Journal of the Meteorological Society of Japan, Vol. 90A, pp. 385--396, 2012.

385

DOI:10.2151/jmsj.2012-A23

NOTES AND CORRESPONDENCE Regional Patterns of Wintertime SLP Change over the North Pacic and Their Uncertainty in CMIP3 Multi-Model Projections

Kazuhiro OSHIMA
Faculty of Environmental Earth Science, Hokkaido University, Sapporo, Japan

Youichi TANIMOTO
Faculty of Environmental Earth Science, Hokkaido University, Sapporo, Japan Research Institute for Global Change, Japan Agency for Marine-Earth Science and Technology, Yokohama, Japan

and Shang-Ping XIE

International Pacic Research Center and Department of Meteorology, School of Ocean and Earth Science and Technology, University of Hawaii at Manoa, Honolulu, Hawaii (Manuscript received 9 May 2011, in nal form 14 October 2011)

Abstract
Regional patterns of wintertime sea level pressure (SLP) trends over the North Pacic and their uncertainty were investigated based on the phase 3 of the Coupled Model Intercomparison Project (CMIP3) multi-model projections under the Special Report on Emissions Scenarios (SRES) A1B emission scenario for the 21st century (20002099). While the 24-model ensemble mean of the 100-yr SLP trend over the North Pacic shows a northward shift of the Aleutian low (AL), regional patterns of the SLP change vary among the models. Projected changes deepen the AL in several models but it shifts northward in some others. The dierent response of the AL results in a large inter-model spread over the North Pacic, which is largest of the Northern Hemisphere and comparable in magnitude to the ensemble mean in the same region. This large spread means a high degree of uncertainty in the 100-yr SLP trend over the North Pacic. For the total uncertainty in the SLP trends over the North Pacic, we examined the relative importance of the internal climate variability and model uncertainty due to dierent treatments of physical processes and computational scheme. To evaluate each of contributions, a single-realization ensemble using a subset of 10 CMIP3 models is compared to a multi-realization ensemble for the same models in the A1B projections. Additionally the control simulations under preindustrial conditions are examined to evaluate the background internal variability in each of the CMIP3 models. Our analysis shows that both the model uncertainty and internal climate variability contribute to the total uncertainty in the 100-yr SLP trends during the 21st century, while the internal climate variability largely explains the total uncertainty in the 50-yr SLP trends during the rst half of the 21st century.

Corresponding author and present aliation: Kazuhiro Oshima, Research Institute for Humanity and Nature, 457-4 Motoyama, Kamigamo, Kita-ku, Kyoto 603-8047, Japan. E-mail: kazuhiro@chikyu.ac.jp 6 2012, Meteorological Society of Japan

386

Journal of the Meteorological Society of Japan

Vol. 90A

The changes in surface heat ux and North Pacic subtropical gyre in association with the dierent response of the AL aect regional patterns of the sea surface temperature trends among models.

1. Introduction To reveal future climate changes in the 21st century, it is important to examine regional patterns in the mean state of the atmosphere and ocean under global warming. Several studies exist on regional changes of North Pacic climate under global warming by using the phase 3 of the Coupled Model Intercomparison Project (CMIP3) model projections in the 21st century (Hori and Ueda 2006). They found that the multi2006; Salathe model ensemble mean of sea level pressure (SLP) trend over the North Pacic shows a northward shift by 34 latitude and deepening by about 0.5 hPa of the Aleutian low (AL) during the 21st century. Yamaguchi and Noda (2006) showed that the individual CMIP3 model projections under the global warming display dierent spatial patterns of the SLP trends over the North Pacic. Because the greenhouse gas (GHG) emission scenario was the same in these model projections, the dierent regional patterns of the SLP trends may be mostly caused by the model uncertainty due to dierent responses to the increased GHG depending on dierent treatments of physical processes and dierent computational schemes (e.g., Deser et al. 2010; Hawkins and Sutton 2009; 2011). In addition to the model uncertainty, internal climate variability in individual models can aect regional SLP projections. Deser et al. (2010) evaluated contributions of internal climate variability to the total uncertainty in the SLP trends over the rst half of the 21st century based on a 40member ensemble using the National Center for Atmospheric Research Community Climate System Model, version 3 (NCAR CCSM3). They revealed that the contribution of internal variability was dominant in the total uncertainty of the SLP trends over mid- and high-latitudes in the Northern Hemisphere. Hawkins and Sutton (2009, 2011) quantitatively evaluated each contribution of the model uncertainty, internal climate variability and emission scenario to the total uncertainty in the trends of surface air temperature (SAT) and precipitation (PR), respectively, using the CMIP3 model projections under the global warming. They indicated that the relative importance of the three uncertainty factors depends on the analyzed variables, selected

region and projected time period of the trends. As for regional SAT trends, internal variability and model uncertainty are the dominant contributors of the total uncertainty in the trends up to 2020 2030. The model uncertainty is of greater importance in the trends of 20202040. As for regional PR trends, while the internal variability is the dominant contributor to the total uncertainty in many selected regions up to 2030, the model uncertainty is generally dominant beyond 2030. In the present study, we attempt to evaluate relative contributions of the model uncertainty and internal climate variability to the total uncertainty in the regional SLP trends over the North Pacic based on the CMIP3 multi-model projections. Several recent studies showed spatial characteristics of sea surface temperature (SST) variability in the North Pacic under the global warming (Furtado et al. 2011; Oshima and Tanimoto 2009; Overland and Wang 2007; Wang et al. 2009). These studies found that the decadal SST variability in the 21st century follows the spatial pattern of the Pacic Decadal Oscillation (PDO) found in the observed records and the 20th century climate in coupled models (20C3M) simulations of the CMIP3 models. As for the SST warming trends in the 21st century, while many studies have examined regional patterns of the trends in the tropical Pacic (e.g., Collins et al. 2005; DiNezio et al. 2009; Liu et al. 2005; Xie et al. 2010), there are few studies in the literature that examines regional patterns of the SST warming trend in the mid- and high-latitude North Pacic and its association with the overlying atmospheric changes. Dierent regional patterns of the SLP may induce the dierent regional patterns of SST. The present study further examines the processes by which regional patterns of SLP trend induce regional patterns of SST trend in the North Pacic. In the rest of this paper, Section 2 introduces datasets and analysis methods to study the projected trends and their uncertainty. Section 3 shows the results of the regional dierences in the SLP trends over the North Pacic in the CMIP3 models. Section 4 evaluates the uncertainty in the SLP trends. Section 5 discusses the relationship in regional spatial patterns between the SLP and SST trends. Finally, Section 6 is summary.

February 2012

K. OSHIMA et al.

387

2. Data and analysis method We used the global warming projections under the SRES A1B emission scenario from 2000 to 2099 and control simulations under preindustrial conditions (PICTL) in the 24 CMIP3 models (Meehl et al. 2007). For simplicity, we named Model A to X as listed in Table 1. Since our focus is on climate change in and around the North Pacic, we dene

the North Pacic sector as the region of 1570 N, 120 E120 W (inset in Fig. 1a). Monthly outputs of SLP, SST, surface zonal wind (UAS), surface turbulent heat ux (SHF, sum of latent and sensible heat uxes) and sea surface height (SSH) were analyzed. SST from the Hadley centre sea Ice and Sea Surface Temperature (HadISST, Rayner et al. 2003), SLP, UAS and SHF based on the Japanese 25-year Reanalysis (JRA25, Onogi et al. 2005) and

Table 1. A list of the 24 CMIP3 model (group A) projections under the SRES A1B scenario and the PICTL simulation used in this study. In the second column, the selected 10 models (group B and C) having more than three ensemble members are marked. In the third column, the number of ensemble members in the individual model projections under the A1B scenario is shown. In the fourth and fth column, models with strong negative and strong positive NPI (North Pacic index) trends are marked. Model A B C D E F G H I J K L M N O P Q R S T U V W X BCCR-BCM2.0 (Norway) CGCM3.1_T47 (Canada) CGCM3.1_T63 (Canada) CNRM-CM3 (France) CSIRO-MK3.0 (Australia) CSIRO-MK3.5 (Australia) GFDL-CM2.0 (USA) GFDL-CM2.1 (USA) GISS-AOM (USA) GISS-EH (USA) GISS-ER (USA) FGOALS-g1.0 (China) INGV-SXG (Italy) INM-CM3.0 (Russia) IPSL-CM4 (France) MIROC3.2_hires (Japan) MIROC3.2_medres (Japan) ECHO-G (Germany/Korea) ECHAM5_MPI-OM (Germany) MRI-CGCM2.3.2 (Japan) NCAR-CCSM3 (USA) NCAR-PCM1(USA) UKMO-HadCM3 (UK) UKMO-HadGEM1 (UK) 24 models 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 10 models Number of ensemble members 1 5 1 1 1 1 1 1 1 3 5 3 1 1 1 1 3 3 4 5 7 4 1 1 3 3 3 3 3 3 3 3 negative NPI trend positive NPI trend

388

Journal of the Meteorological Society of Japan

Vol. 90A

SSH based on the Simple Ocean Data Assimilation reanalysis (SODA, Carton et al. 2000) were used as the observed references of the 20th century climate. All monthly data were bilinearly interpolated onto a common 5 longitude 5 latitude grid for easy comparison. We calculated monthly anomalies by subtracting the monthly climatologies over the 21st century in the A1B scenario. These monthly anomalies were averaged over boreal winter (DJF; December, January and February) and then smoothed with the 5-year running mean. We calculated linear trends at each grid point using the least squares linear regression for all variables from 100-yr (50-yr) DJF mean anomaly records during 20002099 (20002049). Mean and variance of the SLP trends in the 24 CMIP3 models are the multi-model ensemble mean and intermodel spread, respectively. As for analysis of the uncertainty, we employed the following procedure. For a model with several ensemble members, we assume that the multimember mean of the model removes the uncertainty due to internal climate variability. Thus the spread of multi-member ensemble means among individual models is considered to be induced only by model uncertainty. This statistical treatment of uncertainties in the climate trend was used in previous studies (e.g., Deser et al. 2010; Yukimoto and Kodera 2005). In the CMIP3 model projections under the A1B scenario, only 10 models have three or more ensemble members (Models B, J, K, L, Q, R, S, T, U and V, Table 1). The mean and variance of the trends in the multi-member ensembles of these 10 models refer to the multi-member ensemble mean and inter-member spread, respectively. The multi-model ensemble mean and inter-model spread of the SLP trends are calculated in each of the following groups (A, B and C, Table 2). For group A, we chose one simulation each from the
Table 2. group A B

24 CMIP3 models. Groups B and C are used to estimate internal variability. Group B consists of three subsets, each made of 10 runs, one from each of the selected 10 CMIP3 models that contain three or more members. The mean of variance from three subsets of group B represents the uncertainty due to the internal variability and model uncertainty. For group C, we chose the three-member ensemble means of the 10 models. The simulations in the group C are the same as in the group B. The intermodel spreads of the SLP trends in the groups A and B contain the uncertainty both from the internal climate variability and model uncertainty, while the inter-model spread in the group C is assumed to contain only the model uncertainty. For group D, we used 40 members of the model U. Note that each member of model U was integrated only during 20002061 (Deser et al. 2010). The intermember spread in the group D is also assumed to contain the uncertainty only due to the internal climate variability. As an alternative, we estimate the internal climate variability using the PICTL simulations. Group A of the PICTL simulations consists of 24 100-yr records, one each from the 24 models. For the 24 models, we calculated the 100-yr (50-yr) SLP trends from 100-yr (rst 50-yr) records. Group B is made 300-yr runs from a subset of 10 models. We calculated 100-yr (50-yr) SLP trends from each 300-yr PICTL run. For each model, there are 21 such 100-yr samples with the starting years delayed successively by one decade. We used the average of the spreads among these 21 of 100-yr (50-yr) trends as the group-B PICTL simulations. Without GHG increase, the inter-model spread of the PICTL simulation contains the uncertainty only from the internal climate variability. We assume that the amounts of the internal variabilities are comparable both in the PICTL and A1B simulations.

Categories of four groups used in the analysis of the uncertainty. Ensemble

Multi-model ensemble based on 24 of single members from each of the 24 CMIP3 models Multi-model ensemble based on three subsets with 10 of single members from the selected 10 CMIP3 models having more than three ensemble members (marked in the second column in Table 1); Average over the three subsets is used as group B. Multi-model ensemble based on 10 of multi-member ensemble mean averaged over three members from the selected 10 CMIP3 models (marked in the second column in Table 1) Multi-member ensemble based on 40 ensemble members of the model U (NCAR CCSM3)

C D

February 2012

K. OSHIMA et al.

389

Fig. 1. (a) Multi-model ensemble mean (contours at 1 hPa 100yr1 intervals) and inter-model spread (shades in (hPa 100yr1 ) 2 ) of SLP trend during 20002099 in the group A. Supplemental G0.5 hPa 100yr1 contours are added and zero contour is omitted. The solid inset region (15 N70 N, 120 E120 W) indicates the North Pacic sector. (b) Ratio of the number of models with the signicant positive (red contours) and negative (blue contours) SLP trends in the group A. The models with the trend greater (less) than 0.5 (0.5) hPa 100yr1 at each grid point were employed for the ratio calculations. Ratios greater than 30% are shown and the contour interval (CI) is 10%.

3. Regional patterns of SLP trends and their spread over the North Pacic Regional patterns of the wintertime SLP trends during 20002099 are investigated based on the 24 CMIP3 model projections (group A) under the A1B scenario. The multi-model ensemble mean of the group A (contours in Fig. 1a) shows negative SLP trends over the high-latitudes with a minimum of 5 hPa 100yr1 over the northern Bering Sea (65 N, 180 ) and a weak local minima over the Canadian archipelago and Barents Sea. Positive SLP trends are found over the southwest portion of the North Pacic (25 N40 N, 135 E170 W), most of the North Atlantic (30 N55 N, 10 W 60 W) and around the Mediterranean Sea. The inter-model spreads of the SLP trends (gray shades in Fig. 1a) over the North Pacic, Arctic and North Atlantic sectors are comparable in magnitude to the multi-model ensemble mean, showing a large uncertainty in SLP trends over these regions. We note that the inter-model spread of the group A can be induced by both the model uncertainty and internal climate variability. To further examine the polarity of the SLP trends over the Northern Hemi-

sphere, we counted the number of models whose SLP trend is greater (less) than 0.5 (0.5) hPa 100yr1 at each grid point. More than 90% of the CMIP models display the negative SLP trends around the Bering Sea and Canadian archipelago (Fig. 1b), while 5070% of the models display the positive trends over the southwest portion of the North Pacic, the North Atlantic and Mediterranean Sea. This good agreement in the polarity of the SLP trends indicates that the large inter-model spread over these regions is due to dierences in the magnitude of the SLP trends among models rather than the polarity of the trends. Over the North Pacic sector, the multi-model ensemble mean of the group A features a center of the negative SLP trends over the Bering Sea, and a zonal band of the positive trends that extends from east of Japan to the central North Pacic along 35 N (Fig. 1a). This pattern indicates a northward shift of the AL consistent with the previous studies 2006). The inter(Hori and Ueda 2006; Salathe model spread over the North Pacic is largest with a maximum at 12.5 (hPa 100yr1 ) 2 over the northeastern portion of the sector (50 N, 155 W, Fig. 1a), where 3040% of the models display positive SLP

390

Journal of the Meteorological Society of Japan

Vol. 90A

Fig. 2. (a) Spatial pattern of EOF1 (46.0% explained variance) for SLP trend during 20002099 in the group A. CI is 0.5 hPa 100yr1 and the zero contour is omitted. (b) Scatter plot of the EOF1 scores and trends of North Pacic index (NPI), which is dened as an area-weighted mean of SLP anomalies over 3065 N, 160 E140 W, in the group A. Labels of AX represent the 24 CMIP3 models (Table 1). Uppercase letters indicate the 10 models (Models B, J, K, L, Q, R, S, T, U and V) used for the groups B and C. Four models with strong negative (Models C, E, V and X) and four models (Models J, K, Q and R) with strong positive NPI trends are marked with circles. Correlation coecient for this scatter plot is 0.97 above 99% signicant level.

trends while other 3040% of the models display negative ones (Fig. 1b). Thus the large inter-model spread over the northeastern Pacic is due to dierences in polarity of the SLP trends among models. This means that the northward shift of AL found in the multi-model ensemble mean is not a consistent feature among group A models. To identify the dominant pattern that accounts for the large inter-model spread of the SLP trend over the North Pacic sector, we performed an empirical orthogonal function (EOF) analysis on the 100-yr SLP trends of the group A. The rst EOF mode (EOF1, Fig. 2a) explains 46.0% of the total inter-model variance with a center of positive trends over the northeastern portion of the sector (50 N, 160 W), close to the climatological center of the AL (not shown). There are weak centers of negative trends over northern Eurasia and the Arctic Ocean (Fig. 2a). The scores of the EOF1 are positively correlated with the trends of the North Pacic index (NPI, Fig. 2b) among individual models, which is a measure of the AL strength (Trenberth and Hurrell 1994). The correlation coecient between the EOF1 scores and NPI trends is 0.97 above 99% signicant level.

4. Sources of uncertainty in SLP trends over the North Pacic To evaluate relative contributions of the model uncertainty and internal climate variability to the total uncertainty in the SLP trends over the North Pacic sector, we compared the inter-model spreads of the SLP trends during 20002099 among the three groups (A, B, C) of the A1B projections in the CMIP3 multi-models. The spatial pattern of the inter-model spread in the group A over the North Pacic sector (Fig. 3a, same as gray shades in Fig. 1a) is very similar to that in the group B (Fig. 3b), and that in each of the three subsets of the A1B group B (not shown). These results indicate that the inter-model spread is not aected when the number of the models is reduced to 10 from 24. We note that the inter-model spread of the group B, like that of the group A, contains the uncertainty owing to the model uncertainty and internal climate variability. Since the contribution of internal variability can be small in the multimember mean for each of the models, the intermodel spreads in the group C (Fig. 3c) are reduced compared to the groups A and B (Figs. 3a and 3b).

February 2012

K. OSHIMA et al.

391

Fig. 3. (ac) Inter-model spreads (shades; (hPa 100yr1 ) 2 ) of the 100-yr SLP trend during 20002099 over the North Pacic sector in the group (a) A, (b) B, and (c) C. Contours are dierences in the inter-model spread (5 (hPa 100yr1 ) 2 intervals): (b) group A minus B, (c) B minus C. (dg) As in (ac), but for the 50-yr SLP trend (hPa 100yr1 ) 2 during 20002050 in the group (d) A, (e) B, (f ) C, and (g) D. Note that the categories of the four groups are described in Table 2.

Signicant decrease (@30%) is found over the northeastern portion of the North Pacic. On the basin-scale, the spatial pattern of the inter-model spread in the groups A and B looks similar to that in the group C, especially in the western North Pa-

cic. In the basin-average over the North Pacic, the model uncertainty measured by the inter-model spread of the group C (row II in Table 3) accounts for 72% (2.3 (hPa 100yr1 ) 2 ) of the total uncertainty measured by the inter-model spread of the

392

Journal of the Meteorological Society of Japan

Vol. 90A

Table 3. Area-average of the inter-model spread (unit: (hPa 100yr1 ) 2 ) over the North Pacic sector and related statistics of the 100-yr and 50-yr SLP trends in the A1B and PICTL simulations. Analyzed trend period Uncertainty Total uncertainty in A1B Tentative model uncertainty only based on A1B Tentative internal variability only based on A1B Background internal variability in PICTL Better estimation of model uncertainty 100-yr 3.2 2.3 0.9 1.4 1.8 50-yr 11.4 4.7 6.7 7.0 4.4

Row No. (I) (II) (III) (IV) (V)

Statistics and analyzed dataset Spread of group B in A1B Spread of group C in A1B Residual: (I) minus (II) Spread of group B in PICTL Residual: (I) minus (IV)

group B (row I in Table 3), while the internal variability dened as a residual by removing the model uncertainty from the total uncertainty (row III in Table 3) is 0.9 (hPa 100yr1 ) 2 . This result suggests that the model uncertainty largely explains the total uncertainty in the 100-yr SLP trends over most of the North Pacic, while internal climate variability contributes to the total uncertainty over the eastern North Pacic. We further examined uncertainty in the SLP trends for a shorter period in the near-future projection during 20002049 for the same set of four groups in the A1B projections. The inter-model spreads of the 50-yr SLP trends in the groups A, B and C (Figs. 3d, 3e and 3f ) tend to be greater than those of the 100-yr SLP trends (Figs. 3ac), while the spatial patterns of those spreads still show the common feature of a signicant center in the northeastern portion of the North Pacic, as in Figs. 3a c. The dierences in the inter-model spreads of the 50-yr trends between the groups A and B (Figs. 3d and 3e) tend to be larger than those of the 100-yr trends (Figs. 3a and 3b), indicating that the intermodel spread of the 50-yr trends strongly depends on the selection of the models. Such dierences are also seen in each of the three subsets of the group B (not shown). On the basin-scale, the inter-model spread of the 50-yr trends in the group C (Fig. 3f ) is much reduced than in the group B (Fig. 3e). In the area-average over the North Pacic sector, the model uncertainty (the inter-model spread of the group C; row II in Table 3) and the internal variability (the residual; row III in Table 3) account for 41% (4.7 (hPa 100yr1 ) 2 ) and 59% (6.7 (hPa 100yr1 ) 2 ) of the total uncertainty (row I in Table 3), respectively. These results indicate the substantial contribution of the internal climate variability to the total uncertainty in the 50-yr SLP trends compared to the 100-yr trends. Indeed, the inter-

member spread among 40 ensemble members of the model U (Deser et al. 2010) is very large in the northeastern portion of the North Pacic (Fig. 3g). One may ask whether the number of the ensemble members is sucient to remove the internal variability in the group C of the A1B projection. Here we employ the PICTL simulations to estimate the background internal variability of the CMIP3 models. Without GHG increase, the inter-model spreads of the PICTL group B measure the uncertainty from the internal climate variability. As for the 100-yr SLP trends in the PICTL group B, the area-average of the inter-model spread over the North Pacic sector is about 1.5 times larger in the PICTL simulation (1.4 (hPa 100yr1 ) 2 , row IV in Table 3) than in the A1B projection (0.9 (hPa 100yr1 ) 2 , row III in Table 3). The spread for the 50-yr trends in the group B of the PICTL simulation is even larger (7.0 (hPa 100yr1 ) 2 , row IV in Table 3). The residuals in the A1B projection (row III in Table 3) underestimate the internal variability. Using the inter-model spread of the group B in the PICTL simulation as the better estimation of the internal variability, the revised model uncertainty is 1.8 (hPa 100yr1 ) 2 for the 100-yr trend and 4.4 (hPa 100yr1 ) 2 for the 50-yr trend, respectively (row V in Table 3). The inter-model spreads of the 24 models (group A) are comparable to those of the 10 models (group B). This improved statistics show that both the model uncertainty and internal climate variability contribute to the total uncertainty in the 100-yr SLP trends during the 21st century, and that the latter contribution is much greater in the 50-yr trend during the rst half of the 21st century. Based on our analyses of the A1B and PICTL simulations, we note that a large number of the models, each with multi-member integrations, is necessary for a robust near-future projection.

February 2012

K. OSHIMA et al.

393

5. Relationship in regional patterns between SLP and SST trends As documented in Section 3, the polarity of the SLP trend over the North Pacic is dierent among the group A models (Fig. 1b). To examine processes by which the regional dierences in SLP trend induce those in SST trend in the North Pacic, we calculated composites of the trends in four models with strong negative NPI trend (Models C, E, V and X, listed in Table 1 and marked with circles in Fig. 2b) and in four models with strong positive NPI trend (Models J, K, Q and R, listed in Table 1 and marked with circles in Fig. 2b). As expected from the signicant correlation between the EOF1 scores and the NPI trends (Fig. 2b), the former (latter) models show the negative (positive) SLP trends over the North Pacic (Figs. 4a and 4g). In the four models with the strong negative NPI trend, the SLP composite features a negative center of the minimum of 8 hPa 100yr1 over the Bering Sea (Fig. 4a). This corresponds to a deepening trend of the AL (Fig. 4c). In association with this deepened AL, UAS are enhanced by 1.5 m s1 100yr1 (Fig. 4d). This enhanced westerly jet acts

to increase the SHF by 1025 W m 2 100yr1 in the extratropical North Pacic (30 N45 N, Fig. 4e). As a result, the SST trends in the western and central North Pacic are smaller by 0.21 C 100yr1 than the basin-average SST trend of 2.34 C 100yr1 (Figs. 4a and b). While an increase of meridional gradient of the SSH between 30 N45 N suggests an intensication of the subtropical gyre in the North Pacic (Fig. 4f ), its eect on the SST trend seems weak. Overall, patterns of SST change are consistent with changes in prevailing wind, with locally enhanced warming over reduced wind speed. In the four models with strong positive NPI trend, the SLP composite shows a positive center with a maximum of 5 hPa 100yr1 over the central North Pacic (40 N, 175 W, Fig. 4g). The center of the AL moves northward by 5 latitude (Fig. 4i). This shift of the AL induces enhanced (reduced) surface westerly trends in 45 N60 N (20 N 35 N), indicative of a northward shift of the westerly jet by 5 latitude (Figs. 4g and 4j). The shift of the westerly jet is accompanied by an eastward anomalous jet in ocean current centered at 40 N (Fig. 4l) anchoring a locally enhanced warming at the same latitude in the western basin (Fig. 4h). The enhanced SST warming along 40 N takes

Fig. 4. (af ) Composites of the four models with strong negative NPI trends. (a) SLP trend (contours at 1 hPa 100yr1 intervals), regional dierence in the SST trend (colors;  C 100yr1 ), and UAS trend (arrows; m s1 100yr1 ). Note that the SST trend represents deviation from area-average over the North Pacic sector which is 2.34 C 100yr1 . (bf ) Meridional plots of zonal mean present-day climatology (black line), future climatology (red line), and trends (dierences between the future and present-day climatologies, green line). (b) SST, (c) SLP, (d) UAS, (e) SHF and (f ) SSH averaged over 140 E160 W in the North Pacic. Note that the axes for the climatologies are at the bottom of the panel, but the axis for the trend is at the top of the panel. (gl) As in (af ), but for the four models with strong positive NPI trends. Areaaverage of SST trend over the sector is 1.93 C 100yr1 .

394

Journal of the Meteorological Society of Japan

Vol. 90A

place even though the SHF trend features a positive (i.e., enhanced heat release from the ocean) peak at 40 N (Figs. 4h and 4k). This indicates that the enhanced SHF is a result of the SST warming in response to the changes in the subtropical gyre rather than a cause of the SST trend. On the horizontal map, the enhanced SST warming shows a westward intensication with little change in local westerly winds (Fig. 4g). This is consistent with an intensied and northward expanded subtropical gyre. 6. Summary The regional patterns of wintertime SLP trends over the North Pacic and their uncertainty were investigated based on 24 CMIP3 model projections under the A1B scenario and control simulations under pre-industrial conditions (PICTL). To evaluate uncertainty of the SLP trend, we analyzed the multi-model ensemble means and their spread in four groups (groups A, B, C and D, Table 2) of the A1B projections and in the two groups (groups A and B) of the PICTL simulations. In the A1B projection, the multi-model ensemble mean of the 100-yr SLP trends during 20002099 over the North Pacic based on single member runs with 24 CMIP 3 models (group A) shows a deepening and northward shift of the AL (Fig. 1a), consistent with previous studies (Hori and Ueda 2006). Several models in the group 2006; Salathe feature a deepening of the AL while several others a northward shift of the AL. This large inter-model spread of the SLP trend over the North Pacic (Figs. 1 and 2) indicates a large uncertainty of SLP change over this region. We have evaluated the relative contributions of the model uncertainty and internal climate variability to the total uncertainty in the SLP trends using two groups from the selected 10 models that have three or more ensemble members. Group B of the A1B projection, made of three subsets of 10 single runs from dierent models, is used to estimate the total uncertainty, while group C is a multi-member ensemble mean from each of these models, where we assume that the internal variability is removed. If we assume that the internal variability is suciently removed in the group C, the inter-model spreads in the group C of the A1B projections indicate that the model uncertainty dominates the total uncertainty in the 100-yr SLP trend over the North Pacic while internal climate variability plays a minor role (Figs. 3b and c). For the shorter projection of the 50-yr trend, the relative contribution

from internal climate variability increases (Figs. 3eg) in agreement with Deser et al. (2010). Because of the limited number of ensemble members in the A1B group C, we additionally employ the PICTL simulation to evaluate the background internal variability of the SLP trend in the CMIP models. Without GHG increase, the inter-model spreads of the PICTL group B measure the uncertainty from the internal climate variability. Our analysis of the PICTL simulations shows that the limited number of A1B projections underestimates the internal variability over the North Pacic both for the 100-yr and 50-yr trends. The revised estimates indicate that both the model uncertainty and internal climate variability contribute to the total uncertainty in the 100-yr SLP trends over the North Pacic, while the internal variability largely explains the total uncertainty in the 50-yr trends (Table 3). These results have implications for the discussion of future climate change over the North Pacic. Neither multi-member projections with a particular model nor multi-model ensemble mean of single member runs provide a robust projection and associated uncertainty. For a reliable future climate projection in the North Pacic, it is necessary to conduct multi-member ensemble projections using multi-models to reduce the internal variability and model uncertainty, respectively. The dierent response of the projected SLP trend over the Aleutian Low leads to dierent regional patterns of SST trend in the North Pacic (Figs. 4a and 4g). In the four models with the strong negative NPI trend in the group A, the changes in prevailing wind explains spatial pattern of SST warming, with increased SHF from the ocean in association with the deepening of the AL causing reduced SST warming in the central North Pacic (Figs. 4af ). By contrast, in the four models with the strong positive NPI trend in the group A, the northward shift of the subtropical gyre in the North Pacic, in response to the similar northward shift of the AL, induces a band of enhanced SST warming trend in the northwest Pacic along with subtropical-subpoalr gyre boundary (Figs. 4gl). Thus, the inter-model spread of the SST warming pattern in the North Pacic is strongly dependent on the dierences in the SLP trend near the AL. Distinctive regional dierences in ocean response emerge in the North Pacic as shown here and by Xie et al. (2010). Besides the gyre-scale adjustments, recent studies indicate the importance of changes in mode water ventilationa higher verti-

February 2012

K. OSHIMA et al.

395

cal mode processin surface current and SST response to global warming (Xie et al. 2011; Xu et al. 2012). Pacic Ocean adjustments to wind and buoyancy changes need further investigations. The changes in SST and ocean circulation under the global warming are important not only for climate, but for the marine ecosystem (Wang et al. 2010). Recently, Yara et al. (2011) assessed the impact of SST warming and its uncertainty on corals near Japan based on the CMIP3 multi-model projections. Our results about regional dierences in SST warming and changes in the oceanic gyre provide useful information for such assessments of the marine ecosystem change. Acknowledgments This work was supported by the Global Environment Research Fund (S-5) of the Ministry of the Environment, Japan, JAMSTEC, NSF and JSPS Institutional Program for Young Researcher Overseas Visits. We acknowledge the modeling groups for providing their data for analysis, the Program for Climate Model Diagnosis and Intercomparison (PCMDI) for collecting and archiving the model output, and the JSC/CLIVAR Working Group on Coupled Modeling (WGCM) for organizing the model data analysis activity. The multi-model data archive is supported by the Oce of Science, U.S. Department of Energy. We also express our gratitude to the Data Integration and Analysis System Fund for National Key Technology from the Ministry of Education, Culture, Sports, Science and Technology, Japan, for providing us with an invaluable environment for mass data handling. We acknowledge the CCSM Climate Variability and Climate Change Working Groups 21st century CCSM3 Large Ensemble Project, for providing their data. NCAR Command Language (NCL) and GrADS were used for analysis and drawing gures. References
Collins, M., and The CMIP modeling groups, 2005: El Nin o or La Nin a-like climate change? Clim. Dyn., 24, 89104. Carton, J. A., G. Chepurin, and X. Cao, 2000: A Simple Ocean Data Assimilation analysis of the global upper ocean 19501995 Part 2: results. J. Phys. Oceanogr., 30, 311326. Deser, C., A. S. Phillips, V. Bourdette, and H. Teng, 2010: Uncertainty in climate change projections:

The role of internal variability. Clim. Dyn., doi:10.1007/s00382-010-0977-x. DiNezio, P. N., A. C. Clement, G. A. Vecchi, B. J. Soden, B. P. Kirtman, and S.-K. Lee, 2009: Climate response of the equatorial pacic to global warming. J. Climate, 22:18, 48734892. Furtado, J., E. Di Lorenzo, N. Schneider, N. Bond, and J. Overland, 2011: North Pacic Decadal Variability and Climate Change in the IPCC AR4 Models. J. Climate, doi:10.1175/2010JCLI3584.1, in press. Hori, M. E., and H. Ueda, 2006: Impact of global warming on the East Asian winter monsoon as revealed by nine coupled atmosphere-ocean GCMs, Geophys. Res. Lett., 33, L03713, doi:10.1029/ 2005GL024961. Hawkins, E., and R. T. Sutton, 2009: The potential to narrow uncertainty in regional climate predictions. Bull. Amer. Meteor. Soc., 90, 10951107. Hawkins, E., and R. T. Sutton, 2011: The potential to narrow uncertainty in projections of regional precipitation change. Clim. Dyn., doi:10.1007/s00382010-0810-6, in press. Liu, Z., S. Vavrus, F. Fe, N. Wen, and Y. Zhong, 2005: Rethinking tropical ocean response to global warming: The enhanced equatorial warming. J. Climate, 18, 46844700. Meehl, G. A., C. Covey, T. Delworth, M. Latif, B. McAvaney, J. F. B. Mitchell, R. J. Stouer, and K. E. Taylor, 2007: The WCRP CMIP3 multimodel dataset: A new era in climate change research. Bull. Amer. Meteor. Soc., 88, 13831394, doi:10.1175/ BAMS-88-9-1383. Onogi, K., J. Tsutsui, H. Koide, M. Sakamoto, S. Kobayashi, H. Hatsushika, T. Matsumoto, N. Yamazaki, H. Kamahori, K. Takahashi, S. Kadokura, K. Wada, K. Kato, R. Oyama, T. Ose, N. Mannoji, and R. Taira, 2007: The JRA-25 Reanalysis. J. Meteor. Soc. Japan, 85, 3, 369432. doi:10.2151/ jmsj.85.369. Oshima, K., and Y. Tanimoto, 2009: An evaluation of reproducibility of the Pacic Decadal Oscillation in the CMIP3 simulations. J. Meteor. Soc. Japan, 87, 4, 755770. doi:10.2151/jmsj.87.775. Overland, J. E., and M. Wang, 2007: Future climate of the North Pacic Ocean, Eos Trans. AGU, 88, 178, 182. Rayner, N. A., D. E. Parker, E. B. Horton, C. K. Folland, L. V. Alexander, D. P. Rowell, E. C. Kent, and A. Kaplan, 2003: Global analyses of SST, sea ice and night marine air temperature since the late 19th Century. J. Geophys. Res., 108, 4407, doi:10.1029/2002JD002670. , E. P., Jr. 2006: Inuences of a shift in North Salathe Pacic storm tracks on western North American precipitation under global warming. Geophys. Res. Lett., 33, L19820, doi:10.1029/2006GL026882.

396

Journal of the Meteorological Society of Japan

Vol. 90A

Trenberth, K. E., and J. W. Hurrell, 1994: Decadal atmosphere-ocean variations in the Pacic. Climate Dyn., 9, 303319. Wang, M., J. E. Overland, and N. A. Bond, 2010: Climate projections for selected large marine ecosystems. J. Mar. Syst. 79, 258266, doi:10.1016/ j.jmarsys.2008.11.028. Xie, S.-P., C. Deser, G. A. Vecchi, J. Ma, H. Teng, and A. T. Wittenberg, 2010: Global warming pattern formation: sea surface temperature and rainfall. J. Climate, 23, 966986. Xie, S.-P., L.-X. Xu, Q. Liu, and F. Kobashi, 2011: Dynamical role of mode-water ventilation in decadal variability in the central subtropical gyre of the North Pacic. J. Climate, 24, 12121225. Xu, L. X., S.-P. Xie, Q. Liu, and F. Kobashi, 2012: Response of the North Pacic Subtropical Counter-

current and its variability to global warming. J. Oceanogr., in press, doi:10.1007/s10872-011-00316. Yamaguchi, K., and A. Noda, 2006: Global warming patterns over the North Pacic: ENSO versus AO. J. Meteor. Soc. Japan, 84, 221241. Yara, Y., K. Oshima, M. Fujii, H. Yamano, Y. Yamanaka, and N. Okada, 2012: Projection and uncertainty of the poleward range expansion of coral habitats in response to sea surface temperature warming: A multiple climate model study. Galaxea, in press. Yukimoto, S., and K. Kodera, 2005: Interdecadal Arctic Oscillation in twentieth century climate simulations viewed as internal variability and response to external forcing. Geophys. Res. Lett., 32. L03707, doi: 10.1029/2004GL021870.