underhåller norska vattenresurs- och energimyndigheten
en nationell skreddatabas som bland annat tillhandahåller rumslig information om enskilda jord- och
snöskred. Skred registreras dock främst i sina nedre delar, vanligtvis vid ett ställe där vägar eller
dylikt korsas. Syftet med denna uppsats är att utveckla en modell för västra Norge som möjliggör
att bestämma jordskreds begynnelsepunkt så noggrant som möjligt utifrån den rumsliga information
som registrerats i databasen. Modellen använder sig av oskarp logik, till skillnad från dagens
modeller för Norge som visar terrängens benägenhet för jordskred och är baserade på binär logik. I
denna studie bestäms frekvens av förekomst och oskarpa medlemskapsfunktioner för elva relevanta
parametrar. Sedan testas parametervikter och -kombinationer i syfte att bestämma den modell som
ger bäst resultat. Den slutgiltiga modellen har en genomsnittlig distans på 231 m till de faktiska
begynnelsepunkterna och uppnår därmed inte det önskade resultatet. De faktiska begynnelsepunkterna
har dock generellt ett så pass högt oskarpt sanningsvärde att modellen verkar lämpa sig som
underlag för en graderad karta över skredbenägenhet. Detta antagande styrks efter att i ett första
test ha jämfört den graderade kartan med den ograderade som används idag. Det rekommenderas dock ytterligare analyser för att kunna bekräfta detta med säkerhet.

Abstract

Norway experiences hundreds of landslides annually which leads to significant economic losses and
poses a threat to human life. Therefore, Norwegian authorities and scientific institutions are conducting
extensive research and take measures in the fields of documenting, modelling and ultimately
mitigating and preventing landslides. As part of its work, the Norwegian Water Resources and Energy
Directorate maintains a national mass movement database that features spatial information
on landslides and snow avalanches. However, landslides and snow avalanches are often registered
at a downstream location, usually where infrastructure is encountered. In this thesis, a model is
developed for western Norway for the determination of landslide initiation points based on spatial
information from the database with the aim of predicting the actual landslide initiation point as
accurately as possible. A key feature of the model is the application of fuzzy logic, as opposed to
those models based on binary logic that are used in Norway today for the assessment of landslide
susceptibility. In the present thesis, the frequency ratios of eleven landslide-conditioning parameters
are determined and transformed into fuzzy membership functions. Then, multiple parameter
weights and combinations are tested to determine the best-fitting model. The final model yields a
median distance of 231 m, which is larger than the desired result. Nevertheless, the actual landslide
initiation points generally receive a high fuzzy membership degree as intended. This may allow for
the creation of a graded susceptibility map based on the obtained model, and initial comparisons to
the non-graded one used today show promising results, although this requires further investigation.

Popular Abstract

Landslides are a common form of natural disaster in western Norway and a threat to human well-being and infrastructure. In the future, even more landslides are expected because of climate change. Information on landslides in Norway, for example where landslides happen, is collected in a national database. The location in the database is usually the point where a landslide hits buildings or roads. But for landslide research, it is more important to know where the landslide started. In this thesis, a model is developed that takes the location from the landslide database as input and tries to find the starting point. Such a model could allow the use of the database for a lot more purposes than today, and even save time and money if scientists can use the improved database instead of collecting information in the field. The main feature of the model and largest difference to other models used in Norway at the moment is that instead of binary logic (“true”, “false”), fuzzy logic is used (“completely true”, “mainly true”, “rather true”, “rather false”, etc.). Binary logic does not reflect conditions of the “real” world properly. For example, water in a bathtub is not only hot or cold, it can also be lukewarm or rather warm and so on. In the same way, it is unnatural to say that a location in the “real” world either can be or cannot be a place where a landslide can start, because one problem is where the line between “true” and “false” should be drawn. If a landslide location needs to have a minimum slope angle of 25°, can a location with 24.9° not be a starting point? With fuzzy logic, it is possible to differentiate more than with binary because fuzzy logic has, as the name suggests, fuzzy boundaries instead of binary logic’s “hard” boundaries. By doing so, it is not only possible to say if a landslide is likely to start at all in a certain location, but also how likely it is.

Since there is often more than one location on the same slope where a landslide could start or has started in the past, the model turned out to be unable to identify the real starting points of a set of landslides with high enough precision. But even when the model selected a different location as starting point, the real starting point was in most cases still classified as a likely location. For this reason, the model developed in this thesis could be used to create maps that show authorities and the public not only where a landslide might or might not happen (as the map does that is used today), but also where the risk of landslides is higher or lower. This could make it possible to identify for example specific road sections where drivers have to be most alert to the risk of landslides, and authorities can warn the public accordingly or plan targeted countermeasures.

• author: Starke, Tim
• supervisor: Gerhard Barmen
• organization: Civil Engineering (M.Sc.Eng.); Engineering Geology
• alternative title: Statistisk utvärdering och modellering av begynnelsepunkter för jordskred i västra Norge
• course: VTGM01 20201
• year: 2020
• type: H3 - Professional qualifications (4 Years - )
• subject: Technology and Engineering
• keywords: landslide, landslide modelling, debris flow, Norway, fuzzy logic, GIS, susceptibility, hazard, geohazard, natural disaster
• publication/series: ISRN
• report number: ISRN LUTVDG/(TVTG-5167)/1-59/(2020)
• other publication id: ISRN
• language: English
• additional info: Examiner: Jan-Erik Rosberg
• id: 9028692
• date added to LUP: 2020-09-10 08:46:10
• date last changed: 2020-09-10 08:46:10