Interannual variability of tropical Pacific sea level from 1993 to 2014

Abstract

A multimode, linear reduced‐gravity model, driven by ERA‐Interim monthly mean wind stress anomalies, is used to investigate interannual variability in tropical Pacific sea level as seen in satellite altimeter data. The model output is fitted to the altimeter data along the equator, in order to derive the vertical profile for the model forcing, showing that a signature from modes higher than mode 6 cannot be extracted from the altimeter data. It is shown that the model has considerable skill at capturing interannual sea level variability both on and off the equator. The correlation between modeled and satellite‐derived sea level data exceeds 0.8 over a wide range of longitudes along the equator and readily captures the observed ENSO events. Overall, the combination of the first, second, third, and fifth modes can provide a robust estimate of the interannual sea level variability, the second mode being dominant. A remarkable feature of both the model and the altimeter data is the presence of a pivot point in the western Pacific on the equator. We show that the westward displacement of the pivot point from the center of the basin is strongly influenced by the fact that most of the wind stress variance is found in the western part of the basin. We also show that the Sverdrup transport is not fundamental to the dynamics of the recharge/discharge mechanism in our model, although the spatial structure of the wind forcing does play a role in setting the amplitude of the “warm water volume.”

Reconstructing Tropical Pacific Sea Level Variability for the Period 1961–2002 Using a Linear Multimode Model

“…Based on the method of Ebisuzaki (), these correlations are significantly different from zero at the 95% level with the exception of those at Kanton. The drop off in correlation at Kanton is consistent with the spatial pattern of correlation for the period 1993–2014 between AVISO and the model‐computed sea level anomalies noted by Zhu et al () and shown by the color shading in Figure . The region of relatively low correlation near Kanton Island is, in turn, related with the misplacement of the pivot point in the model, a topic we return to in section 3.3. At Christmas Island, the model has skill at capturing events, especially in the ERA‐40 experiment, but generally underestimates the amplitude, consistent with Figure in Zhu et al () (in particular, compare their Figures a and b).…”

“…In our experiments, except for the wind forcing, the model configuration is the same as in Zhu et al (), including the model domain ( S– N, E– W), coastline (300 m isobath), horizontal resolution ( ), boundary conditions (solid walls at the eastern/western boundaries, sponge layers applied to the momentum equations at the northern/southern boundaries with e‐folding scale of in latitude) and a horizontal eddy viscosity of 5,000 m s −1 .…”

“… Locations of tide gauge stations (white squares): Rabaul in the western basin, Kanton Island and Christmas Island in the central basin, Baltra Island and Sancrist Island in the eastern basin. The contour shading denotes the correlation between monthly means of AVISO and model‐computed sea level for the period 1993–2014 (using the same color scale as in Figure 6a of Zhu et al () and note that there are no negative correlations in the region shown).…”

“…The objective of the present study is to reconstruct monthly sea level variability in the equatorial Pacific for the period 1961–2002 using the linear, multimode model of Zhu et al () driven by monthly mean wind stress fields from the NCEP/NCAR and ERA‐40 reanalysis products. The multimodel of Zhu et al () has its origin in the early work by Busalacchi and O'Brien () and Busalacchi et al (). However, unlike those early studies that consider only a single baroclinic normal mode, Zhu et al () consider the first five baroclinic normal modes.…”

“…The multimodel of Zhu et al () has its origin in the early work by Busalacchi and O'Brien () and Busalacchi et al (). However, unlike those early studies that consider only a single baroclinic normal mode, Zhu et al () consider the first five baroclinic normal modes. They also do not specify a vertical structure for the wind forcing term; rather, the vertical structure is derived by fitting model‐computed sea level variability to that seen by satellite data along the equator.…”