Abstract

The interannual variability of tropical cyclone (TC) activity due to El Niño–Southern Oscillation (ENSO) in the main development region of the eastern North Pacific basin has received scant attention. Herein the authors classify years of El Niño, La Niña, and neutral conditions using the multivariate ENSO index (MEI). Storm measurements of the net tropical cyclone activity index and power dissipation index are used to summarize the overall seasonal TC activity and TC intensity between 1971 and 2012. Both measures are found to be statistically dependent on the ENSO phases in the basin’s main development region. However, when the area is longitudinally divided, only the western portion of the development region experienced a significant difference (p < 0.05). Specifically, El Niño years are characterized by more frequent, more intense events compared to La Niña conditions for this subregion. Correlation analyses on the relationships between the MEI and both TC indices demonstrate correlations between ENSO and TC activity and intensity that are statistically significant (p < 0.05) only in the western region. These relationships have the potential to improve the short-term forecast of the local TC activity and intensity on a seasonal basis for public awareness and disaster preparation.

1. Introduction

The El Niño–Southern Oscillation (ENSO) phenomenon is arguably the most dominant control of the interannual climate variability on a global scale (Rasmusson and Carpenter 1982). Atmospheric variations in the ENSO cycle can be monitored by the mean sea level pressure difference between Tahiti and Darwin, Australia, known as the Southern Oscillation index (Ropelewski and Jones 1987). Oceanic anomalies at the Niño-3 (5°N–5°S, 160°E–150°W) and Niño-3.4 (5°N–5°S, 170°–120°W) regions of the equatorial Pacific Ocean during periods of warmer (colder) sea surface temperature are also applied to quantify the extent of El Niño (La Niña) events (Trenberth 1997; Wolter and Timlin 1998). Overall, the El Niño–induced atmospheric instability in the eastern equatorial Pacific Ocean shifts westward when transitioning to the La Niña phase.

Given its feedback on the global atmospheric circulation and its effect on worldwide climate anomalies (Pielke and Landsea 1999; Landsea 2000), ENSO is also critical to the short-term prediction for a range of extreme weather events, including seasonal forecasts of tropical cyclones (TCs) in multiple ocean basins (Camargo et al. 2007). In additional to its role as an important contributor to seasonal TC activity (frequency, intensity, and duration), the impact of ENSO may vary depending on the strength of its signal and the location of the TC development region (Pielke and Landsea 1999). In general, during the La Niña phase, an above-normal TC activity is observed in the southwestern Pacific (Nicholls 1979), the northwestern Pacific (Chan 1985), and North Atlantic (Gray 1984a) TC development basins, while a greater number of storms form during the El Niño phase in the central Pacific basin, west of 140°W–180° (Chu and Wang 1997).

The interannual fluctuation of the ENSO cycle acts as a crucial control to regional storm intensity (Gray 1984b; Bove et al. 1998). For instance, La Niña (El Niño) years are linked to a greater (lower) number of intense TC landfall in the United States during seasons when more (less) Atlantic storms develop (Klotzbach 2011). Since TC potential destruction power can be empirically related and annually aggregated to its intensity (Emanuel 2005), the ENSO cycle plays an important role in determining the potential TC-afflicted societal damages (Chu and Wang 1997; Pielke and Landsea 1999). Because more landfalling storms mature into hurricanes on the eastern United States, the annual average economic damage (Hebert et al. 1997) caused by Atlantic TCs is exacerbated during La Niña years (Pielke and Landsea 1999). However, for the world’s most active tropical storm genesis region when considered on a per unit area and time bases at the eastern North Pacific (ENP) basin (Molinari et al. 2000), the relationship of TC intensity with ENSO has not been explored. Typically, ENP storms dissipate over the open sea. However, when they landfall, they not only affect continental North America but also possess the ability to inflict their strengths on the Hawaiian Islands in the central Pacific basin. For instance, the formation of Iniki (1992) in the ENP basin later reached the category-4 hurricane strength and was accounted for the financial loss of USD 2.5 billion in Hawaii, during the 1991/92 El Niño year (Chu and Wang 1997).

In the Pacific Ocean, where local signatures of El Niño/La Niña conditions are evident, (Nicholls 1992; Trenberth 1997; Wolter and Timlin 1998), strong northwestern Pacific TCs, or typhoons, are suppressed immediately after El Niño years (Chan 1985; Wang and Chan 2002; Chu 2004). Moreover, there is a general consensus that the distribution of TC genesis area in the northwestern Pacific basin has experienced a southeastward drift during the peak typhoon season of El Niño years relative to La Niña years (Chan 1985; Wang and Chan 2002). Thus, El Niño storms tend to travel a greater distance (both westward and northward) prior to experiencing unfavorable conditions to dissipate. Continuing eastward in the North Pacific, we anticipate that the interannual variation of the ENSO cycle would also influence TC activity (frequency, intensity, and duration) in the ENP basin. Specifically, we hypothesize that ENP storm seasons during the El Niño phase will be reflected by a heightened TC activity with more intense TCs, while the La Niña phase will be more subdued.

In the ENP basin where the influence of ENSO on seasonal TC formation and development has remained largely undetermined (Lupo et al. 2008), there is, however, strong evidence of ENSO modification on the longitudinal shift in locations of TC origin and dissipation (Irwin and Davis 1999). Specifically, during El Niño events, more TCs tend to form west of the primary region of TC genesis, while more storms are distributed in the eastern region during the La Niña phase. Though Irwin and Davis (1999) conclude the seasonal total storm count does not deviate between the ENSO phases, Collins (2007) suggests TC frequency of certain storm category is sensitive to the alternation of El Niño and La Niña events only in the western division of the ENP main development region (MDR; 10°–20°N, 85°–140°W). Thus, this unequal east–west distribution of storm genesis contributes to such a difference of spatial sensitivity to environmental conditions (Collins and Mason 2000; Ralph and Gough 2009). In particular, thermodynamic influences at the western subdivision of MDR are argued to enforce such an ENSO-induced difference of TC frequency (Collins 2007, 2010; Klotzbach and Blake 2013).

To date, there remains a gap in the literature that statistically evaluates the relationship of TC climatology in the ENP basin with ENSO. Instead of relying on seasonal storm count alone, an alternative option is to characterize metrics of seasonal TC activity and TC intensity through combining individual measures of storm frequency, intensity, and duration. Though Collins and Roache (2011) integrated these measurements into metrics that better define and quantify seasonal storm activity and intensity, a long-term record of these metrics is needed to establish their relationships with ENSO events. Following the works of Ralph and Gough (2009) and Collins and Mason (2000), our study region will encompass MDR subdivisions to statistically illustrate the spatial sensitivity of regional storm activity as well as the interannual variation due to ENSO influences. Because globally stronger TCs tend to show a stronger relationship with ENSO (Frank and Young 2007), seasonal TC intensity will also be examined as a function of ENSO conditions.

2. Methods

a. Classification of ENSO events

The multivariate ENSO index (MEI) was used to identify ENSO cycles (Wolter and Timlin 1998). Compared to other ENSO indices, MEI is a more holistic approach that reflects the coupled nature of the ocean–atmosphere system through its incorporation of six predictors of sea level pressure, zonal and meridional surface winds, sea surface temperature, surface air temperature, and total cloud fraction. To overlap with the ENP hurricane season, El Niño (La Niña) events are defined by the 10 highest (lowest) years of sliding bimonthly MEI values, averaged from April–May to November–December, with the remaining 22 years defined as neutral years. Since this rank-based approach is less influenced by environmental conditions of a particular season, leading to an asymmetrical number of El Niño and La Niña events, it is widely implemented in not only MEI (Klotzbach 2012), but also anomalies of other sea surface temperature–based ENSO indices (Camargo and Sobel 2005).

b. TC activity and TC intensity

Historical TC track data, including the storm name and its 6-hourly geographic position by longitude and latitude, speed, and surface pressure, are obtained from the northeastern and north-central Pacific hurricane database (HURDAT2) from the National Hurricane Center. As in other works (e.g., Ralph and Gough 2009), TC data after 1971, when the routine use of satellite to monitor storm development, were included. Since the ENP storm formation is spatially unequally distributed, data of TC activity is longitudinally divided at 112°W as the eastern (10°–20°N, 85°–112°W) and western (10°–20°N, 112°–140°W) development regions (EDR and WDR, respectively; Ralph and Gough 2009). During occurrences when EDR storms migrate and establish their peak intensity in WDR, the tabulation of storm count and duration is allocated accordingly, based on the 6-hourly storm-track positions (Collins and Roache 2011).

The seasonal distribution of storm frequency, intensity, and duration during the 42-yr period was taken into account when deriving the two empirically based TC metrics. Based on its surface wind speed, each TC is categorized as a tropical/named storm (18 m s−1), hurricane (33 m s−1), and intense (or major) hurricane (50 m s−1), which corresponds to categories 3, 4, or 5 on a Saffir–Simpson hurricane intensity scale (Simpson 1974). An overall measurement of the net tropical cyclone (NTC; Gray et al. 1994) activity index is defined as

 
formula

where each season’s percentage value is weighted against the entire period mean (1981–2010) and is used for the six measures of seasonal activity: named storms (NS), named storm days (NSD), hurricanes (H), hurricane days (HD), intense hurricanes (IH), and intense hurricane days (IHD). In addition, TC intensity is measured through the power dissipation index (PDI; Emanuel 2005), defined as

 
formula

where the maximum surface wind speed (Vmax) is summed over each storm’s lifetime (τ) every 6 h for all storms of at least tropical storm strength. To make PDI more manageable, seasonal values are accumulated as annual aggregates and displayed at a scale of 107 m3 s−3.

c. Statistical analyses

Seasonal NTC and PDI values for the MDR are stratified into the EDR and WDR (Fig. 1). To account for seasons with no WDR storms, time series for NTC (PDI) are normalized through the square root (natural log) transformation within MDR and its subdivisions. The Mann–Kendall test (Sprent 1989) is utilized to determine if trends of the time series for both storm indices are statistically significant (p < 0.05). The magnitude of the trend is calculated through the Theil–Sen slope (Helsel and Hirsch 1991).

Fig. 1.

The main development region is longitudinally divided into EDR and WDR for eastern North Pacific tropical cyclones. Dots represent areas of major urban centers.

Fig. 1.

The main development region is longitudinally divided into EDR and WDR for eastern North Pacific tropical cyclones. Dots represent areas of major urban centers.

Seasonal NTC and PDI values are statistically compared to reveal any spatial and temporal effects. An analysis of variance (ANOVA) is applied to assess whether there is any statistically significant difference among different groups of MDR storm measurements (NTC and PDI). Seasonal storm measures are compared and grouped according to two conditions: 1) temporal classification of annual ENSO event (El Niño, La Niña, or neutral years) and 2) spatial division of MDR into EDR and WDR. With each storm measurement stratified into groups of ENSO conditions and MDR subdivisions, a two-way ANOVA is applied to test for both individual effects, as well as any interaction from the two factors. An interactive effect from the two factors is examined if relative differences due to ENSO influences are consistent at both EDR and WDR. Since there are three pairings (El Niño–La Niña, El Niño–neutral, and La Niña–neutral) of ENSO phases, a post hoc test using Tukey’s method of multiple comparisons is applied to determine which ENSO pairing(s) is accounted for the different group means (p < 0.05).

A correlation analysis is conducted to measure the degree of relationship for both measures of TC activity and TC intensity with the ENSO index of MEI. A quantile–quantile (Q–Q) plot and the Shapiro–Wilk test are applied to assess the normality of both storm indices prior to the decision of selecting a parametric or nonparametric testing for the correlation analysis. A Q–Q plot demonstrates the fitted distribution of seasonal storm measurements and displays the spread of data deviations from the respective line of normal distribution. Data normality was improved through the best parameter (lambda) estimated from the Box–Cox transformation and quantitatively proven through the Shapiro–Wilk test. A least squares regression is used to correlate MEI values with the two TC measures. A Pearson coefficient of correlation r provides a degree of association between NTC and PDI with MEI. The coefficient of determination r2 is applied to explain the amount of variance that is captured by the linear regression. The correlation analysis is extended when the entire region is subsequently divided into EDR and WDR. Residuals of observed and predicted values from linear regression are assessed for data independence. A Ljung–Box test is applied to examine the null hypothesis of residual independence in a correlogram (autocorrelation plot).

To distinguish the relative contribution of storm frequency, intensity, and duration to PDI, we derive three additional PDI-derived indices. Essentially, the PDI time series is modified so that each of its surrogates only varies with storm frequency, intensity, or duration only. The average intensity (Ii) of a storm during its lifetime (Li) can be represented as Ii = PDIi/Li, where i indices each individual storm. Hence, PDI1, PDI2, and PDI3 can be calculated when the storm intensity, duration, and count are varied, respectively, according to each hurricane season, while the other two terms are averaged across all seasons of the entire 42-yr period (Camargo and Sobel 2005). Correlations calculated with MEI and PDI1, PDI2, and PDI3 can quantitatively determine and separate the contributions of storm intensity, duration, and frequency to PDI.

3. Results

a. Classification of ENSO events

Table 1 highlights the 10 highest (El Niño) and lowest (La Niña) MEI values of bimonthly averages from April–May to November–December, with an additional 22 years classified as neutral years during the 1971–2012 period. Our annual partitioning of ENSO phases is in general agreement with results of Lupo et al. (2008) and Irwin and Davis (1999). In particular, we show the strongest El Niño year (indicated by the highest MEI average) to occur in 1997, when climatological impacts and extreme hydrometeorological events devastated the southwestern United States (Changnon 1999). Nevertheless, the choice of ENSO classification schemes and the length of analysis period could have led to some disagreement in the identification of ENSO events. For instance, using the sea surface temperature anomaly to classify ENSO events, Lupo et al. (2008) deems 1992–94 as neutral years. However, our analysis shows 1992 as a strong El Niño year, associated with the greatest NTC and PDI values reflected by the highest number (24) of storm genesis with the longest duration spent at each (tropical storm, hurricane, and major hurricane) stage. In addition, the extension of our data from that of Lupo et al. (2008) leads to the identification of three La Niña years (2008, 2010, and 2011).

Table 1.

Years from 1971–2012 are classified by ENSO phases based on the averages of the eight sliding bimonthly MEI values from April–May to November–December. The MEI values are provided in parentheses.

Years from 1971–2012 are classified by ENSO phases based on the averages of the eight sliding bimonthly MEI values from April–May to November–December. The MEI values are provided in parentheses.
Years from 1971–2012 are classified by ENSO phases based on the averages of the eight sliding bimonthly MEI values from April–May to November–December. The MEI values are provided in parentheses.

b. Time series of storm activity and intensity

In Fig. 1, the time series for TC measures of frequency, intensity, and duration that define metrics of seasonal storm activity (NTC) and intensity (PDI) are summarized. For comparison purposes, the summary statistics of storm count and lifetime (in days) are binned into storm strengths of tropical/named storm, hurricane, and major hurricane in EDR (Fig. 2) and WDR (Fig. 3). Both EDR and WDR experienced heightened storm activity in terms of the number of tropical storms developed from the early 1980s to 1992 and 1993, respectively, when unusually high proportions of tropical storms strengthened into major hurricanes. However, while it is followed by 1994 and 1999 when no EDR hurricanes grew into major hurricanes, no WDR storms of even NS strength were found in 1996.

Fig. 2.

Time series for the seasonal (a) frequency (TS, H, and MH) and (b) duration (TSD, HD, and MHD) in EDR from 1971 to 2012.

Fig. 2.

Time series for the seasonal (a) frequency (TS, H, and MH) and (b) duration (TSD, HD, and MHD) in EDR from 1971 to 2012.

Fig. 3.

As in Fig. 2, but for WDR.

Fig. 3.

As in Fig. 2, but for WDR.

Overall, the entire region experiences an annual TC count of 15 storms. When MDR is subdivided, while similar seasonal frequencies of NS, H, and MH (Table 2) are shown between EDR and WDR, the origin of a few WDR storms was pertained within WDR boundary. Instead, an average of four WDR storms were initiated in EDR and migrated into WDR, while reaching their maximum intensities. Since these EDR-originated storms establish their peak intensity in WDR, more than half of them strengthened into hurricanes or major hurricanes, thereby greatly extending the length of storm duration during all three storm stages.

Table 2.

The TC activity in the ENP basin in the MDR, EDR, and WDR from 1971 to 2012, while seasonal averages from 1981 to 2010 are incorporated into the derivation of NTC.

The TC activity in the ENP basin in the MDR, EDR, and WDR from 1971 to 2012, while seasonal averages from 1981 to 2010 are incorporated into the derivation of NTC.
The TC activity in the ENP basin in the MDR, EDR, and WDR from 1971 to 2012, while seasonal averages from 1981 to 2010 are incorporated into the derivation of NTC.

The time series of the seasonal overall TC activity (NTC; Fig. 4) and TC intensity (PDI; Fig. 5) are displayed from 1971 to 2012. Although the Mann–Kendall test confirms the Theil–Sen slope values of MDR storm measurements to be statistically insignificant, negative trends of NTC and PDI imply that there are decreasing number of storms with long duration and fewer intense storms of long duration, respectively (Figs. 4a and 5a). Such a recent decrease in the ENP storm development is consistent with findings from Schultz (2007), Lupo et al. (2008), and Wu et al. (2008), but opposite from that experienced in adjacent regions of the Atlantic (Webster et al. 2005; Collins 2010) basin, marked by consecutive (2004 and 2005) peaks of Atlantic hurricane seasons. Overall, this continuous increase of TC activity (Steenhof and Gough 2008) and TC intensity (Wu et al. 2008) in the Atlantic basin over a long time record is widely known to be enhanced by local environmental conditions of thermodynamic and dynamic factors (Emanuel 2005; Goldenberg and Shapiro 1996; Goldenberg et al. 2001).

Fig. 4.

Time series charts for NTC at (a) MDR with horizontal dashed lines showing values at the 25th and 75th percentiles and when TC development region is subdivided into (b) EDR and WDR during 1971–2012. Dashed vertical lines are colored to distinguish El Niño (red), La Niña (blue), and neutral (gray) years, to be consistent with (a).

Fig. 4.

Time series charts for NTC at (a) MDR with horizontal dashed lines showing values at the 25th and 75th percentiles and when TC development region is subdivided into (b) EDR and WDR during 1971–2012. Dashed vertical lines are colored to distinguish El Niño (red), La Niña (blue), and neutral (gray) years, to be consistent with (a).

Fig. 5.

As in Fig. 4, but for PDI (scale of 107 m3 s−3).

Fig. 5.

As in Fig. 4, but for PDI (scale of 107 m3 s−3).

When MDR is subdivided at 112°W (Ralph and Gough 2009), while negative trends of both NTC (Fig. 4b) and PDI (Fig. 5b) values are still maintained in WDR, EDR storm measures have decreased at a slower pace for NTC, with PDI experiencing virtually no change through the 42-yr period. Though trends of the regional contrast are determined as statistically insignificant, there is growing evidence that WDR storm activity is regulated by environmental controls that are different from that in EDR (Collins and Mason 2000; Ralph and Gough 2009).

One such large-scale ocean–atmosphere climate phenomenon that can affect the interannual variability of both NTC (Fig. 4) and PDI (Fig. 5) is the influence of ENSO events. In general, El Niño years are identified with greater values of storm indices derived from seasonal TC activity. On the other hand, La Niña years are associated with some of the least active TC seasons. From Table 3, when seasonal mean values of six NTC measures are binned to the appropriate ENSO events, both the frequency and duration measures are consistently highest during El Niño years, follow by neutral and La Niña years. However, as there is one year (1996) without any WDR storm formation, direct spatial comparison between the six NTC measures under ENSO influences would violate fundamental assumptions (e.g., data normality) of parametric statistics. Presumably, an ENSO-induced difference is greater in WDR where drastic meteorological shifts resonate more with ENSO phase changes (Collins 2007). Although such an assessment of ENSO influences on ENP storms is only qualitative, statistical approaches attributing ENSO events to the regional TC activity and TC intensity using metrics that include storm frequency, maximum sustained wind speed, and duration will follow in the next subsection.

Table 3.

Mean seasonal (1971–2012) values of six NTC measures subdivided into EDR and WDR during three ENSO conditions.

Mean seasonal (1971–2012) values of six NTC measures subdivided into EDR and WDR during three ENSO conditions.
Mean seasonal (1971–2012) values of six NTC measures subdivided into EDR and WDR during three ENSO conditions.

c. Statistical analyses

1) Two-way analysis of variance

The mean differences of the NTC and PDI are compared using a two-way ANOVA under two factors: temporal classification of annual ENSO conditions and spatial division of MDR (Tables 4 and 5). The F test confirms that temporal differences of NTC and PDI can be attributed to ENSO variation though we do not know under which ENSO conditions are both indices different. Since the F test alone cannot detect exactly which ENSO pairing(s) is attributed to a significant difference between both storm measures, Tukey’s method of multiple comparisons identifies NTC and PDI to be only statistically different for pairings of El Niño–La Niña and El Niño–neutral comparisons at p < 0.05 (Table 4). In previous work, though more El Niño storms became intense hurricanes, the analysis of monthly TC statistics does not yield a significantly greater TC activity during El Niño years (Lupo et al. 2008). In contrast, our results demonstrate that both storm measurements are significantly different under both ENSO pairings, which are also acquainted to influence the mean genesis location of ENP storms (Irwin and Davis 1999).

Table 4.

A two-way analysis of variance is tested for effects of ENSO and regional division on the NTC activity index (%).

A two-way analysis of variance is tested for effects of ENSO and regional division on the NTC activity index (%).
A two-way analysis of variance is tested for effects of ENSO and regional division on the NTC activity index (%).
Table 5.

A two-way analysis of variance is tested for effects of ENSO and regional division on the PDI (107 m3 s−3).

A two-way analysis of variance is tested for effects of ENSO and regional division on the PDI (107 m3 s−3).
A two-way analysis of variance is tested for effects of ENSO and regional division on the PDI (107 m3 s−3).

The second main effect of regional difference demonstrates that only seasonal PDI values are statistically different (p < 0.05) between EDR and WDR. A follow-up regional comparison of TC measures using Tukey’s method is shown only as a confirmatory analysis since there are only two levels (EDR and WDR) of MDR subdivisions. Regional variations of TC intensity (PDI) due to ENSO events further confirm that MDR should be viewed as two distinct development regions (Collins and Mason 2000; Ralph and Gough 2009). While the average WDR storm intensity (PDI) is noticeably higher during El Niño years, the mean difference of the overall TC activity (NTC) is not detected (Table 4), possibly because of the derivation of seasonal NTC values. The derivation of the NTC index is sensitive to the seasonal variability of relative measures of all TC components of the index to the respective 1981–2010 climatological means (Table 2). Because the seasonal (1981–2010) averages of storm duration measures are distinctly higher in WDR, its seasonal NTC values during any given year can be comparably similar, statistically indifferent, to EDR values. For instance, 2009 saw the WDR storm duration (95 days) close to be 3 times more than that of EDR (36 days), with a comparable difference for the total storm count. However, owing to the fact that many EDR-originated storms spend the majority of lifetimes in WDR, because long-term averages of all three NTC duration measures in WDR are substantially greater, the regional difference of NTC values in 2009 is virtually the same. As such, even if there is a substantial spatial difference of any individual NTC component in any given year, NTC is collectively bounded by all six NTC parameters’ 30-yr averages (Table 2).

Since there is a combination of effects considered for statistical comparison, an interaction effect of the regional and ENSO-induced temporal (ENSO region; Tables 4 and 5) factors on both NTC and PDI values is assessed. An interactive outcome occurs when rankings of relative NTC and PDI values based on ENSO conditions shift in ranks between regions. Not only do our results show that the average measures of both storm indices are greatest during El Niño years, followed by neutral and La Niña years (Figs. 6a and 6b), but also, both MDR subdivisions maintained such relative ranks of storm metrics due to ENSO modification. This consistency of the result in the F test (p > 0.05) demonstrates the absence of an interaction effect of a combined ENSO region factor across NTC (Table 4) and PDI (Table 5). However, Tukey’s multiple comparisons of the interaction effect demonstrate significant differences between NTC and PDI resulting from different ENSO comparisons that are not equally distributed within MDR subdivisions. In particular, the differences of NTC and PDI values within the ENSO pairing of El Niño–La Niña are maintained only in WDR.

Fig. 6.

(a) The NTC activity index and (b) PDI (scale of 107m3 s−3) under 10 El Niño, La Niña, and neutral years are summarized as boxplots for MDR, WDR, and EDR. Green diamonds represent NTC and PDI averages at each ENSO condition. Results of statistical evaluations of NTC and PDI differences are as in Tables 4 and 5.

Fig. 6.

(a) The NTC activity index and (b) PDI (scale of 107m3 s−3) under 10 El Niño, La Niña, and neutral years are summarized as boxplots for MDR, WDR, and EDR. Green diamonds represent NTC and PDI averages at each ENSO condition. Results of statistical evaluations of NTC and PDI differences are as in Tables 4 and 5.

2) Correlation analysis

Relatively greater NTC and PDI values during El Niño years, followed by neutral and La Niña years suggest that MEI values are positively linked to both indices. However, the use of the parametric correlation analysis is deemed appropriate only when both storm metrics have been normalized. When TC data are binned into EDR and WDR, the data distribution of each storm measure varies. Visual inspections of Q–Q plots show that an absence of WDR storm activity may have caused nonnormal distributions of both NTC (Fig. 7a) and PDI (Fig. 7b) indices. In particular, WDR storm development during 1996 went completely dormant with no storms generated west of 112°W. After both datasets were transformed, distributions of both NTC (Fig. 7c) and PDI (Fig. 7d) are closer to the line of normality. Under the null hypothesis that the sampled data come from normal distributions, a subsequent Shapiro–Wilk normality test was applied to test against the assumption of data normality. The results of the Shapiro–Wilk test calculate a W statistic that is used to calculate the correlation of the sampled data after they have been ordered and standardized and what the samples would have been if they were drawn from a normal distribution and ordered. Significance of the Shapiro–Wilk test subjectively determines if both WDR storm metrics were normalized after data transformations.

Fig. 7.

The (Q–Q) plots for (a) NTC and (b) PDI (scale of 107 m3 s−3) values in WDR have lines of normality passing through the first and third quartiles. (c),(d) Data after transformation are presented. The Shapiro–Wilk test statistic (W), with its p value, is provided before and after data transformations.

Fig. 7.

The (Q–Q) plots for (a) NTC and (b) PDI (scale of 107 m3 s−3) values in WDR have lines of normality passing through the first and third quartiles. (c),(d) Data after transformation are presented. The Shapiro–Wilk test statistic (W), with its p value, is provided before and after data transformations.

A parametric testing of correlation analysis is used to highlight the extent of NTC (Fig. 8) and PDI (Fig. 9) relationships with MEI. Time series of residuals (observed − correlated) their correlograms for the WDR storm metrics are plotted from 1971 to 2012 (Fig. 10). To test for the presence of serial independence of the residuals, subsequent Ljung–Box test statistics (chi-squared values) show that there is no statistical evidence (p > 0.05) for nonzero autocorrelations. Positive associations for both MDR storm measurements with MEI are determined statistically significant for PDI, not NTC (Table 6). However, when MDR is subdivided, such a positive linear relationship is only significant in WDR (p < 0.05). The presence of such statistical significance confirms that the ENSO influence on TC activity is most pronounced in WDR, even though more EDR storms are generated during the local hurricane season. A relatively higher r2 in WDR indicates a greater NTC and PDI variability that is explained by MEI variability. Residuals (observed − predicted) for the WDR storm metrics are plotted in Fig. 10; while, subsequent Ljung–Box test statistics (chi-squared values) show that there is no statistical evidence (p > 0.05) for nonzero autocorrelations.

Fig. 8.

Linear regressions of MEI values with data transformed for NTC at (a) MDR, (b) EDR, and (c) WDR under three ENSO conditions during 1971–2012. The Pearson coefficient of correlation (r) and correlation of variation (r2) are provided in Table 6.

Fig. 8.

Linear regressions of MEI values with data transformed for NTC at (a) MDR, (b) EDR, and (c) WDR under three ENSO conditions during 1971–2012. The Pearson coefficient of correlation (r) and correlation of variation (r2) are provided in Table 6.

Fig. 9.

As in Fig. 8, but for PDI (scale of 107 m3 s−3).

Fig. 9.

As in Fig. 8, but for PDI (scale of 107 m3 s−3).

Fig. 10.

(left) Time series of residuals (observed − predicted) NTC values and (right) correlograms (autocorrelation plots) are evaluated with the Ljung–Box test statistic (chi-squared value), with its p value, based on the first 20 lags in WDR for (a) NTC and (b) PDI (scale of 107 m3 s−3) during 1971–2012.

Fig. 10.

(left) Time series of residuals (observed − predicted) NTC values and (right) correlograms (autocorrelation plots) are evaluated with the Ljung–Box test statistic (chi-squared value), with its p value, based on the first 20 lags in WDR for (a) NTC and (b) PDI (scale of 107 m3 s−3) during 1971–2012.

Table 6.

Correlations of NTC and PDI with MEI are determined as the Pearson coefficient of correlation (r) and the correlation of determination (r2) for Figs. 8 and 9. Boldface values are statistically significant at p < 0.05.

Correlations of NTC and PDI with MEI are determined as the Pearson coefficient of correlation (r) and the correlation of determination (r2) for Figs. 8 and 9. Boldface values are statistically significant at p < 0.05.
Correlations of NTC and PDI with MEI are determined as the Pearson coefficient of correlation (r) and the correlation of determination (r2) for Figs. 8 and 9. Boldface values are statistically significant at p < 0.05.

4. Discussion

The influence of ENSO on different parameters of MDR storm activity at the ENP is statistically addressed in this study. Previously, a lack of such statistical evidence in the basin could be attributed to inconsistent classification schemes in distinguishing ENSO phases of El Niño, La Niña, and neutral conditions, in addition to the absence of reliable TC data on a longer time scale (Collins 2007; Schultz 2007; Lupo et al. 2008). Though maximum NTC and PDI values in 1992 could be related to a greater midlevel moist static energy during the peak (July–September) hurricane season (Wu and Chu 2007), the recent 2007, 2008, and 2010 La Niña years are marked as the three most meager TC years of the entire data analysis (1971–2012). NTC and PDI values for these three years were well below seasonal averages of 97.56% and 1.20 × 107 m3 s−3, respectively, over 10 La Niña years. In terms of the same storm measures of TC activity and intensity, when MDR is subdivided, 2011 is found to exhibit the second least-active TC season of all 10 La Niña years in WDR.

Alternative considerations of MDR divisions from Irwin and Davis (1999) and Collins and Mason (2000) also have attributed to the detection of the difference of regional TC sensitivity to ENSO oscillation. The MDR division at 112°W depicts a substantially greater TC intensity in WDR (Ralph and Gough 2009), where many EDR storms migrate and achieve their peak intensities prior to dissipation. While there are more storms originated from EDR, its overall TC activity and TC intensity are less influenced by environmental changes; rather, it is WDR storms that are more responsive to ENSO-induced environmental influences (Collins and Mason 2000). Such a regional difference of storm sensitivity to changes of the external environment is also consistent with that of the original longitudinal division of MDR storm frequency of at least tropical storm strength into eastern and western geographical boundaries (Collins and Mason 2000).

Another important aspect is revealed when focusing on the correlation strengths of MEI with NTC and PDI in MDR and within its subdivision. Overall, PDI is more sensitive to ENSO influences than NTC (Table 6). Among factors of storm intensity (PDI1), duration (PDI2), and count (PDI3) attributed to PDI variability (Fig. 11), only storm intensity (PDI1) demonstrates a significant (p < 0.05) effect on the correlation of PDI with ENSO signal (Table 7). Such a result is contrary to northwestern Pacific storms where Camargo and Sobel (2005) found that only the duration variability attributes significantly to the ENSO signal on PDI. However, when MDR is subdivided, seasonal variations of all three storm factors are significantly sensitive to the ENSO signal on WDR storm intensity, while their influences are little to minimal in EDR. As the derivation of PDI is directly related to the total storm count, such associations between (PDI1 versus MEI) show comparable strength with the relationship of PDI and MEI (Table 6). Overall, the subdued r2 of ENSO with PDI, and also NTC, suggests other local environmental forcings are involved at explaining the annual variation of seasonal TC measures.

Fig. 11.

Box-and-whisker plot summaries for data distributions of (a) PDI1, (b) PDI2, and (c) PDI3 (scale of 107 m3 s−3) in MDR, EDR, and WDR. Gray diamonds represent the regional averages.

Fig. 11.

Box-and-whisker plot summaries for data distributions of (a) PDI1, (b) PDI2, and (c) PDI3 (scale of 107 m3 s−3) in MDR, EDR, and WDR. Gray diamonds represent the regional averages.

Table 7.

Correlations of PDI1, PDI2, and PDI3 with MEI (May–November) values in MDR, EDR, and WDR from 1971–2012. Boldface values are statistically significant at p < 0.05.

Correlations of PDI1, PDI2, and PDI3 with MEI (May–November) values in MDR, EDR, and WDR from 1971–2012. Boldface values are statistically significant at p < 0.05.
Correlations of PDI1, PDI2, and PDI3 with MEI (May–November) values in MDR, EDR, and WDR from 1971–2012. Boldface values are statistically significant at p < 0.05.

Changes to sea surface temperature due to ENSO modification can be linked to the atmospheric alteration of convective mechanisms for fueling storm development (Collins 2007; Klotzbach and Blake 2013). One of the crucial indicators in determining ENSO signal (i.e., sea surface temperature) has been well known for its local effect on the ENP storms (Ralph and Gough 2009). Since WDR ocean temperatures are closer to the 26.5°C threshold for TC formation (Gray 1968), a warm anomaly arising from ENSO changes may trigger the genesis of more storms and vice versa. In particular, fluctuating sea surface temperature could reinforce changes to the availability of atmospheric moisture and dictate the seasonal variability of WDR storm activity (Collins and Roache 2011). While the seasonal sea surface temperature in EDR is already well above the critical threshold for storm formation (Ralph and Gough 2009), fluctuation of the sea surface temperature–dependent MEI may be less crucial to the region’s overall TC activity and storm intensification. These thermodynamic controls offer explanations to why dramatic differences of the two indices appear only when MDR is subdivided.

The fact that the ENP storm activity and intensity in WDR are more sensitive to changes of ENSO conditions is geographically consistent with neighboring storm development regions. In the Southern Hemisphere, the strength of the correlation between southwestern Pacific storm frequency and the Southern Oscillation index gradually shifts in direction westward from the 170°E longitude of boundary division (Basher and Zheng 1995). This means that in contrast to the ENP basin, more South Pacific storms develop during the La Niña phase. Similarly, at the adjacent northwestern Pacific basin, there are more La Niña storms that reached the strength of a typhoon, equivalent to a hurricane, at its easternmost boundary near the International Date Line (Chan 1985; Wang and Chan 2002). Such a close proximity to the neighboring WDR could result into a greater storm development, which is directly observed through the eastward extension of the monsoon trough crossing the central Pacific boundary during extreme El Niño years (Clark and Chu 2002).

Apart from its direct influences on moist static stability (Malkus and Riehl 1960), ENSO is also indirectly acquainted with changes in the wind direction of the tropospheric profile. As demonstrated in the North Atlantic basin, ENSO is known to control wind shear in other TC basins where such a dynamical factor is primarily responsible for TC intensity (Gray 1984a; Jones and Thorncroft 1998; Landsea 2000; Goldenberg et al. 2001) and is noted to be highly correlated with the primary ENP storm genesis region (Table 6). Because low wind shear is observed over much of the MDR in 1992 during such a strong El Niño event, it could also play an important role at reinforcing WDR convection through the reduction of wind shear by enhancing the upper-level westerly wind and the lower-level easterly wind (Jones and Thorncroft 1998). When wind shear and other dynamical factors are included in the seasonal genesis parameter (Gray 1977), a greater number of El Niño storms is observed to coexist with higher dynamical potential values than during La Niña conditions (Clark and Chu 2002).

5. Conclusions

The longitudinal division of MDR has greatly facilitated our understanding of environmental influences at the ENP basin. We have statistically evaluated and compared the impact of ENSO on TC activity and intensity between the neighboring subdivisions of EDR and WDR. In combination with the classification of years of El Niño, La Niña, and neutral conditions, the spatial and temporal influences on the ENP storm activity and intensity are quantitatively resolved over a 42-yr period. Although the physical forcings accompanying ENSO phase changes are not directly addressed, previous studies have identified a combination of local thermodynamic and dynamic factors in speculating El Niño–induced shifts of more intense storms with longer lifetimes.

Measures of the overall seasonal TC activity and intensity are expressed empirically as indices of NTC and PDI, respectively. Accounting for the nonnormality of storm data in WDR where fewer storms are observed seasonally, both TC indices have been transformed and tested for statistical differences between every pairwise combination of three ENSO phases. Overall, both indices are only proven statistically different between El Niño and La Niña years. When MDR is subdivided, such a contrast is only maintained in WDR. This difference of the regional sensitivity to ENSO is supported by results of the correlation analysis, which demonstrates the strength of correlations in WDR to be stronger for the seasonal TC activity than TC intensity between 1971 and 2012. Statistical comparisons of regional TC indices at different ENSO conditions validate previous findings that the effects of environmental influences over ENP storms would have been overlooked if MDR was not subdivided (Collins and Mason 2000; Ralph and Gough 2009).

Between storm parameters of frequency, intensity, and duration, only the seasonal variability of TC intensity contributes significantly to the ENSO signal on PDI. However, this independent analysis on the influence of PDI due to ENSO-induced storm measurements is also subject to regional variation. When comparing between MDR subdivisions, PDI is found to be significantly sensitive to seasonal fluctuations for all WDR storm parameters, with storm count being relatively more important.

The findings generally underscore the societal importance of ENSO on TC activity and intensity. Although most of the ENP storm development initiated in EDR, many storms did not reach maximum intensity until entering WDR. During El Niño years, there are more WDR storms with higher intensity and longer lifetimes that develop than during La Niña years. A direct consequence is that as the storms are transitioned from EDR; ENSO influences on seasonal WDR environmental conditions could fuel further storm development and prolong storm tracks affecting islands in the central Pacific basin. Hence, the determination and incorporation of ENSO as a predictor for short-term, seasonal storm forecasts could prove to be a practical tool that better anticipates the seasonal outlook of WDR storm activity and intensity and the potential TC-inflicted damages upon landfalls. Since Atlantic storm damage and the relationship with the oscillation of ENSO phases have already been well documented (Hebert et al. 1997; Pielke and Landsea 1999), future work on such a data archive should include additional input on the relationship of ENSO events with the societal cost associated with ENP storm landfalls on the Pacific islands and continental United States.

Acknowledgments

The authors would like to thank the editor and anonymous reviewers for their valuable feedback in greatly improving the paper.

REFERENCES

REFERENCES
Basher
,
R. E.
, and
X.
Zheng
,
1995
:
Tropical cyclones in the Southwest Pacific: Spatial patterns and relationships to Southern Oscillation and sea surface temperature
.
J. Climate
,
8
,
1249
1260
, doi:.
Bove
,
M. C.
,
J. B.
Elsner
,
C. W.
Landsea
,
X.
Niu
, and
J. J.
O’Brien
,
1998
:
Effect of El Niño on U.S. landfalling hurricanes, revisited
.
Bull. Amer. Meteor. Soc.
,
79
,
2477
2482
, doi:.
Camargo
,
S. J.
, and
A. H.
Sobel
,
2005
:
Western north pacific tropical cyclone intensity and ENSO
.
J. Climate
,
18
,
2996
3006
, doi:.
Camargo
,
S. J.
,
K. A.
Emanuel
, and
A. H.
Sobel
,
2007
:
Use of a genesis potential index to diagnose ENSO effects on tropical cyclone genesis
.
J. Climate
,
20
,
4819
4843
, doi:.
Chan
,
J. C. L.
,
1985
:
Tropical cyclone activity in the northwest Pacific in relation to the El Nino/Southern Oscillation phenomenon
.
Mon. Wea. Rev.
,
113
,
599
606
, doi:.
Changnon
,
S. A.
,
1999
:
Impacts of 1997/98 El Niño generated weather in the United States
.
Bull. Amer. Meteor. Soc.
,
80
,
1819
1827
, doi:.
Chu
,
P.-S.
,
2004
: ENSO and tropical cyclone activity. Hurricanes and Typhoons: Past, Present, and Potential, R. J. Murnane and K. B. Liu, Eds., Columbia University Press, 297–332.
Chu
,
P.-S.
, and
J.
Wang
,
1997
:
Tropical cyclone occurrences in the vicinity of Hawaii: Are the differences between El Niño and non-El Niño years significant?
J. Climate
,
10
,
2683
2689
, doi:.
Clark
,
J. D.
, and
P.-S.
Chu
,
2002
:
Interannual variation of tropical cyclone activity over the central North Pacific
.
J. Meteor. Soc. Japan
,
80
,
403
418
, doi:.
Collins
,
J. M.
,
2007
: The relationship of ENSO and relative humidity to interannual variations of hurricane frequency in the North-East Pacific Ocean. Papers Appl. Geogr. Conf.,30,
324
333
.
Collins
,
J. M.
,
2010
:
Contrasting high North-East Pacific tropical cyclone activity
.
Southeast. Geogr.
,
50
,
83
98
, doi:.
Collins
,
J. M.
, and
I. M.
Mason
,
2000
:
Local environmental conditions related to seasonal tropical cyclone activity in the Northeast Pacific basin
.
Geophys. Res. Lett.
,
27
,
3881
3884
, doi:.
Collins
,
J. M.
, and
D. R.
Roache
,
2011
:
The 2009 hurricane season in the eastern North Pacific basin: An analysis of environmental conditions
.
Mon. Wea. Rev.
,
139
,
1673
1681
, doi:.
Emanuel
,
K. A.
,
2005
:
Increasing destructiveness of tropical cyclones over the past 30 years
.
Nature
,
436
,
686
688
, doi:.
Frank
,
W. M.
, and
G. S.
Young
,
2007
:
The interannual variability of tropical cyclones
.
Mon. Wea. Rev.
,
135
,
3587
3598
, doi:.
Goldenberg
,
S. B.
, and
L. J.
Shapiro
,
1996
:
Physical mechanisms for the association of El Niño and West African rainfall with Atlantic major hurricane activity
.
J. Climate
,
9
,
1169
1187
, doi:.
Goldenberg
,
S. B.
,
C. W.
Landsea
,
A. M.
Mestas-Nuñez
, and
W. M.
Gray
,
2001
:
The recent increase in Atlantic hurricane activity: Causes and implications
.
Science
,
293
,
474
479
, doi:.
Gray
,
W. M.
,
1968
:
Global view of the origin of tropical disturbances and storms
.
Mon. Wea. Rev.
,
96
,
669
700
, doi:.
Gray
,
W. M.
,
1977
:
Tropical cyclone genesis in the western North Pacific
.
J. Meteor. Soc. Japan
,
55
,
465
482
.
Gray
,
W. M.
,
1984a
:
Atlantic seasonal hurricane frequency. Part I: El Niño and 30 mb quasi-biennial oscillation influences
.
Mon. Wea. Rev.
,
112
,
1649
1668
, doi:.
Gray
,
W. M.
,
1984b
:
Atlantic seasonal hurricane frequency. Part II: Forecasting its variability
.
Mon. Wea. Rev.
,
112
,
1669
1683
, doi:.
Gray
,
W. M.
,
C. W.
Landsea
,
P. W.
Mielke
Jr.
, and
K. J.
Berry
,
1994
:
Predicting Atlantic basin seasonal tropical cyclone activity by 1 June
.
Wea. Forecasting
,
9
,
103
115
, doi:.
Hebert
,
P. J.
,
J. D.
Jarrell
, and
M.
Mayfield
,
1997
: The deadliest, costliest, and most intense United States hurricanes of this century (and other frequently requested hurricane facts). NOAA Tech. Memo. NWS TPC-1, Miami, FL, 30 pp. [Available from National Hurricane Center, 11691 SW 17th St., Miami, FL 33165.]
Helsel
,
D. R.
, and
R. M.
Hirsch
,
1991
: Statistical methods in water resources. Techniques of Water-Resources Investigations of the United States Geological Survey, USGS, 266 pp.
Irwin
,
R. P.
, III
, and
R. E.
Davis
,
1999
:
The relationship between the Southern Oscillation index and tropical cyclone tracks in the eastern North Pacific
.
Geophys. Res. Lett.
,
26
,
2251
2254
, doi:.
Jones
,
C. G.
, and
C. D.
Thorncroft
,
1998
:
The role of El Niño in Atlantic tropical cyclone activity
.
Weather
,
53
,
324
336
, doi:.
Klotzbach
,
P. J.
,
2011
:
El Niño–Southern Oscillation’s impact on Atlantic basin hurricanes and U.S. landfalls
.
J. Climate
,
24
,
1252
1263
, doi:.
Klotzbach
,
P. J.
,
2012
:
El Niño–Southern Oscillation, the Madden–Julian Oscillation and Atlantic basin tropical cyclone rapid intensification
.
J. Geophys. Res.
,
117
,
D14104
, doi:.
Klotzbach
,
P. J.
, and
S. E.
Blake
,
2013
:
North-central Pacific tropical cyclones: Impacts of El Niño–Southern Oscillation and the Madden–Julian oscillation
.
J. Climate
,
26
,
7720
7733
, doi:.
Landsea
,
C. W.
,
2000
: El Niño/Southern Oscillation and the seasonal predictability of tropical cyclones. El Niño and the Southern Oscillation: Multiscale Variability and Global and Regional Impacts, H. F. Diaz and V. Markgraf, Eds., Cambridge University Press, 149–181.
Lupo
,
A. R.
,
T. K.
Latham
,
T. H.
Magill
,
J. V.
Clark
,
C. J.
Melick
, and
P. S.
Market
,
2008
:
Interannual variability of hurricane activity in the Atlantic and East Pacific regions
.
Natl. Wea. Dig.
,
32
,
119
133
.
Malkus
,
J. S.
, and
H.
Riehl
,
1960
:
On the dynamics and energy transformations in steady-state hurricanes
.
Tellus
,
12A
,
1
20
, doi:.
Molinari
,
J.
,
D.
Vollaro
,
S.
Skubis
, and
M.
Dickinson
,
2000
:
Origins and mechanisms of eastern Pacific tropical cyclogenesis: A case study
.
Mon. Wea. Rev.
,
128
,
125
139
, doi:.
Nicholls
,
N.
,
1979
:
A possible method for predicting seasonal tropical cyclone activity in the Australian region
.
Mon. Wea. Rev.
,
107
,
1221
1224
, doi:.
Nicholls
,
N.
,
1992
:
Recent performance of a method for forecasting Australian seasonal tropical cyclone activity
.
Aust. Meteor. Mag.
,
40
,
105
110
.
Pielke
,
R. A.
, Jr.
, and
C. N.
Landsea
,
1999
:
La Niña, El Niño, and Atlantic hurricane damages in the United States
.
Bull. Amer. Meteor. Soc.
,
80
,
2027
2033
, doi:.
Ralph
,
T. U.
, and
W. A.
Gough
,
2009
:
The influence of sea-surface temperatures on Eastern North Pacific tropical cyclone activity
.
Theor. Appl. Climatol.
,
95
,
257
264
, doi:.
Rasmusson
,
E. M.
, and
T. H.
Carpenter
,
1982
:
Variations in tropical sea surface temperature and surface wind fields associated with the Southern Oscillation/El Niño
.
Mon. Wea. Rev.
,
110
,
354
384
, doi:.
Ropelewski
,
C. F.
, and
P. D.
Jones
,
1987
:
An extension of the Tahiti–Danvin Southern Oscillation index
.
Mon. Wea. Rev.
,
115
,
2161
2165
, doi:.
Schultz
,
L.
,
2007
:
Some climatological aspects of tropical cyclones in the eastern North Pacific
.
Natl. Wea. Dig.
,
32
,
45
54
.
Simpson
,
R. H.
,
1974
:
The hurricane disaster potential scale
.
Weatherwise
,
27
,
169
186
, doi:.
Sprent
,
P.
,
1989
:
Applied Nonparametric Statistical Methods.
Chapman and Hall, 259 pp.
Steenhof
,
P. A.
, and
W. A.
Gough
,
2008
:
The impact of tropical sea surface temperatures on various measures of Atlantic tropical cyclone activity
.
Theor. Appl. Climatol.
,
92
,
249
255
, doi:.
Trenberth
,
K. E.
,
1997
:
The definition of El Niño
.
Bull. Amer. Meteor. Soc.
,
78
,
2771
2777
, doi:.
Wang
,
B.
, and
J. C. L.
Chan
,
2002
:
How strong ENSO events affect tropical storm activity over the western North Pacific
.
J. Climate
,
15
,
1643
1658
, doi:.
Webster
,
P. J.
,
G. J.
Holland
,
J. A.
Curry
, and
H.-R.
Chang
,
2005
:
Changes in tropical cyclone number, duration, and intensity in a warming environment
.
Science
,
309
,
1844
1846
, doi:.
Wolter
,
K.
, and
M. S.
Timlin
,
1998
:
Measuring the strength of ENSO events—How does 1997/98 rank?
Weather
,
53
,
315
324
, doi:.
Wu
,
L.
,
B.
Wang
, and
S. A.
Braun
,
2008
:
Implications of tropical cyclone power dissipation index
.
Int. J. Climatol.
,
28
,
727
731
, doi:.
Wu
,
P.
, and
P.-S.
Chu
,
2007
: Characteristics of tropical cyclone activity over the eastern North Pacific: The extremely active 1992 and the inactive 1977. Tellus,59A, 444–454, doi:.