1. Introduction
Flooding is the second deadliest natural hazard in the United States (National Weather Service), and cumulative U.S. flood damage 1988–2017 is estimated to be nearly $200 billion (Davenport et al. 2021). The northeastern United States (hereafter Northeast) experiences more flood fatalities than elsewhere in the country (Ashley and Ashley 2008a), primarily attributable to the combination of high population density and a large number of extreme precipitation events. The Binghamton, New York National Weather Service Weather Forecast Office, which serves central New York and northeast Pennsylvania, reported 31 fatalities related to flash floods in their subregion between 1996 and 2015 (NWS 2023), highlighting the prevalence of local-scale flood hazards. At the regional scale, six floods between 1980 and 2022 in the Northeast climate region caused damage exceeding $1 billion (NOAA/NCEI 2023).
Extreme precipitation (EP) has increased across the Northeast (Groisman et al. 2005, 2004; Karl et al. 2009; Douglas and Fairbank 2011; Matonse and Frei 2013; Kunkel et al. 2013; Walsh et al. 2014; Huang et al. 2017; Howarth et al. 2019), including changes in the extreme precipitation frequency (e.g., Griffiths and Bradley 2007; Kunkel et al. 2013; Thibeault and Seth 2014; Mallakpour and Villarini 2017; Easterling et al. 2017) and magnitude (e.g., Kunkel et al. 2013; Mallakpour and Villarini 2017; Huang et al. 2017; Easterling et al. 2017; Howarth et al. 2019). For example, Huang et al. (2017) found a 53% increase in extreme precipitation in 1996–2014 compared to 1901–95. The observed extreme precipitation increase across the Northeast has considerable implications for flooding, which affects individuals and society in this densely populated region. However, Ivancic and Shaw (2015) found inconsistent relationships between heavy precipitation and flooding, and variability in flooding is more difficult to discern than changes in precipitation due to the multiple and compounding factors that influence floods (Hayhoe et al. 2018).
EP events in the Northeast are caused by a range of storm types and scales. Marquardt Collow et al. (2016) showed that synoptic-scale pressure systems such as extratropical cyclones contribute to widespread heavy precipitation during the warm season, although some intense precipitation in summer months is associated with mesoscale systems. Consistently, Huang et al. (2018) found that 68% of annual 1979–2016 EP came from storms causing EP events at five or more Global Historical Climatology Network–Daily (GHCN-D) stations. Kunkel et al. (2012) showed that annually 47% of extreme events from 1908 to 2009 were associated with extratropical cyclones near fronts, and 36% are associated with tropical cyclones. In an analysis of the storm types associated with extreme rain events from 1999 to 2003, Schumacher and Johnson (2006) showed that multiple storm types are associated with EP in the Northeast, including mesoscale systems, trailing stratiform systems, synoptic systems, and tropical systems. Together, these studies highlight the range of EP event storm types and spatial scales. Inland areas of the Northeast have more frequent but less intense EP events relative to coastal locations (Agel et al. 2015), with coastal EP event storms frequently associated with the North Atlantic storm track (Agel et al. 2019).
In the eastern United States, Peterson et al. (2013) reported more frequent flood observations in streams and rivers on long time scales (87–125 years). For New England, Collins (2009) found that 25 of 28 gauges with long records (56–104 years) showed increasing magnitudes of annual maximum discharges, and these increases were significant at p < 0.05 at six of the gauges. Similarly, Armstrong et al. (2014) showed increases in the flood magnitude at approximately 70% of 75 minimally altered study gauges in the mid-Atlantic region with at least 60 years of record and statistically significant increases in the flood frequency at 27% of these gauges. Hodgkins et al. (2019) found both increasing and decreasing trends in annual peak flows in the Northeast from 1916 to 2015, across a variety of minimally altered, regulated, and urbanized basins. Armstrong et al. (2012, 2014) showed increasing frequency of river floods in the Northeast. In the central United States, Mallakpour and Villarini (2015) revealed that floods are becoming more frequent but not necessarily larger. Globally, Kundzewicz et al. (2014) found an increasing frequency of floods but not an increase in the magnitude of floods.
Many recent studies of floods employ a peak-over-threshold or annual streamflow maximum to identify floods (e.g., Lins and Slack 1999; Armstrong et al. 2014; Mallakpour and Villarini 2015; Burn and Whitfield 2016; Wasko et al. 2020). While these methods can provide insight into how streamflow interacts with the hydrologic system and how streamflow is changing, floods defined by maximum annual flow and streamflow thresholds that are exceeded multiple times per year are typically within the bank and therefore cause minor to no damage (Collins et al. 2022). Both flood occurrence (likelihood of flooding) and damage (impacts to property, health, or life) are greatly affected by factors beyond precipitation and streamflow. For flood occurrence, a range of nonclimatic flood factors such as surficial materials and land use contribute differently to flood susceptibility in urban, coastal, and rural regions throughout the Northeast (Giovannettone et al. 2018; Narayan et al. 2017). This suggests that the same extreme precipitation event may have different hydrologic impacts in different regions, further complicating the relationship between extreme precipitation and flooding in the Northeast. Furthermore, antecedent conditions can substantially change the likelihood of flood occurrence (e.g., Ivancic and Shaw 2015; Yellen et al. 2016). For flood damage, Kundzewicz et al. (2014) showed that flood-related economic losses are increasing globally, largely due to increases in the exposure of assets to flooding. Duan et al. (2016) found that in parts of China, flood occurrence and economic flood damage have increased due to both urbanization and climate change. Furthermore, Merz et al. (2021) concluded that economic growth and population growth have contributed to increasing trends in economic flood damage globally. These studies show that flood occurrence and flood damage are related not only to the frequency and severity of floods but also to changes in the human–environment interface.
Given the difficulties in quantifying changing flood damage directly, linking extreme precipitation to flood damage could provide critical assessments of changing flood risk. As with flooding, reported characteristics and changes in EP are sensitive to definition. EP is often assessed using a high percentile of wet-day precipitation, frequently the 99th (e.g., Huang et al. 2017; Hayhoe et al. 2018; Howarth et al. 2019). However, Schär et al. (2016) demonstrated how employing different EP definitions, including various wet-day percentiles, all-day percentiles, and frequency indices can affect results regionally and globally. Further review and discussion of how EP definition affects precipitation analyses is provided in Pendergrass (2018) and Barlow et al. (2019). In addition, Konrad (2001) demonstrated that characteristics of EP in the eastern United States from 1950 to 1996 are dependent on the spatial scale and that the results may be slightly different depending on the scale of the datasets employed. These studies demonstrate the importance of a priori decisions on studies of EP and its impacts, as well as the value of further study of EP impacts using multiple EP definitions to understand outcomes, such as flooding, more fully.
Few studies to date have sought to link heavy precipitation directly to flood damage. Kundzewicz et al. (2014) summarized the literature comparing changes in precipitation to increases in flood exposure and damages. Pielke and Downton (2000) related state-level National Weather Service annual economic flood losses with a myriad of precipitation variables, including total precipitation, number of precipitation days exceeding 2 in., and the number of 3-day heavy precipitation events per station. Davenport et al. (2021) used state-level flood damage reports to show that increases in damage are attributable in part to observed increases in precipitation. Looking forward, Rashid et al. (2023) forecast the probability of future flood damage through the late twenty-first century based on relationships between heavy precipitation indicators and flood losses from 1996 to 2019. However, Pielke and Downton (2000) showed that while there is a strong relationship between EP and flood damage, the chosen definition of EP strongly affects the strength of the relationships and that considerable variance between the two remains.
There is clear evidence for increasing EP and changing flood patterns in the Northeast, with resulting enhanced risks to life and economy from flooding in the region. While there have been multiple studies exploring the critical relationships between EP, streamflow, and flooding (e.g., Collins et al. 2022; Wasko et al. 2021; Dickinson et al. 2019; Villarini et al. 2014; Collins et al. 2014), the relationships across observed damaging floods, EP, and extreme antecedent wetness (EW) are largely unexplored. Here, we analyze the association between damaging floods and EP and EW in the Northeast at the county scale to improve the application of both EP and EW thresholds as a proxy for damaging flooding. Our data and methods, including flood selection and calculation of extreme precipitation and wetness, are described in section 2. Section 3 outlines our results, including 1) the number and distribution of damaging flood days, 2) the relationship between EP and EW with damaging floods, 3) the sensitivity of our results to the definition of extreme thresholds, and 4) an optimization of these thresholds. Finally, we discuss our results in section 4, followed by a conclusion in section 5.
2. Data and methods
a. Flood selection
We derive reported damaging flood days for the Northeast (region shown in Fig. 1a) from the NOAA Storm Events Database, maintained by the National Centers for Environmental Information (NOAA 2021), from 1996 (the year floods were first included in the database) to 2020. Events are entered into the system through regional National Weather Service offices, where they are gathered from numerous sources such as trained spotters, law enforcement, and emergency management. Efforts are made to verify the events and impacts to ensure the accuracy of the database, but verification is not guaranteed. This database categorizes floods into flash floods, floods, and coastal floods; we consider flash floods and floods in this study. Flood reports from the database were included in this study if the event caused loss of life, injury, or monetary damage (hereafter damaging floods). Location data for the reports include either a county or an NWS weather forecast zone. Therefore, we translated reports from NWS Weather Forecast Zones to the corresponding county. Flood reports that were unable to be linked to a county were omitted from the study. Multiple damaging flood reports on a single day in a county were merged into a single record, producing a list of damaging flood days for each of the 299 counties in the study area. Therefore, hereafter “damaging floods,” “damaging flood days,” and “damaging flood events” refer to days and events with flood reports, respectively, rather than flood events defined by hydrology. Because of the complex role of snowmelt in the surface hydrology of the Northeast, we examine only damaging flood days occurring in climatological summer (JJA) and fall (SON) seasons (hereafter warm season) in this study. While these months may prefer convective storms, this study period captures many damaging floods, consistent with Jessup and Colucci (2012) and Teale et al. (2017), who found that summer and fall are important seasons for flooding in the Northeast, but will also neglect some large flood events (Collins 2019; Dhakal and Palmer 2020; Collins et al. 2022). Focusing on the warm season excludes floods that are partially or primarily driven by snowmelt and would therefore be expected to be less connected to EP and EW.
(a) Location of the study region (darkened states) within the United States. (b) Map of the number of damaging days by county in warm-season months (JJASON) from 1996 to 2020 and (c) histogram of damaging flood days in the Northeast by month, 1996–2020, with JJASON months in black.
Citation: Journal of Applied Meteorology and Climatology 63, 9; 10.1175/JAMC-D-23-0156.1
Although empirical flood damage data have limitations, they provide a necessary link between the natural environment and societal impacts (Ashley and Ashley 2008b; Wing et al. 2020). This dataset contains a great deal of information on storms in the region, but the voluntary reporting structure creates spatial and temporal variances in reporting practices and likely results in underreporting. This is highlighted by thirteen counties with no damaging floods reported, despite being adjacent to counties reporting numerous damaging flood events. Similar inconsistencies and inaccuracies exist in the magnitude of damage (Downton and Pielke 2005); therefore, we only examine the occurrence of damage, fatalities, and injuries associated with floods, rather than their magnitudes. This bottom-up approach makes our analysis scale and type agnostic, capturing floods minor and major, pluvial and fluvial, and driven local convection and tropical cyclones. We further do not conduct any temporal analyses due to the untraceable inconsistencies in reporting practices described above. Given the limitations of the dataset, as well as our specific study objectives, we regard the damaging floods in the dataset as a sample of damaging floods in the region, as opposed to a comprehensive inventory.
b. Precipitation and antecedent wetness data
We use 4-km gridded daily precipitation totals from Daymet (Thornton et al. 2020) from 1996 to 2020, averaged to the county scale. As noted above, storm events that cause EP vary in spatial and temporal scale. Our use of daily, county-level precipitation data, selected to match the temporal and spatial resolution of the flood damage data, limits our ability to capture and assess localized, high-intensity precipitation, such as summer convective systems, that may still result in damage. Precipitation for each flood event is summed over a 1- and 3-day period to account for single and multiday events, respectively. Because the exact time of the flood damage is unknown, we evaluate precipitation for both the 1- and 3-day time periods ending on the flood day and the day before, with the greater precipitation magnitude defined as flood-related precipitation. For most events, 72% for the 1-day period and 84% for the 3-day period, precipitation ending on the flood day is greater than precipitation ending the day before the flood day.
We calculate 1-day EP for each county as the 99th percentile of wet days during the warm season and 3-day EP as the 99th percentile of nonzero 3-day precipitation accumulations. We evaluate antecedent soil wetness using the standardized precipitation index (SPI), which assesses the standard deviation of precipitation accumulation over a specified time period relative to normal precipitation accumulations (McKee et al. 1993). Here, we use 7-, 15-, and 30-day SPI for the 1996–2020 warm-season study period. For assessments of the SPI on damaging flood days, these time windows end on the flood day. Following McKee et al. (1993), we consider SPI ≥ 2.0 to be the extreme antecedent wetness (EW). Precipitation data for SPI calculations are the same data as those used to calculate flood-related precipitation and EP. We refer to EP and EW collectively as “hydroclimatic extremes.”
c. Associating damaging floods with hydroclimatic extremes
Throughout this study, we link damaging floods with hydroclimatic extremes using two perspectives: 1) the percentage of hydroclimatic extremes associated with damaging floods and 2) the percentage of damaging floods associated with hydroclimatic extremes. The first perspective indicates the predictability of damaging floods given EP and EW (i.e., given a hydroclimatic extreme, what is the likelihood of a damaging flood?). The second perspective indicates the ability of EP and EW to capture damaging floods. Using both perspectives shows a more complete picture of the association between the hydroclimatic extremes and damaging floods than one perspective alone. For example, there might be a high likelihood of damaging flooding given a 1-day EP event, but if only a few damaging floods occur during 1-day EP events, the utility of 1-day EP to indicate damaging flooding is limited.
As detailed above, we analyze EP and EW with commonly used statistical thresholds, which, while convenient for a variety of reasons, are also somewhat arbitrary. To further explore the relationship between EP, EW, and damaging floods, we test the sensitivity of our analysis to EP and EW thresholds. Specifically, we increase the EP threshold from the 90.00th–99.99th percentile of warm-season, wet-day precipitation in each county and assess the relationship between each of the resulting 1000 EP thresholds and damaging floods. We consider floods to be related to EP if the flood occurred on or the day after the EP event. We similarly convert the SPI to percentile by county and repeat this for EW thresholds from 90.00th–99.99th percentiles.
We also identify the EP and EW thresholds that capture 50% of the reported damaging floods. These percentiles indicate what thresholds should be used to encompass half of the damaging flood events analyzed. We calculate these thresholds by county to increase their usability given our highly heterogenous study region.
Finally, we combine EP and EW thresholds to better capture the dual importance of short-term heavy rainfall events and antecedent conditions. For this, we first calculate the EP and EW percentiles associated with each damaging flood. Second, we calculate the number of damaging floods included in a county using every combination of EP and EW from the 50th–99th percentiles in increments of 0.1%. Third, we identify percentile EP–EW threshold pairs that capture 50% or more of the floods in that county. Finally, we define the optimal threshold as the EP–EW threshold pair with percentiles that add to the largest sum, deferring to the higher precipitation percentile in the case of tie. This method allows us to account for sensitivity to precipitation and antecedent wetness in each individual county while also minimizing the number of days that exceed the paired EP–EW thresholds but do not have damaging flooding.
3. Results
a. Damaging flood days
From the warm seasons of 1996–2020, there were 3890 damaging flood county days reported in the 299 counties of the Northeast. Because events are defined by county, heavy precipitation on one calendar day can generate damaging floods in multiple counties (hereafter county days). The variety and spatial heterogeneity in the damaging floods reported in each county (Fig. 1b) suggest that the number of damaging floods is subject to local reporting practices, reinforcing the assumption that damaging floods in this study should be interpreted as a sample of damaging floods in the Northeast rather than a comprehensive inventory. For example, several counties in northern Pennsylvania report substantially fewer damaging floods than adjacent counties in western New York. The maximum number of damaging floods reported in a single county is 97 damaging floods from 1996 to 2020, in Allegheny County, Pennsylvania, where the city of Pittsburgh is located. In 13 counties, no damaging floods were reported during the study period. Most damaging floods in the study region were reported in June and July, with decreasing numbers of reports after July and fewest in November (Fig. 1c).
b. EP and flood-related precipitation and antecedent wetness
Wet-day, warm-season EP thresholds increase in magnitude with proximity to the Atlantic Ocean (Figs. 2a,b). In contrast, the average number of EP events per warm season decreases with proximity to the coast (Figs. 2c,d). This illustrates the spatial gradient of EP across the study region. In particular, the top 1% of precipitation (both 1- and 3-day precipitation) in coastal areas is driven by less frequent but larger precipitation events than in the inland areas during the study period (1996–2020). This is consistent with both the larger amount of total precipitation and the decreasing number of wet days per warm season with proximity to the coast (not shown).
(a),(b) The 99th percentile of wet-day precipitation (EP threshold) for each county for 1- and 3-day precipitation, respectively. (c),(d) Average number of EP events per year, for 1- and 3-day precipitation, respectively. (e),(f) Median precipitation accumulation associated with damaging floods in each county. Flood-related precipitation is the precipitation ending on or the day before the flood report, whichever is greater. Gray shading indicates the counties with no reported damaging floods in the study period. For all panels, the warm season, 1996–2020, is shown.
Citation: Journal of Applied Meteorology and Climatology 63, 9; 10.1175/JAMC-D-23-0156.1
Median flood-related precipitation is the greatest in the mid-Atlantic region of the study area, particularly southern New York and New Jersey, and Long Island (Figs. 2e,f). The greatest 1- and 3-day median flood-related precipitation totals are both located in Rockland County, New York, with precipitation of 219.4 and 248.3 mm, respectively.
We assess the SPI on damaging flood days in 7-, 15-, and 30-day increments, showing the median SPI on damaging flood days for each county in Fig. 3. Across all counties, the median SPI associated with damaging floods is positive, indicating wet antecedent conditions, particularly with shortening SPI time scales. For the Northeast overall, approximately 25% of damaging floods occur on days with SPI ≥ 2.0 (Fig. 3d), the threshold for extreme antecedent wetness following McKee et al. (1993). However, some damaging floods in the Northeast have been associated with negative SPI (dry antecedent conditions), as shown by the tails in Fig. 3d.
(a)–(c) Median 7-, 15-, and 30-day SPI associated with damaging floods in each county, respectively; (d) SPI values associated with all damaging floods in the Northeast.
Citation: Journal of Applied Meteorology and Climatology 63, 9; 10.1175/JAMC-D-23-0156.1
c. Relationship between damaging floods and EP and EW
The percentage of EP events associated with damaging floods (given an EP event, what is the likelihood of a damaging flood?) is shown in Figs. 4a and 4b. Generally, 1-day EP is a better indicator of damaging floods than 3-day EP (Figs. 4a,b), with approximately 15%–30% and 10%–20% of 1- and 3-day EP events, respectively, associated with reported damaging floods in most counties. This suggests that heavy, shorter duration (1-day) rain events are more frequently associated with damaging floods across the Northeast in the warm season than longer-duration rain events. Throughout the region, there are many 1- and 3-day EP events that are not associated with damaging floods.
(a),(b) Percent of 1% EP events occurring on or the day before a flood day, and (c),(d) percent of damaging flood events reported on or the day after an EP event. (a),(c) 1-day EP; (b),(d) 3-day EP [note the difference in the color scale between (a),(b) and (c),(d)].
Citation: Journal of Applied Meteorology and Climatology 63, 9; 10.1175/JAMC-D-23-0156.1
The percentage of damaging floods associated with EP is shown in Figs. 4c and 4d. In most counties, 15%–45% of floods are associated with 1- and 3-day EP. In other words, the majority of damaging floods in the region are not associated with extreme precipitation. However, this relationship varies strongly across the 286 counties that reported damaging floods. For example, all floods are associated with 1-day (3-day) EP events in 3.5% (3.1%) of counties, while no floods are associated with 1-day (3-day) EP events in 9.4% (8.0%) of counties. Generally, the percentage of damaging floods associated with EP (Figs. 4c,d) is higher than the percentage of EP events associated with damaging floods (Figs. 4a,b). That is, damaging flood events are more likely to have an EP event associated with them, than EP events are to have a damaging flood associated with them. However, the moderate relationship between EP occurrence and damaging floods suggests that evaluating the frequency of damaging floods by analyzing 1- and 3-day EP events alone is problematic.
The percentage of EW events associated with damaging floods and the percentage of damaging floods associated with EW are shown in Fig. 5. The percentage of EW events associated with damaging floods is generally higher with 7-day EW (Fig. 5a) than with 15-day EW (Fig. 5b). However, for both metrics, no more than 20% of EW events in the Northeast are associated with damaging floods. This correspondence is generally lower with 30-day SPI (Fig. 1 in the online supplemental material).
(a),(b) Percent of 7- and 15-day EW events associated with floods, respectively. (c),(d) Percent of flood days occurring with 7- and 15-day EW, respectively. For 30-day EW associations with damaging floods, see the supplemental material.
Citation: Journal of Applied Meteorology and Climatology 63, 9; 10.1175/JAMC-D-23-0156.1
Across the three antecedent wetness time scales, four to nine of the 299 counties each report that 100% of damaging floods occur with EW (Figs. 5c,d), demonstrating that EW does capture high percentages of damaging floods in some highly localized, disparate parts of the region, particularly those with few damaging flood reports. However, in most counties, 20%–50% of damaging floods occur during 7-day EW (Fig. 5c), and these rates decrease with longer SPI time windows (Fig. 5d, supplemental Fig. 1). Like EP, the moderate relationship between EW occurrence and damaging flood reports indicates that antecedent wetness is not a clear, standalone proxy of floods that cause damage to life and property throughout the Northeast.
To determine if EP and EW could be combined to create a better predictor of damaging flooding, we examine the percentage days with hydroclimatic extremes, specifically 1) EP and not EW, 2) EW and not EP, 3) both EP and EW, and 4) neither EP nor EW, that are associated with floods. The vast majority of extreme events (days with EP or EW) are not associated with a flood (Fig. 6a). EW events are the least likely to be associated with damaging floods. However, the rate of flood coincidence increases when EP occurred simultaneously with EW, with nearly 20% of these days being associated with a damaging flood in the Northeast. This is a slight improvement upon the rate of damaging floods associated with EP alone, showing that combining EP and EW is a slightly better predictor of damaging flooding than either variable alone. A spatial representation of the percentage of EP events in each of these categories associated with floods is provided in supplemental Fig. 2.
(a) Percent of EP, EW, and synchronous EP and EW events associated with damaging floods, as well as the percent of EP or EW events not associated with damaging floods. (b) Percent of damaging floods associated with extreme hydroclimatic events. Floods are inclusive of all counties in the Northeast. For (a) and (b), all time scales are included: EP includes 1- and/or 3-day time scales, and EW includes 7-, 15-, and/or 30-day SPI.
Citation: Journal of Applied Meteorology and Climatology 63, 9; 10.1175/JAMC-D-23-0156.1
We also calculate the percentage of damaging floods associated with 1) EP and not EW, 2) EW and not EP, 3) both EP and EW, and 4) neither EP nor EW (Fig. 6b). Slightly fewer than half (46%) of damaging floods in the Northeast occur in the presence of hydroclimatic extremes (EP, EW, or both). For the region overall, more damaging floods occur with only EW than with only EP. This demonstrates that in the Northeast, there is clear value in using both EP and EW to predict damaging floods, rather than one or the other individually. A spatial representation of the percentage of damaging floods in the Northeast associated with each of these categories is provided in supplemental Fig. 3.
d. Threshold sensitivity and trade-offs
The previous EP analysis uses a common but somewhat arbitrary EP threshold of 99th percentile wet-day precipitation. To explore the relationship between EP threshold and damaging floods in the Northeast, we calculate EP using the 90th percentile of the warm-season, wet-day precipitation through the 99.99th percentile (Fig. 7a). We then compare each EP threshold’s association with damaging floods in the Northeast using the two perspectives described above: 1) the percentage of hydroclimatic extremes associated with damaging floods and 2) the percentage of damaging floods associated with hydroclimatic extremes. This assesses the trade-offs of each EP threshold as a balance between maximizing the likelihood of damaging floods given an EP event, which increases as the EP threshold percentile increases (Fig. 7a, right axis), and the percentage of damaging floods associated with an EP event, which decreases as the EP threshold percentile increases (Fig. 7a, left axis).
(a) For warm-season, wet-day precipitation percentiles from 90.00th to 99.99th, (left) the percent of damaging floods associated with each percentile and (right) the percent of EP events of that percentile associated with a damaging flood. Floods are associated with EP if they occur on the day of or the day after the EP. (b) As in (a), but for 7-day SPI percentiles ranging from 90.00th to 99.99th. Floods are associated with EW if they occurred on the same day as the EW.
Citation: Journal of Applied Meteorology and Climatology 63, 9; 10.1175/JAMC-D-23-0156.1
The percentage of EP events that are coincident with damaging floods (Fig. 7a, right axis) stays relatively low (<15%) until the 98th percentile and then rapidly increases. This is expected, as increased precipitation is more likely to result in a damaging flood. However, not even the highest EP threshold aligns perfectly with damaging flood frequency. Across the Northeast, only 52% of 99.99th percentile EP events are associated with reports of damaging floods (Fig. 7a, right axis), meaning that damage or loss of life is not reported in nearly half of the most extreme precipitation events. Inversely, the percentage of damaging floods associated with an EP event starts at approximately 60% for the 90th percentile and then dramatically decreases starting at about the 97th percentile, with less than 5% of floods co-occurring with 99.99th percentile EP events (Fig. 7a, left axis). This clearly shows that damaging floods can be associated with relatively low levels of EP, and there are other factors that influence whether low levels of EP will contribute to a damaging flood.
The same pattern follows for the SPI converted to percentiles (Fig. 7b); the lower the EW threshold, the more the floods are associated with EW (left axis), and the higher the EW threshold, the more likely an EW event is associated with a damaging flood (right axis). For instance, the percentage of EW events associated with a damaging flood rises from below 2% when EW is defined at the 90th percentile to nearly 14% when the EW threshold is raised to the 99.99th percentile. Meanwhile, the percentage of floods associated with EW events decreases rapidly with increasing EW threshold.
Across EP and EW, these results highlight the importance of threshold definition, as a low threshold will capture more damaging floods but will have more hydroclimatic extreme events that do not result in a damaging flood, and a high threshold will have more hydroclimatic extreme events that do result in a damaging flood but will capture fewer damaging floods.
Given the analysis above, what thresholds should be used? We rank 1-day precipitation associated with each damaging flood and identified the precipitation magnitude required to capture 50% of damaging floods in each county with ≥10 damaging floods (157 counties). We then calculate the percentile of this precipitation magnitude. We find that in the counties with ≥10 damaging floods, the 75th–99.9th percentiles of wet-day precipitation contribute to 50% of damaging floods (Figs. 8a,c). Conversely, this indicates that 50% of damaging floods occur with precipitation below that percentile, emphasizing the limitations of using the 99th, 95th, or even 90th percentile for damaging flooding assessments. This may also indicate the importance of small, intense precipitation events that contribute to nonextreme daily precipitation when averaged to the county scale but can have damaging impacts at highly localized scales.
(a) Warm-season 1-day precipitation percentile associated with 50% of damaging floods. Counties with gray shading reported fewer than 10 damaging floods during the JJASON 1996–2020 study period and are excluded. (b) As in (a), but for 7-day SPI percentiles. (c),(d) The number of counties with ≥10 damaging floods for which half of the damaging floods are represented by the 1-day precipitation and 7-day SPI percentiles, respectively.
Citation: Journal of Applied Meteorology and Climatology 63, 9; 10.1175/JAMC-D-23-0156.1
We find that the 80th–99.9th percentiles of the SPI capture 50% of damaging flooding in northeastern counties (Figs. 8b,d). The spatial patterns of EW thresholds capturing 50% of damaging floods do not match those of EP thresholds capturing 50% of damaging floods, indicating that some counties may be more sensitive to one extreme and that employing a single threshold of EP or EW may not adequately represent damaging flooding throughout the Northeast.
e. Threshold optimization
To overcome the shortcomings of using EP and EW thresholds in isolation, we combine EP and EW thresholds to create a proxy for damaging flooding at the county scale in the Northeast. We calculate the precipitation and SPI percentile thresholds associated with each damaging flood and then identify the maximum sum of EP and EW thresholds that capture 50% of damaging floods, shown in Figs. 9a and 9b. For each county, the values provided in Figs. 9a and 9b give the thresholds of 1-day precipitation with 7-day SPI that represent half of the damaging floods in that county when EP and EW are used together. For example, half of the damaging flooding in York County in southeastern Pennsylvania is best captured through a relatively high EP threshold and low EW threshold; the reverse is true for Windham County in southeastern Vermont. In contrast, similar, more moderate percentiles for both EP and EW best capture damaging flooding in Essex County in northwestern New York.
(a) 1-day EP threshold and (b) 7-day EW threshold optimized together to show the highest percentile coordinates that are associated with 50% of floods in counties with ≥10 damaging floods. (c) Optimized EP and EW percentiles that capture half of the damaging flooding in counties shown in (a) and (b).
Citation: Journal of Applied Meteorology and Climatology 63, 9; 10.1175/JAMC-D-23-0156.1
The optimal thresholds for each of the 157 counties with ≥10 damaging floods are shown in Fig. 9c; the same is shown with standard unitless dimensions of the SPI in supplemental Fig. 4. The optimal thresholds that capture 50% of damaging floods are both above the 90th percentile for EP and EW for approximately 36% (56/157) of northeastern counties with at least 10 damaging floods and both below the 90th percentile in approximately 27% (43/157) of counties. As with Fig. 8, it should be emphasized that half of the damaging floods occur below these thresholds, meaning that these thresholds will still miss half of the floods that affect people, infrastructure, and society in the Northeast. These county-specific thresholds are compared to the percent of the urbanized land area from the U.S. Census Bureau (U.S. Census Bureau 2023) in supplemental Fig. 5; no relationship between the identified county thresholds and percent of area urbanized was discerned at this scale.
4. Discussion
This study links observations of flood damage to extreme hydroclimatic drivers in the Northeast. We navigate the challenges of observational flood damage data related to the accuracy of damage magnitudes outlined in Downton and Pielke (2005), as well as possible changes in reporting practices across the region over the data collection period, by limiting our analyses to the occurrence of damaging flood reports and by excluding temporal assessments. The damaging flood dataset that we used is therefore a sample of the relationship rather than a comprehensive inventory of socioeconomic damage from flooding in the Northeast. Despite these limitations, we find that the use of flood damage observations provides valuable information to strengthen a growing body of literature that primarily relies on theoretical relationships (e.g., Teng et al. 2017; First Street Foundation 2020; Xie et al. 2021).
Most of the damaging flood events are reported in June and July in the Northeast. This aligns with a previously established flood climatology, which shows the greatest frequency of flash floods in a subregion of the study area in the same months (NWS 2023). However, this distribution differs from Collins (2019) and Collins et al. (2022), likely due to the inclusion of pluvial floods and flood events that cause any amount of damage, as well as the reporting biases in the flood database, differences in basin characteristics (such as the amount of built infrastructure), and time period. In addition, the high evapotranspiration, low base flow, and low soil moisture that often occur during Northeast summers make floods defined by high streamflow less common. We found that the number of damaging floods from 1996 to 2020 across the Northeast is heterogeneous but reiterate that this heterogeneity is influenced by variability in reporting practices as well as variability in drainage and other elements of the built environment.
The spatial pattern of the 99th percentile of wet-day precipitation (EP threshold) is similar to Agel et al. (2015) and Huang et al. (2017), with the 99th percentile of precipitation increasing with proximity to the coastline. This pattern may be partially driven by the higher number of inland wet days, which incorporates more days into the EP threshold and can dilute the largest EP events. Median precipitation associated with damaging flood reports is often higher near the coast than it is inland, although there are many exceptions, including the low median flood-related precipitation in coastal Massachusetts and the high median flood precipitation in portions of central Pennsylvania and central Maine. These exceptions highlight that other factors including antecedent conditions as well as land use, topography, catchment size, and the built environment should not be overlooked in assessments of extreme precipitation impacts and necessitate local, data-driven thresholds in considerations of extreme precipitation. Giovannettone et al. (2018) demonstrated that surficial materials and land use contribute differently to flood susceptibility in the lower Connecticut River Valley region. Liu et al. (2020) found that globally projected increases in flood exposure are largely driven by socioeconomic changes rather than climate change. While the spatial discrepancies in the number of damaging floods may be influenced by reporting practices, our results show that consistent with the studies described above, the same extreme precipitation event may be associated with different sociohydrologic impacts throughout the Northeast, a region with diverse land uses, population densities, and infrastructure.
The heterogeneity in the number of damaging flood reports in the Northeast may also be related to heterogeneity in EP events. Individual EP events in the Northeast are generally not spatially cohesive: Agel et al. (2015) found that 90% of extreme events occur at three or fewer stations and 50% occur at single stations. This suggests that while locally there is some consistency in the EP threshold and number of EP events per warm season, the occurrence of individual EP events and therefore the potential for damaging floods is highly localized.
Most damaging flood reports occur when 7-day SPI, 15-day SPI, or 30-day SPI are moderately to very wet (1.0 ≤ SPI < 2, McKee et al. 1993) rather than extremely wet (SPI ≥ 2.0). However, there are more instances of the SPI within this moderate range than there are in the extreme tail of the distribution. Of the floods that occur with SPI ≥ 2.0, the most occur with 7-day SPI ≥ 2.0 and the fewest occur with 30-day SPI ≥ 2.0. This highlights that long time scales of antecedent moisture have less influence on damaging flood occurrence than antecedent moisture of shorter durations. This finding is consistent with Ashley and Ashley (2008a), who report that most flood fatalities in the Northeast are associated with flash flooding and tropical systems rather than longer-term river flooding. Additionally, this relationship is likely influenced by other climatological factors during the warm season, such as high evapotranspiration rates, as well as the effectiveness of drainage systems.
For both EP and EW, the percentage of damaging floods associated with hydroclimatic extremes is higher than the percentage of hydroclimatic extremes associated with damaging floods. Specifically, while 54% of damaging flood reports are not associated with extremes, more than 90% of extreme events (either precipitation or antecedent wetness) are not associated with flood reports. These rates indicate that there are limitations to the utility of using threshold-based definitions of extremes as proxies for damaging flooding and that traditional metrics of extremes should be used with caution when applied to socioeconomic hazards.
Similar to Pielke and Downton (2000) and Schär et al. (2016), our findings also show that the sensitivity of the relationship between EP and flood damage to the EP definition. As the thresholds for EP and EW are increased from the top 10th to higher percentiles, both hydroclimatic extreme events are more likely to be associated with damaging flood events. However, fewer flood reports are associated with EP and EW events at increasing thresholds.
Setting aside this trade-off, even the highest EP thresholds are only associated with damaging flood occurrences approximately half of the time in the Northeast, demonstrating the difficulty of using extreme precipitation as a direct proxy for flooding. However, some of the EP events not associated with damaging floods could be an artifact of underreporting or reflect flood preparedness. Furthermore, many damaging floods occur with precipitation magnitudes below the lowest EP threshold considered in this study. For example, nearly 60% of floods are associated with the 90th percentile of wet-day, warm-season, 1-day EP; therefore, approximately 40% of damaging floods are related to precipitation below this threshold. The prevalence of nonextreme precipitation associated with damaging floods in this study is consistent with the contribution of small, frequent flooding to flood damage in Germany (Merz et al. 2009) and to flood disasters in China (Duan et al. 2016). This is also consistent with Ashley and Ashley’s (2008b) assertion that the vast majority (85%) of entries in a U.S. flood fatality dataset are small events with fewer than five deaths and without much media coverage. Merz et al. (2021) also show that frequent, low-impact events comprise a large part of average annual flood losses globally. While these analyses use a range of datasets and methods, consistent with our results, they highlight that smaller EP and flood events should not be overlooked in flood risk assessments. Emerging research on low-impact flooding has been outlined in Moftakhari et al. (2018).
The heterogeneity in precipitation percentiles capturing half of the damaging flood reports in Northeast counties demonstrates that differences in the natural and built landscape impede the use of a single optimal EP threshold that can be used as a proxy for flooding throughout the Northeast. Rather, when considering local precipitation impacts, the EP threshold should be considered carefully, incorporating the magnitude of potential flood damage and risk tolerance of the affected population.
Some local relationships between EP, EW, and damaging flooding could be inferred from the spatial distribution of the percentage of damaging floods in the Northeast associated with EP only, EW only, both EP and EW, or neither EP nor EW with additional research. For example, future work could explore if counties along the mid-Atlantic where high percentages of floods are associated with both EP and EW in the warm season are resilient to extremely wet conditions unless EP occurs on top of extremely wet conditions. The differences in these relationships identified in this study demonstrate the importance of investigating the relationship between precipitation and flood damage at the county scale to better understand future flood hazards.
Future work should incorporate additional flood damage datasets with greater spatial and temporal resolution and additional damage reports to better define the relationships between heavy precipitation and damaging flooding in the Northeast. One such example may be the Spatial Hazard Events and Losses Database for the United States (SHELDUS; ASU Center for Emergency Management and Homeland Security 2023). Other sources may include satellite imagery or insurance claims. Extreme precipitation is highly localized (Agel et al. 2015) and is obscured by upscaling. In this analysis, we address only the warm season, but an assessment of winter relationships between antecedent conditions, hydroclimatic extremes, and damaging floods may be useful for infrastructure design and flood mitigation, especially given the projected increases in winter extreme precipitation (Picard et al. 2023). Additionally, future work could include other climatological variables such as evapotranspiration and geographical factors such as land use, topography, and population density to produce more precise, informative relationships between hydroclimatic events and damaging flooding, which could then be upscaled to regions and applied to other locations where flood damage observations are unavailable. Finally, Merz et al. (2010) point out that there are also indirect impacts of flooding, such as the disruption of public service, loss of tax revenue due to companies migrating out of the area, emotional effects of trauma and loss of trust in authorities, degraded water quality, and cumulative impacts of minor damage, which are more difficult to ascertain and quantify but could be incorporated into the definition of damaging floods.
5. Conclusions
This study relates the observations of flood damage, including economic losses, injuries, and fatalities, to extreme precipitation and extremely wet conditions in the Northeast. Exploring these relationships is essential given the observed increases in heavy precipitation across the Northeast and the socioeconomic significance of flooding to the region. We identified reported damaging flood days in each Northeast county in July–November from 1996 to 2020 and related these flood events to the magnitude and frequency of 1- and 3-day wet-day 99th percentile precipitation as well as 7-, 15-, and 30-day antecedent wetness. Most of the damaging flood events in the warm season are reported in June and July in the Northeast and thus are more strongly reflected in our results than in months with fewer floods.
We show that damaging flood reports are not distributed evenly throughout counties of the Northeast. This may be an artifact of reporting practices and may also reflect differences in flood vulnerability. While we find that the magnitude of EP increases and the number of EP events decreases with proximity to the coast, the number of damaging floods did not follow either of these spatial patterns. The majority of reported damaging floods occur during wet antecedent conditions, but some occur during dry antecedent conditions. As with EP, there is little spatial cohesion in the median SPI related to damaging flooding from county to county. Generally, shorter time-scale hydroclimatic extremes are more strongly associated with damaging floods than longer time-scale hydroclimatic extremes, for both precipitation and antecedent wetness. In the case of both EP and EW, a greater percentage of damaging floods are associated with hydroclimatic extremes (likelihood of a hydroclimatic extreme given a damaging flood) than hydroclimatic extremes are associated with damaging floods (likelihood of a damaging flood given a hydroclimatic extreme).
The majority of damaging flood reports are not associated with either EP or EW, indicating that many damaging floods are caused by hydroclimatic events that would not be considered extreme. However, while 54% of damaging floods are not associated with extremes, more than 90% of extreme events (either precipitation or antecedent wetness) are not associated with floods. This finding highlights the limitations of using threshold-based definitions of extremes as proxies for damaging flooding.
We explore these limitations to threshold-based EP and EW definitions by testing a range of thresholds and determining how well each represents damaging flooding. For both EP and EW, extreme events defined by lower thresholds capture more of the damaging floods than extreme events defined by higher thresholds. This may be useful when trying to maximize preparedness for flooding events. However, the majority of hydroclimatic extremes at lower thresholds do not trigger damaging floods. For this reason, these low extreme thresholds are limited as a proxy for damaging flooding. Alternatively, extreme events defined by higher thresholds are more likely to be associated with a damaging flood than extreme events defined by lower thresholds. In this case, higher threshold hydrologic extremes have utility in that they are associated with damaging floods consistently. However, most damaging floods occur during hydroclimatic events below these thresholds. This can have substantial societal implications from preparedness, adaptation, and planning.
Our results show that the association between extreme hydrometeorological conditions and flooding is sensitive to the thresholds employed and that there is no set of EP and EW thresholds that is an excellent proxy for reported damaging floods—that is, there is no identified set of hydroclimatic extremes that is both highly likely to be associated with damaging floods and capture for the vast majority of damaging floods. Even when we include only the half of flood events with the highest precipitation and wettest antecedent conditions in each county, the percentiles associated with damaging flooding are below the traditional thresholds of the 99th percentile. In some counties, precipitation as low as the 75th percentile accounts for half of the damaging flooding, indicating that half of the damaging flooding occurs with precipitation below this moderate threshold. We show that flooding in some counties is more sensitive to precipitation than antecedent wetness, and flooding in other counties is more sensitive to antecedent wetness than precipitation. However, our study is limited by several factors, including biases in the reports included in the Storm Events Database, variable local infrastructure, the exclusion of evapotranspiration in summer months, and the ability of convective events to be captured by the scale of this study.
The sensitivity of damaging flooding to the definition of hydroclimatic extremes has ramifications in the way observed, and projected increases in extreme precipitation are discussed in the Northeast, particularly in the context of a changing climate. There are several important needs for future research in this area, including improved consistency in reporting of damaging floods, improved surveys of flood types, and incorporation of local-scale geographic characteristics, which may improve our ability to adapt to both observed and projected increases in extreme precipitation across the Northeast.
Acknowledgments.
We thank the three reviewers for their constructive feedback that improved the paper. The authors have no conflicts of interest to disclose. This research was supported by funding from the Dartmouth College Society of Fellows, Vermont Established Program for Stimulating Competitive Research (NSF Award OIA 1556770), and Miami University.
Data availability statement.
The flood data used in this manuscript are publicly available at https://www.ncdc.noaa.gov/stormevents/. The precipitation data are publicly available at https://daymet.ornl.gov/.
REFERENCES
Agel, L., M. Barlow, J.-H. Qian, F. Colby, E. Douglas, and T. Eichler, 2015: Climatology of daily precipitation and extreme precipitation events in the northeast United States. J. Hydrometeor., 16, 2537–2557, https://doi.org/10.1175/JHM-D-14-0147.1.
Agel, L., M. Barlow, F. Colby, H. Binder, J. L. Catto, A. Hoell, and J. Cohen, 2019: Dynamical analysis of extreme precipitation in the US northeast based on large-scale meteorological patterns. Climate Dyn., 52, 1739–1760, https://doi.org/10.1007/s00382-018-4223-2.
Armstrong, W. H., M. J. Collins, and N. P. Snyder, 2012: Increased frequency of low-magnitude floods in New England. J. Amer. Water Resour. Assoc., 48, 306–320, https://doi.org/10.1111/j.1752-1688.2011.00613.x.
Armstrong, W. H., M. J. Collins, and N. P. Snyder, 2014: Hydroclimatic flood trends in the northeastern United States and linkages with large-scale atmospheric circulation patterns. Hydrol. Sci. J., 59, 1636–1655, https://doi.org/10.1080/02626667.2013.862339.
Ashley, S. T., and W. S. Ashley, 2008a: The storm morphology of deadly flooding events in the United States. Int. J. Climatol., 28, 493–503, https://doi.org/10.1002/joc.1554.
Ashley, S. T., and W. S. Ashley, 2008b: Flood fatalities in the United States. J. Appl. Meteor. Climatol., 47 (3), 805–818., https://doi.org/10.1175/2007JAMC1611.1.
ASU Center for Emergency Management and Homeland Security, 2023: The spatial hazard events and losses database for the United States, version 21.0. Arizona State University, https://cemhs.asu.edu/sheldus.
Barlow, M., and Coauthors, 2019: North American extreme precipitation events and related large-scale meteorological patterns: A review of statistical methods, dynamics, modeling, and trends. Climate Dyn., 53, 6835–6875, https://doi.org/10.1007/s00382-019-04958-z.
Burn, D. H., and P. H. Whitfield, 2016: Changes in floods and flood regimes in Canada. Can. Water Resour. J., 41, 139–150, https://doi.org/10.1080/07011784.2015.1026844.
Collins, M. J., 2009: Evidence for changing flood risk in New England since the late 20th century. J. Amer. Water Resour. Assoc., 45, 279–290, https://doi.org/10.1111/j.1752-1688.2008.00277.x.
Collins, M. J., 2019: River flood seasonality in the northeast United States: Characterization and trends. Hydrol. Processes, 33, 687–698, https://doi.org/10.1002/hyp.13355.
Collins, M. J., J. P. Kirk, J. Pettit, A. T. DeGaetano, M. S. McCown, T. C. Peterson, T. N. Means, and X. Zhang, 2014: Annual floods in New England (USA) and Atlantic Canada: Synoptic climatology and generating mechanisms. Phys. Geogr., 35, 195–219, https://doi.org/10.1080/02723646.2014.888510.
Collins, M. J., G. A. Hodgkins, S. A. Archfield, and R. M. Hirsch, 2022: The occurrence of large floods in the United States in the modern hydroclimate regime: Seasonality, trends, and large-scale climate associations. Water Resour. Res., 58, e2021WR030480, https://doi.org/10.1029/2021WR030480.
Davenport, F. V., M. Burke, and N. S. Diffenbaugh, 2021: Contribution of historical precipitation change to US flood damages. Proc. Natl. Acad. Sci. USA, 118, e2017524118, https://doi.org/10.1073/pnas.2017524118.
Dhakal, N., and R. N. Palmer, 2020: Changing river flood timing in the northeastern and upper midwest United States: Weakening of seasonality over time? Water, 12, 1951, https://doi.org/10.3390/w12071951.
Dickinson, J. E., T. M. Harden, and G. J. McCabe, 2019: Seasonality of climatic drivers of flood variability in the conterminous United States. Sci. Rep., 9, 15321, https://doi.org/10.1038/s41598-019-51722-8.
Douglas, E. M., and C. A. Fairbank, 2011: Is precipitation in northern New England becoming more extreme? Statistical analysis of extreme rainfall in Massachusetts, New Hampshire, and Maine and updated estimates of the 100-year storm. J. Hydrol. Eng., 16, 203–217, https://doi.org/10.1061/(ASCE)HE.1943-5584.0000303.
Downton, M. W., and R. A. Pielke, 2005: How accurate are disaster loss data? The case of U.S. flood damage. Nat. Hazards, 35, 211–228, https://doi.org/10.1007/s11069-004-4808-4.
Duan, W., B. He, D. Nover, J. Fan, G. Yang, W. Chen, H. Meng, and C. Liu, 2016: Floods and associated socioeconomic damages in China over the last century. Nat. Hazards, 82, 401–413, https://doi.org/10.1007/s11069-016-2207-2.
Easterling, D. R., and Coauthors, 2017: Precipitation change in the United States. Climate Science Special Report: Fourth National Climate Assessment, D. J. Wuebbles et al., Eds., Vol. I, U.S. Global Change Research Program, 207–230, https://doi.org/10.7930/J0H993CC.
First Street Foundation, 2020: The first national flood risk assessment: Defining America’s growing risk. FSF Doc., 163 pp., https://assets.firststreet.org/uploads/2020/06/first_street_foundation__first_national_flood_risk_assessment.pdf.
Giovannettone, J., T. Copenhaver, M. Burns, and S. Choquette, 2018: A statistical approach to mapping flood susceptibility in the lower Connecticut River valley region. Water Resour. Res., 54, 7603–7618, https://doi.org/10.1029/2018WR023018.
Griffiths, M. L., and R. S. Bradley, 2007: Variations of twentieth-century temperature and precipitation extreme indicators in the northeast United States. J. Climate, 20, 5401–5417, https://doi.org/10.1175/2007JCLI1594.1.
Groisman, P. Ya., R. W. Knight, T. R. Karl, D. R. Easterling, B. Sun, and J. H. Lawrimore, 2004: Contemporary changes of the hydrological cycle over the contiguous United States: Trends derived from in situ observations. J. Hydrometeor., 5, 64–85, https://doi.org/10.1175/1525-7541(2004)005<0064:CCOTHC>2.0.CO;2.
Groisman, P. Y., R. W. Knight, D. R. Easterling, T. R. Karl, G. C. Hegerl, and V. N. Razuvaev, 2005: Trends in intense precipitation in the climate record. J. Climate, 18, 1326–1350, https://doi.org/10.1175/JCLI3339.1.
Hayhoe, K., and Coauthors, 2018: Our changing climate. Impacts, Risks, and Adaptation in the United States: Fourth National Climate Assessment, D. R. Reidmiller et al., Eds., Vol. II, U.S. Global Change Research Program, 72–144.
Hodgkins, G. A., R. W. Dudley, S. A. Archfield, and B. Renard, 2019: Effects of climate, regulation, and urbanization on historical flood trends in the United States. J. Hydrol., 573, 697–709, https://doi.org/10.1016/j.jhydrol.2019.03.102.
Howarth, M. E., C. D. Thorncroft, and L. F. Bosart, 2019: Changes in extreme precipitation in the northeast United States: 1979–2014. J. Hydrometeor., 20, 673–689, https://doi.org/10.1175/JHM-D-18-0155.1.
Huang, H., J. M. Winter, E. C. Osterberg, R. M. Horton, and B. Beckage, 2017: Total and extreme precipitation changes over the northeastern United States. J. Hydrometeor., 18, 1783–1798, https://doi.org/10.1175/JHM-D-16-0195.1.
Huang, H., J. M. Winter, and B. Beckage, 2018: Mechanisms of abrupt extreme precipitation change over the northeastern United States. J. Geophys. Res., 123, 7179–7192, https://doi.org/10.1029/2017JD028136.
Ivancic, T. J., and S. B. Shaw, 2015: Examining why trends in very heavy precipitation should not be mistaken for trends in very high river discharge. Climatic Change, 133, 681–693, https://doi.org/10.1007/s10584-015-1476-1.
Jessup, S. M., and S. J. Colucci, 2012: Organization of flash-flood-producing precipitation in the northeast United States. Wea. Forecasting, 27, 345–361, https://doi.org/10.1175/WAF-D-11-00026.1.
Karl, T. R., J. M. Melillo, and T. C. Peterson, 2009: Global Climate Change Impacts in The United States. Cambridge University Press, 196 pp.
Konrad, C. E., II, 2001: The most extreme precipitation events over the eastern United States from 1950 to 1996: Considerations of scale. J. Hydrometeor., 2, 309–325, https://doi.org/10.1175/1525-7541(2001)002<0309:TMEPEO>2.0.CO;2.
Kundzewicz, Z. W., and Coauthors, 2014: Flood risk and climate change: Global and regional perspectives. Hydrol. Sci. J., 59, 1–28, https://doi.org/10.1080/02626667.2013.857411.
Kunkel, K. E., D. R. Easterling, D. A. R. Kristovich, B. Gleason, L. Stoecker, and R. Smith, 2012: Meteorological causes of the secular variations in observed extreme precipitation events for the conterminous United States. J. Hydrometeor., 13, 1131–1141, https://doi.org/10.1175/JHM-D-11-0108.1.
Kunkel, K. E., and Coauthors, 2013: Regional climate trends and scenarios for the U.S. National Climate Assessment. Part 1—Climate of the Northeast U.S. NOAA Tech. Rep. NESDIS 142-1, 87 pp.
Lins, H. F., and J. R. Slack, 1999: Streamflow trends in the United States. Geophys. Res. Lett., 26, 227–230, https://doi.org/10.1029/1998GL900291.
Liu, Y., J. Chen, T. Pan, Y. Liu, Y. Zhang, Q. Ge, P. Ciais, and J. Penuelas, 2020: Global socioeconomic risk of precipitation extremes under climate change. Earth’s Future, 8, e2019EF001331, https://doi.org/10.1029/2019EF001331.
Mallakpour, I., and G. Villarini, 2015: The changing nature of flooding across the central United States. Nat. Climate Change, 5, 250–254, https://doi.org/10.1038/nclimate2516.
Mallakpour, I., and G. Villarini, 2017: Analysis of changes in the magnitude, frequency, and seasonality of heavy precipitation over the contiguous USA. Theor. Appl. Climatol., 30, 345–363, https://doi.org/10.1007/s00704-016-1881-z.
Marquardt Collow, A. B., M. G. Bosilovich, and R. D. Koster, 2016: Large-scale influences on summertime extreme precipitation in the northeastern United States. J. Hydrometeor., 17, 3045–3061, https://doi.org/10.1175/JHM-D-16-0091.1.
Matonse, A. H., and A. Frei, 2013: A seasonal shift in the frequency of extreme hydrological events in southern New York State. J. Climate, 26, 9577–9593, https://doi.org/10.1175/JCLI-D-12-00810.1.
McKee, T. B., N. J. Doesken, and J. Kleist, 1993: The relationship of drought frequency and duration to time scales. Eighth Conf. on Applied Climatology, Anaheim, CA, Amer. Meteor. Soc., 179–184, https://www.droughtmanagement.info/literature/AMS_Relationship_Drought_Frequency_Duration_Time_Scales_1993.pdf.
Merz, B., F. Elmer, and A. H. Thieken, 2009: Significance of “high probability/low damage” versus “low probability/high damage” flood events. Nat. Hazards Earth Syst. Sci., 9, 1033–1046, https://doi.org/10.5194/nhess-9-1033-2009.
Merz, B., H. Kreibich, R. Schwarze, and A. Thieken, 2010: Review article “Assessment of economic flood damage”. Nat. Hazards Earth Syst. Sci., 10, 1697–1724, https://doi.org/10.5194/nhess-10-1697-2010.
Merz, B., and Coauthors, 2021: Causes, impacts and patterns of disastrous river floods. Nat. Rev. Earth Environ., 2, 592–609, https://doi.org/10.1038/s43017-021-00195-3.
Moftakhari, H. R., A. AghaKouchak, B. F. Sanders, M. Allaire, and R. A. Matthew, 2018: What is nuisance flooding? Defining and monitoring an emerging challenge. Water Resour. Res., 54, 4218–4227, https://doi.org/10.1029/2018WR022828.
Narayan, S., and Coauthors, 2017: The value of coastal wetlands for flood damage reduction in the northeastern USA. Sci. Rep., 7, 9463, https://doi.org/10.1038/s41598-017-09269-z.
NOAA, 2021: Storm events database. Accessed 1 October 2021, https://www.ncdc.noaa.gov/stormevents/.
NOAA/NCEI, 2023: U.S. billion-dollar weather and climate disasters. Accessed 31 August 2023, https://doi.org/10.25921/stkw-7w73.
NWS, 2023: Central NY/Northeast PA flash flood climatology, https://www.weather.gov/bgm/hydrologyFlashFloodClimo.
Pendergrass, A. G., 2018: What precipitation is extreme? Science, 360, 1072–1073, https://doi.org/10.1126/science.aat1871.
Peterson, T. C., and Coauthors, 2013: Monitoring and understanding changes in heat waves, cold waves, floods, and droughts in the United States: State of knowledge. Bull. Amer. Meteor. Soc., 94, 821–834, https://doi.org/10.1175/BAMS-D-12-00066.1.
Picard, C. J., J. M. Winter, C. Cockburn, J. Hanrahan, N. G. Teale, P. J. Clemins, and B. Beckage, 2023: Twenty-first century increases in total and extreme precipitation across the northeastern USA. Climatic Change, 176, 72, https://doi.org/10.1007/s10584-023-03545-w.
Pielke, R. A., Jr., and M. W. Downton, 2000: Precipitation and damaging floods: Trends in the United States, 1932–97. J. Climate, 13, 3625–3637, https://doi.org/10.1175/1520-0442(2000)013<3625:PADFTI>2.0.CO;2.
Rashid, M. M., T. Wahl, G. Villarini, and A. Sharma, 2023: Fluvial flood losses in the contiguous United States under climate change. Earth’s Future, 11, e2022EF003328, https://doi.org/10.1029/2022EF003328.
Schär, C., and Coauthors, 2016: Percentile indices for assessing changes in heavy precipitation events. Climatic Change, 137, 201–216, https://doi.org/10.1007/s10584-016-1669-2.
Schumacher, R. S., and R. H. Johnson, 2006: Characteristics of U.S. extreme rain events during 1999–2003. Wea. Forecasting, 21, 69–85, https://doi.org/10.1175/WAF900.1.
Teale, N. G., S. M. Quiring, and T. W. Ford, 2017: Association of synoptic-scale atmospheric patterns with flash flooding in watersheds of the New York City water supply system. Int. J. Climatol., 37, 358–370, https://doi.org/10.1002/joc.4709.
Teng, J., A. J. Jakeman, J. Vaze, B. F. Croke, D. Dutta, and S. J. E. M. Kim, 2017: Flood inundation modelling: A review of methods, recent advances and uncertainty analysis. Environ. Modell. Software, 90, 201–216, https://doi.org/10.1016/j.envsoft.2017.01.006.
Thibeault, J. M., and A. Seth, 2014: Changing climate extremes in the northeast United States: Observations and projections from CMIP5. Climatic Change, 127, 273–287, https://doi.org/10.1007/s10584-014-1257-2.
Thornton, M. M., R. Shrestha, Y. Wei, P. E. Thornton, S. Kao, and B. E. Wilson, 2020: Daymet: Daily surface weather data on a 1-km grid for North America, version 4. ORNL Distributed Active Archive Center, accessed 1 September 2022, https://doi.org/10.3334/ORNLDAAC/1840.
U.S. Census Bureau, 2023: County-level 2020 census urban and rural information for the U.S., Puerto Rico, and Island Areas sorted by state and county FIPS codes, https://www.census.gov/programs-surveys/geography/guidance/geo-areas/urban-rural.html.
Villarini, G., R. Goska, J. A. Smith, and G. A. Vecchi, 2014: North Atlantic tropical cyclones and U.S. flooding. Bull. Amer. Meteor. Soc., 95, 1381–1388, https://doi.org/10.1175/BAMS-D-13-00060.1.
Walsh, J., D. Wuebbles, and K. Hayhoe, 2014: Our changing climate. Climate Change Impacts in the United States: The Third National Climate Assessment, J. M. Melillo et al., Eds., U.S. Global Change Research Program, 19–67, http://nca2014.globalchange.gov/report/our-changing-climate/introduction.
Wasko, C., R. Nathan, and M. C. Peel, 2020: Trends in global flood and streamflow timing based on local water year. Water Resour. Res., 56, e2020WR027233, https://doi.org/10.1029/2020WR027233.
Wasko, C., R. Nathan, L. Stein, and D. O’Shea, 2021: Evidence of shorter more extreme rainfalls and increased flood variability under climate change. J. Hydrol., 603, 126994, https://doi.org/10.1016/j.jhydrol.2021.126994.
Wing, O. E. J., N. Pinter, P. D. Bates, and C. Kousky, 2020: New insights into US flood vulnerability revealed from flood insurance big data. Nat. Commun., 11, 1444, https://doi.org/10.1038/s41467-020-15264-2.
Xie, S., W. Wu, S. Mooser, Q. J. Wang, R. Nathan, and Y. Huang, 2021: Artificial neural network based hybrid modeling approach for flood inundation modeling. J. Hydrol., 592, 125605, https://doi.org/10.1016/j.jhydrol.2020.125605.
Yellen, B., J. D. Woodruff, T. L. Cook, and R. M. Newton, 2016: Historically unprecedented erosion from Tropical Storm Irene due to high antecedent precipitation. Earth Surf. Processes Landforms, 41, 677–684, https://doi.org/10.1002/esp.3896.
