Combined Empirical Orthogonal Function Analysis Research Articles

Role of the Australian High in Seasonal Phase Locking of the Indian Ocean Dipole

AbstractThis paper analyzes the effect of the Australian High (AH) on the seasonal phase locking of Indian Ocean Dipole (IOD) events. The anomalous strong AH associated with the positive phase of the Antarctic Oscillation can cause significant easterly wind anomalies and northward cross‐equatorial flow over the western Maritime Continent (MC) by strengthening the Australian winter monsoon during May–August. The AH‐associated easterly anomalies and northward cross‐equatorial flow can create thermodynamic air‐sea feedback and contribute to a significant cooling anomaly in the western MC and the tropical eastern Indian Ocean. Without considering the effect of ENSO, these processes contribute to the occurrence of positive IOD events, which begin in early summer, peak in late summer, and decay rapidly thereafter. The effect of ENSO can extend the peak period of IOD into the boreal autumn of that year. An anomalous weak AH corresponds to the occurrence and seasonal phase locking of negative IOD events. Through combined empirical orthogonal function analysis, we find that the effect of the AH can well explain the seasonal phase locking of 34 IOD events (40 in total), which provides an important theoretical basis for the prediction of IOD events.

On the Relationship between the North Pacific Climate Variability and the Central Pacific El Niño

Abstract This study examined connections between the North Pacific climate variability and occurrence of the central Pacific (CP) El Niño for the period from 1950 to 2012. A composite analysis indicated that the relationship between the North Pacific sea surface temperature (SST), along with its overlying atmospheric circulation, and the CP El Niño during the developing and mature phases was changed when the occurrence frequency of the CP El Niño significantly increased after 1990. Empirical orthogonal function (EOF) and singular value decomposition (SVD) analyses of variability in the tropical Pacific and its relationship to the North Pacific show that the North Pacific anomalous SST and the atmospheric variability are more closely associated with the occurrence of the CP El Niño after 1990 than before 1990. There were noticeable differences in terms of the atmospheric variability conditions over the North Pacific, such as the North Pacific Oscillation (NPO)-like atmospheric variability during the spring and its associated SST anomalies during the following winter before 1990 and after 1990. In addition, combined EOF analysis also indicated that the NPO-like atmospheric circulation becomes more effective at playing a role in initiating El Niño after 1990. Consequently, such a change might have been associated with the frequent occurrence of the CP El Niño after 1990.

MJO structure associated with the higher-order CEOF modes

The real-time multivariate Madden–Julian oscillation (RMM; MJO) index has been widely employed to monitor the amplitude, phase, and time evolution of MJO events, as the index is formulated from the leading two combined-empirical orthogonal function (CEOF) modes of daily anomalous OLR and 850- and 200-hPa zonal winds, and the modes describe the MJO dynamics well. These two CEOF modes, however, are known to dominate in power spectra at zonal wavenumber one and may underestimate the power and structure at wavenumbers 2–5 where many MJO events are also prominent. This study approximated a baseline for MJO by applying band-pass filters to daily anomalies on 30–100 day periods and at 1–5 eastward propagating waves, as slightly different bands led to the same conclusions. Following the procedures to develop the RMM index, the daily anomalous data were derived and subjected to the CEOF analysis with all modes archived for diagnosis. Different numbers of the leading modes were compared in explained variance, standard deviation (STD), and wavenumber power spectra to describe the overall MJO magnitude and structure, and on the Hovmoller diagrams to represent the evolution of three distinct MJO events. Results show that the two leading CEOF modes explain only a small portion of the power spectra at wavenumbers 2–5. This spectral leakage notably reduces the MJO amplitude, particularly of the OLR in the western Pacific. The CEOF modes 3–10 can withhold power sufficiently such that the anomalies reconstructed by the first 10 modes contribute most of the baseline variance; their structures agree well with the baseline by constituting nearly the same proportion in the region from the central Indian Ocean to the dateline and by providing more complete evolutions of the three MJO events on the Hovmoller diagrams. Meanwhile, these modes introduce a notable amount of power for the equatorial Rossby and Kelvin waves that are partially embedded in the evolution of MJO. The first 50 of the total 432 CEOF modes retain all variance of the baseline MJO, while those higher than 10 contain less information and more noise and can be discarded. Furthermore, this study indicated that the longitudinal STD of the reconstructed anomalies detects the MJO phases and magnitudes in the western Pacific with more physical meaning and in better agreement with the Hovmoller diagrams than the RMM-like amplitude. The results provide an integral figure of the MJO structure from the CEOF analysis and a more robust RMM framework for monitoring the MJO’s evolution in real time and for validating its numerical forecast and simulations.

The Spatiotemporal Structure of Twentieth-Century Climate Variations in Observations and Reanalyses. Part II: Pacific Pan-Decadal Variability

Abstract The spatiotemporal structure of Pacific pan-decadal variability (PDV) is isolated in global long-term surface temperature (ST) datasets and reanalysis atmospheric parameter fields from which El Niño–Southern Oscillation (ENSO) effects have been removed. Empirical orthogonal function (EOF) and combined EOF analysis of the resulting time series identify PDV as one of two primary modes of long-term variability, the other being a global warming (GW) trend, which is addressed in a companion paper (Part I). In this study, it is shown that one of several PDV interdecadal regime shifts occurred during the 1990s. This significant change in the Pacific basin is comparable but antiphase to the well-known 1976 climate regime shift and is consistent with the observed changes in biosystems and ocean circulation. A comprehensive picture of PDV as manifested in the troposphere and at the surface is described. In general, the PDV spatial patterns in different parameter fields share some similarities with the patterns associated with ENSO, but important differences exist. First, the PDV circulation pattern is shifted westward by about 20° and is less zonally extended than that for ENSO. The westward shift of the PDV wave train produces a different North American teleconnection pattern that is more west–east oriented. The lack of a strong PDV surface temperature (ST) signal in the west equatorial Pacific and the relatively strong ST signal in the subtropical regions are consistent with an atmospheric overturning circulation response that differs from the one associated with ENSO. The analysis also suggests that PDV is a combination of decadal and/or interdecadal oscillations interacting through teleconnections.

Derivation of Salinity Profiles in the Indian Ocean from Satellite Surface Observations

Subsurface salinity profiles in the Indian Ocean have been constructed using a combined empirical orthogonal function analysis technique and a nonlinear data-fitting algorithm, which is known as a genetic algorithm. The purpose is to establish a methodology to generate a 3-D structure of salinity using satellite-derived surface observations, which will be useful for assimilating in numerical models. This acquires greater significance in view of the upcoming satellite salinity missions. In this letter, we have been able to generate vertical profiles of salinity using a combination of sea surface temperature derived by the Tropical Rainfall Measuring Mission (TRMM) Microwave Imager onboard TRMM and sea surface salinity from Argo floats. The root mean square error of the retrieved salinity varies from less than 0.05 psu-78 at the surface to about 0.15 psu-78 at greater depths. A sensitivity study of the retrieval algorithm with the anticipated uncertainties in surface salinity from the European Soil Moisture and Ocean Salinity satellite has also been carried out