By Dr. John Galantowicz and Dr. Arindam Samanta
Every day, a fleet of satellites — the eyes in the sky — orbits the earth gathering terabytes of imagery data. Insurance leaders are using sophisticated analyses of satellite imagery in operations to improve their property valuation, risk selection, accumulation management, target marketing, and claims adjustment.
Today’s satellite constellation includes a variety of sensor technologies and operational modes, each of which has been optimized for a specific observational mission. Perhaps most astounding are the very-high-resolution (VHR) sensors that produce the type of imagery widely available in desktop and mobile mapping applications. The utility of such imagery is obvious: Trained image analysts and laymen alike can zoom to virtually any location on earth and immediately have a visual snapshot of that place and its vicinity from at least one point in the recent past.
In addition to VHR, many satellites now offer multispectral sensors with the ability to detect light in wavelength bands beyond the range of regular red-green-blue human vision. The extra bands are carefully chosen for their sensitivity to vegetation cover, water, material type, or surface temperature. Used in the right combinations, data from these bands can detect otherwise hidden properties of a scene, such as storm damage, flood waters, swimming pools, roof types, or vegetation overhanging structures or power lines.
The true analytical power of VHR and multispectral imagery can be realized only when image processing algorithms transform pictures into actionable information. Scientists can assess wind damage and map wildfire and flood areas. Analysts can identify new home additions and detect power outages.
The final step in the creation of comprehensive analytical decision aids is the use of geographic information system (GIS) tools. GIS tools allow spatially complex data sets to be cross-registered for inclusion in location intelligence databases. When converted to GIS formats, satellite imagery analytics can be used to generate property or region-of-interest reports, developed into situational awareness reports during natural disasters, or assimilated directly into GIS-compatible operations tools.
The following examples highlight three areas where advanced sensor data, novel imagery analysis algorithms, location intelligence, and GIS tools have been combined to provide analytics incorporated in industry practices today.
Superstorm Sandy caused wind and storm surge damage that resulted in widespread power outages. Figure 1 demonstrates two types of satellite imagery analysis related to the underlying vulnerability to power infrastructure from such a storm.
The left image in Figure 1 shows a GIS representation of the concentration of power lines vulnerable to tree fall along and inland from the New Jersey shore. The analysis uses two data types: vegetation density derived from pre-storm satellite imagery and specialized knowledge of the transmission and distribution (T&D) network. The right image shows the degree of power outage derived from satellite imagery from a special sensor used to detect nighttime light.
Wildfire is a significant risk to properties in the fire?prone areas of the western United States from the Pacific coast to Texas. Satellite imagery analysis is already affecting two key functions: location based wildfire risk assessment and daily monitoring of wildfire progress and its proximity to properties.
Figure 2 illustrates satellite imagery analyses prior to and during the June 2012 High Park wildfire near Fort Collins, Colorado. The base map represents vegetative fuels analyzed before the fire from multispectral imagery. By combining vegetative fuels, terrain characteristics, and road access types in GIS tools, analysts are making comprehensive, building specific wildfire risk assessments available for underwriting, pricing, claims, and portfolio exposure management needs.
The High Park fire burned for 23 days, affecting almost 90,000 acres before being contained. Satellite imagery provided almost daily coverage of the fire, allowing frequent GIS updates from image analysis.
The three analysis types illustrated in Figure 2 are used for moratoriums and bulk claims reserving during and after the fire. The total burn area is useful for mapping the entire area where damages may have occurred to date. During the fire, the new burn area analysis indicates where the fire most recently spread, which is useful for situational awareness and moratoriums. After the fire, insurers use the new burn area analysis to reassess the fire for verification of underwriting guidelines, for example. Finally, the hot areas analysis provides insight into where damage may be imminent, facilitating staging of claims response resources.
Tornadoes are one of the most destructive forces of nature. A single tornado is capable of causing severe property damage, and numerous individual storms can cause damage across hundreds of miles in a matter of hours. Advanced image analysis techniques are critical for providing rapid assessment of damaged areas.
Figure 3 shows the intense damage associated with the tornado that devastated Tuscaloosa, Alabama, in April 2011. The tornado was one of more than 300 that struck the region over a two-day period, 38 of which were rated EF3 or higher on the five-point Enhanced Fujita scale of tornado damage. The analysis of the destruction shown in Figure 4 was based on change-detection algorithms exploiting the hypersensitivity of multispectral measurements.
If a picture is worth a thousand words, insurers don’t have to imagine how the advances in satellite imagery interpretation alter the landscape for wide-scale property and peril analytics.
John Galantowicz, Ph.D., is senior staff scientist and manager of the Land Surface Remote Sensing group at Atmospheric and Environmental Research (AER).
Arindam Samanta, Ph.D., is a scientist and FireLineTM operations manager at Atmospheric and Environmental Research (AER).