The Effects of Extreme Precipitation Events on Climatology. Pamela Eck and Nicholas Metz. Department of Geoscience, Hobart and William Smith Colleges

Transcription

1 Final Report - August 3, HWS Research Program The Effects of Extreme Precipitation Events on Climatology Pamela Eck and Nicholas Metz Abstract Department of Geoscience, Hobart and William Smith Colleges Average monthly precipitation totals calculated by the National Climatic Data Center do not take into account the distribution of precipitation throughout the entire month. Using a 30-year climatology of daily precipitation totals from in 10 different cities across the United States, this study looks at whether normal monthly cumulative precipitation is made up of several days of little precipitation, or a few days of extreme precipitation. Cities were selected based on their climate, geographic location, the topography of the region, and their proximity to bodies of water. Only the months of April through September were studied due to the increased probability for convection during this time period. For each month, precipitation totals from the middle 20 years were considered to be normal. The percentage of the cumulative monthly precipitation that fell during the month s single largest precipitation event was calculated by dividing the cumulative monthly rainfall amount by the value for the month s largest precipitation event. Months with larger cumulative rainfall totals, more precipitation events per month, and largest monthly precipitation events of comparable magnitude, resulted in lower percentages for the most extreme monthly event. Months with smaller cumulative rainfall totals, fewer precipitation events per month, and largest monthly precipitation events of comparable magnitude, resulted in higher percentages for the most extreme monthly event. This variety in the distribution of precipitation is most likely caused, at least in 1

2 part, by convective episodes such as mesoscale convective systems (MCS) that form in the lee of the Rocky Mountains and are then carried eastward by the upper-level jet stream. 1. Introduction The National Climatic Data Center (NCDC) records and archives the precipitation amount that accumulates at over 2,000 stations across the United States and over 15,000 stations worldwide. These daily precipitation amounts can then be averaged together to determine the average amount of precipitation that falls during each month for any city with a record. However, this simple average does not take into account the distribution of the precipitation throughout the entire month. For example, during May 2007 in Albany, NY, precipitation fell on 8 days, resulting in a cumulative rainfall total of 3.51 inches, just shy of the monthly mean of 3.67 inches (Fig. 1). Three years later during September 2010, Albany only received 0.76 inches of rainfall during the first 29 days of the month. Then, on September 30, 2010, Albany received 2.68 inches of rain resulting from a quasistationary front near the Atlantic coast. Despite this extremely disproportional distribution of precipitation, Albany ended up with a 0.13-inch precipitation surplus for September 2010 and had a total monthly precipitation value that appeared on paper to be nearly normal. However, the monthly distribution was anything but normal. One likely source for these heavier precipitation events could be the warm, moist air from the Gulf of Mexico. Southeasterly flow around the western periphery of the Bermuda High advects maritime tropical air towards the Rocky Mountains. This air is forced upward by the elevated terrain, causing the air to condense out some of its moisture and produce clouds. At this point, convection is carried eastward by the 2

3 prevailing westerlies, resulting in evening and overnight thunderstorm outbreaks in the Midwest (Ahijevych et. al., 2004). In the summertime, these convective events occur on a nearly daily basis and indicate that the topography of the Rockies can have long-range effects on the distribution of precipitation across the U.S as these episodes can extend all the way to 75 W on a daily basis (Fig. 1). Flash flooding can be one of the many results of heavy precipitation events. According to Brooks and Stensrud (1999), flash floods are generally concentrated along the Gulf Coast in the winter months and gradually extend northward as temperatures begin to rise in the summer and high dewpoint air is advected northward (Fig. 2). Since the geographical range where flash floods are likely to occur increases in the summer, the number of flash floods that occur in the summer also increases. Increasing temperatures and larger amounts of moisture result in large convective available potential energy (CAPE), resulting in more convective episodes. This northward increase in flash flood events is in part a reflection of an increased number of convective episodes (Ahijevych et. al., 2004). When conducting research that involves the measurement of precipitation, it is vital to keep in mind the wide variability and numerous issues associated with rain gauges. Some gauges located in windy regions are equipped with wind shields while others contain small, electrical heaters designed to melt snow and ice to obtain a more accurate reading (Groisman, 1994). Evaporation is also thought to be a source of error when measuring precipitation because it is incredibly difficult to measure and there is no means of regulating its measurement. Furthermore, rain gauges are sometime moved from one location to another. This station advection may also cause discontinuities in 3

4 the data. Therefore, data collected from rain gauges is not always completely accurate and are expected to have an average bias of 9% (Groisman, 1994). It is also important to consider the varying lengths of precipitation events. As discussed in Warner and Mass (2012), precipitation events are not limited to 24-hours. Some precipitation events can last up to 96 hours and beyond. Therefore, it can be difficult to determine what measure of time should be used to identify precipitation events because these events are not bounded by daily intervals. While the daily unit might be the easiest way to examine precipitation events, it might be better to try and look at precipitation events in their entirety instead of using predetermined time measurements. The purpose of this paper is to examine normal precipitation months over a 30- year climatology at 10 stations located in different geographic and climatic locations across the U.S. to determine if normal precipitation months are comprised of several days of little precipitation, a few days of extreme precipitation, or a hybrid of these two possibilities. The rest of this paper is organized as follows. Section 2 will outline the criteria used to determine which cities and months were studied along with the method utilized to account for the amount of precipitation and the degree of severity of the individual precipitation events within each month. Section 3 will examine results that have been gathered from 10 cities across the U.S, while Section 4 will discuss the results presented herein relative to previous research and will summarize pertinent results. 2. Data/Methodology The cities looked at in this study include Geneva, NY, Denver, CO, Buffalo, NY, Tampa, FL, New Orleans, LA, Bismarck, ND, Seattle, WA, San Francisco, CA, Norfolk, 4

5 VA, Louisville, KY, and Springfield, MO. These cities were chosen based on their climate, geographic location, the topography of the region, and their proximity to bodies of water (Fig. 3). Geneva was selected because the station is nearby, the weather patterns are well known, and the information is easily accessible. Geneva was the first city to be studied, served as proof of concept, and was considered to be particularly interesting due to the nearby location of Hobart and William Smith Colleges. Denver was selected due to its location near the Rocky Mountains where convective episodes initiate, and its long periods of drought followed by brief, heavy rains. Buffalo was chosen because of its close proximity to the Great Lakes. Tampa and New Orleans were chosen because each city is located on the Gulf of Mexico. Bismarck was selected due to its northerly latitude and relatively high elevation. Seattle was studied because it is positioned in the northwest corner of the continental United States, it is located on the coast of the Pacific Ocean, and it is known for receiving large annual rainfall totals. San Francisco was chosen based on its location on the coast of the Pacific Ocean and its westerly longitude. Norfolk was selected due to its close proximity to the Atlantic Ocean and its easterly longitude. Louisville was studied because of its proximity to the Appalachian Mountains, and Springfield was chosen because it is positioned on the central Plains. However, because the information for Geneva was collected from the Cornell Agricultural Experimentation Station while the information for all other cities was collected from the NCDC database 1, Geneva will not be compared to the other 10 cities. In order to acquire a statistically meaningful dataset, thirty years worth of data was collected from each city between 1981 and Only the warm-seasons spanning 1 5

6 April through September were selected due to the increased likelihood for convective precipitation events. As mentioned in Groisman (1994), some rain gauges include small, electrical heaters that melt any snow or ice that collects in the gauge to help obtain more accurate readings. Thus, selecting months that have a decreased chance for snow and ice accumulation may help to decrease the chance for discontinuities caused by variability in gauges. Daily precipitation amounts, measured in tenths of a millimeter, were collected from NCDC for all cities excluding Geneva. For Geneva, the daily precipitation amounts were collected in hundredths of an inch from the Cornell Agricultural Experimental Station. As discussed in Section 1, examining data in terms of days can potentially cause misleading results because precipitation events can span more than 24-hours. However, at this time, quantifying time in days is a natural first step to collecting and interpreting the data for this research. All daily precipitation data was then converted into inches and summed together to determine the cumulative precipitation for each month. Since the goal of this project is to determine how precipitation is distributed in normal months, all months at each station that were between plus and minus one standard deviation of the mean were kept and considered normal. For the first few cities that were studied, the monthly precipitation was quite close to normally distributed. Therefore, approximately 18 to 22 years (out of 30) were kept for analysis for each month and city. However, upon examining Tampa, FL, it was discovered that some months had such a large standard deviation that months in which 0.03 inches of rain or less had fallen were being considered normal. Therefore, because monthly precipitation was no longer normally distributed, using years that had monthly precipitation totals that were one 6

7 standard deviation away from average was determined to be an inaccurate way of depicting normal. Instead, a new method was formulated in which the median for each month was calculated. From the median value, plus and minus ten years of the median were considered to have normal precipitation. By using plus and minus ten years away from the cumulative monthly median, 20 out of a possible 30 years or 67% of all cases were used for each month at each city, which is comparable to the 68% of cases that would be identified for using plus or minus one standard deviation for normally distributed data. For all normal months at each city, the largest single precipitation event during that month, the number of days it rained during that month, and the average rainfall on the days that it did rain were all recorded. Using this information, two methods were implemented to measure the impact of individual precipitation events on the overall cumulative monthly rainfall. In the first method, each year for a given month and city was sorted into a group based on the largest single precipitation event during that month. The 30-year average cumulative rainfall amount for each month was multiplied by 0.2, 0.4, 0.6, and 0.8 to created five different sorting groups. Months were placed into groups based on whether the largest precipitation event for that month was <20%, between 20-40%, between 40-60%, between 60-80%, or between % of the average cumulative rainfall amount. Eventually, this method was determined to be inferior because some months had aboveaverage precipitation and a large enough precipitation event to fall into a non-existent group of >100% of the average cumulative rainfall. The second method sorted each month and city based on what percentage of the cumulative rainfall amount resulted from the largest precipitation event during that month. This percentage was calculated by 7

8 dividing the largest precipitation event for a single month by the cumulative rainfall amount for the same month and multiplying by 100, yielding a percentage. These percentages were then subdivided into groups of <20%, 20-40%, 40-60%, 60-80%, and % leading to more robust results. 3. Results Although the Geneva data is not from NCDC, results will still be discussed herein but not compared to other cities. On days when it rained, the average amount of rainfall during the warm season in Geneva was 0.26 inches. In addition, Geneva received an average of 3.11 inches of cumulative rainfall per month and rain fell on approximately days each month with the largest precipitation event averaging at 0.96 inches (Table 1). In Geneva, extreme precipitation events within normal" months appear to gradually increase from April to September. In April, the maximum percentage of the rainfall total that was caused by a single precipitation event was 50%, but by September, it had risen to 72%. Similarly, the minimum in April was 8% of the total, which rose to 19% of the total by September (Fig. 4A). Severe weather outbreaks, such as thunderstorms, are usually most common during the middle of the summer due to the increased levels of CAPE in the atmosphere at this time. Although some days in September had precipitation events that accounted for as much as 72% of the cumulative precipitation for that month, the inter-quartile range stayed between 22% and 42% of the cumulative monthly rainfall (Fig. 4B). Thus, normal precipitation months in Geneva typically had a single event that accounted for 1/5 to 2/5 of the total monthly rainfall. 8

9 A perfect example of the stark contrast between months with a few large precipitation events and several small precipitation events occurred in Geneva. Fig. 5A illustrates the amount of precipitation that fell each day in September 2000 in Geneva. During that month, the precipitation was spread out over a period of 15 different days, resulting in a cumulative monthly precipitation total of 3.82 inches. Fig. 5B represents the amount of precipitation that fell each day in September 2001 in Geneva. Unlike September 2000, rainfall was far from equally distributed in September During that month, rain only fell on 11 days and only one of those days had more than 0.50 inches of precipitation. However, the one-day that had precipitation greater than 0.3 inches had 2.95 inches of rain, giving both September 2000 and September 2001 comparable cumulative monthly precipitation values of 3.82 and 4.11, respectively, even though the monthly distribution was markedly different. Additionally, both months had cumulative monthly precipitation totals that exceeded the mean cumulative monthly precipitation value, suggesting that both months had a surplus of precipitation even though September 2001 had minimal precipitation prior to the 25 th of the month resulting in conditions that approached that of a short-term drought. Denver averaged precipitation on 8.78 days each month and on those days, only 1.70 inches of rain fell. These small values indicate that Denver has a dry, arid climate (Table 2). Excluding June, the maximum percentage of average monthly rainfall from a single precipitation event increased into July as temperatures rose, but decreased into September as temperatures began to drop. This pattern is very predictable and logical given that thunderstorms, a major source for precipitation events, also occur most frequently in the mid-summer as temperatures rise and environmental CAPE increases. 9

10 Many of these thunderstorms produce convective episodes that traverse the central U.S. (Ahijevych et. al., 2004). Not only did July have the most extreme maximum single precipitation event of the monthly total at 88%, but it also had the most extreme minimum precipitation event of the monthly total at 13% (Fig. 6A). Denver s interquartile ranges spanned from 28% to 58% of the total monthly precipitation (Fig. 6B). Thus, normal precipitation months in Denver typically had a single event that accounted for 1/3 to 1/2 of the total monthly rainfall. Buffalo averaged precipitation on days each month and on those days, an average of 0.29 inches of rain fell. Buffalo s largest monthly precipitation event was, on average, 1.03 inches (Table 3). In Buffalo, the maxima did not follow the same pattern that occurred in Denver. The maximum precipitation events in Buffalo declined from May through August, but spiked back up in September where one year had a single precipitation event that comprised 80% of the average precipitation (Fig. 7A). Minima were also fairly consistent, remaining confined between 15% and 19% of the total precipitation. The inter-quartile ranges hovered between 21% and 39% of the total precipitation and showed a general increase from April through September (Fig. 7B). Thus, normal precipitation months in Buffalo typically had a single event that accounted for 1/5 to 2/5 of the total monthly rainfall. With an average of 0.43 inches of precipitation, Tampa, Florida proved to have a higher average daily rainfall on days that it did rain than any other city previously studied. One likely cause for this increase could be the fact that Tampa is located significantly further south than Denver and Buffalo. Tampa s southerly latitude causes temperatures to be higher, air to hold more moisture, increasing the CAPE, and leading to 10

11 more precipitation events. Furthermore, Tampa is the first city to be studied that is located on the coast of the Gulf of Mexico, allowing two sea breezes to increase precipitation totals. Typically, Tampa gets precipitation on 9.2 days each month. However, in April and May, the average number of days it rains is 5.15 days and 5.60 days, respectively. By June, this value jumps to days and further increases to greater than 15 days for the following three months (Table 4). Tampa s subtropical location also caused the percentages of total monthly rainfall caused by the single largest precipitation event to form a unique pattern. Instead of following the patterns seen in other cities, percentages peaked in April and May, dropped off in June, and rose again in September (Fig. 8A). Percentages may have peaked in Tampa during the spring instead of the summer because there is a greater land-ocean temperature gradient in the spring than in the summer, causing more extreme precipitation events to develop in the spring. However, these high percentages for April and May (Fig. 8B) can be deceiving if they are not viewed in conjunction with the values in Table 4. Percentages in April and May are larger because the average largest precipitation event during those months was only slightly smaller than the average cumulative rainfall for those months. Percentages in June though September were smaller because the average cumulative rainfall amount increased dramatically, not because the average largest event decreased. Instead, the average largest event for each month remained about the same from April through September. The mean value, 1 st quartile, 3 rd quartile, and inter-quartile range were all indicative of this increasing and then decreasing pattern. The maximum percentage of 100% was reached in May 2007 when 0.35 inches of precipitation fell and accounted for the entire cumulative monthly precipitation value. 11

12 The minimum percentage occurred in June 1999 when only 11% of the total monthly precipitation was caused by the single largest event. The mean percentages for April and May in Tampa fell right around the 60% line, which is an astonishingly high value given that most other cities have 3 rd quartiles that barely reach the 60% line (Fig 8B). Thus, in the early warm season when Tampa does not receive as much rain, normal precipitation months can have a single event that accounts for 1/2 to 3/4 of the total monthly rainfall, while in the later warm season months the largest events are reduced to 1/5 to 2/5 of the total monthly rainfall Also located in the subtropics and on the coast of the Gulf of Mexico, New Orleans had very similar values to those of Tampa. New Orleans average daily rainfall on days that it did rain was 0.51 inches, even higher than that of Tampa. Rain fell on an average of days each month and the average largest monthly precipitation event was the highest yet with 1.82 inches (Table 5). The percentage of total monthly rainfall in New Orleans that was caused by the month s single largest precipitation event shows a pattern comparable to that of Tampa. Percentages peak in April and May, drop in June, and slowly begin to rise from June to September (Fig. 9A). Thus, in the early warm season when New Orleans does not receive as much rain, normal precipitation months can have a single event that accounts for 1/3 to 2/3 of the total monthly rainfall, while in the later warm season months the largest events are reduced to 1/5 to 2/5 of the total monthly rainfall. Furthermore, the mean value, 1 st quartile, 3 rd quartile, and inter-quartile range all follow this exact pattern. However, despite their obvious similarities, there are several differences between Tampa and New Orleans. In April and May, New Orleans percentages are not as high as those in Tampa 12

13 and in June through September, New Orleans percentages are not as low as those in Tampa. Additionally, while Tampa has a maximum and minimum of 100% and 11% respectively, New Orleans only has a maximum and minimum of 94% and 17% respectively (Fig. 9B). These less extreme percentage values in New Orleans are a result of increased average cumulative precipitation in April and May as well as decreased average cumulative precipitation in June, July, August, and September. An explanation for the varying levels of average cumulative precipitation between Tampa and New Orleans could be that New Orleans is not as far south as Tampa, resulting in less heat and a decreased likelihood for convective outbreaks that could result in large-scale precipitation events. Furthermore, while Tampa has two sea breezes coming from the Atlantic Ocean and the Gulf of Mexico, New Orleans only has one sea breeze that comes off the Gulf of Mexico. In Bismarck, the average daily precipitation value on days it rained was 0.22 inches and, on average, precipitation fell on 8.76 days per month. The average cumulative rainfall was 1.88 inches and the average largest event was 0.76 inches (Table 6). Looking at Fig. 10B, maxima occur in June and July, reaching as high as 83% and 87%, respectively. However, the 3 rd quartile of the inter-quartile range reaches its highest percentages in April and September values of 55% and 62%, respectively. June had not only a maximum percentage, but also the minimum percentage of 11%. The median, 1 st quartile, and inter-quartile range all decreased from April to June and then increased from July to September. Thus, normal precipitation months in Bismarck typically had a single event that accounted for 1/3 to 2/3 of the total monthly rainfall. Although percentages decreased from April through June (Fig. 10A), the average number of days that it rained, 13

14 the average rainfall on days that it did rain, the average cumulative rainfall, and the average largest event all increased. This increase indicates that percentages are only decreasing because the average cumulative rainfall and the average largest event are both increasing, but the average cumulative rainfall is decreasing slightly faster than the average largest event, resulting in potentially misleadingly high percentages (Fig. 10B) On days that it did rain in Seattle, the average rainfall total was 0.16 inches and, on average, it rained 8.95 days each month (Table 7). Although Seattle is known for receiving large amounts of precipitation, April through September is Seattle s dry season, explaining why values for Seattle in this study are lower than what might otherwise be expected. The percentage of the cumulative precipitation that fell during the single largest precipitation event increased from April to July and decreased from July through September (Fig. 11A). Thus, normal precipitation months in Seattle typically had a single event that accounted for 1/5 to 2/3 of the total monthly rainfall. However, as percentages rise from April through July, the average number of days that it rained, the average rainfall on days that it did rain, the average cumulative rainfall, and the average largest event all decline as Seattle approaches the height of its dry season. This decline indicates that percentages are only increasing because the average cumulative rainfall and the average largest event are both decreasing, but the average cumulative rainfall is decreasing slightly faster than the average largest event, resulting in potentially misleadingly high percentages (Fig. 11B). The significant lack of precipitation in San Francisco resulted in extremely anomalous results, as seen in Table 8. On average, precipitation only fell on 1.85 days each month and the average rainfall on those days when it did rain was 0.06 inches. Fig. 14

15 12A shows that, it is difficult to determine any kind of pattern over the 6-month period given that during the 20 normal July months and the 20 normal August months, it only rained during one normal August month. In Fig. 12B, inter-quartile ranges span 35-62% in April and 54-78% in May. Thus, normal precipitation months in San Francisco typically had a single event that accounted for 1/3 to 2/3 of the total monthly rainfall in April and 1/2 to 4/5 of the total monthly rainfall in May. Once again, it is important to study Fig. 12B in conjunction with Table 8. Table 8 shows that April and May only receive an average cumulative rainfall total of 1.02 inches and 0.31 inches, respectively. These average cumulative rainfall totals, when divided by their respective average largest event values of 0.53 inches for April and 0.18 inches for May, result in percentages that can be deceivingly high. A similar problem occurs in June and September because the average largest event and the average cumulative rainfall are nearly identical. Precipitation is so scarce that events that would otherwise be considered to be small precipitation events, are accounting for large percentages of the cumulative precipitation totals. July and August have no inter-quartile range because it did not rain more than once for either month. In Norfolk, the average number of days it rained was days per month and on those days, it rained an average of 0.41 inches. The average cumulative rainfall was 4.10 inches and the average largest event was 1.46 inches (Table 9). Similar to Buffalo, there is little fluctuation in the percentage of total monthly rainfall that fell during the single largest event each month from April through September in Norfolk (Fig. 13A). Percentages rise and fall slightly from month to month, but never vary by more than 10%. Inter-quartile ranges hover between 26-47%. Thus, normal precipitation months in 15

16 Norfolk typically had a single event that accounted for 1/4 to 1/2 of the total monthly rainfall. The stable, consistent precipitation totals in Norfolk are likely a result of the nearby Atlantic Ocean, which acts to stabilize the climate in this region. The only values that seem to break away from this steady pattern are the maxima in July and September of 79% and 76%, respectively (Fig. 13B). Similar to Norfolk, in Louisville, the average number of days it rained was days per month and on those days, it rained an average of 0.37 inches. The average cumulative rainfall was 4.10 inches and the average largest event was 1.26 inches (Table 9). The percentage of monthly rainfall that fell during the single largest event slowly increases during the 6-month period of April through September (Fig. 14A). Thus, normal precipitation months in Louisville typically had a single event that accounted for 1/4 to 1/2 of the total monthly rainfall. However, as the percentages are increasing, the average number of days that it rained and the average cumulative rainfall decrease. Since the average largest event remains fairly constant, the decreasing average cumulative rainfall is what causes the percentages to rise. Local maxima occur in July, August, and September with values of 72%, 66%, and 79%, respectively (Fig. 14B). Similar to both Norfolk and Louisville, in Springfield, the average number of days it rained was 9.66 days per month and on those days, it rained an average of 0.45 inches. The average cumulative rainfall was 4.03 inches and the average largest event was 1.52 inches (Table 11). The percentage of monthly cumulative rainfall that fell during the single largest event fluctuates more in Springfield than it did in Norfolk, possibly due to Springfield s close proximity to the Ozark Mountains. However, the values are still somewhat consistent from month to month with inter-quartile ranges staying between 16

17 23% and 51%. Thus, normal precipitation months in Springfield typically had a single event that accounted for 1/4 to 1/2 of the total monthly rainfall. However, there does appear to be a trend of overall increasing values during the 6-month period. As in several other cities, although the percentages increase, the average number of days it rained and the average cumulative rainfall decrease. Like in Norfolk, because the average largest event remains primarily consistent, it is the decreasing average cumulative rainfall that causes percentages to rise, particularly in July, August, and September. It is also important to note that the average rainfall on days that it rained remains consistent from month to month as well. Local maxima occur sporadically in May, July, and September with values of 84%, 76%, and 94%, respectively (Fig. 15A, 15B). 4. Discussion After analyzing all eleven cities, several patterns and trends became apparent. Months with high percentages of rainfall that fell during the single largest precipitation event of the month generally had low average cumulative rainfall values and a low number of days in which rain fell. This is because the average largest event values either remained consistent regardless of the average cumulative rainfall or the average largest event values decreased less severely than the average cumulative rainfall. Therefore, in months with smaller rainfall totals, few events, and largest monthly precipitation events of comparable magnitude, percentages were high. For example, during April in Tampa, the average cumulative rainfall was 1.68 inches, which is significantly lower than other months like August that received 7.49 inches of cumulative rainfall. In April, it only rained an average of 5.15 days each month while August had rain on an average of days each month. However, the average 17

18 largest event for each month remained consistent; April s average largest event was 1.03 inches while August s average largest event was 1.75 inches (Table 4). This resulted in April having very high percentages of the total monthly rainfall that fell during the month s single largest precipitation event and August having very low percentages of the total monthly rainfall that fell during the month s single largest precipitation event. April had an inter-quartile range of 54%-74% while August had an inter-quartile range of 19%- 27% (Fig. 8B). Thus, normal precipitation in April typically had a single event that accounted for 1/2 to 3/4 of the total monthly rainfall while normal precipitation in August typically had a single event that accounted for 1/5 to 1/4 of the total monthly rainfall. Months with low percentages of rainfall that fell during the single largest precipitation event of the month had large average cumulative rainfall values and a large number of days in which rain fell. This is because the average largest event values either remained consistent regardless of the average cumulative rainfall or the average largest event values increased less severely than the average cumulative rainfall. Therefore, in months with larger rainfall totals, more precipitation events, and largest monthly precipitation events of comparable magnitude, percentages were lower. For example, during May in Louisville, the average cumulative rainfall was 4.80 inches, which is significantly higher than other months like September that only received 2.84 inches of cumulative rainfall. In May, it rained an average of days each month while August had rain on an average of 7.95 days each month. However, the average largest event for each month remained consistent; May s average largest event was 1.43 inches while September s average largest event was 1.10 inches (Table 10). This resulted in May having low percentages of the total monthly rainfall that fell during the month s 18

19 single largest precipitation event and August having high percentages of the total monthly rainfall that fell during the month s single largest precipitation event. May had an interquartile range of 23%-35% while September had an inter-quartile range of 31%-57% (Fig. 14B). Thus, normal precipitation in May typically had a single event that accounted for 1/4 to 1/3 of the total monthly rainfall while normal precipitation in September typically had a single event that accounted for 1/3 to 1/2 of the total monthly rainfall. This variety in the distribution of precipitation is in part caused by convective episodes associated with mesoscale convective systems (MCS) that form in the lee of the Rocky Mountains and are then carried eastward with the prevailing westerlies (e.g., Ahijevych et al. (2004)). This is evident by the low percentage of total monthly rainfall that fell during the month s single largest event for cities such as Denver, Springfield, Louisville, and Norfolk. Low percentages correlate to an increased average number of days that it rained each month and an increased average cumulative rainfall for each month, both of which could be the result of passing MCSs. In the months where an average month featured a large event, their convective episodes had the potential to produce flash flooding (e.g., Brooks et al.). In general normal precipitation months varied widely across the 11 locations studied across the U.S., from month to month, and year to year. This study shows the pitfalls of simply using the average to describe the precipitation distribution. 19

20 5. References Ahijevych, D. A., C. A. Davis, R. E. Carbone, J. D. Tuttle, 2004: Initiation of Precipitation Episodes Relative to Elevated Terrain. J. Atmos. Sci., 61, 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, Groisman, P. Y., and D. R. Legates, 1994: The Accuracy of United States Precipitation Data. Bull. Amer. Meteor. Soc., 75, Warner, M. D., and C. F. Mass, 2012: Wintertime Extreme Precipitation Events along the Pacific Northwest Coast: Climatology and Synoptic Evolution. Mon. Wea. Rev., in press. 20

21 6. Tables Month (#of years) average # of days it rained average rainfall average cumulative rainfall April (20) May (19) June (20) July (22) August (22) September (19) Cumulative Average average largest event Table 1 April through September in Geneva, NY from Number of days it rained each month, the average rainfall on days when it rained, cumulative monthly rainfall, and the amount of rain that fell during the largest precipitation event that month (measured in inches). Values discussed in the journal are highlighted in red. Source: Cornell Agricultural Experimental Station Surface Data. Month (# of years) average # of days it rained average rainfall (in.) average cumulative rainfall (in.) April (20) May (19) June (19) July (20) August (19) September (20) Cumulative Average average largest event (in.) Table 2 April through September in Denver, CO from Number of days it rained each month, the average rainfall on days when it rained, cumulative monthly rainfall, and the amount of rain that fell during the largest precipitation event that month (measured in inches). Values discussed in the journal are highlighted in red. Source: NCDC daily precipitation archive. 21

22 Month (# of years) average # of days it rained average rainfall (in.) average cumulative rainfall (in.) April (20) May (20) June (20) July (20) August (20) September (20) Cumulative Average average largest event (in.) Table 3 April through September in Buffalo, NY from Number of days it rained each month, the average rainfall on days when it rained, cumulative monthly rainfall, and the amount of rain that fell during the largest precipitation event that month (measured in inches). Values discussed in the journal are highlighted in red. Source: NCDC daily precipitation archive. Month (# of years) average # of days it rained average rainfall average cumulative rainfall April (20) May (20) June (20) July (20) August (20) September (20) Cumulative Average average largest event Table 4 April through September in Tampa, FL from Number of days it rained each month, the average rainfall on days when it rained, cumulative monthly rainfall, and the amount of rain that fell during the largest precipitation event that month (measured in inches). Values discussed in the journal are highlighted in red. Source: NCDC daily precipitation archive. 22

23 Month (# of years) average # of days it rained average rainfall average cumulative rainfall April (20) May (20) June (20) July (20) August (20) September (20) Cumulative Average average largest event Table 5 April through September in New Orleans, LA from Number of days it rained each month, the average rainfall on days when it rained, cumulative monthly rainfall, and the amount of rain that fell during the largest precipitation event that month (measured in inches). Values discussed in the journal are highlighted in red. Source: NCDC daily precipitation archive. Month (# of years) average # of days it rained average rainfall average cumulative rainfall April (20) May (20) June (20) July (20) August (20) September (20) Cumulative Average average largest event Table 6 April through September in Bismarck, ND from Number of days it rained each month, the average rainfall on days when it rained, cumulative monthly rainfall, and the amount of rain that fell during the largest precipitation event that month (measured in inches). Values discussed in the journal are highlighted in red. Source: NCDC daily precipitation archive. 23

24 Month (# of years) average # of days it rained average rainfall average cumulative rainfall April (20) May (20) June (20) July (20) August (20) September (20) Cumulative Average average largest event Table 7 April through September in Seattle, WA from Number of days it rained each month, the average rainfall on days when it rained, cumulative monthly rainfall, and the amount of rain that fell during the largest precipitation event that month (measured in inches). Values discussed in the journal are highlighted in red. Source: NCDC daily precipitation archive. Month (# of years) average # of days it rained average rainfall average cumulative rainfall April (20) May (20) June (20) July (20) August (20) September (20) Cumulative Average average largest event Table 8 April through September in San Francisco, LA from Number of days it rained each month, the average rainfall on days when it rained, cumulative monthly rainfall, and the amount of rain that fell during the largest precipitation event that month (measured in inches). Values discussed in the journal are highlighted in red. Source: NCDC daily precipitation archive. 24

25 Month (# of years) average # of days it rained average rainfall average cumulative rainfall April (20) May (20) June (20) July (20) August (20) September (20) Cumulative Average average largest event Table 9 April through September in Norfolk, VA from Number of days it rained each month, the average rainfall on days when it rained, cumulative monthly rainfall, and the amount of rain that fell during the largest precipitation event that month (measured in inches). Values discussed in the journal are highlighted in red. Source: NCDC daily precipitation archive. Month (# of years) average # of days it rained average rainfall average cumulative rainfall April (20) May (20) June (20) July (19) August (20) September (20) Cumulative Average average largest event Table 10 April through September in Louisville, KY from Number of days it rained each month, the average rainfall on days when it rained, cumulative monthly rainfall, and the amount of rain that fell during the largest precipitation event that month (measured in inches). Values discussed in the journal are highlighted in red. Source: NCDC daily precipitation archive. 25

26 Month (# of years) average # of days it rained average rainfall average cumulative rainfall April (20) May (20) June (20) July (20) August (20) September (20) Cumulative Average average largest event Table 11 April through September in Springfield, MO from Number of days it rained each month, the average rainfall on days when it rained, cumulative monthly rainfall, and the amount of rain that fell during the largest precipitation event that month (measured in inches). Values discussed in the journal are highlighted in red. Source: NCDC daily precipitation archive. 26

27 7. Figures Fig. 1 Radar-estimated rainfall rate averaged between 30 and 48 N for 4 days in July As precipitating systems move eastward, they appear as narrow rain streaks sloping downward from left to right. Three significant precipitation episodes are evident, all starting at approximately 2200 UTC, 105 W on 18, 20, and 21 July. Source: Ahijevych et. al. (2004). 27

28 Fig. 2 Frequency (events/year) of 1 in. h -1 or larger rainfall totals for each month objectively analyzed to a regular grid from the Hourly Precipitation Dataset (HPD) stations. Contour intervals of 0.1, 0.2, 0.25, 0.33, 0.5, 0.66, 0.75, and 1.0 events year -1. Source: Brooks and Stensrud (1999). 28

29 Fig. 3 Annual precipitation in inches across the U.S. from Black dots indicate cities that were studied. Source: Spatial Climate Analysis Service, Oregon State University. 29

30 Fig. 4A Scatter plot diagram representing the percentage of cumulative monthly precipitation that occurred during the largest precipitation event of each average month from in Geneva, NY. Source: Cornell Agricultural Experimental Station Surface Data. 30

31 Fig. 4B Box and whisker diagram representing the percentage of cumulative monthly precipitation that occurred during the largest precipitation event of each average month from in Geneva, NY. The colored boxes, green triangle, red square, purple X, blue diamond, and turquoise X represent the inter-quartile range, the median, the minimum, the maximum, and the 1 st and 3 rd quartile, respectively. Source: Cornell Agricultural Experimental Station Surface Data. 31

32 Fig. 5A Precipitation in Geneva, NY during September Blue bars represents daily precipitation in inches. Red line represents cumulative precipitation in inches. Green line represents the mean precipitation in inches. 32

33 Fig. 5B Precipitation in Geneva, NY during September Blue bars represents daily precipitation in inches. Red line represents cumulative precipitation in inches. Green line represents the mean precipitation in inches. 33

34 Fig. 6A Scatter plot diagram representing the percentage of cumulative monthly precipitation that occurred during the largest precipitation event of each average month from in Denver, CO. Source: NCDC daily precipitation archive. 34

35 Fig. 6B Box and whisker diagram representing the percentage of cumulative monthly precipitation that occurred during the largest precipitation event of each average month from in Denver, CO. The colored boxes, green triangle, red square, purple X, blue diamond, and turquoise X represent the inter-quartile range, the median, the minimum, the maximum, and the 1 st and 3 rd quartile, respectively. Source: NCDC daily precipitation archive. 35

36 Fig. 7A Scatter plot diagram representing the percentage of cumulative monthly precipitation that occurred during the largest precipitation event of each average month from in Buffalo, NY. Source: NCDC daily precipitation archive. 36

37 Fig. 7B Box and whisker diagram representing the percentage of cumulative monthly precipitation that occurred during the largest precipitation event of each average month from in Buffalo, NY. The colored boxes, green triangle, red square, purple X, blue diamond, and turquoise X represent the inter-quartile range, the median, the minimum, the maximum, and the 1 st and 3 rd percentiles, respectively. Source: NCDC daily precipitation archive. 37

38 Fig. 8A Scatter plot diagram representing the percentage of cumulative monthly precipitation that occurred during the largest precipitation event of each average month from in Tampa, FL. Source: NCDC daily precipitation archive. 38

39 Fig. 8B Box and whisker diagram representing the percentage of cumulative monthly precipitation that occurred during the largest precipitation event of each average month from in Tampa, FL. The colored boxes, green triangle, red square, purple X, blue diamond, and turquoise X represent the inter-quartile range, the median, the minimum, the maximum, and the 1 st and 3 rd percentiles, respectively. Source: NCDC daily precipitation archive. 39

40 Fig. 9A Scatter plot diagram representing the percentage of cumulative monthly precipitation that occurred during the largest precipitation event of each average month from in New Orleans, LA. Source: NCDC daily precipitation archive. 40

41 Fig. 9B Box and whisker diagram representing the percentage of cumulative monthly precipitation that occurred during the largest precipitation event of each average month from in New Orleans, LA. The colored boxes, green triangle, red square, purple X, blue diamond, and turquoise X represent the inter-quartile range, the median, the minimum, the maximum, and the 1 st and 3 rd percentiles, respectively. Source: NCDC daily precipitation archive. 41

42 Fig. 10A Scatter plot diagram representing the percentage of cumulative monthly precipitation that occurred during the largest precipitation event of each average month from in Bismarck, ND. Source: NCDC daily precipitation archive. 42

43 Fig. 10B Box and whisker diagram representing the percentage of cumulative monthly precipitation that occurred during the largest precipitation event of each average month from in Bismarck, ND. The colored boxes, green triangle, red square, purple X, blue diamond, and turquoise X represent the inter-quartile range, the median, the minimum, the maximum, and the 1 st and 3 rd percentiles, respectively. Source: NCDC daily precipitation archive. 43

44 Fig. 11A Scatter plot diagram representing the percentage of cumulative monthly precipitation that occurred during the largest precipitation event of each average month from in Seattle, WA. Source: NCDC daily precipitation archive. 44