posted on 21 November 2016
-- this post authored by Laurel DiSilvestro
We have all seen the dramatic pictures of destruction and loss after a devastating flood event. Usually, the scene is depicted as a sea of water inundating broad areas of a geographic region, impacting communities and homes. However, what we usually don’t understand from these images is the significant regional variations of water depth. This is one of the key indicators of contents and building damage, and therefore integral to understanding flood modeling.
One of the significant factors for accurately assessing flood risk is the site elevation relative to surface water elevation. Determining site elevation often requires parcel or building-level geocoding to obtain the location accuracy required to model flood risk. In addition to location accuracy, modeling with the optimal elevation data resolution is also critical.
With the flood insurance market poised for growth, one common concern is that there is no adequate loss model to characterize the risk from site-specific underwriting through to portfolio management. Accuracy in location, hazard, vulnerability and financial model component resolution is critical for this peril in order to adequately model loss. Furthermore, accuracy at the site level not only improves underwriting confidence, but also leads to greater accuracy in determining aggregate risk.
The following example explores the impact that elevation and location accuracy can have on modeled loss. Figure 1 shows a cross section of the Arroyo de la Laguna watercourse which extends the length of the city of Pleasanton, California, in the eastern portion of the San Francisco Bay Area. This region has a historical record of flooding. The largest recorded flood occurred December 22-24, 1955 and fully inundated the northern part of the region - which back then was mostly farmland. This flood was caused by a series of devastating storms that impacted a broad region along the west coast. There were 67 deaths due to the storms with direct losses of $1.7B in 2016 dollars.1 While no deaths were recorded in the Pleasanton region, the series of storms caused 31 deaths in the seven county San Francisco Bay Area with estimated property damage of over $75M (or approximately $670M in 2016 dollars)1.
Since then, Pleasanton has undergone exponential growth to become a significant urban region with high density residential and commercial developments. Upstream dam construction in 1968 and drainage alterations performed on the watercourse reduced flood risk; however, it has still been subject to flooding, with the most recent event in 1988 causing significant flooding in localized areas. It is important to note that a large component of this recent flood event was caused by flash flooding, not just riverine flooding. So, while the Federal Emergency Management Agency 100-year flood zone boundary matches the width of the engineered waterway channel, the historical flash flooding potential has not been constrained to this boundary.
For this example, “sites" at 10-meter intervals have been created along the cross section. These are analyzed to illustrate the impact on loss of elevation assignments relative to 3-meter, 10-meter, and 30-meter digital elevation model (DEM) data. Notable is that both the 3-meter and 10-meter DEM profiles match one another well. However, the 30-meter DEM profile shows significant differences within the watercourse channel and at the location of the arrow, where the graded ground elevation changes from one property to another.
The elevation differences are less along flat regions as seen on the right side of the cross section. But, when we ask if this will this translate to similar loss potential, the answer is “not always."
For illustrative purposes, a $1 million combined structure and contents reconstruction cost value was modeled at each “site" with the CoreLogic® probabilistic U.S. Inland Flood Model at 10-meter intervals along the cross section. Both the 10-meter and 30-meter DEM data were used in the analysis. Figure 2 shows the modeled mean annual loss, the common basis for assessing premium needed to cover loss over time. At the site marked by the arrow that was referenced earlier, the 10-meter DEM results in a $339 higher pure premium than the 30-meter DEM result - a 33 percent difference. This result could have an impact on both risk selection and pricing.
Focusing on the right side of the cross section, we see that despite the similarity in ground elevation for each of the DEMs, there is a significant loss difference. This is driven by the differences in the channel geometry defined by the 10-meter versus 30-meter DEM data, as well as the downward slope of the topography away from the channel edge up to the berm. The 30-meter channel is narrower at depth, which reduces the volume of water that the channel can hold and increases the velocity of the water. These conditions increase water depth and damage potential, leading to greater loss where water banks up against the higher elevation starting at the endpoint of the cross section.
This example illustrates the potential impact of data accuracy and resolution on not just hazard assessment, but on loss modeling as well - both critical input to accurately managing flood-exposed risks at the site as well as when aggregated to the account or portfolio level.
Flood risk continues to be a significant and growing insured loss threat to the private insurance industry, whether as a policy endorsement, a named peril or an inclusion in a multi-peril policy. As the private insurance industry explores diversifying and growing business with the potential changes in National Flood Insurance Program coverage and rates, as well as an increased perception of risk, insurers are seeking highly granular loss models to address the management of a peril where the amount of loss can be largely influenced by small variations in water depth. Therefore, it is important that insurers leverage data that is current, detailed and highly granular for sound flood risk assessment.
© 2016 CoreLogic, Inc. All rights reserved
>>>>> Scroll down to view and make comments <<<<<<
This Web Page by Steven Hansen ---- Copyright 2010 - 2017 Econintersect LLC - all rights reserved