text01

advertisement
Supplementary Information
“Attributing the increase in Northern Hemisphere hot summers since the late 20th
century”, by Youichi Kamae, Hideo Shiogama, Masahiro Watanabe, and Masahide
Kimoto
Physical mechanism of direct contribution of GHG radiative forcing over
extratropical land areas
The results presented in this study indicate that the direct radiative effect due to
increasing GHG are essential for the JJA-mean land surface warming since the late 20th
century. This conclusion seems to be inconsistent with earlier studies [Joshi et al., 2008;
Compo and Sardeshmukch, 2009; Dommenget, 2009] revealing that SST increase is the
dominant factor for the observed annual-mean land surface warming during recent
decades. Kamae et al. [2014; hereafter K14] demonstrated a possible factor for the
discrepancy. Fig. 1a, e in K14 show free tropospheric warming (at 700 hPa level) and
ΔSAT during JJA in 1% per year CO2 increase experiment (1%CO2) conducted by
coupled general circulation models (CGCMs) [Taylor et al., 2012]. Land–sea warming
contrast appears in ΔSAT in the warming climate [Manabe et al., 1991; Kamae and
Watanabe, 2013]. Joshi et al. [2008, 2013] represented that a uniform free-tropospheric
warming over land and ocean results in a land–sea ΔSAT contrast through boundary
layer amplification mechanisms (Fig. 1 in Joshi et al., 2013). These mechanisms are
supported by results of AMIP-type SST increase experiment (patterned SST +4 K
experiment, hereafter amipFuture) shown in Fig. 1d, h in K14. However, in JJA-mean,
the anomalous free-tropospheric and surface warming over NH middle and high latitude
land areas relative to the ocean are not simulated in amipFuture (Fig. 1d, h in K14).
1
Major difference of the prescribed forcing between 1%CO2 and amipFuture runs is the
increased or fixed CO2 concentration (and sea ice cover; see discussion below). Fig. 1b,
f in K14 shows land surface and free-tropospheric warming due to CO2 quadrupling
simulated in AMIP-type CO2 increase experiment (amip4xCO2) [Kamae et al., 2013].
The results show that the SST increase (CO2 increase) is important for anomalous
land-surface warming over low latitude (middle and high latitude). The latitudinal
dependence of the temperature response to the SST increase can be explained by
difference in sizes of Rossby radiuses of deformation (K14). The direct radiative heating
due to GHG forcing is essential for the land warming over middle and high latitude
projected in the global warming simulations. More details were described in K14.
Prescribed SST and aerosol forcing in model simulations
Figure S2 shows time series of global annual-mean SST [Rayner et al., 2003]
prescribed in the model simulations. The observational SST prescribed in the ALL and
SST runs shows an increasing trend (0.39 K over 63 years) but the SST used in the NAT
run shows a small negative (not statistically significant) trend (–0.02 K over 63 years).
The recent decrease in the SST prescribed in the NAT run indicates that the recent
increases of TNAT and fNAT (Figs. 4c, f, Fig. S5c and h) are not ascribed to the change in
the global-mean but the spatial pattern of the prescribed SST (Fig. 4j).
Radiative forcing associated with aerosols in the model simulations can influence
regional responses of the land SAT in the long-term and short-term changes. Figure S3
shows the net aerosol radiative forcing at tropopause in the long-term mean, long-term
and short-term changes. Climatology of net aerosol radiative forcing (Fig. S3a) is
2
characterized by 1) positive forcing over semi-arid regions (subtropical North Africa,
Arabian Peninsula, and central and South Asia) and 2) negative forcing over the
remaining regions particularly over the land. Anthropogenic influences intensify the
negative forcing over the land (Fig. S3a, e).
The long-term change in the ALL (NAT) run shows a patterned (no) change (Fig.
S3b, f). The change of anthropogenic aerosol forcing act as heating over subtropical
North Africa, Europe, and West Asia and cooling over Canada, Southeast Asia and East
Asia (Fig. S3b). This pattern resembles some parts of the regional characteristics of the
simulated ΔSAT: the larger warming over Europe and a smaller warming over the East
Eurasia (Fig. 3b). This similarity suggests that the anthropogenic aerosol forcing may
influence the spatial pattern of the simulated land warming in the ALL run. The aerosol
forcings also act as a warming factor for the short-term change (Fig. S3c, g).
However, the aerosol forcing alone cannot explain the continuous land surface
warming due to ADIR effect (Figs. 2 and 3d), indicating that the forcings including CO2
play essential roles on TADIR and PADIR in long-term and short-term changes. Note that
the warming effect of natural aerosol forcing in the short-term change (Fig. S3c, d, g, h)
contributes partly to the increases of TNAT and PNAT (Fig. 4c, f), but the peak of the
middle latitude warming (45°N–50°N including North America) cannot be explained by
the aerosol forcing (Figs. 2 and 4).
Dependence of results on selection of observational data
We use the Goddard Institute for Space Studies (GISS) SAT analysis [Hansen et
al., 2010] in this study. Land SAT is also provided as other datasets. For example,
3
CRUTEM4 [Jones et al., 2012] provides global land SAT time series in 5° × 5° grids
since 1850. We confirmed a dependence of the results on a selection of the
observational data. Figure S1a shows the frequency of the hot and very hot summers
derived from the two datasets. The black line shows the result from GISS shown in Fig.
1b. The red line is CRUTEM4, representing a high consistency with GISS. CRUTEM4
shows a smaller increase in recent years than GISS data. Figure S1b–e shows spatial
patterns of hot and very hot summer frequencies during 1991–2011. The maps of
CRUTEM4 show similar results to those of GISS. CRUTEM4 has wider missing area
than GISS because the spatial resolution of the former is larger than that of the latter.
The wider missing area may influence on the smaller increase of frequency of hot
summers. We used GISS temperature data in this study because of the wider data
coverage and the higher spatial resolution of data than CRUTEM4.
Sensitivity of results on data coverage
Hansen et al. [2013] reported that time-series of area-averaging SAT and
frequency of hot summers can be influenced by time-varying data coverage. We used
SAT data over areas where most data are available throughout 1949–2011. Map in Fig.
S4 shows the spatial distribution of the constant masking derived from the observation.
This constant masking was also applied to all the model simulations shown in Figs. 1
and 2. We also examined a sensitivity of the modeled time series to the data coverage.
Figure S4 shows the modeled anomalies of JJA-mean SAT and its dependence on the
data coverage. A time-series of SAT anomaly averaged over the area with
observation-based masking (shown in Fig. 1a) shows no large difference with a
4
time-series averaged over the whole NH land area. A long-term trend of SAT is smaller
than that from the no-masking SAT because of large warming over the masked areas
(e.g. subtropical Africa, tropical South America, and Greenland, Fig. 3b).
Observed and modeled changes in probability density function of SAT
Figure S5 shows the frequency of occurrence of land surface air temperature
(SAT) over the Northern Hemisphere (NH) divided by local standard deviation in the
observation (similar to Fig. 4 in Hansen et al., 2012) and the model simulations. The
probability density function (PDF) of observed SAT (Fig. S5a) shows 1) a warm-ward
shift and 2) a gradual increase in variability as time passed during recent decades
[Hansen et al., 2012]. The probability beyond two standard deviations in the
observation shows a +16% increase in 2001–2011 relative to 1951–1980 (Fig. S5f). The
change of probability can be decomposed into two components: PDF shift effect (a
component reproduced by mean-shift of PDF) and residual (change in shape of PDF;
PDF broadening effect hereafter). The total change is dominated by the PDF shift effect
(12%) but the PDF broadening effect is also substantial (4%). The observed PDF is
simulated accurately in the ALL run (Fig. S5b). Modeled change in probability (Fig.
S5e) and contributions of the two effects are also similar to those of the observation (Fig.
S5g). The NAT run shows small increase in the probability (Fig. S5c, h). Effects of PDF
shift and PDF broadening simulated in the SST run are underestimated compared with
the observation (Fig. S5d, i).
Note that the ALL run underestimates the increase in probability of warm
extremes (larger than three standard deviations, Fig. S5e). Limited spatial resolution,
5
ensemble size, possible biases in land-surface processes, prescribed forcings (see
discussion below) and responses can contribute to this discrepancy.
Model dependences of the results
The results presented in this study were derived from the AGCM simulations
using MIROC5 [Watanabe et al., 2010]. Model sensitivities to GHG radiative forcing,
SST variability, and the natural climate forcings could depend on models. A
multi-model ensemble covering a wide range of model sensitivities could better
evaluate the relative contributions of the three components to the observed variability of
extreme weather events. The model sensitivity to GHG forcing particularly could play a
major role on the results presented in this study. Kamae and Watanabe [2012] examined
the model sensitivities to quadrupling CO2 forcing with prescribed-SST among CMIP5
models [Taylor et al., 2012], representing that MIROC5 shows a similar sensitivity to
the CO2 increase (atmospheric circulation, cloud amount, and hydrological cycle) to the
CMIP5 model average. We also confirmed that land SAT response to CO2 increase
under the fixed-SST condition during JJA shows a similar pattern and magnitude to the
CMIP5 model average (Fig. S7). In addition, the responses in the atmospheric
circulations and land SAT to the recent internal SST variability simulated in MIROC5
(Fig. 4h) are also consistent with those simulated in other AGCMs [Pegion and Kumar,
2010; Kobayashi, 2014; Schubert et al., 2014]. Note that model sensitivities to the other
forcings, e.g. anthropogenic and natural aerosols, volcanic eruptions, and historical land
use change (see discussion below) may also play substantial roles on the results.
6
Decomposing contributions using a set of AMIP-type sensitivity experiments
In this study, the direct (ADIR) and indirect (ASST) effects of anthropogenic
forcing and the effect of natural forcing and SST variability (NAT) on the land surface
warming and variations in occurrence frequency of hot summers are decomposed using
a set of AMIP-type sensitivity experiments. Similar experimental frameworks were used
to attribute the increasing land SAT [Folland et al., 1998; Bracco et al., 2004; Compo
and Sardeshmukh, 2009] and investigate mechanisms of atmospheric circulation
changes [Bracco et al., 2004; Deser et al., 2009] during the recent decades. In this
framework, the atmospheric variability is considered as a superposition of an internal
variability (intrinsic to the dynamical atmosphere) and a forced response to SST
variation [Bracco et al., 2004]. NAT effect defined in this study includes atmospheric
and land surface responses to 1) natural forcing and 2) internal SST variability that is
independent of the anthropogenic forcing.
Folland et al. [1998] assessed the direct effect of anthropogenic radiative forcing
on the recent increase in the land surface and tropospheric temperature and the indirect
warming through the SST increase using a set of experiments conducted in an AGCM.
Andrews [2014] investigated the relative contributions of individual radiative forcings
and SST variation to 30-yr trends in surface and atmospheric temperature during 1979–
2008 in an AGCM. Results of Andrews [2014] revealed that the direct effect of
anthropogenic radiative forcings contributed for ~40% of JJA-mean land SAT trend
over the NH during the 30 years. In the present study, we revealed that ADIR is also an
important factor for the recent increase in frequency of hot summers over the NH land
areas.
7
One might suspect if the sum of ASST and ADIR effects estimated in AGCM
runs reconstructs a response in fully-coupled atmosphere-ocean system to the external
radiative forcing. One of the ways to validate the linear additivity of the two effects is to
estimate the two responses to the external forcing simulated in CGCM and compare
with the AGCM results. Gregory et al. [2004], Andrews et al. [2009], and Bony et al.
[2013] and references therein revealed that the direct and indirect responses to CO2
increase simulated in CGCMs (or atmosphere slab-ocean coupled GCMs) can be
decomposed by using an abrupt CO2 increase experiment (step experiment hereafter). In
this framework, the indirect and direct effects are defined by regression coefficient on
global-mean ΔSAT and regression constant (intercept) in the step experiment,
respectively. In addition, Good et al. [2011, 2013] found that responses to external
forcing simulated in transient warming experiments including transient CO2 increase
run (1%CO2) and future scenario runs can be reconstructed by the abrupt responses in
the step experiment. By using these methods, K14 decomposed JJA-mean land SAT
responses to CO2 increase into the direct and indirect components.
Figure S6 shows 9-model mean of JJA-mean ΔSAT over land simulated in
1%CO2 run and its reconstruction and decomposition by using the step experiment and
AMIP-type experiments (amip4xCO2 and amipFuture) [Taylor et al., 2012]. The
models and experiments were detailed in K14. Global-mean ΔSAT in Fig. S6a is 3.9 K.
Anomalies in Fig. S6c, f are scaled with global-mean ΔSAT so that global-mean ΔSAT
in Fig. S6d, g (sum of Fig. S6b and c, and sum of Fig. S6e and f, respectively) are equal
to 3.9 K. The reconstructions based on the step (Fig. S6d) and AGCM experiments (Fig.
S6f) can simulate general features in 1%CO2 (Fig. S6a) including large warming over
the Amazon and middle and high latitudes (e.g. North America). The indirect effect
8
contributes to general warming with a peak in the low latitude (Fig. S6c) and the direct
effect contributes to middle and high latitude warming (Fig. S6b). These characteristics
can largely be found in the two (Fig. S6e, f) estimated in the AMIP-type experiments.
These results indicate that linear additivity holds for the two effects estimated by
AGCM runs with certain accuracy. Note that sea ice concentration prescribed in
amipFuture [Taylor et al., 2012] run (Fig. S6f) is not identical to that simulated in Fig.
S6a, c, resulting in an underestimation (overestimation) of surface warming over high
latitude (low and middle latitudes) including the Arctic coast (Fig. S6f, g). In addition to
ΔSAT, the relative contributions of the fASST, fNAT, and fADIR shown in Fig. 2d, e are
generally similar to those of ΔSAT (Fig. 2b, c), suggesting that non-linearity terms in f
also do not change the main results drastically (the large contribution of ADIR effect
over the NH extratropics in the long-term change, and the NAT effect over the NH
middle latitude in the short-term change; Fig. 2).
Influence of anthropogenic land use change on SAT
The estimated direct effect of anthropogenic forcing (ADIR) contains effects of
anthropogenic land use changes. Kumar et al. [2013] examined the effects of land use
change in 20th and 21st century by using 16 CMIP5 models including MIROC5 model.
Land use change prescribed in CMIP5 historical simulation (Fig. 1 in Kumar et al.,
2013) influences directly on the local land SAT. The land use change acts as a surface
cooling over land in the 20th century during JJA in MIROC5 (–0.12 K, Fig. 7 and Table
5 in Kumar et al., 2013), representing an opposite and small effects to the estimated
TADIR in 1951–2011 period (Figs. 2b, d, 3d). In MIROC5, the warming effects of other
9
anthropogenic forcings including GHGs radiative forcing is more dominant than the
cooling effect of land use change. Note that impacts of land use change in the 20th
century have a large uncertainty among CMIP5 models [Kumar et al., 2013, and
references therein], resulting in a source of uncertainty in TADIR and fADIR. It is also a
remaining issue to quantify possible roles of land surface feedback (including soil
moisture feedback) [e.g. Seneviratne et al., 2010] that may have large influences on the
variation in the frequency of heat waves.
References
Andrews, T., P. M. Forster, and J. M. Gregory (2009), A surface energy perspective on
climate change, J. Clim., 22, 2557–2570, doi:10.1175/2008JCLI2759.1.
Andrews, T. (2014), Using an AGCM to diagnose historical effective radiative forcing
and mechanisms of recent decadal climate change, J. Clim., 27, 1193–1209,
doi:10.1175/JCLI-D-13-00336.1.
Bony, S., et al. (2013), Robust direct effect of carbon dioxide on tropical circulation and
regional precipitation, Nat. Geosci., 6, 447–451, doi:10.1038/ngeo1799.
Bracco, A., et al. (2004), Internal variability, external forcing and climate trends in
multi-decadal
AGCM
ensembles,
Clim.
Dyn.,
23,
659–678,
doi:10.1007/s00382-004-0465-2.
Compo, G. P., and P. D. Sardeshmukh (2009), Oceanic influences on recent continental
warming, Clim. Dyn., 32, 333–342, doi:10.1007/s00382-008-0448-9.
Deser, C., and A. S. Phillips (2009), Atmospheric circulation trends, 1950–2000: the
relative roles of sea surface temperature forcing and direct atmospheric radiative
forcing, J. Clim., 22, 396–413, doi:10.1175/2008JCLI2453.1.
10
Dommenget, D. (2009), The ocean’s role in continental climate variability and change, J.
Clim., 22, 4939–4952, doi:10.1175/2009JCLI2778.1.
Folland, C. K., et al. (1998), Influences of anthropogenic and oceanic forcing on recent
climate change, Geophys. Res. Lett., 25, 353–356, doi:10.1029/97GL03701.
Good, P., J. M. Gregory, and J. A. Lowe (2011), A step-response simple climate model
to reconstruct and interpret AOGCM projections, Geophys. Res. Lett., 38, L01703,
doi:10.1029/2010GL045208.
Good, P., et al. (2013), Abrupt CO2 experiments as tools for predicting and
understanding CMIP5 representative concentration pathway projections, Clim.
Dyn., 40, 1041–1053, doi:10.1007/s00382-012-1410-4.
Gregory, J. M., et al. (2004), A new method for diagnosing radiative forcing and climate
sensitivity, Geophys. Res. Lett., 31, L03205, doi:10.1029/2003gl018747.
Hansen, J., et al. (2010), Global surface temperature change, Rev. Geophys., 48,
RG4004, doi:10.1029/2010RG000345.
Hansen, J., M. Sato, and R. Ruedy (2012), Perception of climate change. Proc. Natl.
Acad. Sci., 109, E2415–E2423, doi: 10.1073/pnas.1205276109.
Hansen, J., M. Sato, and R. Ruedy (2013), Reply to Rhines and Huybers: Changes in
the frequency of extreme summer heat, Proc. Natl. Acad. Sci., 110, E547–E548,
doi:10.1073/pnas.1220916110.
Jones, P. D., et al. (2012), Hemispheric and large-scale land surface air temperature
variations: An extensive revision and an update to 2010, J. Geophys. Res., 117,
D05127, doi:10.1029/2011JD017139.
Joshi, M. M., et al. (2008), Mechanisms for the land–sea warming contrast exhibited by
simulations
of
climate
change,
doi:10.1007/s00382-007-0306-1.
11
Clim.
Dyn.,
30,
455–465,
Joshi, M. M., F. H. Lambert, and M. J. Webb (2013), An explanation for the difference
between twentieth and twenty-first century land–sea warming ratio in climate
models, Clim. Dyn., 41, 1853–1869, doi:10.1007/s00382-013-1664-5.
Kamae, Y., and M. Watanabe (2012), On the robustness of tropospheric adjustment in
CMIP5 models, Geophys. Res. Lett., 39, L23808, doi:10.1029/2012GL054275.
Kamae, Y., and M. Watanabe (2013), Tropospheric adjustment to increasing CO2: its
timescale and the role of land–sea contrast, Clim. Dyn., 41, 3007–3024,
doi:10.1007/s00382-012-1555-1.
Kamae, Y., et al. (2014), Summertime land–sea thermal contrast and atmospheric
circulation over East Asia in a warming climate–Part II: Importance of
CO2-induced continental warming, Clim. Dyn., doi: 10.1007/s00382-014-2146-0.
Kobayashi, C. (2014), Impact of tropical and subtropical SSTs on mid-latitude
tropospheric
warming
in
the
northern
summer
of
2010,
Clim.
Dyn.,
doi:10.1007/s00382-013-2013-4.
Kumar, S., et al. (2013), Land use/cover change impacts in CMIP5 climate simulations:
A new methodology and 21st century challenges, J. Geophys. Res., 118, 6337–
6353, doi:10.1002/jgrd.50463.
Manabe, S., et al. (1991), Transient responses of a coupled ocean-atmosphere model to
gradual changes of atmospheric CO2. Part I: Annual mean response. J. Clim., 4,
785–818, doi: 10.1175/1520-0442(1991)004<0785:TROACO>2.0.CO;2.
Rayner, N. A., et al. (2003), Global analyses of sea surface temperature, sea ice, and
night marine air temperature since the late nineteenth century, J. Geophys. Res.,
108(D14), 4407, doi:10.1029/2002JD002670.
Seneviratne, S. I., et al. (2010), Investigating soil moisture-climate interactions in a
changing
climate:
A
review,
doi:10.1016/j.earscirev.2010.02.004.
12
Earth-Sci.
Rev.,
99,
125–161,
Schubert, S., et al. (2014), Northern Eurasian heat waves and droughts, J. Clim., 27,
3169–3207, doi:10.1175/JCLI-D-13-00360.1.
Taylor, K. E., R. J. Stouffer, and G. A. Meehl (2012), An overview of CMIP5 and the
experiment
design,
Bull.
Am.
Meteorol.
Soc.,
93,
485–498,
doi:10.1175/BAMS-D-11-00094.1.
Watanabe, M., et al. (2010), Improved climate simulation by MIROC5: Mean states,
variability,
and
climate
sensitivity,
J.
Clim.,
23,
6312–6335,
doi:10.1175/2010JCLI3679.1.
Table captions
Table S1. Observed and modeled trends of ΔSAT and frequency of hot summers (f).
Linear trends (during 1981–2011 and 1997–2011) are calculated from JJA-mean
values averaged over the NH land areas. * and ** denote that trends are statistically
significant at 90% and 95% confidence levels, respectively.
Table S2. List of the CMIP5 models used in Fig. 1.
Figure captions
Fig. S1. Frequency of hot summers derived from GISS and CRUTEM4. (a) Frequencies
of hot and very hot summers defined by SAT anomalies in the different categories
(two and three standard deviations) over NH land areas. Black lines are GISS
temperature (same as Fig. 1b) and red lines are CRUTEM4. (b, c) Fractional
occurrences of hot and (d, e) very hot summers derived from (b, d) GISS and (c,
e) CRUTEM4 during 1991–2011 period.
13
Fig. S2. Global annual-mean SST prescribed in the simulations. Black (magenta) line is
SST used for the ALL and SST runs (NAT run). The values shown are anomalies
relative to 1951–1980 mean of ALL run.
Fig. S3. Net aerosol radiative forcing (shortwave plus longwave, positive downward) at
the tropopause. (a–d) The ALL, (e–h) NAT and SST runs. (a, e) 1951–1980 mean.
(b, f) Differences between 1991–2011 and 1951–1980 (long-term change), and (c,
g) 2001–2011 and 1991–2000 (short-term change). (e, h) Zonal-means of 1951–
1980 mean (black), long-term (red) and short-term changes (blue) over the land
areas.
Fig. S4. Modeled anomalies (relative to 1951–1980) of JJA-mean SAT averaged over
the NH land areas in the ALL run. Black line represents an average over the
whole NH land area. Red line is the average excluding missing data area of GISS
shown as white area in the map (same as Fig. 1a).
Fig. S5. Frequency of occurrence of local SAT anomalies over the Northern Hemisphere
relative to 1951–1980 divided by local standard deviation. (a) Observation, (b)
ALL, (c) NAT and (d) SST runs. (e) Difference of frequency between 2001–2011
and 1951–1980. The thick grey line is 1951–1980 mean in observation. (f–i)
Differences of probability beyond two standard deviations (black) between 2001–
2011 and 1951–1980. Red and blue bars represent PDF shift effect and PDF
14
broadening effect, respectively. Error bars represent 95% confidence limits.
Fig. S6. JJA-mean CMIP5 model ensemble means of land ΔSAT. (a) 1% per year CO2
increase experiment (1%CO2) in coupled general circulation models (CGCMs).
The anomalies are calculated by the difference of 30-year averages between 111–
140 and the corresponding period in pre-industrial control run. (b, c)
decompositions of direct and indirect (via SST change) effects of increasing CO2
estimated by using CO2 quadrupling step experiment conducted in CGCMs. The
indirect and direct effects are defined by regression coefficient on global-mean
ΔSAT and regression constant (intercept), respectively (see Good et al., 2011,
2013; K14). (d) Sum of (b) and (c). (e) Anomaly in AMIP-type CO2 quadrupling
but fixed-SST experiment (amip4xCO2). (f) Anomaly in patterned-SST increase
(global-mean SST increase is 4.0 K) but fixed-CO2 experiment (amipFuture).
Anomalies are calculated by differences of 30-year averages relative to AMIP
control run (amip). Details of the simulations and models are described in K14.
Values in (c, f) are scaled with difference of global-mean ΔSAT between (a) and
(b, e) so that the global-mean ΔSAT in (d, g) are equal to that of (a) (3.9 K).
Fig. S7. Anomalies of JJA-mean SAT in response to CO2 quadrupling in AGCM run. (a)
MIROC5, (b) 13 model ensemble mean from CMIP5 dataset. Experiments and
data used were detailed in Kamae and Watanabe [2012]. (c) Zonal-mean over the
land. Black line and grey shading are CMIP5 ensemble mean and its standard
deviation, and red line is MIROC5.
15
Download