SEARCH

SEARCH BY CITATION

Keywords:

  • extratropical cyclone;
  • air-sea interaction;
  • the Kuroshio Current;
  • large meander

Abstract

  1. Top of page
  2. Abstract
  3. 1 Introduction
  4. 2 Data and Methodology
  5. 3 Results
  6. 4 Discussion and Conclusions
  7. Acknowledgments
  8. References

[1] We examined possible responses of cyclone activities to the bimodal path states of the Kuroshio Current [i.e., large meander (LM) and non-LM (NLM)] by using the long-term reanalysis data and the 20th century hindcast experiment of a high-resolution atmosphere-ocean coupled model. Compared with a seasonal mean cyclone track frequency for the LM and NLM periods, a primary cyclone track shifts southward in association with the meander of Kuroshio Current. Composite analyses of the hindcast experiment showed remarkable atmospheric responses accompanying the Kuroshio LM. The Kuroshio LM causes a decrease in latent heat flux in the south of Japan and a southward shift of the near-surface baroclinic zone. Distinctive decreases in thermodynamic fluxes inhibit the rapid development of cyclones in the meander region, eventually inducing positive sea level pressure anomalies downstream from that region.

1 Introduction

  1. Top of page
  2. Abstract
  3. 1 Introduction
  4. 2 Data and Methodology
  5. 3 Results
  6. 4 Discussion and Conclusions
  7. Acknowledgments
  8. References

[2] Extratropical cyclone passages along the Pacific coast of Japan are one of the major cyclone tracks in the Northern Hemisphere winter. It is well known that a primary path of midlatitude cyclones is located along a sea surface temperature (SST) frontal zone [Nakamura et al., 2004]. According to numerous climatological studies, western boundary currents, such as the Kuroshio Current and Kuroshio-Oyashio Extension (KOE) in the western North Pacific, provide favorable conditions for generating and developing the cyclones along the oceanic frontal zone [Chen et al., 1991; Hayasaki and Kawamura, 2012].

[3] In addition to the climatological studies, atmospheric responses to oceanic variations with interannual to decadal timescales are also pointed out by several studies. For instance, variation of the KOE influences a basin scale atmospheric circulation in the North Pacific [e.g., Taguchi et al., 2012]. However, we still have little understanding of atmospheric response to the large meander (LM) of the Kuroshio due to a deficit in observational data over ocean surface. The Kuroshio LM is characterized by the southward shift of the Kuroshio and by a long lifetime ranging from several months to a few years [Kawabe, 1995]. The Kuroshio LM was rarely observed after 1992, which have an ample oceanic observational data, such as a sea surface height (SSH) and ocean surface wind. Recently, a detailed case study for the latest LM period (2004/2005 cold season) was reported by Xu et al. [2010] using satellite remote sensing measurements and regional atmospheric model simulation. Simulated atmospheric responses in the meander region were in good agreement with the observed characteristics: positive sea level pressure (SLP) anomaly and decreases in precipitation and surface wind speed. According to their study, the difference in the simulated sensible and latent heat fluxes from ocean surface between the LM and non-LM (NLM) experiments exceeds 200 Wm−2. The reduced thermodynamic fluxes may also affect cyclone activities migrating along the Pacific coast of Japan.

[4] More recently, Nakamura et al. [2012] (hereafter N2012) investigated possible changes in cyclone activities in association with the Kuroshio LM. They subjectively identified the cyclones migrating eastward along the Pacific coast of Japan using surface weather charts provided by the Japan Meteorological Agency (JMA) and pointed out that the cyclone tracks disperse and slightly shift southward during the LM period. Despite the modulation of cyclone activities due to the LM, thermodynamical forcing from the underlying ocean and associated atmospheric response were not clarified in their study. Their intriguing observational results need to be validated more objectively, using numerical simulations.

[5] Thus, the purpose of this study is to investigate the dynamic response of cyclone activity to the large meander of the Kuroshio through ocean-to-atmosphere forcing. In order to obtain more reliable results, we use three data sets. Two of those data sets are long-term atmospheric reanalysis data set, and the other one is a hindcast run of 20th century simulation using a high-resolution coupled general circulation model (CGCM).

2 Data and Methodology

  1. Top of page
  2. Abstract
  3. 1 Introduction
  4. 2 Data and Methodology
  5. 3 Results
  6. 4 Discussion and Conclusions
  7. Acknowledgments
  8. References

[6] We use the optimally interpolated SST data from December 1981–2012 [Reynolds et al., 2007]. The 6 hourly atmospheric data are derived from the Japanese 25-year Reanalysis/JMA Climate Data Assimilation System (JRA25/JCDAS) [Onogi et al., 2007] and the National Centers for Environmental Prediction (NCEP) Climate Forecast System Reanalysis (CFSR) [Saha et al., 2010] from 1979 to 2012. Horizontal grid intervals of JRA25 and CFSR are 1.25° and 0.5°, respectively. The high-resolution CGCM simulation is performed by the Model for Interdisciplinary Research on Climate version 4h (MIROC4h) [Sakamoto et al., 2012]. Spatial resolutions of the MIROC4h are 0.5625° in the atmosphere part and 0.28125° zonally and 0.1875° meridionally in the ocean part. We use a hindcast run of 20th century simulation (20C3M) from 1950 through 2007 [Komuro et al., 2012]. The 20C3M run was carried out with historical changes of greenhouse gas concentrations, aerosol, and solar forcing. Model configurations and experimental design are fully documented in Sakamoto et al. [2012] and Komuro et al. [2012].

[7] Detecting and tracking cyclones are based on a classic “nearest-neighbor method” [e.g., Serreze et al., 1993]. Latitude/longitude-gridded SLP is converted to the Equal Area Scalable Earth grid (http://nsidc.org/data/ease/ease_grid.html). A cyclone center is identified when the pressure difference between the center and all its adjacent grids is smaller than −0.5 hPa [Hayasaki and Kawamura, 2012]. This tracking method can identify strong cyclones, but some weak cyclones (e.g., “open” depressions, secondary cyclones on a front) fail to track continuously [Flocas et al., 2010]. In this study, however, most of the traveling cyclones passing near the Kuroshio are in good agreement with the manually identified ones in the JMA weather chart. To compare this study with the results shown in N2012, we extracted cyclones that are passing the vicinity of the Kuroshio (Figure 1, enclosed by dashed lines to the north of 20°N).

image

Figure 1. (blue) Seasonal mean climatology (CI = 20 cm) and (shade with white contour) its interannual variability (CI = 10 cm) of simulated SSH during DJFM 1951–2007. A dashed line denotes a boundary for extracting cyclones in this study.

Download figure to PowerPoint

3 Results

  1. Top of page
  2. Abstract
  3. 1 Introduction
  4. 2 Data and Methodology
  5. 3 Results
  6. 4 Discussion and Conclusions
  7. Acknowledgments
  8. References

3.1 Simulated Mean Cyclone Activities and Kuroshio

[8] First, we examine climatological cyclone activities for a cold season (December–March) in the 20C3M run. Compared with the reanalysis data sets, the simulated cyclone frequency (15.3 per season) is 40% and 20% higher than those in the JRA25 (10.9 per season) and in the CFSR (13.1 per season), respectively. Overestimation of the simulated cyclone frequency is partly caused by the difference in the horizontal grid interval. Despite the fact that the simulated cyclone frequency is higher than that in the reanalysis data set, spatial distribution of the primary cyclone track and the seasonal march of cyclone frequency are roughly consistent with those in the reanalysis (not shown). Thus, we considered that the 20C3M run is able to reproduce cyclone passing along the Pacific coast of Japan.

[9] Next, we examine the climatology and interannual variability of the Kuroshio in the 20C3M run. Figure 1 shows the seasonal mean climatology and interannual variability of SSH during the cold seasons of 1951–2007. Because SSH provides information on the heat content in the upper ocean layers, it is easy to identify a cold-core vortex and to determine the Kuroshio axis. The seasonal mean position of the Kuroshio axis, which is obtained from latitude with the maximum horizontal gradient of SSH, shows good agreement with the observation. Although the 20C3M run reproduces bimodal paths of the Kuroshio (Figures 2a and 2b), the simulated meander path tends to shift westward and southward compared with that of the observation [Kawabe, 1995; N2012]. To determine the LM and NLM years, we calculate latitudinal deviation from a reference latitude of the Kuroshio axis. The reference latitude in each meridian is selected from a coastal path close to the Pacific coast of Japan (133°E–140°E). For the analytical convenience, we use the sixth northernmost latitude of the 57 samples (1951–2007). To make the composite charts for the simulated LM and NLM periods, we select the 10 largest and 10 smallest years of the seasonal mean meridional displacement of the Kuroshio axis (Figure 2c).

image

Figure 2. (solid contour) Monthly mean horizontal gradient of simulated SSH in January (a) 2004 and (b) 2006. Stars indicate the Kuroshio axis in each meridian. A red line denotes a reference latitude of the Kuroshio axis. (c) Time series of a seasonal mean meridional displacement of the Kuroshio axis during 1951–2007. The meridional displacement is obtained by area-averaged deviation of the Kuroshio axis from a reference latitude (133°E–140°E). Blue inverted and red triangles show composite years of LM and non-LM path, respectively.

Download figure to PowerPoint

3.2 Cyclone Activity in the Kuroshio LM Period

[10] To validate the results of N2012, we first perform composite analyses during the LM and NLM years using two reanalysis datasets. Based on the LM definition of JMA, the observed LM years (1982, 1983, 1984, 1987, 1988, 1990, and 2005) and NLM years (1986, 1993, 1994, 1995, 2002, 2003, and 2004) are selected. Both reanalysis datasets reveal that the primary track of cyclones in the LM period shifts southward to the south of Japan compared with that in the NLM period (not shown), which is consistent with N2012. Thus, the difference in the primary cyclone track between the LM and NLM periods is also confirmed by using objectively identified cyclones.

[11] Figure 3 shows the horizontal distributions of a simulated cyclone track frequency during the LM and NLM periods. Composites of seasonal mean SST and the primary cyclone track are also displayed. During the LM period (Figure 3a), the horizontal SST gradient is largest along 30°N in southwestern Japan, whereas the SST front is very close to the Pacific coast of Japan during the NLM period (Figure 3b). The primary cyclone tracks during the LM and NLM periods seem very sensitive to the meridional displacement of the SST front. In addition, cyclone activities in the LM period (maximum development rate, minimum SLP, and total intensification) become weak compared with those in the NLM period (Table 1). The maximum development rate (DR) is calculated by 12 h SLP change (unit: hPahr−1) within the search area (130°E–145°E, the same as in N2012). In order to estimate cyclone DR with different latitudes, SLP change is normalized at 60°N [Yoshiike and Kawamura, 2009], calculated from DR = ((p(t + 6) − p(t − 6))/12)(sin60/sinϕ) where p is the SLP at the cyclone center, t is the time in hour, and ϕ is the latitude at the cyclone center. The minimum SLP is obtained from the center pressure at the grid with the maximum Laplacian of pressure (∇ 2p). The total intensification is measured by ∇ 2p differences between the maximum and initial stages of each cyclone within the search area. To compare the results by N2012, these statistics are calculated based on the extracted cyclones that originated from the East China Sea. Most of these activities show that cyclones tend to weaken during the LM period in both the reanalysis and the 20C3M run, except the maximum DR in the 20C3M run. Weakening of cyclone activities in the LM period is also discussed in the next section.

image

Figure 3. Composite cyclone frequency (unit: count month−1) in the 10 cold seasons (DJFM) for the Kuroshio (a) LM and (b) NLM periods. Area enclosed by thick dashed line is used to determine a primary path of cyclones (triangles). Solid lines over ocean denote the composite SST (CI = 2°C).

Download figure to PowerPoint

Table 1. Mean statistics (maximum development rate (DR; unit: hPa hr−1), minimum SLP (unit: hPa), and total intensification (difference of ∇2p; unit: 10−9 Pam−2) of the extracted cyclones for the Kuroshio LM and NLM periods using JRA25/JCDAS (upper part), NCEP CFSR (middle part), and MIROC4h 20C3M (lower part) datasets. Definitions of these statistics are described in the text.
DatasetperiodcountDRminSLPΔ(∇2p)
JRA25LM480.991002.910.3
NLM351.05999.114.8
total (33-yr)1840.921002.411.4
CFSRLM460.901003.610.7
NLM330.991001.314.1
total (33-yr)1860.891003.011.2
20C3MLM660.761006.312.2
NLM630.731004.119.3
total (57-yr)4040.751005.515.5

[12] Composite maps of the simulated latent heat (LH) flux from the ocean surface during the LM and NLM periods are shown in Figure 4. Decreased LH flux is apparent in southwestern Japan during the LM period (Figure 4a). Composite differences of LH and sensible heat (not shown) fluxes over the corresponding LM region are 100 and 50 Wm−2, respectively. Total differences of turbulent fluxes are comparable to the estimation from the regional model experiments in winter 2004/2005 [Xu et al., 2010]. The distinctive change in surface heat fluxes is expected to affect cyclone activity.

image

Figure 4. (shades with white contour) Composite latent heat flux (CI = 50 Wm−2) and (solid contour) SST (CI = 2°C) for the Kuroshio LM and NLM periods. Triangles show the primary cyclone track (same as in Figure 3).

Download figure to PowerPoint

3.3 Atmospheric Response to the Kuroshio LM

[13] Figure 5 shows the seasonal climatology and composite differences (LM minus NLM) of atmospheric states between the LM and NLM periods. A statistically significant SLP difference appears near the meander region and extends northeastward to the east of Japan (Figure 5a). These significant positive SLP anomalies reflect a weakening of cyclone activities.

image

Figure 5. (red and blue lines) Composite difference of (a) SLP, (b) meridional eddy heat flux at 850 hPa, and (c) horizontal gradient of equivalent potential temperature (inline image) at 925 hPa between the Kuroshio LM and NLM periods (former minus latter). Black line represents seasonal climatology. Light and heavy shades indicate a statistically significant area (5% and 1% significance level).

Download figure to PowerPoint

[14] In fact, total development of cyclones and minimum SLP in the LM period is weaker than that in the NLM period (Table 1). Furthermore, meridional sensible heat flux by subweekly scale eddies (vT′; 2–8 day cycle) significantly decreases in the vicinity of the LM region (Figure 5b). During the LM period, a strong near-surface baroclinicity zone (characterized by large horizontal gradient of equivalent potential temperature; | ∇ θe| at the 925 hPa level) is also displaced southward (~ 30°N) along with the meander axis of the Kuroshio (Figure 5c). The composite analyses show that the Kuroshio LM induces the significant positive SLP anomaly over the western North Pacific. Reduced thermodynamic fluxes (Figure 4) and southward displacement of the strong baroclinicity provide unfavorable conditions for developing cyclones.

4 Discussion and Conclusions

  1. Top of page
  2. Abstract
  3. 1 Introduction
  4. 2 Data and Methodology
  5. 3 Results
  6. 4 Discussion and Conclusions
  7. Acknowledgments
  8. References

[15] As mentioned in N2012, the primary track of cyclones, which passes along the Pacific coast of Japan, shifts southward in association with the Kuroshio LM. This study confirms the southward displacement of the cyclone track in both the long-term reanalysis data and the hindcast experiment using high-resolution CGCM. Moreover, the composite analyses with respect to the simulated LM and NLM periods suggest that thermodynamic forcing from ocean to atmosphere is important to modulate the cyclone activities. The pronounced change in the surface heat fluxes modifies near-surface baroclinicity in a marine atmospheric boundary layer overlying the LM region. Similar ocean-to-atmosphere forcing by a mesoscale oceanic eddy was also reported in the unusual development of the Hurricane Catarina [Vianna et al., 2010]. Due to a southward shift of the baroclinic zone, the LM region becomes unfavorable for developing cyclones. As a consequence, the primary cyclone track shifts southward to the south of Japan. These results strongly suggest that regional scale SST anomalies along the Kuroshio have a strong potential to change cyclone activity along the oceanic frontal zone.

[16] The findings in this study show the importance of ocean-to-atmosphere forcing to the cyclone activity along the Kuroshio. However, cyclone activity along the Kuroshio is also influenced by other factors. For instance, interannual to decadal scale climate change (e.g., El Niño–Southern Oscillation, the Pacific Decadal Oscillation) has a large impact on the background atmospheric circulation states in the North Pacific. Furthermore, it is known that the spatial distribution of rapidly deepening cyclones in the western North Pacific is also influenced by the East Asian winter monsoon activity [Yoshiike and Kawamura, 2009; Iizuka et al., 2013]. To precisely evaluate the significant influence of oceanic variation to cyclone activity, further studies will be required using high-resolution AGCM and CGCM experiments.

Acknowledgments

  1. Top of page
  2. Abstract
  3. 1 Introduction
  4. 2 Data and Methodology
  5. 3 Results
  6. 4 Discussion and Conclusions
  7. Acknowledgments
  8. References

[17] This study was supported by the grant-in-aid for scientific research 22106005 by the Ministry of Education, Culture, Sports, Science and Technology (MEXT), Japan.

[18] The Editor thanks two anonymous reviewers for their assistance in evaluating this paper.

References

  1. Top of page
  2. Abstract
  3. 1 Introduction
  4. 2 Data and Methodology
  5. 3 Results
  6. 4 Discussion and Conclusions
  7. Acknowledgments
  8. References