1
Absence of 21st century warming on Antarctic Peninsula
1
consistent with natural variability
2
John Turner, Hua Lu, Ian White, John C. King, Tony Phillips, J. Scott Hosking,
3
Thomas J. Bracegirdle, Gareth J. Marshall, Robert Mulvaney and Pranab Deb
4
5
British Antarctic Survey, Natural Environment Research Council, High Cross, Madingley
6
Road, Cambridge, CB3 0ET, UK
7
Corresponding author: Prof. John Turner (jtu@bas.ac.uk)
8
9
Since the 1950s research stations on the Antarctic Peninsula have recorded some of the
10
largest increases in near-surface air temperature in the Southern Hemisphere
1
. This
11
warming has contributed to the regional retreat of glaciers
2
, disintegration of floating
12
ice shelves
3
and a ‘greening’ through the expansion in range of various flora
4
. Several
13
interlinked processes have been suggested as contributing to the warming, including
14
stratospheric ozone depletion
5
, local sea ice loss
6
, an increase in the westerly winds
5, 7
,
15
and changes in the strength and location of low-high latitude atmospheric
16
teleconnections
8, 9
. Here we use a stacked temperature record to show an absence of
17
regional warming since the late 1990s. The annual mean temperature has decreased at a
18
statistically significant rate, with the most rapid cooling during the Austral summer.
19
Temperatures have decreased as a consequence of a greater frequency of cold, east to
20
south-easterly winds resulting from more cyclonic conditions in the northern Weddell
21
Sea associated with a strengthening mid-latitude jet. These circulation changes have
22
also increased the advection of sea ice towards the east coast of the peninsula,
23
amplifying their effects. Our findings cover only 1% of the Antarctic continent and
24
2
emphasise that decadal temperature changes in this region are not primarily associated
25
with the drivers of global temperature change but, rather, reflect the extreme natural
26
internal variability of the regional atmospheric circulation.
27
While global mean surface air temperature (SAT) has increased over recent decades, the rate
28
of regional warming has varied markedly
10
, with some of the most rapid SAT increases
29
recorded in the polar regions
11-13
. In Antarctica, the largest SAT increases have been
30
observed in the Antarctic Peninsula (AP) and especially on its west coast
1
: in particular,
31
Vernadsky (formerly Faraday) station (Fig. 1) experienced an increase in annual mean SAT
32
of 2.8° C between 1951 and 2000.
33
The AP is a challenging area for the attribution of the causes of climate change
34
because of the shortness of the in-situ records, the large inter-annual circulation variability
14
35
and the sensitivity to local interactions between the atmosphere, ocean and ice. In addition,
36
the atmospheric circulation of the AP and South Pacific are quite different between summer
37
(December - February) and the remainder of the year.
38
Since the late 1970s the springtime loss of stratospheric ozone has contributed to the
39
warming of the AP, particularly during summer
7
. However, during the extended winter
40
period of March – September, when teleconnections between the tropics and high southern
41
latitudes are strongest
15
, tropical sea surface temperature (SST) anomalies in the Pacific and
42
Atlantic Oceans
16
can strongly modulate the climate of the AP. The teleconnections are
43
further affected by the mid-latitude jet, which influences regional cyclonic activity and AP
44
SATs. While the jet is strong for most of the year, during the summer it is weaker, there are
45
fewer cyclones, and tropical forcing plays little part in AP climate variability.
46
The annual mean SAT records from six coastal stations located in the northern AP
47
(Fig. 1) show a warming through the second half of the Twentieth Century, followed by little
48
change or a decrease during the first part of the Twenty First Century
17
. We investigate the
49
3
differences in high and low latitude forcing on the climate of the AP during what we
50
henceforth term the ‘warming’ and ‘cooling’ periods, focussing particularly on the period
51
since 1979, since this marks the start of the availability of reliable, gridded atmospheric
52
analyses and fields of sea ice concentration (SIC). We use a stacked and normalized SAT
53
anomaly record (Fig. 2a) based on the six station SAT time series (see Methods) in order to
54
investigate the broad-scale changes that have affected the northern AP since 1979. To provide
55
an objective measure of the timing of the change in trend we used the sequential Mann-
56
Kendall test (see Methods). This identified the middle of 1998 to early 1999 as the most likely
57
turning point between the warming and cooling periods (indicated by shading on Fig. 2). The
58
trends in the stacked SAT during the warming (0.32 ± 0.20 dec
-1
, 1979 - 1997) and cooling (-0.47
59
± 0.25 dec
-1
, 1999 - 2014) periods are both statistically significant at p < 0.05 (Extended Data
60
Table. 1). To confirm that the change in trend is not simply an artefact of the extreme El Niño
61
conditions during 1997 – 1998, we repeated the analysis for 1979 – 1996 and 2000 – 2014. The
62
trends were still significant at p < 0.05, although magnitudes were slightly smaller.
63
While the stacked SAT increased in all seasons during the warming period of 1979 –
64
1997 (Extended Data Table 1), the warming was largest during the summer, although the
65
significance of the trend is lower (p < 0.10). During this period there was a positive trend in
66
the Southern Annular Mode (SAM)
5
, primarily during summer (Extended Data Fig. 1) in
67
response to stratospheric ozone depletion and increasing greenhouse gas concentrations
5,18
.
68
The trend in the SAM led to a greater flow of mild, north-westerly air onto the AP (Extended
69
Data Fig. 2a), with SAT on the northeastern side increasing most because of amplification
70
through the foehn effect
7
. This atmospheric circulation trend contributed to the large decrease
71
in SIC in summer (Extended Data Fig. 3a) and for the year as a whole (Fig. 3a). However,
72
there was no significant trend in annual mean sea level pressure (SLP) across the AP during
73
4
the warming period (Fig. 3b). During the summer, tropical climate variability had little
74
influence on the AP SATs
15
and the trend in the SAM had the greatest impact.
75
Over the cooling period of 1999 – 2014 there was a significant increase in annual
76
mean SIC around the northern AP and across the northern part of the Weddell Sea (Fig. 3c).
77
This occurred as a result of increasingly cyclonic conditions in the Drake Passage and north-
78
western Weddell Sea (Fig. 3d), associated with a strengthening mid-latitude jet, which
79
advected cold air towards the AP.
80
The stacked SAT decreased in all seasons over 1999 - 2014, but with the greatest
81
cooling during summer, when the trend was moderately significant (p < 0.10) (Extended Data
82
Table 1). In this season the trends during the warming and cooling periods are different at a
83
90% confidence level (see Methods).
84
During the summer the SAM index remained predominantly positive (Extended Data
85
Fig. 1), and SLP was on average lower over the Antarctic than during the warming period
86
(Fig. 4a). However, there was no significant trend in the SAM, likely due to there being little
87
change in the depth of the ozone hole.
88
The summer cooling resulted from strong east to south-easterly near-surface flow
89
towards the AP as SLP decreased (increased) over the South Atlantic (Bellingshausen Sea)
90
(Extended Data Fig. 4a). The greater cyclonic conditions over the South Atlantic occurred in
91
association with a mid-latitude jet that was significantly (p < 0.05) stronger than during the
92
warming period (Fig. 4b), and which was located at the northern limit of a cold trough
93
extending from Antarctica. The stronger east to south-easterly flow advected sea ice towards
94
the east coast of the AP, giving a positive SIC trend that extended across the whole of the
95
northern Weddell Sea (Extended Data Fig. 5a). This greater ice extent limited the flux of heat
96
from the ocean and amplified the effects of the circulation changes.
97
5
During the extended winter, the circulation over the South Pacific is marked by a
98
clearly defined climatological split jet structure, with the Sub-tropical Jet (STJ) near 30° S
99
and the Polar Front Jet (PFJ) close to 60° S
19
. The PFJ is sensitive to both zonally symmetric
100
forcing, such as SAM variability, and regional factors, such as the stationary Rossby waves
101
propagating from the tropical Pacific Ocean
15
, and the meridional gradient of extratropical
102
SSTs. In general, the STJ (PFJ) is stronger during El Niño/positive Interdecadal Pacific
103
Oscillation (IPO) (La Niña/negative IPO)
20,21
. Nevertheless, their combined effect on the AP
104
regional circulation is complex and may involve nonlinear wave - mean flow feedbacks
22
.
105
The tropical – high latitude linkages during the extended winter were examined
106
through an analysis of large-scale Rossby wave propagation via the horizontal stationary
107
Eliassen-Palm (EP) fluxes
23
at 300 hPa (see Methods section). During the 1979 - 1997
108
warming period, tropical SSTs were characterized by relatively high SSTs across the eastern
109
tropical Pacific Ocean with relatively more frequent El Niño conditions and a more positive
110
IPO (Fig. 2b), and thus a relatively weak (strong) PFJ (STJ). The generation of quasi-
111
stationary Rossby waves occurred primarily close to 180° E, consistent with higher SSTs in
112
this area, and the wave propagation from the tropics to the South Pacific was limited by a
113
strengthened STJ.
114
During the 1999 - 2014 cooling period, a major change in tropical Pacific SSTs
115
occurred with SSTs higher over the Maritime Continent and lower over the eastern Pacific
116
Ocean (Fig. 3e). There was also enhanced transmission of quasi-stationary Rossby waves
117
from the tropics towards the Antarctic, which can be seen in the differences in 300 hPa
118
streamfunction (Fig. 4f). However, at higher latitudes wave propagation was
119
reduced/prohibited because of enhanced equatorward wave refraction/reflection from the PFJ
120
region (Fig. 4f). The PFJ was significantly stronger across the South Pacific during the
121
cooling period (Fig. 4e), which is consistent with the higher frequency of La Niña-like
122