Slip Slidin' Away: GIS-Based Risk Analysis for Landslides and Avalanches in Norway

KelmanArticles, Disasters, Earth Observation, Original

Ilan Kelman, Bård Romstad, and Håkon Sælen
Center for International Climate and Environmental Research – Oslo
Slide Processes and GIS in Norway
On 26 March 2008, the coastal Norwegian city of Ålesund was stunned by a rock face giving way, falling onto an apartment block constructed four years earlier, and knocking the building off its foundations, collapsing the bottom two floors. Five people were killed, bringing to the fore the challenge of contending with hazards when living in a mountainous country.
A slide-related disaster with so many casualties, and which could have caused many more deaths, is relatively rare in Norway. On average, Norway witnesses 10-20 landslide and avalanche fatalities per year, less than half the annual homicide rate and equivalent to the monthly death toll on the roads. Yet as new development expands urban limits, encroaching into less stable topography, and as affluence leads people to back country skiing and hiking, the dangers from slide processes are ever-present. Even if casualties could be entirely averted, economic costs would be incurred from evacuations, road closures, and bridge repairs-plus the expected cost of damage would need to be compared to mitigation costs.

Figure 1: A debris flow near Bergen, Norway during autumn 2006. Photo: NGI.

Figure 1: A debris flow near Bergen, Norway during autumn 2006. Photo: NGI.

To investigate and address these challenges, computer-based tools for analyzing risks, and for using the research to inform policy and practice, are becoming increasingly helpful. Tools such as Geographic Information Systems (GIS) now routinely appear on desktops and laptops in planning and environmental departments, with user interfaces that make the software appear simple to use-which often masks its power and complexity. The capabilities of GIS as an analytical tool can, in fact, be overlooked in favor of data storage and linkage, along with visualization.
All these functions are currently being harnessed for slide disasters in Norway through a research team developing a GIS-based model to analyze risks and to further understand possible limits of and extensions to available GIS tools. The model accommodates all the most important slide processes in Norway, namely avalanches, rock falls, debris flows (Figure 1), shallow landslides, and quick-clay slides. The main outputs are the threats posed to human life and possible economic damage.
This form of GIS-based risk analysis is one of the first applications for Norway, although it draws on experience from, and adjusts to the Norwegian context. Similar work is underway on slide processes from around the world, including Iceland [1] and Switzerland [2]. Similar methods have also been applied to other potential disasters, such as those involving volcanoes [3] and floods [4].
The Model’s Architecture
Risk is generally defined as a combination of (i) hazard, incorporating the environmental process, which for this case is slide processes in Norway, and (ii) vulnerability, incorporating the harm to society that could result, which for this case is loss of life and tangible economic damage, and why that potential for harm exists. GIS provides a solid tool for combining different layers of hazard and vulnerability information into a risk calculation. The model then automates map generation from the risk analysis.
For the first investigations using this model, a partially qualitative approach is being pioneered and evaluated for the hazard representation. For a given slide process, the model takes its input as a set of polygons defining hazard zones based on field surveys, historical studies, and average recurrence intervals of damaging events. This partially qualitative approach is powerful because it bypasses the usual limitations of strict return period calculations, such as assuming a constant baseline, incomplete data sets, and applicability of extreme value statistical models.
For the vulnerability factor, buildings that intersect the hazard zones are extracted from a national geodatabase maintained by the Norwegian Mapping Authority. The database contains the size and type or use of each building. Economic value per square meter for each building class in each municipality is calculated based on information provided by Norway’s largest property insurer, and entered into a database in the GIS.
This database also includes estimates of the probability distributions of different event intensities. For each event intensity class of each slide process, other database entries are:
• The probability of impact on each building given an event in the hazard zone.
• The vulnerability of each building class.
• The vulnerability of people in each building class, representing another part of the vulnerability input and indicating the probability of death. For the moment, every individual is assigned the same probability of death, but that would later need to be differentiated by characteristics such as age, gender, illness, and disability.
Table 1: Study areas in Norway

Table 1: Study areas in Norway

A significant uncertainty is the assumption regarding how many people are in each building. While residential occupancy data could potentially be known precisely from census data, that information is protected by law. In any case, these values fluctuate rapidly. Cycles include day/night as people attend work and school, weekends and holidays, and trends such as children being born, young adults seasonally living in university residences, and workers retiring. Various social reasons for living in different locations and with different lifestyles mean that property size correlates poorly with the number of people living in that property.
Such values, therefore, have high uncertainty, but the database can easily be updated as better data become available. Additionally, toggling the values permits sensitivity analyses to be run or scenarios to be compared. Examples are (i) an event during a holiday weekend at night versus a working day at noon; (ii) an area populated by young couples versus retirees; (iii) the fatality consequences for different levels of occupant disability; or (iv) poor versus competent maintenance affecting building type vulnerability.
Once all these parameters are set, for every building, the risk to life and the expected average annual damage is calculated for each type of geohazard individually and for the combination of all geohazards. In its most detailed form, building-by-building and person-by-person results could be presented together with the original map of buildings. For a more synoptic and potentially more accurate (although less precise) view, the numbers could be aggregated over grid cells of a user specified size or characteristic.
Figure 2: Model screenshots from Otta in south-central Norway. The left panel shows the input data (buildings and hazard zones) and the right panel shows the model output in terms of estimated average annual damage to buildings aggregated over 100 m grid cells. NOK 100 ≈ EUR 12.5

Figure 2: Model screenshots from Otta in south-central Norway. The left panel shows the input data (buildings and hazard zones) and the right panel shows the model output in terms of estimated average annual damage to buildings aggregated over 100 m grid cells. NOK 100 ≈ EUR 12.5

As part of the model’s development, a national database of input parameters is being constructed, capturing the demographic and economic variations among different localities. Notwithstanding changes over time, this can reduce the amount of field work and parameterization necessary in each location.
The high flexibility and wide scope of the model ensures that different locations, different slide processes, and different scenarios can be analyzed consistently, effectively from the same baseline. That assists in determining high priority areas for disaster risk reduction measures along with an estimate of how different measures could affect the casualties and building damage.
So far, the model has been applied to four different areas in Norway that experience different slide processes (Table 1). Figure 2 provides a sample from Otta in south-central Norway.
Expanding the Model
In presenting the model and results to policy makers and practitioners, care must be taken to note that the usefulness of the output depends to large extent on confidence in the inputs. Ranges encompassing the uncertainties or sensitivities of outputs to specific inputs could be a useful display, with the GIS maps presenting those ranges rather than absolute values.
Additionally, results can be provided for only the slide processes considered, not for all possible challenges from slide process that Norwegians face. The greatest slide process threat to Norway has yet to be incorporated into the model: tsunamis generated from slides plunging into lakes or fjords [5] or from giant underwater landslides in the North Sea that have devastated Norway’s western shores in the geological record [6]. Both casualties and damage would be enormous and the GIS model would need to be further developed to consider, for example, different evacuation scenarios under different warning scenarios.
[1] R. Bell and T. Glade, “Quantitative Risk Analysis for Landslides – Examples from Bildudalur, NW Iceland,” Natural Hazards and Earth System Sciences, vol. 4, pp. 117-131, 2004.
[2] H.R. Heinimann, Risikoanalyse bei gravitativen Naturgefahren – Fallbeispiele und Daten, Umwelt-Materialen nr. 107/II, Bern: Bundesamt für Umwelt, Wald und Landschaft (BUWAL), 1999.
[3] R. Spence, I. Kelman, E. Calogero, G. Toyos, P. Baxter, and J.-C. Komorowski, “Modelling expected physical impacts and human casualties from explosive volcanic eruptions,” Natural Hazards and Earth Systems Sciences, vol. 5, pp. 1003-1015, 2005.
[4] T. Thumerer, A.P. Jones, and D. Brown, “A GIS based coastal management system for climate change associated flood risk assessment on the east coast of England,” International Journal of Geographical Information Science, vol. 14, pp. 265-281, 2000.
[5] A. Braathen, L.H. Blikra, S.S. Berg, and F. Karlsen. “Rock-slope failures in Norway; type, geometry, deformation mechanisms and stability,” Norwegian Journal of Geology, vol. 84, pp. 67-88, 2004.
[6] T. Bugge, R.H. Belderson, and N.H. Kenyon, “The Storegga Slide”, Philosophical Transactions of the Royal Society A, vol. 325, pp. 357-388, 1988.
Ilan Kelman is a Senior Research Fellow at CICERO with a PhD from the University of Cambridge examining disaster risk reduction. His research covers vulnerability reduction in the context of sustainability and development processes, with a particular focus on island case studies and disaster diplomacy. More information is at
Bård Romstad is a Senior Research Fellow at CICERO and holds an MSc degree in Physical Geography from the University of Oslo. His current research focuses on the use of GIS and remote sensing to study the interface between biophysical and socioeconomic impacts of natural hazards and climate change.
Håkon Sælen is a Research Fellow at CICERO. He holds an MSc in Environmental Change and Management from the University of Oxford. His recent research has focused on economic consequences of natural hazards in the context of a changing climate.