This content has been downloaded from IOPscience. Please scroll down to see the full text. Download details: IP Address: 207.162.87.194 This content was downloaded on 28/06/2017 at 15:26 Please note that terms and conditions apply. Seasonal climate change patterns due to cumulative CO 2 emissions View the table of contents for this issue, or go to the journal homepage for more 2017 Environ. Res. Lett. 12 075002 (http://iopscience.iop.org/1748-9326/12/7/075002) Home Search Collections Journals About Contact us My IOPscience
Transcript
This content has been downloaded from IOPscience. Please scroll down to see the full text.
Download details:
IP Address: 207.162.87.194
This content was downloaded on 28/06/2017 at 15:26
Please note that terms and conditions apply.
Seasonal climate change patterns due to cumulative CO2 emissions
View the table of contents for this issue, or go to the journal homepage for more
2017 Environ. Res. Lett. 12 075002
(http://iopscience.iop.org/1748-9326/12/7/075002)
Home Search Collections Journals About Contact us My IOPscience
Supplementary material for this article is available online
AbstractCumulative CO2 emissions are near linearly related to both global and regional changes inannual-mean surface temperature. These relationships are known as the transient climateresponse to cumulative CO2 emissions (TCRE) and the regional TCRE (RTCRE), and have beenshown to remain approximately constant over a wide range of cumulative emissions. Here, weassessed how well this relationship holds for seasonal patterns of temperature change, as well asfor annual-mean and seasonal precipitation patterns. We analyzed an idealized scenario with CO2
concentration growing at an annual rate of 1% using data from 12 Earth system models from theCoupled Model Intercomparison Project Phase 5 (CMIP5). Seasonal RTCRE values fortemperature varied considerably, with the highest seasonal variation evident in the Arctic, whereRTCRE was about 5.5 °C per Tt C for boreal winter and about 2.0 °C per Tt C for borealsummer. Also the precipitation response in the Arctic during boreal winter was stronger thanduring other seasons. We found that emission-normalized seasonal patterns of temperaturechange were relatively robust with respect to time, though they were sub-linear with respect toemissions particularly near the Arctic. Moreover, RTCRE patterns for precipitation could not bequantified robustly due to the large internal variability of precipitation. Our results suggest thatcumulative CO2 emissions are a useful metric to predict regional and seasonal changes inprecipitation and temperature. This extension of the TCRE framework to seasonal and regionalclimate change is helpful for communicating the link between emissions and climate change topolicy-makers and the general public, and is well-suited for impact studies that could make useof estimated regional-scale climate changes that are consistent with the carbon budgets associatedwith global temperature targets.
1. Introduction
Cumulative CO2 emissions have been shown to be auseful metric to estimate the global-mean temperaturechange resulting from human CO2 emissions (Mat-thews et al 2009, Allen et al 2009). A growing body ofresearch has confirmed that global temperaturesrespond approximately linearly to cumulative emis-sions (Matthews et al 2009, Gillett et al 2013, Collinset al 2013, Leduc et al 2015) which has led to theformal definition of this relationship as the ‘transient
climate response to cumulative CO2 emissions’(TCRE) (Gillett et al 2013, Collins et al 2013).Estimates of the TCRE have now been widely used tocompare model responses to cumulative emissions, aswell as to estimate the total allowable emissions (orcarbon budgets) associated with different levels ofglobal-mean temperature change (Meinshaisen et al2009, Zickfeld et al 2009, Friedlingstein et al 2014,Rogelj et al 2016).
While useful to inform efforts to meet climatemitigation targets such as 2 °C of global warming, the
Table 1. CMIP5 models and the institutions that provided the model data used in this study.
Model name Modeling Center or Group
BNU-ESM College of Global Change and Earth System Science, Beijing Normal University
CanESM2 Canadian Centre for Climate Modelling and Analysis
CESM1-BGC Community Earth System Model Contributors
HadGEM2-ES Met Office Hadley Centre (additional HadGEM2-ES realizations contributed by InstitutoNacional de
PesquisasEspaciais)
INM-CM4 Institute for Numerical Mathematics
IPSL-CM5A-LR Institut Pierre-Simon Laplace
IPSL-CM5A-MR Institut Pierre-Simon Laplace
IPSL-CM5B-LR Institut Pierre-Simon Laplace
MIROC-ESM Japan Agency for Marine-Earth Science and Technology, Atmosphere and Ocean Research Institute (The
University of Tokyo), and National Institute for Environmental Studies
MPI-ESM-LR Max-Planck-InstitutfurMeteorologie (Max Planck Institute for Meteorology)
MPI-ESM-MR Max-Planck-InstitutfurMeteorologie (Max Planck Institute for Meteorology)
NorESM1-ME Norwegian Climate Centre
Environ. Res. Lett. 12 (2017) 075002
TCRE is limited in its ability to estimate the likelihoodof climate impacts that manifest at the local (ratherthan global) scale. To expand the utility of the TCREframework for climate impact assessments, Leduc et al(2016) recently demonstrated that regional changes inannual-mean temperature can also be linearly relatedto cumulative CO2 emissions. This analysis suggeststhat there is considerable potential to quantify regionalTCRE (or RTCRE) values based on the patterns ofclimate response to cumulative emissions as anextension of the current global TCRE framework.
Furthermore, there is reason to expect thatseasonal temperature change patterns, as well aspatterns of annual-mean and seasonal precipitation,might also scale approximately linearly as a function ofcumulative emissions. For many years, researchershave employed a technique known as pattern-scalingto estimate the regional temperature and precipitationchanges associated with a given level of global meanwarming (Santer et al 1990, National ResearchCouncil et al 2011, Tebaldi and Arblaster 2014). Thistechnique is premised on the finding that climatepatterns normalized by global-mean temperatureremain approximately constant across a wide rangeof global temperature changes, which has been shownto hold for both annual-mean and seasonal patterns ofboth temperature and precipitation (National Re-search Council et al 2011, Tebaldi and Arblaster 2014,Mitchell 2003). Given a linear global temperatureresponse to cumulative emissions, we therefore expectthat it should be possible to extend the pattern-scalingapproach from global-mean temperature to cumula-tive emissions, and apply this to both annual-meanand regional temperature and precipitation changepatterns.
In this paper, we calculate annual-mean andseasonal regional temperature and precipitationresponses to cumulative CO2 emissions based on anensemble of models from the Coupled ModelIntercomparison Project Phase 5 (CMIP5) (Tayloret al 2012). In addition to the RTCRE patterns, wequantify the robustness of the patterns in terms of
2
inter-model spread, calculate the contribution ofinternal variability to inter-model spread, and assessto what extent these RTCRE patterns remain stableover time.
2. Methods
We used simulation data from the Coupled ModelIntercomparison Project Phase 5 (CMIP5) (Tayloret al 2012). Specifically, we analyzed temperature andprecipitation data from a scenario where CO2
concentration was rising annually by 1 percent frompreindustrial levels (about 285 ppm) to fourfold thislevel after 140 years (experiment 1pct CO2) (Gillettet al 2013). As non-CO2 forcings were not consideredin this experiment, it allows an excellent opportunityto calculate the effects of cumulative CO2 emissionsalone. We used data from 12 earth system models;this model ensemble (table 1) was the same as thatused by Leduc et al (2016), and for each model, onlyone realization of the experiment was available. Inaddition to using annual mean values (ANN), we didour analysis using also the means over four seasons:December–January–February (DJF), March–April–May (MAM), June–July–August (JJA), and Septem-ber–October–November (SON). Regional changes intemperature and precipitation were averaged season-ally, but the scaling was still done using annual globaltotal CO2 emissions.
The data processing methodology that we used isbased on that of Leduc et al (2016). To study the globalmean response, we evaluated the TCRE for tempera-ture (TCRET) as:
TCRET m; tð Þ ¼ ΔT gm m; tð ÞE m; tð Þ ð1Þ
where ΔTgm m; tð Þ is change in global mean tempera-ture for modelm at time t, and E(m, t) is the diagnosedcumulative CO2 emissions. The metric for precipita-tion (TCREP) was calculated in the same way as a ratioof change in global mean precipitation and cumulative
Environ. Res. Lett. 12 (2017) 075002
emissions. We evaluated TCRE by comparing themeans over 20-year average windows at the start of thesimulation and at the time of doubling of CO2
concentration, using the model ensemble mean as thebest estimate of the TCRE.
RTCRE for temperature (RTCRET) was evaluatedfor each model m, grid-cell x, and time t as:
RTCRET m; x; tð Þ ¼ ΔT m; x; tð ÞE m; tð Þ ð2Þ
where DT is the local change in temperature and E isthe diagnosed cumulative CO2 emissions. The metricfor precipitation (RTCREP) was calculated in thesame way. As for global mean response, we used20-year average windows centered at the time ofdoubling of CO2 concentration to evaluate theannual and seasonal means local changes, which isstandard practice when evaluating climate changepatterns. As for TCRE, we used the model ensemblemean as the best estimate for RTCRE. We calculatedthe spatial multi-model means by first interpolatingall model data onto the grid of CanESM2 (2.8° ofresolution).
To test whether the RTCRE stays constant overtime, we calculated RTCRE for different 20-yearwindows relative to a same reference period (the first20 years of the time series). We then show to whatextent the zonal mean temperature and relativeprecipitation change remains stable over the 140 yearduration of these simulations.
To assess the robustness of the climate changepatterns, we calculated the ratio of the RTCRE andinter-model spread defined as:
R xð Þ ¼ jRTCRE xð ÞjsIMS xð Þ ð3Þ
where RTCRE(x) is the ensemble mean RTCREchange in location x and sIMSðxÞ is the standarddeviation of RTCRE in the model ensemble. Both areevaluated at the time of doubling of the CO2
concentration as described above. The ratio is basicallya signal to inter-model spread ratio, which is aconvenient measure since the inter-model spread isoften the largest where the signal is also large (Leducet al 2016).
The previous measure of robustness can beinterpreted as a standard signal to noise ratio in caseswhere the internal variability dominates the inter-model spread. To account for this effect, we calculatedhow the internal variability of the climate affects theRTCRE. As in Leduc et al (2016), we estimated,usingthe variance inflation factor from Wilks (2011), thatover a N-year period, the effect of internal variabilityon RTCRE is:
s ðxÞ2 ¼ 2s1ðxÞ2 1þ fðxÞ ð4Þ
IVNE2 1� fðxÞ
3
where s1 xð Þ2 is the inter-annual variance of thedetrended time series of annual mean temperature orrelative change in precipitation with respect to the 20year reference period. E is the cumulative CO2
emissions at the time of doubling of the CO2
concentration, and f(x) is autocorrelation of thedetrended time series with lag of one year. Autocorre-lation with lag of two years was negligible for bothvariables, and therefore higher-order terms are notneeded to calculate the internal variability. Wedetrended using a fourth-order polynomial fit(Hawkins and Sutton 2009, Leduc et al 2016). Toquantify the contribution of the internal variability tothe inter-model spread, we thus calculated the ratio ofthe effect of internal variability on the RTCRE to theinter-model spread in RTCRE:
G xð Þ ¼ sIV xð ÞsIMS xð Þ : ð5Þ
3. Results
3.1. Annual mean temperature and precipitationresponseFigures 1(a) and (b) show the change in global meantemperature and precipitation as a function ofcumulative CO2 emissions for all 12 models and forthe ensemblemean. The plots also contain straight linesrepresenting ensemble mean TCRE at the time ofdoubling ofCO2 concentration. In general,models witha strong temperature response had also a strongprecipitation response. The ensemble mean TCRETand TCREP were 1.68 °C per TtC and 2.5% per TtC,respectively. Following the convention established inprevious works (Matthews et al 2009, Gillett et al 2013,Leduc et al 2015 and 2016), these values were calculatedas a change (in temperature or precipitation) percumulative CO2 emissions at the time when theatmospheric CO2 concentration has doubled its initialconcentration for each model individually beforecalculating the ensemble mean. Although evaluatingTCRE at the time of doubling of CO2 concentration hasbeen shown to be more stable than if it is calculated at acertain amount of cumulative emissions (Leduc et al2015), the ensemblemeanwas very similar usingbothofthese two methods (compare solid and dashed blacklines in figures 1 (a) and (b)).
Figures 1(c) and (d) show the ensemble meanannual mean RTCRE for temperature and precipita-tion. As discussed in Leduc et al (2016), land areas andespecially the Arctic showed a strong temperatureresponse to CO2 emissions mainly due to sea-icefeedback (Kumar et al 2010, Screen and Simmonds2010). RTCRET values were greater than two times theinter-model spread (sIMS) over most of the globe(hatched regions in figure 1(c)). Here we also show theRTCREP as a relative change in precipitation per Tt Cemitted. The strongest positive response took place at
(a) (b)
(c) (d)
TCRET for ANN
RTCRET for ANN RTCREP for ANN
TCREP for ANN5
4
3
2
1
0
10
8
6
4
2
0
–2
6.05.55.04.54.03.53.02.52.01.51.00.50.0
302520151050–5–10–15–20–25–30
0.0 0.5 1.0 1.5 2.0 0.0 0.5 1.0 1.5 2.0Cumulative emissions (Tt C) Cumulative emissions (Tt C)
Δ T
(°C
)
Δ P
(%)
°C p
er T
t C
% p
er T
t C
Ensemble mean at CO2 doublingEnsemble meanBNU-ESMCanESM2CESM1-BGCHadGEM2-ESinmcm4IPSL-CM5A-LRIPSL-CM5A-MRIPSL-CM5B-LRMIROC-ESMMPI-ESM-LRMPI-ESM-MRNorESM1-ME
Figure 1. Ensemble mean of annual TCRE for (a) temperature and (b) precipitation and RTCRE for (c) temperature (reproducedfigure 1(a) from Leduc et al (2016) with hatching added) and (d) precipitation. Hatching indicates areas where |RTCRE| > 2sIMS.(c) Reprinted by permission from Macmillan Publishers Ltd: Nat. Clim. Change 6 474–8, copyright 2016.
(a) (b)
(c) (d)
TCRET for DJF
RTCRET for DJF
TCRET for JJA
RTCRET for JJA
Cumulative emissions (Tt C) Cumulative emissions (Tt C)
Δ T
(°C
)
Δ T
(°C
)
5
4
3
2
1
0
5
4
3
2
1
0
0.0 0.5 1.0 1.5 2.0 0.0 0.5 1.0 1.5 2.0
Ensemble mean at CO2 doublingEnsemble meanBNU-ESMCanESM2CESM1-BGCHadGEM2-ESinmcm4IPSL-CM5A-LRIPSL-CM5A-MRIPSL-CM5B-LRMIROC-ESMMPI-ESM-LRMPI-ESM-MRNorESM1-ME
9876543210
9876543210
°C p
er T
t C
°C p
er T
t C
Figure 2. Seasonal TCRET in (a) DJF and (b) JJA and seasonal RTCRET in (c) DJF and (d) JJA. Hatching indicates areas where |RTCRE| > 2sIMS.
Environ. Res. Lett. 12 (2017) 075002
the Equator in the Pacific Ocean and NortheasternAfrica, where RTCREP was about 30%–40% per Tt C.Over land areas, precipitation decreased considerably(by 8%–32% per Tt C) in Southern Europe, North-western Africa, and Southern parts of North America.RTCREP was mostly positive (about 5%–25% perTt C) at both poles. Unlike for temperature, only oversmall regions RTCREP was greater than 2sIMS,indicating considerably lower robustness as discussedin more detail in section 3.4.
3.2. Seasonal temperature and precipitation responseFigures 2(a) and (b) show seasonal TCRET in DJFand in JJA, respectively. Global mean responses were
4
fairly similar to annual mean response for all seasonsin terms of both model spread and ensemble mean.TCRET was 1.72, 1.63, 1.64, and 1.73 °C per Tt C forDJF, MAM, JJA, and SON, respectively. See figures S1(a) and (c) available at stacks.iop.org/ERL/12/075002/mmedia for TCRET for MAM and SON, respectively.
Figures 2(c) and (d) show the RTCRET duringDJF and JJA, respectively. Arctic amplification wasmost visible during DJF, where the highest value ofRTCRET was 8.9 °C per Tt C in some parts of theArctic. Everywhere below latitude 50 °N, RTCRETwas below 3.8 °C per Tt C. There was little variationinside the latitude band between 40 °S and 30 °N: thezonal mean RTCRET was between 1.3 and 1.6 °C per
Cumulative emissions (Tt C)Cumulative emissions (Tt C)
Δ P
(%)
Δ P
(%)
10
8
6
4
2
0
–2
10
8
6
4
2
0
–20.0 0.5 1.0 1.5 2.0 0.0 0.5 1.0 1.5 2.0
Ensemble mean at CO2 doublingEnsemble meanBNU-ESMCanESM2CESM1-BGCHadGEM2-ESinmcm4IPSL-CM5A-LRIPSL-CM5A-MRIPSL-CM5B-LRMIROC-ESMMPI-ESM-LRMPI-ESM-MRNorESM1-ME
302520151050–5–10–15–20–25–30
302520151050–5–10–15–20–25–30
% p
er T
t C
% p
er T
t C
Figure 3. Seasonal TCREP in (a) DJF and (b) JJA and seasonal RTCRET in (c) DJF and (d) JJA. Hatching indicates areas where |RTCRE|> 2sIMS.
Environ. Res. Lett. 12 (2017) 075002
Tt C within this region. For other seasons, the zonalvariability of TCRET was considerably lower butland-sea contrasts were still apparent. For JJA,RTCRET was below 3.3 °C per Tt C everywhere,with the largest values occurring in Northern Eurasiaand over parts of the Southern Ocean.
TCREP was 2.66, 2.40, 2.22, and 2.55 % per Tt Cfor DJF, MAM, JJA, and SON, respectively. RTCREPfor annual means (figure 1(b)), DJF (figure 3(a)),MAM (figure S2(b)), JJA (figure 3(b)), and SON(figure S2(d)) were relatively similar in mid-latitudesand near the Equator. The most significant example ofseasonal variability was the stronger precipitationresponse in the Arctic during DJF and SON. RTCREPin the Arctic (above 60°N) was mostly around about20%–40% per Tt C in DJF, whereas for other seasonsthe values were mostly below 24% per Tt C. Therewere clear seasonal differences also in Southern Africa:a slight positive response (0%–8% per Tt C) in DJFand mainly negative response in other seasons (up toabout �20% per Tt C). Also parts of Australia,Northern America, and Southern America had adifferent sign of RTCREP in different seasons.
As discussed in section 3.4 and shown with thelimited extent of the hatched areas in figures 3(c) and(d) and figures S3(b) and (d), precipitation responsehad larger inter-model spread and was more sensitiveto interannual variability than the temperatureresponse. Therefore, the results on precipitationshould be interpreted with caution.
3.3. Stability of climate change patterns over timeFigure 4 shows zonal mean of RTCRE over time forboth temperature and precipitation. RTCRET stayedrelatively stable over time. However, there was clearlynoticeable decrease especially in the Arctic region, butalso elsewhere in both annual and seasonal mean
5
values (figures 4(a)–(c), figures S3(a) and (b)). Thisgeneral decrease is consistent with previous work thathas shown that TCRE deviates from linearity at veryhigh levels of cumulative emissions (e.g. Leduc et al2015, Herrington and Zickfeld 2014) The largerdecrease of RTCRET in the Arctic can be attributedmostly to the weakening of the sea-ice albedo feedback(Leduc et al 2016). Although the Arctic amplificationmanifested mostly during DJF, the relative decreasewith respect to time of zonal mean of RTCRE wassimilar for all DJF and JJA, suggesting that seasonalsea-ice albedo feedback changes are of comparablemagnitude to the annual-mean change.
RTCREP stayed also relatively constant over time(figures 4(d) and (e), figures S3(c) and (d)). Differentseasons had changes partly in opposite directions thatmade the annual mean RTCREP more stable. Forexample, in DJF, precipitation increase over the Arcticwas stronger at later years, and weaker in SON. Therewas also some non-monotonic behavior in theprecipitation response. During JJA, at latitudes around50 °N, precipitation response was positive for the timewindow of the years 20–40, then negative again for theyears 40–80, and changed again to positive for the restof the experiment. At least part of this variation of signwas related to noise from the early years when both thesignal and the emissions were small. Therefore, weshow results beginning from the years 60–80 infigure 4. In general, there was not such a clear trend forprecipitation as seen in the decreased temperatureresponse over time.
3.4. Robustness of RTCRE patterns and thecontribution of internal variabilityIn the following, the ratio of RTCRET to inter-modelspread (RT, equation (3)) is used as a measure ofrobustness of the models’ climate change patterns. In
(a) (b) (c)
(d) (e) (f)
RTCRET for ANN RTCRET for DJF RTCRET for JJA
RTCREP for ANN RTCREP for DJF RTCREP for JJAYear
Year
Year
Year
Year
Year
Latit
ude
Latit
ude
Latit
ude
Latit
ude
Latit
ude
Latit
ude
°C p
er T
t C%
per
Tt C
°C p
er T
t C%
per
Tt C
°C p
er T
t C%
per
Tt C
50
0
–50
50
0
–50
50
0
–50
50
0
–50
50
0
–50
50
0
–50
60 80 100 120 140
60 80 100 120 140
60 80 100 120 140
60 80 100 120 140
60 80 100 120 140
60 80 100 120 140
6.05.55.04.54.03.53.02.52.01.51.00.50.0
10
9
8
7
6
5
4
3
2
1
0
3.5
3.0
2.5
2.0
1.5
1.0
0.5
0.0
40
32
24
16
8
0
–8
–16
–24
–32
–40
40
32
24
16
8
0
–8
–16
–24
–32
–40
40
32
24
16
8
0
–8
–16
–24
–32
–40
Figure 4. Ensemble mean of zonal mean RTCRE for temperature for (a) annual means, (b) DJF, (c) JJA, and for precipitation in (d)annual means, (e) DJF, and JJA.
(a) (b) (c)
(d) (e) (f)
(g) (h) (i)
(j) (k) (l)
RT for ANN RT for DJF RT for JJA
GT for ANN GT for DJF GT for JJA
RP for ANN RP for DJF RP for JJA
GP for ANN GP for DJF GP for JJA
109876543210
109876543210
109876543210
1.00.90.80.70.60.50.40.30.20.10.0
1.00.90.80.70.60.50.40.30.20.10.0
1.00.90.80.70.60.50.40.30.20.10.0
1.00.90.80.70.60.50.40.30.20.10.0
1.00.90.80.70.60.50.40.30.20.10.0
1.00.90.80.70.60.50.40.30.20.10.0
4.03.53.02.52.01.51.00.50.0
4.03.53.02.52.01.51.00.50.0
4.03.53.02.52.01.51.00.50.0
Figure 5. The robustness of the RTCRE patterns and the contribution of internal variability to inter-model spread. The first two rowsshow the robustness (ratio of ensemble mean RTCRE to inter-model spread of RTCRE for temperature, RT) for (a) annual means(ANN), (b) DJF, (c) JJA, and ratio of ensemble mean of the internal variability of RTCRE to the ensemble standard deviation ofRTCRE (GT) for (d) annual means, (e) DJF, and (f) JJA. The two bottom rows show the robustness for precipitation (RP) for (g)annual means, (h) DJF, (i) JJA, and ratio of the ensemble mean of the internal variability of RTCRE to the ensemble standard deviationof RTCRE (GP) for (j) annual means, (k) DJF, and (l) JJA.
Environ. Res. Lett. 12 (2017) 075002
general, models agreed relatively well in their RTCRETpatterns within the mid-latitudes and the Equator (RTmostly between 4 and 8, see figures 5(a)–(c)). Eventhough the signal in the annual means (figure 1(c))and in DJF (figure 2(c)) was largest at high latitudes,
6
also the model spread was large and thus RT valueswere lower there than at mid-latitudes and theRTCRET values were lower than 2sIMS. Patterns showimportant disagreements in the Southern Ocean andin the poles (RTmostly below 3). figures 5(d)–(f) show
Environ. Res. Lett. 12 (2017) 075002
the ratio of internal variability to the inter-modelspread (GT, equation (5)) for temperature. Relativeimportance of the internal variability was considerablyhigher for seasonal means than for the annual mean.Especially in North America, Australia, and Europe inDJF, the internal variability component was importantrelative to the whole inter-model spread with valuesof GT being around 1 (note that GT occasionallyexceeded 1, because one ensemble member per modelis not enough to capture all internal variability andtherefore there is some sampling error that affect theevaluation of equations (4) and (5)). In the areaswhere internal variability dominated the inter-modelspread, more (than 1) ensemble members forindividual models would help to produce more robustpatterns. Conversely, in regions where internalvariability was small compared to the inter-modelspread, having more members would likely notincrease the robustness of the climate change patterns.Rather, the latter case refers to the situation where animprovement of the patterns involves a reduction ofthe overall uncertainty in regional climate sensitivity.In JJA, the contribution of internal variability to theinter-model spread was small over most of land areaswith exception of Australia, Uruguay, and eastern partof the United States.
The robustness (RP) (figures 5(g)–(i)) wasconsiderably lower for precipitation than for temper-ature. The values for RP were mostly below 3, whichmeans that the signal is relatively weak compared tothe model spread over almost all regions. In particular,in most regions where RP was smaller than 0.5, lessthan 75% of the models agreed on the sign of thechange. The robustness was higher in SouthernEurope (where precipitation decreased) during JJAand in Siberia and Hudson Bay (where precipitationincreased) during DJF with RP values of about 2–4.
Internal variability also had a large effect on therobustness of the TCREP metric (figures 5(j)–(l)).Already for the annual results, the ratio between theeffect of internal variability on RTCREP and inter-model spread (GP) was over 0.8 over large regionsreaching values of over 3 in certain areas. Note thatregions with GP > 1 were larger for precipitation,because internal variability was more difficult tocalculate accurately for precipitation than for temper-ature. The large effect of internal variability was clearlyvisible, when we plotted the regional change inprecipitation against cumulative CO2 emissions overGiorgi regions (figure S4, Giorgi and Francisco 2000).For some regions, such as Australia (figure S5(a)),Sahara (figure S5(o)), Central Asia (figure S5(s)), itwas practically impossible to evaluate the meanRTCREP because of the noise generated by internalvariability in these low-precipitation regions. Meanscalculated over larger areas are more likely to be stable.As the Giorgi regions had different areas, ranking ofregions with similar amount of noise should be donewith caution.
7
Internal variability had an even larger role in theseasonal analysis. Regions with GP> 1 were larger andthe values of GP were considerably greater everywherecompared to the annual results (figures 5(j)–(l)).Thenoise from internal variability was also very evident inthe regional mean plots (figures S6–7), suggesting thatit may be difficult to detect the effect of cumulativeCO2 emissions on seasonal precipitation changes overmost areas. However, there were some exceptions thatshow a more robust relationship, notably Greenland(figure S6(i) and figure S7(i)), Alaska (figure S6(h) andfigure S7(h)), and North Asia (in JJA, figure S7(u)) hadall clear increase in precipitation.
4. Discussion
The patterns of temperature and precipitation changescaled with cumulative CO2 emissions look visuallyvery similar to those scaled with global meantemperature change (Tebaldi and Arblaster 2014).This is not surprising, given the near linear relation-ship between cumulative emissions and global meantemperature change as well as the consistent use ofCMIP5 data (although using different scenarios anddifferent model ensembles). This gives us confidencethat our approach can be presented as a robust andstraightforward extension of established and widelyused pattern-scaling techniques.
Scaling climate change patterns with cumulativeCO2 emissions rather than global mean temperatureprovides a more direct link between human actionsand climate change. Estimating global mean tempera-ture change with TCRE provides an easy-to-under-stand way of framing the climate response to CO2
emissions. Our work here, together with Leduc et al(2016) presents a generalized TCRE framework, whichlinks cumulative CO2 emissions directly to regionaland seasonal change in temperature and precipitation.With traditional pattern scaling with global meantemperature, a simple climate model is required topredict the evolution of the global mean temperatureaccording to a given amount of CO2 emissions (e.g.Goodwin et al 2015). The generalized TCREframework presented here provides a simple methodto link emissions to expected climate change basedsimply on estimates of anticipated cumulativeemissions associated with a given scenario, thereforeremoving the need to use a simple climate modelbefore pattern scaling.
In theory, this extended TCRE framework couldbe improved further by employing more advancedmethods of pattern scaling. For example, the TimeShift Approach (Herger et al 2015) could be used inalmost identical way to the approach we have usedhere. In this method, the climate change patterns atcertain amount of cumulative CO2 emissions would bedirectly taken from another simulation at a time of theequivalent cumulative CO2 emissions. This would
Environ. Res. Lett. 12 (2017) 075002
preserve the internal consistency of temperature andprecipitation patterns and implicitly take into accountnon-linearities involved in the retreat of Arctic sea ice.However, the improved accuracy would reducetheconceptual simplicity of scaling climate changepatterns with cumulative CO2 emissions and makeit less usable for climate impact studies.
Our analysis here was based on CO2-only forcingscenarios. The role of non-CO2 greenhouse gases,aerosols, and land-use change could be evaluated bycomparing the results here to analysis of the RCPscenarios with other forcings included. This compari-son will be addressed in future work.
5. Conclusions
We analyzed data from an ensemble of 12 CMIP5models to assess to what extent regional and seasonalchanges in temperature and precipitation can bedescribed with a linear relationship to cumulative CO2
emissions. Overall, seasonal RTCRET was a fairlyrobust measure across models (signal was greater thantwo inter-model standard deviations over most of theglobe), time and emissions. Thus, cumulative CO2
emissions can be used to predict temperature changeboth at global and regional scale for different seasonswith almost similar robustness as for the annualmeans.
For precipitation the picture is more complicated.Although global total precipitation change scaled wellwith cumulative CO2 emissions for both annual andseasonal means, the regional change in precipitationwas much more uncertain. First, disagreementsbetween different models were large, even on thesign of change. Second, large internal variability madeit difficult to extract a clear signal. This was true evenwhen averaged over large regions. This large role ofinternal variability in precipitation response is,however, not specific to scaling the precipitationresponse with cumulative CO2 emissions. The samelack of robustness applies to the patterns scaled withglobal mean temperature change (Tebaldi andArblaster 2014), and our results support the hypothe-sis by Tebaldi and Arblaster (2014) that internalvariability can contribute tomodel disagreements evenwhen 20 year averages are used.
Our analysis demonstrates that cumulative CO2
emissions can be linearly linked to regional andseasonal changes in temperature and to some degreealso to changes in precipitation. This relationship canbe used in assessing carbon budgets (van Vuuren et al2016) also at regional and seasonal level, and in climateimpact studies without using even a simple climatemodel before pattern scaling, and is also a usefulconcept in communicating with the general publicabout how climate will respond to continuedanthropogenic CO2 emissions.
8
Acknowledgments
A.-I. Partanen was supported by a research grant fromEmil Aaltonen Foundation. We acknowledge theWorld Climate Research Programme's WorkingGroup on Coupled Modelling, which is responsiblefor CMIP, and we thank the climate modeling groups(listed in table 1 of this paper) for producing andmaking available their model output. For CMIP theUS Department of Energy’s Program for ClimateModel Diagnosis and Intercomparison providescoordinating support and led development of softwareinfrastructure in partnership with the Global Organi-zation for Earth System Science Portals.
References
Allen M R, Frame D J, Huntingford C, Jones C D, Lowe J A,Meinshausen M and Meinshausen N 2009 Warming causedby cumulative carbon emissions towards the trillionthtonne Nature 458 1163–6
Collins M et al 2013 Long-term climate change: projections,commitments and irreversibility Climate Change 2013: ThePhysical Science Basis. Contribution of Working Group I tothe Fifth Assessment Report of the Intergovernmental Panelon Climate Change (Cambridge: Cambridge UniversityPress) pp 1029–136
Friedlingstein P et al 2014 Persistent growth of CO2 emissionsand implications for reaching climate targets Nat. Geosci 7709–15
Gillett N P, Arora V K, Matthews D and Allen M R 2013Constraining the ratio of global warming to cumulative CO2
emissions using CMIP5 simulations J. Clim. 26 6844–58Giorgi F and Francisco R 2000 Uncertainties in regional climate
change prediction: a regional analysis of ensemblesimulations with the HADCM2 coupled AOGCM Clim.Dyn. 16 169–82
Goodwin P, Williams R G and Ridgwell A 2015 Sensitivity ofclimate to cumulative carbon emissions due tocompensation of ocean heat and carbon uptake Nat.Geosci. 8 29–34
Hawkins E and Sutton R 2009 The potential to narrowuncertainty in regional climate predictions Bull. Am.Meteorol. Soc. 90 1095–107
Herger N, Sanderson B M and Knutti R 2015 Improved patternscaling approaches for the use in climate impact studiesGeophys. Res. Lett. 42 3486–94
Herrington T and Zickfeld K 2014 Path dependence of climateand carbon cycle response over a broad range ofcumulative carbon emissions Earth Syst. Dyn. 5 409–22
Kumar A, Perlwitz J, Eischeid J, Quan X, Xu T, Zhang T,Hoerling M, Jha B and Wang W 2010 Contribution of seaice loss to Arctic amplification Geophys. Res. Lett. 37L21701
Leduc M, Matthews H D and de Elía R 2015 Quantifying thelimits of a linear temperature response to cumulative CO2
emissions J. Clim. 28 9955–68Leduc M, Matthews H D and de Elía R 2016 Regional estimates
of the transient climate response to cumulative CO2
emissions Nat. Clim. Change 6 474–8Matthews H D, Gillett N P, Stott P A and Zickfeld K 2009 The
proportionality of global warming to cumulative carbonemissions Nature 459 829–32
Meinshausen M, Meinshausen N, Hare W, Raper S C B, FrielerK, Knutti R, Frame D J and Allen M R 2009 Greenhouse-gas emission targets for limiting global warming to 2 °CNature 458 1158–62
Mitchell T D 2003 Pattern scaling: an examination of theaccuracy of the technique for describing future climatesClim. Change 60 217–42
National Research Council 2011 Climate Stabilization Targets:Emissions, Concentrations, and Impacts over Decades toMillennia (Washington, DC: The National AcademiesPress)
Rogelj J, Schaeffer M, Friedlingstein P, Gillett N P, van Vuuren DP, Riahi K, Allen M and Knutti R 2016 Differencesbetween carbon budget estimates unravelled Nat. Clim.Change 6 245–52
Santer B D, Wigley M L T, Schlesinger M E and Mitchell J F B1990 Developing Climate Scenarios from Equilibrium GCMResults (Hamburg: Max Planck Institute for Meteorology)
Screen J A and Simmonds I 2010 The central role of diminishingsea ice in recent Arctic temperature amplification Nature464 1334–1337
9
Taylor K E, Stouffer R J and Meehl G A 2012 An overview ofCMIP5 and the experiment design Bull. Am. Meteorol. Soc.93 485–98
Tebaldi C and Arblaster J M 2014 Pattern scaling: its strengthsand limitations, and an update on the latest modelsimulations Clim. Change 122 459–71
van Vuuren D P, van Soest H, Riahi K, Clarke L, Krey V,Kriegler E, Rogelj J, Schaeffer M and Tavoni M 2016Carbon budgets and energy transition pathways Environ.Res. Lett. 11 075002
Wilks D S 2011 Statistical Methods in the Atmospheric Sciences(International Geophysics Series vol 100) 3rd edn(Cambridge, MA: Academic)
Zickfeld K, Eby M, Matthews H D and Weaver A J 2009 Settingcumulative emissions targets to reduce the risk ofdangerous climate change Proc. Natl Acad. Sci. 10616129–34
