Oceanic forcing for the East Asian precipitation in pacemaker AGCM experiments

Authors


Abstract

[1] The impact of the tropical oceanic forcing on the East Asian winter-to-early spring climate is investigated by implementing the pacemaker technique in the slab mixed-layer version of the Geophysical Fluid Dynamics Laboratory AM2.1 atmospheric general circulation model. The results demonstrate that oceanic forcing from the deep tropical eastern Pacific (DTEP) can instigate the Pacific-East Asia teleconnection and that the Philippine Sea anticyclone and the associated air-sea interaction are crucial for the realization of the impact of the teleconnection over East Asia. Comparison among cases in which the pace-maker is designated over the DTEP region, DTEP plus the Indian Ocean, and the whole tropical oceans indicates that tropical oceanic forcing outside of the Niño regions can also exert significant influence on East Asian climate. As a result, a total of 30% of the variance of the East Asian precipitation index can be accounted for by the tropical oceanic forcing.

1. Introduction

[2] It is well established that the El Niño-Southern Oscillation (ENSO), the most prominent mode of interannual variability of the atmosphere-ocean coupled climate system, can influence the East Asian precipitation through atmospheric teleconnections [e.g., Huang and Wu, 1989; Zhang et al., 1999; Zhang and Sumi, 2002; Lau and Nath, 2000; Wang et al., 2000; Lau et al., 2004]. A crucial process that conveys the impact of El Niño “upstream” (with respect to the direction of the midlatitude westerlies) to East Asia is an anomalous low-level circulation feature termed Philippine Sea anticyclone (PSAC). The formation and maintenance mechanisms for the PSAC and its important role in prolonging the ENSO influence on East Asia-western North Pacific monsoon have been extensively studied [e.g., Wang et al., 2000; Wang and Zhang, 2002; Lau and Nath, 2000; Lau et al., 2004; Lau and Nath, 2006]. Here, through the implementation of a pacemaker technique in a slab mixed-layer version of an atmospheric general circulation model (AGCM), we unambiguously identify the deep tropical eastern Pacific (DTEP) as the source for the Pacific-East Asian (PEA) teleconnection and the PSAC as an indispensable link in the PEA that leads to the precipitation anomalies over East Asia, corroborating previous studies [Wang et al., 2000; Lau and Nath, 2000].

[3] Although it is well known that there exists an atmospheric bridge that links the tropical Pacific to East Asia, to the best of our knowledge, no model study has explicitly demonstrated significant skill in capturing the observed variability of the East Asian precipitation and quantified the role of the Niño forcing, or the tropical oceanic forcing in general. One reason is the challenges the global climate models face in accurately simulating regional features like the precipitation over East Asia; another arises from the fact that SST variability outside of the Niño regions is partially driven remotely by the Niño forcing and thus can obscure any attempt of attribution. Here, by specifying the q-flux (as described below) over different regions of the tropical ocean, and in so doing making each of the regions act as a pace maker for global climate variability, we attempt to quantitatively estimate the skill of a state-of-the-art AGCM in simulating the interannual variability of the East Asian precipitation. The pacemaker experiments also serve as a unique framework for the attribution of the simulated East Asian precipitation variability to different sub-regions of the tropical ocean. Even with the pacemaker framework, it remains a challenge for the AGCM to simulate the extent, intensity, and the orientation of the East Asian summer rain band [cf. Kang et al., 2002], we therefore restrict our diagnosis only to winter and early spring months. Results of the experiments indicate that tropical oceanic forcing other than the Niño regions can exert detectable influence on East Asian climate during the boreal winter and spring. As a result, up to 30% of the variance of the East Asian precipitation index (EAPI hereafter) can be accounted for by the prescribed oceanic forcing in the whole tropics. To the extent that the upper limit of predictability is set by how much skill a numerical model can achieve with the best possible information of the forcing, our result suggests a substantial amount of the East Asian winter/spring climate variability is potentially predictable.

2. Pacemaker With q-Flux

[4] The use of the pacemaker technique in a slab mixed-layer model is motivated by the concern that the method of specifying SST globally is deficient [Wang et al., 2005; Douville, 2005] in that it, in many cases (such as over the monsoon regions [Wu and Kirtman, 2005]) erroneously represents the actual surface energetics between the atmosphere and ocean, a vital aspect of the monsoon dynamics. In the meantime, coupled model is not used here because coupled models quickly diverge from the observed climate towards the biased model state, preventing us from evaluating the model results against the observations directly. To overcome these difficulties, a quantity termed q-flux is calculated and prescribed as an external forcing term for the slab ocean-atmosphere coupled system. This q-flux field is calculated from a separate integration with the 50-m mixed-layer temperature tendency equation driven by a relaxation term with a timescale of 1 day towards the observed SST. This term represents the heating/cooling associated with the oceanic processes missing in the slab model such as advection by ocean currents, upwelling, Ekman pumping, and propagation of modal structures. It also acts to nudge the model state towards the observational state so as to prevent model drift. This approach is validated by the replication of the observed history of the SSTs as the q-flux is added back to drive the slab ocean (not shown). We stress that the SST in the slab model is a prognostic quantity determined by the thermodynamic interaction between the slab ocean and the atmosphere.

[5] Three pacemaker ensemble experiments are conducted: (i) deep tropical eastern/central Pacific (DTEP) experiment in which the full monthly q-flux fields from 1950 through 1999 are prescribed over the domain between 8°S and 8°N from 172°E to the South American coast, elsewhere monthly climatological q-flux is specified (note that the area of pace-maker is much narrower than the DTEP run by Lau et al. [2004]); (ii) DTEP+IO experiment, which is similar to (i) except that the q-flux is inter-annually varying over the tropical Indian Ocean basin as well; (iii) TROP experiment, in which the full monthly q-flux fields are prescribed over the whole tropical band between 30°S and 30°N. Each ensemble consists of 8 members, each differing in the atmospheric initial condition. The atmospheric model used for these experiments is the Geophysical Fluid Dynamics Laboratory AM2.1 and viewers are referred to Global Atmospheric Model Development Team [2004] and Delworth et al. [2006] for detailed description of the model.

[6] An atmosphere-only experiment (referred to as ENSO SST experiment) is also performed to further verify if air-sea thermal coupling is indispensible for the PEA teleconnection. By prescribing the positive SST anomalies only over the central and eastern tropical Pacific with SST elsewhere specified as monthly climatology, we can single out the response to the SST forcing alone with the remote feedback due to air-sea interaction excluded. The pattern of the SST anomalies is almost identical to that used by Wang et al. [2000, see Figure 14c]; it is derived as the first EOF of the winter month (Dec–Feb) SST from 1970 through 1999 [Fu et al., 2008] and specified at a magnitude of 1 standard deviation of the corresponding principal component.

[7] All experiments performed for the purpose of the current study are listed in Table 1. Data output from a long control coupled simulation with the GFDL CM2.1 under 1860 pre-industrial climate forcing [Wittenberg, 2009] is also utilized for the purposes of robust statistics and evaluation against the pacemaker results.

Table 1. Summary of Experiments
NomenclatureDurationNumber of MembersDescription
Slab Pacemaker Experiments
DTEP1950–19998Full monthly q-flux over [8°S–8°N, 172°-S. Amer. coast], elsewhere monthly climatology
DTEP-IO1950–19998Full q-flux over DTEP and tropical Indian Ocean
TROP1950–19998Full q-flux over the whole tropics between 30°S and 30°N
 
Atmosphere-Only Experiment
Control SST26 yrs1Fixed monthly SST climatology everywhere
ENSO SST26 yrs1Perturbed by one standard deviation of the leading EOF of Dec–Feb SSTs over the central and eastern Pacific
 
Fully Coupled Experiment
 1000 yrs1Control run with CM2.1 under fixed 1860 pre-industrial radiative forcings

3. Data

[8] Two precipitation datasets are used to validate model simulations. To obtain the spatial distribution of the interannual variability of East Asian precipitation and the ENSO's influence thereon, we make use of the National Oceanic and Atmospheric Administration (NOAA) the Climate Prediction Center Merged Analysis of Precipitation (CMAP) [Xie and Arkin, 1997]. CMAP has a global coverage but only for the period since 1979. To evaluate the variability of the whole model period we revert to station records collected by the National Climate Center of Chinese Meteorological Administration (CMA) during the period from January 1951 through 2008.

[9] The primary fields used for assessing the atmospheric circulation include the 850 hPa winds and 200 hPa streamfunction from the National Centers for Environmental Prediction–National Center for Atmospheric Research (NCEP–NCAR) global atmospheric reanalysis dataset [Kalnay et al., 1996] spanning a 58-yr period from January 1951 to December 2008. The monthly HadISST 1.1 of the Met Office Hadley Centre [Rayner et al., 2003] for the 1950–2008 period is used to compute the q-flux for the pacemaker experiments and to validate model SST output.

4. Results

[10] To quantify the skill of the pacemaker experiments in simulating the observed record of East Asian rainfall, we define the winter-to-early spring EAPI as the December–March rainfall averaged within the region outlined in purple in Figures 1a and 1b for the observations and the model, respectively. The region is chosen based on the pattern of the leading Empirical Orthogonal Function (EOF) of the DJFM rainfall over the domain (15°–55°N, 100°–145°E) for the CMAP observation and the model separately. Note that all the three pacemaker experiments have almost identical first EOF pattern over that domain, therefore only the result of TROP is shown. Both EOF patterns are characteristic of a southwest–northeast oriented rain band spanning from southern China through the Korean peninsula and southern Japan. The EOF pattern of the model is displaced several degrees latitude northward relative to its observational counterpart, likely owing to the poleward bias in the background mean rain distribution over the East Asian region. Nevertheless, in both model and the observations, the East Asian EOF pattern stands robust even after the rainfall variability associated with the ENSO is removed through linear regression (not shown), suggesting that the existence of the EOF pattern itself is not dependent on the ENSO forcing, although, as shown by the El Niño minus La Niña composites (shadings) in Figure 1, the impact of ENSO can project strongly onto the pattern.

Figure 1.

El Niño–La Niña composite of the DJFM precipitation with a magnitude corresponding to 1K warming of the Niño 3.4 SST (shadings, units are mm/day). The EAPI is computed by averaging the land precipitation over the area demarcated by the purple line, which indicates the zero isopleth of the first EOF of rainfall over the East Asia-West Pacific regions. The gray contours indicate the distribution of the mean precipitation during DJFM months at intervals of 1 mm/day. (a) The CMAP and (b) the ensemble mean precipitation in the TROP pacemaker experiment. The fraction of the variance explained by the first EOF is also displayed in parenthesis.

[11] We start by examining the skill of the three pacemaker experiments in capturing the observed variability of the EAPI. Note that because of the lack of good-quality ocean precipitation data prior to 1979, the EAPI is estimated using land precipitation data of CMA spanning from 1951 through 2008 over 35 stations encircled by the purple contour in Figure 1a. The modeled EAPI is computed similarly using precipitation data at 36 land grid points. All three experiments simulate significant amount of variance of the observed EAPI variability (see Figure 2) and the influence of ENSO can be clearly discerned therein as well as in the observation (Figure 2). The prominence of the ENSO forcing in the interannual variability of the EAPI is further elucidated in Figure 3, in which the model EAPI index is compared with the Niño 3.4 index in all the three pacemaker runs (Figures 3a3c), the fully coupled CM2.1 pre-industrial control run (Figure 3d), and the observations (Figure 3e). There clearly exists a positive linear relationship between the SST warming in the equatorial eastern Pacific and the EAPI [see also Wu et al., 2003; Zhang et al., 1999]. The correlation coefficients, as displayed in the parentheses at the upper left corner of each panel, are significant at the 99% confidence level for all cases. The correlation, if calculated with the ENSO years excluded, would have largely diminished, suggesting the prominent role of ENSO in modulating the precipitation over the East Asian region. Quantitatively similar result is also found in the fully coupled CM2.1 control run and the observations. Moreover, a hint of asymmetry between the El Niño years and La Niña years is discernable in the model simulations as well as in the observations. The significance of this intriguing aspect is further substantiated by analyzing a 1000-year long control simulation with the fully coupled CM2.1 (see Figure S1 of the auxiliary material for details).

Figure 2.

Time series of the EAPI from the CMA observations (gray bars), the TROP (red), DTEP-IO (dashed green), and DTEP (dashed blue) pacemaker experiments. Only the ensemble mean values of the model simulations are shown. Correlations with the observations are also displayed. The red and blue vertical lines demarcate the El Niño and La Niña winters, respectively.

Figure 3.

Relationship between the anomalous EAPI and the Niño 3.4 SST anomalies. (a–c) For the pacemaker runs, both the SST index and the EAPI are computed using 8-member ensemble average. (d) For the CM2.1 coupled simulation, a single-member 100-year data are used. (e) The observational EAPI is based on the CMA data.

[12] The mechanisms for the PEA teleconnection and the important role the PSAC have been extensively studied [Wang et al., 2000; Lau and Nath, 2000, 2003; Lau and Wang, 2006] and will not be reiterated here. We just stress that the pacemaker simulations can capture the key features associated with this teleconnection with considerable realism (see Figures 4c4e). See Figure S2 for the corresponding SST and 200 hPa streamfunction fields. The fully coupled CM2.1 (Figure 4b) can also simulate this teleconnection. However, all the key features are displaced westward compared to their observational and pacemaker counterparts, a manifestation of the inherent biases in the mean and the interannual variability of the coupled model. The origin and the maintenance of the PSAC in the PEA teleconnection have been attributed to the thermal coupling between the ocean and atmosphere and the radiative feedback [Wang et al., 2000]. To test if air-sea thermal coupling is a necessary condition for the PEA teleconnection, we examine the response in the atmosphere-only experiment forced by the central and eastern tropical Pacific SST anomalies associated with the warm phase of ENSO (Expt. ENSO SST). The overall characteristics of the precipitation response (Figure 4f) are not dissimilar to the patterns derived from the similar SST forcing in Wang et al. [2000, Figure 14b]. It is striking to see that the center of the original PSAC is now displaced westward to the southern tip of Indian sub-continent and as a consequence the southwesterlies along the margin of the East Asian continent largely diminish. Thus, in this experiment, despite the presence of the instigator of the PEA–the prescribed SST anomalies–the climatic impact over East Asian region cannot be materialized without the air-sea thermal coupling over the west Pacific. This result strongly supports the notion that the air-sea interaction, which is only captured by the pacemaker or the fully coupled configuration of experiments, is of crucial importance to the formation of the PSAC and hence the teleconnection from the Niño region to East Asia.

Figure 4.

El Niño–La Niña composite of the DJFM precipitation (shadings, mm/day), 850 hPa wind (arrows), and 850 hPa streamfunction (purple, 105 m2/s) from (a) NCEP/CMAP; (b) CM2.1 100-year control run; (c) TROP; (d) DTEP-IO; (e) DTEP; and (f) fixed Nino SST experiments. All the fields are normalized with respect to 1 K warming of the Niño 3.4 index.

[13] The ENSO teleconnection is not the only game in town. When the q-flux other than the Niño region is added to the model in TROP case, the correlation between the simulated and the observed EAPI during the period 1951-1998 is augmented from 0.358 for the DTEP case to 0.543 for the TROP case, an increase that is significant at the 87% confidence level according to Fisher's Z test [Spiegel, 1961]. Thus, up to ∼30% of the variance in the interannual variability of the EAPI may be accounted for by the specified tropical q-flux. Assuming linearity, subtracting the DTEP from the TROP gives rise to a realization of climate variability in the absence of the DTEP oceanic forcing. In this case, the EAPI, despite its negative correlation with the EAPI from DTEP, remains marginally correlated (r = 0.28) with the observed index. Although the PSAC still plays a crucial role in the moistening of the East Asian region, circulation elsewhere and the precipitation anomalies over the Indian Ocean are now distinct from those associated with the ENSO (Figures S3a and S3c). In comparison with the slab-coupled ENSO teleconnection (see Figures S2cS2e), the upper level low in Figure S3c is displaced eastward by 30° and the north Indian Ocean is occupied by an anticyclonic circulation. Interestingly, all the circulation, precipitation, and SST patterns associated with the EAPI in the TROP-DTEP case bear a close resemblance to those accompanying the 2nd EOF of the internal variability of the precipitation over the domain (15°–55°N, 100°–145°E) (Figures S3b and S3d). The internal variability is identified as the inter-member spread deviated from the ensemble mean in the pacemaker realizations. The 2nd EOF accounts for 13% of the internal precipitation variability and is separable from the 1st and the 3rd EOFs according to the criterion of North et al. [1982]. We also note that this 2nd EOF resembles strikingly the “meridional mode” of Chang et al. [2007], which arguably arises from wind-evaporation-SST feedback in the absence of the dynamical coupling to the ocean. Taken together, the marginal improvement of skill in capturing the EAPI variability in the TROP simulation upon the ENSO influence might be thought of as the excitation of the 2nd EOF of the (slab-coupled) internal variability by the oceanic forcing over the west Pacific and South China Sea. The nature of the 2nd EOF of variability is still under investigation and the result will be reported elsewhere.

5. Summary and Discussion

[14] By prescribing a q-flux in an AGCM coupled with a slab mixed-layer to represent the heating associated with oceanic processes over the equatorial eastern Pacific (Expt. DTEP), we demonstrate that the essential mechanism of the PEA teleconnection can be replicated by this setting and that this mechanism accounts for a substantial amount of the variability of the East Asian precipitation during boreal winter and early spring. Further experiment with the air-sea coupling disabled reveals that the PSAC, an entity originated from the air-sea interaction, is indispensible and acts as a conduit for the effect of Nino forcing to impact the East Asian precipitation. The pacemaker experiments are also valuable for the purpose of attribution. In comparing the time series of the EAPI simulated by three pacemaker experiments, we found that the oceanic forcing outside of the DTEP region can contribute substantially to the observed variation of the EAPI independent of the ENSO variability. As a result, a total of 30% of the variance of the winter-to-spring EAPI can be accounted for by the tropical oceanic forcing. A quadratic relationship between the EAPI and the Niño forcing is also identified from the pacemaker runs and further confirmed by a long coupled simulation.

[15] The possibility of the southwest–northeast oriented rain band over East Asia being a part of an internal mode of variability of East Asia/West Pacific climate warrants further investigation, in view of that both ENSO and the oceanic forcing independent of ENSO can project onto it. If some hardwired positive feedback (through slab coupling) does exist as the maintaining mechanism for the internal variability of the EAPI, according to the recent studies of applying the fluctuation-dissipation theory to climate mode variability [e.g., Gritsun et al., 2008], this might be the underlying reason for why East Asia is one the regions prone to the ENSO forcing and by corollary, the East Asian winter-to-spring precipitation might be preferably predictable.

Acknowledgments

[16] The authors thank the two anonymous reviewers for their perceptive comments, which helps improve the manuscript substantially. Discussion with Drs. Bin Wang and Renguang Wu has been very valuable. Dr. Andrew Wittenberg provided the data of the long control simulation of CM2.1. J.L. is supported by the startup fund at George Mason University and the COLA omnibus fund from NSF Grant 830068, NOAA Grant NA09OAR4310058, and NASA Grant NNX09AN50G. M.Z. and S.L. are supported by the Special Fund in the Public Interest of China Meteorological Administration (Grant No. GYHY201006022) and the Knowledge Innovation Program of the Chinese Academy of Sciences (Grant No. KZCX2-YW-BR-14).

[17] The Editor thanks the anonymous reviewer for their assistance in evaluating this paper.