Volume 44, Issue 2 p. 1167-1174
Research Letter
Free Access

A teleconnection between Atlantic sea surface temperature and eastern and central North Pacific tropical cyclones

Christina M. Patricola

Corresponding Author

Christina M. Patricola

Department of Atmospheric Sciences, Texas A&M University, College Station, Texas, USA

Climate and Ecosystem Sciences Division, Lawrence Berkeley National Laboratory, Berkeley, California, USA

Correspondence to: C. M. Patricola,

[email protected]

Search for more papers by this author
R. Saravanan

R. Saravanan

Department of Atmospheric Sciences, Texas A&M University, College Station, Texas, USA

Search for more papers by this author
Ping Chang

Ping Chang

Department of Atmospheric Sciences, Texas A&M University, College Station, Texas, USA

Department of Oceanography, Texas A&M University, College Station, Texas, USA

Physical Oceanography Laboratory/Qingdao Collaborative Innovation Center of Marine Science and Technology, Ocean University of China, Qingdao, China

Search for more papers by this author
First published: 23 December 2016
Citations: 27
This manuscript has been authored by an author at Lawrence Berkeley National Laboratory under Contract No. DE-AC02-05CH11231 with the U.S. Department of Energy. The U.S. Government retains, and the publisher, by accepting the article for publication, acknowledges, that the U.S. Government retains a non-exclusive, paid-up, irrevocable, world-wide license to publish or reproduce the published form of this manuscript, or allow others to do so, for U.S. Government purposes has been authored by an author at Lawrence Berkeley National Laboratory under Contract No. DE-AC02-05CH11231 with the U.S. Department of Energy. The U.S. Government retains, and the publisher, by accepting the article for publication, acknowledges, that the U.S. Government retains a non-exclusive, paid-up, irrevocable, world-wide license to publish or reproduce the published form of this manuscript, or allow others to do so, for U.S. Government purposes.

Abstract

The El Niño–Southern Oscillation (ENSO) is a major source of seasonal tropical cyclone (TC) predictability in both local and remote ocean basins. Unusually warm eastern-central equatorial Pacific sea surface temperature (SST) during El Niño tends to enhance eastern and central North Pacific (ECNP) TCs and suppress Atlantic TCs. Here we demonstrate that Atlantic SST variability likewise influences remote TC activity in the eastern-central Pacific through a Walker Circulation-type response analogous to the ENSO-Atlantic TC teleconnection, using observations and 27 km resolution tropical channel model (TCM) simulations. Observed and simulated ECNP TC activity is reduced during the positive Atlantic Meridional Mode (AMM), which is characterized by warm northern and cool southern tropical Atlantic SST anomalies, and vice versa during the negative AMM. Large ensembles of TCM simulations indicate that SST variability, rather than internal atmospheric variability, drives extreme ECNP hurricane seasons.

Key Points

  • A teleconnection exists between tropical Atlantic sea surface temperature and seasonal eastern-central Pacific tropical cyclone activity
  • Both the Atlantic Meridional Mode and ENSO provide seasonal eastern-central Pacific tropical cyclone predictability
  • Tropical Pacific and Atlantic sea surface temperature, not internal atmospheric variability, drives extreme East Pacific hurricane seasons

1 Introduction

Interannual to multidecadal oceanic modes of climate variability are a major source of seasonal tropical cyclone (TC) predictability, as sea surface temperature (SST) patterns are often predictable months ahead of the hurricane season and can influence TCs in local and remote ocean basins. For example, eastern and central North Pacific (ECNP) tropical cyclone activity is strongly connected to SST variability within its own basin, especially that characterized by the El Niño–Southern Oscillation (ENSO). During El Niño, when eastern-central equatorial Pacific SSTs are unusually warm, TC activity tends to be enhanced in the offshore eastern Pacific and central Pacific in association with decreased vertical wind shear, increased potential intensity, and observed TC genesis farther west than usual [Chu and Wang, 1997; Clark and Chu, 2002; Ralph and Gough, 2009; Jien et al., 2015; Fu et al., 2016; Camargo et al., 2007; Irwin and Davis, 1999]. Atlantic TCs are also influenced by SST variability within their own basin [e.g., Emanuel, 2005]. Atlantic hurricane seasons are more active during the positive phase of the Atlantic Multidecadal Oscillation (AMO) and the Atlantic Meridional Mode (AMM), which, respectively, characterize multidecadal variability in North Atlantic SST and interannual to decadal variability in the meridional gradient in tropical Atlantic SST [Landsea et al., 1999; Goldenberg et al., 2001; Vitart and Anderson, 2001; Vimont and Kossin, 2007; Kossin and Vimont, 2007].

In addition to these local oceanic influences on Atlantic TC activity, Atlantic hurricane seasons are also remotely influenced by tropical Pacific SST variability via ENSO teleconnection patterns [Bove et al., 1998; Pielke and Landsea, 1999; Smith et al., 2007]. The physical mechanisms for ENSO's influence on Atlantic TCs operate through both a Walker Circulation-type response and an upper tropospheric temperature response [Tang and Neelin, 2004]. Warmer than usual equatorial Pacific SST during El Niño drives an eastward shift in tropical Pacific convection and an upper tropospheric zonal wind response that enhances tropical Atlantic vertical wind shear, leading to TC suppression [Arkin, 1982; Gray, 1984; Goldenberg and Shapiro, 1996; Zhu et al., 2012]. The location of El Niño plays a critical role in the deep convection response and the degree of Atlantic TC suppression and ECNP TC enhancement [Patricola et al., 2016].

Since ENSO influences TC activity in both the ECNP and Atlantic basins, it is not surprising that an inverse relationship between TC activity in the two basins has been observed [Wang and Lee, 2009]. However, this relationship operates on interannual to decadal time scales (i.e., time scales characteristic of ENSO and longer) suggesting that decadal climate variations, perhaps those in the Atlantic, can play a role in driving ECNP TCs. The anticorrelation between Atlantic and ECNP TC activity, together with the importance of local and remote SST variability in controlling seasonal Atlantic TC activity [Bell and Chelliah, 2006; Klotzbach, 2011; Patricola et al., 2014], leads us to ask: Is there a remote influence of Atlantic SST on ECNP TC activity that is analogous to the ENSO-Atlantic TC relationship? Understanding this question can improve seasonal forecasts and future projections of ECNP TC activity.

2 Data and Methodology

We next describe the observational data sets and climate model used in this study, which are the same as in Patricola et al. [2014] and Patricola et al. [2016], respectively. Additional details are in the supporting information.

2.1 Observational Data and Climate Indices

TC observations are from the Revised Hurricane Database (HURDAT2) [Landsea et al., 2004; Landsea and Franklin, 2013], which contains 6-hourly data from the National Hurricane Center “best tracks” from 1851 to present for the Atlantic and eastern and central North Pacific basins. We quantified seasonal TC activity using the sum of accumulated cyclone energy (ACE) [Bell et al., 2000] calculated for all TCs each year. TCs that originate in the Central Pacific (180°–140°W) and East Pacific (east of 140°W) are considered together and referred to as ECNP TCs.

ENSO is represented by the Niño 3.4 index of the National Oceanic and Atmospheric Administration (NOAA) Climate Prediction Center (CPC), which is the area-average of eastern-central equatorial Pacific (5°S–5°N, 170°W–120°W) monthly SST. The AMM index is described by Chiang and Vimont [2004].

For the observationally based analysis, we categorized each year from 1950 to 2015 according to ENSO and AMM phase using the May–October averaged Niño 3.4 and AMM indices. A hurricane season is defined as having occurred during a negative (positive) climate mode phase if the May–October averaged climate mode index is less than or equal to (exceeds or is equal to) the 25th (75th) percentile of the May–October averaged index over the 1950–2015 period (Table S1 in the supporting information).

2.2 Tropical Channel Model and Simulations

The climate modeling approach controls for SST variability outside the basin of interest and generates a sample size sufficient to evaluate statistical significance and quantify internal atmospheric variability, which is needed to assess the seasonal TC predictability gained from SST patterns. We conducted simulations with the Weather Research and Forecasting model [Skamarock et al., 2008] configured as a tropical channel model (TCM) with TC-permitting horizontal resolution of 27 km, as in Patricola et al. [2016]. The TCM represents TCs and covers a model domain (Figure 1) that extends around the equator from 30°S to 50°N, allowing atmospheric disturbances to propagate throughout the tropics and part of the northern midlatitudes.

Details are in the caption following the image
(a) Positive AMM, (b) negative AMM, (c) cold tongue El Niño, and (d) La Niña SST (°C) forcings on the TCM domain. The color scale for Figures 1c and 1d covers 4 times the range of the scale for Figures 1a and 1b).

The control or “climatology” simulation uses the monthly climatology (1950–2011) of SST from the 1.0° × 1.0° HadISST data set of the Met Office Hadley Centre [Rayner et al., 2003]. It is the same control simulation as in Patricola et al. [2016], which presented model validation. Four AMM and ENSO experiments (“AMM+,” “AMM-,” “La Niña,” and “Cold Tongue El Niño”) differ from one another in prescribed SST forcing (Figure 1) and are designed by adding a SST anomaly characteristic of each climate mode phase and constant in time to the SST of the climatology simulation. The SST forcings are based on HadISST.

The AMM forcing is constructed by identifying the strongest 10% of observed positive and negative AMM events during August–October (i.e., the peak of the Western Hemisphere hurricane season) over the 1950–2014 period according to the AMM index (Table S2). These choices were made to generate a clear simulated TC activity response. We then calculated the deviation from the August–October 1950–2014 mean SST for composites of each positive and negative AMM events and created an idealized AMM forcing that is the average of the positive AMM SST anomaly and the opposite of the negative AMM SST anomaly, where significant (10% level). Therefore, one SST pattern is created, allowing us to characterize the positive (Figure 1a) and negative (Figure 1b) AMM SST patterns with equal magnitude and opposite sign. The La Niña forcing (Figure 1d) is the average deviation from the August–October 1950–2014 mean SST for observed cases (Table S2) that are less than or equal to the 10th percentile of the August–October averaged Niño 3.4 index during the 1950–2014 period, where significant (10% level). The Cold Tongue El Niño forcing (Figure 1c) is described in Patricola et al. [2016].

The 16-member and 22-member ensemble simulations are generated as described in the supporting information.

3 Results

3.1 Tropical Cyclone Activity

The observations and TCM simulations solidly support a significant remote influence of tropical Atlantic SST patterns on eastern-central North Pacific TCs. From scatterplots of seasonal Atlantic and ECNP ACE marked by ENSO and AMM, it is evident that the observed anticorrelation in seasonal TC activity between the two basins (R =−0.38) is associated not only with ENSO (Figure 2a) but also with the AMM (Figure 2b). The TCM simulations support this finding (Figure S2) and show less scatter than observations due to the isolation of ENSO or AMM in the SST forcings. More active than average ECNP hurricane seasons tend to occur during the negative AMM phase, and vice versa (Figures 2 and S2). In fact, the observed correlation between seasonal ECNP ACE and the May–October averaged AMM index (R =−0.32) is significant and of similar magnitude but opposite sign to the correlation between seasonal ECNP ACE and the May–October Niño 3.4 index (R = 0.33; Table S3) over the 1950–2015 period, supporting the idea that both tropical Pacific and Atlantic SST variability are important drivers of ECNP TC activity. As expected, the observations show that El Niño tends to drive more active ECNP hurricane seasons but less active Atlantic hurricane seasons than average (Figure 2a).

Details are in the caption following the image
Scatterplots of observed seasonal Atlantic and ECNP ACE (104 knots2) from 1950 to 2015 for (a) neutral ENSO (black), La Niña (blue), and El Niño (red) hurricane seasons, and (b) neutral AMM (black), negative AMM (blue), and positive AMM (red) hurricane seasons, as defined according to the 25th and 75th percentiles of the May–October averaged SST indices in Table S1. Dashed lines denote climatological averages of Atlantic and ECNP ACE.

Since tropical Pacific and Atlantic SST patterns are not entirely independent [Enfield and Mayer, 1997; Klein et al., 1999; Saravanan and Chang, 2000; Mo and Häkkinen, 2001; Chang et al., 2006; Wang et al., 2011], we examined the observed composite SST anomalies (Figure S3) for co-occurring patterns, as described in the supporting information.

Observed and simulated TCs in the Atlantic and ECNP exhibit coherent spatial patterns in response to ENSO and AMM (Figure 3). During the positive AMM, TC track density is decreased over the ECNP and increased over the Atlantic (Figures 3a and 3e), and vice versa during the negative AMM (Figures 3b and 3f). The TC track density response to AMM is generally similar between model and observations, except for a greater magnitude response in the TCM simulations that is likely related to the greater magnitude of SST forcing prescribed in the model than in the observed composite (Figures 1a and 1b versus Figures S3a and S3b).

Details are in the caption following the image
Observed TC track density (number of TCs day−1 per 10 seasons) from observed composites of (a) positive AMM, (b) negative AMM, (c) El Niño, (d) La Niña hurricane seasons minus climatology, according to the 25th and 75th percentiles of the May–October averaged SST indices in Table S1. Simulated TC track density from (e) AMM+, (f) AMM−, (g) cold tongue El Niño, and (h) La Niña simulations minus the climatology simulation. TC track density is calculated over 2° × 2° boxes at a frequency of 6 h during 1 June to 1 December.

Consistent with previous studies, observed and simulated Atlantic TC track density is reduced during El Niño (Figures 3c and 3g) and enhanced during La Niña (Figures 3d and 3h), with the greatest response in the western Atlantic and Gulf of Mexico. The observed spatial pattern of the TC track density response to ENSO is more complex in the Pacific, with increases in track density in the central and offshore eastern Pacific and track density decreases in the near-coastal eastern Pacific during El Niño. This intrabasin variability in eastern Pacific TC activity driven by ENSO has been documented and linked to the topographically locked Central American Gap Winds that give rise to low-level vorticity anomalies [Fu et al., 2016]. The observed ECNP TC track density response to AMM also exhibits an intrabasin pattern. However, the climate model fails to represent these localized near-coastal responses. This difference between the model and observations is likely related to model biases in Gap Winds representation—a subject currently under investigation. Therefore, simulated changes in eastern Pacific TC landfall should be interpreted cautiously.

The observations and model simulations provide evidence that AMM is, like ENSO, an important driver of ECNP TC activity. This is supported by the similar magnitudes in responses of basin-averaged seasonal ECNP TC statistics to ENSO and AMM in observations (Figure 4a) and the TCM (Figure 4b), and by the statistical TC model of Caron et al. [2015]. During the positive AMM, both observations and model simulations show reduced ACE (−33% and−38%, respectively) and number of major hurricanes (−41% and−53%, respectively) relative to their respective climatologies. Likewise, the negative AMM drives increased seasonal ECNP ACE and number of major hurricanes. The observed and simulated responses in ACE and a number of major hurricanes are of similar magnitude between both AMM and ENSO (Figure 4).

Details are in the caption following the image
(a) Average deviation (percent) from the 1950 to 2015 mean in observed seasonal ECNP ACE and number of tropical storms, hurricanes (category 1–5), and major hurricanes (category 3–5) for composites according to the 25th and 75th percentiles of the May–October averaged AMM and ENSO phases, defined in Table S1. (b) Ensemble mean deviations from the climatology simulation (percent) in the same quantities for La Niña, cold tongue El Niño, AMM+, and AMM− TCM simulations. All simulated deviations are significant (10% level).

The influence of each AMM and ENSO individually on seasonal ECNP TC activity motivates investigating the influence of the climate modes jointly on ECNP TC activity (Figure S5). Although binning data jointly by AMM and ENSO results in some combinations with a small sample size, the observations support the finding that both AMM and ENSO are important drivers of ECNP TC activity (Figure S5). Furthermore, this suggests that AMM and ENSO can exert compensating or constructive influences on ECNP TC activity, similar to the case for Atlantic TCs [Patricola et al., 2014]. Of particular note is that ENSO alone is insufficient to describe seasonal ECNP TC variability; ECNP ACE and a number of major hurricanes are less than average during La Niña only together with a neutral or positive AMM and are greater than average during El Niño only when combined with a neutral or negative AMM. The response in ECNP TC activity to El Niño is noticeably greater with neutral compared to negative AMM likely because the strong 1997 and 2015 El Niño events occurred with a neutral AMM (Table S1).

3.2 Physical Teleconnection Mechanism

The physical mechanism that supports the AMM-ECNP TC relationship is analogous to the ENSO-Atlantic TC teleconnection in several ways. Both the ENSO- and AMM-forced changes in favorability for ECNP and Atlantic TCs, diagnosed with the genesis potential index [Emanuel and Nolan, 2004], are associated primarily with vertical wind shear, with secondary contributions from relative humidity (Figure S6). Indeed, observations and TCM simulations show increased (decreased) vertical wind shear in the eastern Pacific during the positive (negative) AMM (Figures S7a, S7b, S7e, and S7f) consistent with the seasonal TC activity responses (Figures 3 and 4). In addition, an anticorrelation in the vertical wind shear response between the Atlantic and eastern Pacific is apparent in both the AMM and ENSO simulations (Figure S7), consistent with the observed relationship between vertical wind shear and the AMM [Kossin and Vimont, 2007, Figure 6]. Furthermore, the positive AMM drives not only local anomalous ascent in the northern tropical Atlantic but also remote anomalous descent in the eastern North Pacific (Figure S8a), and vice versa for the negative AMM (Figure S8b), consistent with Zhang et al. [2016]. These changes in tropical Atlantic and eastern Pacific vertical motion are largely qualitatively similar to those driven by La Niña and Cold Tongue El Niño (Figures S8c and S8d), respectively. In addition, there are co-occurring changes in eastern Pacific and tropical Atlantic upper and lower tropospheric zonal winds that are qualitatively similar for the positive AMM and La Niña (Figures S9a and S9c and S10a and S10c), and for the negative AMM and El Niño (Figures S9b and S9d and S10b and S10d). Together, these circulation responses suggest that the AMM-ECNP TC teleconnection operates through a Walker Circulation-type response similar to the ENSO-Atlantic TC teleconnection.

3.3 Internal Atmospheric Variability

The significant mean responses in TC activity to AMM and ENSO fulfill one of two criteria needed to gain seasonal TC predictability from skillful ENSO and AMM predictions. The second criterion is comparatively small internal atmospheric variability, since high predictability requires a low ratio of signal to noise (i.e., mean response to internal atmospheric variability). The statistical summaries of seasonal ACE from the full ensemble of each TCM simulation (Figure 5) illustrate the approximate mean ACE response (by comparing the median ACE of the climatology simulation and an experiment) and internal atmospheric variability (estimated from the intraensemble spread in ACE from a TCM experiment). In the positive AMM simulation, there is a decrease in median ACE relative to the climatology simulation, as well as no overlap between the 25th and 75th percentile ranges of the positive AMM and climatology simulations, indicating relatively small internal atmospheric variability. The same is true for the negative AMM and ENSO simulations, suggesting that tropical Atlantic SST variability provides seasonal TC predictability. If a predictability measure is estimated using the ratio of the absolute value of the mean ACE response to the standard deviation of ACE in an experiment (Table S5), we find that values for the AMM simulations (2.0 and 2.3) are close to those from the ENSO simulations (1.7 and 2.4), further suggesting that ENSO and AMM provide similar value to seasonal ECNP TC prediction given a “perfect” SST forecast. However, in the practical sense it is worth noting that ENSO may be more predictable than the AMM [Chen et al., 2004; Jin et al., 2008; Vimont, 2012].

Details are in the caption following the image
Boxplots of seasonal (1 June to 1 December) ACE (104 knots2) showing minimum, first quartile, median, third quartile, and maximum values for TCs generated over the eastern and central North Pacific from HURDAT2 observations and the climatology, cold tongue El Niño, La Niña, AMM+, and AMM− TCM simulations.

The large ensemble of TCM simulations indicates that interannual tropical Pacific and Atlantic SST variability drives the extremes in seasonal ECNP TC activity (i.e., very active or inactive hurricane seasons). This is suggested by comparison of the boxplot of observed seasonal ECNP ACE from the 1950–2015 period with boxplots of the TCM experiments. The climatology simulation produces a median ACE value that is close to observed; however, the range in ACE covered by the climatology simulation is far less than observed. The ACE of the least and most active 25% of observed hurricane seasons approximately coincides with the ACE values from the AMM and ENSO simulations. This implies that reliable future projection of extremely active and inactive ECNP hurricane seasons relies on skillful future projection of the frequency, intensity, and location of ENSO and AMM.

4 Discussion

This study presents a newly discovered teleconnection between tropical Atlantic SST variability and eastern and central North Pacific TC activity. Observations from 1950 to 2015 indicate that seasonal ECNP TC activity is remotely suppressed by positive AMM conditions and enhanced by the negative AMM, and that AMM is, like ENSO, an important driver of ECNP TC activity. A large ensemble of tropical channel model simulations forced with idealized AMM and ENSO SST patterns supports these conclusions and reveals that the mean response in ECNP TC activity is sufficiently large compared to internal atmospheric variability to gain significant seasonal TC predictability from skillful AMM predictions. Furthermore, the TCM simulations indicate that interannual tropical Pacific and Atlantic SST variability, rather than internal atmospheric variability, drives extremely active and inactive ECNP hurricane seasons. Observations also suggest that AMM and ENSO may exert compensating or constructive influences on ECNP TC activity.

The mechanism by which Atlantic SST influences ECNP TCs is in many ways analogous to the Walker Circulation response associated with the ENSO-Atlantic TC teleconnection. Both ENSO and AMM modify the favorability for ECNP and Atlantic TCs largely through vertical wind shear. In addition, the responses in eastern Pacific and tropical Atlantic TC activity, vertical velocity, and upper and lower tropospheric zonal wind are similar between the positive AMM and La Niña, and between the negative AMM and El Niño. However, the western North Pacific circulation response is different between the AMM and ENSO. This work highlights the importance of considering changes in interannual modes of tropical Atlantic and Pacific SST variability in seasonal and future predictions of eastern-central North Pacific TC activity.

Acknowledgments

This research was supported by U.S. National Science Foundation grant 1347808 and used the Extreme Science and Engineering Discovery Environment (XSEDE), which is supported by National Science Foundation grant ACI-1053575. This material is based upon work supported by the U.S. Department of Energy, Office of Science, Office of Biological and Environmental Research, Climate and Environmental Sciences Division, Regional and Global Climate Modeling Program, under award number DE-AC02-05CH11231. High-performance computing resources provided by the Texas Advanced Computing Center (TACC) at the University of Texas at Austin and by Texas A&M Supercomputing. HURDAT2 was provided by Atlantic Oceanographic and Meteorological Laboratory (AOML)/NOAA Hurricane Research Division. AMM index was calculated by Daniel Vimont at University of Wisconsin-Madison and provided by NOAA Earth System Research Laboratory (ESRL). Niño 3.4 index was provided by NOAA CPC. NCEP-II was obtained from NOAA National Operational Model Archive and Distribution System (NOMADS). Climate model output was available by request to C.M.P. We thank Jim Kossin and one anonymous reviewer for insightful comments that improved the manuscript.