369557 Reanalysis of the Extended Multivariate ENSO Index

Tuesday, 14 January 2020
Hall B1 (Boston Convention and Exhibition Center)
Eric Webb, University of North Carolina - Charlotte (UNCC), Charlotte, NC; and B. Magi

The El Nino Southern Oscillation (ENSO) is the most robust, coupled ocean-atmospheric component of intraseasonal-interannual variability on the globe. The development of simple, one or two-dimensional ENSO indices to characterize the state of the tropical Pacific coincided with the rise in research on ENSO during the 1980s and 1990s, becoming a backbone for quantitatively diagnosing ENSO teleconnections. Rasmusson and Carpenter (1982) developed NINO indices (NINO 1-2, NINO 3, and NINO 4) based on ship tracks over the east-central equatorial Pacific, and with improved analyses as well as additional data, Barnston et al (1997) further defined the NINO 3.4 index as representing the “core” of ENSO-related SST anomalies in the equatorial Pacific.

In concert with the rise of ENSO indices near the turn of the millennia, the Multivariate ENSO Index (MEI) became one of the most widely used ENSO indices, being utilized to study not only ENSO, variability in the northern annular mode, stratospheric temperature changes in the satellite era, global sea level reconstructions, hydroclimate variability in proxy tree-ring and coral records, marine heat waves, wildfire activity, extreme weather events, and climate sensitivity in GCMs. Version 1 of the MEI (MEIv1) was defined as the first unrotated Principal Component (PC) of six combined, primary observed surface variables over the Equatorial Pacific Ocean (SST, surface air temperature, sea level pressure, meridional winds, zonal winds, and total cloudiness fraction of the sky) computed separately over 12 sliding bi-monthly seasons from the International Comprehensive Ocean Atmosphere Dataset (ICOADS). By using Principal Component Analysis (PCA) across multiple variables, this allows the components MEIv1 to vary with the seasonal cycle, and the inclusion of more variables makes MEIv1 less susceptible to instrumentation and reconstruction errors that may be contained within individual variables from singular datasets. Therefore, MEIv1 is inherently more stable than other geographically fixed, univariate ENSO indices such as NINO 3.4 SSTs and ONI (Barnston et al (1997), Ropelowski & Jones (1987), Wright (1989)), and the Southern Oscillation Index (SOI) (Allan et al (1991)).

The original Extended Multivariate ENSO Index (MEI.ext) described in Wolter & Timlin (2011) adapts the same methodology as the MEIv1 and extends the index backwards in time to range from 1871 to 2005. The MEI.ext is defined by Wolter and Timlin (2011) as the combined, bi-monthly, leading principal component analysis (PCA) of Sea Level Pressure (SLP) and SST over the Tropical Pacific domain (30S-30N, 100E-100W) in the UK Met Office Hadley Centre SST dataset version 2 (HADSST2) (Rayner et al (2006)) and the UK Met Office Hadley Centre SLP Dataset version 2 (HADSLP2) (Allan and Ansell (2006)) datasets. The exclusion of other variables used in the original MEI, such as OLR, zonal and meridional wind, is largely attributed to the lack of data and datasets describing them before the satellite era. Therefore, as was the case with the original MEI.ext, only SST and SLP will be utilized in this reanalysis of the Extended MEI.

This reanalysis of MEI.ext will provide a number of major improvements to this popular index. These improvements include a longer available record ((1865-2019) vs (1871-2005)), the utilization of sliding base periods and detrending to account for significant background mean state changes in both the climate and observations since the mid-19th century, consideration of the relative uncertainties in SST and SLP data, and using an ensemble of observationally-based and reanalysis datasets. The latter involves the projection of SST and SLP data from various available datasets onto each other’s leading Empirical Orthogonal Function (EOFs) of SST and SLP in the Tropical Pacific, allowing for the retrieval of a very large set of potential ENSO realizations. Experiments that altered the resolution and temporal scale of the datasets and EOFs respectively, were also conducted to test the sensitivity of the index to these parameters, and also increase the stability, confidence, and reliability in this new version of the MEI.ext. In conjunction with randomization experiments, such an analysis provides important clues on how the datasets respond to inhomogeneities and uncertainties that emerge from varying observational input over the observed record. Furthermore, this allows for exploration of structural uncertainties that emerge between available datasets, as well as the sensitivity of the MEI.ext to these uncertainties and confidence in their output.

Since the release of MEI.ext (Wolter & Timlin (2011)), the number of SST datasets available to the construction of this reanalyzed version of the MEI.ext has nearly doubled, and multiple revisions of ICOADS have been released and, along with it, the addition of tens of millions of new observations in the 19th and 20th centuries. This clearly suggests that significantly more, new information regarding ENSO behavior in the instrumental record (1850-present) is available for analysis and that a major revision of MEI.ext is clearly warranted.

Preliminary results from this reanalyzed version of the MEI also generally confirm some of Wolter & Timlin (2011)’s findings and closely follow other available ENSO indices, which generally portend that ENSO was more active near the turn of the 20th and 21st centuries and relatively more quiescent in the middle of the 20th century. It was also discovered that astounding similarities exist in the interannual evolution of ENSO behavior between mid-late 2010s and in the mid-late 1870s and early 1880s.

This new version of the MEI.ext is intended to be a major improvement upon the work of Wolter & Timlin (2011) and effectively synthesize the current state of knowledge regarding ENSO during the “observational era” (1850 - present). This version of the MEI.ext uses as much observationally-based data as possible, while also implementing various uncertainty and climate base state adjustments to improve the quality and reliability of the data. Given that a majority of the datasets utilized in the construction of this reanalyzed version of the MEI.ext were released within just the past several years, this kind of work would not possible until recently, so this will provide a very modern retrospective on ENSO behavior from the mid-19th century to the present.

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner