Flood Damage in the United States, 1926-2003
A Reanalysis of National Weather Service Estimates

2. Sources of Flood Damage Estimates, 1926-2003

For nearly a century, the NWS and its predecessor, the U.S. Weather Bureau, have collected flood damage estimates through a nationwide system of field offices. The quality of the flood damage estimates is uneven, depending on operational constraints at particular field offices and diverse sources of damage reports. Policies and procedures for collecting and compiling the estimates have changed somewhat in the course of time.

A. Overview of Historical NWS Estimates

The NWS has published flood damage estimates almost annually since 1933. From 1933 to 1975, reporting units were defined by natural boundaries (river basins), which could be useful for local planning on issues such as water supply, agriculture, and flood control. In 1955, annual summaries of damage by state were added. Consistent administration, methodology, and format of the published reports suggest that these data form a reasonably homogeneous time series.

From 1976 through 1979, reduction of funding led to cutbacks in the compilation of flood damage data. Data collection was consistent with prior years, but there appears to have been less checking and updating of initial damage information. Publication of annual summaries ceased. In 1980, compilation of flood damage estimates was discontinued entirely.

In 1983, Congress ordered the U.S. Army Corps of Engineers (USACE) to provide annual reports of flood damage suffered in the U.S. The USACE contracted with the NWS to provide the required data. NWS estimates of flood damage in each state have been published annually since 1983 by the USACE. The NWS Hydrologic Information Center (NWS-HIC) has gradually improved its procedures for compiling and checking the damage estimates.

The long-term consistency in collection of flood damage data results from its connection to weather forecasting and storm warning operations of the NWS. Since at least 1950, reports on severe storms have been submitted regularly to NWS headquarters from field offices distributed across the U.S. The reports include descriptions of severe storms and associated deaths and damage. Since 1959, these reports have been published monthly in a NOAA periodical, Storm Data, and have provided the initial information used in compiling flood damage estimates. However, the field office reports are filed soon after the storm events and receive only minimal quality control before publication, thus the damage estimates provided are preliminary and incomplete. Staff at NWS headquarters perform considerable checking and follow-up to produce final flood damage estimates.

This brief overview highlights a major change in the purpose and format of the flood damage data. Before 1980, the NWS compiled damage estimates for meteorological and hydrological purposes, based on natural units such as watersheds. Annual estimates were compiled by calendar year. Since 1983, the USACE and NWS have prepared flood damage information for Congress, whose members focus on the state as a political unit. Estimates are compiled by federal fiscal year.

B. Present Methods of Compiling Flood Damage Estimates

The staff of NWS-HIC willingly answered our questions about methods used in recent years to collect and compile damage estimates. However, none had direct experience with the methods used before 1989. They provided to us copies of their flood damage data sets and made available all of the materials in their historical archives, including publications of federal agencies, files containing flood reports submitted monthly by the NWS field offices, and notes made by former staff who compiled the data into annual reports.

The NWS operates approximately 120 field offices distributed across the U.S. and its territories. Each office provides weather and hydrological forecasts for an assigned area and issues warnings during severe weather and flood events. Most offices have a Warning Coordination Meteorologist (WCM) who issues storm and flood warnings in the forecast area. The WCM is also responsible for submitting monthly reports on severe storm events to the NWS, including deaths and estimates of damage to property and crops. The descriptions, deaths, and damage estimates are published monthly in Storm Data.

Compiling estimates of storm damage is a minor part of the job, receiving little attention from many WCMs (Frank Richards, NWS-HIC, personal communication, 2/16/00). Field offices differ greatly in the regularity and completeness of their damage reports. Their staff obtain damage estimates from numerous local sources, and cannot always know how those estimates were made and what is included.

A meteorologist at NWS-HIC is responsible for collecting flood damage reports from all of the field offices and checking the damage estimates. NWS-HIC staff are in a good position to track damaging floods because they receive the first flood and flash flood warnings issued by all of the field offices and produce the daily National Flood Summary (NWS-HIC website under Current Flooding). They also receive monthly summaries of significant hydrological events from the field offices. Hence the meteorologist is aware of most flooding events as they occur, receives narrative descriptions monthly, and can check whether estimates are received for all severe floods.

Floods that appear to involve less than $50,000 in damage are entered into the database but generally not checked for accuracy or completeness. When it appears that damage could exceed $50,000, and estimates are missing or seem unreasonable based on descriptions of weather and flood conditions, other reports (e.g. news accounts), and prior experience in compiling damage records, the meteorologist contacts the field office and asks for more information and better estimates. In practice, it is often difficult to clearly separate the estimates of damage to property and crops. Therefore, in recent years, NWS-HIC has combined the estimates of property and crop damage into a single damage estimate.

In most cases, damage information is collected within three months after the flood event. It is most difficult to get the information for large floods because attention in the field office is focused on other more urgent tasks related to the event.

Historically, field office personnel obtained their damage estimates primarily from newspapers (Paul Polger, NWS, pers. comm., 2/16/00). Today, however, they obtain estimates through a variety of contacts in their area such as emergency managers, insurance agents, and local officials. Many offices also subscribe to a newspaper service, which allows the staff to search for any story having to do with weather.

Newspapers and emergency managers are the best sources of information, according to a WCM in Boulder, Colorado (Robert Glancy, NWS, pers. comm., 8/24/01). If a flood has received a presidential disaster declaration, information can be obtained from damage assessments by Federal Emergency Management Agency (FEMA) storm survey teams that travel to the flood scene. Estimates of damage to insured property can be obtained from local insurance agents. However, the estimation process is not performed with rigorous attention to accuracy. One WCM described using the following procedure: Since the largest insurer handles about 25% of the insured property in the local area, an estimate of insured losses is obtained by getting a cost estimate from that insurer and multiplying by four (John Ogren, NWS, pers. Comm., 8/29/01). A full survey of each damaged structure does not take place; instead, in many cases a simplifying formula is used to estimate damage (John Ogren, pers. comm., 8/29/01).

Crop damage estimates are obtained from U.S. Department of Agriculture (USDA) agents or from monthly "flash" reports that are compiled from claims that farmers make to USDA. Damage is calculated based on expected return on the crop: Average yield is multiplied by the number of acres damaged, the estimated percentage of the crop lost, and the expected sale price based on the market at the time of event (John Ogren, NWS, pers. comm., 8/29/01). Unlike property damage, the estimates of crop damage rely on self-reporting by farmers and permit reports to be submitted up to 60 days after the event. After a major flood event market prices often rise so that, by the time of filing, the market price claimed may be higher than the market price at the time of the flood event.

Storm Data's compilers vary widely in terms of training and expertise (Frank Richards, pers. comm., 6/27/01). NWS provides operations manuals to its staff, which explain how to collect and report flood damage. However, one compiler reports that he received most of his training from previous employees who had experience with Storm Data compilation. He was referred to NWS manuals after he had been doing the job for some time (Frank Cooper, pers. comm. 8/27/01).

Instructions for estimating damage have changed in successive versions of the NWS Operations Manual. For example, the 1985 revised manual required that damage estimates be entered by checking off damage categories (though actual dollar amounts could be entered in the narrative section of a report), and specified that damage below $5,000 could be omitted or entered as zero. Furthermore, the manual stated, "Damage resulting from flash floods and floods should be reported only if it is the result of local rainfall but not if it is the result of heavy rain upstream, i.e., that which fell more than 24 to 48 hours in advance of the flooding" (NWS 1985, chap. 42, p. 14). In other words, NWS wished to collect damage estimates only for floods that were the result of localized precipitation. It is uncertain how widely this rule was followed, but it was eliminated less than a decade later. In the 1994 revised manual, instructions simply state, "Damage resulting from flash floods and floods should be reported by each office in whose county area of forecast responsibility the damage was reported." The 1994 revision also eliminated the use of damage categories, specifying that damage estimates should be entered as actual dollar amounts, rounded to three significant digits. The manual further advised, "Focus attention on providing reasonable estimates of larger events (damages greater than $100,000)" (NWS 1994, chap. 42, p. 10).

The field office procedures for collecting flood damage data have some notable strengths and weaknesses. Damage estimators trained by their predecessors are likely to maintain continuity in the data sets, because the training ensures that collection methodology does not change from employee to employee. However, since the NWS operations manual is not always used for guidance, employees may overlook changes in official NWS data collection policies.

C. Sources of Historical NWS Estimates

The NWS and the U.S. Weather Bureau published flood reports regularly in five publications from 1918 through 2001. Table 2-1 summarizes the time periods covered and the information provided by each of these sources. In the early years, damage estimates were published only after major flood events. Annual reporting of flood damage throughout the U.S. commenced in 1933.

From 1934 to 1975, the River and Flood Service published monthly flood reports and annual summaries of flood damage by river basin, first in The Monthly Weather Review and later in Climatological Data National Summary. Two formats were consistently used for the annual summaries, one during 1934-1947, the other during 1948-1975. Annual damage estimates by state for calendar years 1955-1975, and monthly damage estimates for the nation during 1925-1975, were calculated and published in later reports (NWS 1975, 1977).

The 1978 annual summary issue of Climatological Data National Summary announced "Compilation of the General Summary of National Flood Events and Flood Damage Statistics has been delayed. These data will be published later." However publication of Climatological Data National Summary ceased the following year.

For several years after the demise of Climatological Data National Summary, the only published NWS records of flood damage were those included in Storm Data monthly reports. As noted above, these reports often were incomplete and received little checking. Until 1995, most damage estimates were indicated by marking a damage category. (Difficulties of using estimates based on the damage categories are discussed in Section 4.) Until the mid-1970s, the cause of damage was often listed as "heavy rain", rather than "flood", even when flood damage was mentioned in the description. Flood descriptions gradually became more detailed in the 1980s. In general, the flood descriptions provide ample information about precipitation and river flows, but only brief mention of damage.

Table 2-1. Published sources of flood damage estimates from the NWS and U.S. Weather Bureau (WB).

In 1983, when Congress asked the USACE for annual reports of flood damage suffered, Storm Data was the only available nationwide source of damage estimates. Under contract to USACE to provide estimates, NWS-HIC compiled the limited information available. In the years that followed, methods of compiling and checking the estimates were established and gradually improved. These estimates are published annually in the U.S. Army Corps of Engineers Annual Flood Damage Report to Congress (USACE 1983-2001).

In the USACE damage reports from 1983 to 1988, narrative descriptions of floods are quite brief (1/2 to 3/4 page). Many states have no damage estimate but an asterisk (*) indicates that flooding occurred. The 1984 report explains that the table gives a summation of all major flood events but that damage estimates are unavailable for minor flood events. After 1988, the descriptions of flooding and flood damage are more detailed. Beginning in 1991, the asterisk is no longer used and there are few zero entries in the tables. It appears that considerably more record keeping and analysis has gone into damage reports since 1989.

Table 2-2 lists the types of flood loss reported in each of the above publications. From 1933 to 1977, estimates were divided into several categories, separated into property and agricultural damage, compiled by river basin, and presented by calendar year. In 1983, the loss categories, spatial scale, and time period changed. Estimates were summarized by state and fiscal year. In 1993, the distinction between property and agricultural damage was eliminated. Throughout the entire period, estimates focused on direct physical damage, though some data on loss of business and wages were included before 1947. Little is known about the methods used to compile and check the estimates prior to 1980. The published reports themselves show an intent to include all parts of the United States and all types of physical damage.

D. Additional Sources of Flood Damage Estimates

To compile and evaluate a continuous time series of damage estimates, we supplemented the NWS estimates with comparable data from other sources. Comparable estimates should represent direct physical damage in significant flood events. Extensive information would be required to fill the 1976-1982 gap in the state and national estimates. In addition, independent estimates or cost information were needed to assess the accuracy of the estimates. Reports from many sources were used to confirm damage estimates and to provide information about specific floods.

Reports by Federal Agencies and Task Forces:

Several federal agencies prepare reports after severe flood events, in order to study the causes of particular floods and recommend improvements in systems of flood monitoring, warning, or control. Some of these reports include descriptions of earlier floods in the community, and some provide damage estimates.

Table 2-2. Types of flood loss reported during each era.

Post-flood reports prepared by district offices of the U.S. Army Corps of Engineers (USACE) often provide fairly detailed damage estimates that are more complete than NWS estimates because they are compiled many months after the flood event. The Tennessee Valley Authority (TVA) publishes post-flood reports, similar to USACE reports, for areas of the southeastern U.S. under its jurisdiction. Post-flood reports from USGS, NOAA, and the U.S. Weather Bureau usually focus on hydrological and meteorological conditions preceding and during the flood event, with only brief mention of damage. If damage estimates are provided, often they are obtained from the NWS or the USACE.

FEMA has appointed special task forces to study particular major floods and recommend mitigation measures (for example, Interagency Hazard Mitigation Teams for each state affected by the 1993 Midwest flood). Their reports often contain damage estimates.

National Water Summary 1988-1989: Hydrologic Events and Floods and Droughts (USGS 1991) provides historical flood information for all fifty states through 1989. In particular, floods that are considered major historical events for each state are listed, including some damage estimates for individual floods.

State Reports:

State government agencies occasionally publish post-flood reports after particular flood events. To obtain additional, perhaps unpublished, information, we wrote to emergency management agencies in each state, asking them to provide information about historical flood damage. Five states were able to provide long-term historical summaries of their damaging floods, and these proved invaluable for analyzing the accuracy of the NWS estimates (see Section 5). Other states sent shorter-term information which provided useful examples.

Unpublished NWS Damage Information:

The NWS-HIC staff provided copies of their state and national flood damage data sets. These data sets included unpublished estimates for 1976-1982; however, the state and national estimates were found to be incompatible, as described in Section 3. Staff members also gave us access to the historical archives at their office in Silver Spring, MD. Two sets of files proved helpful in understanding how damage estimates were compiled in the past, and were used to supplement estimates for 1976-1982.

Monthly files for 1971-1995 contain the original flood reports from field offices all over the U.S., in no particular order. (These were discontinued when electronic submission of reports began in 1996.) The reports often contain descriptions of damage, but only occasionally provide damage estimates. They do not provide a basis for computing total damage by state or river basin.

Yearly files contain notes made by the people who compiled damage estimates, as well as news clippings and agency communications during the year. These are extremely helpful in developing estimates for 1976-1979, as they contain preliminary annual damage estimates with notes on when and where major floods occurred.

Articles on flash flood damage in 1978 and 1979, published in the journal Weatherwise (Marrero 1979, 1980), were written by José Marrero who had been responsible for collecting the flood damage data formerly published in Climatological Data, National Summary. These articles provide many of our state damage estimates for those years.

E. Summary

The NWS effort to collect flood damage estimates has been remarkably consistent across the nation and over long time periods, resulting in the only source of long-term national flood damage information available in the United States. Similar procedures have been used to obtain estimates from field offices throughout the country, at least since 1950 and perhaps longer. Annual summaries were compiled using consistent methodologies and published in uniform formats during two extended periods, from 1933 through 1975, and from 1983 up to the present.

To create continuous time series of state and national damage estimates requires obtaining compatible estimates for the missing years, 1976-1982. It would also be desirable to base all the data on the same calendar, either fiscal years or calendar years. These tasks are addressed in Section 3.

The accuracy of the damage estimates is uncertain. Methods used to obtain the estimates suggest that they are often educated guesses. For many years they came primarily from newspaper reports. Today, short cuts are often used to extrapolate from a few good sources to make an estimate for an entire community. Evaluation of the accuracy of the estimates is undertaken in Sections 4 and 5.

[ Back ]   [ Next ]