The Science of PerilsBy Arindam Samanta | May 5, 2015
The newly released Prometrix® Peril and Incident Report brings to commercial insurers, for the first time, the most granular and up-to-date analytics on a range of perils – wind, lightning, hail and wildfire – that are sources of multi-billion dollar losses. These analytics represent decades of scientific and technological expertise on insights gleaned from a variety of sources, and which has been developed in a truly big data framework.
Our work distills all the essential characteristics of big data – variety, velocity and volume. Variety is evident in the diverse data streams we collect across the entire continental US – dense network of ground-based radars (for wind and hail), lightning detection sensors (for lightning), satellite and airborne remote sensing (for wildfire fuels mapping, terrain and road network analysis), weather and climate models, as well as ground observation stations (for data-analytic validation). And, these data assets are unprecedented in their granularity, which covers a wide spectrum from 50 cm imagery to 1km weather data - a key requirement for assessment of risk at property- and neighborhood-levels. Together these sensors collect data at an astounding velocity – for instance observations of fast-moving storms is made every 2-5 minutes by radar networks, and over 99% of all cloud-to-ground strikes are recorded by lightning sensors.
Such timely information is essential for an accurate record of perils at property locations, and for assessing trends in hazard occurrences as well. This tremendous variety and velocity of data collection results in humongous volumes of data, which can be in the order of terabytes each day or over 4 petabytes in a decade, which is equivalent to 10 years worth of HD-TV videos.
We harness the power of these big data assets via a range of technologies from statistics, weather and climate sciences, image and computer vision, and machine learning. And, more importantly, we incorporate insights from industry-wide claims data to create analytics that can answer key business questions facing (commercial) underwriters such as:
- How likely is a property to have pre-existing damage?
- Which properties should be inspected?
- Which properties are at higher risk from perils than others?
- What is the impact on underwriting eligibility?
- What is the effect on (non-modeled) loss costs?
Read the press release: Verisk Insurance Solutions Announces Peril and Incident Report