Abstract

Extreme precipitation can have significant adverse impacts on infrastructure and property, human health, and local economies. This paper examines recent changes in extreme precipitation in the northeast United States. Daily station data from 58 stations missing less than 5% of days for the years 1979–2014 from the U.S. Historical Climatology Network were used to analyze extreme precipitation, defined as the top 1% of days with precipitation. A statistically significant (95% confidence level) increasing trend of the threshold for the top 1% of extreme precipitation events was found (0.3 mm yr−1). This increasing trend was due to both an increase in the frequency of extreme events and the magnitude of extreme events. Rainfall events ≥ 150 mm (24-h accumulation) increased in frequency from 6 events between 1979 and 1996 to 25 events between 1997 and 2014, a 317% increase. The annual daily maximum precipitation, or the highest recorded precipitation amount in a given year, increased by an average of 1.6 mm yr−1, a total increase of 58.0 mm. Decreasing trends in extreme precipitation were observed east of Lake Erie during the warm season. Increasing trends in extreme precipitation were most robust during the fall months of September, October, and November, and particularly at locations further inland. The analysis showed that increases in events that were tropical in nature, or associated with tropical moisture, led to the observed increase in extreme precipitation during the fall months.

1. Introduction

Extreme precipitation significantly impacts infrastructure and property, health, and local economies. In recent years, the Mother’s Day Flood of 2006 in the Northeast cost $1.8 billion and killed 20 people (NCEI 2017) and Hurricane Irene in 2011 cost the Northeast $2.6 billion, left 3.6 million people without power, and killed 22 people (The Associated Press 2012). The northeast United States, which here includes the states of Maine, New Hampshire, Vermont, New York, Massachusetts, Connecticut, Rhode Island, Pennsylvania, and New Jersey, can experience extreme precipitation year-round, with large seasonal and spatial variability, due to snowmelt/heavy snow, convective storms, extratropical cyclones, and tropical cyclones (Agel et al. 2015). Under a changing climate, extreme precipitation is increasing in intensity and frequency across the United States (Walsh et al. 2014), and studies have shown that extreme precipitation is increasing substantially in the Northeast (e.g., Griffiths and Bradley 2007; Brown et al. 2010; Walsh et al. 2014; Frei et al. 2015).

Because of the high-impact nature of extreme precipitation events, it is essential that the temporal changes in the characteristics of these events, including frequency, magnitude, storm type, and how these changes vary spatially and seasonally, are well understood. This paper examines the regional, spatial, and seasonal changes of extreme precipitation in the northeast United States and discusses possible causes of observed increases.

Changes in weather and climate extremes (tropical cyclones, heat waves, cold waves, extreme precipitation, etc.) under the influence of a changing climate are of great concern due to their high-impact nature and their observed increases in intensity and frequency (Walsh et al. 2014). Studies have focused on recent changes in heavy precipitation in the United States (e.g., Karl and Knight 1998; Groisman et al. 2004; Kunkel et al. 1999, 2010, 2013; Peterson et al. 2013; Walsh et al. 2014; Hoerling et al. 2016), as well as changes in the northeastern United States (e.g., Huntington et al. 2004; Griffiths and Bradley 2007; Brown et al. 2010; Frei et al. 2015; Huang et al. 2017). Average precipitation increased 5%–7% in amount nationally over the past century, particularly in the warmer seasons of spring, summer, and fall during the last three decades (Groisman et al. 2004; Griffiths and Bradley 2007; Walsh et al. 2014). However, many studies have shown that the majority of changes in precipitation are occurring as increasing amounts in the upper end of the precipitation distribution, particularly in the extremes (e.g., Karl and Knight 1998; Kunkel et al. 1999; Groisman et al. 1999; Easterling et al. 2000; Groisman et al. 2001; IPCC 2001; Semenov and Bengtsson 2002; Kunkel 2003; Groisman et al. 2004). In addition, several studies, including the Third U.S. National Climate Assessment (https://nca2014.globalchange.gov), have concluded that extreme precipitation itself is increasing in frequency and intensity nationwide (e.g., Karl and Knight 1998; Kunkel et al. 2013; Walsh et al. 2014; Hoerling et al. 2016).

Walsh et al. (2014) found that nationwide average precipitation is increasing the most in the Northeast, which has experienced an 8% increase in precipitation since 1991 (relative to 1901–60). In addition, the amount of precipitation falling during very heavy precipitation events (99th percentile) was found to have increased 71% in the Northeast from 1958 to 2007, the greatest increase nationwide. The frequency of very heavy precipitation also increased 58% from 1958 to 2007, again the largest increase in the United States. If current emissions continue, CMIP5 model ensembles predict the frequency of daily extreme precipitation events will increase 3%–4% in the Northeast in the latter part of this century (2081–2100) compared to the later part of last century (1981–2000; Walsh et al. 2014).

Several studies have specifically examined changes in extreme precipitation in the Northeast and have determined that there is a tendency toward wetter conditions. Griffiths and Bradley (2007) examined changes in heavy precipitation days (days with precipitation ≥ 10 mm), consecutive dry days (maximum number of days with <1 mm of precipitation), very wet days (fraction of annual precipitation due to days exceeding the 95th percentile for the years 1961–90), maximum precipitation in a 5-day period, and daily intensity using 38 U.S. Historical Climatology Network (USHCN) recording stations for the years 1926–2000. Results showed a trend toward wetter conditions in all categories, especially in the later years. The most statistically significant results were associated with heavy precipitation days (95th and 99th percentiles; Griffiths and Bradley 2007).

Brown et al. (2010) assessed long-term (1870–2005) precipitation and temperature trends at 40 stations in the northeast United States and found a trend toward warmer and wetter conditions as well. In addition, the frequency of heavy precipitation days, consecutive wet days, and annual precipitation were all observed to increase during the study period. Since 1893, total precipitation increased 9 mm decade−1, and the trend was significant at nearly one-third of the stations. Though many of the wet-day indices, such as the 95th and 99th percentile, did show an increasing trend, the number and spatial cohesiveness of significant trends was relatively low. The largest spatial trend was observed in coastal stations, where extreme precipitation is most extreme (Agel et al. 2015; Frei et al. 2015), and many had significant increases since 1893 (Brown et al. 2010).

A recent study by Huang et al. (2017) examined both total and extreme precipitation changes from 1901 to 2014 in the Northeast using station data derived from the Global Historical Climatology Network Daily (GHCN-D) dataset as well as gridded observations from Livneh et al. (2015). It was concluded that extreme precipitation increased more than total precipitation [2.4 mm decade−1 (3.6%), and extreme precipitation was 53% higher from 1996 to 2014 than from 1901 to 1995]. Spring (March–May) and fall (September–November) were found to have the largest increases in extreme precipitation (Huang et al. 2017) and the frequency of extreme precipitation was found to have increased most during the warm season (Frei et al. 2015), in which extreme precipitation occurs most often and is most intense (Agel et al. 2015). Frei et al. (2015) also concluded that the increase in extreme precipitation during the warm season (June–October) was due more to an increase in frequency than an increase in magnitude of events (Frei et al. 2015).

Though many studies have observed increases in extreme precipitation in the northeast United States, few have completed a thorough and comprehensive examination of both the observed changes in extreme precipitation and how these changes are manifesting themselves. This paper examines these factors both seasonally and spatially across the Northeast, as well as the possible meteorological causes of the observed changes. This paper focuses on the observed changes of extreme precipitation in the northeast United States regionally, seasonally, and the variability within the region. A thorough examination of temporal changes in the threshold for extreme precipitation, the frequency of extreme precipitation, and the magnitude of extreme precipitation are examined both across the Northeast and the spatial variability of these changes within the Northeast, and on a seasonal basis. In addition, a discussion and preliminary analysis of causes of the observed changes in extreme precipitation is included. These analyses provide a detailed understanding of how and why extreme precipitation is changing. This paper builds a comprehensive understanding of extreme precipitation in the northeast United States in order to increase scientific knowledge and support resiliency planning. The remainder of this paper follows as such 1) methodology of this research, 2) temporal changes in the overall distribution of precipitation, 3) temporal changes in extreme precipitation both regionally and seasonally, 4) the spatial variability of changes in extreme precipitation, 5) possible meteorological causes of extreme precipitation and 6) summary and conclusions.

2. Data and methods

a. Selection of station data

The Northeast is defined as the states of New Hampshire, Maine, Massachusetts, Vermont, New York, Pennsylvania, Rhode Island, Connecticut, and New Jersey. Station data were selected for this study in order to have a spatially comprehensive analysis of precipitation in the Northeast. The USHCN was chosen for its quality-controlled daily precipitation totals taken from National Weather Service first-order weather stations and the U.S. Cooperative Observer Network weather stations where factors such as length of record, number of station moves, and amount of missing data, are taken into account (Easterling et al. 2000). In total, 58 stations reporting daily precipitation and missing less than 5% of days for the years 1979–2014 were chosen (Fig. 1). This is similar to the methods used in Agel et al. (2015), where 35 stations missing less than 1% of days for the years 1979–2008 were chosen from the USHCN database (Agel et al. 2015). For the current study, the criteria for station inclusion was modified to increase station density and the time period was extended to include more recent years (to 2014). Unfortunately, the USHCN does not include years after 2014, and the time period cannot be extended earlier without losing station density. Though the inherent biases in reported observations, such as undercatching snow events or overreporting precipitation totals, are recognized (Daly et al. 2007), the individual station biases and reporting errors are beyond the scope of this study. Using the USHCN database is the most quality-controlled station data that is readily available, and the focus on the long-term trends and large-scale changes in extreme precipitation mitigates individual station biases that may be present.

Fig. 1.

Map of selected stations (58), where black dots indicate station locations.

Fig. 1.

Map of selected stations (58), where black dots indicate station locations.

b. Definition of extreme events and statistical testing

The definition of an extreme precipitation event, for the purposes of this study, is the top 1% (99th percentile) of daily accumulated precipitation on days where precipitation was recorded (days with no precipitation were not included in defining this distribution). The average precipitation was also calculated including only days with precipitation. In addition, for the analysis of seasonal characteristics and changes in extreme precipitation, seasons were defined using the meteorological designation where winter includes the months of December–February (DJF), spring includes the months of March–May (MAM), summer includes the months of June–August (JJA), and fall includes the months of September–November (SON). More details about the methodology for individual analyses will be described in their respective sections.

Though stations were missing less than 5% of days, results were normalized in order to remove biases caused by different amounts of data being included in a given year. To do so, the occurrence of a given variable was divided by the number of available data points. For example, if 20 days in a given year reached the top 1% threshold, that would be divided by the number of available data points within that year.

Monte Carlo methods were used to determine if trends in examined indices were statistically significant. The Monte Carlo methods rely on repeated random sampling of the distribution and remove the possibility that random chance alone explains the observed trend. This method was chosen as extreme events are often rare occurrences (e.g., landfalling tropical cyclones), and these single-occurrence events could, theoretically, strongly skew an observed trend in precipitation. The Monte Carlo methods were applied by first generating 10 000 random samples of the data selected for the given analysis, and then computing the slope of each random sample and creating a distribution of these slopes. If 95% of these slopes are positive, the trend is significant. If the observed trend is still present after the Monte Carlo methods are applied, it can be concluded that the trend is significant and not due to random chance. Trends were proven statistically significant at the 95% confidence level. All temporal figures displayed with 5-yr running averages were determined whether or not they were statistically significant using the Monte Carlo methods.

The Mann–Whitney U test was used to determine if the probability distribution function (PDF) of precipitation amounts between the first half of the time period (1979–96) were statistically different from the PDF of precipitation amounts in the second half of the time period (1997–2014). The Mann–Whitney U test was chosen as it does not require an assumption of normality in the distribution and tests the probability that a randomly selected value from one sample (1979–96) will be less than a randomly selected value from a second sample (1997–2014). The distributions were also proven statistically different at the 95% confidence level.

3. Temporal changes in the distribution of precipitation amounts

It is important to also analyze the entire distribution of precipitation amounts to gain a comprehensive understanding of changes in the characteristics of precipitation from trace to extreme precipitation amounts. The PDFs of precipitation amounts give insight into whether precipitation became heavier on days with precipitation during the time period. The top 1% of precipitation on days with precipitation across all stations for the years 1979–2014 was 51 mm. The time period (1979–2014) was divided into two halves (1979–96 and 1997–2014) in order to objectively observe shifts in the distribution of precipitation during the time period. The PDFs of the bottom 99% of precipitation (<51 mm) and the top 1% of precipitation (≥51 mm) for the first half of the time period (1979–96) and the second half of the time period (1997–2014) were examined (Fig. 2). The PDF for the bottom 99% of precipitation amounts showed a decrease from the first half of the time period to the second half in precipitation amounts of 10–15 mm. However, changes between 15 and 51 mm were not as robust or consistent. These two distributions were not statistically different from each other, indicating there was not a significant change in precipitation amounts in the bottom 99% between the first half of the time period and the second. The PDF of precipitation amounts in the top 1% exhibits a decrease in the percentage of events between 51 and 100 mm from the first half of the time period to the second, but an increase in precipitation amounts above 100 mm from the first half of the time period to the second. The two distributions were shown to be statistically different from each other between the first half of the time period and the second. The significant changes in the two distributions indicates that when extreme events (in the top 1%) occurred, they were more extreme in the second half of the time period than the first.

Fig. 2.

The probability distribution functions (%) for precipitation amounts in the bottom 99% (<51 mm; solid lines; bottom x axis and left y axis) and for precipitation amounts in the top 1% (≥51 mm; dashed lines; top x axis and right y axis) for 1979–96 (blue) and 1997–2014 (red).

Fig. 2.

The probability distribution functions (%) for precipitation amounts in the bottom 99% (<51 mm; solid lines; bottom x axis and left y axis) and for precipitation amounts in the top 1% (≥51 mm; dashed lines; top x axis and right y axis) for 1979–96 (blue) and 1997–2014 (red).

The northeast United States is characterized by high seasonal variability. Therefore, the above analysis was repeated for each season (Fig. 3). The PDFs for the bottom 99% of precipitation in all seasons displayed very little difference between the first half of the time period and the second, and the PDFs were not statistically different. The PDFs for the top 1% of precipitation increased from the first half to the second half of the time period in all seasons. These increases were largest during SON (Fig. 3d). During SON, precipitation occurrences were lower from 59 to 100 mm but higher at precipitation amounts greater than 100 mm. The distributions of the top 1% in SON were statistically different from the first time period to the second.

Fig. 3.

The probability distribution functions (%) for precipitation amounts in the bottom 99% (solid lines; bottom x axis and left y axis) and for precipitation amounts in the top 1% (dashed lines; top x axis and right y axis) for 1979–96 (blue) and 1997–2014 (red) for (a) DJF, (b) MAM, (c) JJA, and (d) SON. Thresholds for the top 1% of precipitation were defined in each season.

Fig. 3.

The probability distribution functions (%) for precipitation amounts in the bottom 99% (solid lines; bottom x axis and left y axis) and for precipitation amounts in the top 1% (dashed lines; top x axis and right y axis) for 1979–96 (blue) and 1997–2014 (red) for (a) DJF, (b) MAM, (c) JJA, and (d) SON. Thresholds for the top 1% of precipitation were defined in each season.

Because the most robust changes occurred in the extreme tail of the distribution, it is imperative that these changes are examined and further motivates the purpose of this paper. The following sections will focus on trends in these extreme events to gain a better understanding of how these events are changing.

4. Temporal changes of extreme precipitation threshold, frequency, and magnitude

Three components of extreme precipitation where chosen in order to examine temporal changes; the threshold for the top 1% of precipitation, the frequency of events at or above the top 1% threshold, and the annual daily maximum of precipitation. These three variables provide insight into the frequency and the magnitude of extreme precipitation and how these factors may be changing.

The magnitude of the top 1% threshold increased 0.3 mm yr−1 with a total increase of 10.8 mm, based on the 5-yr running average, which was determined to be statistically significant (Fig. 4a). To increase the threshold of the top 1% of precipitation, the frequency of events above this threshold must increase, the magnitude of events above this threshold must increase, or both must increase. Therefore, both changes in frequency and magnitude of extreme precipitation will be examined in order to thoroughly investigate the increase in the top 1% threshold.

Fig. 4.

(a) The top 1% threshold each year (mm), (b) extreme event frequency each year (an extreme event is counted when the top 1% threshold being met or exceeded at a station, with the top 1% threshold defined at the individual station), and (c) the annual daily maximum precipitation each year (mm). Annual values are represented with blue dots and the 5-yr running average is represented with a red line for the years 1979–2014.

Fig. 4.

(a) The top 1% threshold each year (mm), (b) extreme event frequency each year (an extreme event is counted when the top 1% threshold being met or exceeded at a station, with the top 1% threshold defined at the individual station), and (c) the annual daily maximum precipitation each year (mm). Annual values are represented with blue dots and the 5-yr running average is represented with a red line for the years 1979–2014.

To determine the frequency of extreme events, an extreme precipitation event was defined as one or more station locations reaching or exceeding the top 1% of precipitation as defined at that particular location. The frequency of extreme events was normalized (dividing the number of extreme days by the number of days included in each year) to remove any influence of more data being recorded in certain years. However, the observed trends and statistical significance of these trends remained unchanged compared to the nonnormalized frequency (not shown). Therefore, the nonnormalized frequency of extreme events was used for its simplicity of understanding. The frequency of extreme precipitation events increased 0.4 events yr−1 based on the 5-yr running average (Fig. 4b). This is a total average increase of 15 events yr−1 and was a statistically significant increase.

To measure the magnitude of extreme events, the maximum daily precipitation amount reached in the domain on any day within the year (annual daily maximum) was used. The annual daily maximum precipitation increased on average 1.61 mm yr−1, a total increase of 58.0 mm, which was statistically significant (Fig. 4c). The increase in magnitude of this parameter indicates that the top end of the distribution (top 1%) is increasing, but that the maximum value of precipitation is rising to a greater degree. The intensification of extreme precipitation amounts and the increase in frequency of extreme precipitation events indicates a likely upsurge in both magnitude and frequency of the associated weather patterns/events that produce extreme precipitation (i.e., extreme weather events such as hurricanes and thunderstorms may be growing stronger and becoming more frequent).

There was large interannual variability in extreme events, both in amount and frequency. Annually, the threshold for extreme events (top 1%) ranged from 41.9 (1997) to 66.0 mm (2010), and the annual daily maximum ranged from 91.4 (1993) to 298.2 mm (1996). The frequency of extreme events ranged from 20 (1986) to 52 events (2005). The large interannual variability in the frequency of extreme events is due to the inherent nature of extreme events, in that the frequency and amount of precipitation depends on the spatial extent and the area of impact, both of which are dependent on the event and the event type. Because of this variability, Monte Carlo methods to test for significance were chosen, and this variability is why determining the types of weather events that cause extreme precipitation and the observed changes is important.

To further examine changes in the extreme end of the precipitation distribution, extreme event thresholds were determined. Objective definitions of 50, 100, and 150 mm, where 50 mm is the approximate top 1% threshold across all years, were selected. These event thresholds were defined and the number of days with precipitation that met or exceeded these thresholds. If any location in the domain reached the designated threshold on a given day, the event was counted once, regardless of how many station locations were effected. Events were also counted based on the number of locations that reached the designated threshold. Under this definition, a single weather event on a given day could be counted multiple times if several station locations met the designated threshold (not shown). By using these two definitions, a sense of the trend in both size and frequency of extreme events could be determined. Each definition showed similar increases and no change in the statistical significance of trends. The similar increase in both of these definitions indicates that the spatial extent of extreme events is not growing, but the number of events is. If this were the reverse, only an increase in the first definition would be seen. Results of the first definition (where events are only counted once) is presented below.

The number of events of ≥50, ≥100, and ≥150 mm all increased at a statistically significant rate. In addition, the number of events in the first half of the included time period (1979–96) to the second half (1997–2014) was compared (Fig. 5). Though the current study examined the two halves of the time period, future work could further analyze temporal trends by dividing the time period into more groups or determining changepoints in the observed trends. At the 50-mm threshold, there were 594 events in the first half and 719 events in the second half, an increase of 125 events, or 21% (Fig. 5a). At the 100-mm threshold, there were 56 events in the first half and 89 events in the second half, an increase of 33 events (59%; Fig. 5b). At the 150-mm threshold, the most extreme, there were only six events in the first half, but 25 events in the second half, an increase of 19 events (317%; Fig. 5c). The increases in the frequency of these thresholds were statistically significant and indicate a substantial increase in events more severe than the top 1% that have large societal impacts. For example, Hurricane Irene (2011) dropped nearly 300 mm of precipitation in New Jersey (Avila and Cangialosi 2011), and the Mother’s Day Flood (2006) recorded over 350 mm of precipitation in New Hampshire (Olson 2007). Both events fall within the top 1% percentile of the distribution.

Fig. 5.

The number of events (a) ≥50, (b) ≥100, and (c) ≥150 mm. The years 1979–96 are represented in blue, and the years 1997–2014 are represented in green.

Fig. 5.

The number of events (a) ≥50, (b) ≥100, and (c) ≥150 mm. The years 1979–96 are represented in blue, and the years 1997–2014 are represented in green.

Analyzing a histogram of precipitation events ≥ 50 mm gives insight into both changes in magnitude and frequency of these extreme events. The histogram is similar to the probability distributions (Fig. 2), but examines the raw data rather than a percentage calculation. Unlike Fig. 5, these events were counted based on the number of reports (so an event could be counted multiple times if more than one station reached the designated threshold on a given day), and the time period (1979–2014) was divided into two halves where the first half included the years 1979–96 and the second half included the years 1997–2014. In comparing the two halves of the time period, an increase in both frequency and magnitude of events was observed (Fig. 6).

Fig. 6.

The frequency of precipitation events ≥ 50 mm for 1979–96 (gray) and 1997–2014 (blue). Frequency of events is plotted on a log scale.

Fig. 6.

The frequency of precipitation events ≥ 50 mm for 1979–96 (gray) and 1997–2014 (blue). Frequency of events is plotted on a log scale.

The median rainfall in events ≥ 50 mm during the first half of the time period was 63.5 mm, while in the second half of the time period it was 64.3 mm. However, in examining only the events of ≥100 mm, the median during the first half of the time period was 112.1 mm, while during the second half it increased to 123.9 mm. It is important to remember that these are high-impact events, where 50 mm is equivalent to 2 in. and 100 mm is approximately 4 in. An increase in frequency of 50–60-mm events of approximately 150 events from the first half to the second half is substantial, as these events can cause flooding. Though not as robust, particularly because these are infrequent events, events of ≥100 mm increased as well. For example, events of 150 mm or greater (particularly events of 200 mm or greater) occurred either rarely or never during the first half of the time period, but occurred up to 10 times during the second half of the time period. Though the increase of these events was not as large as the increase in lesser extreme events (50–100 mm), the impact would be equal, if not more, substantial because of their flooding potential (i.e., Hurricane Irene and the Mother’s Day Flood). This clearly indicates that both an increase in the frequency and the magnitude of extreme events are increasing and causing the increase in the top 1% threshold.

In addition to the PDF of seasonal precipitation (Fig. 3), a histogram of the top 1% of precipitation events (≥50 mm) was completed for each season (Fig. 7). Every season experienced an increase in both the intensity and frequency of extreme precipitation events. Though the temporal changes in the top 1% threshold, the frequency of extreme events, and the magnitude of extreme events was not directly examined seasonally in this paper, values for changes in these variables can be found in Table 1 and support the conclusions based on both the seasonal PDFs (Fig. 3) in section 3 and the seasonal histograms discussed below (Fig. 7).

Fig. 7.

The frequency of events ≥ 50 mm for (a) DJF, (b) MAM, (c) JJA, and (d) SON for 1979–96 (gray) and 1997–2014 (blue). Frequency displayed on a log scale.

Fig. 7.

The frequency of events ≥ 50 mm for (a) DJF, (b) MAM, (c) JJA, and (d) SON for 1979–96 (gray) and 1997–2014 (blue). Frequency displayed on a log scale.

Table 1.

Table of changes in the top 1% threshold, the frequency of extreme events, and the annual daily maximum for each season. The average annual change and the total change for the time period are displayed. Statistically significant trends are marked with an asterisk.

Table of changes in the top 1% threshold, the frequency of extreme events, and the annual daily maximum for each season. The average annual change and the total change for the time period are displayed. Statistically significant trends are marked with an asterisk.
Table of changes in the top 1% threshold, the frequency of extreme events, and the annual daily maximum for each season. The average annual change and the total change for the time period are displayed. Statistically significant trends are marked with an asterisk.

DJF displayed the smallest change between the two time periods, where the median magnitude of events ≥ 50 mm in the first half was 58.9 mm and the median in the second half was 59.7 mm. For events ≥ 100 mm, the median in the first half was 104.4 mm and the median decreased in the second half to 103.4 mm (Fig. 7a). Extreme events of this magnitude rarely occur during the cold season and have shown to be increasing the least compared to the other seasons, as shown in previous results (Table 1). MAM experienced very little change in the median of events ≥ 50 mm (59.9 mm, 60.7 mm) but the median of events ≥ 100 mm increased from 107.3 to 126.2 mm, the largest increase across all of the seasons (Fig. 7b). JJA also had a small increase in the median for precipitation events ≥ 50 mm (61.5 mm, 61.7 mm), and a slightly larger increase for events ≥ 100 mm (114.8 mm, 118.9 mm; Fig. 7c). SON had the largest shift toward more extreme events in the second half of the time period (Fig. 7d). The median of precipitation events ≥ 50 mm for 1979–96 was 60.8 mm, while the median from 1997 to 2014 was 64.5 mm. The median for events ≥ 100 mm from 1979 to 1996 was 118.4 mm and increased to 126.8 mm in the second half. Though MAM had the largest increase in the median, this is likely due to the smaller number of extreme events, and as the tail end of these extreme events increased, this in turn easily increased the median. Events of these magnitudes were more common during SON, so while the median did not increase as drastically as in MAM, the increase in the most extreme events was more robust.

In summary, both the frequency and magnitude of extreme events increased from 1979 to 2014 at a statistically significant rate, particularly in the warm season. Because the Northeast is characterized by high spatial variability in its weather systems, the following section provides a spatial analysis of these trends.

5. Spatial analysis of changes in extreme precipitation

The spatial variability of observed changes in the top 1% provides insight into the possible changes in the associated weather patterns, such as changes in storm track or inland extent. To examine the spatial variability of changes in extreme precipitation, the statistical significance and direction of the trend (positive or negative) is calculated and displayed at each location. The same variables as in section 4 were examined. In the top 1% threshold 6 of the 58 stations displayed negative trends in the top 1% threshold, though these locations were not statistically significant (Fig. 8). These stations were located east of Lake Erie in New York and Pennsylvania, in the northern Adirondack region of New York, and in southern Vermont east of the Green Mountain Range. Eighteen (18) of the stations had statistically significant positive trends in the top 1% threshold. Many of these stations were located away from the coast in New York, Pennsylvania, and New Jersey, while several locations with statistically significant positive trends were located in Maine, northern Vermont, and New York. However, the large inland grouping of statistically significant increases in the extreme precipitation threshold suggests a further inland extent of coastal-track storm systems. The area of significant increases is also an area of complex terrain (northeast Appalachian Mountains), which could affect precipitation due to enhanced uplift in favored windward locations. An investigation into changes in weather patterns will have to be completed to confirm these inferences.

Fig. 8.

The statistical significance of changes in the top 1% at each station location. Blue dots represent negative trends, while red dots represent positive trends. Filled dots indicate a statistically significant trend, and hollow dots represent a trend that is not statistically significant.

Fig. 8.

The statistical significance of changes in the top 1% at each station location. Blue dots represent negative trends, while red dots represent positive trends. Filled dots indicate a statistically significant trend, and hollow dots represent a trend that is not statistically significant.

The spatial variability of changes in the frequency of events at or above the top 1% displayed a similar distribution to that of changes in the top 1% (Fig. 9). There was a grouping of statistically significant increases in frequency in eastern Pennsylvania, New Jersey, and central New York, as in Fig. 8. In total, 15 stations had statistically significant increases in frequency of extreme precipitation events, and five stations had decreases in the frequency of extreme events, one significant. Again, these decreases in frequency were focused east of Lake Erie. The near collocation of the trends in both magnitude and frequency of extreme events suggests that the increase in frequency of extreme events acted to increase the top 1% threshold.

Fig. 9.

The statistical significance of changes in the frequency of events at or above the top 1% at each station location. Blue dots represent negative trends, while red dots represent positive trends. Filled dots indicate a statistically significant trend, and hollow dots represent a trend that is not statistically significant.

Fig. 9.

The statistical significance of changes in the frequency of events at or above the top 1% at each station location. Blue dots represent negative trends, while red dots represent positive trends. Filled dots indicate a statistically significant trend, and hollow dots represent a trend that is not statistically significant.

The annual daily maximum also displayed a similar distribution to both the top 1% threshold and the frequency of extreme events (Fig. 10). Twenty-four (24) stations had statistically significant increases in the annual daily maximum, again concentrated in Pennsylvania and New York. In addition, seven stations had negative trends in the annual daily maximum (one significant) east of Lake Erie, and an additional station was located in Connecticut. The consistent pattern of decreasing trends east of Lake Erie and increasing trends in New York and Pennsylvania across these three variables indicates that the weather patterns that bring extreme events to these two regions are changing.

Fig. 10.

The statistical significance of changes in the annual daily maximum at each station location. Blue dots represent negative trends, while red dots represent positive trends. Filled dots indicate a statistically significant trend, and hollow dots represent a trend that is not statistically significant.

Fig. 10.

The statistical significance of changes in the annual daily maximum at each station location. Blue dots represent negative trends, while red dots represent positive trends. Filled dots indicate a statistically significant trend, and hollow dots represent a trend that is not statistically significant.

The spatial distribution of changes in the top 1% threshold and their statistical significance was also examined seasonally (Fig. 11). DJF had three stations with statistically significant increases in the top 1% threshold, 41 stations with insignificant increases, and 14 stations with insignificant decreases in the top 1% threshold (Fig. 11a). MAM had eight stations with statistically significant increases, concentrated in central New York and northern Vermont, 38 stations with insignificant increases, and 12 stations with insignificant decreases, most located in the eastern portion of the domain (Fig. 11b). JJA had 13 stations with statistically significant increases, spread throughout the eastern half of the domain, 34 stations with insignificant increases, 3 stations with significant decreases southeast of Lake Erie, and 8 stations with insignificant decreases, three of which were also east of Lake Erie (Fig. 11c). SON had the most stations (15) with statistically significant increases in the top 1% threshold (Fig. 11d). These stations were concentrated in the eastern portion of the domain slightly inland from the coast, largely in central New York and eastern Pennsylvania. Thirty-two (32) stations had insignificant increases. Eleven (11) stations decreased in the top 1% threshold (one significant). All but one of these stations was east of the Great Lakes and in the Northern Adirondack region.

Fig. 11.

The statistical significance of changes in the top 1% at each station location. Blue dots represent negative trends, while red dots represent positive trends. Filled dots indicate a statistically significant trend, and hollow dots represent a trend that is not statistically significant. (a) DJF, (b) MAM, (c) JJA, and (d) SON.

Fig. 11.

The statistical significance of changes in the top 1% at each station location. Blue dots represent negative trends, while red dots represent positive trends. Filled dots indicate a statistically significant trend, and hollow dots represent a trend that is not statistically significant. (a) DJF, (b) MAM, (c) JJA, and (d) SON.

The spatial distribution of changes in the frequency of extreme events was not examined seasonally due to the small sample size of extreme events at each station location. However, the spatial distribution of changes in the annual daily maximum of precipitation was examined (Fig. 12). Very few stations had significant trends during DJF and MAM (Figs. 12a,b). There were three statistically significant increasing trends in DJF and nine in MAM, though these were sporadic except for a grouping in northern New York and Vermont in MAM. In JJA, there were 4 stations with statistically significant decreases east of Lake Erie, and 12 stations with statistically significant increases in annual daily maximum precipitation throughout the domain (Fig. 12c). In SON, there was again a grouping of negative trends east of Lake Erie. In addition, 15 stations had statistically significant positive trends during SON (Fig. 12d), a large portion of which occurred in New York and Pennsylvania, similar to the trends in the top 1% (Fig. 11d) during SON and the annual frequency of extreme events (Fig. 9).

Fig. 12.

The statistical significance of changes in the annual daily maximum at each station location. Blue dots represent negative trends, while red dots represent positive trends. Filled dots indicate a statistically significant trend, and hollow dots represent a trend that is not statistically significant. For (a) DJF, (b) MAM, (c) JJA, and (d) SON.

Fig. 12.

The statistical significance of changes in the annual daily maximum at each station location. Blue dots represent negative trends, while red dots represent positive trends. Filled dots indicate a statistically significant trend, and hollow dots represent a trend that is not statistically significant. For (a) DJF, (b) MAM, (c) JJA, and (d) SON.

The consistent decreasing trend in extreme precipitation east of Lake Erie, both annually and during JJA and SON, suggests that local processes surrounding Lake Erie may be modifying extreme weather events. Under a warming climate, lake temperatures are warming (U.S. Environmental Protection Agency 2016). Warming lake temperatures may negatively impact warm season convection traversing Lake Erie, as warming surface temperatures decrease the stable near-surface inversion over the lake. This inversion is present in the warmer months due to warmer air being advected over the cooler lake surface. If this inversion is strong, convection is able to traverse the lake unaffected by the cool lake temperatures, as the inversion becomes the effective surface. A weaker near-surface inversion can allow mixing of stable air into the convective cells, decreasing their strength and production of precipitation once on land (Metz 2011). In addition, lake breezes may also impact the development and severity of convective cells. As lake temperatures warm lake breezes may weaken, decreasing onshore convergence and the strength of convective cells that develop along this boundary. However, it is unclear how these circulations may (or may not be) changing over time.

In addition, consistent increases were observed in the top 1% threshold, extreme event frequency and magnitude in Pennsylvania and New York, both annually and during the warm seasons of JJA and SON. This suggests a possible increase in inland intrusion of strong coastal systems, such as tropical storms, or an increase in topographic interactions, such as upslope flow. Unfortunately, thoroughly investigating the aforementioned hypotheses behind the observed trends are beyond the scope of this paper and future work is needed.

6. Discussion: Possible causes of observed changes in extreme precipitation

a. Weather types

Increases in extreme precipitation were most robust for events within the top 1%, most notably, events ≥ 150 mm (Fig. 5). To further examine the meteorological causes of these events and their respective changes, they were cataloged into the following weather types: extratropical cyclones, convection, events associated with tropical moisture, and hurricanes/tropical storms. Extratropical systems were dominated by a frontal pattern. Convective events may have been associated with a frontal pattern/extratropical cyclone, but the main mechanism driving extreme precipitation were convective cells. An event was considered to be associated with tropical moisture when a nearby tropical system (e.g., a tropical cyclone) was directly responsible for providing moisture to produce heavy precipitation, regardless of the synoptic characteristics of the event. An event was considered to be a hurricane/tropical storm if a hurricane or tropical storm directly contributed to extreme precipitation in the Northeast by either making landfall, or coming close to the Northeast. The category of each event was determined through credible archived reports and published journal articles summarizing the details of the event (see  appendix).

There were 24 individual weather events ≥ 150 mm occurring across 31 days during 1979–2014 (Table 2). Of the 24 events, five events were caused by extratropical cyclones, 3 events were related to convection, 5 events had moisture sources from tropical systems, 10 events were caused by landfalling hurricanes and tropical storms, and 1 event could not be identified (4 August 1979). Precipitation events with 150 mm or greater were most common in July–October, with a peak occurrence during September (Fig. 13a). Hurricanes/tropical storms, and tropical moisture feeds most commonly contributed to 150-mm precipitation events, particularly in August–October. Only 7 of the 24 events of 150 mm or greater occurred over multiple days (150-mm precipitation amounts recorded on consecutive days), and no 150-mm precipitation events lasted longer than 2 days.

Table 2.

Table of extreme precipitation events ≥ 150 mm between 1979 and 2014.

Table of extreme precipitation events ≥ 150 mm between 1979 and 2014.
Table of extreme precipitation events ≥ 150 mm between 1979 and 2014.
Fig. 13.

(a) Event type by month, (b) the number of 150-mm events in 1979–96 (blue) and 1997–2014 (green) by month, and (c) the number of 150-mm events in 1979–96 (blue) and 1997–2014 (green) by event type. The number of 150-mm events in each category is divided by season, where DJF is represented with vertical lines, MAM is represented with the square pattern, JJA is diagonally hatched, and SON is solid.

Fig. 13.

(a) Event type by month, (b) the number of 150-mm events in 1979–96 (blue) and 1997–2014 (green) by month, and (c) the number of 150-mm events in 1979–96 (blue) and 1997–2014 (green) by event type. The number of 150-mm events in each category is divided by season, where DJF is represented with vertical lines, MAM is represented with the square pattern, JJA is diagonally hatched, and SON is solid.

b. Observed changes in weather types

From 1979 to 1996, 150-mm events were only recorded in June, August, and October (four events total), with two events occurring during August (Fig. 13b). From 1997 to 2014, 150-mm precipitation events were also recorded in March, April, July, and September. The largest increase in 150-mm precipitation events from the first half of the time period to the second half occurred during SON. During SON, only one 150-mm precipitation event occurred in the first half of the time period (in October), and ten 150-mm precipitation events occurred in the second half (six in September and four in October). During JJA, three 150-mm precipitation events occurred in the first half of the time period (one in June, two in August). In the second half of the time period, eight 150-mm precipitation events were recorded (one in June, four in July, and three in August). During MAM, there were no 150-mm precipitation events in the first half of the time period, and two were recorded during the second half of the time period (one in March, one in April). DJF and May did not have any 150-mm precipitation events observed in either half of the time period.

Changes in 150-mm event types from the first half of the time period to the second were also compared. From Fig. 13a, all 150-mm precipitation events that occurred during SON were either from tropical moisture sources or were caused by landfalling hurricanes/tropical storms. In comparing changes in the frequency of 150-mm precipitation events by event type, there were substantial increases from the first to the second half of the time period in the number of landfalling hurricane/tropical storms and tropical moisture feeds causing 150-mm precipitation events (Fig. 13c). The increase in tropical events was most pronounced during SON, where all ten 150-mm precipitation events in the second half of the time period were caused by landfalling hurricanes/tropical storms (six 150-mm events) or by tropical moisture feeds (four 150-mm precipitation events). During JJA, there was an increase in extratropical cyclone events (an increase of two 150-mm events), convective events (an increase of three 150-mm events), and landfalling hurricane/tropical storm events (an increase of three 150-mm events). During MAM, there was an increase of two extratropical cyclone events between the two halves of the time period.

The increases of tropical events in SON align with previously discussed results. SON experienced the largest increase of extreme precipitation events in both frequency and magnitude (Fig. 7, Table 1). The increase in frequency and magnitude was also observed in the distributions of the top 1% of precipitation during SON (Figs. 3d, 7d). The increases in extreme precipitation were most robust at locations further inland, as seen in the spatial distributions (Figs. 11, 12). The increase in the frequency of tropical events impacting the Northeast during SON is apparent in Fig. 13c. Coupled with the observed increases in the top 1% threshold, extreme event frequency, and the annual daily maximum at inland locations, suggests that more tropical events are making landfall and traveling further inland during SON.

Though the magnitude of extreme precipitation events within the top 1% is increasing (Figs. 3, 7), it cannot be concluded that events of a tropical nature are also increasing in magnitude without further investigation. The increases in magnitude of extreme events impacting the Northeast may be a by-product of the increasing frequency of these incredibly strong tropical precipitation events that previously did not impact the Northeast to this degree. Determining trends in tropical cyclones is difficult as changes in the number of events, the magnitude of events, and the track of the events can all have an impact. Tropical events affecting the Northeast are also rare, and natural variability on a long-term time scale can obscure trends in these events as well. In addition, further analysis of the weather types associated with extreme events is needed, particularly for extreme events below 150 mm, as well as analysis of atmospheric components that influence extreme precipitation, such as precipitable water and upward motion/forcing.

7. Summary and conclusions

This paper examined recent changes in extreme precipitation in the northeast United States. Daily station data from 58 stations missing less than 5% of days for the years 1979–2014 from the U.S. Historical Climatology Network were used to analyze both total and extreme precipitation, with the latter defined as the top 1% of days with precipitation. A detailed analysis of temporal trends in extreme precipitation was presented from regional, seasonal, and spatial viewpoints. Results showed statistically significant (95% confidence level) increases in the top 1% threshold of precipitation, and extreme precipitation frequency and magnitude. Increasing trends in extreme precipitation were most robust during the fall months of September, October, and November, and particularly at locations further inland. Analysis showed that increases in events that were tropical in nature, or associated with tropical moisture, caused the observed increase in extreme precipitation during the fall months.

The top 1% threshold increased 11 mm between 1979 and 2014, based on the 5-yr running average, a statistically significant increase at the 95% confidence level. The increase in the top 1% threshold was shown to be caused by both an increase in the frequency and in the magnitude of extreme precipitation events. The frequency of extreme events increased 15 events yr−1 (statistically significant at the 95% confidence level), and the annual daily maximum increased 58.0 mm (also statistically significant at the 95% confidence level) from 1979 to 2014. The observed trends in extreme precipitation were most robust during the warm season, particularly during SON.

Extreme precipitation, and associated trends, displayed high spatial variability in the Northeast. In examining the spatial distribution of trends in extreme precipitation, decreases in the top 1% threshold, frequency, and annual daily maximum, were observed east of Lake Erie. However, statistically significant increases in these variables were observed in eastern Pennsylvania and New York. The spatial pattern in the observed trends were, again, most robust during the warm seasons of JJA and SON. Though further analysis needs to be completed, the decreases east of Lake Erie may be due to decreased convective activity as the lake inversion typically present over Lake Erie during the warmer months weakens with warming water temperatures. Furthermore, the statistically significant increase of extreme precipitation in inland Pennsylvania and New York suggested further encroachment inland of coastal storms.

As increases in extreme precipitation frequency were largest at higher precipitation thresholds above the top 1% precipitation threshold, a brief weather-typing analysis was completed for precipitation events 150-mm or greater to examine whether there was an attributable trend in weather types to the observed trends in extreme precipitation. Increases in frequency of 150-mm precipitation events were most pronounced during July–October, particularly in September. Events with tropical moisture sources and landfalling hurricanes/tropical storms were the main contributors to the observed increases in the frequency of these most extreme events. This is consistent with the significant increases at inland locations during the fall months.

Though this study provides an in-depth examination of changes in precipitation in the northeast United States, much work is still to be done in determining the causes of the observed increases in extreme precipitation. Though preliminary investigation shows an increase in landfalling tropical storms and tropical moisture events during the fall months, this is an area that needs further study. More detailed weather typing, inclusion of less extreme events, and an analysis of contributions to extreme precipitation, such as trends in atmospheric water content or upward vertical motion, are necessary.

Acknowledgments

This research was supported by the National Science Foundation Partnership for International Research and Education Program between the United States and Taiwan, OISE-1545917, awarded to the University at Albany, SUNY.

APPENDIX

Weather Event References

a. Extratropical events
b. Convective events
c. Tropical moisture events
d. Hurricane/tropical storm events

REFERENCES

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.
Avila
,
L. A.
, and
J.
Cangialosi
,
2011
: Tropical cyclone report: Hurricane Irene, 21–28 August 2011. Rep. AL092011, National Hurricane Center, 45 pp., https://www.nhc.noaa.gov/data/tcr/AL092011_Irene.pdf.
Brown
,
P. J.
,
R. S.
Bradley
, and
F. T.
Keimig
,
2010
:
Changes in extreme climate indices for the northeastern UNITED STATES, 1870–2005
.
J. Climate
,
23
,
6555
6572
, https://doi.org/10.1175/2010JCLI3363.1.
Daly
,
W. P.
,
G. H.
Taylor
,
M. K.
Doggett
, and
J. I.
Smith
,
2007
:
Observer bias in daily precipitation measurements at United States cooperative network stations
.
Bull. Amer. Meteor. Soc.
,
88
,
899
912
, https://doi.org/10.1175/BAMS-88-6-899.
Easterling
,
D. R.
,
J. L.
Evans
,
P. Y.
Groisman
,
T. R.
Karl
,
K. E.
Kunkel
, and
P.
Ambenje
,
2000
:
Observed variability and trends in extreme climate events: A brief review
.
Bull. Amer. Meteor. Soc.
,
81
,
417
425
, https://doi.org/10.1175/1520-0477(2000)081<0417:OVATIE>2.3.CO;2.
Frei
,
A.
,
K. E.
Kunkel
, and
A.
Matonse
,
2015
:
The seasonal nature of extreme hydrological events in the northeastern united states
.
J. Hydrometeor.
,
16
,
2065
2085
, https://doi.org/10.1175/JHM-D-14-0237.1.
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.
, and Coauthors
,
1999
:
Changes in the probability of heavy precipitation: Important indicators of climatic change
.
Climatic Change
,
42
,
243
283
, https://doi.org/10.1023/A:1005432803188.
Groisman
,
P. Ya.
,
R. W.
Knight
, and
T. R.
Karl
,
2001
:
Heavy precipitation and high streamflow in the contiguous United States: Trends in the twentieth century
.
Bull. Amer. Meteor. Soc.
,
82
,
219
246
, https://doi.org/10.1175/1520-0477(2001)082<0219:HPAHSI>2.3.CO;2.
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.
Hoerling
,
M.
,
J.
Eischeid
,
J.
Perlwitz
,
X.-W.
Quan
,
K.
Wolter
, and
L.
Cheng
,
2016
:
Characterizing recent trends in U.S. heavy precipitation
.
J. Climate
,
29
,
2313
2332
, https://doi.org/10.1175/JCLI-D-15-0441.1.
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.
Huntington
,
T. G.
,
G. A.
Hodgkins
,
B. D.
Keim
, and
R. W.
Dudley
,
2004
:
Changes in the proportion of precipitation occurring as snow in New England (1949–2000)
.
J. Climate
,
17
,
2626
2636
, https://doi.org/10.1175/1520-0442(2004)017<2626:CITPOP>2.0.CO;2.
IPCC
,
2001
: Climate Change 2001: The Scientific Basis. J. T. Houghton et al., Eds. Cambridge University Press, 881 pp.
Karl
,
T. R.
, and
R. W.
Knight
,
1998
:
Secular trends of precipitation amount, frequency, and intensity in the United States
.
Bull. Amer. Meteor. Soc.
,
79
,
231
241
, https://doi.org/10.1175/1520-0477(1998)079<0231:STOPAF>2.0.CO;2.
Kunkel
,
K. E.
,
2003
:
North American trends in extreme precipitation
.
Nat. Hazards
,
29
,
291
305
, https://doi.org/10.1023/A:1023694115864.
Kunkel
,
K. E.
,
K.
Andsager
, and
D. R.
Easterling
,
1999
:
Long-term trends in extreme precipitation events over the conterminous United States and Canada
.
J. Climate
,
12
,
2515
2527
, https://doi.org/10.1175/1520-0442(1999)012<2515:LTTIEP>2.0.CO;2.
Kunkel
,
K. E.
,
D. R.
Easterling
,
D. A.
Kristovich
,
B.
Gleason
,
L.
Stoecker
, and
R.
Smith
,
2010
:
Recent increases in U.S. heavy precipitation associated with tropical cyclones
.
Geophys. Res. Lett.
,
37
,
L24706
, https://doi.org/10.1029/2010GL045164.
Kunkel
,
K. E.
, and Coauthors
,
2013
:
Monitoring and understanding trends in extreme storms: State of knowledge
.
Bull. Amer. Meteor. Soc.
,
94
,
499
514
, https://doi.org/10.1175/BAMS-D-11-00262.1.
Livneh
,
B.
,
T. J.
Boho
,
D. W.
Pierce
,
F.
Munoz-Arriola
,
B.
Nijssen
,
R.
Voss
,
D. R.
Cayan
, and
L.
Brekke
,
2015
:
A spatially comprehensive, hydrometeorological data set for Mexico, the U.S., and southern Canada 1950–2013
.
Sci. Data
,
2
, 150042, https://doi.org/10.1038/sdata.2015.42.
Metz
,
N. D.
,
2011
: Persistence and dissipation of Lake Michigan-crossing mesoscale convective systems. Ph.D. thesis, University at Albany, State University of New York, 237 pp.
NCEI
,
2017
: Billion-dollar weather and climate disasters: Table of events. NOAA/NCEI, https://www.ncdc.noaa.gov/billions/.
Olson
,
S. A.
,
2007
: Flood of May 2006 in New Hampshire. USGS Open-File Rep. 2007-1122, 40 pp, https://pubs.usgs.gov/of/2007/1122/.
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.
Semenov
,
V.
, and
L.
Bengtsson
,
2002
:
Secular trends in daily precipitation characteristics: Greenhouse gas simulation with a coupled AOGCM
.
Climate Dyn.
,
19
,
123
140
, https://doi.org/10.1007/s00382-001-0218-4.
The Associated Press
,
2012
: Hurricane Irene one year later: Storm cost $15.8 in damage from Florida to New York to the Caribbean. New York Daily News, 27 August, http://www.nydailynews.com/new-york/hurricane-irene-year-storm-cost-15-8-damage-florida-new-york-caribbean-article-1.1145302.
U.S. Environmental Protection Agency
,
2016
: Climate change indicators: Great lakes water levels and temperatures. Environmental Protection Agency, https://www.epa.gov/climate-indicators/great-lakes#ref7.
Walsh
,
J.
, and Coauthors
,
2014
: Our changing climate. Climate change impacts in the United States: The Third National Climate Assessment, J. M. Melillo, T. C. Richmond, and G. W. Yohe, Eds., U.S. Global Change Research Program Rep., 19–67, https://doi.org/10.7930/J0KW5CXT.

Footnotes

© 2019 American Meteorological Society. For information regarding reuse of this content and general copyright information, consult the AMS Copyright Policy (www.ametsoc.org/PUBSReuseLicenses).