The landslide elements of Day 2 at EGU were split between two sessions on landslide forecasting and two on landslide risk. The latter is of comparatively little interest to me, and I had a load of work to do and meetings to attend, so I only attended the morning sessions.

In terms of landslide forecasting, there was some pretty good stuff presented. I would say that some speakers need to think a little more about their audience – using endless meteorological acronyms might work well if you are talking to weather specialists, but when the audience is mostly composed of landslide geologists it is a surefire way to lose the focus of your audience.

The first talk that caught my eye was that of Brunetti and colleagues from Italy, who used datasets culled from the literature to look at the statistical properties of landslide volumes. This sounds pretty terminal, but actually it is interesting as the distribution of landslide sizes in any given area follows a very specific relationship – a so-called power law. The paper examined the caharctersitics of this power law for a range of landslides, finding that there were consistent patterns that appeared to be determined by the mechanism of failure – i.e. soil slides formed a group, irrespective of location, that was distinctly different from the group associated with rockfalls. This felt like a substantial step forward – power lay relationships have been around for a while but we have struggled to understand what this tells us. Detailed studies like this will help greatly.

The next talk was a slightly odd one, by Peter Lehmann and his co-author on self-organised criticality. This relates to the power law issue above, but here the starting point appeared to be that avalanches in a sand pile also show power law behaviour and that this is associated with self-organised criticality. Ergo, landslides occur because of self-organised criticality, which means that the characteristics associated with SOC can be used to look at precursors to slope failure. This step may be problematic because the SOC displayed by sand piles is associated with frictional systems, whereas landslides are generally cohesive. Therefore I remain to be convinced that tje same precursors will occur in SOC for natural slopes. Clearly there is more work to do here, so I will watch with interest. This is one of those things that could be brilliant or it could be very esoteric.

Matthias Jakob went next, talking about the design of a debris flow warning system for N. Vancouver in Canada. His starting point was that given the risk of debris flows some sort of warning system is needed, but the cost of a full blown deterministic system is too high given the area covered. So, using 30 years of very high quality data, they had looked at understanding the relationship between debris flows and rainfall. Interestingly, they have rejected the standard intensity – duration relationship, instead undertaking a discriminant function analysis on the data to find that the three key factors are:

• long term antecedent rainfall (fills up the groundwater stores)
• medium term antecedent rainfall (tops up groundwater)
• short term rainfall intensity (triggers failure)

The upshot was an equation that allows warnings to be issued. Three alert levels will be used (no debris flows, debris flow watch, debris flow warning). The authors thought that on average five warnings will be issued each year, of which two on average will generate debris flows. The system has worked well in trials this year, but it will be interesting to see how the community reacts to so many warnings.

A rather peculiar presentation was given by Thiebes and his colleagues from the Department of Geography at Vienna. The talk was well delivered and the topic was both interesting and scientifically valid. So what was odd? Well, the team are part of the ILEWS consortium that is trying to develop early warning systems for landslides. To do so they have instrumented a landslide in the Swabian Alb area. The aim is to use the instruments to drive an online data collection and analysis tool that incorporates end-user driven modelling via the CHASM code. This is great – and I fully support such initiatives. The odd thing is that the landslide that they are instrumenting has moved 1 cm in the last 2.5 years – this hardly sounds like a slide that needs an early warning system! This is a shame as there are so many slides around that do need such a system.

Serval Miller from Chester University presented a very detailed analysis of a landslide susceptibility mapping exercise that he had undertaken in Jamaica. Interestingly, they had tested a range of GIS based techniques, concluding that a Bayesian Model provided the best results. Second best was a straight forward layer combination model. This was quite interesting, but with this and other presentations on landslide susceptibility analysis I do end up wondering whether the time spent would be better used to provide a geomorphological map that indicated where landslides are considered likely. I wonder whether this would really be any less accurate?

Finally, I would like to note the work of Devoli and her colleagues, who have been trying, with some success, to implement a landslide risk reduction programme in Nicaragua in the aftermath of Hurricane Mitch a decade ago. Although the in-country team is small (three geologists), a huge amount appears to have been achieved through SINAPRED, the national emergency commission. This presentation highlighted two web resources that are well worth a look:

Georiesgos-ca.info, which provides georisk information across all of Central America (in Spanish)
http://mapserver.ineter.gob.ni/website/mapas/Estudios/viewer.htm, which is a mapping server that provides access to hazard maps for Nicaragua.