DEC/JAN 2014/15 - Insurance News (the magazine)

Page 34

INMAG DEC 14_page layouts 2/12/2014 8:08 pm Page 34

“We’re over-modelled. We have too many models and they are too complicated, with too many bells and too many whistles.”

He notes a rapid rise in the use of geospatial technology through platforms such as Google Earth and geographic information systems. “They are getting a lot more popular and a lot more open. We are also getting much higher resolution and more open information and data sets nowadays.” Terrain models used about 15 years ago could provide images from 30-90 metres, but now global coverage is available at 12 metres. “There’s really a need now for the rapid acquisition of this kind of information,” Mr Mead told the seminar. “That’s where aerial platforms such as drones, which we are using, or balloons or applications with kites are very fast and cheap to deploy.” US insurer the United Services Automobile Association has applied for approval to fly unmanned drone aircraft after floods or other disasters to perform rapid damage assessment. After Typhoon Haiyan in the Philippines last year the Red Cross flew aerial drones to capture images of the damage. “They put all this information online and they effectively got crowd-sourced damage assessment from volunteers,” Mr Mead says. “People would log on to have a look at this map and start making out what buildings were damaged and where and to what level – slight or heavy.” High-resolution cameras are now cheap and light, and can provide better, faster damage assessment, he says. After the Christchurch earthquake in 2010, “before” and “after” pictures of the city’s cathedral were compared to assess damage. “We used an algorithm to try to match the images together,” Mr Mead says. “This kind of stuff, if you apply it broadly, will allow you to make more quick and efficient analysis of information for something such as rapid disaster processing.” Mr Mead is undertaking a PhD on visualisation technologies that can be used with numerical model34

ling to create 3D models of flood events. “It’s quite clear that the higher-resolution data we have nowadays means we can get a finer-scale understanding of risk and get it much faster,” he says. “This is leading more towards getting street-level, household-level information. But with these data sources it’s very important that you also use appropriate analytics to add value.” Big Data in its raw and natural state is not really a data stream, Mr Mead says. “There’s a lot of rubbish associated with it and you need to refine it, you need to re-map it and you need to code it to get your valuable information.” Data has been “democratised” online and governments are now releasing more, he says. “It reduces barriers to entry and when you combine that with cloud computing and your more scaleable access to computing power, there’s very little barrier to entry. “It’s all in the appropriate analytics you’re doing and is not necessarily about being able to handle Big Data sizes. Data is plentiful, but meaningful analytics are still relatively scarce.” Professor Blong told the seminar there is room for improvement in modelling. “My view of the present day is that we’re over-modelled. We have too many models and they are too complicated, with too many bells and too many whistles. “We don’t put nearly enough time into analysing the losses we have from past events and we put almost no time into understanding what is actually exposed to losses.” Professor McAneney agrees there are some areas that could be refined. “Risk Frontiers is lucky in having good relationships with sponsor companies that have been very generous in providing us with claims data,” he says. “However, we should be going right back to the assessors’ reports to get a deeper understanding of * vulnerability. We haven’t done enough of this.”

insuranceNEWS

December 2014/January 2015


Issuu converts static files into: digital portfolios, online yearbooks, online catalogs, digital photo albums and more. Sign up and create your flipbook.