Dataset DOI: 10.5061/dryad.np5hqc07p

Dataset overview

This dataset accompanies the study: “Oceanic mesoscale eddies enhance the Pacific Decadal Oscillation and its predictability”.

The dataset contains MATLAB scripts (.m) and processed data files (.mat) used to reproduce all figures in the manuscript.

Scientific context

The Pacific Decadal Oscillation (PDO) profoundly influences marine ecosystems, fisheries, and global hydroclimate. While traditionally interpreted as basin-scale oceanic responses to atmospheric stochastic forcing, whether its dynamics involve active ocean feedbacks remains unresolved. Using an unprecedented multi-century eddy-resolving global climate simulation, we find that mesoscale eddy-driven atmospheric anomalies in the Kuroshio Extension (KE) region are critical to PDO evolution. During the PDO cold phase, the northward-shifted meandering KE generates mesoscale sea surface temperature (SST) anomalies that intensify the lateral diabatic heating gradient, driving deep updrafts that cool the mid-troposphere and weaken its northern baroclinicity. This suppresses transient eddy momentum flux and facilitates a basin-scale low-pressure anomaly, initiating transition to the warm phase. Concurrently, mesoscale eddy-induced vertical heat transport sustains SST anomalies, providing additional PDO memory. These coupled processes substantially enhance the PDO’s predictability. Our findings highlight the previously underappreciated upscale effects of oceanic eddies, offering important insight into decadal climate variability.

The analysis is based on:

• CESM high-resolution simulations (eddy-resolving; CESM-HR)
• CESM low-resolution simulations (CESM-LR)
• observational and reanalysis datasets (e.g., HadISST, 20CRv3)
• paleoclimate proxy reconstructions

All data provided here are processed outputs used directly for figure generation.

File structure and detailed file descriptions

The repository is organized by figure number. Each figure folder contains a MATLAB script (SA_FigureX.m) and the corresponding data files (.mat).

Figure1 folder

Purpose: PDO characteristics and model evaluation (Fig. 1A to 1E)

SA_Figure1.m - MATLAB script to generate Figure 1

SST_reg_PDO_*.mat - Winter (DJF) SST regression onto PDO index. Three files: HadISST observations (1870-2020), CESM-HR (years 146-645), CESM-LR (years 50-549)

SLP_reg_PDO_*.mat - Winter (DJF) SLP regression onto PDO index. Three files: 20CRv3 reanalysis (1870-2015), CESM-HR, CESM-LR

TS_reg_PDO_*.mat - Winter (DJF) surface air temperature (SAT) regression onto PDO index. Three files: 20CRv3 reanalysis, CESM-HR, CESM-LR

Boxplot_HR_LR_vs_HadISST_150yr.mat - Boxplot data for skill score comparison between CESM-HR, CESM-LR, and HadISST

Proxy_location.mat - Geographic coordinates of paleo-proxy sites (used in Fig. 1C).

PDO_Index_CESM_Proxy.mat - Processed PDO indices from CESM-HR, CESM-LR, and paleo-proxy reconstructions. Original Paleo-proxy data used in this study are available from the NOAA National Centers for Environmental Information (NCEI) Paleoclimatology Program.

Figure2 folder

Purpose: Coupled ocean-atmosphere evolution in the KE region (Fig. 2A and 2B)

SA_Figure2.m - MATLAB script to generate Figure 2

SST_Lag_PDO_146_645_DJF_FFT1030.mat - Winter (DJF) SST lag regression onto PDO index, CESM-HR, 10-30 year bandpass filtered

SSH_Lag_PDO_146_645_DJF_FFT1030.mat - Winter SSH lag regression onto PDO index, CESM-HR, 10-30 year bandpass filtered

SLP_Lag_PDO_146_645_DJF_FFT1030.mat - Winter SLP lag regression onto PDO index, CESM-HR, 10-30 year bandpass filtered

Figure3 folder

Purpose: Basin-scale atmospheric response to KE variability (Fig. 3A to 3G)

SA_Figure3.m - MATLAB script to generate Figure 3.

TP_removed_*Lag_KEI_*DJF.mat - All data files contain lag regressions onto the normalized Kuroshio Extension jet (KEJ) index, with tropical teleconnections removed (indicated by "TP_removed"). Two model resolutions are included: CESM-HR (years 146-645) and CESM-LR (years 50-549). A lag of -1 year indicates KE leads atmosphere; +1 year indicates atmosphere leads KE. File naming conventions indicate variable, lag, and model name: GPH250A (250-hPa geopotential height anomaly), SLPA (sea level pressure anomaly), UV (transient eddy momentum flux, u'v'), WAF250 (250-hPa wave activity flux).

Zonal160220_L300850_146_646_DJF.mat - Zonal-mean (160°E to 140°W) and vertical-mean (300-850 hPa) diagnostics for Figure 3G. Contains: u (zonal wind, m/s), e (poleward gradient of u'v'), dzdtv (geopotential height tendency from transient eddy vorticity flux convergence, m/day), gph (geopotential height, m).

Figure4 folder

Purpose: Mesoscale air-sea coupling within the boundary layer (Fig. 4A to 4C)

SA_Figure4.m - MATLAB script to generate Figure 4.

*_Lag_KESSH_01yr_147_646_Feb.mat - All data files are lag +1 year regressions onto the KEJ index, for February. Files with "GS600X300" prefix are high-pass filtered (6°lon × 3°lat) to isolate mesoscale signals; files without this prefix are unfiltered fields (shown as insets in Fig. 4). Variables: SSH (sea surface height anomaly, m), SST (sea surface temperature anomaly, °C), THF (turbulent heat flux, W/m²), TMQ (column-integrated specific humidity, kg/m²), Qint (column-integrated diabatic heating, Pa K/day), OMEGA700 (700-hPa vertical velocity, -ω, Pa/s), UVdiv (surface wind convergence, s⁻¹).

2.5 Figure5 folder

Purpose: Tropospheric ocean-atmosphere coupling at oceanic mesoscales (Fig. 5A to 5C)

SA_Figure5.m - MATLAB script to generate Figure 5.

145165E_*_Lag_KEJ_01yr_146_646_Feb.mat - All data files are lag +1 year regressions onto the KEJ index, for February, averaged over the zonal band 145°E to 165°E. Variables: OMEGA (700-hPa vertical velocity, -ω, Pa/s), LapQdiab (Laplacian of diabatic heating, 1/(Pa·s³)), T (air temperature, K), Ty (meridional temperature gradient, K/m), Tbudget (thermodynamic budget terms, K/day), including: Ttend (total temperature tendency), Qdiab (diabatic heating), Hor (total horizontal heat transport), hor_mean (horizontal mean flow), hor_eddy (horizontal eddy), Ver (total vertical heat transport), ver_mean (vertical mean flow), ver_mean_adiab (vertical adiabatic), ver_mean_advec (vertical advective), ver_eddy (vertical eddy).

Figure6 folder

Purpose: Heat budget analysis for SST persistence (Fig. 6A and 6B)

SA_Figure6.m - MATLAB script to generate Figure 6.

KE_Budget_terms_reg_PDO.mat - KE region heat budget terms regressed onto the PDO index (W/m²), including: TD (temperature tendency), SST (SST tendency), Qsurf (surface heat flux), Qbasin (basin-scale advection), Qeddy_h (horizontal eddy heat transport), Qeddy_v (vertical eddy heat transport), Qturb_v (vertical turbulent mixing).

KE_Qbasin_advec_decom_terms_reg_PDO.mat - Decomposition of mean flow advection regressed onto the PDO index (W/m²), including: UcTc/VcTc/WcTc (background flow advecting background temperature), UcTa/VcTa/WcTa (background flow advecting anomalous temperature), UaTc/VaTc/WaTc (anomalous flow advecting background temperature), UaTa/VaTa/WaTa (anomalous flow advecting anomalous temperature).

Figure7 folder

Purpose: PDO predictability analysis (Fig. 7A to 7C)

SA_Figure7.m - MATLAB script to generate Figure 7.

APT1_Space_5yr.mat - Leading Average Predictability Time (APT) mode spatial pattern (5-year moving averaged)

APT1_R2_5yr.mat - APT1 potential predictability: R-squared as a function of lead time

R2_Damping_persistence_5yr.mat - Damped persistence forecast skill (reference, dashed lines in Fig. 7C)

colorbar folder

Contains colormap files for figure generation.

Variable definitions

Common variables across all .mat files:

• lon: Longitude (degrees east)
• lat: Latitude (degrees north)
• time: Time (years)
• coef: Regression coefficient (units vary by variable)
• pval: Statistical significance p-value (dimensionless)

Code/software

MATLAB (tested with R2020b and later versions)