Reconciling warming trends

Abstract

Climate models projected stronger warming over the past 15 years than has been seen in observations. Conspiring factors of errors in volcanic and solar inputs, representations of aerosols, and El Niño evolution, may explain most of the discrepancy.

Full text

158 NATURE GEOSCIENCE | VOL 7 | MARCH 2014 | www.nature.com/naturegeoscience
commentary
Reconciling warming trends
Gavin A. Schmidt, Drew T. Shindell and Kostas Tsigaridis
Climate models projected stronger warming over the past 15years than has been seen in observations.
Conspiring factors of errors in volcanic and solar inputs, representations of aerosols, and ElNiño
evolution, may explain most of the discrepancy.
Any divergence between real-world
climate phenomena and prior
expectations poses interesting
science questions. e case of the
apparent slow-down of warming since
the record El Niño event in 1997/1998is
no exception. e global mean surface
temperature trend was smaller1 between
1997and 2013 (0.07±0.08°C per decade)
than over the last 50years (0.16±0.02°C
per decade), highlighting questions about
the mechanisms that regulate decadal
variability in the Earth’s temperature. In
addition, the warming trend in the most
recent 15-year period is near the lower
edge of the 5–95% range of projections
from a collection of climate models
that were part of the Coupled Model
Intercomparison Project Phase 5 (CMIP5).
Why most of the model simulations
suggest more warming than has been
observed is a second question that deserves
further exploration.
Short-term uctuations in global mean
surface temperature anomalies have been
a perennial focus of public discussions
related to climate change. We should
expect to see climate changes, by denition,
only in the long-term trends that average
over stochastic weather and year-to-year
uctuations such as those associated with
the ElNiño/Southern Oscillation (ENSO),
which favoured a cool LaNiña phase in
the past few years. On decadal timescales,
global mean surface temperatures are
expected to vary, too. One cause might
be the chaotic internal variability of the
coupled system of oceans and atmosphere,
for example in the tropical Pacic Ocean2,
or in variability in deep ocean circulation3.
Alternatively, decadal-scale temperature
variations can be a response of the climate
system to external inuences, such as
volcanic eruptions or the solar cycle. Both
kinds of inuence will continue to operate
and be superimposed on any long-term
warming trend from rising atmospheric
greenhouse gas concentrations. For shorter
time periods — and a 15-year interval
may just about qualify as such — natural
and externally driven uctuations can,
in principle, lead to a change in warming
trend. But it is not obvious that either
eect should lead to a discrepancy
between climate model projections
and observations.
Here we argue that a combination
of factors, by coincidence, conspired
to dampen warming trends in the real
world aer about 1992. CMIP5 model
simulations were based on historical
estimates of external inuences on the
climate only to 2000or 2005, and used
scenarios (Representative Concentration
Pathways, or RCPs) thereaer4. Any
recent improvements in these estimates
or updates to the present day were not
Well-mixed greenhouse gases. We
linearly interpolated observations of
atmospheric concentrations of CO2, N2O,
CH4 and chlorouorocarbons (CFCs)
between 2005and 2012, and calculated
the radiative eect of these greenhouse gas
levels using standard simplied formulae.
Concentrations of CO2, N2O and CH4
were very slightly lower than used in the
CMIP5 runs, whereas CFCs were a little
higher, leading to a small cool bias in the
models (Fig.1a).
Solar irradiance. Because solar irradiance
at the last solar-cycle minimum was
lower and the present weak solar cycle
started later12 than had been assumed in
the CMIP5 simulations13, solar drivers
were too positive in the CMIP5 runs. We
multiplied the dierence in total solar
irradiance forcing by an estimated factor
of 2, based on a preliminary analysis
of solar-only transient simulations, to
account for the increased response over
a basic energy balance calculation when
whole-atmosphere chemistry mechanisms
are included.
Human-made tropospheric aerosols. Only
two CMIP5 models include nitrate aerosol
forcing, and half did not include aerosol
indirect eects. Given the recent increases
in Asian sources, nitrate eects amount
to perhaps −0.04Wm−2 of additional
forcing14. Exclusion of aerosol indirect
eects15 might add another −0.06 Wm−2
globally since 2000. We assumed that this
forcing increased slowly in the 1990s, and
faster in the 2000s following Asian NO2
emissions16, peaking at −0.1Wm−2 in
2010. Underestimates of secondary organic
aerosols are an additional missing factor,
but are harder to quantify17.
Volcanic stratospheric aerosols. A
reanalysis of satellite retrievals in the
1990s and additional observations in the
2000s has produced a record of aerosol
optical depth9 (updated to version 2.0;
ref.10) that diers signicantly from that
used in CMIP5 (ref.18). In particular,
the magnitude of the impact of Pinatubo
in the early 1990s is downgraded, and
larger contributions are suggested for
the twenty-rst century. Radiative
forcing is calculated by assuming no
post-2000 volcanic aerosols in CMIP5
and multiplying the dierence in aerosol
optical depth by −25 to derive a forcing
in Wm–2.
Box 1 | Adjustments to the CMIP5 external influences
A combination of factors, by
coincidence, conspired to
dampen warming trends in the
real world after about 1992.
© 2014 Macmillan Publishers Limited. All rights reserved

NATURE GEOSCIENCE | VOL 7 | MARCH 2014 | www.nature.com/naturegeoscience 159
commentary
taken into account in these simulations.
Specically, the inuence of volcanic
eruptions, aerosols in the atmosphere
and solar activity all took unexpected
turns over the 2000s. e climate model
simulations, eectively, were run with the
assumption that conditions were broadly
going to continue along established
trajectories. Additionally, the exact phasing
of any natural uctuations will not be
captured in any specic model.
Updated estimates of climate drivers
To test this hypothesis, we have collated
up-to-date information on volcanic aerosol
concentrations, solar activity and well-
mixed greenhouse gases in the 1990s and
2000s (Box1). ese updates include both
newly observed data and also reanalyses
of earlier 1990s data on volcanic aerosols
based on improved satellite retrievals.
Comparing the updated information
with the data used in the CMIP5 climate
model simulations (Fig.1a) suggests that
in the mid-1990s, cooling factors were
overestimated (largely an eect of an
overestimate of the stratospheric aerosol
optical depth in the wake of the Mount
Pinatubo eruption), making the models
cooler than the real world. Conversely, in
the 2000s, multiple cooling factors were
underestimated, so the real world would
be expected to have been cooler than in
the models.
We estimate how simulated global
mean surface temperature would have
been dierent in the CMIP5 runs if two
eects had been included: rst, if ENSO
in each model had been in phase with
observations, and second, if the ensemble
mean were adjusted using results from a
simple impulse-response model5 with our
updated information on external drivers
(Fig.1b). We nd that this procedure
reduces the dierences between observed
and simulated temperatures by a factor
of 3. We conclude that the coincident
underestimates of cooling factors in the
2000s — that is, of volcanic aerosols,
solar irradiance and eects of human-
made aerosols — have combined to bias
the model ensemble. According to this
estimate, about a third of the dierence
between simulated and observed trends
(in the GISTEMP analysis6) between
1997to 2013is a result of underestimated
volcanic emissions; about one-seventh of
the discrepancy comes from overestimates
of solar activity and dierences in ENSO
phasing between climate models and
the real world; and just under a quarter
could be related to human-made aerosols.
ese estimates leave only a small residual
between models and observations that
can be easily accounted for by internal
variability unrelated to ENSO, or to any
further misspecications of external
inuences. In comparison to an alternative
temperature analysis7 the observed
trend matches the adjusted simulated
temperature increase even more closely
(Fig.1b).
Prediction
uncertainty
GISTEMP LOTI
HadCRUT4 (infilled)
CMIP5 ensemble (5–95%)
CMIP5 with obs. ENSO
and with updated forcings
Adjusted CMIP5 ensemble
Year
Year
Temperature anomaly (°C)
Global mean temperature anomalies
1992 1996 2000 2004 2008 2012
1992 1996 2000 2004 2008 2012
–0.4
–0.2
0.0
0.2
0.4
0.6
0.8
–0.4
–0.3
–0.2
–0.1
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
Radiative forcing (W m–2)
WMGHG
Solar
Aerosols
Volcanic
Updates to climate model drivers
a
b
Figure 1 | Updated external influences on climate and their impact on the CMIP5 model runs. a, The
latest reconstructions of optical depth for volcanic aerosols9,10 from the Mount Pinatubo eruption in
1991 suggest that the cooling eect of the eruption (1991–1993) was overestimated in the CMIP5 runs,
making the simulated temperatures too cool. From about 1998 onwards, however, the cooling eects
of solar activity (red), human-made tropospheric aerosols (green) and volcanic eruptions (pink) were
all underestimated. WMGHG, well-mixed greenhouse gases. b, Global mean surface temperature
anomalies, with respect to 1980–1999, in the CMIP5 ensemble (mean: solid blue line; pale blue
shading: 5–95% spread of simulations) on average exceeded two independent reconstructions from
observations (GISTEMP Land–Ocean Temperature Index (LOTI)6, solid red; HadCRUT4 with spatial
infilling7, dashed red) from about 1998. Adjusting for the phase of ENSO by regressing the observed
temperature against the ENSO index11 adds interannual variability to the CMIP5 ensemble mean
(dashed blue), and adjusting for updated external influences as in a further reduces the discrepancy
between model and data from 1998 (black). The adjusted ensemble spread (dashed grey) clearly
shows the decadal impact of the updated drivers. As an aside, we note that although it is convenient
to use the CMIP5 ensemble to assess expected spreads in possible trends, the ensemble is not a true
probabilistic sample.
© 2014 Macmillan Publishers Limited. All rights reserved

160 NATURE GEOSCIENCE | VOL 7 | MARCH 2014 | www.nature.com/naturegeoscience
commentary
Reconciliation
We conclude that use of the latest
information on external inuences on the
climate system and adjusting for internal
variability associated with ENSO can almost
completely reconcile the trends in global
mean surface temperature in CMIP5 models
and observations. Nevertheless, attributing
climate trends over relatively short periods,
such as 10 to 15years, will always be
problematic, and it is inherently unsatisfying
to nd model–data agreement only with the
benet of hindsight. We see no indication,
however, that transient climate response is
systematically overestimated in the CMIP5
climate models as has been speculated8, or
that decadal variability across the ensemble
of models is systematically underestimated,
although at least some individual models
probably fall short in this respect.
Most importantly, our analysis implies
that signicant warming trends are likely to
resume, because the dominant long-term
warming eect of well-mixed greenhouse
gases continues to rise. Asian pollution levels
are likely to stabilize and perhaps decrease,
although lower solar activity may persist and
volcanic eruptions are unpredictable. ENSO
will eventually move back into a positive
phase and the simultaneous coincidence of
multiple cooling eects will cease. Further
warming is very likely to be the result. ❐
Gavin A.Schmidt, Drew T.Shindell and
Kostas Tsigardidis are at the NASA Goddard Institute
for Space Studies, 2880 Broadway, New York, New York
10025, USA. Kostas Tsigaridis is also at the Center for
Climate Systems Research, Columbia University,
2880 Broadway, New York, New York 10025, USA.
e-mail: gavin.a.schmidt@nasa.gov
References
1. Fyfe, J.C., Gillett, N.P. & Zwiers, F.W. Nature Clim. Change
3, 767–769 (2013).
2. Kosaka, Y. & Xie, S-P. Nature 501, 403–407 (2013).
3. Trenberth, K.E. & Fasullo, J.T. Earth’s Future 1, 19–32 (2013).
4. Taylor, K.E., Stouer, R. & Meehl, G. Bull. Am. Meteorol. Soc.
93, 485–498 (2012).
5. Boucher, O. & Reddy, M.S. Energy Policy 36, 193–200 (2008).
6. Hansen, J., Ruedy, R., Sato, M. & Lo, K. Rev. Geophys.
48, RG4004 (2010).
7. Cowtan, K. & Way, R.G. Q.J.R.Meteorol. Soc.
http://dx.doi.org/10.1002/qj.2297 (2013).
8. Otto, A. et al. Nature Geosci. 6, 415–416 (2013).
9. Vernier, J-P. et al. Geophys. Res. Lett. 38, L12807 (2011).
10. Santer, B.D. et al. Nature Geosci. 7, 185–189 (2014).
11. Wolter, K. & Timlin, M.S. Weat he r 53, 315–324 (1998).
12. Fröhlich, C. & Lean, J. Geophys. Res. Lett.
25, 4377–4380 (1998).
13. Lean, J. Calculations of Solar Irradiance (2009);
http://go.nature.com/VbldYY
14. Shindell, D.T. et al. Atmos. Chem. Phys. 13, 2939–2974 (2013).
15. Bellouin, N. et al. J.Geophys. Res. 116, D20206 (2011).
16. Ma, J.Z. et al. Atmos. Chem. Phys. 13, 1547–1567 (2013).
17. Hoyle, C.R. et al. Atmos. Chem. Phys. 11, 321–343 (2011).
18. Sato, M., Hansen, J.E., McCormick, M.P. & Pollack, J.B.
J.Geophys. Res. 98, 22987–22994 (1993).
Bumpy path to a warmer world
Martin Visbeck
Decadal climate variability has long received limited attention. With the slow-down in surface warming
since the late 1990s, the decadal scale has rightly become a focus of attention: for assessing climate
change and its impacts, it is of critical importance.
A
changing climate matters most to
our generation on timescales from
seasons to decades. Decadal climate
information, in particular, is essential
for societies in developing to become
climate-resilient.
e developed world has established
a wealth of adaptation mechanisms to
cope with the eects of a changing climate
on seasonal and interannual timescales.
Advanced water management systems,
capital reserves and widespread insurance
all help to buer the loss, damage and gains
from short-term climate uctuations. By
contrast, a 10-year drought, or a decade
of high tropical storm activity, may easily
exhaust the coping strategies of even
the most developed economies. In the
developing world, coping mechanisms
are less abundant, and the same change
in the climate oen leads to great
societal hardship.
Beyond the decadal timescale, on the
other hand, the world is developing long-
term adaptation strategies to cope with
persistent climate change. ese include the
development of large-scale infrastructure
such as seawalls and water reservoirs,
alternative responses such as retreat
from aected land, and adjustments in
economic and legislative structures such as
insurance frameworks.
Climate variability on scales of 10 to
30years — too long to sit out, yet too short
to adapt to fully — is therefore particularly
important to our generation. Nevertheless,
the slow-down in warming over the
past 15years or so — despite unabated
emissions of greenhouse gases and global
climate model projections that suggest
much steeper warming — has caught parts
of the scientic community by surprise. We
should have expected such decadal swings
if more emphasis had been devoted to
understanding the decadal scale.
Weak and regional
e most energetic frequency range in
almost any long-term climate record is the
annual cycle. In the tropics, the annual
band is followed by climate variability on
timescales of 3 to 7years associated with
the El Niño/Southern Oscillation (ENSO),
a large-scale climate see-saw that is rooted
Figure 1 | Ocean observations. Martin Visbeck
deploying an Argo float in the ocean from the
research vessel MV Maria S.Merian. Argo is an
international network of more than 3,000 robotic
floats that measure the temperature and salinity of
the ocean’s upper 2,000m.
© MARTIN VISBECK GEOMAR
© 2014 Macmillan Publishers Limited. All rights reserved