Science - USA (2020-09-25)

Atlantic 2017 MHW, which caused fish mass
mortalities ( 32 ); and the Southern Ocean
2016 MHW, which has been linked to strong
decreases in sea ice ( 8 ).
With the exception of the Southwest Atlan-
tic 2017 MHW, all analyzed MHWs have been
the longest and most intense MHWs in their
respective regions within satellite tempera-
ture recording (fig. S1). This raises the ques-
tion of whether their occurrence has been
influenced by anthropogenic climate change.
By calculating the FAR, we determine to what
extent the occurrence probability of heatwaves
equaling or exceeding the observed dura-
tion, intensity, and cumulative intensity has
already been affected by anthropogenic cli-
mate change.
The FAR of MHW intensity is generally
very high, with point estimates between 0.97
and 1.0 for all but one of the MHWs that could
be attributed and with narrow confidence
intervals for most of the heatwaves (Table 1).
These high FAR values indicate a more than
20-fold increase in occurrence probability
of similarly warm MHWs caused by anthro-
pogenic climate change. Additionally, accord-
ing to our results, it is very unlikely that the
MHWs could have reached the high temper-
atures that were measured without the influ-
ence of climate change. The only exceptions
are the Western Australian 2011 MHW, which
we could not attribute (materials and meth-
ods), and the Southern Ocean 2016, which has
a FAR close to zero, indicating little change
in occurrence probability as a result of cli-
mate change.
FAR values of heatwave duration are like-
wise very high. The point estimates are above
0.79 for five of the six heatwaves that could be
attributed (Table 1). Exceptions are the Indo-
Australian 2016 MHW, for which we could not

attribute the duration (materials and meth-

ods), and the Southern Ocean 2016 MHW,

where we found a decreased occurrence prob-

ability in the present-day climate. In the South-

ern Ocean, several models simulate only a

modest warming or even decreases in SST

over the historical period. Additionally, many

models show a reduction of interannual var-

iability there (figs. S2 to S8). The lack of sur-

face ocean warming in the Southern Ocean

reflects large-scale, interhemispheric asymme-

tries in the mean ocean circulation. The South-

ern Ocean circulation is dominated by strong

upwelling, which nearly anchors SST at pre-

industrial levels under transient global warm-

ing ( 33 ). However, simulating SST in the

Southern Ocean is challenging for models,

as the processes leading to the formation

of sea ice and the formation of deep-water

masses are currently not well represented,

which decreases model fidelity in this highly

dynamic region ( 34 , 35 ).

FAR values of cumulative intensity are >0.96

for three of the four heatwaves that could be

attributed (Table 1). Again, the only exception

is the Southern Ocean, where we find a nega-

tive FAR value—that is, a lower occurrence

probability in the present-day climate. Overall,

our results clearly demonstrate that anthro-

pogenic climate change has already had a sub-

stantial impact on the probability of individual

MHW occurrences in terms of intensity, dura-

tion, and cumulative intensity.

To illustrate the effect of future climate

change on the occurrence of the seven large

MHWs, we calculated return periods for dif-

ferent levels of global warming (Fig. 2). For

MHWs equaling or exceeding the duration,

intensity, or cumulative intensity of the ob-

served ones, return periods sharply decrease

under global warming. MHWs were extremely

rare events during preindustrial times, with

expected return periods of hundreds to thou-

sands of years for North Atlantic MHWs and

>10,000 years for all other MHWs in terms

of their intensity, duration, and cumulative

intensity.

Under the condition of 1.5°C of global warm-

ing, the return periods of all MHWs are re-

duced to ten to hundreds of years, with the

exception of heatwaves that are as intense as

the Western Australian MHW and the South-

ern Ocean MHW, which have slightly higher

return periods. Under the condition of 3°C of

warming, the return periods of all MHWs are

only 1 to 10 years—again with the exception of

the intensity of the Western Australian MHW

and the Southern Ocean MHW, which have

slightly higher return periods.

The intensity of the Southern Ocean MHW

has a longer return period at 0.5°C of warm-

ing compared with preindustrial conditions.

This is explained by high preindustrial tem-

perature variability in many models and a

modest warming in the present day (fig. S7).

At higher levels of warming, the return periods

shorten in the Southern Ocean as well, reach-

ing return periods of <10 years for both heat-

wave duration and cumulative intensity at 3°C

of warming. In a 3°C-warmer world, the re-

gions in which the Northeast Pacific, the South-

west Atlantic, and the Indo-Australian Basin

MHWs occurred all reach SSTs above the pre-

industrial 99th temperature percentile for >97%

of the data, which means that they are essen-

tially in a continuous, extreme heatwave state.

Limiting global warming to 1.5° or 2°C will

temporarily relieve these regions from heat

pressure, such that expected return periods

between 5 and 20 years (for heatwave dura-

tion, intensity, and cumulative intensity) might

leave some limited time for recovery.

1624 25 SEPTEMBER 2020•VOL 369 ISSUE 6511 sciencemag.org SCIENCE

Table 1. Attribution of duration, intensity, and cumulative intensity

of seven MHWs to anthropogenic climate change.In contrast to the

three-dimensional heatwave definition used in Fig. 1, the heatwaves

are defined here on one-dimensional time series, which represent regional

averages across the areas where the respective heatwaves occurred

(materials and methods). Duration denotes the number of consecutive

days above the 99th percentile, heatwave intensity is the average

temperature anomaly above the baseline climatology in degrees Celsius,

and cumulative intensity is the sum of all temperature anomalies over the

duration of the heatwave. We present the point estimate for the attributable

risk (FAR) and the 95% confidence interval in brackets. Heatwaves that

we cannot attribute are denoted with a dash (materials and methods).

Heatwave

number

Time and

location

Intensity

(°C)

FAR

intensity

Duration

(days)

FAR

duration

Cumulative intensity

(°C days 1000 km^2 )

FAR cumulative

intensity

...................................................................................................................................^1 Western Australian 2011 2.26 ......................................................................................................................................–^101 0.79 [−0.55, 0.97]^9 .......................................................–

...................................................................................................................................^2 Northwest Atlantic 2012 2.15 0.97 [0.92, 0.99]......................................................................................................................................^57 0.96 [0.94, 0.97]^41 .......................................................0.96 [0.94, 0.98]

...................................................................................................................................^3 Northeast Pacific 2013 to 2015 1.56 1.0 [0.97, 1.0]......................................................................................................................................^357 1.0 [0.99, 10]^530 .......................................................1.0 [0.99, 1.0]

...................................................................................................................................^4 Tasman Sea 2015 and 2016 1.49 0.98 [0.92, 0.99]......................................................................................................................................^175 1.0 [0.49, 1.0]^37 .......................................................–

...................................................................................................................................^5 Indo-Australian Basin 2016 1.67 1.0 [0.77, 1.0]......................................................................................................................................^90 –^13 .......................................................–

...................................................................................................................................^6 Southern Ocean 2016* 1.0 0.03 [......................................................................................................................................−2.71, 0.74]^183 −0.6 [−2.6, 0.26]^47 .......................................................−6.5 [−42.26,−0.34]

...................................................................................................................................^7 Southwest Atlantic 2017 1.96 1.0 [0.74, 1.0]......................................................................................................................................^82 1.0 [0.91, 1.0]^11 .......................................................1.0 [0.87, 1.0]

*Represents two spatially distinct heatwaves that occurred simultaneously in different parts of the Southern Ocean.

RESEARCH | REPORTS