‘Challenges for GIS remain around the uncertainty and availability of data’

‘Challenges for GIS remain around the uncertainty and availability of data’

SHARE

Tina Thomson, a young professional working in a global insurance and reinsurance company, shared the challenges insurers commonly faced in accessing geospatial information for risk assessment and model validation.

It is paramount in the insurance industry to manage exposure to natural and man-made catastrophes, which can easily wipe out an entire region. Obtaining an accurate and complete inventory of exposures remains a challenge¹ and data quality plays an increasingly important role in assessing risks for clients to help insurers respond quickly and fairly in the event of claims while improving the accuracy of pricing.

Geographical Information Systems (GIS) can help insurers and reinsurers like Amlin to identify exposure concentrations as well as enhance existing information through reference to other data. We can better understand the model sensitivity of exposure codings by location while visualising the geography of damage ratios (relating to hazard intensity) for the various physical characteristics of a risk, such as a building’s occupancy, construction and age. Capturing and geographically verifying the distinction of exposures is therefore very important to assess more precisely the financial risks from windstorms, floods and other hazards.

In combination with visualising footprints from historical events, among other things, GIS can help to assess whether losses appear reasonable and in line with expectations. For example, Figure 1 illustrates an external event footprint² for Windstorm Klaus, which affected Western Europe in January 2007, overlaid with a portfolio of campsite losses for the same modelled event. The high losses correspond with higher wind speeds along the coast and the event track. Comparing the losses against incurred claims further helps to support the loss figures that modelling produces.

Figure 1: Klaus raw hazard footprint overlaid with the modelled event loss²

Following a significant CAT event, information is often requested about potential exposures. By identifying these affected areas and applying probable maximum loss (PML) scenarios, we can start to estimate, albeit crudely, the potential impact on clients before the first claims arrives. GIS skills are vital for such spatial analysis, from overlaying information, calculating exposures within the footprint and producing maps visualising the affected areas.

This is more difficult with a large nationwide portfolio when complete coverage of the affected areas is required. Flooding, for example, is a particularly difficult peril to estimate due to its localised, highly granular nature.

During the December 2013 to January 2014 floods in the UK, the Environment Agency flood warnings and alerts provided an initial indication of affected areas but did not show the exact inundated areas. Model vendors may also release information from affected postcode sectors to delineated satellite footprints but again, the quality and geographical resolution are varied and only accessible if models are being licensed. Other limitations include coverage and visibility (cloud cover) while flood depth information is only available if ground-truthed or hydraulically modelled.

Regarding UK flood risk, the majority of insurance companies practise prudent underwriting due to the adverse selection of risks. An underwriter has to check that every risk has been assessed for its flood potential prior to quotation, acceptance or renewal of a policy and a lack of previous flooding in a location is not evidence enough; the full spectrum of possible, stochastically modelled events or design return period hazard maps needs to be considered.

Information from previously settled claims can be used for comparison against modelled losses to help assess underwriting performance and improve accuracy in the future. Indeed, an assessment of the UK flood claims confirmed the prudent underwriting practice of not insuring higher-risk flood locations while raising questions about the adequacies of the modelled hazard maps.

The December 2013 to January 2014 floods were the result of a number of strong rainfall events, exacerbated by saturated grounds and full ground water tables. Nonetheless, analysis of such events using GIS, especially with a number of different model outputs, helps us to understand the different models’ views of risks, and provide insight into differences in the underlying hazard resolution and defence assumptions used in the models.

Challenges for GIS remain around the uncertainty and availability of data, integration and interoperability of systems. In-house expertise for analysing information beyond simple mapping and visualisation is scarce and it is rare for a dedicated GIS team to exist within an insurance organisation.

Costs can prevent full use of GIS data with free data often only available for disaster relief efforts following a catastrophic event. Even with open-source data, such as the Copernicus EU programme, it is not straightforward to use free data in a commercial setting due to complex license terms and conditions. Other free data sources, such as satellite altimetry to measure water heights from flooded areas³, may provide long time series data (40 years), but data are also limited by satellite tracks and frequency of beams (resolution). And while insurers can purchase value added services from companies that offer catastrophe event response, often budgets for such expenditure do not exist.

More collaboration directly with the Earth Observation market could help achieve progress in the application of data in insurance. Some Learned Societies (RSPSoc, AGI) are starting to work with the insurers and have created special interest groups for disaster management, insurance and risk. The Willis Research Network is also well known for investing in this area, but the cost of data, available expertise and cost value benefit for the organisation remain significant.

Geographic information is undeniably valuable for integrating, organising and understanding data. New catastrophe modelling platforms are already integrating simple GIS capabilities for underwriters to harness the power of geographical information. If we want to make even more use of GIS beyond mapping and visualising, further investment is warranted in training, systems and data.

As the saying goes: “A picture is worth a thousand words.”

Notes:

  1. Ernst & Young (2008). Raising the bar on catastrophe data. The Ernst & Young 2008 Catastrophe Exposure Data Quality Survey.
  2. Extreme Wind Storm Catalogue, (c) Copyright Met Office, University of Reading and University of Exeter, available at URL: https://www.europeanwindstorms.org/.
  3. Lavendar, S. (2014). Using satellite altimetry to measure water height in estuaries, rivers and lakes. Geo Business Conference May 2014, London.