Precipitation Magnitude and Frequency Data Development in the U. S

During the past decade, the Hydrometeorological Design Studies Center (HDSC) of the National Oceanic and Atmospheric Administration (NOAA) of the U.S. embarked on the enormous task of updating frequency analyses for precipitation extremes and development of Atlas 14 volumes (documents) (Bonnin et al., 2006, 2011) for the entire U.S. Frequency analyses were carried out on partial duration series (PDS) for the following 19 durations: 5 minutes, 10 minutes, 15 minutes, 30 minutes, 1 hour, 2 hours, 3 hours, 6 hours, 12 hours, 1 day, 2 days, 3 days, 4 days, 7 days, 10 days, 20 days, 30 days, 45 days, and 60 days. The precipitation magnitudes are available for ten return periods of 1, 2, 5, 10, 25, 50, 100, 200, 500, and 1000 years. Figure 9.1 shows the completed data for precipitation magnitudes for different frequencies at the time this chapter was written (NOAA, 2020b). A total of 11 volumes of Atlas 14 comprising all the states (shown shaded in Figure 9.1) were completed at the time of writing of this chapter.

The precipitation magnitudes for different return periods are available to the public through the web site https://hdsc.nws.noaa.gov/hdsc/pfds/index.html referred to as Precipitation Frequency Data Server (PFDS). Users can provide the location information to obtain the estimates and the corresponding confidence intervals. The precipitation data used by the NOAA for frequency analysis are mainly obtained from the National Centres for Environmental Information - NCEI (formerly National Climatic Data Center - NCDC). The NOAA Atlas 14 time-series data (NOAA, 2020a) are available online as annual maximum series (AMS) and partial duration series (PDS).

Figure 9.1 Map showing the states of the U.S. for which precipitation frequency data are updated (NOAA, 2020).

The values provided for AMS and PDS are constrained and the factors for converting them to unconstrained values are provided in separate reports by the NOAA (2020a). A constrained observation refers to a precipitation measurement bound by clock hours as defined by the NOAA glossary (https://www.nws.noaa.gov/oh/hdsc/glossary. html). The AMS values are available for different temporal scales for 11 different regions (shown in Figure 9.1), and they are Semiarid Southwest (1), Ohio River Basin and surrounding states (2), Puerto Rico and the U.S. Virgin Islands (3), Hawaiian Islands (4), Selected Pacific Islands (5), California (6), Alaska (7), Midwestern States (8), Southeastern States (9), Northeastern States (10), and Texas (11). The PDS values are available only for two regions (1 and 2).

Trends in Historical Precipitation Extremes in the U.S

The NOAA Atlas 14 has developed IDF estimates under the assumption of stationar- ity while providing 90% confidence intervals for all the precipitation magnitudes for different return periods. To evaluate the stationarity assumption used for the development of IDF estimates by the NOAA, trend analysis using Spearman’s Rho (SR) test is conducted for the AMS of historical precipitation datasets available at https://hdsc. nws.noaa.gov/hdsc/pfds/pfds_series.html. Analysis of precipitation extremes for three different durations is presented in this section. The precipitation data used are quality controlled AMS data provided by the NOAA. The SR test is conducted at a 5% significance level. Details of the Spearman’s Rho test can be obtained from Teegavarapu (2018) and Teegavarapu et al. (2019). Trends in the annual daily maximum precipitation totals at the available 10,853 sites in the continental U.S. are shown in Figure 9.2. Only 5.1% and 1.3% of the sites show increasing and decreasing trends, respectively. Small clusters of sites in different regions suggest increasing trends, especially in the northeastern United States. Data at a small number of sites also point to decreasing trends.

Figure 9.2 Spatial variation of trends in annual daily maximum precipitation depths across the majority of continental U.S.

Results from the trend analysis for annual two-day maximum precipitation totals at 10,816 sites are shown in Figure 9.3. Only 4.8% and 0.86% of the sites show increasing and decreasing trends, respectively.

Trends in annual hourly maximum precipitation depths are shown in Figure 9.4 with a total of 3506 sites. Only 4.5% and 1.3% of the sites show increasing and decreasing trends, respectively. A cluster of sites with increasing trends in the northeastern United States and multiple states (viz., Florida, Texas, and California) are noted. While the assumption of stationarity is valid for PFA at most of the sites, the development of IDF relationships at the remaining sites using this assumption may lead to inaccurate estimates of precipitation magnitudes for different return periods. Several sites with

Figure 9.3 Spatial variation of trends in annual two-day maximum precipitation depths across the majority of continental U.S.

Figure 9.4 Spatial variation of trends in annual hourly maximum precipitation depth.

increasing trends are non-uniformly distributed across the country. At most of the sites other than those with increasing or decreasing trends or confirmed nonstation- arity due to regions experiencing abrupt or gradual changes, the development and use of IDF relationships under stationary climate are acceptable for infrastructure design.

Intensity–Duration–Frequency Relationships

Many state agencies, including local or regional water management agencies in the U.S., develop and publish from time to time the IDF relationships or curves for water resources professionals and hydrologists to use in hydrologic modeling and for hydrologic design. Figure 9.5 shows an example of IDF relationships developed for different return periods by the Florida Department of Transportation (FDOT). The precipitation intensity associated with a 25-year 24-hour return period-based precipitation depth for a site (Ft. Lauderdale International Airport, latitude: 26.0719° and longitude: -80.1536° and elevation of 11 ft) in Florida, U.S.A., obtained using the IDF curve shown in Figure 9.3 is 10.56 inches (268 mm).

The precipitation magnitude for the same return period and duration from the recently developed NOAA-based estimates (from Atlas 14, volume 9 version 2) as provided in Table 9.1 is 11.5 inches (292 mm) with a 90% confidence interval of 233 and 381 mm (9.19 and 15.0 inches). The difference in the magnitudes from two different sources can be attributed to (i) statistical methods employed in the development of IDF relationships; (ii) the number of rain gauges considered and the data length considered at the time of development of these relationships.

Design Rainfall Distributions

Temporal distributions referred to as synthetic rainfall distributions or legacy distributions are still being used for hydrologic design purposes in the U.S. These distributions were developed by the Natural Resources Conservation Service (NRCS) and are referred

Figure 9.5 Intensity-Duration-Frequency (IDF) curves for a specific region in Florida, U.S.A. (Florida Department of Transportation (FDOT, available in the public domain https://www.fdot.gov/roadway/drainage/manualsandhand- books.shtm.)

to as Type I, IA, II, and III appropriate for different regions in the contiguous United States. The Type I, Type II, and Type III rainfall distributions shown in Figure 9.6 were developed from rainfall-frequency data contained in the United States (US) Department of Commerce publications US Weather Bureau Technical Paper 40 and National Weather Service (NWS) Hydro-35 (Merkel et al., 2015). The “S” shaped curves show the ratio of accumulated precipitation until a specific time to the total precipitation in a 24-hour storm event. Development of Type I and II distributions has been documented in a report, TP- 149, from the Soil Conservation Service (SCS) in 1973, and information about Type IA and III distributions can be found in Woodward (1974) and Cronshey and Woodward (1989).

Several states (e.g., Florida) in the U.S. have developed regional temporal distributions of rainfall considering the lack of updates to the NRCS rainfall distributions. For example, the State of Florida has developed temporal distributions for different durations (viz., 1, 2, 4, 8, 24, 72, 168, and 120 hours). Rainfall distributions for durations longer than 24 hours are needed as the state is affected by slow-moving hurricane landfalls in the w'et season. There is a need for updating these distributions as and when extreme precipitation data become available. Teegavarapu (2013) documented the changes in the distributions based on the data collected more recently than the

Table 9.1 Precipitation depths for different return periods estimated using AMS by the NOAA for a site in South Florida, U.S.A.

Figure 9.6 NRCS synthetic distributions that apply to different parts of the U.S.

Figure 9.7 NRCS synthetic distribution Type III and distribution curve based on most recent precipitation extremes for a site (Boca Raton) in Florida, U.S.A.

time when these distributions were originally developed. Differences in the distribution values based on the most recent precipitation extremes and synthetic rainfall distribution provided by the FDOT for a site in Florida, U.S.A., are noticeable in Figure 9.7. Inclusion of the most recent precipitation storm data in the calculation of distributions can be the cause of the difference in the distributions. Changes to the characteristics (i.e., early, central, and late peaking) of distributions will have an impact on stormwater management.

The NRCS is currently replacing the legacy rainfall distributions with distributions based on the NOAA Atlas 14 precipitation-frequency data. This effort of the development of temporal distributions by the N RCS is expected to provide more accurate and site-specific distributions that will improve the estimates of peak discharges and hydrographs required for the design.

Changes in Precipitation Extremes

A compilation of results from different studies by the United States Environmental Protection Agency, USEPA (2020), documents the following key points. The points from the USEPA (2020) are listed verbatim here.

• • In recent years, a larger percentage of precipitation has come in the form of intense single-day events. Nine of the top 10 years for extreme one-day precipitation events have occurred since 1990.
• • The prevalence of extreme single-day precipitation events remained fairly steady between 1910 and the 1980s but has risen substantially since then. Over the entire period from 1910 to 2015, the portion of the country experiencing extreme single-day precipitation events increased at a rate of about half a percentage point per decade.
• • The prevalence of extreme single-day precipitation events remained fairly steady between 1910 and the 1980s but has risen substantially since then. Over the entire period from 1910 to 2015, the portion of the country experiencing extreme single-day precipitation events increased at a rate of about half a percentage point per decade.

The insights developed and listed by the USEPA (2020) here are based on the data until 2015. Figure 9.8 shows the changes over time in the percent of land area in the contiguous 48 states of the U.S. receiving one-day maximum precipitation.

Figure 9.9 shows the rate of change in total annual precipitation in different parts of the U.S. for different climate divisions defined by the NOAA. The change is estimated based on data since 1901 for the contiguous 48 states and from 1925 for Alaska. Wright et al. (2019) indicate that conventional analyses neglect trends in extreme rainfall

Figure 9.8 Change in the percent of land area in the contiguous 48 states of the U.S.

receivingone-day precipitation extremes. (Data obtained from the NOAA, 2016.)

Figure 9.9 Changes in the total annual precipitation in different climate divisions. (USEPA, available in the public domain.)

events and the existing hydrological infrastructure and analyses in much of the U.S. may be underperforming due to increases in storm activity. An increase in frequency and intensity of heavy precipitation events in several parts of the U.S. since 1901 has been documented in a recent study by Easterling et al. (2017).

Precipitation Extremes and Floods

In a recent opinion article, Sharma et al. (2018) point to several issues that need to be considered to understand the links between precipitation and flood extremes or peak discharges. These issues include changes to antecedent hydrologic conditions and their impact on flood response; changes in the proportion and persistence of storms arising from different causative mechanisms, such as an increased proportion and frequency of convective extremes; interaction among catchment size and geometry and changing storm characteristics including extent, intensity, and duration; snow cover and snow volume changes and their changing contributions to flood extremes in a warmer climate; the role of land cover change (especially, but not only, urbanization) and the interaction of land cover change with climatic factors. According to Swain et al. (2018), increasing precipitation volatility leading to a rapid transition from multi-year dryness to extreme wetness can have serious consequences on the existing water storage, the resilience of conveyance, and flood control infrastructure.

Increases in precipitation frequency and magnitude have been attributed to increases in temperature or warming trends in many parts of the world. Warmer temperatures generally contribute to water content in the atmosphere via evaporation and increase its capacity to hold water. Changes in precipitation and temperature at a national- and region-scale in the conterminous United States (CONUS) can be evaluated using different long-term observation datasets. Long-term datasets for over 100 years at several sites are available from the U.S. Historical Climatology Network (USHCN) referred to as the USHCN version 2.5 dataset (USHCN v2.5) (Menne et al., 2009). This dataset for mean monthly maximum, minimum temperatures, and total precipitation values was developed using National Climate Data Centre (NCDC) datasets. The current dataset consists of observations from 1218 stations distributed across the U.S. Variation of correlation between positive monthly precipitation totals and the average temperature at 1218 sites for the period 1996-2019 across the U.S. is shown in Figure 9.10. It is evident from the correlations (i.e., Pearson correlation coefficient values) that strong association exists between temperature and precipitation in many regions of the U.S. Also, changes in spatial variations in the correlations from the earlier part of the last century to most recent decades were noted. Antecedent moisture conditions (AMC) of soil influenced by preceding rainfall and extreme precipitation events can influence the intensity of floods. Analyses of joint variations of temperature and precipitation at different temporal resolutions may provide insights into the nature of these changes over time. It is also important to note that precipitation and temperature relationships may also be changing due to climate variability influences.

Figure 9.10 Variation of correlation between monthly precipitation totals and average temperatures at different sites in the conterminous U.S.

Precipitation extremes: influences of climate variability

Climate variability manifested in the form of coupled oceanic and atmospheric interactions referred to as oscillations or teleconnections influences regional hy- droclimatic extremes. Major oscillations influencing precipitation, temperature, and other essential climatic variables in many regions of the world include the Atlantic Multidecadal Oscillation (AMO), Pacific Decadal Oscillation (PDO), North Atlantic Oscillation (NAO), Arctic Oscillation (AO), El Nino Southern Oscillation (ENSO), and Indian Ocean Dipole (IOD). Attempts at understanding these oscillations and their influences in earlier studies either focused on large spatial scales or coarser temporal resolutions with analysis limited to one or two duration-specific precipitation extremes. Understanding the combined influences of two or more oscillations modulating each other with regional hydroclimatology defining the spatial extent of these influences is also critical for hydrological analysis, design, and flood control management.

Several recent studies by the author (Teegavarapu, 2012; Teegavarapu et al., 2013; Goly and Teegavarapu, 2014; Teegavarapu, 2018) have evaluated the influences of one or more oscillations on regional precipitation extremes. These studies have pointed out that two or more phases (viz., warm, cool, and neutral) of oscillations are linked to the (i) temporal shifts in the occurrences of extremes, (ii) spatial variation of extremes, (iii) changes in the magnitudes of extremes of different durations, (iv) changes in the intra-storm temporal distributions of rainfall, (v) changes in depth-duration- frequency (DDF) relationships, (vi) variations in the long-duration rainfall associated with increased frequency and landfall of tropical and extra tropical storms (i.e., hurricanes, typhoons, or cyclones), (vii) changes in the distribution characteristics of precipitation, (viii) antecedent precipitation preceding extreme storm events, (ix) spatial and temporal variations in occurrences of droughts, (x) inter-annual and intra-annual variations, (xi) transition states (as defined by dichotomous events: wet and dry based on specific precipitation magnitude thresholds), (xii) changes in seasonality, (xiii) nature of precipitation extreme storm events (e.g., frontal, convective, and cyclonic), (xiv) variations in persistence, (xv) extreme precipitation magnitudes defined by multiple precipitation indices, (xvi) inter-event time definition based metrics - depth, duration, and gap between events, and (xvii) increased frequency of droughts assessed using the standard precipitation index (SPI). Evidence of some of these changes in precipitation extremes and characteristics due to climate variability is documented in the following select studies. Teegavarapu et al. (2013) in their study of evaluation of influences of the AMO on regional precipitation extremes in the state of Florida, U.S.A., indicated that statistically significant higher precipitation totals were noted in the AMO warm phase compared to the cool phase. Temporal occurrences of extremes shifted from the wet season in the warm phase to the dry season during the cool phase. Goly and Teegavarapu (2014) in their study in the same region (i.e., Florida) reported an increase in the number of hurricanes and tropical storm landfalls and higher antecedent rainfall totals before extreme precipitation storm events in the warm phase of the AMO compared to those in the cool phase. Higher precipitation extremes were noted during El Nino compared to the La Nina phase. Differences in the rainfall temporal distributions for extreme storm events were noted in El Nino and La Nina phases. Coupled influences of the AMO and ENSO on droughts are also reported. Teegavarapu (2018) documented the changes in extremes and temporal occurrences of extremes and spatial variation of droughts in Japan influenced by the PDO and ENSO.

Parametric and non-parametric statistical hypothesis tests are required to confirm statistically significant changes. The non-uniform spatial and temporal influences of coupled oceanic and atmospheric oscillations and changing climate on hydroclimatic variables will have implications for hydrologic design, stormwater and water resources management, and adaptation measures used for protecting regions with adequate infrastructure from evolving climate in different regions of the world.