副熱帶東北太平洋海溫年際及年代際變化以及其對亞洲-太平洋之影響

全文

(1)Department of Earth Sciences National Taiwan Normal University Doctoral Dissertation. Interannual and Interdecadal variations of Subtropical Eastern North Pacific SST and its Impact on Pacific Asia Climate. Yi-Kai Wu. Advisor: Prof. Cheng-Ta Chen Prof. Chi-Cherng Hong. 108 8 August 2019.

(2) p. d. 5. e. t t. ，. t. 5. r. （t t. T. t. t ：t. nt t. T. t 5. 5. 5E. 5E. ）5. 5. 、. E. I. 5. 5. 5. T.

(3) Abstract The subtropical Eastern North Pacific (SENP) sea surface temperature (SST) is persistent warming from 2013. Notably, the impact of the SENP SST on weather and climate are also significant and demonstrated in several recent studies. Compare to the equatorial eastern Pacific SST, the characteristics of SENP SST are less discussed. In this study, we aired to diagnose the characteristics, warming mechanism, and possible impact of SENP SST. It reveals that the SENP SST is associated with Pacific meridional mode (PMM) and global warming. The warming trend contributes approximately 15% to the variability of SENP SST. Wavelet analysis further shows that SENP SST exhibits the interannual and interdecadal variations. The regression analysis shows that SENP and ENSO-like (El Niño–Southern oscillation) SST warm simultaneously in the interannual time scale. The Pacific decadal oscillation (PDO), North Pacific gyre oscillation (NPGO), Atlantic multi-decadal oscillation (AMO) have an impact on SENP SST variation in the interdecadal time scale. The mixed layer heat budget analysis suggested that the SENP SST is warming through wind–evaporation–SST mechanism. Furthermore, the impact of SENP SST on ENSO and tropical cyclone (TC) activity are also showed in the two case studies, which hinders the development of 2014 El Niño and enhances the 2016 TC activity in western North Pacific.. Key words: subtropical Eastern North Pacific SST, Interannual, Interdecadal, ENSO, PMM, Global Warming. II.

(4) 5. o. 5. 。. r r. r. S. T. 5. i. 。. 5. % W. o. o. o. r. 5. w. o T. v. o. 5 .. ,3 0 1 2 . 5. 5. o. 。. r. 5 T. o. 5S. ( ). 。. 5. a. v. o. i. III.

(5) Catalog ............................................................................................................ I Abstract ..................................................................................................... II ..........................................................................................................III Catalog .................................................................................................... IV Ch1. Introduction .......................................................................................1 1.1 Oceanic and Atmospheric Background in Tropical Pacific ..............1 1.2 Oceanic and Atmospheric Background in Subtropical Pacific .........2 1.3 Motivation .........................................................................................3 Ch2. Data and Method ...............................................................................6 Ch3. Spatial–Temporal Characteristics of Subtropical Eastern North Pacific SST .................................................................................................8 3.1 SST Variance, Background, and SENP Definition ...........................8 3.2 Spatial–Temporal Evolution .............................................................9 3.3 Comparison between SENP and PMM ...........................................10 3.4 Effect of Warming Trend ................................................................11 Ch4 Interannual and Interdecadal Variations ...........................................13 4.1 Wavelet Analysis ............................................................................13 4.2 Interannual Variation ......................................................................13 4.2.1 SST Variance .............................................................................................. 14 4.2.2 Spatial–Temporal Evolution ....................................................................... 14. 4.3 Interdecadal Variation .....................................................................15 4.3.1 SST Variance .............................................................................................. 15 4.3.2 Relationship of PDO, NPGO, and AMO, and SENP SST ......................... 15. Ch5 Mechanisms of SENP Warming .......................................................18 5.1 Mixed Layer Heat Budget Analysis ................................................18 IV.

(6) 5.2 Wind–evaporation–SST ..................................................................20 Ch6 Impact of SENP on ENSO and TC ...................................................21 6.1. ENSO:. Meridional. Dipole. Cross-Equatorial Flow in the. of. SSTA. and. Associated. Tropical Eastern Pacific in. Terminating the 2014 El Niño Development ........................................21 6.1.1 Comparison between 2014 Conditions and the 1997 El Niño Event.......... 21 6.1.2 Role of the SSTA Dipole and the Cross-Equatorial Flow in the EP .......... 23 6.1.3 Numerical experiment ................................................................................. 25. 6.2 Tropical cyclone: Distinct effects of the two strong El Niño events in 2015–2016 and 1997–1998 on the western North Pacific monsoon and tropical cyclone activity: Role of subtropical eastern North Pacific warm SSTA ...........................................................................................28 6.2.1 Distinctions between 2016 and 1998 .......................................................... 29 6.2.1.1 Summer monsoon and TC activity in the WNP....................................... 29 6.2.3 Numerical experiments ............................................................................... 33. Ch7 Conclusion and Discussion ...............................................................36 Reference ..................................................................................................40 Table Captions ..........................................................................................48 Figure Captions ........................................................................................51. V.

(7) Ch1. Introduction About 71 percent of the Earth’s surface is water-covered, and the oceans hold about 96.5 percent of all Earth’s water. Moreover, water has a high specific heat capacity, therefore, sea surface temperatures (SSTs) are always considered important heating source forcing the atmospheric circulation, especially in low-latitude area. Several SSTs in specific region, such as ENSO (Rasmusson and Carpenter 1982; Philander 1990; McPhaden et al., 2006) and PMM (Chiang and Vimont, 2004; Chang et al., 2007), were found have a great impact on our weather and climate. Fig. 1 shows the SST variance from 1950–2010, which can reflects the amplitude of SST variation. There are two larger SST variance signal in the low-latitude Pacific; one is in the tropical Pacific and another is in the subtropical eastern North Pacific (SENP).. 1.1 Oceanic and Atmospheric Background in Tropical Pacific Because of the different angle of incidence of the sun and isolation duration, the net heat flux of earth exhibits a north-south asymmetric pattern and tropic absorbs the lion’s share of solar radiation. The asymmetric heat flux and earth rotation further causes three-cell meridional atmospheric circulation (Hadley, 1735; Ferrel, 1856; Persson, 2006). Additionally, the land-sea contrast also makes non-uniform SST and atmospheric circulation in different regions. The tropical SST is non-uniform, which is warmer in Indian Ocean, western Pacific, and western Atlantic Ocean. The low-level wind over equator is easterly (westerly) in Pacific and Atlantic Ocean (Indian Ocean). The mean state of large scale circulation and SST structure is closely related to the variation of SST. One of the most important SST variations, ENSO, consists of warm phase (El Niño) and cold phase (La Niña), is located in eastern tropical Pacific. It’s an air–sea coupled mode that the SST warming accompanying strong westerly. The SST warming forced an anomalous walker circulation which is ascending in eastern Pacific and descending in western Pacific. The anticlockwise circulation further enhances the low-level 1.

(8) westerly. In the ocean, the warm SST deepens the thermocline and the westerly wind stress weakens the upwelling, which has a feedback to enhance the SST warming. This atmosphere and ocean coupled system is an important mechanism for the development of ENSO (Bjerknes, 1969). Another feature of ENSO is phase locking. It usually initials in the boreal spring, develops in summer to autumn, mature in the winter, and rapidly decay in its following spring. The life cycle and phase locking of ENSO is closely related to the annual cycle of our climate and be mentioned in lots of studies (Neelin et al., 2000). Besides the special features of ENSO itself, the impact and interaction of ENSO and our climate is also an interesting issue, such as the relationship of ENSO and monsoon, tropical cyclone (TC), and temperature (Ropelewski and Halpert, 1986; Torrence and Webster, 1999; Chia and Ropelewski, 2002).. 1.2 Oceanic and Atmospheric Background in Subtropical Pacific In addition to the tropic, subtropics are another interesting region where the strong atmospheric and oceanic meridional structure is. The warmer SST is in SENP but the colder is in subtropical eastern South Pacific. The precipitation is shown similar structure with more precipitation in SENP, which is also known as Intertropical Convergence Zone (ITCZ, Waliser and Gautier, 1993). It is noted that the asymmetric SST associated with precipitation accompany with cross-equatorial flow is strongest in boreal summer, i.e., June to August. The cross-equatorial flow is an unstable mode of the air–sea coupled system in the tropical eastern Pacific (EP), which is closely associated with the annual cycle and the geometries of continents (Chang and Philander 1994; Li and Philander 1996).. Besides the asymmetric climatology in EP, another meridional SST variation is also found which is called Pacific merdional mode (PMM). PMM is an air–sea coupled leading mode without ENSO signal and warming trend, which is characterized with tilted warm SST in the SENP and cold SST in the equatorial EP accompanying a 2.

(9) cross-equatorial flow (Chiang and Vimont, 2004). The cross-equatorial flow associated with PMM is an anomalous southwesterly in SENP, which is against the trade wind and then reduces the evaporation and warms the SST. The warm SST then has feedback to enhance the wind. This is also called wind–evaporation–SST (WES) mechanism (Xie and Philander, 1994). Several studies demonstrated the PMM might occur prior to El Niño (Chang et al., 2007; Di Lorenzo et al., 2015). The correlation of boreal spring PMM index and winter cold tongue index is 0.68. These statistical results are more robust than the MJO–ENSO relationship (Slingo et al., 1999; Hendon et al., 1999). Recently, Yu et al. (2011) further suggested that forcing from the extratropical atmosphere might be particularly important to the generation of the central Pacific (CP) El Niño. Additionally, CP El Niño also could increase the PMM amplitude (Stuecker, 2018).. These SSTs variation have great impact on our weather and climate. To understand their mechanism and how to affect the climate could improve our climate forecast. It is noted that the large variance is also found in SENP region. Although the region is similar to that of PMM, it shows several differences between PMM and SENP, the 2014 El Niño in particular. Our research found that the SENP warm SST in summer of 2014 is associated with the bad development of El Niño. The detail will further descript in the section 6.. 1.3 Motivation In the boreal spring of 2014, the oceanic and atmospheric conditions were favorable for an El Niño’s development. Scientists and state-of-the-art models predicted that in 2014 a super El Niño would initiate. The super El Niño only occurred two times in the record with accurate observation and had great impact on global climate. Therefore, it drew lots of attention to this event. However, the growth rate of the sea surface temperature anomaly (SSTA) in the equatorial EP suddenly declined in the boreal summer. In the end of the year, the El Niño was not 3.

(10) materialized. The possible physical processes responsible for hindering the 2014 El Niño’s development have been noted in several studies (e.g., Menkes et al., 2014; Min et al., 2015; Hu and Fedorov, 2016; Zhu et al., 2016). Among these studies, the unusual warming SST occurred in SENP was also found and had possible impact on hindering the 2014 El Niño event. The region of SENP warm SST is similar to that of PMM, however, the value of SENP SST is high and the PMM index is near normal in the summer of 2014. It indicates the warm SENP SST in the summer of 2014 may not affect by PMM. Previous studies calculated PMM index without ENSO signal and warming trend. We found that the impact of warming trend on SENP region is increase in recent decade. Moreover SENP SST not only exhibits the interannual but also the interdecadal variation. Although several studies mentioned that the PMM or CP index also have interdecadal variation (Sullivan et al., 2016). It is still need more analysis to explain the relationship of SENP, PMM, and warming trend. It is also noted that the interaction among tropical, subtropical, and high-latitude. A series of works are focused and discussed about this issue. If we start from tropic, ENSO is the most famous climate variability, which is further separated into CP El Niño and EP El Niño. The EP El Niño affects the Aleutian low (AL) by atmospheric bridge (Alexander et al., 2002); AL then influence the PDO. The CP El Niño affects the NPO which then affects NPGO and central Pacific warming. PDO and NPGO are regarded as large scale interdecadal variation and the coverage include tropic to high-latitude. It is suggested that PDO and NPGO may influence the variation of SENP SST. Besides the impact of warm SENP SST in 2014 on El Niño, this SST also influences TC activity. One dramatic example is in 2015; both the genesis position of TC in western and eastern North Pacific is close to the International Date Line. This location shift is associated with the positive PMM-like SST (Murakami et al., 2017; Hong et al., 2018). In addition to TC genesis position, Zhang et al., (2016) pointed that the positive PMM phase favors the occurrence of TC in the WNP while the 4.

(11) negative PMM phase inhibits the occurrence of TC. There are lots of studies associated with PMM and its impact, however, it remains a question is the SENP SST and PMM the same phenomena? If not, what is the relationship between the SENP SST and PMM? In addition to PMM, does other factor influence the SENP SST? We found that the SENP SST persisted almost positive from 2013 to present. Many studies suggested that it is associated with the recent climate variability. Such as we mentioned before, Wu et al. (2018) pointed that the warm SSTA in boreal summer (June to August, JJA) hindering the El Niño development in 2014. The genesis location of TC closed to the International Date Line both in western and eastern North Pacific (Murakami et al., 2017; Hong et al., 2018). The warm SST is also observed in 2015/2016 and the TC activity is quite different to the 1997/1998 in the El Nino decaying year. The persisted warming of SENP suggested the SST might exist interdecadel variation which is less discussed. In this study, we aim to understand the characteristic of SENP SST in the interannual and interdecadal timescale. Previous studies calculated PMM index after removing the ENSO component. However, we found the ENSO extended SST also have an impact on the SENP SST. To understand the interannual SENP warming structure and its connection with tropic and mid-latitude is one of our scopes. On the other hand, PDO and NPGO affects the Pacific climate in interdecadal timescale are addressed in many studies. The relationship of SENP SSTA and PDO/NPGO and how does the PDO/NPGO affects the SENP SST are other interesting questions. Furtherly, we also analyze the impact of SENP SST on the East Asia climate.. 5.

(12) Ch2. Data and Method The datasets employed in this study comprised: (1) the monthly SST data from the Met Office Hadley Centre sea ice and SST data sets (Rayner et al., 2003); (2) the (1/4)° daily SST data from optimum interpolation sea surface temperature (OISST) version 2 (Reynolds et al., 2007); (3) the monthly SST data from Extended Reconstructed Sea Surface Temperature (ERSST) version 3b (Smith et al., 2008); (4) the zonal and meridional wind of ERA-Interim reanalysis (Dee et al., 2011); (5) the ocean reanalysis data including the ocean currents and temperature from the National Centers for Environmental Prediction (NCEP) global ocean data assimilation system (GODAS) (Behringer and Xue, 2004). It has (1/3)° latitude-longitude grid within 10°S–10°N and 10 m vertical resolution in the upper ocean; (6) the ocean reanalysis data including the ocean currents and temperature from the ECMWF ocean reanalysis system (ORAS4) (Balmaseda et al., 2013); and (7) the National Centers for Environmental Prediction (NCEP)/National Center for Atmospheric Research’s Reanalysis data set (Kalnay et al., 1996). The 3 SST data are used in this study due to different part of the work. What dataset is used is further being mentioned in the figure caption.. Two models are employed in this study. One is FGOALS-g2 (Li et al. 2013), which is a climate system-coupled general circulation model (CGCM). We used FGOALS-g2 to test the role of the north–south dipole of SSTA in preventing the 2014 El Niño development. Another is ECHAM5 (Roeckner et al., 2003), which is an atmospheric general circulation model (AGCM). It is used to infer the relative influence of the ENSO-like and PMM-like SSTA on the WNPAC. The details of the experimental design are described in section 6.1.3 and 6.2.3, respectively.. The climate indices used in this study include (1) PMM (Chiang and Vimont, 2004), 6.

(13) (2) PDO (Zang et al., 1997), (3) NPGO (Di Lorenzo et al., 2008), and (4) AMO (Enfield et al., 2001). The data are downloaded from NOAA earth system research laboratory and the homepage of Emanuele Di Lorenzo (http://www.ord.org/npgo).. 7.

(14) Ch3. Spatial–Temporal Characteristics of Subtropical Eastern North Pacific SST In this section, we will show the characteristics of SENP SST, which seems to be more influential in recent year. How we define the SENP index and what is its spatial–temporal pattern is described in the first two sections. In section 3.3, we show the similar and difference between SENP and PMM. Finally, the effect of warming trend on SENP is illustrated in section 3.4.. 3.1 SST Variance, Background, and SENP Definition Fig.1 depicts the SST variance during 1950–2010. The strong variance are found in various region, such as equatorial CP to EP, mid-latitude northern Pacific around 40°N, SENP from Baja California to south of Hawaii, and western North Atlantic ocean around 50°N. The large variance in tropical EP is associated with ENSO mode that is introduced in many previous studies. The mid-latitude variance in Pacific and Atlantic Ocean may relate to extended Kuroshio and Gulf Stream, respectively. It is noted that the relative large variance is in SENP region. Previous studies linked the SST warming with the anomalous southwesterly in SENP to an air–sea interaction phenomena, that is, PMM. Chiang and Vimont (2004) pointed that the anomalous southwesterly against the trade wind, therefore, the reduced latent heat warm the SST. However, PMM is calculated without warming trend and ENSO signal. It is suggested that the variance in SENP region may not only contribute by PMM. Fig.2 shows large scale background of Pacific, the climatology of SST, SLP, and 850 hPa wind is displayed as shaded, contour, and vector, respectively. The SST in EP is asymmetric, which the maximum is located in North Pacific around 10°N. It is also found that the Pacific subtropical high, which the center is in 140°W, 30N. The SENP region, also as the southeast edge of subtropical high in North Pacific, accompanies the trade wind (northeasterly). The large scale background, location of 8.

(15) maximum SST and subtropical high, may associate with the SENP warming and will be described later. To analyze the variation of SENP SST completely, the SSTA averaged over 105°W–145°W, 15°N–30°N are defined as SENP index. Fig.3 shows the time series of SENP from 1910–2010. The interannual and interdecadal variation is clear seen and the amplitude is larger in the later period than it is in prior period. The larger amplitude may be associated with interdecadal variation, for example, SENP is always positive after late 1970s. Besides interdecadal variation, warming trend also influence SENP SST. The upward linear trend is found in Fig.3 approximately 1°C/100yr. Particularly, the amplitude are reached a new high after 2013. The larger and larger SENP SST is suggested that it may have more influence on global climate.. 3.2 Spatial–Temporal Evolution We displayed the variance and background of SENP in section 3.1. In this section, we will show its spatial–temporal evolution. The lead-lag regression map of SENP and SSTA (shaded) and low-level wind (vector) is depicted in Fig. 4a (SENP lag 8 month) – Fig.4i (SENP lead 8 month). Fig. 4e shows the simultaneous regression map that the maxima SST accompanying with southwesterly wind is clear seen in the SENP region. The southwesterly wind is part of a cyclonic circulation locating in north Pacific. It can also see a cold SST to the northwest of the SENP warm SST. The cold and warm SST may associate with the cyclonic circulation, which are descripted later. In addition to extratopic, a warm SST and westerly exist in the central Pacific. The tropical and extratopical SST develop and decay are simultaneously. Before the maximum SENP SST 7 months, the weak warm SST exists in SENP and CP and the cyclonic circulation and westerly are also seen. The warm SST and wind become stronger gradually few months later. The SST evolution shows that the larger SSTs are exist initially in SENP and CP; then the SSTs of these two regions become stronger and seem connected each other. When the SENP SST decays (Fig. 4f–i), the CP SST become weaker, too. 9.

(16) The SST and wind change simultaneously indicate that it may be a coupled system. Previous studies demonstrated that the SST and wind in tropical Pacific exist a close relationship. The well relationship is also seen in our result of regression. Besides the tropic, the cyclonic circulation over north Pacific shows the good relationship with the SENP SST. The cyclonic circulation forces a southwesterly wind in SENP region and enhances the warm SST. After the mature phase (Fig .4e), the cyclonic circulation disappears quickly and then the SENP SST decays gradually. It is suggested that the SENP warm SST is related to the atmospheric circulation and tropical Pacific SST. The SST pattern is partly similar to the PMM structure. We will show the comparison between SENP and PMM in the next section and how does the cyclonic circulation affect the SENP SST in the next chapter.. 3.3 Comparison between SENP and PMM We mentioned that the PMM is calculated without warming trend and ENSO signal, which is related to SENP SST. Although the SENP index is similar to PMM, it exhibits several differences. Fig.5a (5b) depicts the regression map of PMM (SENP) and, SSTA and 850 hPa wind from 1950–2010. The tilted warm SST from Baja California to south of Hawaii accompanying southwesterly is both in PMM and SENP pattern. It is also found a cyclonic circulation in North Pacific and the southwesterly wind in SENP region seems part of the cyclonic circulation. However, it exhibits several differences in tropic Pacific. The regression of PMM is negative SST in EP and warm SST in near 175°E. The easterly is also in EP and merges with southwesterly in SENP region. As to the regression of SENP, the El Niño-like SST with westerly is found in the CP. The result is consistent with the SST evolution in section 3.2 and is suggested that the SENP SST may be related to the tropical SST. Fig.5c shows the time series of PMM (red) and SENP (black) from 1950–2010. The trend of SENP and PMM is quite similar and the correlation coefficient is 0.58, however, but it is still difference in some year. For example, the SENP is positive in 1997, but the PMM is negative. It is also noted that the 10.

(17) relationship is better in prior period than it is in later period. Particularly, SENP and PMM exhibit obvious difference in recent decade. The interannual variation is more significant in PMM. Another discrepancy between SENP and PMM is in the seasonal variance. Fig.6 is the variance of SENP and PMM from January to December from 1950–2018. The maximum variance of PMM is in boreal spring, which is consistent with previous studies. The maximum variance of SENP is later, which the maximum is in June. The later maximum variance of SENP and different regression map suggests that the SENP SST may relate to PMM, but not totally the same. Notably, the variance of SENP and PMM changed before and after the late 1980s. It is suggested that the characteristic of SENP and PMM may different in various periods.. 3.4 Effect of Warming Trend The global warming is a hot issue and effect different way of our nature. Fig.7 shows the linear trend of SST from 1950–2018. The trend exhibits warming in most of area but it is not uniform. It is noted that the large warming in Indian Ocean, western Pacific, and Atlantic Ocean. In EP, the warming trend is stronger in subtropics; one of them is SENP region. The linear trend of SENP is 1 K/100yr and also shows in the Fig.3. To remove the warming trend effect on SENP, the SST variance without linear warming trend are calculated (Fig. 8a). The pattern is similar to the result calculating with raw data, except the strength. Fig. 8b shows the difference of variance between the SST without and with the linear trend. Most of areas are positive, which is indicating the linear trend would enhance the SST variation. Where the larger variance is, such as Indian Ocean, western Pacific, and Atlantic Ocean, are becoming weaker, except the equatorial CP to EP. It is suggested that the variation in the equatorial CP to EP is slightly or not affected by the warming trend. However, the SST variance in SENP is weakening almost 15%. The SST variance of SENP with and without linear trend is 0.35 and 0.41, respectively (Fig. 8c). 11.

(18) The warming trend might enhance the SST variance in SENP region. Moreover, the global warming trend is larger in recent period than it is in prior period. It is suggested that the impact of warming trend on SENP SST is stronger in recent decade or in the future. The SST warming pattern exhibits a significant signal in SENP region in CMIP5 model result. More detail of SENP warming in CMIP5 model will discuss in the section 7.. 12.

(19) Ch4 Interannual and Interdecadal Variations We have showed the characteristic of SENP SST and its relationship with PMM and linear warming trend. In this chapter, we will demonstrate that the SENP SST exists different timescale variation. The wavelet analysis is showed in the section 4.1; the interannual and interdecadal variation will show in 4.2 and 4.3, respectively.. 4.1 Wavelet Analysis Fig. 9 depicts the (a) normalized time series, (b) wavelet analysis, and (c) power spectrum of SENP SST from 1950–2010. The normalized SENP SST exhibits clearly interannual variation and reaches a new high after 2013 (Fig. 9a). It also shows the interdecadal variation, which is not noticed before (black curve, Fig. 9a). The power spectrum analysis further shows three significant cycles, that is, 2–8 years, 8–16 years, and up to 20 year. Because the data length, we only focus on the interannual and interdecadal timescale. Wavelet analysis not only shows the significant frequency, but also shows that when is the significant frequency. Fig. 9b shows that the larger signals are more than 2 years. One is in 2–8 years, which the strongest signals are before 1980 and after 1995. Another is in 8–16 years, which is significant in whole period and become stronger after late 1990s. These results indicate that the SENP SST variation possibly had more influence in recent decade. The SENP SST will be separate into interannual (0–8 years) and interdecadal (more than 8 year) time scale in the following sections. 4.2 Interannual Variation The result of power spectrum and wavelet analysis showed that the SENP SST exhibits interannual variation. The 0–8 years filtered SST is used to analyze the interannual variation.. 13.

(20) 4.2.1 SST Variance Fig.10 depicts the interannual SST variance during 1950–2010. The variance in several regions is similar to that of raw data. The tropical, North Pacific, and North Atlantic Ocean exhibit strong variance in interannual time scale. It indicates that the strong interannual variation exist in those region, for example, ENSO exhibits significant interannual variation. In the SENP region, the variance is slightly weaker in interannual time scale, but it is still relatively larger than the variance in other subtropical regions. The relatively larger variance in SENP suggested that the SENP SST also exbibits interannual variation. The result is consistent with the interannual variation of PMM. Other regions where the interannual variance is weak, such as, tropical western Pacific, tropical western Atlantic Ocean and tropical Indian Ocean, suggested that the SST do not exhibit the interannual variance.. 4.2.2 Spatial–Temporal Evolution The spatial–temporal evolution of SENP SST is showed again but in interannual time scale (Fig.11). The result is similar to that we showed in section 3.2, however, it also exhibits several differences, which we will descript in this section. In interannual time scale, the cold North Pacific SST, warm SENP and CP SST still occur at the same time simultaneously (Fig. 11e). The SSTs accompany with an anticyclone and westerly also be found in SENP and CP. These SSTs and wind are found several months before the maximum SENP SST. The SENP and CP SST develop gradually to its peak phase (Fig. 11a–e) and then decay (Fig. 11f–i). Notably, the interannual SENP SST is weaker in its decay phase than that of raw data. The anticyclonic circulation seems to not disappear but moves northward. The. 14.

(21) CP SST is also decay quickly and there is no SSTA in tropic when the SENP SST lags 8 months. The regression of CP SST (105°W–145°W, 15°N–30°N) and SST are showed in fig.12. It is suggested that the CP SST may enhance the SENP SST warming.. 4.3 Interdecadal Variation The result of power spectrum and wavelet analysis showed that the SENP SST exhibits interdecadal variation. The up to 8 years filtered SST is used to analyze the interdecadal variation.. 4.3.1 SST Variance Fig.13 depicts the interdecadal SST variance of sea surface temperature during 1950–2010. The variance of SENP (tropical EP) SST is stronger (weaker) in the interdecadal time scale than it is in interannual time scale. These results indicate that the SENP (tropical EP) SST exhibits stronger (weaker) variation in interdecadal time scale. Several regions also exist large SST variance, such as, central North Pacific, the coastline of North America and East Asia, and tropical Indian Ocean. The SENP SST variation seems associated with three regions, that is, mid-latitude North America coastline, central North Pacific, and tropical CP. The SST variance structure in mid-latitude North America coastline is similar to SSTA pattern of NPGO. The North Pacific SST variance may associate with PDO. Sullivan et al. (2016) also demonstrated that the CP El Niño exist an interdecadal variation. In the next section, we will show the relationship between these interdecadal indices and SENP SST.. 4.3.2 Relationship of PDO, NPGO, and AMO, and SENP SST The interdecadal SST variance is significant in SENP region; therefore, the. 15.

(22) regression map of SENP index and SST and 850 hPa wind is calculated and showed in Fig. 14. The regression of SSTA and 850 hPa wind also exhibits two special structure like interannual SST pattern, one is CP warming with westerly wind, and another is the tilted SST warming with a southwest wind on the SENP region. It is also noted that this subtropical warm SSTA is extended along the coast and there is a cold SSTA in the west accompanying cyclonic circulation. This pattern is similar to the correlation map of SSTA on PDO (Fig. 15a) and NPGO (Fig. 15b) suggesting that PDO and NPGO might be associated with the variation of SENP SST. Moreover, previous studies also showed that the AMO have influence on Pacific Ocean. Therefore, we analyzed the possible impact on SENP region in interdecadal time scale. Fig.15 depicts the correlation map of SSTA and (a) PDO, (b) NPGO, and (c) AMO from 1950–2010. The PDO pattern shows the ENSO-like SSTA pattern but the correlation coefficient is stronger in subtropical than it is in tropic, especially in SENP region. The cold SSTA also be found in the central north Pacific. The similar structure between SENP regression map and PDO correlation map suggests that PDO have an influence on SENP SST variation. The NPGO pattern shows the tilted SST from equatorial CP to SENP region and a north–south dipole structure in North Pacific. It is noted that the correlation map of SSTA and NPGO is multiply by -1 for compare conveniently. The larger correlation coefficient in SENP of NPGO is not totally the same with that of PDO. It extends closely to the equatorial central tropical Pacific but weaker in the coastline of western America. Fig. 15c also shows negative correlation in SENP and tropical EP, although it also exhibits positive correlation in small region. The correlation result suggested that these climate indices might have impact on SENP. PDO, NPGO and, AMO all have impact on SENP, however, what is the relative influence? To answer this question, we future calculated the partial correlation of SST and these indices (Fig. 16, Baba et al., 2004). The partial correlation of PDO here (fig. 16a) indicates the correlation of SST and PDO without the influence of NPGO and 16.

(23) AMO. The PDO and NPGO structure in SENP is unchanged, but the AMO structure in SENP is from negative to positive. The result indicates that PDO, NPGO, and AMO are related to SENP SST in interdecadal time scale. Particularly, the positive AMO may induce positive SST in SENP. The difference pattern of AMO in correlation and partial correlation map suggested that the AMO might also relate to the PDO and NPGO.. 17.

(24) Ch5 Mechanisms of SENP Warming In this chapter, we aim to understand the mechanism of SENP warming. The mixed layer heat budge analysis is showed in section 5.1. The mechanism of SENP warming by wind–evaporation–SST is shown in section 5.2.. 5.1 Mixed Layer Heat Budget Analysis The mixed layer temperature tendency equation can be written as (Li et al. 2002; Hong et al. 2008; Wang et al. 2015; Chen et al. 2015, 2017b). Q'net ∂T ' ' +R = −(V ⋅∇T ) + ρ CP H ∂t. (1). Where T denotes the mixed layer temperature, V = (u, v, w) denotes the 3D ocean current (i.e., the vertical average from the surface to the bottom of the mixed layer), ∇ = (∂ / ∂x, ∂ / ∂y, ∂ / ∂z ) denotes the 3D gradient operator, ( ) ' denotes anomaly. variable departure from the climatological mean, and −(V ⋅∇T ) ' denotes anomalous 3D temperature advections. Furthermore, Qnet represents the summation of the net heat flux, R represents the residual term, ρ (= 103 kg m −3 ) represents the density of water, CP (= 4000 J kg K −1 ) represents the specific heat of water, and H represents the mixed layer depth. A positive heat flux indicates heating to the ocean. H is defined as the depth at which the temperature is 0.8 K lower than the SST, following Wang et al. (2012). The fig. 17 shows the lead-lag regression map of the tendency of SENP and total advection from 1980–2014. Fig. 17a–17i represents the SENP tendency lags 6 months to leads 2 month. The regression result is fragmented and insignificant indicating that the advection is not the main warming term. Before SENP SST matures 6 to 1 months (fig. 17a–fig. 17f), the ocean advection is fragmented. Several small signals, including positive and negative advection, are 18.

(25) found in the SENP region. When the advection lag 0 to 1 month (fig. 17g–fig. 17h), the positive advection appears along the coastline. It is suggested that the ocean advection may partially contributes to SENP warming along the coastline. It is consistent with the evolution of SENP warming; the relative large SST warming is also found near the coastline (fig. 4). A further analysis, box averaged ocean advection, is depicts in the fig. 18. The advection averaged from SENP region (black) is small, but it is relative large averaged from coastline (red). The regression is largest when advection leads 0–3 months. In addition to oceanic advection, the atmospheric heat flux is another factor which may warm the SST. Fig. 19 is the same as fig. 17 but the advection is replaced by total atmospheric heat flux and 850 hPa wind. The total heat flux here is the sum of short wave radiation, long wave radiation, latent heat flux, and sensible heat flux at the surface. The downward direction is positive, which indicates the positive value of heat flux warm the SST. The result is suggested that atmospheric heat flux is an important factor warming the SENP SST. When the heat flux leads the SENP tendency 1–0 months, the heat flux is almost positive in the SENP region, which is heating the SST. The positive heat flux is associated with the cyclonic circulation over 20°–30°N. The 850 hPa wind is not significant when the heat flux leads the SENP tendency 6–4 months. In following 2 months, the anti-cyclonic circulation is found accompanying with negative heat flux. When heat flux leads 1 month, the positive heat flux is in SENP and it becomes stronger in next 2 months. In fig.19g, the maximum positive heat flux accompanies the strongest cyclonic circulation with the southwesterly wind is located in the SENP region. Fig. 20 shows the lead–lag regression of SENP index and atmospheric heat flux (black), short wave radiation (blue), long wave radiation (green), latent heat flux (red), and sensible heat flux (light blue) in SENP region from 1980–2014. It is consistent with the regression map result, the heat flux heat the SENP SST when it leads 0-9 months. The time series of heat flux is similar to that of latent heat flux (red), which 19.

(26) indicates the latent heat is mainly factor warming the SENP SST.. 5.2 Wind–evaporation–SST The relationship between SST and the detail of heat flux is further analyzed. The result shows that the latent heat flux is the important factor, which warms the SST. Fig. 21 shows the lead-lag regression map of SENP tendency and latent heat flux. The pattern of latent heat flux (fig. 21) is similar to that of total net heat flux (fig. 19). In fig.21g, the maximum positive latent heat flux accompanies the strongest cyclonic circulation with the southwesterly wind is located in the SENP region. The climatology of low-level wind is the northeasterly wind in SENP and shows in Fig. 2. The background of low-level wind may reduce the latent heat flux to go to the atmosphere by Wind–evaporation–SST mechanism (Xie and Philander, 1994). The regression map shows a cyclonic circulation which is southwesterly wind in the SENP. This southwesterly wind against the climatological northeasterly wind and reduce the wind speed, therefore, the latent heat flux warm the SST. The warm SST further enhances the southwesterly wind. The SST warming in SENP is through Wind–evaporation–SST mechanism, which is similar to that of PMM. This is associated with the background of low-level wind. The SST pattern of global warming also exhibits relative large warming in SENP. It is suggested that the SENP warming may be more important. We will discuss it in section 7.. 20.

(27) Ch6 Impact of SENP on ENSO and TC We will show the impact of SENP SST on ENSO and tropical cyclone (TC) activity in the two case studies, which hinders the development of 2014 El Niño (Wu et al., 2018a) and enhances the 2016 TC activity in western North Pacific (Wu et al., 2018b). Notably, the definition of SENP region is similar but not totally the same because it is case study. However the result is similar which is suggested the SENP SST is not sensitive. 6.1 ENSO: Meridional Dipole of SSTA and Associated Cross-Equatorial Flow in the Tropical Eastern Pacific in Terminating the 2014 El Niño Development 6.1.1 Comparison between 2014 Conditions and the 1997 El Niño Event Fig. 22 presents a Hovmüller diagram (averaged over 2°S–2°N) of the SSTA (contour) and SSTA tendency (shaded) in 1997 and 2014. During the late boreal winter to the ensuing spring (February–April), the Niño 3 SST growth rates in 2014 and 1997 were approximately 0.35 and 0.45 K/month, respectively. The SSTA growth rate in 2014 was pronounced and was similar to that in 1997. The SSTA grew continually in the boreal summer of 1997; however, the SSTA growth terminated and SSTA tendency rapidly changed from positive to negative (−0.1 K/month) in the boreal summer of 2014. Such a decline in SSTA tendency is responsible for the prevention of El Niño development in the boreal summer of 2014. Fig. 23 depicts the spatial distributions of the anomalous SST and low-level wind in the boreal summer [i.e., June–August (JJA)] of 1997 and 2014. A significant positive SSTA in the equatorial EP and pronounced westerly wind anomalies in the equatorial western Pacific were clearly identified in 1997 (Fig. 23b). In contrast, a weak positive SSTA and westerly anomaly were observed in 2014 (Fig. 23a). Among the differences between 1997 and 2014 conditions, the north–south dipole of SSTA in 2014 [i.e., the 21.

(28) positive (negative) SSTA in the northeastern (southeastern) subtropical Pacific] was particularly notable (see the boxes of ENP and ESP in Fig. 23a). This distinction was further evidenced by the meridional profiles of SSTA in the JJA of 1997, 2014, and the composite El Niño (Fig. 24). The SSTA in both 1997 and the composite El Niño were symmetric to the equator; that is, the meridional SSTA gradient between the northeastern pole and southeastern pole was approximately zero (bars shown in the right upper corner of Fig. 24). However, a pronounced SSTA asymmetric to the equator was observed in 2014. The SSTA dipole was accompanied with an enhanced cross-equatorial flow, starting from ESP and ending at the ENP, where the climatological ITCZ is approximately located. The cross-equatorial flow in the ENP is an unstable mode of the air–sea coupled system in the tropical EP, which is closely associated with the annual cycle and the geometries of continents (Chang and Philander 1994; Li and Philander 1996). Details of the physical processes responsible for the asymmetry of the cross-equatorial flow were provided by Chang and Philander (1994). The enhanced cross-equatorial flow led to a rainfall surplus in the ENP (not shown), exceeding normal levels by approximately 30%. Further analysis revealed that the anomalous rainfall associated with diabatic heating generated a meridional overturning circulation anomaly, ascending (descending) in the eastern North (South) Pacific (not shown), in which the low-level returning flows had a positive feedback to enhance the cross-equatorial flow. In the following, we show the relationship between the SSTA dipole and the associated cross-equatorial flow. Fig. 25a shows the surface wind stress regressed on the SSTA dipole index in boreal summer (i.e., JJA); here the SSTA dipole index was defined as the SSTA in the ENP minus the SSTA in the ESP (boxes in Fig. 23a). It is clearly shown that the SSTA dipole matched well with the cross-equatorial flow, characterizing by a southeasterly anomaly in the Southern Hemisphere and a southwesterly anomaly in the Northern Hemisphere. A westward low-level wind anomaly in the equatorial central–eastern Pacific was clearly observed during the 22.

(29) positive SSTA dipole phase. As shown in Fig. 25b, the SSTA dipole index and the cross-equatorial flow (whose strength is measured by the 925 hPa meridional wind anomaly averaged over 80°W–140°W, 10°S–10°N) show a positive correlation relationship (the correlation coefficient is approximately 0.65, which is at the 95% confidence level). Fig. 25b also shows that the SSTA dipole index in JJA of 2014 exhibits the highest value during the period 1979 to 2014. All the aforementioned results indicate that the enhanced cross-equatorial flow in 2014 is closely related to the unusual meridional dipole of SSTA in the EP. It is worth mentioning that such cross-equatorial flow comes from not only the contribution of the cold SSTA over ESP but also the contribution of warm SSTA over ENP. Additionally, it is noted that the SSTA dipole index is statistically negative correlated with Niño 3 index (cr = −0.43). This indicates that the oceanic conditions associated with the SSTA dipole might not favor the El Niño development.. 6.1.2 Role of the SSTA Dipole and the Cross-Equatorial Flow in the EP The role of cross-equatorial flow in suppressing the El Niño development through modulating the oceanic conditions was documented in this section. During the boreal spring, a marked positive temperature anomaly along the thermocline was clearly observed in 2014 and 1997 (Fig. 26a, b, respectively). The temperature perturbation in the thermocline was associated with a deepening D20 (i.e., depth at 20 °C) and eastward current anomalies in the upper ocean, providing the optimal oceanic conditions for El Niño development. Oceanic conditions favoring El Niño development were persistent in the boreal summer of 1997 (Fig. 26d); however, the oceanic conditions have changed in the boreal summer of 2014, e.g., the upper oceanic current anomalies turned from eastward to westward in response to the westward surface wind stress anomaly in the equatorial central–eastern Pacific (Fig. 26a, c). This westward surface wind stress anomaly may induce westward and. 23.

(30) upwelling oceanic current anomalies in the equatorial central–eastern Pacific (Fig. 26c), blocking the Bjerknes positive feedback and hence preventing El Niño development. In the following, we conduct the oceanic mixed layer heat budget to reveal how the enhanced cross-equatorial flow hinder El Niño development through regulating the three-dimensional (3D) oceanic currents. Firstly, corresponding to the decrease in the positive SSTA in JJA compared to that in MAM, the magnitude of the negative heat flux become weaker in JJA than MAM (i.e., larger positive SSTA generally corresponds to larger negative heat flux term, and vice versa), the thermodynamic term associated with the surface heat flux makes a negative contribution to the decrease of SSTA tendency (not shown) from boreal spring to summer. Thus, in the following we mainly focused on the dynamic terms, which are essential in determining the SSTA tendency during El Niño development (Li 1997; Jin and An 1999; Chen et al. 2015, 2016a). These important dynamic terms are associated with Bjerknes feedback processes, including the advection of mean temperature by anomalous zonal current ( −u '∂T / ∂x , hereinafter anomalous zonal advection term), the advection of anomalous temperature by mean upwelling −w∂T '/ ∂z , hereinafter mean upwelling advection term), and the advection of mean. temperature by anomalous upwelling ( −w'∂T / ∂z , hereinafter anomalous upwelling advection term). Fig. 27 illustrates a comparison of the heat budget results of 1997 and 2014. The bars, from left to right, represent the mixed layer temperature tendency, the anomalous zonal advection term, anomalous upwelling advection term, mean upwelling advection term, and anomalous meridional advection term. The temperature tendency in the boreal spring of 2014 was approximately 0.35 K/month, showing a comparable magnitude to that of 1997. Here all the four dynamic terms made a positive contribution to the positive temperature tendency. Nevertheless, the four dynamic terms either changed from positive to negative or decreased substantially in the boreal summer of 2014, resulting in negative temperature tendency. In contrast, the positive temperature tendency and associated Bjerknes 24.

(31) feedback processes were sustained in the boreal summer of 1997. The weakening in the oceanic dynamic terms that lead to the termination of positive temperature tendency in the boreal summer of 2014, could be traced back to the anomalous cross-equatorial flow in the boreal summer of 2014. The chain of physical causation is described as follows. (1) The cross-equatorial flow induced westward surface wind stress anomaly near the equator, driving an anomalous westward current ( u ' < 0 ) and upwelling ( w ' > 0 ) in the equatorial central–eastern Pacific. Consequently, negative zonal ( −u ' ∂T / ∂x < 0 ) and vertical ( −w ' ∂T / ∂z < 0 ) temperature advections were produced. (2) D20 shoaled and returned to the climatological level because of the anomalous westward and upwelling currents; the temperature perturbation in the thermocline decreased, and thus the mean upwelling advection term ( −w∂T '/ ∂z ) weakened substantially. (3) The northward current anomaly ( v ' > 0 ) that is also induced by anomalous cross-equatorial flow, transported a substantial quantity of climatological cold water from the South Pacific to the equator ( −v '∂T / ∂y < 0 ), especially over the region east of 90°W (Fig. 28), contributing to a decrease in temperature tendency. This explains how the cross-equatorial flow modulated the 3D ocean currents and then prevented the 2014 El Niño development. In next section, we further investigate such a possible mechanism through a set of numerical experiments.. 6.1.3 Numerical experiment To test the role of the north–south dipole of SSTA in preventing the 2014 El Niño development, a series of numerical experiments were conducted by employing a climate system-coupled general circulation model (CGCM). This CGCM was named FGOALS-g2 (Li et al. 2013) and was applied in the Coupled Model Intercomparison Project Phase 5 (CMIP 5); it showed excellent performance in ENSO simulation among both CMIP3 and CMIP5 models (Chen et al. 2013, 2016b; Bellenger et al.. 25.

(32) 2014; Chen and Yu 2014; Yu et al. 2014). Through an SSTA restoring method (e.g., Luo et al. 2005; Yan et al. 2009, 2010), the SSTA field of the CGCM was nudged to the OISST daily SSTA field at each time step from 1982 to 2014. Through this coupled nudging approach, the model could represent realistic SSTA field and relatively reasonable ENSO-related anomaly fields (e.g., zonal wind stress anomaly and thermocline depth anomaly). Employing the outcome from the aforementioned nudging approach as the initial condition field at a certain time, namely May 1, 2014, the CGCM was freely integrated without the SSTA-nudging scheme (i.e., the nudging scheme is closed when running the CGCM) since May 1, 2014, which is referred to as the control hindcast experiment (hereinafter CTL). In the CTL hindcast experiment, the predicted Niño 3 SSTA increased during May 2014, but decreased since June 2014 (solid blue curve in Fig. 30), indicating that this CGCM is able to reproduce the transition feature of the SSTA (i.e., the SSTA started to decrease from June 2014). Parallel to the CTL experiment, a sensitivity experiment, named “non-dipole”, was performed. All steps in non-dipole were same as those in CTL, except for the generation of the initial condition in the nudging step. Specifically, in the non-dipole experiment, the SSTA field of the CGCM was nudged to a modified SSTA field as shown in Fig. 29b. The modified SSTA field was same as the previously observed daily SSTA field used for nudging in CTL (Fig. 29a), except that the north–south dipole of SSTA was removed during March–April in 2014. Therefore, we obtained a new initial field on May 1, 2014, in which the warming over the eastern equatorial Pacific remained the same as that of CTL, but the meridional asymmetric SSTAs over ENP and ESP were removed (similar to the SSTA pattern in Fig. 29b). Using this new initial field to integrate the CGCM from May 1, 2014, we conducted the non-dipole experiment. Notably, the Niño 3 SSTA increased during May 2014 and continued to grow throughout 2014 (solid red curve in Fig. 30). This SSTA evolution feature in the non-dipole experiment was distinctly different from the SSTA transition feature in the CTL experiment, indicating that the SSTA dipole played a critical role in the 26.

(33) prevention of the 2014 El Niño event. The relative contribution of the southeastern-pole and northeastern-pole SSTAs in hindering the Niño3 SST growth rate was evaluated by two additional experiments, namely, non-ENP-pole and non-ESP-pole. The non-ENP-pole (non-ESP-pole) experiment was conducted in the same way as that for the non-dipole experiment, except that the SSTAs over ENP (ESP) were removed from the target nudging field, as shown in Fig. 29c, d. Details of the design of the numerical experiments are shown in Table 1. The numerical experiment results revealed that the southeastern-pole SSTA alone appeared to be insufficient in terminating the El Niño development (see non-ENP-pole experiment: dashed purple curve in Fig. 30) compared with the result in CTL experiment; the northeastern-pole SSTA had a considerable effect on suppressing the SST growth rate in Niño 3 (see non-ESP-pole experiment: dashed brown curve in Fig. 30). The impact of northeastern pole SSTA on suppressing the El Niño’s growth seems to be comparable to that due to the southeastern pole (comparing the dashed brown curve with the dashed purple curve in Fig. 30). Additionally, the ENP pole’s dynamical role was examined through comparing the non-ENP-pole experiment with CTL experiment (non-ENP-pole minus CTL). It is found that without the ENP pole’s role, the cross-equatorial flow in the summer was suppressed especially to the north of the equator and the easterly wind was weakened near the equator (not shown). This indicates the warm SSTA over ENP plays a role in reinforcing the cross-equatorial flow, which could modulate the oceanic temperature advection terms and regulate the ocean temperature. To sum up, the experiment results indicate that both the warm SSTA over ENP and the cold SSTA over ESP make a contribution to the suppression of El Niño’s development in boreal summer of 2014. It’s worth mentioning that one recent study (Zhu et al. 2016) also focused on the role of off-equatorial SSTA in the 2014 El Niño’s growth. Through conducting a series of hindcast experiments with CFSv2 (Saha et al. 2010; Zhu and Shukla 2013; Zhu et al. 2014), Zhu et al. (2016) found that 40% of the amplitude error (i.e., the 27.

(34) overestimated amplitude) in peak phase of 2014 El Niño is due to the lack of prediction of negative SSTA over southeastern Pacific. In the view of the impact of the negative SSTA over ESP on suppressing the 2014 El Niño’s development, the experiment results in the present study is generally consistent with that in Zhu et al. (2016). In addition, Zhu et al. (2016) also found that the positive SSTA over the tropical western North Pacific (120°E–180°E, 10°S–20°N; hereafter TNWP) region could cause the increase of predicted amplitude (this means the positive SSTA over TNWP may be favorable for 2014 El Niño development). However, such TNWP region is entirely different from the ENP region analyzed in this study. Another relevant study (Su et al. 2014) suggested the termination of 2012 Pacific warming is attributed to the cold SSTA over both the northern and the southern subtropical Pacific. The role of cold SSTA over the southeastern Pacific in suppressing the 2012 Pacific warming is generally consistent with the results in the present study. Despite of cold SSTA during boreal summer of 2012 over the northern subtropical region (which is mainly confined to the west of 125°W), a warm SSTA signal is noted over the region of 80°–125°W, 10°–20°N (see Fig. 4c, d in Su et al. 2014), which overlaps most of the ENP region in this study (80°–140°W, 10°–20°N). Therefore, more in-depth study with idealized numerical experiments may be needed to carefully examine the specific impacts of the different SSTA signals over different northern off-equatorial regions on the special case in 2012. The mechanism of meridional dipole of SSTA terminating 2014 El Niño was presented in the section 6.1. The main results are also illustrated in the schematic diagram (Fig. 31). In next subsection, we will show another case, the TC activity, which was influenced by SENP warm SST. 6.2 Tropical cyclone: Distinct effects of the two strong El Niño events in 2015–2016 and 1997–1998 on the western North Pacific monsoon and tropical cyclone activity: Role of subtropical eastern North Pacific warm SSTA. 28.

(35) 6.2.1 Distinctions between 2016 and 1998 6.2.1.1 Summer monsoon and TC activity in the WNP Fig. 32 provides a comparison of the TC genesis number and summer monsoon intensity in the WNP between 1998 and 2016. We focused on the El Niño decaying summer because the major distinctions between the two strong El Niño events primarily occurred in decaying summer. A positive low-level (850 hPa) streamfunction anomaly (i.e., WNPAC, shading in Fig. 32a) was clearly observed in June, July and August (JJA) 1998. The WNP monsoon trough (black dashed line, Fig. 31a) was weakened and retreated westward to approximately 120°E in response to the WNPAC. In contrast to 1998, a negative streamfunction anomaly was observed in JJA 2016 (Fig. 32b). The monsoon trough was strengthened and extended eastward to 140°E because of the negative streamfunction anomaly in the WNP. The enhancement of the WNP monsoon trough provided a favorable environment for TC genesis, and it possibly resulted in an above-normal TC genesis number in JJA 2016 (Fig. 32c); contrastingly, the weakening of the WNP monsoon trough was unfavorable for TC genesis, and a below-normal TC genesis in JJA 1998 was observed. Among the distinctions, the increase of TC genesis number in August and September 2016 was particularly high, approximately 1 σ higher than the climatological mean. Because the boreal summer was the focus, the JJA mean was considered in the analysis. Overall, the distinctions of TC-genesis number between 1998 and 2016 in JJA, typhoon season mean (JJASON), and even the entire year exhibits similar results (Table 2). In addition to the distinct TC activity, the WNP summer monsoon index (Wang and Fan, 1999), a measure of monsoon strength, exhibited a significant difference between 1998 and 2016 (Fig. 32d). Whereas both events showed consistency in the delay of monsoon onset, the WNP summer monsoon was much stronger in 2016 than in 1998. Particularly, the WNP summer monsoon in 2016 was stronger than the climatological. 29.

(36) mean after August. Because the Niño 3.4 index was approximately the same during both strong El Niño events, the SSTA other than the Niño 3.4 region, is likely a contributor if the oceanic forcing is a candidate for the distinct effects on the WNP’s climate. In the next section, we will demonstrate that the positive SSTA in the SENP is crucial for resulting in these distinctions.. 6.2.1.2 Large scale circulation and SST anomalies A comparison of the SSTA in the decaying summers of 1997–1998 and 2015–2016 revealed that a southwest–northeast-tilted positive SSTA in the SENP, which resembles the PMM, was apparent in 2016 (Fig. 33). This subtropical positive SSTA was accompanied by a southwesterly anomaly in the subtropical eastern North Pacific (~150°W, 20°N). The southwest–northeast-tilted SSTA was not observed in JJA 1998 (Fig. 33). It is evident that the major difference in large scale atmospheric and oceanic conditions between 2016 and 1998 was featured by a cyclonic anomaly in the WNP, which resembles the southwest–northeast-tilted PMM-like SSTA. Because the PMM is an air–sea coupling mode phase-locked with the annual cycle (Chiang and Vimont, 2004), singular value decomposition (SVD) of the covariance of the SSTA and low-level wind in the region 80°–180°W, 30°S–40°N was analyzed to investigate the possible air–sea coupling processes correlated with these distinctions. The leading mode exhibited an El Niño-like pattern, positive equatorial eastern SSTA-associated equatorial westerly anomalies in the tropical eastern Pacific (Fig. 34a, referred as ENSO mode hereafter). The second mode resembled the PMM structure [the correlation coefficient between PMM index (Chiang and Vimont, 2004) and PC2 is 0.91], a meridional dipole-like SSTA accompanied by a cross-equatorial flow in the eastern Pacific (Fig. 34b, referred as PMM mode hereafter). These two leading models explain approximately 88% of the total covariance of the SSTA and low-level wind. The time series of the expansion coefficient of the leading and second modes (referred to as PC1 and PC2 hereafter) revealed that PC1 had approximately the same 30.

(37) amplitude during the lifecycle between 1997–1998 and 2015–2016 (Fig. 34c, 34e). The amplitude of ENSO (PC1) was nearly identical between the two strong events, but PMM (PC2) exhibited a contrasting characteristic: a positive anomaly in 2015– 2016 but negative in 1997–1998 (Fig. 34d, 34e). Because the air–sea coupling structure in the Pacific was primarily determined by the first two SVDs, the aforementioned results indicated that the PMM substantially contributed to the differences in large scale oceanic and atmospheric conditions, particularly in the central and eastern north Pacific, between 1997–1998 and 2015–2016.. 6.2.2 Effect of PMM associated SST on the distinctions 6.2.2.1 Regression analysis The regression of the SSTA and 850 hPa wind on ENSO and PMM was analyzed to investigate the relative effects of the ENSO and PMM on the oceanic and atmospheric anomalies in the WNP, respectively, during the El Niño decaying summer. Because ENSO and PMM typically peak in winter and spring respectively, the lag regression was calculated respective to the peak phase. The lag regression reveals that ENSO is lag-correlated with the WNPAC (Fig. 35a). The spatial distribution of SST and low-level wind anomalies, WNPAC is accompanied by a positive (negative) SSTA occurs in west (east) of the center WNPAC, indicates that the air–sea interaction is crucial in maintaining the WNPAC (Wang and Zhang, 2002). By contrast, PMM is correlated with a cyclonic circulation anomaly in the WNP and a westerly anomaly in the off-equator region of the CP and WNP (Fig. 35b). The lag regression indicates that the PMM-related cyclonic circulation and westerly anomalies in the WNP might offset, at least partially, the ENSO-related WNPAC. Because PMM showed contrasting trends between 1997–1998 and 2015–2016, the WNPAC is expected to be enhanced in 1998 and decreased in 2016 due to PMM. The differences in the SSTA and low-level wind anomalies in the decaying summer between the two strong El Niño events resembled PMM (pattern correlation coefficient in the ENP was approximately 0.82; Fig. 35b and 35c), indicating that the distinctions in the WNP 31.

(38) monsoon and TC activity between the two El Niño events possibly resulted from the PMM associated SSTA.. 6.2.2.2 GPI analysis How the PMM affects TC activity in the WNP was investigated using the GPI analysis. Fig. 36a-b depicts the GPI regressed onto ENSO and PMM. It reveals that the ENSO associated SSTA is negatively correlated with TC activity in the WNP. Conversely, the PMM associated SSTA is positively correlated with a southeast– northwest-tilted GPI, extending from equatorial central Pacific to South Taiwan, and negatively correlated with a GPI in North Taiwan and South Japan. A comparison of observed GPI (2016 subtracted from 1998, Fig. 36c) with regression map indicates that the increase of GPI in the WNP is primarily from the PMM associated SSTA. By contrast, the decrease of GPI in South Japan (approximately in 130°E, 30°N) accounts for the ENSO and PMM associated SSTAs. The GPI may separate to two parts, the dynamic (e.g., vertical wind shear and low-level vorticity) and thermodynamic (e.g., mid-level relative humidity). The regression of 850 hPa streamfunction, vertical zonal wind shear (U 200 - U 850) and 700 hPa relative humidity on ENSO and PMM was applied to investigate the relative contribution of ENSO and PMM on the large scale dynamics and thermodynamics. It depicts that the ENSO and PMM are correlated with a low-level anti-cyclonic and cyclonic circulation anomaly in the WNP respectively (Fig. 37a, 37b). It is evident that the low-level cyclonic circulation anomaly (2016 subtracted from 1998, Fig. 37c) in the WNP was related with PMM. The regression of zonal wind vertical shear (Fig. 37d, 37e) shows consistent result with that of 850 hPa streamfunction: The PMM is negatively correlated with vertical zonal wind shear in western Pacific (approximately in 130°E–150°W, 15°S–15°N), and positively correlated with vertical zonal wind shear in east Taiwan (approximately in 130°E, 20°N), in where the climatology of vertical zonal wind shear is negative. Therefore, the PMM associated negative (positive) vertical zonal wind shear in the western Pacific (east Taiwan) is favorable 32.

(39) for TC genesis. In contrast to PMM, the ENSO is positively (negatively) correlated with the vertical zonal wind shear in Philippine, Brunei, and Indonesia (east Taiwan). That is ENSO associated anti-cyclone anomaly and zonal wind shear in the WNP tends to suppress the TC genesis. The 700 hPa relative humidity regressed on ENSO reveals a positive and negative anomaly in the tropical western Pacific (i.e., Indonesia and Sumatra) and central Pacific (~170°E~150°W) respectively (Fig. 38a). By contrast, the PMM depicts a southeast–northwest-tilted structure from equatorial central Pacific to South Taiwan (Fig. 38b), which resembles the regression pattern of GPI (Fig. 36b). The regression analysis reveals the PMM associated thermodynamics and dynamics in the WNP both are favorable for TC genesis, suggesting the enhancement of GPI in the WNP in 2016 (Fig. 36c) is likely resulted from the PMM associated thermodynamics and dynamics.. 6.2.3 Numerical experiments Numerical experiments were conducted to investigate the remote effects of the PMM-like SSTA on the atmospheric circulation anomaly in the WNP. Two experiments were executed. In the control experiment (CTL), the model was forced by the observed climatological monthly sea surface temperature (time mean during 1981–2010). In the PMM run, the PMM-like SSTA (thick trapezium line, Fig. 35c), as well as the climatological mean SST were used to force the model. The PMM run was integrated from January 2015 to December 2016. Ten-member ensemble runs with different initial conditions, initiated on 1 January of arbitrary year from long-term climate run, were performed in the PMM experiment, and the ensemble mean was analyzed. Fig. 39 shows the dynamical and thermodynamical anomalies of experiment results (PMM run minus CTL run). Overall, the large scale cyclonic circulation anomaly in the WNP and the anticyclone anomaly in the eastern tropical Pacific were successfully simulated (Fig. 39a, 37b). In particular, the westerly. 33.

(40) anomaly in the off-equator region of the western Pacific and CP was realistically reproduced. The westerly in the off-equator offsets the prevailing easterly trade wind, therefore, the vertical wind shear reduce in western to central Pacific (Fig. 39b, 37e). For the thermodynamics, the mid-level humid anomaly in the subtropical central Pacific was successfully captured (Fig. 39c, 38b). The major bias of simulation was the cyclonic circulation anomaly and the vertical wind shear shift eastward compared with the observation. Additionally, the model fails in capturing the humid anomaly in the South China Sea and Philippine Sea and therefore the simulated GPI in the WNP is underestimated. A PMM-coupled run is further conducted to investigate whether the bias of simulation is due to the lack of local air–sea interaction. In PMM-coupled run, the model is coupled to a slab ocean in the WNP and forced by observed SST in the PMM region and climatological SST elsewhere PMM and WNP. It reveals that the simulation of PMM-coupled does not show essential improvement, indicating the eastward bias is not likely caused by the lack of local air–sea interaction in the WNP. A possibility for the bias of simulation is that seven TC genesis events clustered in southern Japan in August 2016 (Fig. 40). The TCs may create remarkably upscale energy feedback to the background mean flow through TC mean flow interaction (Hsu et al., 2008; Ko et al., 2012), which may modify the large scale cyclonic circulation anomaly in the WNP to move northeastward. Because the TC genesis and mean flow interaction cannot be resolved using the low-resolution model, the model fails in capturing the northwestward shift of the large scale anomalous cyclone, which substantially contributes to the eastward bias of the westerly winds. The other possible cause is the intraseasonal oscillation that also cannot be realistically simulated in the low-resolution atmospheric model. The Hovmöller diagram of low-pass filtered (30−60 days) 250 hPa velocity potential anomaly reveals that the higher TC genesis number in the WNP in August 2016 occurred in active Madden-Julian Oscillation (MJO, Madden and Julian, 1971) phase (Fig. 41). A comparison of low-pass filtered and non-filtered fields indicates that enhancement of cyclonic circulation anomaly in the WNP (Fig. 41b, 41c) was partially contributed 34.

(41) from the MJO-associated cyclone anomaly. The effect of MJO on the TC activity during the typhoon season of 2016 was also reported by Zhan et al. (2017). Besides the aforementioned possibility, other causes, such as the model’s horizontal resolution, the long-term not precisely included in the PMM-associated SST forcing field, may also contribute to the bias of simulation.. 35.

(42) Ch7 Conclusion and Discussion In this study, we diagnose the characteristics, warming mechanism, and possible impact of SENP SST. The SENP SST is persistent warming from 2013. Observation shows that the SST here exhibits strong variability in interannual and interdecadal time scale. It is associated with PMM in interannual time scale; PDO, NPGO, and AMO in interdecadal time scale. The mixed layer heat budget analysis further reveals that the warming mechanism of SENP SST is through Wind–evaporation–SST mechanism. The main results are summarized as follows:. 1) The strong variance are found in SENP region and the SSTA averaged over 105°W–145°W, 15°N–30°N are defined as SENP index. The lead-lag regression map shows that the SENP SST is warming accompanying with southwesterly wind. It is also found the warm SST in the equatorial central Pacific simultaneously. 2) The variance of SENP SST is associated with PMM and global warming. The regression result of SST, PMM and SENP index is similar but exhibits several differences in tropic Pacific. The SST of SENP (PMM) is warm (cold) in equatorial eastern Pacific. The seasonal variance also shows difference between SENP and PMM, that is, the maximum variance of SENP (PMM) is in June (boreal spring). Notably, the linear warming trend contributes approximately 15% to the SST variance in SENP. The SST variance of SENP with and without linear trend is 0.35 and 0.41, respectively. 3) Wavelet analysis reveals that SENP index exhibits interannual (2–8 years) and interdecadal (8–16 years) variation. The regression analysis shows that SENP and ENSO-like (El Niño–Southern oscillation) SST warm simultaneously in the interannual time scale. It is noted that the variation becomes stronger after late 1990s in interdecadal time scale. The partial regression shows that the Pacific 36.

(43) decadal oscillation (PDO), North Pacific gyre oscillation (NPGO), Atlantic multi-decadal oscillation (AMO) have an impact on SENP SST variation in the interdecadal time scale. 4) Mixed layer heat budget analysis shows that the atmospheric heat flux is the main factor, which is warming the SENP SST. Notably, the latent heat flux is the most important warming factor, which is heating the SENP SST through the Wind–evaporation–SST mechanism. Furthermore, the oceanic advection partially contributors to the SST warming along the coastline. 5) An unusual meridional dipole of SSTA and an enhanced cross-equatorial flow were observed in the EP in the boreal summer of 2014. The enhanced cross-equatorial flow prevented the El Niño event through modulating the 3D oceanic currents as follows. (1) The cross-equatorial flow forced an anomalous westward current ( u ' < 0 ) and an anomalous upwelling ( w ' > 0 ) in the equatorial central–eastern Pacific that respectively led to a negative anomalous zonal advection term ( −u ' ∂T / ∂x < 0 ) and anomalous upwelling advection term ( −w ' ∂T / ∂z < 0 ). (2) D20 shoaled and returned to the climatological level because of the anomalous westward and upwelling currents, which substantially weakened the mean upwelling advection term ( −w∂T '/ ∂z ). (iii) The anomalous cross-equatorial flow induced a northward current anomaly ( v ' > 0 ) that produced a meridionally cold temperature advection ( −v '∂T / ∂y < 0 ) east of 90°W, which also contributed to the decline of SST growth. 6) TC activity and summer monsoon experienced a different behavior between two strong El Niño events, that is, the summertime TC genesis number and monsoon circulation declined during 1998; conversely, a higher TC genesis number and nearly normal strength of the summer monsoon were observed in JJA 2016. The major difference in the SSTA and low-level wind is that a PMM-like subtropical warm SSTA in the ENP associated with a pronounced westerly anomaly in the off-equator region of the western and central Pacific was observed in JJA 2016. The regression and GPI analysis indicated that the PMM SSTA was correlated 37.