There’s an interesting graph on the web site of the Univ. of Colorado sea level page comparing de-trended sea level to the multivariate el Niño index:
There’s also a note about a recent publication titled “The 2011 La Niña: So strong, the oceans fell” (Boening et al. 2012, GRL, 39, doi: 10.1029/2012GL053055). They note that because of its effect on patterns of evaporation and precipitation, the el Niño southern oscillation (ENSO) can cause short-term fluctuations in sea level.
It’s certainly possible to estimate the impact of ENSO on sea level using multiple regression. This is not what Boening et al. did in their publication — but let’s see what happens anyway. One of the complications is that satellite data for sea level contains a spurious oscillation with period about 59 days due to a periodic change in the satellite orbits. This causes extra variance of the sea level data, which is not due to genuine variance of sea level:
The U. Colo. group removed that spurious variation by applying a 60-day moving average (boxcar) filter. That does a fine job, although a small residue of the fluctuation remains because the period isn’t exactly 60 days and the data are not evenly spaced in time. I decided to do it a different way, by estimating the actual oscillation with Fourier analysis, then subtracting that out:
Then I fit these data to a linear time trend and the multivariate el Niño index (MEI). This enabled me to estimate the impact of ENSO on sea level, which can then be subtracted to create adjusted sea level data. Here’s the model itself compared to the unadjusted data:
Note that it does a fine job modelling the rise of sea level due to the 1998 el Niño, but it only captures part of the dip in sea level due to the 2010/2011 la Niña. Removing the estimated el Niño impact gives adjusted data:
One interesting result is that the raw sea level data exhibit a linear trend of 3.1 mm/yr sea level rise, but taking the el Niño adjustment into account the trend is a little faster at 3.24 mm/yr — although the difference isn’t statistically significant.
One of the reasons that simple multiple regression doesn’t capture much of the 2010/2011 sea level dip, may be that the impact of ENSO on sea level can depend on the season at which it occurs, as its effect on evaporation and precipitation, and therefore the movement of water between ocean and land, depends on the time of year.
Boening et al. didn’t relate sea level fluctuations to ENSO directly. Instead, they directly estimated the movement of water between ocean and land using gravity measurements from the GRACE satellite mission, and the effect of temperature variations on sea level using data from the Argo floats. They were able to show that the 2011 sea level drop was mainly due to a change in ocean mass, stating in their abstract:
Using a combination of satellite and in situ data, we show that the decline in ocean mass, which explains the sea level drop, coincides with an equivalent increase in terrestrial water storage, primarily over Australia, northern South America, and Southeast Asia. This temporary shift of water from the ocean to land is closely related to the transition from El Niño conditions in 2009/10 to a strong 2010/11 La Niña, which affected precipitation patterns world-wide.
The GRACE satellite data enable them to estimate the gain in water content over land during this time period (comparing the Jan-Mar 2010 average to the Mar-May 2011 average):
They particularly note the water increase over Australia, southeast Asia, and northern South America. These terrestrial water gains match the decline in ocean mass during this time span:
It must be noted that for this comparison, the ocean mass and terrestrial water storage figures include the loss of ice from the Greenland and Antarctic ice sheets.
Clearly, water transfer from ocean to land is responsible for the 2010/2011 sea level drop. But is that water transfer due to ENSO? The relationship between precipitation patterns and ENSO has already been established:
The increase in precipitation over land, and simultaneous decrease over the ocean, are consistent with findings of Gu et al. , who showed similar anomalies occurring during La Niña events in general. This indicates a strong connection between the transition to the 2010/11 La Niña, the changes in TWS and mass related sea level.
In this particular case, the vast quantity of water transfer from ocean to land seems to be related both to the extremity of the 2010/2011 la Niña, and to the seasonal timing of the event:
The time series of ENSO events represented by the Southern Oscillation Index (SOI) indicates that the 2010/11 La Niña was the strongest over the altimetry period starting in 1992 – and one of the strongest La Niña events for that season in the last 80 years. High precipitation events leading to flooding in Australia, Pakistan and China have been associated with the 2010/11 La Niña and also with record high sea surfaces temperatures in the Arabian Sea and north of Australia [Trenberth and Fasullo, 2012]. The cumulative influence of related synoptic events appears to have transported enough water to the continents to explain the 2010 drop in GMSL.
When the changes in sea level due to ocean mass (according to GRACE data) and thermal (steric) sea level change (according to Argo data) are combined, they explain the recent sea level changes exceptionally well, including the strong dip in 2010/2011:
The top panel compares observed (black) sea level since 2005 to that computed (red) from changes in ocean mass and temperature. The bottom panel shows the individual components making up the computation.
This new understanding of the 2010/2011 sea level drop, one of the more notable short-term fluctuations in sea level, raises hope that we may finally be getting a handle, not only on the slower changes due to global warming, but the faster changes due to fluctuation effects like ENSO. The match between satellite altimetry for sea level, satellite data for ocean mass (and land water storage), and Argo float data for ocean temperature, is nothing short of impressive. Better understanding of the short-term fluctuations can only help us understand the long-term trends, and therefore better anticipate the dangerous rise of sea level expected this century.
It also shows the folly of hopes that the 2010/2011 sea level drop should allay fears of continued global-warming induced sea level rise. As the authors say in their conclusion:
The connection to ENSO and the fact that most of the additional water on the continents at low and mid latitudes will be subject to runoff suggests a rather short-lived hiatus in GMSL rise. Indeed, the most recent data suggest a recovery of more than 5 mm (Figure 1) in the last few months of the GMSL time series despite the subsequent La Niña in 2011/12. ENSO-driven changes in GMSL like the one described here might mask GMSL variations related to anthropogenic forcing over short time periods, but as expected from the lag of continental freshwater outflows relative to precipitation anomalies, they are unable to curtail the longer timescale trends associated with persistent ice melt and ocean warming as observed in recent decades.
Predicting future rates of sea level change and detecting any acceleration in GMSL rise will require the ability to distinguish such events from increases in the net heat content of the ocean, as well as rapid changes in the amount of ice lost from the glaciers and ice sheets. This underscores the importance of complementary global observing systems such as Jason, GRACE and Argo, without which such distinctions would be impossible.