Abstract

Temporal trends in precipitation associated with 500-hPa low pressure systems are assessed for consistency with overall precipitation trends in the northeastern United States. Increases in precipitation associated with closed upper lows can be attributed to an increase in the number of closed lows and/or an increase in the average precipitation occurring in association with closed lows. A climatological description of 500-hPa closed lows was developed, and precipitation was analyzed using data from the U.S. Cooperative Observer Program. Closed 500-hPa lows were identified within a rectangular geographic region bounded by 35° and 50°N and 65° and 90°W. Precipitation observations within 1.25° of latitude and 1.25° of longitude of the closed-low center were attributed to the closed low. Statistical testing procedures were conducted to evaluate whether a long-term trend existed in the closed-low frequency, the average precipitation occurring near closed lows, or the percentage of >2.54-cm precipitation observations associated with the lows. Regional trends (1948–2007) in the frequency of closed lows were evident, and a statistically significant increase in average precipitation near closed lows was found. Likewise, the percentage of precipitation totals in excess of 2.54 cm (and separately 5.08 cm) showed a statistically significant increase with time. In both cases, recent (2007) values were nearly 50% higher than the comparable value in 1948. Increases were particularly evident during the winter months. These trends are consistent with increases in tropospheric water vapor due to increased global mean temperature.

1. Introduction

a. Climate change and precipitation trends

In recent decades numerous studies have documented changes in the observational climate record. Particular attention has been paid to the precipitation record, with specific interest in extremes. Climate models have indicated that in a warmer future climate global mean precipitation and extreme precipitation frequency will increase (Trenberth et al. 2003; Meehl et al. 2005). The causes of this increase can be attributed to thermodynamic processes involving increased global mean precipitable water (Ross and Elliott 2001), and also to dynamic processes (Emori and Brown 2005). On a regional level, it is expected that annual precipitation will increase over the northeastern United States, with a greater percentage increase during the winter months (Schoof et al. 2010).

There is evidence showing that these precipitation increases have already begun over much of the continental United States over the past few decades, including the northeastern United States. Karl and Knight (1998) found that precipitation has increased by about 10% over the continental United States since 1910. Groisman et al. (2004) found a nationwide increase in annual precipitation of about 7% (100 yr)−1. Higgins et al. (2000) showed a positive trend in precipitation in the continental United States across all seasons.

It has also been shown that increases in the frequency of heavy-precipitation events are more pronounced than increases in mean precipitation. An increase in the intensity of precipitation events is contributing about one-half of the overall precipitation increase (Karl and Knight 1998). A statistically significant increase in the occurrence of precipitation events in excess of the 1-yr recurrence interval amount has been recently documented (Kunkel et al. 1999). This and similar increases in extreme rainfall parameters have been most pronounced in the northeastern United States and the western Great Lakes (DeGaetano 2009; Groisman et al. 2004).

Preliminary work analyzed partial-duration daily extreme-rainfall series at three major northeastern U.S. stations: General Edward Lawrence Logan International Airport in Boston, Massachusetts (BOS); John F. Kennedy International Airport in New York City, New York (JFK); and Washington Dulles International Airport in Chantilly, Virginia (IAD), as well as annual maximum flow data from a network of New England stream gauges (Collins 2009). The synoptic environment associated with these events was subjectively classified based on archived surface and upper-air (500 hPa) weather maps as tropical cyclones, closed upper lows, stationary fronts, convective storms, strong synoptic systems, and coastal lows. In the New England study, nearly 25% of the maximum annual flows were associated with a closed low, and the station-based precipitation totals indicated that more than one-third of the daily extremes at IAD occurred in conjunction with closed lows. A statistically significant increase in the number of heavy-precipitation events associated with closed lows was identified whereas little change in extreme-rainfall frequency was found with the other storm types. Across the southeastern United States, however, Knight and Davis (2009) found that extreme precipitation from tropical cyclones has been increasing in recent decades.

b. Motivation

This preliminary finding was the motivation for our current study: analyzing precipitation trends associated with closed upper lows. An increase in total precipitation associated with closed lows could be attributed to either one of two causes (or possibly both): 1) an increase in the number of closed lows over the northeastern United States or 2) an increase in the average amount of precipitation produced by a closed low. The purpose of this study was to determine which, if either, of these two mechanisms was responsible for the increases seen in the preliminary study. In section 1c, a number of previous studies that have developed climatological descriptions (“climatologies”) of closed lows are discussed. In section 2, the adaptation of these previous methods to this study is presented, as are the procedures for assigning daily rainfall to each closed low and other statistical methods. A presentation of results follows in section 3, and the major findings are discussed in section 4 and summarized in section 5.

c. Previous closed-low literature

Various definitions of a “closed upper low” exist in the literature. Nieto et al. (2005) used three parameters—200-hPa geopotential height, the pattern of thickness in relation to the low, and the temperature gradient pattern at 200 hPa—to identify “cutoff lows.” Bell and Bosart (1989) defined a “closed low” as a low at the 500-hPa level that contained at least one closed 30-m contour around a central minimum geopotential height. Parker et al. (1989) required at least one closed 60-m contour at the 500-hPa level. It is important to note the distinction between a cutoff low and a closed low—the former requires the condition that it be displaced from the main band of midlatitude westerlies, whereas the latter does not have to be displaced from the westerlies. Indeed, closed lows can often be especially deep synoptic-scale systems within very strong westerlies. Furthermore, closed lows can be associated with strong horizontal temperature gradients, baroclinic structure, and significant midlevel temperature advection, whereas cutoff lows tend to be equivalent barotropic.

The occurrence of closed 500-hPa lows is relatively infrequent. Parker et al. (1989) found that in any given 10° × 10° latitude–longitude quadrangle a closed low occurred generally less than 10% of the time. The frequency of closed lows generally increases with increasing latitude, albeit with some regional variations (Parker et al. 1989). Bell and Bosart (1989) found a closed-low frequency maximum in northeastern Canada that shifts southeastward to the Canadian Maritimes from January to March. Another frequency maximum was found in southwestern North America, especially during the autumn and spring months. This maximum was also identified by Parker et al. (1989) and was found to be slightly upstream of a major surface cyclogenesis region in the lee of the Rocky Mountains. A closed-low frequency and genesis maximum was observed in the Upper Midwest and Great Lakes region. Bell and Bosart (1989) attribute this maximum to a surface cyclogenesis maximum during late winter and spring in the lee of the Rockies (Whittaker and Horn 1981), because these surface cyclones often deepen significantly and become closed at the 500-hPa level. A similar surface cyclogenesis region off the east coast of North America could also be a contributing factor to the 500-hPa cyclogenesis and frequency maximum near the Canadian Maritimes.

2. Methods

A climatology of 500-hPa closed lows affecting the northeastern United States was formed using the National Centers for Environmental Prediction–National Center for Atmospheric Research (NCEP–NCAR) reanalysis dataset (Kalnay et al. 1996). Closed lows were identified by a computerized routine using 500-hPa geopotential height data available 4 times per day (0000, 0600, 1200, and 1800 UTC). The data are available at a resolution of 2.5° latitude × 2.5° longitude.

a. Diagnosis of 500-hPa closed lows

The closed-low centers were diagnosed within a rectangular geographical region bounded by 50°N to the north, 35°N to the south, 65°W to the east, and 90°W to the west (Fig. 1). Because of the labor-intensive nature of the methods used by Nieto et al. (2005) that required the identification of specific thickness and temperature gradient patterns that are not easily resolved in the reanalysis data, closed lows were identified using the 500-hPa level, and a 30-m closed height-contour requirement as used by Bell and Bosart (1989). This latter threshold was chosen because the 60-m requirement of Parker et al. (1989) yielded significantly fewer lows.

Fig. 1.

Boundaries of the geographical area used to identify 500-hPa closed lows. Dashed lines depict the division of the region into four analysis quadrants. Available stations for a 20 Feb 2007 closed-low event are indicated by small dots and show the representative distribution of precipitation reports.

Fig. 1.

Boundaries of the geographical area used to identify 500-hPa closed lows. Dashed lines depict the division of the region into four analysis quadrants. Available stations for a 20 Feb 2007 closed-low event are indicated by small dots and show the representative distribution of precipitation reports.

For each 6-h time interval, the lowest geopotential height was found in each 2.5°-latitude band within the study region. If the minimum height within a given latitude band was less than that of both the latitude to the north and the latitude to the south (regardless of the longitude), the location of the minimum height was stored as a “possible” closed low. A possible closed low was identified on the southern border (35°N) if the adjacent latitude row (37.5°N) had a higher minimum height value. If neither condition was met, the location of the minimum height on the northern border (50°N) was considered to be a possible closed low.

When more than one possible closed low was identified within the same latitude band, the one with the greatest height difference between that latitude and the adjacent one to the north (at the same longitude) was taken as the possible closed low. Because these instances were very rare, it was not labor intensive to analyze these cases manually to determine whether there were indeed two closed lows occurring simultaneously within the broad region. During the 60-yr evaluation period, there were only a few instances of two closed lows occurring simultaneously within the same latitude band, and in these cases the subsequent steps were applied to both low centers.

As an initial screening, the geopotential height at each proposed low center was compared with the heights of the surrounding eight grid points. If that height value was not lower than all of the surrounding grid points, it was not classified as a closed low. This step was necessary to address possible closed lows on the borders of the domain. A test, based on the work of Bell and Bosart (1989), was applied to the remaining possible closed lows. Sixteen radial arms were extended from the possible low center to a radius of 10° (latitude × longitude) at 45° compass-point intervals beginning at north and to a radius of 10° × 5° (latitude–longitude × longitude–latitude) at the intermediate 22.5° compass points (e.g., north-northeast) (Fig. 2). If the height difference between the maximum geopotential height on the radial arm and the minimum geopotential height of the low center was less than 30 m in any of these directions, the low was considered to be “not closed.” If all 16 radial arms met the 30-m standard, the system was considered to be “closed” at that particular time step. A less-stringent constraint was considered, as in Hirsch et al. (2001), which required >80% of the arms to meet the height difference criterion. This did not change the number of identified lows substantially.

Fig. 2.

The 16-radial-arms test, as applied to (a) a closed low and (b) an open trough. In (a), all 16 radial arms have a grid point that has a height of at least 3 dam (30 m) greater than the low center, but in (b) the radial arm pointing due north does not.

Fig. 2.

The 16-radial-arms test, as applied to (a) a closed low and (b) an open trough. In (a), all 16 radial arms have a grid point that has a height of at least 3 dam (30 m) greater than the low center, but in (b) the radial arm pointing due north does not.

b. Counting closed lows

To produce a count of closed lows, a method was devised that determined which time periods contained a closed low that was identified during an earlier time period. This was intended to ensure that a closed low occurring at time step t would only be counted at the initial time of occurrence and not be counted again in subsequent time steps t + 6 h, t + 12 h, and so on.

Distinguishing “old” from “new” closed lows was not straightforward. Although closed lows in the midlatitudes usually move along with the westerlies in a zonal west-to-east direction, there can often be exceptions. These include, but are not limited to, a low moving meridionally, a low that is quasi stationary, or a low that retrogrades. In addition, the speed of motion of these closed lows can vary greatly. Despite this variability, a set of rules was established to indicate the formation of a new low. If a closed low existed during consecutive time steps (6 h apart), it was assumed that it was the same closed low during both of these time steps if the low center moved no more than 10° to the west, east, or north and no more than 5° to the south. Farther movement to the south was considered unlikely and often indicated a second, separate, closed low. This condition required that the closed-low center be within approximately 715–1400 km of its position 6 h previous, or a forward speed of less than 119–233 km h−1. Because the vast majority of midlatitude synoptic-scale systems possess forward speeds lower than this range, this restraint erred on the side of counting too few individual lows rather than too many.

There were many cases in which a closed low was intermittently closed during a series of time periods. Any closed low that was identified more than 72 h after the dissipation of the previous closed low was considered to be a new closed low. This was based on the assumption that within a 72-h period it was very likely that the closed low moved to a location outside the analyzed regional grid. For closed lows that were identified less than 72 h after the previous closed-low time step, the position of the not-necessarily-closed-low center was evaluated at intermediate time steps to determine whether the center moved more than 10° to the west, east, or north or more than 5° to the south in any 6-h time period. Note that although a given low was not closed during the interval its center was identified and tracked by the existence of a local geopotential height minimum (the possible-closed-low location mentioned previously).

The series of low positions and times of occurrence for each year were manually screened to identify cases in which the automated routine may have mistakenly counted a single closed-low event as two discrete systems. Within the 60-yr period, there were 18 total corrections that were made to the automated list. Manual evaluation allowed features such as trough axes to be identified and facilitated the tracking of a single low based on its correspondence with the trough.

c. Determining precipitation associated with closed lows

Precipitation associated with 500-hPa closed lows was determined from U.S. Cooperative Observer Program stations located within 1.25° latitude and 1.25° longitude of the closed-low center. These data, contained in the National Climatic Data Center TD3200 database, are relatively free of instrument biases and were obtained from the Northeast Regional Climate Center. These station locations encompassed a 2.5° latitude × 2.5° longitude (approximately 275 km × 210 km) grid box (analogous to the resolution of the NCEP–NCAR reanalysis dataset). The use of a larger grid box would have increased the potential for including precipitation that occurred outside the closed low’s influence, and a smaller grid box would have potentially eliminated rainfall associated with the storm. Nonetheless, analyses were repeated using a larger domain encompassing stations within 2.5° of the closed-low center. Because the use of this larger domain did not change the results of subsequent analyses, only the original 1.25° results are presented.

For stations within the grid, two distinct time periods were used to determine the amount of precipitation associated with the closed low. The first involved precipitation for a 24-h period. In all cases, the UTC date of the 6-hourly closed-low occurrence was used to identify the proper daily precipitation report to assign to the low. A large proportion (>50%) of daily precipitation observations take place around 1200 UTC (between 0700 and 0900 local time). Observations taken near 1800 UTC are the next most frequent, and 0600 UTC readings occur at some stations but are less common (DeGaetano and Wilks 2008). When based on 1200 and 1800 UTC observations, daily precipitation allocated in this manner preferentially included precipitation occurring before the passage of the low, at which point the dynamical and thermal processes associated with vertical motion favor precipitation. These processes tend to be less favorable following the passage of the low. For instance, Fig. 3 shows that when a 1200 UTC low occurrence is paired with a 1200 UTC daily precipitation report all of the reported precipitation falls before the low occurrence. When a low occurs at 1800 UTC, the 1200 UTC precipitation report reflects precipitation occurring during the period from 30 to 6 h prior to the low and excludes all precipitation that falls after the low is identified and within the interval from 1200 to 1800 UTC before the occurrence of the low (Fig. 3).

Fig. 3.

Schematic diagram of daily rainfall accumulation periods corresponding to closed-low positions on days from d − 1 through d + 2. Closed-low occurrence times are given in the shaded boxes. The white boxes indicate the daily precipitation accumulation periods for observations taken at 1200, 1800, and 0600 UTC.

Fig. 3.

Schematic diagram of daily rainfall accumulation periods corresponding to closed-low positions on days from d − 1 through d + 2. Closed-low occurrence times are given in the shaded boxes. The white boxes indicate the daily precipitation accumulation periods for observations taken at 1200, 1800, and 0600 UTC.

The second time period considered precipitation over a 48-h period. This method was used because of the concern over the somewhat arbitrary nature of the divisions made in the 1-day method. A wider time range ensured a higher likelihood that any precipitation occurring with a closed low would be captured but also increased the probability of capturing precipitation that was not associated with the closed low. In addition to the daily precipitation total used in the 24-h-period analysis, a second daily precipitation report was included. For a closed low occurring on day d at 0600, 1200, or 1800 UTC, the 24-h precipitation for day d + 1 was included. Lows occurring at 0000 UTC used the 48-h total encompassing precipitation reports from days d and d − 1 (Fig. 3). Thus, for cases in which a low occurred at 1800 UTC, the use of two daily 1200 UTC precipitation reports included rainfall within the interval from 30 h prior to 18 h after the low occurrence (Fig. 3).

In both the 24- and 48-h cases, closed lows that were present in multiple time periods during a given precipitation-reporting period corresponded to a single daily precipitation report. In cases in which the 6-hourly closed-low occurrence times spanned two daily precipitation-reporting periods, however, the precipitation reports from both days were used. For instance, if a closed low was present at 1800 UTC on day d and 0000 UTC on day d + 1, the 24-h total for the 1800 UTC occurrence represented one precipitation report (the daily total for day d) and the 0000 UTC occurrence represented a different 24-h total representing day d + 1. Both of these values were included in determining the average precipitation associated with the closed low.

The average station precipitation per closed low was computed as

 
formula

where P is the 24- or 48-h precipitation accumulation at station n for closed low k. Computing average precipitation in this fashion eliminated possible biases associated with year-to-year differences in the number of stations located near closed lows, because each closed low was weighted equally regardless of the number of stations that were available. Figure 4 shows that since 1970 the average number of precipitation stations within the 2.5° × 2.5° domain centered on the closed low was approximately 200, with the geographic distribution of stations being relatively even over the study region (Fig. 1). This number is lower than that in earlier years when the number of available stations often exceeded 300. The selected computation method mitigated any resulting biases that were due to this changing station density.

Fig. 4.

Average number of precipitation reports per closed low for each year in the study period. The 5-yr running mean is shown by the heavy black line.

Fig. 4.

Average number of precipitation reports per closed low for each year in the study period. The 5-yr running mean is shown by the heavy black line.

A second measure of closed-low precipitation was also formulated. For each low, the number of stations reporting ≥2.54 cm (and separately ≥5.08 cm) was tallied and divided by the total number of available stations under the low. These percentages were then averaged for all lows in a season to give the average annual percentage of stations receiving ≥2.54 cm (≥5.08 cm) of precipitation from closed lows. The use of percentages aided in mitigating the subtle change in station density through the study period.

d. Statistical testing

Statistical testing was based on Karl et al. (1987), who used difference series to evaluate temperature trends such that each observation received equal weighting. Time series of annual, seasonal, monthly, or regional average precipitation (and closed-low counts) spanning the 60-yr (1948–2007) period were converted to z values. Using the z values, 10-yr running means were computed and were used to form a series of differences between subsequent running-mean values (i.e., the 10-yr mean for years from ti to ti+10 was subtracted from that for years from ti+1 to ti+11). This resulted in a set of fifty 10-yr first-difference values. A Student’s t test was then conducted on the set of difference values to determine whether there were any significant long-term trends in the data, with the null hypothesis being that no long-term trend existed.

A nonparametric bootstrapping procedure was used to generate an estimated null distribution of these t values, against which the actual t value computed from the original data could be compared (Wilks 2006, 166–171). The original dataset was shuffled, creating a new artificial dataset composed of the same original values (but in a different order). This reshuffling occurred 10 000 times, generating 10 000 artificial datasets and subsequently 10 000 different t values. The actual t value was then compared with the estimated null distribution of t values, and a percentile was calculated and compared with the two-tailed significance level to determine whether the trend was statistically significant.

3. Results

a. Closed-low climatology

During the 1948–2007 period, an average of 31.6 closed lows occurred per year (Table 1) within the broad rectangular region defined in Fig. 1. A considerable amount of year-to-year variability in the annual count of closed lows exists (Fig. 5). There are four discernible peaks in the 5-yr running mean, occurring in the mid-1950s, mid-1960s, late 1970–early 1980s, and in the late 1990s. Minimums occur during the late 1950s, early 1970s, and the late 1980s–early 1990s. In terms of seasons, the number of closed lows peaks during the spring and falls to a minimum during the summer months (Table 1).

Table 1.

Summary statistics for 500-hPa closed-low frequency (p indicates significance level), by season and compass quadrant. The seasons are defined as winter = December–February, spring = March–May, summer = June–August, and autumn = September–November.

Summary statistics for 500-hPa closed-low frequency (p indicates significance level), by season and compass quadrant. The seasons are defined as winter = December–February, spring = March–May, summer = June–August, and autumn = September–November.
Summary statistics for 500-hPa closed-low frequency (p indicates significance level), by season and compass quadrant. The seasons are defined as winter = December–February, spring = March–May, summer = June–August, and autumn = September–November.
Fig. 5.

Time series of annual closed-low counts (dotted line) with 5-yr running mean (heavy solid line) and linear trend (light solid line) superimposed. Counts represent lows over the entire study domain outlined in Fig. 1.

Fig. 5.

Time series of annual closed-low counts (dotted line) with 5-yr running mean (heavy solid line) and linear trend (light solid line) superimposed. Counts represent lows over the entire study domain outlined in Fig. 1.

Closed-low frequency was also analyzed in four subregions of the broad geographical region defining the northeastern United States (Fig. 1). The broad geographical region was not divided up into equal geographic fourths because experimentation with this method yielded very few precipitation stations in the northeastern (NE) quadrant because a large portion of that quadrant covers parts of Ontario, Quebec, and New Brunswick, Canada, whose precipitation stations were not evaluated in this study. Conversely, the area of the northwestern (NW) quadrant was decreased such that it corresponded to the closed-low genesis region identified by Bell and Bosart (1989). Taking into account the varying geographic area of each quadrant by normalizing the counts by the area of the corresponding subregion, it is seen that closed-low frequency was maximized in the NW quadrant (Table 1).

There was a statistically significant increase in the number of closed lows occurring through time in the eastern quadrants (Fig. 6; Table 1). In the NE quadrant, the annual average number of closed lows increased by 2 (about 10%) over the 60-yr study period. The percentage increase was similar in the southeastern (SE) quadrant. As expected from the lack of a statistically significant increase in the overall annual count of closed lows, there was no corresponding increase in the NW and southwestern (SW) quadrants (Table 1).

Fig. 6.

As in Fig. 5, but for the (a) NE and (b) SE quadrants of the study region.

Fig. 6.

As in Fig. 5, but for the (a) NE and (b) SE quadrants of the study region.

b. Closed-low average precipitation

The average amount of precipitation produced by a closed low in a given year increased during the 60-yr period (Fig. 7) with both the 24-h method (p = 0.9996) and the 48-h method (p = 0.9773) (Table 2). On the basis of the 24-h totals, there is nearly a 40% increase in the amount of average rainfall associated with closed lows; there is an approximately 20% increase on the basis of the 48-h totals. Both methods indicated positive seasonal trends in closed-low precipitation during all seasons with statistically significant trends indicated in winter (Fig. 8). During summer and autumn, the interannual variability of average closed-low precipitation is considerably higher than in the other months (Fig. 8). In some years, average closed-low precipitation in summer exceeded 2.54 cm. This is likely due to the convective nature of the precipitation occurring during this season and fewer overall 500-hPa closed lows. During autumn, a number of dissipating tropical cyclones met the criteria for a closed 500-hPa low and therefore were counted in this study.

Fig. 7.

Time series of average annual precipitation associated with closed lows, using 24-h (light solid lines) and 48-h (dark solid lines) totals.

Fig. 7.

Time series of average annual precipitation associated with closed lows, using 24-h (light solid lines) and 48-h (dark solid lines) totals.

Table 2.

Summary statistics for average precipitation associated with 500-hPa closed lows.

Summary statistics for average precipitation associated with 500-hPa closed lows.
Summary statistics for average precipitation associated with 500-hPa closed lows.
Fig. 8.

Time series of 48-h annual average closed-low precipitation (thin lines) with 5-yr running mean (thick solid lines) and linear trend (straight) lines superimposed for (a) winter, (b) spring, (c) summer, and (d) autumn.

Fig. 8.

Time series of 48-h annual average closed-low precipitation (thin lines) with 5-yr running mean (thick solid lines) and linear trend (straight) lines superimposed for (a) winter, (b) spring, (c) summer, and (d) autumn.

c. Closed-low extreme precipitation

The average annual percentage of stations reporting closed-low 2-day rainfall totals of ≥2.54 cm (and ≥5.08 cm) increased significantly from 1948 to 2007 (Fig. 9; Table 3). From 1948 to 2007, the percentage of stations observing extreme precipitation increased by 50% for stations reporting at least 2.54 cm of precipitation (i.e., the average percentage of stations meeting the criterion increased from 10% to 15%) and nearly doubled when the 5.08-cm threshold was used.

Fig. 9.

Time series of average annual percentage of stations receiving ≥2.54 cm (top plot) and ≥5.08 cm (bottom plot) of precipitation during closed-low events. Linear trend lines are superimposed for each threshold.

Fig. 9.

Time series of average annual percentage of stations receiving ≥2.54 cm (top plot) and ≥5.08 cm (bottom plot) of precipitation during closed-low events. Linear trend lines are superimposed for each threshold.

Table 3.

Summary statistics for percentage of stations with rainfall totals exceeding 2.54 and 5.08 cm during 500-hPa closed lows.

Summary statistics for percentage of stations with rainfall totals exceeding 2.54 and 5.08 cm during 500-hPa closed lows.
Summary statistics for percentage of stations with rainfall totals exceeding 2.54 and 5.08 cm during 500-hPa closed lows.

Seasonal trends, although always increasing, varied in statistical significance (Table 3). Using 48-h accumulations, significant (p > 0.9) trends were present in the average percentage of stations reporting ≥2.54 cm during summer, and in terms of 5.08-cm totals such a trend was present in the spring. During each season, the average percentage of stations receiving 2.54 cm of precipitation from closed lows increased from about 10 in 1948 to nearly 15 in 2007, an indication that such heavy rainfall has become more widespread over time (Fig. 10). This trend is mirrored in the higher 5.08-cm threshold (Fig. 10).

Fig. 10.

As in Fig. 9, but for (a) winter, (b) spring, (c) summer, and (d) autumn. The 5-yr running means are also given (thicker jagged lines).

Fig. 10.

As in Fig. 9, but for (a) winter, (b) spring, (c) summer, and (d) autumn. The 5-yr running means are also given (thicker jagged lines).

4. Discussion

a. Comparing 500-hPa closed-low climatology with previous findings

The seasonal variation of closed-low frequency found in our results is consistent with Bell and Bosart (1989) and Parker et al. (1989), who both found a maximum frequency occurring during spring and a minimum frequency occurring during summer. There were some differences in the geographic distribution of closed lows, however. Because of the presence of the Hudson Bay low, its southeast shift during the late winter and spring, and the tendency for closed lows to develop and propagate around the circumference of the Hudson Bay low (Bell and Bosart 1989), one would expect the NE quadrant to experience the highest closed-low frequency. Instead, the maximum closed-low frequency was found in the NW quadrant, with closed lows being nearly 2 times as frequent there as in the NE quadrant (Table 1). Note that Bell and Bosart (1989) and Parker et al. (1989) evaluated closed-low frequency by counting the number of analysis periods that contained a closed low, whereas our method counted the number of distinct closed lows regardless of life span. Therefore, it is possible that, because of the semipermanent nature of the Hudson Bay low, closed lows that occurred in the NE quadrant simply lasted for a longer period of time and contributed less to the count of distinct closed lows.

The lack of statistically significant findings with regard to temporal trends in closed-low frequency (except in the eastern quadrants) is consistent with Parker et al. (1989), who found a very small decline in closed-low frequency in North America during the 1950–70 period, followed by a very small increase in frequency during 1971–85. Although this is consistent with Fig. 5, given that our study encompassed a much smaller spatial domain, this comparison is not particularly revealing.

b. Comparing precipitation results with previous findings

The increase in precipitation associated with closed lows is consistent with the overall increase in precipitation across the United States over the past few decades, as found by Karl and Knight (1998), Higgins et al. (2000), and Groisman et al. (2004). It is important to note that closed-low precipitation trends may not necessarily concur with overall precipitation trends. At any given station, closed lows may pass within 1.25° on only a few days in a year. Thus, the total precipitation produced by closed lows at a particular station constitutes only a relatively small fraction of the annual precipitation. There are many other atmospheric processes that can produce precipitation and contribute to the overall increasing annual trend.

The greatest increase in precipitation associated with closed lows occurred in the winter months. This result differs qualitatively from the findings of Karl and Knight (1998) who found increases to be magnified during the spring and autumn months but minimized during the winter months. In addition, Higgins et al. (2000) found an increase in precipitation in the northeastern United States during the April–December time period but no increases during January–March. Given that Karl and Knight (1998) included the entire contiguous United States, however, their findings may not necessarily be reflective of trends in the northeastern U.S. region. Palecki et al. (2005), however, did find slight increases in mean storm total precipitation in the northeastern United States during the winter and spring seasons, with no change in the summer and a slight decrease in the autumn. Note that the closed lows analyzed here are just one of the possible mechanisms responsible for the overall precipitation trends assessed in these previous studies.

It is also possible that these differences result from changes in station location (latitude, longitude, and elevation) through time as well as changes in station density through the study period. The correspondence between trends in average precipitation, which are most likely affected by these nonclimatic artifacts, and the percentage of stations reporting rainfall occurrence above a fixed threshold indicates that the sign of the trends is robust across the region, because the percentage value is more resilient to nonclimatic factors—especially station density. Nonetheless, Keim and Russo Fischer (2005) warn that differences in spatial density can result in spurious precipitation trends. Their work, however, is at a much smaller spatial scale (U.S. climate divisions) and considers samples of stations that are almost two orders of magnitude smaller than are used to characterize closed-low precipitation. Furthermore, across the southeastern United States, where elevation differences are minimal, average precipitation trends are affected by differences in station density to a much smaller degree (Allard et al. 2009).

The increase in precipitation associated with closed lows could be attributed to an increase in tropospheric water vapor. Given the rise in global temperatures, it is expected that tropospheric water vapor will increase (Meehl et al. 2007), and there have already been indications that it has. Ross and Elliott (1996) found an increase in surface–500-hPa precipitable water over most of North America of about 2 mm during the 1973–95 period. They also found a corresponding increase in 850-hPa dewpoint and an increase in 850-hPa specific humidity of about 0.5 g kg−1over the eastern United States. Dai (2006) also found statistically significant increases in surface (2 m) relative humidity and surface specific humidity in the eastern United States. Sun et al. (2000) found prominent increases in atmospheric humidity over North America. These increases in tropospheric water vapor were more pronounced in the summer months (Ross and Elliott 1996; Dai 2006). This is despite the fact that specific humidity and temperature were strongly correlated (Dai 2006), and temperature increases were magnified during the winter months. Increased tropospheric moisture does not necessarily translate into increased precipitation, however, without the presence of a lifting mechanism.

There is some indication that the increase in precipitation associated with closed lows contributes to changes in extreme-precipitation events. The increased frequency of stations receiving at least 2.54 cm of precipitation from closed lows is consistent with the findings of Karl and Knight (1998). It is unclear how these results contribute to the decrease in return frequency for specific extreme rainfall amounts identified by DeGaetano (2009), however. The 5.08- and, to a lesser extent, 2.54-cm thresholds are consistent with the definition of extreme rainfall used in the previous studies; such events are typically associated with relatively short (1 or 2 yr) recurrence intervals, however. An examination the frequency of higher-rainfall amounts as well as annual average rainfall maxima from closed lows gave noisy and inclusive results. DeGaetano (2009) found that it is unlikely that the increase in extreme-rainfall events in general was due to increased spatial coverage of extreme rainfall associated with a given storm. If a similar spatial nontrend were to be assumed with overall precipitation occurrence, one could conclude that the precipitation increase can be attributed to an increase in precipitation intensity near closed lows, as opposed to an increase in spatial coverage.

c. Topics of further research

Attributing the increase in precipitation associated with closed lows to a particular mechanism is beyond the scope of this work. It can be theorized that the increase in tropospheric water vapor in the presence of ascent could be a cause, but it would be revealing to examine closed-low environments to determine whether tropospheric water vapor is truly increasing. Changes in instability associated with closed lows could also be examined by determining whether any changes in low–midtropospheric lapse rates have occurred. Dynamical processes could also be responsible for the increase in precipitation, and so an analysis of vorticity advection or upper-air mass divergence trends could be useful. Such analyses may require a spatial resolution that is finer than that of the 2.5° reanalysis data used here.

It would be interesting to examine the origins of 500-hPa closed lows in the northeastern United States. An analysis of the tracks of 500-hPa closed lows could be conducted to determine whether there are more closed lows originating in the subtropics and moving poleward, bringing increased moisture and precipitation with them, or whether the proportion of closed lows resulting from deepening surface lows originating in the lee of the Rocky Mountains, from East Coast winter storms (Hirsch et al. 2001), from the Hudson Bay low, and from other such causes has changed. The frequency of closed lows during the winter and spring seasons may also be related to the frequency of Arctic outbreaks, because the northwesterly flow around the lows can usher in very cold air from Canada and the Arctic latitudes. Such changes could be rooted in oscillatory patterns such as the North Atlantic Oscillation. A more detailed examination of the spatial distribution of precipitation associated with closed lows could also shed light on the cause of the precipitation increase.

5. Conclusions

Closed upper-level lows are one of the synoptic mechanisms associated with extreme rainfall over the northeastern United States. An analysis of the climatology of these systems, the average precipitation associated with them, and the occurrence of heavy rainfall on days with a closed low present indicated

  1. that there have been statistically significant increasing trends in annual closed-low frequency in the eastern United States east of 81°W,

  2. that the average precipitation associated with closed lows showed a statistically significant increase, particularly in winter, and

  3. that the percentage of stations experiencing ≥2.54- and ≥5.08-cm rainfall events in association with closed lows increased by nearly 50%, particularly in winter, which is a statistically significant trend.

Taken together, these findings indicate that changes in the amount and occurrence of extreme rainfall in association with closed lows are consistent with those reported for precipitation from all causes. These overall increases in average and extreme rainfall may be in part due to an increase in the frequency of closed lows, particularly in eastern portions of the study region. In addition, individual storms have become wetter throughout the region, in part contributing to the increasing overall precipitation trends.

Acknowledgments

On the basis of a preliminary version of this paper, the first author (LTN) was named as the recipient of the American Meteorological Society’s 2009 Father James B. Macelwane Award for best undergraduate research paper; he thanks the Society for its recognition and the accompanying stipend. Our thanks are given to Yolanda Roberts for invaluable preliminary research. This work was partially supported by the Northeast Regional Climate Center under NOAA Contract EA133E07CN0090.

REFERENCES

REFERENCES
Allard
,
J.
,
B. D.
Keim
,
J. E.
Chassereau
, and
D.
Sathiaraj
,
2009
:
Spuriously induced precipitation trends in the southeast United States
.
Theor. Appl. Climatol.
,
96
,
173
177
.
Bell
,
G. D.
, and
L. F.
Bosart
,
1989
:
A 15-year climatology of Northern Hemisphere 500 mb closed cyclone and anticyclone centers
.
Mon. Wea. Rev.
,
117
,
2142
2164
.
Collins
,
M. J.
,
2009
:
Evidence for changing flood risk in New England since the late 20th century
.
J. Amer. Water Resour. Assoc.
,
45
,
279
290
.
Dai
,
A.
,
2006
:
Recent climatology, variability, and trends in global surface humidity
.
J. Climate
,
19
,
3589
3606
.
DeGaetano
,
A. T.
,
2009
:
Time-dependent changes in extreme-precipitation return-period amounts in the continental United States
.
J. Appl. Meteor. Climatol.
,
48
,
2086
2099
.
DeGaetano
,
A. T.
, and
D. S.
Wilks
,
2008
:
Radar-guided interpolation of climatological precipitation data
.
Int. J. Climatol.
,
29
,
185
196
.
Emori
,
S.
, and
S. J.
Brown
,
2005
:
Dynamic and thermodynamic changes in mean and extreme precipitation under changed climate
.
Geophys. Res. Lett.
,
32
,
L17706
,
doi:10.1029/2005GL023272
.
Groisman
,
P. Ya.
,
R. W.
Knight
,
T. R.
Karl
,
D. R.
Easterling
,
B.
Sun
, and
J. H.
Lawrimore
,
2004
:
Contemporary changes of the hydrological cycle over the contiguous United States: Trends derived from in situ observations
.
J. Hydrometeor.
,
5
,
64
85
.
Higgins
,
R. W.
,
A.
Leetmaa
,
Y.
Xue
, and
A.
Barnston
,
2000
:
Dominant factors influencing the seasonal predictability of U.S. precipitation and surface air temperature
.
J. Climate
,
13
,
3994
4017
.
Hirsch
,
M. E.
,
A. T.
DeGaetano
, and
S. J.
Colucci
,
2001
:
An East Coast winter storm climatology
.
J. Climate
,
14
,
882
899
.
Kalnay
,
E.
, and
Coauthors
,
1996
:
The NCEP/NCAR 40-Year Reanalysis Project
.
Bull. Amer. Meteor. Soc.
,
77
,
437
471
.
Karl
,
T. R.
, and
R. W.
Knight
,
1998
:
Secular trends of precipitation amount, frequency, and intensity in the United States
.
Bull. Amer. Meteor. Soc.
,
79
,
231
241
.
Karl
,
T. R.
,
G.
Kukla
, and
J.
Gavin
,
1987
:
Recent temperature changes during overcast and clear skies in the United States
.
J. Climate Appl. Meteor.
,
26
,
698
711
.
Keim
,
B. D.
, and
M.
Russo Fischer
,
2005
:
Are there spurious precipitation trends in the United States Climate Division database?
Geophys. Res. Lett.
,
32
,
L04702
,
doi:10.1029/2004GL021985
.
Knight
,
D. B.
, and
R. E.
Davis
,
2009
:
Contribution of tropical cyclones to extreme rainfall events in the southeastern United States
.
J. Geophys. Res.
,
114
,
D23102
,
doi:10.1029/2009JD012511
.
Kunkel
,
K. E.
,
K. A.
Andsager
, and
D. R.
Easterling
,
1999
:
Long-term trends in extreme precipitation events over the conterminous United States and Canada
.
J. Climate
,
12
,
2515
2527
.
Meehl
,
G. A.
,
J. M.
Arblaster
, and
C.
Tebaldi
,
2005
:
Understanding future patterns of increased precipitation intensity in climate model simulations
.
Geophys. Res. Lett.
,
32
,
L18719
,
doi:10.1029/2005GL023680
.
Meehl
,
G. A.
, and
Coauthors
,
2007
:
Global climate projections
.
Climate Change 2007: The Physical Science Basis, S. Solomon et al., Eds., Cambridge University Press, 747–846
.
Nieto
,
R.
, and
Coauthors
,
2005
:
Climatological features of cutoff low systems in the Northern Hemisphere
.
J. Climate
,
18
,
3085
3103
.
Palecki
,
M. A.
,
J. A.
Angel
, and
S. E.
Hollinger
,
2005
:
Storm precipitation in the United States. Part I: Meteorological characteristics
.
J. Appl. Meteor.
,
44
,
933
946
.
Parker
,
S. S.
,
J. T.
Hawes
,
S. J.
Colucci
, and
B. P.
Hayden
,
1989
:
Climatology of 500 mb cyclones and anticyclones, 1950–85
.
Mon. Wea. Rev.
,
117
,
558
570
.
Ross
,
R. J.
, and
W. P.
Elliott
,
1996
:
Tropospheric water vapor climatology and trends over North America: 1973–93
.
J. Climate
,
9
,
3561
3574
.
Ross
,
R. J.
, and
W. P.
Elliott
,
2001
:
Radiosonde-based Northern Hemisphere tropospheric water vapor trends
.
J. Climate
,
14
,
1602
1612
.
Schoof
,
J. T.
,
S. C.
Pryor
, and
J.
Surprenant
,
2010
:
Development of daily precipitation projections for the United States based on probabilistic downscaling
.
J. Geophys. Res.
,
115
,
D13106
,
doi:10.1029/2009JD013030
.
Sun
,
B.
,
P. Y.
Groisman
,
R. S.
Bradley
, and
F. T.
Keimig
,
2000
:
Temporal changes in the observed relationship between cloud cover and surface air temperature
.
J. Climate
,
13
,
4341
4357
.
Trenberth
,
K. E.
,
A.
Dai
,
R. M.
Rasmussen
, and
D. B.
Parsons
,
2003
:
The changing character of precipitation
.
Bull. Amer. Meteor. Soc.
,
84
,
1205
1217
.
Whittaker
,
L. M.
, and
L. H.
Horn
,
1981
:
Geographical and seasonal distribution of North American cyclogenesis, 1958–1977
.
Mon. Wea. Rev.
,
109
,
2312
2322
.
Wilks
,
D. S.
,
2006
.
Statistical Methods in the Atmospheric Sciences
.
Academic Press, 467 pp
.

Footnotes

*

Current affiliation: Department of Atmospheric and Environmental Sciences, University at Albany, State University of New York, Albany, New York.