A.W. Robertson

Interdecadal variability in the north Pacific region is investigated in a 500-yr control integration of the Hamburg ECHAM+LSG coupled general circulation model. The spectrum is predominantly red, but a significant peak with a period of abou t 18 yrs is detected in the spectrum of sea surface temperature (SST). This peak is shown to be associated with an irregular oscillation that involves both the model ocean and atmosphere. The SST, sea-level pressure, and geopotential height at 500hPa all undergo a primarily standing oscillation with an extensive monopo le structure centered near the date line. The surface anticyclone is situated to the northeast of the warm SST anomaly, and there is a small westward tilt with height; temporal changes are approximately in phase. The anomalous surface heat flux accompanying the warm phase of SST is primarily out of the ocean, but is la rgely balanced by anomalous horizontal thermal advection by surface currents, al lowing the SST anomaly to persist. Oceanic thermocline anomalies propagate north ward in the western Pacific, and lag the atmosphere indicating a disequilibrium with the atmosphere; sub-surface thermal advection appears to play an important role. A comparison is made between the model's 18-yr oscillation and oscillatory components identified in an analysis of the GISST observational SST dataset, wh ich have periods of about 6 and 30 yrs

**Figure 1:** Standard deviation of annual means of (a) SST (K), (b) upper-oc
ean heat content (the temperature averaged between 50-575m depth) (K), and (c) g
eopotential height of 500mb surface (gpm).

**Figure 2: **Monte-Carlo significance test of coupled GCM's North Pacific SS
T (4-yr means). Shown are projections of North Pacific SST onto the AR(1) null-h
ypothesis basis, versus the dominant frequency of the basis vectors. The latter
were computed via a reduced Fourier transform. The error bars show the 90% confi
dence interval of an AR(1) noise null hypothesis, generated using 1000 segments
of noise, see Appendix A for details.

**Figure 3: **Combined 7-variable M-SSA of North Pacific region, (a) eigenval
ue spectrum, (b) leading pair of temporal principal components (T-PCs). The vari
ance associated with each T-PC is given in the top-right corner.

**Figure 4: **(a) Indices of spatial averages of RCs 1-2 for SST over (30-40
oN, 160-180oE) (solid line), and 500hPa geopotential height over (30-60oN, 160-1
80oE) (dotted). (b) Power spectrum of SST index, computed using MEM with 20 pole
s.

**Figure 5:** (a) Time series of \122raw\123 SST averaged over the region (30
oN-40oN, 160o-180oE) (thick line), together with RCs 1-2 from Fig. 4a. A 5-yr ru
nning smoother has been applied to the raw data. (b) Power spectrum of raw SST i
ndex, computed using MEM with 40 poles.

**Figure 6: **Composite cycle of RCs 1-2 for (a) SST over (30-40oN, 160-180oE
) (solid line), and 500hPa geopotential height over (30-60oN, 160-180oE) (dotted
).

**Figure 7: **Composite maps of RCs 1-2 at yrs 0 and +4 of the cycle in Fig.
6. (a) and (b): SST (contour interval 0.02K); (c) and (d) sea-level pressure (co
ntour interval 0.05mb); (e) and (f): geopotential height at 500hPa (contour inte
rval 0.5gpm). Negative anomalies dashed. Stippling indicates those grid boxes wh
ere a t-test on the mean of the 10 events in the composite exceeds the 99% signi
ficance level.

**Figure 8: **Composite maps of RCs 1-2: oceanic heat content (contours every
0.01K) and current vectors at 150m at yrs -2, 0, +2, +4, +6, and +8 of cycle in
Fig. 6. As Fig. 7.

**Figure 9: **Climatological mean ocean heat content (contours every 1oC) and
currents at 150m depth.

**Figure 10: **As Fig. 7, but for (a) and (b): net heat flux into the ocean;
(c) and (d) surface-layer advective temperature tendency; (e) and (f) advective
tendency at 150m. Contour interval is 0.2 K/decade, with the surface heat flux
assumed to be distributed over a 50m depth of water.

**Figure 11: **Composite cycle of 150m temperature (solid line) and advective
tendency (dotted line) averaged over the regions (a) (160-180oE, 30-40oN), and
(b) (140-160oE, 25-35oN). Crosses denote the residual of the heat budget.

**Figure 12: **Composite cycle of surface layer temperature (SST, solid line)
, advective tendency (dotted line), and heat flux into the ocean (dashed line) a
veraged over the regions (a) (160-180oE, 30-40oN), and (b) (180-160oW, 30-40oN).
Crosses denote the residual of the heat budget. The surface heat flux is assume
d to be distributed over a 50m depth of water.

**Figure 13: **Monte-Carlo significance test of North Pacific GISST annual me
an SSTs. The error bars indicate the 90% confidence interval of the AR(1) noise
null hypothesis. As Fig. 2.

**Figure 14: **Indices of the leading two GISST RC pairs over the 5o-box (35-
40oN, 150-155oW). (a) RCs1-2 (thick line), and RCs3-4 (thin line); (b) raw GISST
SST (thick line), and the sum RCs1-4 (thin line).

p>
**Figure 15: **Composite maps of GISST RCs1-2 (30-yr component) and RCs3-4 (6
-yr component), for the peak positive phase of the cycle and the following quadr
ature transition phase. (a) RCs1-2 peak phase, (b) RCs1-2 transition phase, (c)
RCs3-4 peak phase, (d) RCs3-4 transition phase. Contour interval is 0.04K. Stipp
ling indicates those grid boxes where a t-test on the mean of the events in the
composite (3 for RCs 1-2, 4 for RCs 3-4) exceeds the 99% significance level.

~