A Vehicle OverTurning (VOT) Model: How Can An Impact Based Model Be Verified?

Numerical models are an integral part of weather forecasting, both in every day life and during extreme weather events. These models allow warnings to be issued and prepare emergency responders for the upcoming event. Numerical weather prediction has been done since the 1950s however forecasting the impact of weather on society is still in a developing phase. Impact-based forecasts may provide more relevant information for decision-makers and the public. This means there is a need for impact models that highlight the areas or communities which are at highest risk during an extreme weather event. These areas can then prepare for, and mitigate against the impact ahead of the event. However in order for these impact models to become trusted their output needs to be verified against the observed impact. Generally impact reports are difficult to acquire and often subjective, not standardised and lacking detail. This means that traditional verification techniques cannot easily be adopted.

This presentation will demonstrate ideas for how impact models can be verified, using the Vehicle OverTurning (VOT) Model as an example. This model has been developed as part of a Hazard Impact Model (HIM) under the auspices of the Natural Hazards Partnership, which is a collaboration between a number of UK agencies including the Cabinet Office. The aim of the HIM is to produce early warnings for severe events that allow us to generate an overall picture of the risk to society based on probability and impact. The VOT model uses uses the high resolution MOGREPS-UK ensemble to generate a probabilistic risk value for likely disruption to the road network based on wind gust speed and direction. This probabilistic hazard value is then combined with vulnerability and exposure values to give an overall risk value termed 'Risk of Disruption' which indicates the severity of road disruption due to a vehicle overturn.

During the winter 2013/2014 the UK experienced an exceptional number of high impact wind storms. These storms have provided a number of interesting case studies to use as verification for the VOT model, however finding data on actual vehicle overturning events to do this verification has proved laborious with inconsistent and limited results. Despite this, a number of data sources have been identified including news reports, social media reports, including twitter, police data and the Met Office Weather Observations Website (WOW, http://wow.metoffice.gov.uk/). Overall the VOT model appears to verify well using the currently collected data suggesting it provides useful guidance on the risk of disruption during high wind events. It is hoped that a more time-effective method of verification, allowing continual verification, will be developed in the near future using impact data from multiple sources.