1. Introduction
Extreme rainfall, which often results in flash flooding, poses a risk to agriculture, life, and property. It tends to be poorly predicted by numerical models (e.g., Fritsch and Carbone 2004; Novak et al. 2011) and is also of interest in the context of the global climate system. As a result, there continues to be a need for the quantification and understanding of precipitation extremes.
Many recent studies have attempted to objectively classify extreme rainfall events in the United States, particularly east of the Rocky Mountains, in order to construct a climatology of such events. Brooks and Stensrud (2000) analyzed heavy precipitation events with ≥1 in. (25.4 mm) h−1 using the Hourly Precipitation Dataset from the National Climatic Data Center (NCDC). Konrad (2001) identified the most extreme precipitation events over a 47-yr period using rain gauge data from the cooperative observer network. Extreme rainfall events for the Konrad study were determined by ranking the highest mean 2-day precipitation over various spatial scales from 2500 to 500 000 km2 and over four regions in the eastern United States. Two issues with the sole use of rain gauges to determine extreme rainfall are the inconsistent number of gauges over time and the scarcity of gauges in some regions. Hitchens et al. (2012) analyzed extreme precipitation events in the midwestern United States using the National Centers for Environmental Prediction (NCEP) stage-II hourly precipitation dataset, which combines both rain gauge measurements and radar-derived precipitation. The Hitchens et al. study is one of the first to incorporate radar-derived precipitation into an extreme rainfall climatological analysis; however, the study used a fixed threshold to identify hourly extreme rain events and was limited to the midwestern states. Hitchens et al. (2013) reexamined the Brooks and Stensrud study, for the area spanning the eastern two-thirds of the United States, with the inclusion of radar rainfall estimates from two high-resolution precipitation datasets: a 10-yr period of NCEP’s stage-IV precipitation analysis and a 3-yr period of the Next Generation Multisensor Quantitative Precipitation Estimate (Q2). Again, the study was limited to hourly rain events.
The present study is motivated by previous work from Schumacher and Johnson (2006; hereafter SJ06). SJ06 examined the characteristics and temporal variability of extreme rainfall events during 1999–2003 using the National Weather Service/cooperative rain gauge network in the eastern two-thirds of the United States. Their analysis revealed most extreme rainfall events occurred during the summer month of July and more than half were associated with mesoscale convective systems (MCSs). Studies have also found that most extreme rainfall events are nocturnal in the central and eastern United States (Winkler et al. 1988; SJ06).
Although a large number of extreme rain events were analyzed, the study by SJ06 had several limitations, including the wide gaps between rain gauges in some parts of the country, the consideration of only one accumulation duration (24 h) in the selection of cases, and the use of a 5-yr period of study, which is relatively short for the examination of extreme events. This study aims to address some of these limitations and extend the SJ06 analysis by using 10 years of data from a gridded multisensor precipitation analysis at multiple temporal scales.
NCEP released their stage-IV precipitation analysis (Lin and Mitchell 2005) product starting in December 2001. Unlike previous precipitation analyses, the stage IV combines both radar-estimated rainfall and rain gauge data to provide better coverage over the United States. This product includes 6- and 24-hourly analyses that are consistently quality controlled manually by the 12 National Weather Service River Forecast Centers, along with 1-hourly analyses that undergo less consistent quality control.
Like SJ06, this study will identify extreme precipitation by identifying points where the accumulated rainfall exceeded a historical recurrence interval threshold. The two recurrence intervals (i.e., return periods) used in this study are the 50 and 100 year. A “50-yr rain event” is defined as the 2% chance that precipitation accumulation in a given time span will exceed the threshold in any given year at any given point. Similarly, the term “100-yr rain event” refers to the 1% chance the aforementioned will occur. Using even a relatively short data record, the amount of precipitation that corresponds to these probabilities can be determined by fitting the observed distribution to a mathematical function. Hershfield (1961) reported one of the first calculations of these recurrence intervals for the United States. The 50- and 100-yr recurrence intervals were chosen because these are the two highest thresholds constructed by Hershfield, thus they represent the most extreme rainfall thresholds available. Although these thresholds might have limitations due to the lack of data coverage in some areas at the time of their construction, they still represented the official values across much of the United States as this research was being conducted. Therefore, despite these potential limitations, we will use these values for the identification of extreme precipitation. Precipitation frequency estimates for the United States were in the process of being updated at the time of this study but were not yet complete.
This research intends to provide a detailed survey of extreme rainfall events in the eastern two-thirds of the United States over a 10-yr period during 2002–11. Previous studies have not examined this large of an area over multiple accumulation durations using combined rain gauge and radar-estimated precipitation. The analysis will include 50- and 100-yr rain events that occur on a 1-, 6-, and 24-h time interval. Extreme rainfall for the 37 states east of the Rocky Mountains is investigated both collectively and regionally (Fig. 1).
The regional divisions considered for this study, consistent with SJ06. The regions are as follows: 1) plains, 2) North, 3) Northeast, 4) Ohio–Mississippi Valley, 5) South, and 6) Southeast.
Citation: Monthly Weather Review 142, 9; 10.1175/MWR-D-13-00345.1
The regional divisions considered for this study, consistent with SJ06. The regions are as follows: 1) plains, 2) North, 3) Northeast, 4) Ohio–Mississippi Valley, 5) South, and 6) Southeast.
Citation: Monthly Weather Review 142, 9; 10.1175/MWR-D-13-00345.1
The regional divisions considered for this study, consistent with SJ06. The regions are as follows: 1) plains, 2) North, 3) Northeast, 4) Ohio–Mississippi Valley, 5) South, and 6) Southeast.
Citation: Monthly Weather Review 142, 9; 10.1175/MWR-D-13-00345.1
In section 2, the data and methods are described. The spatial and temporal distributions of extreme rainfall on the 50-yr recurrence interval are explored in section 3. Since the 50- and 100-yr results had many similarities, section 4 gives a brief overview of the results from the 100-yr recurrence intervals. The unique ranking of extreme rain events using this method are described in section 5. Section 6 analyzes the type of systems that caused the extreme rainfall and investigates the environmental differences between events with larger and smaller spatial extents of extreme rainfall. The results of this study are summarized in section 7.
2. Data and methods
a. Identifying extreme rainfall
The stage-IV precipitation analysis gridded data files for 1-, 6-, and 24-h time intervals for the years 2002–11 were obtained from the National Center for Atmospheric Research (NCAR) Earth Observing Laboratory (EOL). The gridded threshold data for the given time intervals on the 50- and 100-yr recurrence intervals were obtained from the Automated Geospatial Watershed Assessment Tool from the University of Arizona (see Fig. 2). The stage-IV analysis data have been coarsened slightly for this study, to a grid with approximately 8.25-km horizontal spacing, to correspond with the gridded threshold data. We have found that this helps to eliminate the identification of spurious events, although it may also remove legitimate events that occurred at very small spatial scales.
The threshold values (mm) for the (left) 50- and (right) 100-yr recurrence intervals on the (top) 1-, (middle) 6-, and (bottom) 24-h time intervals. The amount of precipitation needed to exceed the threshold is contoured.
Citation: Monthly Weather Review 142, 9; 10.1175/MWR-D-13-00345.1
The threshold values (mm) for the (left) 50- and (right) 100-yr recurrence intervals on the (top) 1-, (middle) 6-, and (bottom) 24-h time intervals. The amount of precipitation needed to exceed the threshold is contoured.
Citation: Monthly Weather Review 142, 9; 10.1175/MWR-D-13-00345.1
The threshold values (mm) for the (left) 50- and (right) 100-yr recurrence intervals on the (top) 1-, (middle) 6-, and (bottom) 24-h time intervals. The amount of precipitation needed to exceed the threshold is contoured.
Citation: Monthly Weather Review 142, 9; 10.1175/MWR-D-13-00345.1
An automated method compared the stage-IV precipitation amounts to the thresholds. Each “point” identified through this automated method represents a point on the 8.25-km grid where the rainfall exceeded the amount set by the specified threshold. The 24-h precipitation from the stage-IV analysis represents the 24-h period ending at 1200 UTC each day, and the 6-h precipitation represents the 6-h periods ending at 0000, 0600, 1200, and 1800 UTC each day. The precipitation amounts ending at these times are used in the analysis of 6- and 24-h extreme rainfall, which may result in an undercounting of rainfall events that straddle these analysis times (e.g., where heavy rain occurs between 0900 and 1500 UTC). However, given the often poor quality of the hourly analyses (discussed below), we did not attempt to identify events that straddled the 6- and 24-h analysis times by summing hourly analyses.
b. Quality control
Although the stage-IV analysis product from NCEP is regularly manually quality controlled, many clearly spurious extreme rainfall points were found, particularly in the 1-h data. Therefore, we augmented the NCEP quality control by comparing the Next Generation Weather Radar (NEXRAD) National Mosaic Reflectivity animations to the stage-IV precipitation analyses and eliminating the points not supported by the radar observations. Relatively few points were discarded in the 6- and 24-h analyses; however, over 80% of the 1-h points identified by the automated method were not supported in the NEXRAD reflectivity animations (Table 1). Several of the eliminated points were not in the vicinity of any rainfall. This suggests a need for greater quality control in the 1-h analyses. Many of the spurious points were small-scale “bull’s-eyes” with unrealistically large rainfall amounts, though the reasons that they were present in the analyses are unknown. Hitchens et al. (2013) similarly identified large numbers of spurious points exceeding 25.4 mm (h−1) in the hourly stage-IV precipitation analyses.
The percentage of points discarded for this study for the two different recurrence intervals (50 and 100 yr) and three different time intervals (1, 6, and 24 h).
c. Sorting the data for analysis
All the identified “points” that exceeded the recurrence interval threshold were subjectively combined into “events” based on their association with the same weather system. Most events were produced by either synoptic systems, tropical systems, or MCSs (discussed later); however, some extreme rainfall events were individual convective cells, especially on the 1-h time interval. NEXRAD National Mosaic Reflectivity animations were employed to discern the number of events at each time and recurrence interval. Multiple concurrent events were possible and counted as separate events at some of the analysis periods. Our analyses on both the total number of points and the total number of events will be discussed, as both classification methods provide valuable information about the character of extreme precipitation.Seasonal, monthly, and hourly patterns were also analyzed. The 1-h points were transformed from coordinated universal time (UTC) to local standard time (LST) since the local diurnal cycle has an impact on when convection occurs. These transformations into LST were calculated by subtracting five, six, and seven from the UTC time of an extreme rainfall point in the eastern, central, and mountain time zones, respectively. In addition to examining these patterns over the entire domain, regional analysis was also performed. Because of the difficulty of separating events (i.e., group of points) by region, since many events span multiple regions, no regional analysis on the event totals were performed. The six regions, replicated from the SJ06 study, include the plains, North, Northeast, Southeast, South, and Ohio–Mississippi Valley (Fig. 1). These regions group the states with the most similar climates together. The precipitation amounts necessary for a point to be classified as extreme for each recurrence and time interval are shown in Fig. 2.
To examine the similarities and differences between the type of events identified by our method and the one used by SJ06, we categorized the 100-yr, 24-h events into the three categories of weather systems used by SJ06: 1) synoptic, 2) tropical, and 3) MCS. We chose to categorize the 100-yr, 24-h events because these are the least frequent and most extreme events identified in this study. Although SJ06 used the 50-yr, 24-h threshold, the number of events they observed strictly from rain gauge observations is comparable to the number of events from our 100-yr, 24-h threshold that uses both rain gauges and radar-estimated precipitation amounts (discussed later). The subjective classification of these 100-yr, 24-h events was performed using both radar animations and surface analyses. The classification method mirrored that of SJ06. Synoptic events included those with strong large-scale ascent typically associated with synoptic-scale (i.e., ≥1000 km) features, like extratropical cyclones, and/or those that lasted longer than 24 h. Although MCSs and other mesoscale aspects can contribute to the extreme rainfall within a synoptic-scale system, we will classify those events meeting our synoptic definition as a synoptic event. MCS events included mesoscale (i.e., 100 km) features with convective echos (≥40 dBZ; Parker and Johnson 2000) on radar that persisted for 3–24 h (SJ06). For more details on radar patterns typically observed in MCSs [e.g., backbuilding, trailing stratiform, training line/adjoining stratiform (TL/AS)], refer to Schumacher and Johnson (2005). The last category, tropical, included tropical cyclones and the associated remnants as it moved inland or along the coast. Those events which did not fit one of these three categories were not classified.
d. Environmental conditions
To illustrate the usefulness of this objectively determined database of extreme rain events, the TL/AS MCS events exceeding the 100-yr, 24-h threshold in the Great Plains and midwestern United States were selected for further analysis in section 6b. We chose TL/AS systems because, as later shown, a majority of the MCS events with the most extreme rainfall points fell into this category. Schumacher and Johnson (2005) found that the highest frequency of extreme-rain-producing MCSs were TL/AS systems. We quantified widespread TL/AS events as those containing 10 or more points, with at least 10 contiguous points, and concentrated TL/AS events as those containing fewer than 10 points. Gridded North American Regional Reanalysis (NARR; Mesinger et al. 2006) datasets provided the meteorological data used to investigate the environments in both the widespread and concentrated extreme rainfall events. The NARR has a horizontal resolution of 32 km. Our storm-centered composites were created by defining a 31 × 31 point (approximately 1000 km × 1000 km) domain centered at the time and location of the point that experienced the highest hourly rainfall amount in each event, and averaging these fields over multiple events. This method for composite analysis is generally similar to that used by Schumacher and Johnson (2005) and Peters and Schumacher (2014).
3. 50-yr recurrence interval analysis
a. 1-h duration
In the 1-h time interval, the spatial distribution of points across the central and eastern United States was more densely populated in the western half of the plains, along the Ohio River valley, and along the eastern coastline. Relatively few extreme rainfall points occurred in between these regions (Fig. 3). We believe the high density of points along the western half of the plains were a result of poor sampling in the pre-1960s era when the thresholds were created; thus, the true frequency of extreme rainfall was likely underestimated for this region in the historical thresholds (cf. Figs. 8 and 9 of Hershfield 1961). Therefore, the 1-h results in these areas should be interpreted with some caution.
Spatial distribution of points for 50-yr, 1-h threshold during 2002–11. There are 2075 total points. The color of the crisscross (×) represents the month during which the point occurred. The total number of points in each month is displayed in the legend.
Citation: Monthly Weather Review 142, 9; 10.1175/MWR-D-13-00345.1
Spatial distribution of points for 50-yr, 1-h threshold during 2002–11. There are 2075 total points. The color of the crisscross (×) represents the month during which the point occurred. The total number of points in each month is displayed in the legend.
Citation: Monthly Weather Review 142, 9; 10.1175/MWR-D-13-00345.1
Spatial distribution of points for 50-yr, 1-h threshold during 2002–11. There are 2075 total points. The color of the crisscross (×) represents the month during which the point occurred. The total number of points in each month is displayed in the legend.
Citation: Monthly Weather Review 142, 9; 10.1175/MWR-D-13-00345.1
On average, around 93 events occurred each year that exceeded the 50-yr, 1-h threshold, with an average of two points per event. The month with the greatest number of points and events varied from year to year, but the months of June–August (JJA) consistently produced the maximum number of points and events and December–February (DJF) produced the minimum (Fig. 4). It is not surprising for the maximum number of events to occur in the warm season since most extreme rainfall is convective (Maddox et al. 1979). Combining all of the years, the maximum number of points and events for the 1-h time interval is in July (Fig. 5). The monthly distribution patterns shown match almost identically to the Brooks and Stensrud (2000) study that analyzed ≥1 in. h−1 events. Similar to Brooks and Stensrud’s 45-yr period and the results of Hitchens et al. (2013) with two different datasets, we observed that about 20% of events occurred in the month of July. Maddox et al. (1979) similarly found 25% of their flash flood events occurred in July. No clear annual trend was found in the 1-h analysis (not shown).
The 50-yr, 1-h number of points for each month of each year (color bars), the average number of points per month over all years (black line), and the average number of events per month over all years (dashed black line).
Citation: Monthly Weather Review 142, 9; 10.1175/MWR-D-13-00345.1
The 50-yr, 1-h number of points for each month of each year (color bars), the average number of points per month over all years (black line), and the average number of events per month over all years (dashed black line).
Citation: Monthly Weather Review 142, 9; 10.1175/MWR-D-13-00345.1
The 50-yr, 1-h number of points for each month of each year (color bars), the average number of points per month over all years (black line), and the average number of events per month over all years (dashed black line).
Citation: Monthly Weather Review 142, 9; 10.1175/MWR-D-13-00345.1
The 50-yr, 1-h monthly distributions of total points by (top) region and (bottom) total events for the years 2002–11.
Citation: Monthly Weather Review 142, 9; 10.1175/MWR-D-13-00345.1
The 50-yr, 1-h monthly distributions of total points by (top) region and (bottom) total events for the years 2002–11.
Citation: Monthly Weather Review 142, 9; 10.1175/MWR-D-13-00345.1
The 50-yr, 1-h monthly distributions of total points by (top) region and (bottom) total events for the years 2002–11.
Citation: Monthly Weather Review 142, 9; 10.1175/MWR-D-13-00345.1
The monthly point peaks in extreme rainfall varied regionally (Fig. 5). Five of the six regions had a peak in JJA, consistent with the results of the combined regions mentioned above. The plains and South both peaked the earliest in June. The North and Northeast both peaked a month later in July. Hershfield (1961) noted a similar July maximum for the Northeast on the 1-h duration. August held the most points in the Ohio–Mississippi Valley. The only region that did not have a summer maximum was the Southeast. Most points occurred in September, resulting from the vast span of coastline in this region and the increased tropical cyclone activity during this month; over 65% of the 1-h points in this month were a direct result of a tropical cyclone.
When analyzing the point data by the time of day it occurred, the late-evening to overnight hours during 1600–0000 LST was the most active period (Fig. 6); almost 70% of the extreme rainfall points occurred during these hours. The hours 2300, 1700, and 1100 LST emerged as the most frequent, second most frequent, and least frequent times, respectively. A nocturnal maximum is consistent with previous studies of rainfall over the United States (Wallace 1975; Maddox et al. 1979; Winkler et al. 1988; Dai et al. 1999; Konrad 1997; SJ06). However, for 25 mm (h−1) events, Hitchens et al. (2013) found a maximum in the late afternoon. This suggests there may be a tendency for extreme 1-h events (as defined here; see Fig. 2) to occur later than the 25-mm events they studied.
The 50-yr, 1-h distribution of time of day (LST) for all points during 2002–11.
Citation: Monthly Weather Review 142, 9; 10.1175/MWR-D-13-00345.1
The 50-yr, 1-h distribution of time of day (LST) for all points during 2002–11.
Citation: Monthly Weather Review 142, 9; 10.1175/MWR-D-13-00345.1
The 50-yr, 1-h distribution of time of day (LST) for all points during 2002–11.
Citation: Monthly Weather Review 142, 9; 10.1175/MWR-D-13-00345.1
The time of day when extreme rainfall most frequently occurred also varied regionally (Fig. 7). A pattern emerges in the peak times as one moves farther south and east across the regions: the plains at 1700 LST, the North and Northeast at 2100 LST, and the Ohio–Mississippi Valley and South at 2300 LST. Though these are the hours with the highest number of points, other times were also active with extreme rainfall. The 1700 LST peak in the plains was followed by secondary maxima during the nighttime hours of 2200 and 0100 LST. The North and Northeast had very similar late-afternoon to nocturnal active periods, though the North’s active period began slightly earlier in the afternoon at 1600 LST as opposed to 1800 LST in the Northeast. The South, in addition to the maximum at 2300 LST, had a very noticeable afternoon maximum at 1600 LST. The Southeast exhibited different time of day characteristics of extreme rainfall than the other regions. The most active period was 1300–2000 LST, with 2000 LST holding the maximum. This early afternoon through early nighttime maximum is due to the frequent scattered convection in this region (e.g., Wallace 1975). Wallace found a late-afternoon maximum in the Southeast, which is earlier than the 2000 LST maximum found in our analysis. Konrad (1997) studied heavy rainfall in the Southeast United States and found the lighter events tended to occur in the late afternoon, while the heavier events were more nocturnal.
The 50-yr, 1-h distribution of time of day (LST) by region for 2002–11, expressed as a percentage of the total points for the region.
Citation: Monthly Weather Review 142, 9; 10.1175/MWR-D-13-00345.1
The 50-yr, 1-h distribution of time of day (LST) by region for 2002–11, expressed as a percentage of the total points for the region.
Citation: Monthly Weather Review 142, 9; 10.1175/MWR-D-13-00345.1
The 50-yr, 1-h distribution of time of day (LST) by region for 2002–11, expressed as a percentage of the total points for the region.
Citation: Monthly Weather Review 142, 9; 10.1175/MWR-D-13-00345.1
b. 6-h duration
In the 6-h time interval, points were more uniformly distributed (Fig. 8), though a high density of points still existed in the western plains region similar to the 1-h data. We believe this is likely still a reflection of the poor sampling of the shorter time scale rainfall in this region when the thresholds were created. The number of points and events were highly variable from year to year; a period longer than 10 years is necessary to assess trends in the number of extreme rainfall points and events. On average, approximately 530 points and 87 events (an average of 6 points per event) occurred in any given year on the 6-h time interval. Combining all 10 years in the study, the maximum number of points on this temporal scale occurred in August while the maximum number of events occurred in July (Fig. 9). June and August followed closely behind in event totals, consistent with the summer maximum observed in the 1-h data. The time of day variation produced the same patterns as the 1-h data, with the late-afternoon to overnight hours being the most active and the mid- to late-morning hours being the least (not shown).
As in Fig. 3, but for the 50-yr, 6-h threshold. There are 5445 total points.
Citation: Monthly Weather Review 142, 9; 10.1175/MWR-D-13-00345.1
As in Fig. 3, but for the 50-yr, 6-h threshold. There are 5445 total points.
Citation: Monthly Weather Review 142, 9; 10.1175/MWR-D-13-00345.1
As in Fig. 3, but for the 50-yr, 6-h threshold. There are 5445 total points.
Citation: Monthly Weather Review 142, 9; 10.1175/MWR-D-13-00345.1
As in Fig. 5, but for the 50-yr, 6-h threshold.
Citation: Monthly Weather Review 142, 9; 10.1175/MWR-D-13-00345.1
As in Fig. 5, but for the 50-yr, 6-h threshold.
Citation: Monthly Weather Review 142, 9; 10.1175/MWR-D-13-00345.1
As in Fig. 5, but for the 50-yr, 6-h threshold.
Citation: Monthly Weather Review 142, 9; 10.1175/MWR-D-13-00345.1
Regionally, the monthly point maxima in the Ohio–Mississippi Valley, Southeast, Northeast, and plains matched that of the 1-h threshold. The South, which had a June point peak in the 1-h duration, exhibited a September point maximum in the 6-h duration. June had the second most points in the South on this time interval. In the North, monthly point peaks do not necessarily differ between the two time intervals, rather the 6 h had roughly the same number of points in JJA while the 1 h had a clearer peak in July.
The point database generated allows for the comparison of points and events between the different time intervals. Of the 6-h points, 9.3% corresponded to a 1-h point, and 31% of the 6-h events matched a 1-h event. This implies a large majority of the 6-h points (90%) and events (69%) are not represented on the 1-h extreme rainfall threshold. However, the low percentage of overlap between the 1- and 6-h points and events does not indicate that the relative hourly rainfall contributions to the 6-h points and events were insignificant, as our analysis did not investigate this aspect (though this is suggested for future work).
c. 24-h duration
For the 24-h time interval, the distribution of points across the area was fairly uniform, with the lowest concentration per square kilometer occurring in the South (Fig. 10). More uniform daily (i.e., 24 h) rain gauge records were available when Hershfield (1961) created the thresholds (see their Fig. 9), thus the 24-h recurrence intervals are likely closer to the true frequency than some of the less sampled 1- and 6-h recurrence intervals. The average number of points per year over all the years for the 24-h data was higher than the other two time intervals studied; however, the average number of events per year was less at 47. This indicates the average number of points per event, 16.5, is greater at the 24-h analysis than the other two time durations. The difference in the average number of points per event at the different temporal scales reveal that a 24-h point is typically associated with an organized precipitation system (thus making it likely that adjacent grid points are also receiving heavy precipitation), whereas a 1-h point can come from a few relatively isolated convective cells (i.e., only causing extreme rainfall on 1–2 grid points). While only 28% of the 24-h events were not captured on the 6-h threshold, over 65% were not captured on the 1-h time interval. As further evidence that 24-h events are more organized precipitation systems than 1-h events, only 4% of the 24-h points corresponded to a 1-h point and only 16% of the 1-h points corresponded to a 24-h point.
As in Figs. 3 and 8, but for the 50-yr, 24-h threshold. There are 7549 total points.
Citation: Monthly Weather Review 142, 9; 10.1175/MWR-D-13-00345.1
As in Figs. 3 and 8, but for the 50-yr, 24-h threshold. There are 7549 total points.
Citation: Monthly Weather Review 142, 9; 10.1175/MWR-D-13-00345.1
As in Figs. 3 and 8, but for the 50-yr, 24-h threshold. There are 7549 total points.
Citation: Monthly Weather Review 142, 9; 10.1175/MWR-D-13-00345.1
The months that experienced the maximum number of points for the 24-h time interval varied from year to year (Fig. 11) and differed from the other time intervals. Most years had a high number of points in September and October, largely due to precipitation from tropical systems; 70% of the 24-h September points and 58% of October points resulted from tropical cyclones. This expanded the range of the month that holds the maximum number of events to June through September for this time interval. Aggregated over the 10-yr period, the peak in points occurred in September, but the maximum for events shifted to June, with the following three months closely behind (Fig. 12). Again, this is consistent with the idea that individual tropical cyclones, which are categorized here as a single event, commonly produce rainfall amounts that exceed the threshold at a large number of grid points.
As in Fig. 4, but for the 50-yr, 24-h threshold.
Citation: Monthly Weather Review 142, 9; 10.1175/MWR-D-13-00345.1
As in Fig. 4, but for the 50-yr, 24-h threshold.
Citation: Monthly Weather Review 142, 9; 10.1175/MWR-D-13-00345.1
As in Fig. 4, but for the 50-yr, 24-h threshold.
Citation: Monthly Weather Review 142, 9; 10.1175/MWR-D-13-00345.1
As in Figs. 5 and 9, but for the 50-yr, 24-h threshold.
Citation: Monthly Weather Review 142, 9; 10.1175/MWR-D-13-00345.1
As in Figs. 5 and 9, but for the 50-yr, 24-h threshold.
Citation: Monthly Weather Review 142, 9; 10.1175/MWR-D-13-00345.1
As in Figs. 5 and 9, but for the 50-yr, 24-h threshold.
Citation: Monthly Weather Review 142, 9; 10.1175/MWR-D-13-00345.1
Monthly peaks in the 24-h duration also varied regionally. The North had a peak number of points in August (Fig. 12). The plains and Ohio–Mississippi Valley regions had the maximum number of points in May. Almost 90% of the May points in the Ohio–Mississippi Valley region occurred during the major flooding event in 2010, which will be discussed more in a later section. With this one event aside, this region’s maximum shifts to September. The remaining four regions used in the study, all with oceanic coastlines, experienced the most number of points over this 10-yr time span in September. Many of the point maxima in September could partially be correlated to peak tropical cyclone activity in the Atlantic basin. The mean fraction of total monthly precipitation due to tropical cyclones approaches one-fifth in some areas of the south in September, with the maximum fraction near one in some years (Larson et al. 2005). Shepherd et al. (2007) also found extreme rainfall days are most likely during the peak of hurricane season along the coastal regions of the Southeast United States. In the Northeast, October had a high number of points compared to the other regions. However, most of the October points were attributed to two events where the remnants of Tropical Storm Tammy (2005) and Tropical Storm Nicole (2010) interacted with a passing cold front. The Northeast experiences extreme rainfall of 100 mm or more due to tropical cyclones every 10–20 years (Hart and Evans 2001). A rain event with 115 mm or more would meet the 24-h threshold in parts of the Northeast (see Fig. 2). Hurricane Irene and the remnants of Tropical Storm Lee both affected the region in 2011 and were responsible for large numbers of points exceeding the threshold.
4. 100-yr recurrence interval analysis
The 100-yr spatial distributions for all time intervals follow closely to that of the 50 yr, although there were fewer points overall. No clear year-to-year trend could be determined from this short climatology. On average, there were about 52 one-hour events, 56 six-hour events, and 29 twenty-four-hour events that occurred in any given year. These numbers were lower than the 50-yr equivalents for the same time durations mentioned in the previous section. The seasonal variation of points and events followed the same pattern as the 50 yr, with the summer months JJA being the most active and the winter months DJF being the least active (Fig. 13).
The 100-yr seasonal variation for the 1-, 6-, and 24-h durations based on (top) total number of points and (bottom) number of events for 2002–11.
Citation: Monthly Weather Review 142, 9; 10.1175/MWR-D-13-00345.1
The 100-yr seasonal variation for the 1-, 6-, and 24-h durations based on (top) total number of points and (bottom) number of events for 2002–11.
Citation: Monthly Weather Review 142, 9; 10.1175/MWR-D-13-00345.1
The 100-yr seasonal variation for the 1-, 6-, and 24-h durations based on (top) total number of points and (bottom) number of events for 2002–11.
Citation: Monthly Weather Review 142, 9; 10.1175/MWR-D-13-00345.1
The months with the most events for the 1-, 6-, and 24-h time intervals were July, August, and June, respectively (not shown). Hitchens et al. (2012) found a similar July maximum in 1-h events when evaluating the Midwest region. The most points occurred in a similar warm season, summer months on the 100-yr recurrence interval except for the 24-h duration, which again showed a September maximum. Unlike the 50-yr regional analysis, the plains had a maximum number of points in the month of July and the North had its maximum in June (not shown). All of the remaining regions had similar monthly point peaks for both the 100- and 50-yr hourly data. The time of day results for the 100-yr period mirrored that of the 50-yr period.
5. Objective ranking of extreme rain events
A unique aspect of this analysis method is the ability to rank the events. An event with more points indicates a larger areal extent (i.e., widespread) of the extreme rainfall. Such events can heavily influence the statistics mentioned above. An example of this occurs in the 24-h monthly distributions. The combined 2002–11 data showed a substantial number of points in May. However, this is mostly due to the major Nashville flood in 2010, which has been analyzed by Moore et al. (2012), Durkee et al. (2012), Lackmann (2013), and Lynch and Schumacher (2014). This event ranked second for most points in a single event during the 10-yr time span studied (Table 2). A large number of points, 420 (256), surpassed the 50-yr (100 yr), 24-h threshold with this event. Hurricane Irene (2011) had the most points for a single event over the 10-yr time span. It recorded 522 (295) points exceeding the 50-yr (100 yr), 24-h threshold. The remaining events had about half as many points or less.
The top 10 events from the 100-yr, 24-h threshold, ranked by the highest number of points per event. Dates, regional locations of the extreme rainfall events, system types, and the minimum and maximum precipitation totals for the extreme rainfall points in the event are also shown. For tropical events, the name of the tropical cyclone is noted in parentheses.
Of the top 10 events with the most points during the period studied, 5 were the result of a tropical cyclone. This suggests that most of the largest areal impact events are caused by tropical cyclones in the central and eastern United States. The remaining five events included three that were classified as synoptic systems, and two as MCSs. Both of the MCS events fit into the subcategory of TL/AS systems.
6. Weather system types and environmental conditions
This objectively defined database of extreme rainfall can provide information about the type of weather systems that contribute most frequently to extreme rain events, and when they are most likely to occur. Our database is also a great resource that can be utilized to identify different types of events and examine the meteorological conditions associated with them. Although it is beyond the scope of this study to compare the environments of events at all the spatial and temporal thresholds (though this is suggested for future work), an example of this type of analysis is presented here in section 6b. Section 6a will describe the monthly distributions and other characteristics of each of the three types of weather systems. Since the spatial distributions of the 50- and 100-yr events at the 24-h time scale were relatively similar (except for the total number of events), we will examine the environmental conditions characteristic of 100-yr, 24-h events. These represent the most extreme events examined over this 10-yr period.
a. Overall characteristics
Each of the 100-yr, 24-h events were classified into one of three categories based on definitions described in section 2c: synoptic systems, MCS, or tropical systems. The monthly distribution of each type of system is shown in Fig. 14. Of the events that fell into these classifications, approximately 7% were tropical, 30% were synoptic, and 63% were MCSs. These are in close agreement with the percentages found by SJ06. The monthly distribution pattern of these types of systems is also similar to SJ06, with a peak in MCSs in the summer, synoptic systems evenly distributed throughout the year, and tropical events occurring most commonly in the fall. Tropical events experienced the most points per event in comparison to the other two types. As previously mentioned, Hurricane Irene accounted for the most points in a single event during the period analyzed. Most of the tropical events had 20 or more points. Synoptic and MCS events tended to have fewer points per event, with most having five or less points. However, this does not indicate that synoptic and MCS events are always the smaller and less widespread events. The widespread flooding near Tennessee and Kentucky in 2010, which ranked second in points per event for this study, provide an example of a case with both mesoscale- and synoptic-scale forcing mechanisms (Lynch and Schumacher 2014; Moore et al. 2012; note: because this event persisted for more than 24 h, it was classified as a synoptic event in our study).
The 100-yr, 24-h monthly distribution of system type for 2002–11 events.
Citation: Monthly Weather Review 142, 9; 10.1175/MWR-D-13-00345.1
The 100-yr, 24-h monthly distribution of system type for 2002–11 events.
Citation: Monthly Weather Review 142, 9; 10.1175/MWR-D-13-00345.1
The 100-yr, 24-h monthly distribution of system type for 2002–11 events.
Citation: Monthly Weather Review 142, 9; 10.1175/MWR-D-13-00345.1
b. Training line/adjoining stratiform (TL/AS) events
We chose to further analyze the environmental differences between larger, widespread events and smaller, concentrated events for the TL/AS MCSs. TL/AS events have a convective line and adjoining stratiform region and move parallel to the convective line, thus allowing prolonged heavy convective rainfall at locations along the convective line (Schumacher and Johnson 2005). This particular kind of MCS accounted for many of the MCS events that experienced 10 or more points in an event. A total of 7 of the top 10 MCS events were characterized by TL/AS radar patterns. As previously mentioned, two of the top 10 events during this 10-yr period of study were TL/AS MCS events in the North region: 19 August 2007, a predecessor rain event (PRE) associated with tropical cyclone Erin (Schumacher et al. 2011; Galarneau et al. 2010), and 15 September 2004 (see Table 2). There were 23 widespread events and 26 concentrated events (Fig. 15) incorporated into each respective composite analysis.
Location of the points within widespread (blue) and concentrated (red) TL/AS mesoscale events used in the composite analyses.
Citation: Monthly Weather Review 142, 9; 10.1175/MWR-D-13-00345.1
Location of the points within widespread (blue) and concentrated (red) TL/AS mesoscale events used in the composite analyses.
Citation: Monthly Weather Review 142, 9; 10.1175/MWR-D-13-00345.1
Location of the points within widespread (blue) and concentrated (red) TL/AS mesoscale events used in the composite analyses.
Citation: Monthly Weather Review 142, 9; 10.1175/MWR-D-13-00345.1
Figure 16 shows the composite analysis of the environment at 900 hPa for the widespread (top) and concentrated (bottom) events. Since these figures are composites of many events, an arbitrary map centered over Iowa, located in the middle of all the TL/AS events, was chosen. One prominent difference is the higher-θe air and stronger winds in the widespread events upwind of the most extreme rainfall point. The winds are predominately from the southwest at this level in both figures, reflecting the low-level jet that is common to MCS environments in the central United States (e.g., Laing and Fritsch 2000; Schumacher and Johnson 2005). As height increases, the wind direction rotates clockwise, becoming more westerly at 500 hPa (not shown). The θe distributions are fairly similar, but there is a larger area of high θe and a tighter θe gradient in the widespread composite. The tighter θe gradient and stronger winds suggest stronger moisture advection, stronger warm-air advection, and destabilization in the widespread events. Junker et al. (1999) found that the scale and intensity of rainfall was related to the magnitude of warm advection, among other factors. The stronger wind gradient in the widespread events is also associated with stronger and more widespread low-level convergence. This is reflected in the magnitude and expansiveness of frontogenesis (gray contours) shown in Fig. 16. The difference in frontogenesis between the widespread and concentrated cases (Fig. 17) shows a region of higher low-level frontogenesis in the widespread cases surrounding the point with the most rainfall. Coniglio et al. (2010) found a similar elongated region of frontogenesis in the inflow of long-lived MCSs when compared to short-lived MCSs.
Composite of the 900-hPa environment that includes (top) 23 widespread TL/AS events and (bottom) 26 concentrated TL/AS events. The black box in the center represents the location of the most extreme rainfall point. Displayed are equivalent potential temperature (K; color fill), wind speed (m s−1; contours), and wind barbs (kt, 1 kt = 0.5144 m s−1). Gray contours show areas of frontogenesis; these contours are every 0.25 K (100 km)−1 (3 h)−1 starting at 0.25 K (100 km)−1 (3 h)−1.
Citation: Monthly Weather Review 142, 9; 10.1175/MWR-D-13-00345.1
Composite of the 900-hPa environment that includes (top) 23 widespread TL/AS events and (bottom) 26 concentrated TL/AS events. The black box in the center represents the location of the most extreme rainfall point. Displayed are equivalent potential temperature (K; color fill), wind speed (m s−1; contours), and wind barbs (kt, 1 kt = 0.5144 m s−1). Gray contours show areas of frontogenesis; these contours are every 0.25 K (100 km)−1 (3 h)−1 starting at 0.25 K (100 km)−1 (3 h)−1.
Citation: Monthly Weather Review 142, 9; 10.1175/MWR-D-13-00345.1
Composite of the 900-hPa environment that includes (top) 23 widespread TL/AS events and (bottom) 26 concentrated TL/AS events. The black box in the center represents the location of the most extreme rainfall point. Displayed are equivalent potential temperature (K; color fill), wind speed (m s−1; contours), and wind barbs (kt, 1 kt = 0.5144 m s−1). Gray contours show areas of frontogenesis; these contours are every 0.25 K (100 km)−1 (3 h)−1 starting at 0.25 K (100 km)−1 (3 h)−1.
Citation: Monthly Weather Review 142, 9; 10.1175/MWR-D-13-00345.1
The low-level (900 hPa) frontogenesis difference field (widespread − concentrated) expressed as K (100 km)−1 (3 h)−1. Positive (red) represents larger frontogenesis in the widespread events.
Citation: Monthly Weather Review 142, 9; 10.1175/MWR-D-13-00345.1
The low-level (900 hPa) frontogenesis difference field (widespread − concentrated) expressed as K (100 km)−1 (3 h)−1. Positive (red) represents larger frontogenesis in the widespread events.
Citation: Monthly Weather Review 142, 9; 10.1175/MWR-D-13-00345.1
The low-level (900 hPa) frontogenesis difference field (widespread − concentrated) expressed as K (100 km)−1 (3 h)−1. Positive (red) represents larger frontogenesis in the widespread events.
Citation: Monthly Weather Review 142, 9; 10.1175/MWR-D-13-00345.1
Composites of total precipitable water (PW) show a maximum in PW near the location of the extreme rainfall with the composite maximum values between 45 and 50 mm for both the widespread and concentrated events (Fig. 18). The mean and median PW at the point of the heaviest rainfall were very similar between the two subsets of events (Fig. 19), but there is a slightly broader region of PW greater than 45 mm in the widespread composite (Fig. 18). For the widespread events, the mean PW at the point with the most extreme rainfall was 49.3 mm and the median PW was 48.9 mm. For the concentrated events, the mean PW was 47.0 mm and the median PW was 47.5 mm. There were seven concentrated events, but only one widespread event, that had PW less than 40 mm at the point with the most extreme rainfall. Overall, the widespread events tended to have a slightly higher PW than the concentrated events, though this result is not statistically significant. When observing the difference field between widespread and concentrated events, a corridor of higher moisture content and stronger winds is evident in the inflow region to the most extreme rainfall point (Fig. 20). It is suggested that this area of larger PW in the inflow region plays a greater role in the areal extent of the extreme rainfall than the PW in the column directly above the most extreme rainfall point.
Composites for total precipitable water for (top) widespread and (bottom) concentrated TL/AS events. Total precipitable water (mm) is the color fill and contours.
Citation: Monthly Weather Review 142, 9; 10.1175/MWR-D-13-00345.1
Composites for total precipitable water for (top) widespread and (bottom) concentrated TL/AS events. Total precipitable water (mm) is the color fill and contours.
Citation: Monthly Weather Review 142, 9; 10.1175/MWR-D-13-00345.1
Composites for total precipitable water for (top) widespread and (bottom) concentrated TL/AS events. Total precipitable water (mm) is the color fill and contours.
Citation: Monthly Weather Review 142, 9; 10.1175/MWR-D-13-00345.1
The blue (red) dots represent the total precipitable water (mm) in the column located at the point where the maximum precipitation in each of the TL/AS widespread (concentrated) events. The box and whisker plot for each shows the 75th percentile, median, and 25th percentile in thick black lines. The whiskers represent the 90th and 10th percentile. The white dashed line represents the mean.
Citation: Monthly Weather Review 142, 9; 10.1175/MWR-D-13-00345.1
The blue (red) dots represent the total precipitable water (mm) in the column located at the point where the maximum precipitation in each of the TL/AS widespread (concentrated) events. The box and whisker plot for each shows the 75th percentile, median, and 25th percentile in thick black lines. The whiskers represent the 90th and 10th percentile. The white dashed line represents the mean.
Citation: Monthly Weather Review 142, 9; 10.1175/MWR-D-13-00345.1
The blue (red) dots represent the total precipitable water (mm) in the column located at the point where the maximum precipitation in each of the TL/AS widespread (concentrated) events. The box and whisker plot for each shows the 75th percentile, median, and 25th percentile in thick black lines. The whiskers represent the 90th and 10th percentile. The white dashed line represents the mean.
Citation: Monthly Weather Review 142, 9; 10.1175/MWR-D-13-00345.1
The difference fields of moisture (PW) and wind between widespread and concentrated (widespread − concentrated) at 900 hPa. The color fill shows the differences in PW (mm). The thick black contours are the difference in winds (m s−1).
Citation: Monthly Weather Review 142, 9; 10.1175/MWR-D-13-00345.1
The difference fields of moisture (PW) and wind between widespread and concentrated (widespread − concentrated) at 900 hPa. The color fill shows the differences in PW (mm). The thick black contours are the difference in winds (m s−1).
Citation: Monthly Weather Review 142, 9; 10.1175/MWR-D-13-00345.1
The difference fields of moisture (PW) and wind between widespread and concentrated (widespread − concentrated) at 900 hPa. The color fill shows the differences in PW (mm). The thick black contours are the difference in winds (m s−1).
Citation: Monthly Weather Review 142, 9; 10.1175/MWR-D-13-00345.1
In both the widespread and concentrated composites, the necessary ingredients of moisture, instability, and lift were in place to support heavy-rain-producing MCSs. However, each of these ingredients was either more intense, more expansive, or both, for the widespread events.
7. Summary
This study provided a 10-yr survey of extreme rainfall events in the central and eastern United States on three different time durations: 1, 6, and 24 h, using two established recurrence intervals: 50 and 100 yr. The study used NCEP stage-IV precipitation analyses that combined both radar-estimates and rain gauge precipitation totals. Extreme rainfall “points” were where the observed precipitation from the stage-IV analysis exceeded either of the recurrence interval thresholds for any of the three time durations. Points that occurred with the same weather system were classified into “events.”
The spatial and seasonal patterns of extreme rainfall points and events between the 50- and 100-yr recurrence intervals were very similar. For the 1-h duration, there was a high density of points along the East Coast, in the western plains, and along the Ohio River valley, though these results may reflect limitations in the historical recurrence interval thresholds. Approximately 93 (52) events occurred in any given year for the 50-yr (100 yr) recurrence interval at this time duration. The late-evening to overnight hours (1600–0000 LST) were the most active for extreme rainfall. The 50-yr (100 yr), 6-h recurrence interval averaged 87 (55) events in any given year with a point maximum in August over the 10-yr period. The location of the 24-h recurrence interval points were the most uniformly distributed across the central and eastern United States. This time duration averaged about 46 (29) events per year for the 50-yr (100 yr) recurrence interval; however, the 24-h events had more points per event. This resulted in the overall point maximum falling in September in connection with the annual maximum in tropical cyclone activity. Conversely, the event maximum for the 24-h duration occurred in June, with July and August following closely behind. Over all three time durations, the most events occurred in the summer months of JJA, and the least occurred in the winter months of DJF.
Most of the extreme rainfall events at the 100-yr, 24-h duration were caused by mesoscale convective systems (MCSs), which peaked in JJA, consistent with previous studies. Approximately 30% were caused by synoptic systems with a fairly uniform monthly distribution, and 7% were tropical with a peak in September.
This extreme rainfall identification method allows for the unique analysis of events, including the ability to objectively rank the extremity of events (i.e., the most points occurring within a single event). During 2002–11, the top two events were Hurricane Irene (2011) and the Nashville flood event (2010). Four additional top 10 events were the result of tropical cyclones. Two of the top 10 events over this time period were TL/AS MCS events. Further analysis showed the larger, widespread TL/AS events had stronger low-level winds, stronger warm air advection, and stronger and more expansive frontogenesis in the inflow than the smaller, concentrated events. Total precipitable water levels were indistinguishable between widespread and concentrated events in the column above the most extreme rainfall point; however, the composite of the widespread events had an area of larger total precipitable water in the inflow layer when compared to the concentrated events.
Finally, since the stage-IV analysis is produced in near–real time, it is also possible to use the same methods described in this manuscript to automatically identify extreme rainfall events shortly after they occur. (The near-real-time results, as well as an archive of all events identified in the past, are displayed online at http://schumacher.atmos.colostate.edu/precip_monitor.) This database provides many avenues for future exploration. We suggest extending our analysis beyond the 10-yr period analyzed in this paper and including the use of the new threshold maps that were in development at the time this study was performed. Future research will also include expanding the analysis to the western United States and using the database of events for further meteorological understanding.
Acknowledgments
We thank the three anonymous reviewers for their comments that helped improve the paper. The stage-IV precipitation analyses were obtained from the National Center for Atmospheric Research(NCAR) Earth Observing Laboratory (EOL). The North American Regional Reanalysis (NARR) was obtained from the National Climatic Data Center (NCDC). Gridded threshold data for the different recurrence intervals and durations were obtained from The Automated Geospatial Watershed Assessment Tool from the University of Arizona. This research was supported by National Science Foundation Grants AGS-0954908 and AGS-1157425.
REFERENCES
Brooks, H. E., and D. J. Stensrud, 2000: Climatology of heavy rain events in the United States from hourly precipitation observations. Mon. Wea. Rev., 128, 1194–1201, doi:10.1175/1520-0493(2000)128<1194:COHREI>2.0.CO;2.
Coniglio, M. C., J. Y. Hwang, and D. J. Stensrud, 2010: Environmental factors in the upscale growth and longevity of MCSs derived from Rapid Update Cycle analyses. Mon. Wea. Rev., 138, 3514–3539, doi:10.1175/2010MWR3233.1.
Dai, A., F. Giorgi, and K. E. Trenberth, 1999: Observed and model-simulated diurnal cycles of precipitation over the contiguous United States. J. Geophys. Res., 104, 6377–6402, doi:10.1029/98JD02720.
Durkee, J. D., L. Campbell, K. Berry, D. Jordan, G. Goodrich, R. Mahmood, and S. Foster, 2012: A synoptic perspective of the record 1–2 May 2010 Mid-South heavy precipitation event. Bull. Amer. Meteor. Soc., 93, 611–620, doi:10.1175/BAMS-D-11-00076.1.
Fritsch, J. M., and R. E. Carbone, 2004: Research and development of improved quantitative precipitation forecasts in the warm season. Bull. Amer. Meteor. Soc., 85, 955–965, doi:10.1175/BAMS-85-7-955.
Galarneau, T. J., Jr., L. F. Bosart, and R. S. Schumacher, 2010: Predecessor rain events ahead of tropical cyclones. Mon. Wea. Rev., 138, 3272–3297, doi:10.1175/2010MWR3243.1.
Hart, R. E., and J. L. Evans, 2001: A climatology of the extratropical transition of Atlantic tropical cyclones. J. Climate, 14, 546–564, doi:10.1175/1520-0442(2001)014<0546:ACOTET>2.0.CO;2.
Hershfield, D. M., 1961: Rainfall frequency atlas of the United States for durations from 30 minutes to 24 hours and return periods from 10 to 100 years. U.S. Weather Bureau Tech. Paper 40, 61 pp.
Hitchens, N. M., M. E. Baldwin, and R. J. Trapp, 2012: An object-oriented characterization of extreme precipitation-producing convective systems in the midwestern United States. Mon. Wea. Rev., 140, 1356–1366, doi:10.1175/MWR-D-11-00153.1.
Hitchens, N. M., H. E. Brooks, and R. S. Schumacher, 2013: Spatial and temporal characteristics of heavy hourly rainfall in the United States. Mon. Wea. Rev., 141, 4564–4575, doi:10.1175/MWR-D-12-00297.1.
Junker, N. W., R. S. Schneider, and S. L. Fauver, 1999: A study of heavy rainfall events during the Great Midwest Flood of 1993. Wea. Forecasting, 14, 701–712, doi:10.1175/1520-0434(1999)014<0701:ASOHRE>2.0.CO;2.
Konrad, C. E., 1997: Synoptic-scale features associated with warm season heavy rainfall over the interior southeastern United States. Wea. Forecasting, 12, 557–571, doi:10.1175/1520-0434(1997)012<0557:SSFAWW>2.0.CO;2.
Konrad, C. E., 2001: The most extreme precipitation events over the eastern United States from 1950 to 1996: Consideration of scales. J. Hydrometeor., 2, 309–325, doi:10.1175/1525-7541(2001)002<0309:TMEPEO>2.0.CO;2.
Lackmann, G. M., 2013: The south-central U.S. flood of May 2010: Present and future. J. Climate, 26, 4688–4709, doi:10.1175/JCLI-D-12-00392.1.
Laing, A. G., and J. M. Fritsch, 2000: The large-scale environments of the global populations of mesoscale convective complexes. Mon. Wea. Rev., 128, 2756–2776, doi:10.1175/1520-0493(2000)128<2756:TLSEOT>2.0.CO;2.
Larson, J., Y. Zhou, and R. W. Higgins, 2005: Characteristics of landfalling tropical cyclones in the United States and Mexico: Climatology and interannual variability. J. Climate, 18, 1247–1262, doi:10.1175/JCLI3317.1.
Lin, Y., and K. E. Mitchell, 2005: The NCEP Stage II/IV hourly precipitation analyses: Development and applications. 19th Conf. on Hydrology, San Diego, CA, Amer. Meteor. Soc., 1.2. [Available online at https://ams.confex.com/ams/Annual2005/techprogram/paper_83847.htm.]
Lynch, S. L., and R. S. Schumacher, 2014: Ensemble-based analysis of the May 2010 extreme rainfall in Tennessee and Kentucky. Mon. Wea. Rev., 142, 222–239, doi:10.1175/MWR-D-13-00020.1.
Maddox, R. A., C. F. Chappell, and L. R. Hoxit, 1979: Synoptic and meso-α-scale aspects of flash flood events. Bull. Amer. Meteor. Soc., 60, 115–123, doi:10.1175/1520-0477-60.2.115.
Mesinger, F., and Coauthors, 2006: North American Regional Reanalysis. Bull. Amer. Meteor. Soc., 87, 343–360, doi:10.1175/BAMS-87-3-343.
Moore, B. J., P. J. Neiman, F. M. Ralph, and F. E. Barthold, 2012: Physical processes associated with heavy flooding rainfall in Nashville, Tennessee, and vicinity during 1–2 May 2010: The role of an atmospheric river and mesoscale convective systems. Mon. Wea. Rev., 140, 358–378, doi:10.1175/MWR-D-11-00126.1.
Novak, D. R., F. E. Barthold, M. J. Bodner, K. F. Brill, and M. Eckert, 2011: Quantifying extreme rainfall threats at the Hydrometeorological Prediction Center. 24th Conf. on Weather and Forecasting/20th Conf. on NWP, Seattle, WA, Amer. Meteor. Soc., 14A.4. [Available online at https://ams.confex.com/ams/91Annual/webprogram/Paper179714.html.]
Parker, M. D., and R. H. Johnson, 2000: Organizational modes of midlatitude mesoscale convective systems. Mon. Wea. Rev., 128, 3413–3436, doi:10.1175/1520-0493(2001)129<3413:OMOMMC>2.0.CO;2.
Peters, J. M., and R. S. Schumacher, 2014: Objective categorization of heavy-rain-producing MCS synoptic types by rotated principal component analysis. Mon. Wea. Rev., 142, 1716–1737, doi:10.1175/MWR-D-13-00295.1.
Schumacher, R. S., and R. H. Johnson, 2005: Organization and environmental properties of extreme-rain-producing mesoscale convective systems. Mon. Wea. Rev., 133, 961–976, doi:10.1175/MWR2899.1.
Schumacher, R. S., and R. H. Johnson, 2006: Charateristics of U.S. extreme rain events during 1999–2003. Wea. Forecasting, 21, 69–85, doi:10.1175/WAF900.1.
Schumacher, R. S., T. J. Galarneau Jr., and L. F. Bosart, 2011: Distant effects of a recurving tropical cyclones on rainfall in a midlatitude convective system: A high-impact predecessor rain event. Mon. Wea. Rev., 139, 650–667, doi:10.1175/2010MWR3453.1.
Shepherd, J. M., A. Grundstein, and T. L. Mote, 2007: Quantifying the contribution of tropical cyclones to extreme rainfall along the coastal southeastern United States. Geophys. Res. Lett.,34, L23810, doi:10.1029/2007GL031694.
Wallace, J. M., 1975: Diurnal variations in precipitation and thunderstorm frequency over the conterminous United States. Mon. Wea. Rev., 103, 406–419, doi:10.1175/1520-0493(1975)103<0406:DVIPAT>2.0.CO;2.
Winkler, J. A., B. R. Skeeter, and P. D. Yamamoto, 1988: Seasonal variations in the diurnal characteristics of heavy hourly precipitation across the United States. Mon. Wea. Rev., 116, 1641–1658, doi:10.1175/1520-0493(1988)116<1641:SVITDC>2.0.CO;2.
