This study examines the observed interannual variability of the cyclonic activity along the U.S. Pacific coast and quantifies its impact on the characteristics of both the winter total and extreme precipitation in the western United States. A cyclonic activity function (CAF) was derived from a dataset of objectively identified cyclone tracks in 27 winters (1979/80–2005/06). The leading empirical orthogonal function (EOF1) of the CAF was found to be responsible for the EOF1 of the winter precipitation in the western United States, which is a monopole mode centered over the Pacific Northwest and northern California. On the other hand, the EOF2 of the CAF contributes to the EOF2 of the winter precipitation, which indicates that above-normal precipitation in the Pacific Northwest and its immediate inland regions tends to be accompanied by below-normal precipitation in California and the southwestern United States and vice versa. The first two EOFs of CAF (precipitation) account for about 70% (78%) of the total interannual variance of CAF (precipitation). The second EOF modes of both the CAF and precipitation are significantly linked to the ENSO signal on interannual time scales. A composite analysis further reveals that the leading CAF modes increase (decrease) the winter total precipitation by increasing (decreasing) both the number of rainy days per winter and the extremeness of precipitation. The latter was quantified in terms of the 95th percentile of the daily rain rate and the probability of precipitation being heavy given a rainy day. The implications of the leading CAF modes for the water resources and the occurrence of extreme hydrologic events in the western United States, as well as their dynamical linkages to the Pacific storm track and various atmospheric low-frequency modes (i.e., teleconnection patterns), are also discussed.
A large portion of the winter precipitation in the western United States is directly or indirectly related to the extratropical cyclones that develop over the central and eastern North Pacific. Upon approaching the land, these cyclones can bring remarkable amounts of precipitation to the coastal area in a short time period and often cause flooding and mudslides in the Pacific Northwest and California (Ely et al. 1993, 1994). Furthermore, these cyclones can directly contribute to or trigger significant snowfall events in the Sierra Nevada and Rocky Mountains, therefore partly controlling the depth of mountain snowpack that serves as a crucial water resource in spring and summer for the southwestern United States (Serreze et al. 1999). The characteristics of cyclonic activity along the U.S. Pacific coast, including location, frequency, and intensity, are therefore expected to have strong implications for the natural hazards mitigation and water resource management in the western United States. Recent modeling studies have shown that the Pacific storm track, where most West Coast cyclones originate, tends to shift poleward in a warm climate (Hall et al. 1994; Yin 2005). It is of great socioeconomic significance for us to understand how this poleward shift projects onto the cyclonic activity along the West Coast and how the changed pattern of the cyclonic activity affects the trend and variability of winter precipitation in the western United States. To achieve this goal, a first step would be to identify in observations the interannual variation of the cyclonic activity, establish quantitatively its linkage to the characteristics of the winter precipitation, and explore large-scale dynamical processes that control the identified variability.
During the past three decades, numerous studies have examined the climatology of the winter cyclones focusing on either their spatial distributions or their frequency variations across various time scales (e.g., Colucci 1976; Sanders and Gyakum 1980; Parker et al. 1989; Agee 1991; Hodges 1994, 1995, 1996; Sinclair 1994; Lefevre and Nielsen-Gammon 1995; Blender et al. 1997; Serreze et al. 1997; Key and Chan 1999; Hoskins and Hodges 2002). However, despite the feature-tracking methods employed and various datasets of cyclone tracks produced in these studies, few of them discussed the variability of the coastal cyclone characteristics and its impact on the regional precipitation variability. Raphael and Mills (1996) reported that the total precipitation received at ten California stations in two El Niño winters was drastically different because of different effects of local circulation anomalies on cyclone tracks. Raphael and Cheung (1998) further pointed out that the dryness of the second half of the 1980s in California was related to the diminished extent of cyclonic activity over the east Pacific near California. Even though the length of period considered in these two studies is rather short, their results suggest that the variability of coastal cyclones largely determines the variability of the winter precipitation in the western United States on interannual time scales. Such influences are also expected to be modulated by tropical signals such as the El Niño–Southern Oscillation (ENSO).
The relationship between the U.S. precipitation and tropical sea surface temperature anomalies (SSTA), especially those related to ENSO events, has been extensively studied (Ropelewski and Halpert 1986; Andrade and Sellers 1988; Kiladis and Diaz 1989; Schonher and Nicholson 1989; Dracup and Kahya 1994; Cayan 1996; Livezey et al. 1997; Piechota et al. 1997; Dettinger et al. 1998; Cayan et al. 1999; Seager 2007). In particular, the southern (northern) part of the western United States was found to be wetter than normal during El Niño (La Niña) winters (Schonher and Nicholson 1989; Dettinger et al. 1998; Mo and Higgins 1998a). The wet anomalies over the southwestern United States in El Niño winters were also shown to be more pronounced when anomalously cold SSTs occur in the northwest Pacific, which characterizes a warm phase of the Pacific decadal oscillation (PDO) (Latif and Barnett 1994; McCabe and Dettinger 1999) or equivalently a high phase of the North Pacific oscillation (NPO) (Gershunov and Barnett 1998).
The poleward propagation of Rossby waves excited by anomalous convective heating in the tropics is an important route through which tropical processes like ENSO affect extratropical climate. The remote impacts established in such a way are often called “teleconnections” (Trenberth et al. 1988). On interannual time scales, ENSO-induced tropical heating anomalies over the central and eastern Pacific are often linked to teleconnection patterns in the Northern Hemisphere known as Pacific/North American (PNA), tropical Northern Hemisphere (TNH), and western Pacific oscillation (WPO) (Mo and Livezey 1986; Livezey and Mo 1987). While the dominant teleconnection patterns associated with an individual ENSO event depend on the strength, timing, and spatial distribution of the heating anomaly, they all project strongly onto the midlatitude climate and surface weather. For example, the positive phase of PNA, normally associated with Pacific warm episodes (El Niño), features a positive (negative) height anomaly over west North America (southeastern United States). This pattern tends to induce warmer (through large-scale subsidence and advection of warm maritime air) and drier (by blocking cyclone activities) conditions over the northwestern United States (Cayan and Peterson 1989; Leathers et al. 1991; Redmond and Koch 1991; McCabe and Legates 1995). Meanwhile, the positive phase of PNA is also characterized by an eastward and equatorward shift of the subtropical jet over the North Pacific and the Pacific storm track downstream of the jet (Trenberth and Hurrell 1994; Straus and Shukla 1997). Such shifts imply more frequent cyclone activities along the California coast and thus more precipitation in the southwestern United States. Mo and Higgins (1998a,b) further showed that the longitudes of the tropical convection (heating) anomalies strongly affect the western U.S. winter precipitation. Specifically, enhanced convection west (east) of the date line tends to increase (suppress) precipitation over the Pacific Northwest and suppress (increase) precipitation over the southwestern United States. In both cases, the westward (eastward) shift of the subtropical jet and the Pacific storm track play a key role in setting up the precipitation anomalies. In the case of enhanced convection west of the date line, a substantial part of the convection variability came from the intraseasonal bands, which hints at a relationship with the Madden–Julian oscillation (MJO) (Madden and Julian 1994).
This study aims to examine the interannual variability of the cyclonic activity along the U.S. Pacific coast and quantify its impact on the characteristics of both the winter total and extreme precipitation in the western United States. Since precipitation characteristics in the western United States tend to be influenced not only by landfalling cyclones but also by bypassing cyclones that get close enough to the U.S. west coastline, both type of cyclones (coastal cyclones, hereafter) were taken into consideration in this study. The contribution of the precipitation induced by the coastal cyclones to the total winter precipitation in the western United States will first be measured. It also addresses the question of how ENSO modulates the variability of the western U.S. winter precipitation partly through inducing changes in the cyclonic activity over the western United States and the adjacent Pacific Ocean.
The paper is organized as follows: section 2 defines the study region and outlines the data and diagnostic methods used in this study. Section 3 presents the major results of the study. The observed interannual variability of the coastal cyclonic activity will be quantified based on a predefined cyclonic activity function (CAF). The relationship of cyclonic activity over the coastal region with variability of mean and extremeness of precipitation will be established. The ENSO impact on such a relationship, as well as the relationship between major modes of CAF variability and various atmospheric teleconnection patterns, will be discussed. Section 4 summarizes the results and provides concluding remarks.
2. Data and methods
We focus our analysis on the recent 27 winters [December–February (DJF) 1979/80–2005/06]. The cyclone tracking data (Sinclair 1997) used in this study are based on the National Centers for Environmental Prediction–National Center for Atmospheric Research (NCEP–NCAR) reanalysis (Kalnay et al. 1996; Kistler et al. 2001) and consist of track (12-hourly latitude and longitude) and intensity [sea level pressure (SLP) and 1000-mb geostrophic relative vorticity] information of individual cyclones for the period from 1979 to 2006. To accurately quantify the characteristics of the coastal cyclones, the tracks and intensities of cyclones were first linearly interpolated onto 1-h intervals. As mentioned previously, cyclones considered in this study include those making actual landfall over the western United States and those bypassing the west coastline but still producing precipitation over the land. The coastal cyclones are therefore defined as those whose centers moved within a zonal distance of 475 km from the coastline at least once. On selecting the zonal distance threshold, various values ranging between 400 and 600 km have been considered and 475 km gives the most robust linkage of the coastal cyclonic activity with western U.S. winter precipitation. The ratio of the total number of cyclones that make actual landfall to the total number of the coastal cyclones is around 0.42 based on the above definition. For each coastal cyclone identified, the latitude, longitude, SLP, and vorticity at the first track point that gets within 475 km of the coastline were recorded for further analysis.
Based on the latitude and vorticity of the coastal cyclones, a CAF was constructed to quantify the latitudinal variation of the coastal cyclonic activity along the U.S. Pacific coast. Specifically, the CAF is defined as the accumulated intensity of the coastal cyclones (e.g., sum of the vorticities of the coastal cyclones) in a winter in each 1° latitude interval from 26° to 52°N. A moving window with a width of 10° latitude was applied to the CAF to smooth the variation across latitudes. Replacing the accumulated intensity with the total number of the coastal cyclones gives roughly the same latitudinal variation, but the former shows higher correlations with other variables such as the winter total precipitation.
Daily precipitation data for the study period were obtained from the North American Regional Reanalysis (NARR) (Mesinger et al. 2006). NARR is a grid dataset with approximately 32-km horizontal resolution and 45 vertical levels covering North America and adjacent oceans. It is produced by NCEP’s Eta Model and employs gridded analyses of rain gauge precipitation in its data assimilation scheme. The precipitation assimilation results in excellent representation of precipitation from diurnal to annual time scales (Bukovsky and Karoly 2007). The study region of precipitation consists of two rectangular boxes, 29°–41°N–124°–102°W and 41°–49°N–124°–115°W, which are indicated by gray shading in Fig. 1. This study region includes a total of nine states (California, Arizona, New Mexico, Nevada, Utah, Colorado, Oregon, Washington, and Idaho).
Contributions of the cyclone-induced precipitation to the total winter precipitation were estimated by mapping the NARR daily accumulated precipitation onto the area under a cyclone’s precipitation sectors. Such an area is assumed to encompass 4/cosθ to the west, 8/cosθ° to the east, and 7.5° to the north and south from the center of the cyclone (i.e., the track point), where θ is the latitude of the track point. This assumption accounts for all the precipitation within about 444 km to the west, 888 km to the east, and 833 km to the south and north of the cyclone center. It essentially assumes a constant size for cyclones’ precipitation sectors and allows the eastern sectors to be larger.
Empirical orthogonal function (EOF) analyses were applied to the CAF and the DJF-averaged precipitation in 27 winters to identify the principal modes of the interannual variability of the coastal cyclonic activity and the western U.S. winter precipitation, respectively. Linear correlation analysis was used to quantify the relationship among the CAF, winter total precipitation, and their principal modes of variability. To examine the influences of ENSO and other teleconnection patterns on the characteristics of the coastal cyclones and winter precipitation, additional linear analysis was carried out between the principal modes of cyclone/precipitation variability and indices of ENSO [6-month (October–March) averaged Niño-3.4 SST], Pacific/North American, tropical Northern Hemisphere, Arctic Oscillation (AO), and North Pacific pattern (NP). All the indices were obtained from the National Oceanic and Atmospheric Administration (NOAA)/Earth System Research Laboratory (http://www.cdc.noaa.gov/data/climateindices/) except for the TNH index, which came from the Web archive at the NOAA Climate Prediction Center (http://www.cpc.noaa.gov/data/teledoc/telecontents.shtml). Finally, a composite analysis based on the principal modes of the CAF variability was performed to quantify the contribution of the anomalous coastal cyclonic activity to the anomalies of the frequency and intensity of winter precipitation events in the western United States.
3. Results and discussion
a. Variability of the cyclonic activity
The locations of all the coastal cyclones that occurred in the 27 winters (1979/80–2005/06) are indicated by “+” signs in Fig. 1. The histogram on the left shows the frequency distribution of cyclones across different latitudes. Though cyclones have appeared nearly everywhere along the west coast of North America, the peak of frequency is located in northern Canada at 60°N. The southern and northern U.S. borders are another two regions with relatively high frequency of cyclone passages.
Figure 2 shows the four leading EOF modes of the CAF (i.e., EOFCAF1, EOFCAF2, EOFCAF3, and EOFCAF4). They account for 40.7%, 29.6%, 14.4%, and 3.9% of the total interannual variance, respectively. EOFCAF1 is characterized by a monopole with a peak amplitude between 40° and 50°N (Fig. 2a). This mode thus may have great implications for the precipitation variability in the Pacific Northwest and part of northern California. EOFCAF2 (Fig. 2b) on the other hand has a dipole structure with opposite signs between the south (35°–42°N) and north (45°–52°N). It indicates that above-normal cyclonic activity along the northwest coast of the United States tends to be accompanied by below-normal activity along the southwest coast and vice versa. The third mode (Fig. 2c) shows opposite signs between 40°–44°N and 26°–38°N–47°–52°N. Enhanced cyclonic activity over the central west coast of the United States (40°–44°N) is associated with weakened activities over Southern California and the United States–Canada border and vice versa. The fourth mode (Fig. 2d) is characterized by a quadruple with peak amplitudes occurring north of 40°N.
The time series of the principal components (PCs) corresponding to the four EOFs (i.e., PCCAF1, PCCAF2, PCCAF3, and PCCAF4) and their spectra are displayed in Fig. 3. While the accuracy of the spectra estimate is limited by the sample size, the variability of PCCAF1 is found to be concentrated in the periods greater than 7 yr with a secondary peak at the period of about 2.7 yr (Figs. 3a,b). PCCAF2 is characterized by oscillations with periods between 2.5 and 5.5 yr (Figs. 3c,d). The spectra of PCCAF3 show maximum values at periods of 6–10 yr (Figs. 3e,f). The variability of PCCAF4 (Figs. 3g,h) is more high-frequency in nature with a secondary peak at the period of 4–5 yr.
As mentioned in section 2, defining the CAF based on the total number of coastal cyclones instead of the accumulated intensity does not make substantial differences with respect to the results of the EOF and spectral analysis presented above. This suggests that the latitudinal and interannual variability of the coastal cyclonic activity is primarily controlled by the frequency and that the intensity of cyclones only plays a secondary role.
b. Precipitation variability in the western United States and its linkage to the variability of coastal cyclonic activity
Although it is a common belief that winter precipitation in the western United States is closely linked to cyclones that exit the Pacific storm track and propagate over the U.S. Pacific coast, the contribution of cyclone-induced precipitation to total winter precipitation has not been quantitatively evaluated. In this section, we aim to quantify the importance of cyclones to the characteristics of western U.S. winter precipitation from both the climatology and the interannual variability perspective.
Figure 4 shows the ratio of the precipitation induced by the coastal cyclones to the total precipitation in the 27 winters. The value of this ratio increases rapidly from about 0.45 in the inland region to about 0.65 in the coastal area with maximum values at the Pacific Northwest and Southern California exceeding 0.7. On average, this ratio is above 0.6 at most coastal areas indicating that more than 60% of the total winter precipitation is associated with the Pacific cyclones. If only the cyclones that made actual landfalls were considered, this ratio dropped to about 0.3, showing the substantial contribution of those bypassing cyclones to the total winter precipitation. An additional point we want to make here is that the cyclone-tracking algorithm employed in this study tried to pick up only robust (with minimum SLP less than 995 mb) and relatively long-lasting (typically above 2 days) cyclonic features in the circulation. This implies that some weak, transient, but precipitation-producing cyclones were not included in our analysis. The ratio presented here is thus a first-order approximation to and a slight underestimate of the actual value. In fact, Eichler and Higgins (2006) computed a similar quantity composited for various ENSO phases for bypassing and landfalling cyclones and found that the ratio is less than 0.15 in the western United States when using a 9° × 9° box to represent the area with cyclone-induced precipitation. The choice of the 9° × 9° box tends to exclude precipitation from the cold or occluded fronts occurring south of the cyclone center (Eichler and Higgins 2006). In addition, the longitudinal range of 600–800 km that varies with the latitude of the cyclone track might be too limited to capture the whole precipitation system of a Pacific cyclone (note that the longitudinal range considered in this study is about 1332 km). While uncertainty remains in the precise way of tracking cyclones and defining precipitation zones, Fig. 4 demonstrates that the winter precipitation in the coastal states is largely determined by coastal cyclonic activity.
Figures 5a and 5b are the first and second EOF of the DJF-averaged precipitation in our study region as defined in section 2. They explain about 53% and 25% of the total interannual variance, respectively. The first mode (EOFPRECIP1) is characterized by a monopole with the maximum amplitude over the Pacific Northwest and northern California. The second mode (EOFPRECIP2) shows a dipole structure with opposite signs between the southwest and the northwest of the study region. It indicates that above-normal precipitation in the Pacific Northwest and its immediate inland regions tends to be accompanied by below-normal precipitation in California and the southwestern United States and vice versa.
To investigate how the coastal cyclonic activity influences the primary EOF modes of precipitation, a linear correlation analysis was performed between the CAF and the principal components of precipitation EOFs (PCPRECIP1 and PCPRECIP2). At each latitude, a correlation coefficient (CC) between the CAF values in 27 winters and the time series of the precipitation principal component was computed and the resulting latitudinal distribution of the coefficients was plotted in Fig. 6. For PCPRECIP1, positive correlations are found between 38° and 47°N (Fig. 6a), which are statistically significant at the 95% level (indicated by the dashed horizontal line in the figure). This result indicates that above-normal (below normal) cyclonic activity at 38°–47°N significantly contributes to the occurrence of the positive (negative) phase of EOFPRECIP1. Similarly for PCPRECIP2 (Fig. 6b), enhanced (suppressed) cyclonic activity at 37°–41°N and suppressed (enhanced) activity at 48°–52°N tend to induce the positive (negative) phase of EOFPRECIP2.
The effect of coastal cyclonic activity on the interannual variability of the precipitation can also be shown in terms of the spatial distribution of the correlation coefficients between the DJF-averaged precipitation and the principal components of the CAF (Fig. 7). The winter precipitation in the northwestern United States is positively correlated with PCCAF1, not only in the coastal region but also in the far inland areas (Fig. 7a). This result is consistent with the enhanced cyclone activities along the northwest coast of United States, as revealed by EOFCAF1 in Fig. 2a. The correlations between PCCAF2 and the precipitation have a dipole structure with opposite signs over the southwestern and northwestern United States (Fig. 7b), which are again in agreement with the spatial structure of EOFCAF2 (Fig. 2b). Compared to the first and second EOF, the relationship of the precipitation to PCCAF3 and PCCAF4 (Figs. 7c,d) is relatively weak.
Finally, correlation coefficients between the principal components of the precipitation and those of CAF (Table 1) reveal that EOFPRECIP1 and EOFPRECIP2 are significantly related to EOFCAF1 and EOFCAF2 at the 95% level, respectively. Such linkages are consistent with the results presented in Figs. 6 and 7 and demonstrate that the interannual variability of the characteristics of the cyclonic activity along the U.S. Pacific coast largely determines the interannual variability of the winter precipitation in the western United States.
c. Effects of ENSO and major teleconnection patterns on the variability of coastal cyclonic activity and precipitation
Figure 8 shows the relationship between Niño-3.4 SST and the principal components of winter precipitation (PCPRECIP1 and PCPRECIP2). Niño-3.4 is barely linked with PCPRECIP1 (Fig. 8a), but positively correlated with PCPRECIP2 at the 95% level (Fig. 8b). Given the dipole structure of EOFPRECIP2 (Fig. 5b), this indicates that higher Niño-3.4 SST favors an increase (decrease) of precipitation in the southwestern (northwestern) United States.
One very likely route through which ENSO can contribute to EOFPRECIP2 is to modulate the coastal cyclonic activity. To test this hypothesis, linear correlations were computed between Niño-3.4 SST and the principal components of CAF, which is shown in Fig. 9. Among the four principal components, only PCCAF2 is significantly correlated with Niño-3.4 at the 95% level with a correlation coefficient of −0.38. The negative sign indicates that cyclonic activity over the southwestern (northwestern) United States increases (decreases) when El Niño occurs. This feature is consistent with the result of the spectral analysis of the CAF principal components in section 3a. The spectrum of PCCAF2 shown in Fig. 3d indicates that EOFCAF2 is characterized by the period of 2.5–5.5 yr, which is consistent with the dominant periods of the ENSO cycle (Setoh et al. 1999; An and Wang 2000). Noel and Changnon (1998) reported a decrease of cyclone-track density over the northwest and an increase over the southwest when a warm phase of ENSO develops over the tropical Pacific. Our results further quantify such changes and indicate that the preferred cyclone paths shift equatorward in El Niño winters.
The close relationships of Niño-3.4 with the principal components of both CAF (PCCAF2 in Fig. 9b) and winter precipitation (PCPRECIP2 in Fig. 8b) support the speculation that ENSO contributes to the dipole of precipitation variability by inducing changes in the preferred latitudes of cyclone tracks. Assuming the latitudes of the coastal cyclones are tied to the mean latitude of the Pacific storm track, the above results are consistent with previous discoveries that the Pacific storm track tends to shift eastward and equatorward during El Niño winters (Trenberth and Hurrell 1994; Straus and Shukla 1997).
Potential linkages were also explored between the variability of the coastal cyclonic activity and major atmospheric teleconnection patterns that have proven to have an impact on the wintertime weather over the western United States. The corresponding correlation coefficients are shown in Table 2. Occurrences of the positive PNA or negative TNH phase have traditionally been linked to El Niño and both are featured by a negative height anomaly west of North America and a strengthened/eastward-displaced subtropical jet over the North Pacific (Mo and Livezey 1986; Livezey and Mo 1987). Such an anomalous circulation pattern tends to bring more storms to the low latitudes. A relatively weaker correlation (−0.31) was found between PCCAF2 and PNA, in contrast with a more significant one (0.59) between PCCAF2 and TNH. This difference is actually consistent with the fact that PCCAF2 is strongly modulated by ENSO and TNH is a more direct response to ENSO compared to PNA (Mo and Livezey 1986; Livezey and Mo 1987). Negative AO and NP indices both correspond to the deepening of the Aleutian low and a stronger subtropical jet over the North Pacific in winter. This pattern could enhance the eastward propagation of synoptic eddies over the eastern North Pacific and therefore increase cyclonic activity at the Pacific Northwest and northern California. The significant, negative correlations of PCCAF3 with AO (−0.47) and NP (−0.5) are consistent with such circulation anomalies. Given the obvious impacts of teleconnection patterns on regional cyclonic activity, it is equally important to recognize that anomalous cyclonic activity often provides positive feedback to a teleconnection pattern and plays a key role in initiating and maintaining large-scale flow anomalies characteristic of such a teleconnection (Holopainen and Fortelius 1987; Mullen 1987; Lau and Nath 1991; Nakamura et al. 1997).
d. Changes of the characteristics of extreme precipitation in relation to the variability of coastal cyclonic activity
The significant impact of the coastal cyclones on precipitation characteristics shown in section 3b suggests that different phases of the leading EOF modes of the CAF would induce substantial differences of the precipitation pattern in their effective latitudes. Figure 10 shows the differences of the winter mean precipitation between the composite PC+ and PC− winters, which are defined respectively as the average over the five winters with the highest and lowest values of each principal component of CAF from PC1 to PC3. PC4 was excluded because of its relatively small impact on precipitation variability (Fig. 7d). The winters selected to create the composites are shown in Table 3.
For PCCAF1 (Fig. 10a), substantial increases of precipitation at 38°–49°N from PC1− to PC1+ winter are clearly caused by the significant change of the cyclone activity at 40°–50°N (Fig. 2a). Over most part of the northwestern coastal region, the increase is greater than 2.5 mm day−1, which is about 40% of the winter mean daily rain rate in the corresponding area. For PCCAF2 (Fig. 10b), the positive phase of EOFCAF2 is responsible for more precipitation in the Pacific Northwest and less precipitation in the mountain area of California. This dipole structure can also be found in Fig. 7b and is clearly caused by the north–south contrast of the cyclonic activity between 35°–42°N and 45°–52°N (Fig. 2b). The maximum decrease from PC2− to PC2+ winter in northern California is equivalent to a 58% reduction of the winter mean daily rain rate. A relatively weak dipole pattern in the northwest and the southwest occurs between PC3+ and PC3− winter as well (Fig. 10c). Although the differences of the winter mean precipitation for PCCAF3 are very localized, a 110% increase (68% decrease) of the winter mean daily rain rate from PC3− to PC3+ winter takes place in northern California (western Washington).
Figure 10 and the results presented in section 3b clearly indicate that the phases and amplitudes of the leading EOF modes of the CAF contribute significantly to the variation of the winter mean (total) precipitation in the western United States. In addition, one may wonder whether the variation of the mean (total) precipitation is caused by changes in the number of rainy days per winter (i.e., precipitation frequency) and/or changes in the extremeness of precipitation (i.e., precipitation intensity). To answer this question, the number of rainy days per winter and the measures of precipitation extremes (e.g., 95th percentile of the rain rate and probability of precipitation being heavy given a rainy day) were examined and compared for the previously chosen composite winters. A rainy day and a heavy precipitation day are defined as a day during which the total accumulated precipitation is greater than or equal to 1 and 20 mm, respectively (Frei et al. 1998; Cayan et al. 1999).
The average number of rainy days per winter based on the data of the 27 winters is depicted in Fig. 11a. It increases from the southeast to the northwest in our study domain and exceeds 50 days per winter over the Pacific Northwest. Figure 11b shows the observed 95th percentile of the daily rain rate, which measures the intensity of the extreme precipitation. Daily rain rates greater than 20 mm day−1 are found along the northwest coast and the adjacent inland regions including northern California with the maximum values exceeding 60 mm day−1 in some areas. We can also quantify the characteristics of the extreme precipitation through the probability of precipitation being heavy given a rainy day. This probability can be estimated by 1) the ratio of the number of heavy precipitation days to the number of rainy days (hereafter the number probability) and 2) the ratio of the precipitation amount related to heavy precipitation to the total precipitation amount. Since the estimates of the two quantities are very similar, we focus on the number probability in the subsequent analyses. Figure 11c displays the spatial distribution of this probability based on data of the 27 winters. The highest value is greater than 0.25 in several mountain areas. For all the fields shown in Fig. 11, the northwest coast and its immediate inland regions have distinctively higher values compared to the southwest coast and its inland regions, indicating that the area where it rains more frequently also tends to be the area experiencing more heavy precipitation events.
Figure 12 shows the differences of the number of rainy days per winter (Figs. 12a,d,g), the 95th percentile of the daily rain rate (Figs. 12b,e,h), and the number probability between PC1+–PC1−, PC2+–PC2−, and PC3+–PC3− winters (Figs. 12c,f,i). Between PC1+ and PC1−, drastic differences at 38°–49°N (Fig. 12a) are associated with the significant change of the cyclonic activity at 40°–50°N (Fig. 2a). In addition, the spatial distribution of the differences mimics the corresponding correlation pattern between PCCAF1 and the precipitation very well (Fig. 7a). The maximum difference exceeds 14 days (out of 90 days per winter) at certain locations, which is equivalent to 48% of the average number of rainy days per winter at those locations (Fig. 11a). This result suggests that above- (below) normal precipitation in an individual winter is at least partly contributed to by an increase (decrease) of the number of rainy days.
For the 95th percentile of the rain rate (Fig. 12b), significant differences are found along the northwest coast and in its adjacent inland regions. Changes are greater than 16 mm day−1 in northern California and southern Oregon. The spatial pattern is similar to the spatial distribution of the differences of winter mean precipitation between PC1+ and PC1− winters (Fig. 10b). Since the 95th percentile measures the intensity of precipitation extremes, this result demonstrates that the interannual variation of the winter mean (total) precipitation is also caused by the change of the extremeness of precipitation in addition to the change of the precipitation frequency measured by the number of rainy days (Fig. 12a).
For the differences in the probability of precipitation being heavy given a rainy day (Fig. 12c), an increase of the probability in the northwestern United States from PC1− to PC1+ winter is also evident. Over the northwestern coastal area, this increase is equal to about 32% of the corresponding average number probability as shown in Fig. 11c. This further confirms that changes of the DJF-averaged precipitation are often accompanied by consistent changes of the probability of the occurrence of extreme precipitation events.
Figures 12d–f (Figs. 12g–i) are the same as Figs. 12a–c but between PC2+ and PC2− winters (PC3+ and PC3− winters). For PC2+–PC2−, the most distinctive feature is the dipole mode between the northwestern and the southwestern United States, which is consistent with the corresponding patterns in Fig. 7b and Fig. 10b. Negative phases of EOFCAF2 are clearly associated with increases of not only the winter total precipitation but also the extremeness of precipitation in the southwestern United States. The differences found for PC3 (Figs. 12g–i) are also in agreement with the pattern of the mean precipitation change related to EOFCAF3 (Fig. 10c). In central California, while the increase in the number of rainy days (Fig. 12g) is equivalent to 20%–45% of its mean values, changes greater than 100% of the corresponding mean values are found for the 95th percentile of the daily rain rate. The composite analysis based on the different phases of the CAF EOFs suggests that coastal cyclonic activity affects the winter total (DJF averaged) precipitation by inducing changes not only in the number of rainy days per winter but also in the extremeness of precipitation quantified in terms of the 95th percentile of the daily rain rate and in terms of the probability of precipitation being heavy given a rainy day.
4. Summary and concluding remarks
This study quantifies the interannual variability of the cyclonic activity along the U.S. Pacific coast. Its impact on various aspects of the western U.S. winter precipitation has been carefully examined, including the total precipitation amount and the frequency and extremeness of precipitation events. The role of ENSO and other major teleconnection patterns in modulating the cyclonic activity and cyclone-induced precipitation variability has been firmly established. It is also found that the cyclone-induced precipitation accounts for more than 60% of the total winter precipitation in most coastal areas with the maximum ratio being about 74%.
Based on the objectively identified cyclone tracks in 27 winters (1979/80–2005/06), the statistics of cyclonic activity along the west Coast was derived first. A CAF is defined as the accumulated intensity of cyclones that move within 475 km west of the coastline in winter in each 1° latitude interval from 26° to 52°N. The interannual variations of the CAF and the winter precipitation are well coupled through their leading EOFs. The first principal component of the CAF is significantly correlated with that of the precipitation and the structure of EOFCAF1 clearly contributes to the monopole precipitation mode (EOFPRECIP1) with the peak amplitude over the Pacific Northwest and northern California. EOF2 of the winter precipitation has a dipole structure indicating that above-normal precipitation in the Pacific Northwest and its immediate inland regions tends to be accompanied by below-normal precipitation in California and the southwestern United States and vice versa. This precipitation dipole was found to be significantly contributed to by the EOF2 of the CAF.
Correlation analysis using Niño-3.4 SST and winter-averaged indices of several teleconnection patterns suggests that the EOF2s of the CAF and winter precipitation are closely tied to ENSO and TNH variability while the EOF3 of the CAF is associated with the AO and NP signal. The dominant periods of the PCCAF2 are ENSO-like at 2.5–5.5 yr. The significant correlations of ENSO not only with EOFCAF2 but also with EOFPRECIP2 and the linkage between PCCAF2 and PCPRECIP2 indicate that ENSO affects the southwestern U.S. winter precipitation at least partly through modulating the cyclonic activity across the latitudes. Since very similar results were obtained when using the total number of cyclones to define the CAF, the ENSO influence is actually on the preferred cyclone-track locations (latitudes) and the effects on cyclone intensities are secondary. Therefore, the results presented here further demonstrate the following process through which the tropical Pacific SSTA modulate the winter precipitation in the western United States: El Niño (La Niña) events induce an equatorward and eastward (poleward and westward) shift of the Pacific storm track, thereby move the preferred cyclone-track locations southward (northward), and lead to more (less) precipitation in the southwestern United States.
Additional composite analysis reveals that the above- (below) normal precipitation associated with different phases of the CAF EOFs is caused by 1) the increase (decrease) of the total number of rainy days in winter and 2) the increase (decrease) of the extremeness of the precipitation. The positive phases of EOFCAF1 and EOFCAF2 (Figs. 7a,b) are potentially linked to the occurrence of floods and mudslides in the northwestern United States. EOFCAF2 could also contribute substantially to the spring and summer droughts in the southwestern United States because of its close linkage with EOFPRECIP2 (i.e., the precipitation dipole). Given the above results and the poleward shift of the Pacific storm track in a warm climate as predicted by multiple coupled general circulation models (Hall et al. 1994; Yin 2005), such a shift would imply increased frequency and intensity of flooding events in the northwestern United States and possibly more severe droughts in the Southwest. In this regard, an improved understanding of the large-scale circulation features that control characteristics of the coastal cyclones and their sensitivity to the enhanced greenhouse gas forcing will certainly benefit society from perspectives of both natural hazards mitigation and water resource management.
The authors thank Dr. Mark Sinclair for providing the global cyclone-tracking data. We also thank the two anonymous reviewers for comments that led to major improvements of the manuscript. This research was in part supported by NASA Grant NNX09AJ36G.
Corresponding author address: Yi Deng, School of Earth and Atmospheric Sciences, Georgia Institute of Technology, 311 Ferst Drive, Atlanta, GA 30332-0340. Email: firstname.lastname@example.org