38 Universities Council on Water Resources Issue 138, Pages 38-44, April 2008 Assessment of Flood Losses in the United States Stanley A. Changnon University of Illinois: Chief Emeritus, Illinois State Water Survey Private property insurance data on losses resulting from floods in the U.S. during 1972-2006 provide useful new measures of flood losses across the nation, as well as variations over this 35-year period, for floods labeled as catastrophes 531 events causing $1 million or more in losses. Assessment of the flood loss data reveals four types of damaging floods: hurricane-related floods, snow-melt floods, floods associated with major convective storms, and flood-only events. The regional aspects of flood events and their losses as well as temporal patterns were assessed. Planning for and responding to floods requires quality measures of losses created by past floods (Changnon 1999). Unfortunately, much of the flood loss data and information readily available have been based on estimates of losses. Quantified long-term loss data created by assessors who measured each damaged entity have not been available. Existing measures of flood losses include estimates of damages to facilities or crops plus information in news media reports and in government agency calculations of response costs. These sources collectively serve as an approximation of actual flood losses (Pielke 2000). Insurance loss data from the private sector is a more accurate measure of flood losses (National Research Council 1999), but it has not been accessible until provided for this study. The government-operated National Flood Insurance Program does not have long-term data and is not as extensive coverage as the private sector data. Since 1972, the property insurance industry has systematically collected data on major weather-caused loss events, ones they labeled as catastrophes (Property Claims Services 2007). These were losses of $1 million or more per event, and each listed the causative weather conditions such as floods, hail, tornadoes, heavy snow, icing, and hurricanes. The catastrophe loss data from 1972-2006 were used to analyze four types of damaging floods: events caused solely by floods; events caused by snow-melt floods; flood losses from hurricanes; and floods associated with major convective storms including those that also produced tornadoes, hail, and high winds. The values are based on losses to insured property, which includes vehicles, buildings, and other structures, and they represent 90 percent of all storm-produced property losses (Roth 1996, Changnon and Hewings 2001). This assessment addresses the magnitudes of flood losses, flood loss locations within the nation, and the temporal distributions of flood losses. Flood incidences and losses were assessed for each flood type and for all flood catastrophes. Data and Analysis The available data for each catastrophe show the cause of the losses, the amount of loss, the states with losses, and dates of the storm for the 1972-2006 period. In each year since 1972, experts in the property-casualty insurance industry have systematically analyzed the historical catastrophe data to update the past catastrophe values to match the current year conditions. This annual assessment, updated to 2007, resulted in a database allowing one to perform an unbiased comparison of recent catastrophe losses with those in past years, and thus to assess shifting risk of losses in any part of the nation. This adjusted catastrophe data base was provided to the author for this research.
Assessment of Flood Losses in the United States 39 Table 1. Ten states with the highest flood losses from various types of floods during 1972-2006. Rank Floods only Snow-melt floods Hurricane floods Floods & Major storms Total floods 1 Texas New York Florida Texas Texas 2 California Pennsylvania N. Carolina Oklahoma New York 3 Pennsylvania New Jersey S. Carolina Louisiana Florida 4 New York Connecticut Georgia Missouri Pennsylvania 5 New Jersey Maryland Virginia Illinois Missouri 6 Massachusetts Virginia Louisiana Arkansas Louisiana 7 Washington Ohio Alabama Nebraska Virginia 8 Florida Illinois New York Mississippi Illinois 9 Michigan Indiana Pennsylvania Kansas Ohio 10 Oregon Missouri Texas Ohio Oklahoma Figure 1. Frequency of four types of flood catastrophes in each climate region during 1972-2006. This annual loss adjustment effort was a sizable and complex task, requiring assessment of each past event. Three adjustment calculations were made to the original loss value for the locations of each catastrophe. One adjustment corrected for time changes in property values and the cost of repairs and replacements and, hence, this also adjusted for inflation. The second adjustment addressed the relative change in the size of the property market in the areas affected by the catastrophe using census data, property records, and insurance records. This action adjusted losses for shifts in the insured property between the year of a given storm s occurrence and the updated year (2006 in this study). The third adjustment was based on estimates of the relative changes in the share of the total property market that was insured against weather perils in the loss areas; this was done using insurance sales records. These adjustments were used to calculate a revised monetary loss value for each catastrophe to make it comparable to current year values. Thus, adjustments made in 2006 for all past catastrophes dating back to 1972 allowed a comparative temporal assessment of the losses resulting from them (Changnon 2004). For example, a floodrelated loss in Pennsylvania during 1978 was adjusted by the insurance experts upwards by
40 Changnon a factor of 31.3, whereas a 1978 flood loss in Oregon, where coverage and other conditions differed from those in Pennsylvania, was adjusted by 37.8. In the resulting loss values, the loss from a catastrophe, whether in 1976, 1993, or 2004, could be assessed in terms of 2006 conditions. The cause of loss in the insurance data base for 1972-2006 was used to identify those catastrophes caused solely by floods and those caused by floods occurring with other conditions such as heavy snow or hurricanes, permitting spatial and temporal analyses. During this 35-year period there were 531 catastrophes that listed flooding as one of the weather perils causing the event, including 116 events caused by floods alone. Heavy snows produced snow-melt floods in 48 cases; floods occurring with hurricanes totaled 55; and floods with major convective storms, including tornadoes, hail, and high winds, totaled 312. The analysis of the floods was performed for each of these four types of floods plus the 531 total catastrophes. All losses presented are in 2006 dollar values. Flood Losses Flood-Only Events The 116 flood-only events caused a total loss of $9.256 billion for the 35-year period, an average of $79.8 million per flood. The lowest flood loss was $5.9 million and the highest was $1.157 billion. Texas and California had the largest losses, and four states in the Northeast (NY, PA, NJ, and MA) also had high losses from flood-only events (Table 1). The number of states affected by each event was assessed, showing that in 59 events (51 percent of the total floods) losses occurred in only one state. Nineteen floods were in 2 states, 14 in 3 states, 7 events in 4 states, and the remaining 17 floods were in 5 to 7 states. Spatial analysis showed the flood-only events were typically small-scale events, averaging losses in 1 state per event. Many were flash floods. Assessment of the flood-only locations, according to occurrence in the nation s nine climate regions, revealed a maximum in the South (Figure 1). High incidences also occurred in the Central, Northeast, Southeast, and West districts. The regional totals include some catastrophes large enough to have produced losses in 2 or 3 regions, and hence the sum of the events of the nine regions is larger than the nation s 531 floods. The temporal distribution of the flood-only catastrophes during 1972-2006 (Figure 2) shows considerable year-to-year fluctuations. A peak of 9 events came in 2006 with 8 events in 1978. No long-term up or down trend existed, and the annual average was 3.3 events. Figure 2. Temporal distributions of flood-only floods and hurricane floods during 1972-2006.
Assessment of Flood Losses in the United States 41 Losses Caused by Floods from Heavy Snows There were 48 catastrophes when flooding occurred as a result of heavy snows. Flood losses were found to be 32 percent of the total losses of the 48 catastrophes. The snow-melt floods caused $3.365 billion in losses during 1972-2006, an average loss of $69 million per event. Sizes of the snowmelt floods were often large with an average of 8 states affected per event. Assessment of the number of states with losses per event showed the most frequent outcome was 4 states (8 floods) with 6 events affecting 2 states. However, 12 of the 48 catastrophes (25 percent) produced flood losses across 11 to 17 states. The snow-melt flood losses, when sorted according to climate regions in the U.S., show the greatest frequency in the Northeast (Figure 1). The Central district had the second highest number and 13 snow-melt flood catastrophes occurred in the West district. The top ten states based on their losses (Table 1) revealed that the top six were all in the Northeast with peak values in New York and Pennsylvania. The temporal distribution of the 48 catastrophes Figure 3. Temporal distributions of snow-melt floods and floods with major convective storms during 1972-2006. Figure 4. Annual flood losses in the U.S. during 1932-2006.
42 Changnon during 1972-2006 (Figure 3) had no marked up or down trend over time. The annual average was 1.3 events with a peak of 6 snowmelt flood events in 1996. Nine years had no snowmelt flood catastrophes. Losses Caused by Floods with Hurricanes Damaging floods occurred with 55 hurricanes during 1972-2006, averaging 1.7 events per year. Two years, 1985 and 2005, had six hurricane-flood catastrophes. Assessment of the loss magnitudes revealed that the flooding losses were 47 percent of the total hurricane catastrophe loss, and the total losses just from the floods were $53.515 billion. This resulted in an average loss per event of $973 million. Table 1 lists the top ten states based on their losses from hurricane floods. This reveals that losses peaked in Florida and five other states in the Southeast. The areal extent of flood losses averaged 4 states per event. Fourteen events produced losses in only one state, nine events had losses in only 2 states, but 7 catastrophes produced flood losses across 10 to 13 states. The regional distribution of the losses, as expected, showed a maximum in the Southeast climate district with 41 loss events (Figure 1). The South experienced 19 flood events from hurricanes and the Northeast had 13 events. The temporal distribution of the hurricanerelated flood events (Figure 2) has no long-term up or down trend. The early years, 1972-1982, had very few flood events, but thereafter the frequencies increased and showed only a flat time-trend. Assessment of the flood-related catastrophes with losses of $1 billion or more revealed there were 20 during 1972-2006. They totaled $61 billion in flood losses. Fourteen came with hurricanes, 1 with a heavy snow, 4 with largescale convective storms, and 1 flood-only case. Losses from Floods Occurring with Major Convective Storm Systems Floods often occurred with strong convective storm systems that also produced tornadoes, hailstorms, and high winds. Multi-day moderate to heavy rainfall often produced these floods, and several such events produced the record 1993 flood in the Midwest (Kunkel 1996). During the 1972-2006 period, 312 such catastrophes occurred, an average of 9 per year. Flood losses represented 53 percent of the total losses from these 312 catastrophes. The flood losses totaled $27 billion, an average of $88 million per event. These catastrophes were often sizable with flood losses extending across 5 states on average. Forty of the 312 catastrophes produced losses in only one state, and this was a greater frequency than came with 2, 3, 4, or more states with losses. Losses extended into 3 states in 39 floods, and 38 loss events affected 5 states. Eight of the 312 catastrophes had flood losses that extended across 9 to 12 states. A peak of 167 flood losses occurred in the South with a secondary high in the Central region (Figure 1). Losses were highest in three adjacent states (Texas, Oklahoma, and Louisiana) (Table 1). Other nearby states were also among those in the top ten including those in Arkansas, Mississippi, Kansas, Nebraska, and Missouri. The temporal distribution of the 312 catastrophes during 1972-2006 (Figure 3) shows relatively low annual values from 1972 to 1988, with a major peak in floods during 1990-1992 when 53 floods occurred. The 35-year temporal distribution has an upward linear trend over time that was statistically significant at the 0.01 level and agrees with other assessments of the time distributions of major floods (Pielke and Downton 2000). The temporal increase results from three factors. One is the upward trend in annual precipitation in most of the nation (Karl and Knight 1998) and a marked increase in heavy rain events (1 to 7 days) from 1931 to 1998 (Kunkel et al. 1999). Another cause relates to the fact that society has developed growing vulnerability to weather extremes including floods over the past 30 years (Changnon et al. 2000). Platt (2007) has also identified the degradation of aging flood control structures as a factor leading to ever-increasing losses. Total Catastrophes with Floods The 531 flood catastrophes during 1972-2006 show an average of 15 per year. One year (1991) had 30 events and another (1996) had 29; the lowest annual value was 2 floods in 1974. The incidence of flood catastrophes has a distinct upward trend with time during 1972-2006. Flood losses from the 531 events totaled $94 billion, translating into an event average of $176
Assessment of Flood Losses in the United States 43 Table 2. Number of flood-related catastrophes occurring in climate regions during 1972-2006. Climate region (see Figure 1 to identify states in each region) Total losses ($ millions) Percent of catastrophes with flood losses in only one state South 23,769 38 Central 20,439 17 Southeast 15,347 12 Northeast 11,230 7 West North Central 6,083 20 East North Central 5,989 16 Southwest 4,960 35 West 3,610 47 Northwest 2,152 30 million and an annual average of $2.67 billion. Annual insured flood losses during 1972-2006 are shown in Figure 4, depicting large losses in individual years, including 1989, 1992, 1995, and 2005. The values have a distinct upward trend over time that was statistically significant at the 0.01 level. The average area with losses for the 531 catastrophes was 4 states. However, 117 floods (22 percent) produced losses in only one state, the most frequent of all possible state combinations (Table 1). One-state loss floods were more frequent in the West (47 percent) and South (38 percent) regions, and were infrequent in the Southeast, Central, and East North Central regions. Flood losses extended over 12 to 17 states in 24 cases. Flood losses were greatest in the eastern half of the nation and least in the far west with the highest in the South and Central regions. Texas had the greatest number of flood losses, followed by New York, Florida, and Pennsylvania (Table 2). Flood losses are a function of flood frequency and intensity, and also a function of population density and wealth. The high losses in New York, Pennsylvania, Illinois, and Ohio partially reflect high population densities and great societal vulnerability (Changnon et al. 2000). The temporal increase in the 531 flood catastrophes and their losses (Figure 4) reflect shifting societal factors as well as temporal increases in heavy rainfall and the degrading of flood control structures over time. Summary The nation s total property losses from floods 1972-2006 totaled $94 billion, representing an annual average of $2.67 billion and an average of $176 million per event. Assessment of four types of catastrophe-producing floods revealed major spatial and temporal variations in the incidences and loss magnitudes. The average loss for the flood-only catastrophes was $80 million, with $88 million for convective storm floods, $70 million for snow-melt floods, and $973 million for hurricanerelated floods. The average size of loss areas was 1 state for the flood-only cases, whereas the snow-melt floods averaged losses across 8 states. Hurricane floods averaged 4 states with losses, and convective storm floods averaged losses across 5 states. The most frequent loss size was only one state for hurricane, flood-only, and convective storm floods, but 4 states was the most frequent storm size for snow-melt floods. Snow melt floods covered large areas (8 states on average) but had relatively low losses compared to other floods. Hurricane floods had much larger losses but covered smaller areas (4 states). In general, flood studies show that snow melt floods do not create river levels as high as those created by hurricanes (U.S. Geological Survey 2007). The distribution of floods and their losses across the nation varied by type of flood. As expected, hurricane floods were most frequent in the Southeast, whereas snow-melt floods were most common in the Northeast and Central regions. The flood-only cases and the convective storm-related floods were both most frequent in the South and second most frequent in the Central region. Texas was the state with the greatest number of flood-related losses, being first for the flood-only and convective storm cases and being the tenth-ranked state for hurricane flood losses. Other high loss ranked states included New York, Florida, Pennsylvania, and Missouri. Losses were greatest in the South ($23.7 billion) and Central ($20.4 billion) regions, and losses were least in the Northwest and West regions. The temporal distributions of three flood types flood-only, hurricane, and snow-melt exhibited no up or down trend for the 35-year period of
44 Changnon study. Each type had years with no events and 1 or 2 years with 5 or 6 events. In contrast, the convective-storm floods showed a marked upward trend over time. Since they were 312 of the 531 flood catastrophes, this led to a statistically significant upward time trend in the nation s total flood events during 1972-2006. This outcome is not unexpected for three reasons. Climate studies have shown upward trends in annual precipitation and heavy rain events since the 1930s; other studies have identified growing societal vulnerability to floods; and the nation s flood control structures have been degrading with time. Author Bio and Contact Information Stanley A. Changnon has pursued and directed atmospheric and hydrospheric research for 56 years. He directed the atmospheric research program of the Illinois State Water Survey for 15 years and served as the Survey s Chief for six years. As Chief Emeritus, he also serves as a professor in the Geography Department at the University of Illinois. His interests include hydrology; weather and climate extremes; climate variability and change; atmospheric effects on agriculture, water resources, and society; and weather modification. He can be contacted at Illinois State Water Survey, 2204 Griffith Drive, Champaign, IL 61820. E- mail: schangno@uiuc.edu. Telephone (217) 586-5691. References Kunkel, K. E. 1996. A hydroclimatological assessment of the rainfall. In The Great Flood of 1993, ed. Stanley Changnon, 52-67. Westview Press, Boulder, CO. Kunkel, K. E., K. Andsager, and D. Easterling. 1999. Long-term trends in heavy precipitation events over North America. Journal of Climate 12: 2513-2525. National Research Council. 1999. The Costs of Natural Disasters: A Framework for Assessment. National Academy Press, Washington, DC. Pielke, R. A., Jr. 2000. Flood Impacts on Society. Floods, Volume 1. Chapter 8, 133-155. Pielke, R. A., Jr. and M. Downton. 2000. Precipitation and damaging floods: Trends in the US, 1932-1997. Journal of Climate 13: 3625-3337. Platt, R. 2007. Comments on National Flood Insurance Program Evaluation Final Report. Natural Hazards, November, 11-12. Property Claims Services. 2007: The Catastrophe Record for 2006. The Fact Book 2006: Property Casualty Insurance, Insurance Information Institute, New York, NY, 28pp. Roth, R. J. 1996. The property insurance industry. Impacts and Responses of the Weather Insurance Industry Recent Weather Extremes, Changnon Climatologist, Mahomet, IL, 63-82. United States Geological Survey. 2007. Large Floods in the United States: Where They Happen and Why. Circular 1245, Washington, DC. 13 pp. Changnon, S. A. 1999. The historical struggle with floods in the Mississippi River Basin: Impacts of recent floods and lessons for future flood management and policy. Water International 23: 263-271. Changnon, S. A. 2004. Present and future economic impacts of climate extremes in the U.S. Environmental Hazards 5: 47-50. Changnon, S. A., R. A. Pielke, Jr., D. Changnon, R. Sylves, and R. Pulwarty. 2000. Human factors explain the losses from weather and climate extremes. Bulletin American Meteorological Society 81: 437-442. Changnon, S. A., and J. Hewings. 2001. Losses from weather extremes in the United States. Natural Hazards Review 2: 113-123. Karl, T. R., and R. W. Knight. 1998. Secular trends of precipitation amount, frequency and intensity in the United States. Bulletin American Meteorological Society 79: 231-242.