- Home Page
- Full Report
- Flood Damage Data
- Links of Interest
- For Further Reading
- Search this Website
- Site Map
A. Why We Need Historical Flood Damage Data
The National Weather Service (NWS) estimates that flooding caused approximately $50 billion damage in the U.S. in the 1990s (NWS-HIC 2001). Although flood damage fluctuates greatly from year to year, estimates indicate that there has been an increasing trend over the past century (Pielke and Downton 2000). Some have speculated that the trend is indicative of a change in climate (e.g., Hamburger 1997), some blame population growth and development (e.g, Kerwin and Verrengia 1997), others place the blame on federal policies (e.g., Coyle 1993), and still others suggest that the trend distracts from the larger success of the nation’s flood policies (e.g, Labaton 1993).
To understand increasing damage and assess implications for policy, decision makers need to resolve the independent and interdependent influences of climate, population growth and development, and policy on trends in damage. Increased flood damage due to changing climate requires different policy actions than would damage increases due to implementation of flood policies.
The available records of historical flood damage are inadequate for policy evaluation, scientific analysis, and disaster mitigation planning. There are no uniform guidelines for estimating flood losses, and there is no central clearinghouse to collect, evaluate, and report flood damage. The data that exist are rough approximations, compiled by the NWS from damage estimates that are reported in many different ways. Moreover, most published summaries of the damage estimates focus primarily on aggregate national damage totals.
Scientists need historical flood damage data at a variety of spatial scales to analyze variations in flood damage and what contributes to them. For example, during El Niño years, southern California receives more precipitation than in the typical year. Conventional wisdom suggests that the increase in precipitation should result in an increase in damaging floods. If California’s emergency planners knew this to be the case, they could prepare for the floods that come with El Niño, possibly reducing damage. In this case, scientists looking for a causal relationship would want to determine to what degree historical high damage years in southern California are associated with El Niño events. This requires sub-state-level data sets, rather than a national data set.
Social scientists looking at the effect of policies designed to reduce flood damage also need access to historical data at regional and local scales. Take the example of the National Flood Insurance Program, created in 1968 to “assist in reducing damage caused by floods” (42 U.S.C. # 4102 (c)(3)). Researchers evaluating the program would like to isolate the effect of the program from all other factors influencing flood damage in particular areas. At the river basin or community level, the effect of a federal policy implemented in 1968 might be isolated and measured.
In sum, historical damage data are essential for any study that seeks to understand the role that climate, population growth and development, and policy play in determining trends in flood damage. Some studies might require data at the national level, and others at the state or local level. Moreover, researchers need guidance to use the data effectively. Some data sets are not accurate enough for certain types of analysis.
B. Sources of Historical Flood Damage Data
Ideally, a national database of historical flood damage should cover the entire country over a long time period, using consistent criteria and methods in all times and places. Table 1-1 compares possible sources of damage data. The National Weather Service is the only organization that has maintained a long-term and fairly comprehensive record of flood damage throughout the U.S. Insurance company records include only insured property. Records of the Federal Emergency Management Agency (FEMA) include only property that qualifies for federal assistance in presidentially declared disasters. Few state and local governments maintain damage records beyond those required by FEMA. Only in newspaper archives from cities and towns across the nation might one find more complete reporting of historical flood damage. Indeed, a newspaper archive could be the best source of information on flood damage in a particular locale. But the parochial nature of such data makes aggregation problematic.
For long-term coverage of the entire nation, and of most states, the NWS data sets appear to be the best available source of flood damage estimates. However, the scope, accuracy and consistency of the data must be evaluated to determine how they can be appropriately used and interpreted.
C. Scope of the NWS Flood Damage Data
The NWS Hydrologic Information Center (NWS-HIC 2001) describes the data as “loss estimates for significant flooding events,” providing estimates of “direct damages due to flooding that results from rainfall and/or snowmelt.” However, key concepts such as “flood” and “flood loss” are defined differently by various agencies and researchers depending on their objectives. Appropriate use of NWS damage data requires understanding of what is and is not included.
Types of Flooding:
Ward (1990) defines a flood broadly as “a body of water which rises to overflow land which is not normally submerged.” This definition covers river and coastal flooding, rainwater flooding on level surfaces and low-gradient slopes, flooding in shallow depressions which is caused by water-table rise, and flooding caused by the backing-up or overflow of artificial drainage systems.
The NWS includes damage from most types of flooding listed above, but excludes ocean floods caused by severe wind (storm surge) or tectonic activity (tsunami). These are excluded because, although they result in water inundation, they are not hydrometeorological events. In addition, the NWS excludes damage that results from mudslides because, though they are caused by excess precipitation, they are considered primarily a geologic hazard.
Table 1-1. Sources of flood damage estimates.
|National Weather Service flood damage data sets||1925-present||
|Estimates of direct physical damage from significant flooding events that result from rainfall or snowmelt|
|Insurance records (National Flood Insurance Program, private insurers)||1969-present||
|Personal property claims made by individuals holding flood insurance|
|Disaster assistance records (Federal Emergency Management Agency)||1992-present||
|Federal and state outlays for public assistance, individual assistance, and temporary housing in presidentially declared disasters|
|State and local government records||Varies||State||Varies|
Definition of Loss, Damage, and Damage Estimates:
Researchers specializing in natural hazards have expressed a need for more complete documentation of losses, including both direct and indirect costs associated with flooding (Mileti 1999; National Research Council 1999; Heinz Center 2000). Direct costs are closely connected to a flood event and the resulting physical damage. In addition to immediate losses and repair costs they include short-term costs stemming directly from the flood event, such as flood fighting, temporary housing, and administrative assistance. By contrast, indirect costs are incurred in an extended time period following a flood. They include loss of business and personal income (including permanent loss of employment), reduction in property values, increased insurance costs, loss of tax revenue, psychological trauma, and disturbance to ecosystems. They tend to be more difficult to account for than direct costs (Heinz Center 2000).
The NWS describes its flood loss data as estimates of “direct damages” including, for example, loss of property and crops and costs of repairing damaged buildings, roads, and bridges. The NWS estimates have usually been restricted to direct physical damage, a subset of the losses generally considered to be direct costs.
The dollar figures in the NWS damage data are estimates compiled soon after each flood event, before the actual costs of repair and replacement can be known. They are not verified by comparison with actual expenditures. The estimates are gathered from diverse sources, some who use accurate estimation methods (e.g. insurance companies) and others who do not (e.g. newspapers). Therefore, NWS damage data are best described, not as “loss data”, but as “damage estimates.”
D. Purpose and Methods
Objectives of this study are (1) to assemble a national database of historical flood damage based on NWS damage estimates, making it as complete and consistent as possible; (2) to describe what the estimates represent; (3) to evaluate the accuracy and consistency of the estimates; and (4) to develop guidelines for use of the data and make it widely available to users. Steps followed to achieve these objectives are described below.
The NWS Hydrologic Information Center (NWS-HIC) is responsible for compiling and archiving flood damage estimates collected from NWS field offices throughout the U.S. Its staff members provided several data sets and access to files and publications archived in their office at Silver Spring, Maryland. This report augments published NWS data with information from NWS files and reports of other federal and state agencies. The following data sets are presented:
In interviews, staff of NWS-HIC and two NWS field offices described their data and recent data collection procedures. NWS-HIC documents and several editions of the NWS Operations Manual provided additional information on past and present procedures. This report describes the nature of the damage estimates and provides a guide to their interpretation and use.
This report critically examines criteria and methods used by the NWS in collecting past and present damage estimates to identify likely sources of inaccuracy. To understand the inaccuracy generally inherent in damage estimation, the report uses statistical comparison methods to assess a California data set containing both preliminary damage estimates and actual cost information. Then it uses similar statistical methods to compare NWS damage estimates with independent estimates from state sources to evaluate the variability in flood damage estimates. Finally, it assesses the impacts of errors and omissions on aggregated damage estimates.
Evaluation results show substantial errors in many of the damage estimates. Uncertainty about the accuracy of the estimates implies that comparisons of flood damage estimates from different flood events or different locations must be undertaken with caution. The report presents examples that illustrate appropriate and inappropriate ways of using the damage data and suggests ways of reducing the impact of errors.
This report is organized as follows. Section 2 describes NWS procedures for obtaining damage estimates and other sources used in compiling the reanalyzed data sets. Section 3 presents the reanalyzed data sets and explains how they were developed. Section 4 describes the types of inaccuracy users should expect in the damage estimates. Section 5 compares damage estimates from different sources and analyzes the accuracy of the estimates. Section 6 suggests ways of dealing with data omissions and inconsistencies. Section 7 provides guidance for use and interpretation of the reanalyzed data, with examples and warnings, and concludes with recommendations regarding future collection and dissemination of flood damage estimates.