The Forcing of the Pacific Decadal Oscillation Schneider and Cornuelle, 2005 Patrick Shaw 3/29/06
Overlying Questions What processes statistically influence variability in the PDO? On what time scales do they influence the PDO? Is the PDO a single dynamical phenomenon, or something else?
The Goal Problem - PDO index characterizes variability in climate, but gives nothing about the causes Solution use statistics to determine how SST, SLP and winds impact the patterns in time and space
PDO Review Variability: interannual to decadal Correlated with other anomalies (precip., streamflow, SST, monsoons, ENSO) Horseshoe pattern of opposite anomalies Several additional processes complicate picture - statistics needed to resolve forcing issue
Indices N. Pacific SST has variation contributions from tropical ENSO and local patterns AR-1 process uses SLP time series as forcing NPI describes anomalies of Aleutian Low (average of SLP where skill exceeds.55 w/ Nino3.4 removed) and shows variability on all time scales
Figure 3 - Correlation (10x) of observed with reconstructed PDO from AR-1 process forced by SLP variation Top panel includes full SLP and Nino3.4 index Bottom panel Nino3.4 removed by taking out Areas of high correlation show SLP anomalies are most linear significant regression w/ observed of PDO Nino3.4 SLP has skill at reconstructing PDO in N. Pacific, SE Pacific, W Pacific warm pool, E. Indian Ocean and tropical Atlantic. Tropical skill associated with ENSO, so removal of Nino3.4 eliminates all but N. Pacific skill
The KOE Kuroshio-Oyashio Extension zonally oriented SST frontal zone associated with strong eastward surface currents, forming the boundary of the oceanic subtropical and the subpolar gyres (Nonaka, et al. 2005)
Impacts on KOE Rossby waves cause Ekman pumping to affect thermocline depth (p) which affect position and sharpness of KOE and SST Two ways to affect KOE SST Direct Changes in p - average between 35-40 N gives P AVG index Zonal advection - difference of zonal averaged p between 38 N and 40 N gives P DEL index P AVG and P DEL most affective on decadal timescales, so only account for some variability in SSTA
Figure 1 time evolution of different indices PDO leading EOF of monthly anomalies in SST poleward of 20 N Solid line on PDO July-June annual averages of SSTA (R=.99 with PDO) NINO 3.4 SST anomaly (5 S-5 N, 170-120 W) NPI N. Pacific index of average SLP in Aleutian low P DEL thermocline depth index in KOE (170-140 W, averaged between 35-40 N) P A VG zonal advection index, by taking differences of zonal
SSTA Reconstruction Forced by red noise time series of each index i? is spatially dependent damping term (assumed constant)? is spatial footprint of index
Figure 2 spatial distribution of PDO pattern contours of.2 K for unit deviation of PC Horseshoe pattern in temp. anomalies surround opposite sign in central N. Pacific Reconstructions closely mimic observed PDO in shape
Reconstruction Skill Correlations with observed PDO
Figure 5 Correlation (10x) of hindcast AR-1 SSTA model and observed PDO for each forcing Nino3.4 skill has most impact on eastern coast NPI has most impact over wider area of Pacific P DEL skillful in KoE
Rotated Indices Because time series of indices are not independent (due to covariances and teleconnections), indices must be orthogonalized to remove mutual impacts For example, Nino3.4 Nino3.4* Nino3.4 and NPI negatively correlated
Forcing Footprints Figure 8 forcing footprints, SST anomalies in K Nino3.4 eastern/coastal impact due to waves and teleconncections -NPI (note minus sign)
Figure 9 footprints of P DEL and P AVG KOE extension, advection Seasonal dipole of mixed-layer depth
Figure 10 Rotated footprints Effects of ENSO spread west because of NPI additions NPI looks much the same, P DEL and P AVG have little impact
Implications NPI* and Nino3.4* have nearly equal but opposite weights ENSO affects PDO by changing the NPI Warm E. Pacific yields cool central N. Pacific, etc. Positive NPI* (warmer central N. Pacific, lower pressures) means negative anomalies in PDO
PDO Time Evolution PDO is the sum of forcing of the four previously derived indices P AVG * has insignificant affect NPI accounts for most of variance, followed by Nino3.4, which also affects the NPI
PDO Power Spectrum NPI* and Nino3.4* affect PDO on interannual (10 0 ) frequencies * NPI*, Nino3.4*, P DEL on decadal (10-1 ) frequencies
Some Side Conclusions 1976/77 shift has tropical origin, as seen in symmetry in N and S hemispheres PDO plus ENSO has less effect on teleconnections compared to combination of ENSO and NPI Tree ring growth depends more on ENSO and NPI than it does on PDO alone A decadal PDO reconstruction would miss KOE anomalies
Answering Overlying Questions SSTA and PDO reconstructed from AR model which is forced by intrinsic variability in N. Pacific, changes in Aleutian Low, ENSO and KOE Different impacts have different frequencies: < year: intrinsic N. Pacific variability interannual: NPI and ENSO decadal: KOE wave anomalies PDO not a single climate mode, but a superposition of SST influences
Remaining Problems Does this mean ENSO influences PDO more than PDO influences ENSO? Is ENSO independent of PDO? Are these statistics valid? (still no causality) Who is correct? Schneider and Cornuelle or Karspeck and Cane