The climatology and interannual variability of North American extratropical cyclones are examined using 6-hourly sea level pressure data from the NCEP–NCAR reanalysis for the period 1950–2002 and ECMWF 40-yr Re-Analysis (ERA-40) data from 1971 to 2000. The climatology includes an evaluation of the seasonal frequency and intensity of storms as well as an analysis of extreme event intensity. ENSO variability is evaluated by ENSO phase with emphasis on boreal winter. Results show an enhanced East Coast storm track during El Niño as well as an equatorward shift in storm tracks in the North Pacific for storms generated from both the NCEP–NCAR reanalysis and ERA-40 datasets. Observed precipitation close to a storm’s center is used to determine which phase of the ENSO cycle is associated with the most productive storms and where they occur. During El Niño winters, a precipitation maximum is located east of the Appalachians and is associated with an enhanced East Coast storm track. During La Niña winters, the precipitation maximum shifts to the Ohio Valley and is associated with an enhanced Great Lakes storm track. Along the U.S. west coast, there is a precipitation maximum in the Pacific Northwest during La Niña winters, which is due to a storm track west of Washington State.
The mean location of synoptic-scale extratropical cyclone activity is often referred to as the storm track. The location of the storm track is dictated by the configuration of the large-scale circulation pattern made up of ridges and troughs. Areas immediately downstream of a mean trough tend to have an increased frequency of cyclones while areas immediately downstream of ridges will have a decreased frequency. The intensity of individual surface events is typically related to the amplitude of the upper-level circulation pattern, with more amplified, negatively tilted troughs favoring stronger surface storms. Since extratropical cyclones are a major contributor to seasonal precipitation, their mean location has a direct bearing on many human activities (e.g., agriculture and water management), as well as public safety, life, and property.
Some of the pioneering work on extratropical cyclone tracks was done by Petterssen (1956), who used daily weather maps from 1899 to 1939 to study extratropical cyclones and anticyclones for both boreal winter and summer. Klein (1957) used the same data but generated results for each month and included mean tracks of cyclones and anticyclones. More recently Reitan (1974) calculated the mean extratropical cyclone frequency for North America using data from 1951 to 1970. He found a maximum frequency of storms in the latitude band 40°–50°N, with the most active areas for extratropical cyclone activity in the Gulf of Alaska, the Great Lakes, and over the Gulf Stream off the east coast of the United States. Zishka and Smith (1980) examined storm tracks during January and July and demonstrated an equatorward shift of the storm track in January relative to July, with principal areas of cyclogenesis in the lee of the Rocky Mountains and along the U.S. east coast. They also studied trends in extratropical cyclones for the United States and found an overall decrease in the frequency of storms from 1950 to 1977. Whittaker and Horn (1981) also found a decreasing trend in the frequency of storms for this period in North America, but no significant trend was evident for the Northern Hemisphere. Finally, Sanders and Gyakum (1980) examined extreme events defined as storms whose central pressure falls 1 mb h−1 (normalized for latitude) in 24 h (also known as bombs). In the period from September 1976 to May 1979, they found that the majority of extreme events occurred over the westernmost portions of the Atlantic and Pacific Oceans over or just north of the Gulf Stream and Kuroshio Current.
In addition to the North American studies mentioned above, there have been several notable regional studies. Mather et al. (1964) used daily synoptic weather charts from 1921 to 1962 to classify eight types of East Coast storms based on specific synoptic situations. Examples of these classifications include storms that move northeastward west of the coastal margin and wave developments in the central and western Gulf of Mexico. Colucci (1976) calculated storm-track frequencies during the period November–April 1964–73 over the eastern United States and found that the storm tracks were strongly influenced by the northern edge of the Gulf Stream and were not favored in the Appalachian Mountains except in cases of strong upper-level forcing. Hirsch et al. (2001) developed an East Coast cyclone climatology for October–April using the National Centers for Environmental Prediction–National Center for Atmospheric Research (NCEP–NCAR) reanalysis data (1948, 1951–97) and found that an average of 12 East Coast winter storms occur each year. They also found a small decrease in storms over the period studied although a larger decrease was found for intense storms. Hirsch et al. (2001) also examined interannual variability in East Coast extratropical cyclone activity and found a significant increase in storms during El Niño and a significant decrease during La Niña.
In this study we examine the climatology and interannual variability of storm tracks in general. We exploit observed 6-hourly sea level pressure (SLP) data from the NCEP–NCAR reanalysis (Kalnay et al. 1996) in conjunction with observed daily precipitation data from the Unified Raingauge Dataset (URD) of Higgins et al. (2000) for a 50-yr period (1950–99); the higher-frequency and longer temporal duration of the sea level pressure data allows for the presentation and analysis of a more detailed climatology than that of Klein (1957), Reitan (1974), and Zishka and Smith (1980). Because the data are global, the results allow a broader perspective of interannual variability than that presented by Mather et al. (1964), Colucci (1976), or Hirsch et al. (2001) who focused their analyses on the United States east coast. An analysis of the interannual variability of precipitation associated with storm tracks also gives a perspective on the storm track–precipitation relationship as a function of ENSO phase.
Data and methods used to identify and track extratropical cyclones are given in section 2. A seasonal climatology of extratropical cyclone tracks, including a climatology of extreme events, is presented in section 3. Interannual variability in extratropical cyclone frequency and intensity tied to the phase of the ENSO cycle is examined in section 4. A summary follows in section 5.
2. Data and methods
The storm tracks presented in this study are generated using an algorithm employed at the Climate Diagnostics Center (see information online at http://www.cdc.noaa.gov/map/clim/st_data.html for more details) and described in Serreze (1995) and Serreze et al. (1997). The algorithm is used to track storms in the NCEP–NCAR reanalysis 6-h SLP data for the period 1950–2002 on a 2.5° latitude × 2.5° longitude grid. Storm tracks are identified by locating grid points in which the SLP is less than its surrounding grid points. The detection threshold is flexible, with lower thresholds detecting more storms than higher thresholds. For this study, the threshold is chosen as 1 hPa. To track storms, the tracking software analyzes the position of systems between time steps with a maximum distance threshold between candidate pairings. For the 6-hourly NCEP–NCAR reanalysis data, it is set to 800 km, which means that the maximum distance cyclone can move between steps is 800 km (133 km h−1). Although this is too fast for a normal cyclone, the 800-km threshold allows for “center jumps” (i.e., secondary storms that develop at a distance greater than the average speed of propagation). It also reflects the fact that the input SLP data are gridded and therefore are not continuous in time or space. More information on the software is available online (ftp://ftp.cdc.noaa.gov/Datasets.other/map/storm/README.txtftp). A typical output from the storm-tracks program from the El Niño winter of 1983 is shown in Fig. 1.
For this study, storm-track frequency is defined as the number of storms crossing a 5° latitude × 5° longitude box. While other box sizes were tried, a 5° latitude × 5° longitude box was deemed appropriate for the following reasons. 1) Smaller grids result in boxes being skipped due to the propagation speed of storms. 2) Larger grids capture too many storms resulting in a smoothing out of features. The area of the box is not area normalized as this could lead to frequency bias at higher latitudes (see Paciorek et al. 2002; Changnon et al. 1995). To compute the storm-track frequency, storms were binned into 5° latitude × 5° longitude boxes for four seasons [defined as January–March (JFM), April–June (AMJ), July–September (JAS), and October–December (OND) for winter, spring, summer, and fall, respectively] during the period 1950–2002. To ensure accuracy, individual storms are only included once in the various calculations (e.g., storm frequency). In addition, only storms that propagated at least 5° longitude were included in order to eliminate possible effects of quasi-stationary SLP minima such as thermal lows in the Southwest during boreal summer.
To assess storm intensity, a monthly gridded linear regression analysis was applied to the reanalysis data spanning 1950–2002 and was subtracted from the storm’s intensity. Regression analysis was performed to account for long-term trends in the atmospheric circulation, such as changes to the Arctic Oscillation (AO) or North Atlantic Oscillation (Hurrell 1995, 1996; Thompson et al. 2000) or to the Aleutian low (Hurrell 1996). The procedure for computing the storm’s intensity is as follows: first, the gridded monthly SLP climatology from 1950 to 2002 was linearly detrended; next, the gridded storm’s SLP was obtained by binning the storm’s SLP into 5° latitude × 5° longitude boxes; the gridded intensity of storms was then computed by subtracting the monthly detrended climatology from the gridded storm’s SLP; finally, seasonal SLP storm anomalies were computed by averaging over the 3 months of average storm intensity for a respective season. An effort was also made to analyze the seasonal climatology of extreme storm events. Any storm that had a minimum sea level pressure anomaly (computed from the procedure outlined above) exceeding 1.5 standard deviations below the mean seasonal anomaly was binned into 5° latitude × 5° longitude boxes and seasonally averaged.
To test how robust the storm-track climatology generated from NCEP–NCAR reanalysis data was, seasonal storm-track climatologies were also generated from SLP derived from the European Centre for Medium-Range Weather Forecasts 40-yr Re-Analysis (ERA-40) dataset, which is based on the ECMWF model (information online at http://www.ecmwf.int/products/data/archive/descriptions/e4/). To ensure an accurate comparison, the ERA-40 data used the same temporal sampling (i.e., 4 times daily) and spatial domain (2.5° latitude × 2.5° longitude grid). Since the ERA-40 data does not go as far back as the NCEP–NCAR reanalysis data, the NCEP–NCAR reanalysis frequency climatology was recomputed from 1971 to 2000 for winter and compared with the identical period from ERA-40.
ENSO impacts on storm tracks were assessed using composites by ENSO phase.
Here we define warm (El Niño) and cold (La Niña) events based on the ENSO intensity scale (EIS) developed by Kousky and Higgins (2006, manuscript submitted to Wea. Forecasting, hereafter KH). The EIS is a four-class intensity scale that is calculated by doubling the oceanic Niño index (ONI), defined as the 3-month running mean of SST anomalies for the Niño-3.4 region. In KH it is stated that the advantage of using the EIS system is that it allows a comparison of current ENSO events to previous episodes without regard to external impacts. Storms were composited according to the EIS using the following categories: strong El Niño, weak/moderate El Niño, neutral conditions, weak/moderate La Niña, and strong La Niña cases. Storms were also composited according to the EIS for the strong El Niño and strong La Niña cases for the reanalysis data and the ERA-40 data for winter to assess how similar the two datasets were in duplicating the impact of ENSO on storm tracks.
Finally, the relationships between the storm tracks as defined by sea level pressure minima and precipitation as a function of ENSO phase were examined by mapping the precipitation data of Higgins et al. (2000) onto a 9° latitude × 9° longitude grid using the minimum sea level pressure of a storm as a center point. Since a 9° latitude × 9° longitude grid fails to capture much of the precipitation on the West Coast, the analysis was repeated using a 13° latitude × 9° longitude grid to test how much precipitation is accounted for by cold fronts south of the main storm center. Care was taken to include each precipitation event only once in the composites.
3. Seasonal climatology
The seasonal climatology was obtained using gridded sea level pressure data at a horizontal resolution of 2.5° latitude × 2.5° longitude and binning it into a resolution of 5° latitude × 5° longitude for each season during the period 1950–2002. Mean seasonal storm-track frequency is shown in Fig. 2. Frequency maxima during winter (JFM) occur in the North Pacific, the Great Lakes, and along the Eastern Seaboard (Fig. 2a). The frequency and location of East Coast winter storms agree well with previous studies (e.g., Colucci 1976; Hirsch et al. 2001) that found an average of roughly seven storms per grid box off the northeast coast of the United States. Storm frequency maxima over western Canada and the Rocky Mountains imply that many of the storms that cross the Great Lakes have their origin either from “Alberta clippers” or from lee Rocky cyclogenesis. From winter to summer, the axes of maximum storm frequency shift poleward with the zone of maximum baroclinicity (Figs. 2b and 2c). The storm tracks shift south and east during the transition from summer to autumn as the baroclinic zone begins its equatorward migration.
To check the robustness of the storm-track frequency climatology from the NCEP–NCAR reanalysis, we also examined the storm-track frequency climatology from the ERA-40 data using 4-times-daily SLP data (Figs. 2e–h for winter through fall, respectively). Even though the frequency seasonal climatologies were obtained over a different period than the reanalysis data (1971–2000 for the ERA-40 data and 1950–2002 for the reanalysis data), excellent agreement can be seen between the seasonal storm-track frequency climatologies for the reanalysis data (Figs. 2a–d for JFM, AMJ, JAS, and OND, respectively) and the ERA-40 data (Figs. 2e–h for JFM, AMJ, JAS, and OND, respectively) Specifically, the ERA-40 analysis shows three main storm tracks for winter (JFM, Fig. 2e): the North Pacific, the Great Lakes, and along the East Coast. As with the reanalysis data, the storm-track areas retreat poleward from spring through summer (Figs. 2f and 2g) then begin to move equatorward in the fall (Fig. 2g). To examine differences between the ERA-40 and reanalysis climatologies more closely, a winter (JFM) storm-track frequency climatology was computed for the same time span as the ERA-40 data (1971–2000). The JFM average 1971–2000 frequency climatologies for the NCEP–NCAR reanalysis data (Fig. 3a), the ERA-40 data (Fig. 3b), and the difference (ERA-40 − NCEP–NCAR; Fig. 3c) are shown. It is clear that the ERA-40 JFM climatology is consistent with the reanalysis JFM climatology; the SLP climatology produces slightly less frequent storms in the North Pacific and the Rockies and slightly more over Japan and the eastern half of the United States, although all differences are less than one storm.
The frequency and location of East Coast winter storms obtained by the reanalysis and ERA-40 analyses agree well with previous studies (e.g., Colucci 1976; Hirsch et al. 2001) that found an average of roughly seven storms per grid box off the northeast coast of the United States. However, these results are contrasted by Hodges et al. (2003) who found that storm activity generated from ECMWF 15-yr reanalysis (ERA-15) 850-hPa vorticity was generally greater relative to NCEP–NCAR reanalysis data. Paciorek et al. (2002) generated a winter storm-track climatology by defining a storm as a minimum in the 1000-hPa geopotential height field from NCEP–NCAR reanalysis data; they also found a maximum winter frequency of more than 10 storms in the North Pacific at approximately 50°N from 150°E to 160°W. Although Paciorek et al. found a greater frequency of storms over the North Pacific than the results in this study, they also found that their storm tracks extended to the coast of British Columbia–Washington, which is in good agreement with our results. For the North Atlantic storm track, NCEP, ERA-40, Colucci (1976), Hirsch et al. (2001), and Paciorek et al. (2002) are all in general agreement although Paciorek et al. (2002) has slightly greater maxima (> nine storms) than the NCEP reanalysis data (> seven storms). Since all of these studies use different methodologies and different data resolution [e.g., Hodges et al. (2003) used ERA-15 data], further work is needed to ascertain the best method (or combination of methods) in defining and assessing storm tracks.
In addition to testing the robustness of the storm tracks by comparing two independent datasets (ERA-40 and reanalysis), it is useful to test how sensitive the algorithm is to changing parameters. Two tests were performed using the 1950–2002 reanalysis data: 1) defining a storm using 2 hPa as a minimum and 2) defining a storm using 4 hPa as a minimum. The seasonal frequency climatologies obtained by using a 2-hPa minimum (Figs. 4a–d for JFM, AMJ, JAS, and OND, respectively) yield similar results to the seasonal storm-track frequency climatologies using a 1-hPa difference (refer to Figs. 2a–d). However, using a 4-hPa difference (Figs. 5a–d for JFM, AMJ, JAS, and OND, respectively) results in a considerable reduction in the number of storms; So for the given spatial resolution of the data, the results are robust up to 2 hPa. For finer-resolution data, it is likely that the algorithm will be more sensitive; that is, a smaller pressure difference with respect to surrounding grid points will yield different frequency climatologies.
The seasonal intensities of storms as defined in section 2 are illustrated in Fig. 6, which emphasizes the seasonal changes associated with active storm-track regions in the North Pacific and North Atlantic. As would be expected, the seasonal climatology of SLP associated with storms for winter shows lower pressure associated with the North Pacific and North Atlantic storm tracks with SLP averaging more than 20 hPa below the average daily climatological average (Fig. 6a). From spring through summer (Figs. 6b and 6c), the storms weaken with a minimum of approximately 16 hPa confined to small areas near the Aleutians and south of Iceland. By autumn, the North Pacific and North Atlantic lows become more intense with SLP anomalies of more than 24 hPa below the average south of the Aleutians and Iceland. It is interesting to note that in the Pacific, storms are stronger in fall rather than winter while for the Atlantic, storms are stronger in winter. This is in agreement with Nakamura (1992) who found that baroclinic wave energy peaked in late fall in the North Pacific and in January for the Atlantic.
Since the deepest storms often produce the largest impacts, it is useful to investigate the climatology of extreme storm events. To generate an extreme storm climatology, all storms with sea level pressure anomalies exceeding 1.5 standard deviations below normal were averaged over each grid box and computed for each season (Fig. 7). During winter, the most intense storms are connected with the North Pacific and North Atlantic storm tracks, with a mean SLP up to 35 hPa below the climatological mean in the North Pacific and 45 hPa in the North Atlantic (Fig. 7a). For spring and summer (Figs. 7b and 7c, respectively), the anomalies of the most extreme events weaken with the seasonal cycle with extremes of 30 hPa noted over the Aleutians and south of Iceland by summer (Fig. 7c). Finally, the intensity of extreme events increases in the fall with pressures exceeding 40 hPa below climatological mean southwest of the Aleutians and south of Iceland (Fig. 7d).
4. Variability and ENSO
One of the primary responses to ENSO-induced changes in the patterns of deep tropical convection is a shift of the midlatitude jet streams, especially during the NH winter season. As SSTs increase in the eastern equatorial Pacific, convection that is usually confined to the equatorial western Pacific expands eastward. Compensatory sinking motion poleward of the convection results in an eastward expansion of the subtropical highs across the Pacific equatorward of its normal position. Westerlies on the poleward side of the subtropical high result in an intensified subtropical jet stream equatorward of their normal position. An increased thermal gradient on the poleward side of the NH subtropical anticyclone results in an intensified subtropical jet stream equatorward and to the east of its usual position. These changes alter the midlatitude storm tracks, which in turn contribute to significant temperature and precipitation anomalies in many regions.
Relationships between ENSO and the NH winter storm tracks were examined using precipitation composites keyed to ENSO phase as defined by KH. For convenience, the EIS category and the particular cases in each composite are shown in Table 1. For the strongest El Niño events (Fig. 8a), the greatest frequencies are over the North Pacific and from the southeast coast of the United States to the Maritime Provinces with a weaker track evident from the lee of the Rocky Mountains to the Great Lakes. For weak/moderate El Niño events (Fig. 8b), the storm tracks are similar to the strong El Niño case although there is a slight increase in the number of storms over the northern plains. During ENSO-neutral conditions (Fig. 8c), the frequency of storms decreases over the mid-Atlantic with the maximum frequency now extending from east of New Jersey to Nova Scotia. As La Niña strength increases, the frequency maxima of East Coast storms shifts poleward, the North Pacific storm track extends eastward toward the Pacific Northwest, and the frequency of the Great Lakes storms becomes more pronounced (Figs. 8d and 8e). The difference between strong El Niño events and neutral events and strong La Niña events and neutral events are shown in Fig. 9; analysis of the ERA-40 data was included to assess the ERA-40-induced response to ENSO. For the reanalysis data, strong El Niño was associated with an increased frequency of storms over the north-central Pacific, the Gulf of Mexico, and the east coast of the United States (Fig. 9a). The ERA-40 data show a similar response, although a greater number of storms extend westward into New Mexico relative to the reanalysis (Fig. 9b). Strong La Niña events are associated with a decrease in storm-track frequency for the East Coast and central North Pacific storm track and an increase in storm-track frequency in the northern North Pacific and the northern plains/Great Lakes for the reanalysis data (Fig. 9c). For ERA-40, a strong La Niña was once again similar to the reanalysis (Fig. 9d) although the Great Lakes track extended farther south relative to the reanalysis; it should be noted that the slight differences between the datasets may be due to different sampling periods for the ERA-40 and reanalysis data (1951–2000 and 1971–2000 for reanalysis and ERA-40 data, respectively). Despite the different temporal periods, both datasets produce an excellent response to El Niño. For the reanalysis and ERA-40 results, the increase in storm frequency along the East Coast during strong El Niño episodes is in agreement with Hirsch et al. (2001) and Noel and Changnon (1998), who also found an increase in the frequency of NH winter storms along the East Coast during El Niño and a decrease along the southeast coast during La Niña. Not coincidentally, Janowiak and Bell (1999) found an increase in the number of heavy precipitation days for the East Coast and Patten et al. (2003) found an increase in the frequency of occurrence of heavy snowfall events for the Northeast Corridor and New England consistent with the idea of an enhanced East Coast storm track during El Niño. The decrease (increase) of the Great Lakes storms during El Niño (La Niña) is supported by Angel and Isard (1998) who found a negative correlation between Great Lakes storm-track frequency and the Pacific–North American (PNA) index. Since El Niño winters are often characterized by positive values of the PNA index, the implication is that less (more) Great Lakes storms are produced during El Niño (La Niña). The stronger Great Lakes storm track during La Niña relative to El Niño is also supported by Janowiak and Bell (1999) who indicated that an increased frequency of heavy precipitation events in the Midwest during La Niña was likely associated with a storm track displaced west of the Appalachians. Although the above analysis speaks to the general changes in seasonal storm tracks, it does not address the fact that storm tracks may vary from El Niño to El Niño. For example, Compo et al. (2001) utilized a 60-member ensemble of integrations from the NCEP AGCM using prescribed global SSTs for one El Niño (1987) and one La Niña (1989) and concluded that considerable variability of storm tracks occurred especially over the Atlantic and Europe. Compo et al. (2001) also showed that it is possible that a different picture will emerge if looking at different temporal scales; for example, during El Niño, intraseasonal height anomalies in the North Atlantic were of similar sign, and antisymmetric about the equator, which is the opposite of what is expected on monthly/seasonal time scales.
Relative changes in the storm tracks during El Niño and La Niña are related to changes in the global circulation. It is well known that ENSO causes changes in the strength of the winds at 200 hPa in the subtropics (e.g., Arkin 1982). During El Niño, the subtropical jet stream is enhanced from the Pacific eastward to the subtropical Atlantic. Active storm tracks in the North Pacific and along the east coast of the United States are located on the cyclonic shear side of the subtropical jet wind maxima. During strong La Niña episodes, the subtropical jet stream weakens resulting in an overall poleward shift in the storm tracks.
To determine the relationship between storm tracks and precipitation during phases of the ENSO cycle, the precipitation dataset of Higgins et al. (2000) was employed at a horizontal resolution of 1° latitude × 1° longitude. Precipitation near a storm center was defined for a box measuring 9° latitude × 9° longitude on a resolution of 1° latitude × 1° longitude with the storm center located at the center of the box. The EIS scale of KH was then used to obtain composite mean precipitation associated with the storm tracks by ENSO phase. Kocin and Uccellini (1990) found that for East Coast snowstorms, the heaviest snow accumulations occur in a band 100–300 km to the left and generally parallel to the storm track. The choice of a 9° latitude × 9° longitude box accounts for all precipitation within 285 km of the storm center at 50°N ranging up to 385 km for a storm center at 30°N; so the range is sufficient to capture a large portion of large-scale precipitation close to a storm while being small enough so as not to allow contamination from another separate storm system. Tracking was also tested using a 13° latitude × 9° longitude box (i.e., extending the 9° × 9° box 2° equatorward and 2° poleward) to account for the possibility that precipitation from fronts can extend hundreds of kilometers equatorward from a storm system. A climatology of precipitation associated with storms by ENSO phase was also computed. During strong El Niño events (Fig. 10a), the heaviest precipitation extends from the Gulf Coast states along the Eastern Seaboard including the Northeast Corridor (shaded), consistent with the position of the axis of highest storm-track frequency along the East Coast (contoured data). For weak/moderate El Niño and ENSO neutral events (Figs. 10b and 10c), the precipitation amounts decrease in the southeastern United States and increase over the Ohio Valley as the East Coast storm track shifts northward and the Great Lakes storm track becomes more prominent. During weak/moderate La Niña (Fig. 10d), conditions are similar to neutral although for strong La Niña (Fig. 10e) the Great Lakes storm track is well defined while the East Coast storm track is shifted to the east of New England. The heaviest precipitation close to storms now shifts west of the Appalachians as the East Coast storm track shifts farther north and east relative to weak La Niña and neutral-condition East Coast storm tracks. Note that in this analysis, precipitation along the West Coast is less prominent although an increase in the Pacific Northwest is seen as La Niña increases. As noted by Compo and Sardeshmukh (2004), the sensitivity of storm tracks to ENSO may have a predictability signal that would have a potentially large impact on seasonal precipitation forecasts.
Some questions raised by this analysis are 1) how much of the total observed precipitation occurs close to storms centers and 2) why does the relationship between precipitation and storm tracks weaken on the West Coast? To evaluate how much of the total precipitation is close to storms, composites were generated by EIS phase for JFM with one important exception to the previous analysis: all precipitation without regard to storm tracks was composited (Fig. 11). A comparison of Fig. 11 with precipitation close to storms (Fig. 10) gives an estimate of the fraction of the winter’s precipitation budget that is near surface low pressure systems. The overall spatial pattern of precipitation for strong El Niño events (Fig. 11a) compared to that near storms (Fig. 10a) is similar along the East Coast although greater in magnitude. As one proceeds from strong El Niño through ENSO-neutral conditions to strong La Niña, the precipitation pattern shifts west of the Appalachians similar to the precipitation pattern near storms (Fig. 11e); note, however that the magnitudes are greater.
The fraction of the total NH winter precipitation close to storms (expressed as a percentage) is calculated by taking the precipitation associated with storms and dividing it by the total precipitation for each ENSO phase. For strong El Niño the greatest contribution of precipitation close to storms is along the East Coast and in the Great Lakes region, which coincides with two of the major climatological storm tracks (Fig. 12a). As El Niño weakens, the fraction of total precipitation close to storms decreases along the East Coast as the storm track becomes less prominent (proceeding from Figs. 12b to 12e). The Great Lakes storm track remains a major contributor to the precipitation during weak/moderate La Niña events although the precipitation is less for strong La Niña events likely due to the poleward shift of the jet stream. Note that for all ENSO phases, the percentage of precipitation close to storms drops to less than 15% along the West Coast. A likely explanation for this is that the majority of storms move well north of California (e.g., Fig. 1) suggesting that El Niño–related precipitation in southern California is triggered by either upper-level disturbances associated with the subtropical jet stream or the southern end of cold fronts arching southward from storms moving toward the Gulf of Alaska. Therefore, it is not efficient to evaluate West Coast precipitation by tracking storms defined by their SLP minima.
To test whether cold or occluded fronts are a major contributor to precipitation along the West Coast, precipitation was also tracked using a 13° latitude × 9° longitude box. The precipitation composites (i.e., for each ENSO phase) of the 13° latitude × 9° longitude box was then subtracted from the precipitation composites of the 9° latitude × 9° longitude box in order to analyze precipitation from long, cold fronts occurring equatorward of the storm center (Fig. 13). For all phases of ENSO (Figs. 13a–e), the 13° latitude × 9° longitude box captures more rainfall relative to the 9° latitude × 9° longitude box in Oregon–Washington (approximately 20 mm per winter more). A trend can also be seen as more precipitation is captured southward along the coast of California for increasingly negative ENSO phase; for the strongest La Niñas (Fig. 13e), the 13° latitude × 9° longitude box captures at least 20 mm per winter more precipitation as far south as south-central California while for strong El Niños (Fig. 13a) the 20 mm per winter contour only gets as far south as extreme northern California. A likely reason for this trend is that during La Niña the storm track is closer to the Pacific Northwest (refer to Figs. 9c and 9d, e.g., which show the increase in storms during La Niña across the Pacific Northwest) so that the West Coast is subject to precipitation from the cold/occluded fronts from these systems. Also note that for each ENSO phase, at least 20 mm day−1 more precipitation per winter occurs east of the Mississippi River. A possible reason for this is that the 13° latitude × 9° longitude box is capturing precipitation from frontal systems from the Great Lakes storm track (although precipitation southwest of mature coastal storms could also be contributing).
The climatology computed on a 5° latitude × 5° longitude horizontal grid gives a reasonable distribution of the frequency and intensity of storms with appropriate seasonal shifts. The NCEP–NCAR reanalysis is in excellent agreement with the ERA-40 analysis in the location and magnitude of seasonal frequency maxima. Given the spatial resolution of the data, the algorithm yielded similar results if the SLP minima criterion was increased from 1 to 2 hPa; however, increasing to 4 hPa resulted in a substantial decrease in frequency. Storm intensity followed the seasonal cycle with stronger storms over the North Pacific and North Atlantic during the cold half of the year with weakening and a slight poleward shift noted in the warm half of the year. The climatology for storms that achieved SLP pressure anomalies less than 1.5 standard deviations below the mean seasonal anomaly indicated that the most extreme storm events were located in the North Pacific and North Atlantic. The most intense storms (up to 45 hPa below the climatological mean) occurred in winter for the Atlantic and fall for the Pacific with weaker storms in the summer.
An examination of relationships between ENSO and winter storms showed that for both the reanalysis and ERA-40 data, El Niño was associated with an equatorward shift in the storm track in the North Pacific and an enhanced East Coast storm track. La Niña produced a poleward shift in the storm track (relative to El Niño) in the North Pacific and along the U.S. east coast. La Niña also produced a stronger Midwest/Great Lakes storm track. Precipitation close to winter storm tracks show that East Coast storms during El Niño contribute the largest fraction of the total while West Coast storms contribute the least. In southern California, precipitation is likely associated with other triggers for precipitation such as upper-level jet disturbances, particularly during El Niño. From northern California to Washington State, rainfall from cold/occluded fronts is a major contributor to the precipitation total; this is especially true during La Niña, which features a storm track closer to the Pacific Northwest coastline relative to El Niño. Topographical influences associated with onshore flow are also of particular importance in the Pacific Northwest, which lies to the south of the North Pacific storm track. Any study designed to monitor storm impacts should keep in mind that storms defined by SLP minimum will be most successful in the central and eastern United States, while in the western United States other methods, such as monitoring jet disturbances or pressure gradients, should be employed.
The authors are grateful to Mr. John Janowiak and Drs. Marco Carrera, Hyun-Kyung Kim, and Vernon Kousky for assistance with scientific and technical aspects of this manuscript. The authors also thank the three anonymous reviewers for their comments and suggestions, which greatly improved the manuscript.
Corresponding author address: Dr. Timothy Eichler, RSIS/Climate Prediction Center, NOAA/NWS/NCEP, Camp Springs, MD 20746. Email: email@example.com