2The National Center for Atmospheric Research is sponsored by the National Science Foundation.
Accordingly, it has been very difficult to define El Niño or an El Niño event. The term has changed meaning: some scientists confine the term to the coastal phenomenon, while others use it to refer to the basin wide phenomenon, and the public does not draw any distinction. There is considerable confusion, and past attempts to define El Niño have not led to general acceptance. Clearly, the term El Niño covers a diverse range of phenomena. An earlier general review of the terminology confusion is given by Aceituno (1992).
Glantz (1996) has formally put forward a definition of El Niño as it should appear in a dictionary:
El Niño \ 'el ne nyo noun [spanish] \ 1: The Christ Child 2: the name given by Peruvian sailors to a seasonal, warm southward-moving current along the Peruvian coast <la corriente del niño> 3: name given to the occasional return of unusually warm water in the normally cold water [upwelling] region along the Peruvian coast, disrupting local fish and bird populations 4: name given to a Pacific basin-wide increase in both sea surface temperatures in the central and/or eastern equatorial Pacific Ocean and in sea level atmospheric pressure in the western Pacific (Southern Oscillation) 5: used interchangeably with ENSO (El Niño-Southern Oscillation) which describes the basin-wide changes in air-sea interaction in the equatorial Pacific region 6: ENSO warm event synonym warm event antonym La Niña \ [Spanish] \ the young girl; cold event; ENSO cold event; non-El Niño year; anti-El Niño or anti-ENSO (pejorative); El Viejo \ 'el vya ho \ noun [Spanish] \ the old man
from Glantz 1996). This definition
reflects the multitude of uses for the term but is not quantitative. Several
attempts have also been made to make quantitative definitions, although always
by choosing just one of the myriad of possibilities, and therefore falling
short for general acceptance.
Quinn et al. (1978) provided a listing of El Niño events and a measure of
event intensity on a scale of 1 to 4 (strong, moderate, weak, and very weak)
beginning in 1726. The measures used to define the El Niño
and its intensity
were primarily based on phenomena along the coast of South America and were
often qualitative.
In the early 1980s, a Scientific Committee for Ocean Research working group
SCOR WG 55 was set up to define El Niño (SCOR 1983)
and came up with the following:
``El Niño is the appearance of anomalously warm water along the coast of Ecuador and Peru as far south as Lima (12°S). This means a normalized sea surface temperature (SST) anomaly exceeding one standard deviation for at least four (4) consecutive months. This normalized SST anomaly should occur at least at three (3) of five (5) Peruvian coastal stations.''
This definition clearly refers to the event right along the South American
coast and it did not achieve acceptance from the scientific community.
For some time, the focus of much activity
related to ENSO has been the Niño 3
region, defined as 5°N to 5°S, 150°W to 90°W. This has been, for
instance, the primary predicted ENSO-related quantity by models verified by
observed data. The following is the working definition used by the Japan
Meteorological Agency (JMA). It is an objective procedure and the obtained El
Niño periods are quite consistent with the consensus of the ENSO research
community.
The 5 month running mean of SST anomalies is made in order to smooth
out the possible intraseasonal variations in the tropical ocean. The periods
that qualify define the El Niño
periods and provide a quantitative measure
of the intensity. The results of this technique
applied to the Niño 3 data
from NOAA are shown in Fig. 1.
Kiladis and van Loon (1988) used a standard Southern Oscillation index (SOI)
combined with an SST anomaly index for the eastern tropical Pacific (within
4° of the equator from 160°W to the South American coast) to define an
``event'' and required that the SST anomaly had to be positive for at least
three seasons and be at least 0.5°C above the mean, while the SOI had to
remain negative and below -1.0 for the same duration. They provide a listing
of ``Warm'' and ``Cold'' events from 1877 to 1982.
Various versions of the SOI exist although, in recent years, most deal only
with atmospheric pressures and usually only those at Darwin and Tahiti. In
using the SOI based on just two stations, it must be recognized that there
are many small scale and high frequency phenomena in the atmosphere, such as
the Madden-Julian Oscillation, that can influence the pressures at stations
involved in forming the SOI, but which do not reflect the Southern
Oscillation itself. Accordingly, the SOI should only be used when monthly
means are appropriately smoothed (Trenberth 1984, Trenberth and Hoar 1996a).
For many years, Tahiti data were available only after 1935.
Ropelewski and Jones (1987) outline an extension of the SOI prior to then
using newly discovered Tahiti data, and they also discuss different ways of
standardizing the data for use in the SOI. However, there are questions
about the integrity of the Tahiti data prior to 1935 (Trenberth and Hoar
1996a), as the Tahiti-Darwin correlation is much lower in spite of strong
evidence that the SO was present from other stations, and the noise level and
variance in the early Tahiti data is higher than in the more recent period.
Although Ropelewski and Jones (1987) state that they believe that months with
data contained a full complement of daily values, several monthly means are
missing and the monthly means that are present for the earlier period are
more consistent with the view that they originate from an incomplete dataset
in which values contributing to monthly means are missing. There are also
questions about whether the diurnal cycle in surface pressure, which contains
a strong semidiurnal tide component, has been adequately dealt with in
forming the means. Ropelewski and Jones (1987) use a 5 month running mean to
define their indices (as is done in NOAA's Climate Diagnostics
Bulletin).
In recent years, it has been apparent that the key region for coupled
atmospheric-ocean interactions in ENSO is somewhat farther west (Trenberth
1996a, Trenberth and Hoar 1996a). Thus Trenberth and Hoar (1996b) proposed a
Niño 3.5 region as the region 180°
to 120°W, 5°N to 10°S, which
straddles part of the Niño 3
and Niño 4 regions and extends farther into
the Southern Hemisphere as the key region for ENSO. For this region, ENSO
events might more appropriately be defined by a threshold of 0.3°C. This
slight shift in focus in the area of SST of importance has also been found in
the Climate Prediction Center of NOAA's National Centers for Environmental
Prediction, where since April 1996 they have introduced a new SST index
called Niño 3.4 for the region 5°N
to 5°S 170°W to 120°W in
their monthly Climate Diagnostics Bulletin. Similar criteria to the
JMA definition could be applied to such a region although a threshold lower
than 0.5°C might be appropriate.
Because the Niño 3.4 values are now
widely disseminated, we use those and the Niño 3 values to examine the
record after 1950.
The monthly standard deviations in Table 1 are of considerable
interest. Largest standard deviations of comparable values for both indices
occur in the northern winter of slightly over 1°C, while lowest values
occur in the northern spring. In Niño 3 the
minimum in March of 0.44°C
is less than half the November, December and January values while the standard
deviations are only slightly larger in the Niño 3.4 region. From May to
October, the reverse is the case; the standard deviations in the
Niño 3 region are slightly larger. This annual cycle of variance in the
SSTs is perhaps a primary reason for a focus on the northern winter for ENSO
that has been evident in much of the literature, as that is clearly when the
largest SST changes occur in the central Pacific. However, the annual cycle
in standard deviations of outgoing longwave radiation (OLR) in the tropics,
which indicates the direct convective response in the atmosphere, is
displaced one to two months later (see Trenberth et al. 1997). For OLR there
is a distinct minimum in interannual variance in September, corresponding to
the time of year when the east-west SST gradients are strongest and so SST
anomalies have less effect on the location of atmospheric convection. In
March and April on the other hand, small SST anomalies in Niño 3.4 can
greatly influence the location of the warmest water and thus the preferred
locations of convection. Accordingly, the background climatology sets the
stage for the atmospheric response.
Overall, the standard deviations of Niño 3
and Niño 3.4 monthly anomalies
for 1950-79 are 0.79 and 0.77°C respectively and for 5-month running means
both of these drop to 0.71°C. If the entire period from 1950 to March
1997 is considered then the standard deviations increase by 0.05 to 0.07°C
The reasons for this increase can be seen from
Fig. 2(presented below).
In selecting a single threshold and duration to define ENSO events, it was
considered desirable to keep the criteria as simple as possible and for the
results to match conventional wisdom as to what have historically been
considered as events. In addition, both cold and warm phases of ENSO are
considered. Moreover, we use the same threshold for each although this is
not necessarily required. We adopt the JMA duration which requires a minimum
of 6 months for the 5 month running mean to exceed the threshold. We have
examined in detail three thresholds for the Niño 3 and 3.4 indices as
indicators of ENSO events. These are 0.3, 0.4 and 0.5°C which were chosen
as rounded numbers rather than fractional standard deviation values. As seen
in Table 1, the latter vary with the annual cycle. For the top threshold,
several single historic events are broken up into multiple events because the
index drops below the threshold for a month or two. For the bottom threshold,
the duration of events seems excessive. Overall the best match with
historical judgements was achieved for the Niño 3.4 index for a threshold
of 0.4°C. [While the 3.4 really refers to 40% of the way between the
Niño 3 and Niño 4 regions, this could give new meaning to the .4
adjunct.]
El Niño events identified using
Niño 3 and 0.5°C threshold include the
following years (relative to 1900): 51-52, 53, 57-58, 63-64, 65-66, 68-69-70,
72-73, 76-77, 79, 82-83, 86-87-88, 91-92; see Fig.1. Using this index and
threshold, there are exactly 6 months from July to December 1979 that exceed
the threshold and thus this
qualifies 1979 as an El Niño event whereas from
the JMA analyses it is not so qualified, and it is not normally so
identified. It is one ``event'' that could be eliminated if a more recent base
period were chosen. Also, in both 1993 and at the end of 1994, 5 consecutive
months exceed the threshold but not 6. However, examining -0.5°C as a
threshold for La Niña events does not work very well.
The 1950 La Niña
enters only at the 0.3°C level. The 54-56 event is broken into two pieces.
1964 qualifies, but so too does 67-68 which is not usually identified. The
events during 70-71, 73 and 75 qualify, as do 84-85, 88-89, and 95-96.
Most of these exceptions are remedied
if the 0.4°C threshold and Niño
3.4 is chosen (Fig. 1). The exact periods of events so defined are given in
Table 2.
Here we see the El Niños as 51-52, 53, 57-58, 63-64, 65-66,
68-69-70, 72-73, 76-77-78, 79-80, 82-83, 86-87-88, 91-92, 93, and 94-95. The
La Niñas are 50-51, 54-56, 64-65,
70-71-72, 73-74, 74-75-76, 84-85, 88-89,
95-96. 1979-80 still qualifies as an El Niño with a duration of 7 months,
although it would drop out at a threshold of 0.5°C where the duration is 4
months. But whereas 1953 qualifies here for 9 months, it would not qualify
at the higher threshold, where the longest duration exceeding 0.5°C is 3
months. April 1956 breaks the 1954-56 La Niña into two parts
and the fairly
strong La Niñas in 73-74 and 74-76 are not broken by a return to above
normal SSTs. Similarly, the 1976-77 El Niño event extends to January 1978
with a brief break from April to June 1977.
The perspective of whether the 1990 to 1995 period was one long event with
three more active phases or three events varies considerably depending upon
the index used. At the 0.3°C threshold, a prolonged El Niño begins in
March 1990 until April 1995 with brief breaks from August to December 1992
and December 1993 to April 1994 when the index remained positive. At the
0.4°C threshold this event began in March 1991 and the breaks are a bit
longer. From the perspective of the Niño 4 region farther west the SST
anomalies were more uniformly positive although with smaller amplitude and
the SOI remained of one sign (Trenberth and Hoar 1996a).
With these quantitative assignments of events, it is readily seen that there
are 177 out of 567 months (January 1950 to March 1997) assigned as El
Niño (= 31%) and 133 months assigned as La Niña months (= 23%).
Thus
55% of the time there is either an El Niño or La Niña
underway and only
45% of the time is one not present. Considering that there is a broad 3—6
year period spectral peak in ENSO indices, it is reasonable to consider
generically that the average time between events is 4 years or so and, because
each event typically lasts a year, therefore 50% of the time would be
assigned to either an El Niño or La Niña.
This can also be examined via
the histogram of Niño SST anomalies (Fig.2). The distribution of
monthly anomalies for January 1950 to March 1997 is shown along with those
from the post 1979 period, and the corresponding normal distribution. Here
the zero corresponds to the mean for the entire period, and a comparison of
the shaded post-1979 months shows the recent warm bias. Also readily apparent
here is the much greater spread in the more recent period with both tails of
the distribution being filled out.
The short record means that there is uncertainty in the histogram, as can
also be seen by comparing the 1950-79 period with the post 1979 period.
There is evidence for a flat distribution for Niño 3.4 anomalies between
about -0.6 and +0.4°C and perhaps even a bimodal character. This is not
so evident for Niño 3 where instead the overall distribution is strongly
skewed so that there is a deficiency of negative anomalies less than -1.6°C,
a surfeit of negative anomalies of -0.2 to -0.6°C and a surfeit of
positive anomalies above 1.2°C relative to the normal distribution. The
long tail on the right provides evidence supporting the attention given to
excessive warm conditions: the El Niño events. This is also why the list
of La Niña events is much
shorter than that for El Niños in Table 2.
Clearly the varying amplitude of events throughout the record, combined with
the fact that the peak values are not sustained for very long (compared with
a sine curve), means that in spite of the quasi-periodicity in ENSO, the
distribution is not that far from normal.
For Niño 3.4 it may be bimodal.
And for SSTs farther east it is skewed, presumably because of the shallowness
of the thermocline. Note that the normal distribution with the same standard
deviation as for the observed data would have 60.0% of the values exceeding
±0.4°C. This is more than the 55% assigned as ENSO events in part
because of those SST values that exceed the threshold but which do not endure
for longer than 6 months and are therefore not included.
One further point worthy of note in Table 2 is that the starting dates of
events are not uniformly distributed throughout the year. Most events begin
between March and September. The exceptions are the 1979 case, which is very
weak, and the 1993 El Niño which was a continuation of the previous warm
phase. A preferred ending time for events is February-March. Perhaps it is
not surprising that the transition times occur in the northern summer half
year while the peak amplitudes of the interannual variability (Table 1) are
in the northern winter. However, these observations show that there is a
distinct seasonality to the onset and occurrence of ENSO events which seems
to be difficult for models to get right.
It is clear from the standpoint of quantifying El Niño and related
phenomena, that the definition is still evolving, and in any case needs to
recognize the richness of the phenomenon. If a definition is needed, then
the one proposed by Glantz should be promulgated although this is not
quantitative. Precision can only be achieved if the particular definition is
identified in each use, and this is to to be recommended in all cases. For
more quantitative purposes, the JMA definition is suitable in most cases,
although it is suggested that it should be modified to apply to Niño 3.4
and with a threshold of 0.4°C. However, we have concluded that
it does not appear to be appropriate for any particular definition to be
officially recognized by CLIVAR.
Acknowledgements. I thank David Stepaniak for performing the
calculations and the figures. This research was sponsored by NOAA Office of
Global Programs.
Aceituno, P., 1992: El Niño, the Southern Oscillation, and ENSO:
Confusing names for a complex ocean-atmosphere interaction. Bull. Amer.
Meteor. Soc., 73, 483-485.
Glantz, M. H., 1996. Currents of change: El Niño's impact on climate and
society. Cambridge University Press. 194 pp.
Kiladis, G. N., and H. van Loon 1988: The Southern Oscillation. Pt VII:
Meteorological anomalies over the Indian and Pacific sectors associated with
the extremes of the oscillation. Mon. Wea. Rev., 116, 120-136.
Quinn, W. H., D. O. Zopf, K. S. Short, and R. T. W. Yang Kuo, 1978: Historical
trends and statistics of the
Southern Oscillation, El Niño, and Indonesian
droughts. Fish. Bull., 76, 663-678.
Ropelewski, C. F., and P. D. Jones, 1987: An extension of the Tahiti-Darwin
Southern Oscillation Index. Mon. Wea. Rev., 115, 2161-2165.
Scientific Committee on Oceanic research (SCOR) 1983: Prediction of El
Niño. Proceedings No. 19 Paris. Annex VI, SCOR WG 55 47-51.
Trenberth, K. E., 1984: Signal versus noise in the Southern
Oscillation. Mon. Wea. Rev., 112, 326-332.
Trenberth, K. E., and T. J. Hoar, 1996a: The 1990-1995 El Niño-Southern
Oscillation Event: Longest on record. Geophys. Res. Lttrs.,
23, 57-60.
Trenberth, K. E., and T. J. Hoar, 1996b: The 1990--1995 El Niño-Southern
Oscillation event: Longest on record. Proc. Symposium on Global
Ocean-Atmosphere-Land System (GOALS). Atlanta, 28 January-2 February 1996,
84-87.
Trenberth, K. E., G. W. Branstator, D. Karoly, A. Kumar, N-C. Lau, and C.
Ropelewski, 1997: Progress during TOGA in understanding and modeling
global teleconnections associated with tropical sea surface temperatures.
J. Geophys. Res. (special TOGA issue), (in press).
3. The recent ENSO record
Figure 1 shows the five month running mean
SST time series for the Niño 3
and 3.4 regions relative to a base period climatology of 1950-1979 given in
Table 1.
The base period can make a difference. This standard 30 year base
period is chosen as it is representative of the record this century, whereas
the period after 1979 has been biased warm and dominated by
El Niño events
(Trenberth and Hoar 1996a). Mean temperatures are higher in the Niño 3.4
region than in Niño 3 and its proximity to the Pacific warm pool and main
centers of convection is the reason for the physical importance of Niño
3.4.
4. Recommendation
It is hoped that the above provides a quantification of ENSO events in
several ways including when they have occurred, their duration and, from
Fig. 1, a measure of their amplitude all of which may be useful for some
purposes. A listing of the
duration of the El Niño and La Niña events
after 1950 is provided. Nevertheless these measures are not unique and
alternative criteria can be used. In particular, different criteria might be
used if the interest is just the coast of South America where the term El
Niño originated.
It is useful to know that ENSO events occur about 55% of
the time with the definition used above, so that most of the time the
tropics is regarded as being in one phase or the other and average conditions
are less common.
References
Back