A closer look at the CVFPP (2): Consequence modeling brings out uncertainty in risk estimates

A closer look at consequence estimation.

Learning more about how the Dutch assess flood risk and evaluate costs and benefits over the past eight months has  taught me lot about estimating loss-of-life and other consequences. Doing so from outside of the United States has allowed me to look more critically at how we perform them and how we might improve our own methods.

Part 2 of looking more closely at the CVFPP concerns Appendix 8G- Life Risk Analysis and Appendix 8F- Flood Damage Analysis. A first-round review suggests that the risk analysis methods used in the CVFPP may have underestimated the true risk facing the Central Valley (including life loss and other damages).


Section 3.3 of the CVFPP Life Risk Analysis explores various approaches to estimating life safety (including the Dutch approach), and in Section 3.4, the CVFPP proposes to use its’ own new method. A new approach should be peer-reviewed, however, especially one that my serve are the basis for critical decision-making and the investment of billions of dollars of state and federal funds.

On pages 3-27 and 3-28, DWR acknowledges limits to the study, but the following points highlight additional uncertainties in the analysis and could have effects on the real loss of life and damages experienced during a flood.

  • Inputs to the DWR Life Risk method are based on 2000 Census data and don’t account for new development behind levees since 2000, leading to an underestimation of the potential loss of life during a flood (and those that would require emergency services).  DWR’s model requires input data on a person-per-structure-level of detail and the 2010 Census data is not available at this level yet. However, conversations with US Army Corps personnel indicate that it is possible to account for the growth and still maintain the desired fine-grained data (FEMA has a method to account for it in HAZ-US as well) despite DWRs’ reluctance to do so.
  • The evacuation efficiency used is questionable for the following reasons
  • The warning times used for various impact areas may be optimistic as compared to other studies.[1]
  • One report cites,

“Furthermore, Sacramento residents have little or no experience with evacuations, and thus it is difficult to predict the fraction of at-risk populations who would choose to evacuate with much certainty.” [2]

  • Though DWR justifies this by assuming no levee breach, and therefore a 21 hour warning time and flooding only from the Sacramento River as opposed to the American(page 3-18), the point is that another scenario with 0 hours of warning time (as in the Comprehensive study) could produce a significantly different results.  This analysis therefore fails to capture uncertainty.
    • The model assumes a 100% willingness to evacuate but a recent study[3] showed 17% of residents of one Central Valley neighborhood would not evacuate if advised.
    •  Media reports during floods often mention individuals try to “wait it out” and have to be rescued in the end.  Consider the National Public Radio Broadcast from the 2011 floods:

“A thousand residents of Poplar Bluff, Missouri, were told to evacuate yesterday, when floodwaters started topping the levees there. Butler County Sheriff Mark Dobbs says one of the levees breached today, but despite repeated warnings, some residents did not evacuate and had to be rescued by boat.”  (NPR 2011)[4]

  • The model assumes a single mortality rate for an entire impact area despite the reality that mortality rates for areas adjacent levee breaches are higher than the rates in the rest of the floodplain. This is due to the fast moving and fast rising rate of water which can sweep people off their feet and cause buildings to collapse. [5]
  • Discrepancies between CVFPP and other Life Loss Analyses: 1) CVFPP Life Risk Analysis for Natomas Basin, 2) a 2005 “Urban Flood Scenario”[2] by SAFECA/David Ford Consultants, and a 3) recent assessment by Jonkman et al. in 20121 suggest  uncertainty in the Life Risk Assessment Method and the potential for DWR to have overlook critical factors. For example, the DWR Natomas impact area (SAC 36) assessment estimates a 2.5 Annualized Life Risk (which would be about 250 fatalities in a 100 year flood), while the Jonkman study suggests a significantly higher range around 500-1000 fatalities in a 100-year flood.

The Flood Damage Analysis does not consider nor communicate “societal risk.”  It  does not consider nor discuss the full range of consequences seen in large flood events like Katrina,[6] or used in risk assessment methods by other countries like the Netherlands[7]including:

    • ASCE (2007): Mass Migration from New Orleans

      Massive job loss

    • Migration of hundreds of thousands of people to other parts of the country and subsequent loss of culture
    • Short and long terms health effects due to contaminated waters
    • Analysis does not provide adequate discussion on availability of emergency shelters
    • Analysis does not appear to consider time/costs for pumping dry and rehabilitating areas/levees such as was done in the ARkStorm Report by the USGS[8].
    • It is important to note that though emergency response can be an effective tool, more Katrina fatalities occurred during evacuation than due to flood exposure.[9]

In summary, the Flood Damage and Life Risk analyses do not capture nor communicate a full societal Risk.  Annualized deaths and damages do not convey the same message as large numbers and losses from a single event.

Why is this all important?  It is important to accurately communicate the threat and potential consequences of a large flood to decision makers (CVFPB, land use planners, local governments) and to the public at large so they can make informed decisions about how best to prepare, prevent, and reduce risk. While uncertainties will always exist (which is the reason the field of risk management developed), efforts could be made to reduce uncertainty and consequently reduce risk by having more accurate information displayed using more comprehensive analyses with which to make decisions. Additionally, if DWR intends to use these analyses/methods to choose between alternatives in the future, limitations of the methods, and not accounting for uncertainties could bias certain risk reduction measures over others.

[1] Jonkman, SN. Hiel, L., Bea, R., Foster, H., Tsioulou, A., Arroyo, P., Stallard, T., and Lyndsie Harris (2012) Integrated Risk Assessment for the Natomas Basin (CA):Analysis of Loss of Life and Emergency Management for Floods.  Natural Hazards Review doi:http://dx.doi.org/10.1061/(ASCE)NH.1527-6996.0000079

[2] David Ford Consultants (2005). Urban Flood Scenario: Sacramento Area Levee Breach, Department of Water Resources http://www.ford-consulting.com/portfolio.php

[3] Ludy, J. & Kondolf, GM (2012). Flood risk perception in lands “protected” by 100-year levees. Natural Hazards. 61:(2), 829-842 DOI: 10.1007/s11069-011-0072-6

[4] Schaper, David (April 26, 2011). Severe Weather Wreaks Havoc in Midwest, South. National Public Radio  http://www.npr.org/2011/04/26/135745211/severe-weather-wreaks-havoc-in-midwest-south

[5] Rijkswaterstaat (prepared by HKV). (2004). Standard Method Damage and Casualties Caused by Flooding. DWW-2005-009.

[6] American Society of Civil Engineers (2007). The New Orleans Hurricane Protection System: What went wrong and why? http://www.asce.org/uploadedfiles/publications/asce_news/2009/04_april/erpreport.pdf

[7] Vrijling, H 2001. Probabilistic design of water defenses in the Netherlands. Reliability Engineering and System Safety 74 :337-344

[8] U.S. Geological Survey/MultiHazards Demonstration Project.  An ARkSTorm Scenario. http://pubs.usgs.gov/of/2010/1312/of2010-1312_text.pdf  

[9]Boyd, Ezra. 2011.  Fatalities Due to Hurricane Katrina’s Impacts in Louisiana.  Dissertation, University of New Orleans http://etd.lsu.edu/docs/available/etd-06092011-084046/


About wateraway

Jessica Ludy is an independent researcher with Unesco-IHE, a Water Resources Planner with ARCADIS, and a Fulbright Scholar in Flood Risk Management at Delft Technical University, the Netherlands. She was formerly the Associate Director of Flood Management at American Rivers in Berkeley, California, where she worked to promote sustainable flood management strategies in regional, and state, and federal agencies. Jessica has also been a lecturer and researcher at UC Berkeley in the Department of Landscape Architecture and Environmental Planning. Jessica grew up on the Great Lakes in Michigan, earned a BSc at the University of Vermont in Environmental Science, and an MSc in Environmental Planning from the University of California, Berkeley. This website is not affiliated with the Fulbright commission, ARCADIS, TU-Delft, Unesco-IHE, or American Rivers.
This entry was posted in Flood Risk Management, Flood Risk Perception, Loss of Life and tagged , , , , . Bookmark the permalink.

One Response to A closer look at the CVFPP (2): Consequence modeling brings out uncertainty in risk estimates

  1. Pingback: A Few Good Reads (5/7/12): Flooding in Great Britain » Hydraulically Inclined

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s