There are a number of assumptions that are made to complete the analysis that determines the 100-year flood. First, the extreme events observed in each year must be independent from one year to the next. In other words, the 1984 peak river flow cannot be significantly correlated with the observed flow in 1985, which cannot be correlated with 1986, and so on. The second assumption is that the observed extreme events must come from the same probability density function. The third assumption is that the probability distribution relates to the largest storm (measurement of rainfall or river flow) that occurs in any year. The fourth assumption is that the probability distribution function is stationary, meaning that the mean (average), standard deviation, and maximum and minimum values ​​are not increasing or decreasing over time. This concept is known as stationarity. [3]​.

The first assumption is often, but not always, valid and must be tested on a case-by-case basis. The second assumption is usually valid if extreme events are observed under similar climatic conditions. For example, if all recorded extreme events come from late summer thunderstorms (as is the case in the southwestern US), or from snowmelt (as is the case in the north-central US), then this assumption should be valid. However, if there are some extreme events taken from thunderstorms, others from melting ice, and others from hurricanes, then this assumption is most likely not valid. The third assumption is only a problem when trying to forecast a peak but low flow event (e.g., an event less than a 2-year flood). Since this is not usually an objective in extreme analysis or civil engineering design, the situation rarely arises.

The final assumption about stationarity is difficult to test from single-site data due to large uncertainties in even the longest flood records (see next section). More broadly, substantial evidence from climate change strongly suggests that the probability distribution is also changing [4]​ and that managing flood risks in the future will become even more difficult. The simplest implication of this is that most historical data represent the climate of the century and may not be valid for the analysis of extreme events in the century.

When these assumptions are violated, an unknown amount of uncertainty is introduced into the reported value of what the 100-year flood means in terms of rainfall intensity or flood depth. When all inputs are known, uncertainty can be measured in the form of a confidence interval. For example, you could say that there is a 95% probability that the 100-year flood will be greater than X, but less than Y.

Direct statistical analysis [3]​[5]​ to estimate the 100-year river flood is only possible in the relatively few places where an annual series of maximum instantaneous flood discharges has been recorded. In the United States, as of 2014, taxpayers have supported such registries for at least 60 years in fewer than 2,600 locations, for at least 90 years in fewer than 500, and for at least 120 years in just 11. For comparison, the nation's total area is about 3,800,000 square miles (9,800,000 km²), so there are perhaps 3,000 stream reaches draining 1,000 square mile (2,600 km²) watersheds and 300,000 reaches draining 10 square miles (26 km²). In urban areas, 100-year flood estimates are needed for watersheds as small as 1 square mile (2.6 km²). For reaches without sufficient data for direct analysis, 100-year flood estimates are derived from indirect statistical analysis of flood records at other locations in a hydrologically similar region or from other hydrological models. Similarly for coastal flooding, tide gauge data exists for only about 1,450 sites worldwide, of which only about 950 added information to the global data center between January 2010 and March 2016.[6]​.

Much longer records of flood elevations exist in some locations around the world, such as the Danube River in Passau, Germany, but they must be carefully evaluated for accuracy and completeness before any statistical interpretation.

For an individual stream reach, the uncertainties in any analysis can be large, so 100-year flood estimates have large individual uncertainties for most stream reaches. For the largest flood recorded at any specific location, or any potentially largest event, the recurrence interval is always poorly known.[7] Spatial variability adds further uncertainty, because a flood peak observed at different locations on the same stream during the same event commonly represents a different recurrence interval at each location.[7] If an extreme storm drops enough rain on one branch of a river to cause a 100-year flood, but no rain falls on another branch, the flood wave downstream of its junction could have a recurrence interval of only 10 years. In contrast, a storm that produces a 25-year flood simultaneously on each branch could form a 100-year flood downstream. During a flood period, news reports necessarily simplify the story by reporting the greatest damage and greatest recurrence interval estimated at any location. The public can easily and incorrectly conclude that the recurrence interval applies to all stream reaches in the floodplain.

[7]​.

Maximum elevations of 14 floods since 1501 on the Danube River in Passau, Germany, reveal large variability in the actual intervals between floods. Flood events larger than the 50-year flood occurred at intervals of 4 to 192 years since 1501, and the 50-year flood of 2002 was followed only 11 years later by a 500-year flood. Only half of the 50- and 100-year flood intervals were within 50 percent of the nominal average interval. Similarly, intervals between 5-year floods during 1955 to 2007 ranged from 5 months to 16 years, and only half were within 2.5 to 7.5 years.

In the United States, the 100-year flood provides the risk basis for flood insurance rates. A regulatory flood or base flood is routinely established for river reaches through a science-based regulatory process, targeting a 100-year flood at the historical average recurrence interval. In addition to historical flood data, the process takes into account previously established regulatory values, the effects of flood control reservoirs, and changes in land use in the basin. Coastal flood hazards have been mapped with a similar approach that includes relevant physical processes. Most areas where severe flooding can occur in the United States have been consistently mapped in this way. On national average, those 100-year flood estimates are sufficient for the purposes of the National Flood Insurance Program (NFIP) and provide reasonable estimates of future flood risk, if the future is like the past. Approximately 3% of the US population lives in areas subject to coastal flood hazard with a 1% annual chance.[8]​.

In theory, removing homes and businesses from areas that repeatedly flood can protect people and reduce insurance losses, but in practice it is difficult for people to remove themselves from established neighborhoods.[9]​.

The most serious centenary flood recently recorded in Spain occurred in late October 2024 in the Valencia region and surrounding areas. This catastrophe, caused by an Isolated High Level Depression (DANA), resulted in one of the most devastating floods in the country's recent history.

In late October 2024, heavy flooding affected several regions of Spain, especially the province of Valencia, due to an High Level Isolated Depression (DANA).[10]​[11]​ The October 2024 flood in the region of Valencia and surrounding areas was an event of considerable proportions that can be compared to other significant hydrological disasters in the recent history of Spain.[12]​ Two notable events that serve as reference points for contextualization. The magnitude of the 2024 flood:

There are a number of assumptions that are made to complete the analysis that determines the 100-year flood. First, the extreme events observed in each year must be independent from one year to the next. In other words, the 1984 peak river flow cannot be significantly correlated with the observed flow in 1985, which cannot be correlated with 1986, and so on. The second assumption is that the observed extreme events must come from the same probability density function. The third assumption is that the probability distribution relates to the largest storm (measurement of rainfall or river flow) that occurs in any year. The fourth assumption is that the probability distribution function is stationary, meaning that the mean (average), standard deviation, and maximum and minimum values ​​are not increasing or decreasing over time. This concept is known as stationarity. [3]​.

The first assumption is often, but not always, valid and must be tested on a case-by-case basis. The second assumption is usually valid if extreme events are observed under similar climatic conditions. For example, if all recorded extreme events come from late summer thunderstorms (as is the case in the southwestern US), or from snowmelt (as is the case in the north-central US), then this assumption should be valid. However, if there are some extreme events taken from thunderstorms, others from melting ice, and others from hurricanes, then this assumption is most likely not valid. The third assumption is only a problem when trying to forecast a peak but low flow event (e.g., an event less than a 2-year flood). Since this is not usually an objective in extreme analysis or civil engineering design, the situation rarely arises.

The final assumption about stationarity is difficult to test from single-site data due to large uncertainties in even the longest flood records (see next section). More broadly, substantial evidence from climate change strongly suggests that the probability distribution is also changing [4]​ and that managing flood risks in the future will become even more difficult. The simplest implication of this is that most historical data represent the climate of the century and may not be valid for the analysis of extreme events in the century.

When these assumptions are violated, an unknown amount of uncertainty is introduced into the reported value of what the 100-year flood means in terms of rainfall intensity or flood depth. When all inputs are known, uncertainty can be measured in the form of a confidence interval. For example, you could say that there is a 95% probability that the 100-year flood will be greater than X, but less than Y.

Direct statistical analysis [3]​[5]​ to estimate the 100-year river flood is only possible in the relatively few places where an annual series of maximum instantaneous flood discharges has been recorded. In the United States, as of 2014, taxpayers have supported such registries for at least 60 years in fewer than 2,600 locations, for at least 90 years in fewer than 500, and for at least 120 years in just 11. For comparison, the nation's total area is about 3,800,000 square miles (9,800,000 km²), so there are perhaps 3,000 stream reaches draining 1,000 square mile (2,600 km²) watersheds and 300,000 reaches draining 10 square miles (26 km²). In urban areas, 100-year flood estimates are needed for watersheds as small as 1 square mile (2.6 km²). For reaches without sufficient data for direct analysis, 100-year flood estimates are derived from indirect statistical analysis of flood records at other locations in a hydrologically similar region or from other hydrological models. Similarly for coastal flooding, tide gauge data exists for only about 1,450 sites worldwide, of which only about 950 added information to the global data center between January 2010 and March 2016.[6]​.

Much longer records of flood elevations exist in some locations around the world, such as the Danube River in Passau, Germany, but they must be carefully evaluated for accuracy and completeness before any statistical interpretation.

For an individual stream reach, the uncertainties in any analysis can be large, so 100-year flood estimates have large individual uncertainties for most stream reaches. For the largest flood recorded at any specific location, or any potentially largest event, the recurrence interval is always poorly known.[7] Spatial variability adds further uncertainty, because a flood peak observed at different locations on the same stream during the same event commonly represents a different recurrence interval at each location.[7] If an extreme storm drops enough rain on one branch of a river to cause a 100-year flood, but no rain falls on another branch, the flood wave downstream of its junction could have a recurrence interval of only 10 years. In contrast, a storm that produces a 25-year flood simultaneously on each branch could form a 100-year flood downstream. During a flood period, news reports necessarily simplify the story by reporting the greatest damage and greatest recurrence interval estimated at any location. The public can easily and incorrectly conclude that the recurrence interval applies to all stream reaches in the floodplain.

[7]​.

Maximum elevations of 14 floods since 1501 on the Danube River in Passau, Germany, reveal large variability in the actual intervals between floods. Flood events larger than the 50-year flood occurred at intervals of 4 to 192 years since 1501, and the 50-year flood of 2002 was followed only 11 years later by a 500-year flood. Only half of the 50- and 100-year flood intervals were within 50 percent of the nominal average interval. Similarly, intervals between 5-year floods during 1955 to 2007 ranged from 5 months to 16 years, and only half were within 2.5 to 7.5 years.

In the United States, the 100-year flood provides the risk basis for flood insurance rates. A regulatory flood or base flood is routinely established for river reaches through a science-based regulatory process, targeting a 100-year flood at the historical average recurrence interval. In addition to historical flood data, the process takes into account previously established regulatory values, the effects of flood control reservoirs, and changes in land use in the basin. Coastal flood hazards have been mapped with a similar approach that includes relevant physical processes. Most areas where severe flooding can occur in the United States have been consistently mapped in this way. On national average, those 100-year flood estimates are sufficient for the purposes of the National Flood Insurance Program (NFIP) and provide reasonable estimates of future flood risk, if the future is like the past. Approximately 3% of the US population lives in areas subject to coastal flood hazard with a 1% annual chance.[8]​.

In theory, removing homes and businesses from areas that repeatedly flood can protect people and reduce insurance losses, but in practice it is difficult for people to remove themselves from established neighborhoods.[9]​.

The most serious centenary flood recently recorded in Spain occurred in late October 2024 in the Valencia region and surrounding areas. This catastrophe, caused by an Isolated High Level Depression (DANA), resulted in one of the most devastating floods in the country's recent history.

In late October 2024, heavy flooding affected several regions of Spain, especially the province of Valencia, due to an High Level Isolated Depression (DANA).[10]​[11]​ The October 2024 flood in the region of Valencia and surrounding areas was an event of considerable proportions that can be compared to other significant hydrological disasters in the recent history of Spain.[12]​ Two notable events that serve as reference points for contextualization. The magnitude of the 2024 flood: