AGCI Session II: Characterizing and Communicating Scientific Uncertainty
Session Chairs: Dr. Richard H. Moss and Dr. Stephen H. Schneider
July 31 to August 8, 1996
Recent Changes in El Niño/Southern Oscillation and Climatic Implications
Kevin Trenberth
National Center for Atmospheric Research
Boulder, Colorado
Trenberth focused on the climatic significance of and changes in the El Niño/Southern Oscillation (ENSO) phenomenon. He also discussed the "mini global warming" that accompanies ENSO and which affects the Northern Hemisphere temperature record. In reference to James Hansen's data that show 1995 to be the warmest year on record, Trenberth highlights the pattern of the warming, which came largely in the first six months of the year and largely over North America and Eurasia. This pattern suggests that the observed warming may not be entirely due to the enhanced greenhouse effect because atmospheric circulation patterns, specifically the North Atlantic Oscillation and ENSO, appear to play a contributory role.
Northern Hemisphere winter temperature data for 1977-1993 reveal an overall warming anomaly relative to a base period (1951-1989) with a characteristic pattern related to atmospheric circulation. Trenberth showed data which suggest that changes in the North Atlantic Oscillation and the Southern Oscillation account for the major portion of this observed anomaly. Why are these atmospheric circulation patterns changing as they are? Gaining a fuller understanding of natural variability modes in the climate system is of central importance in this inquiry.
The focus here
is on time series analyses which attempt to quantify how unusual
recent behavior in ENSO has been.
Much of Trenberth's recent work has focused on analyzing how unusual the recent behavior of ENSO has been. An article entitled, "What is Happening to El Niño?" by Trenberth appears in the 1997 Yearbook of Science and the Future, Encyclopedia Britannica, and deals more comprehensively with this subject. The original scientific article in which this work is reported appears in Geophysical Research Letters , 23:57-60, 1996. The focus here is on time series analyses which attempt to quantify how unusual recent behavior in ENSO has been.
"El Niño" technically refers only to the ocean warming off South America, but it is linked to important basin-scale changes in temperature in the tropical Pacific Ocean. In one extreme El Niño event in December of 1982, the Western Pacific warm pool extended well out into the tropical Pacific. Temperature gradients are reflected in surface pressure gradients which determine wind patterns and thus rain and atmospheric circulation patterns. The Southern Oscillation is a global wave pattern that is the atmospheric part of this phenomenon. Together, these make up a whole cycle of warming and cooling which we call ENSO, a natural mode of the coupled tropical Pacific ocean-atmosphere system arising from air-sea interactions. This is basically the mechanism by which El Niño influences climate around the world.
El Niño events occur on average every 3 to 6 years; paleoclimatic data (from corals and ice cores) show that such events have been going on for millennia. Characteristically, we see a warming in the traditional El Niño region of the tropical Pacific that is connected to a warming across the basin and to a boomerang-shaped cooling in the North and South Pacific caused directly by changes in atmospheric circulation (SO). La Niña is a cold event in which a cold tongue of water extends along the equator into the warmer water and causes changes in rainfall patterns in the tropical Pacific and changes in temperature patterns, storm tracks, and atmospheric circulation in North America and elsewhere.
Correlations associated with the SO were discussed, related to annual mean sea level pressure. When pressure is high at Darwin, Australia, it tends to be high across the entire region extending across Africa all the way to Brazil. At the same time, there are low pressure conditions across the central and eastern Pacific, north Pacific and southern oceans which lead to cloudy, rainy weather (while high pressure in other regions leads to more dry and settled conditions). This pattern leads to droughts in Australia, Africa and parts of Brazil and Columbia, and excessive rains in other regions. There is also a structure in this pattern, associated with wintertime conditions, that is called the Pacific /North American teleconnection pattern.
El Niño
events occur on average every 3 to 6 years; paleoclimatic data (from
corals and ice cores) show that such events have been going on for
millennia.
The two data-collecting stations that have the highest negative correlations of any in the world are Darwin and Tahiti. At these two stations, sea level pressures vary, very distinctly, in opposite directions (see Figure 2.17). The difference between these two stations is thus often used as an index of the Southern Oscillation. One way of completely recovering the information from two stations is to take both the difference and the sum. The SO signal is the difference and the sum, a measure of noise, or non-SO variations. The signal to noise ratio for these two monthly series is 2.0. An appropriate smoother is used to improve the signal to noise ratio to 6.4 and this helps to reveal all the El Niño and La Niña events that correspond to the fluctuations in the Southern Oscillation. Many noise features are simply related to individual small-scale temperature events that are not related to the see-saw of the SO events. The key in analyzing this phenomenon is to focus on the phenomenon itself, and not get hung up on all the noise inherent in the system.
The two
data-collecting stations that have the highest negative correlations
of any in the world are Darwin and Tahiti. At these two stations, sea
level pressures vary, very distinctly, in opposite directions.
One problem is that the early record from the Tahiti station is noisy, incomplete and unreliable due to the fact that records were stored in a documentary form that was affected by bugs, mold, etc., so Trenberth doesn't use this older data. There are almost no sea surface temperature (SST) data from the tropical Pacific prior to 1950. It is thus very difficult to get reliable measurements of EN during that period of time. At the same time, it is clearly important to have a homogeneous time series of EN behavior in recent times. The SO can be reconstructed back to 1882.
After about 1976, a series of unprecedented events occurred. One La Niña and several El Niño events, including the very long 1990-1995 EN event, are unlike anything else in the record. Questions are thus raised regarding how unusual this behavior actually is and why it occurred. Prior to describing in more detail the time series analyses used to research these questions, Trenberth showed a correlation of SSTs with the SO from 1950 to 1994 in the Central Pacific which reveal the characteristic boomerang pattern and with warming along the coast of California associated with EN events (see Figure 2.18). The SO is most strongly related to SSTs in the central Pacific, not along the coast of South America.
When we look at SSTs in the region Trenberth calls "Niño 3 and a half" (5°N-10°S, 180°-120°W) we see a tendency for more EN behavior recently including the prolonged event of 1990-1995, while along the coast of South America, the traditional EN region, we see more ups and downs and less tendency for typical EN behavior. Activity in this central Pacific "Niño 3 and a half" region seems to generate most of the global consequences. Looking at the recent prolonged EN event of 1990-1995, the question arises: is this a manifestation of global warming?
Looking at recent behavior in the context of the past record, we can see that the recent prolonged EN lasted from October 1989 to June 1995, 5 years and 9 months. The next longest EN in the record was from 1911 to 1915, 4.11 years. From 1906 to 1911 there was a 5-year event of opposite sign. Looking at SSTs from 1950 forward, we see that in the period from September 1989 to August 1995, sea surface temperatures were above normal. In terms of seasonal values for the Southern Oscillation, the most recent event was 22 seasons all of one sign (negative), from 1989 to 1995. The longest previous such event lasted for 15 seasons, from 1894 to 1897, and the longest such event of opposite sign lasted 13 seasons. In this context, recent behavior looks very unusual, so statistical tests were conducted to try to quantify just how unusual this behavior is.
Allowing for appropriate degrees of freedom based on the effective time between independent samples, the test results reveal a strong statistical significance of these recent events. For the 66 months from December 1989 to May 1995, a high statistical significance of t=2.82, departure 0.87 mb, significance at 0.5 percent was found. For the period of March 1977 to May 1995, t=2.68, significant at 1 percent.
After about
1976, a series of unprecedented events occurred. One La Niña
and several El Niño events, including the very long 1990 -1995
EN event, are unlike anything else in the record.
Then, autoregressive-moving average (ARMA) modeling was used to better deal with persistence in the time series, using seasonal anomalies in the Darwin station data. The model was used to generate a synthetic one million-year record that contains many simulated ENSO events, averaging one about every four years, that can be tested to see how unusual the recent behavior is. Summarizing the statistical conclusions that emerge from this work: the low frequency variability and the negative trend in the Southern Oscillation Index observed in recent decades are quite unusual. Likelihood of occurrence of the 1990-1995 ENSO, given the first 100 years of the record, is once in 1,500 to 3,000 years. The change beginning about 1976 toward more frequent ENSO events is also unlikely, about a once in 2,000 year expectation. These two events are not independent and the results suggest a non-stationary influence such as would arise from a change in climate.
How will ENSO events change with global warming? ENSO involves a build up and depletion of heat as well as major redistribution of heat content in the ocean and the atmosphere during the course of events. Because GHGs trap heat, they alter the heat budget. As such, they can possibly expand the Pacific warm pool and enhance the rate of recharge of heat losses. Models which include crude simulations of ENSO show that conditions become more ENSO-like with increased GHGs. Second, the hydrological cycle is expected to speed up with increased GHGs. Increased evaporation enhances the moisture content of the atmosphere which makes moisture more available for rainfall. ENSO-related droughts are apt to be more severe and last longer, while floods are likely to be enhanced.
In discussion, MacCracken pointed out that this work demonstrates that what may seem like relatively small and subtle changes in temperature can cause dramatic changes in elements of the climate system like the ENSO phenomenon and in impacts such as precipitation. Further, he pointed out that while we often speak of phenomena like ENSO and changes in them as "natural," it makes sense that as the climate warms, it would take advantage of such elements of the system, perturbing these natural modes of the system first and evidencing itself in this way.
How will ENSO
events change with global warming? ... ENSO-related droughts are apt
to be more severe and last longer, while floods are likely to be
enhanced.