Abstract
The usefulness of a set of climate change projections largely depends on how well it spans a range of outcomes consistent with known uncertainties. Here, we present exploratory work towards developing a strategy to select variants of a state-of-the-art but expensive climate model suitable for climate projection studies. The strategy combines information from a set of relatively cheap, idealized perturbed parameter ensemble (PPE) and CMIP5 multi-model ensemble (MME) experiments, and uses two criteria as the basis to select model variants for a PPE suitable for future projections: (a) acceptable model performance at two different timescales, and (b) maintaining diversity in model response to climate change. This second part of a pair of papers builds upon Part I in which we established a strong relationship between model errors at weather and climate timescales across a PPE for a variety of key variables. This relationship is used to filter out parts of parameter space that do not give credible simulations of present day climate, while minimizing the impact on ranges in forcings and feedbacks that drive model responses to climate change. We use statistical emulation to explore the parameter space thoroughly, and demonstrate that about 90% can be filtered out without affecting diversity in global-scale climate change responses. This leads to the identification of plausible parts of parameter space from which model variants can be selected for projection studies. We selected and ran 50 variants from the plausible parameter combinations and validated the emulator predictions. Comparisons with the CMIP5 MME demonstrate that our approach can produce a set of plausible model variants that span a relatively wide range in model response to climate change. We also highlight how the prior expert-specified ranges for uncertain model parameters are constrained as a result of our methodology, and discuss recommendations for future work.
Similar content being viewed by others
References
Adler RF, Huffman GJ, Chang A, Ferraro R, Xie PP, Janowiak J, Rudolf B, Schneider U, Curtis S, Bolvin D (2003) The version-2 global precipitation climatology project (GPCP) monthly precipitation analysis (1979-present). J Hydrometeorol 4(6):1147–1167
Allan R, Ansell T (2006) A new globally complete monthly historical gridded mean sea level pressure dataset (HadSLP2): 1850–2004. J Clim 19(22):5816–5842
Andrews T, Gregory JM, Webb MJ, Taylor KE (2012) Forcing, feedbacks and climate sensitivity in CMIP5 coupled atmosphere-ocean climate models. Geophys Res Lett 39(9):L10702. https://doi.org/10.1029/2012GL051942
Bellouin N, Collins W, Culverwell I, Halloran P, Hardiman S, Hinton T, Jones C, McDonald R, McLaren A, O’Connor F (2011) The HadGEM2 family of Met Office Unified Model climate configurations. Geosci Model Dev 4(3):723–757
Bony S, Colman R, Kattsov VM, Allan RP, Bretherton CS, Dufresne JL, Hall A, Hallegatte S, Holland MM, Ingram W (2006) How well do we understand and evaluate climate change feedback processes? J Clim 19(15):3445–3482
Bony S, Webb M, Stevens B, Bretherton C, Klein S, Tselioudis G (2009) The cloud feedback model intercomparison project: summary of activities and recommendations for advancing assessments of cloud-climate feedbacks. CFMIP Doc
Booth BB, Jones CD, Collins M, Totterdell IJ, Cox PM, Sitch S, Huntingford C, Betts RA, Harris GR, Lloyd J (2012) High sensitivity of future global warming to land carbon cycle processes. Environ Res Lett 7(2):024,002
Brown A, Milton S, Cullen M, Golding B, Mitchell J, Shelly A (2012) Unified modeling and prediction of weather and climate: a 25-year journey. Bull Am Meteorol Soc 93(12):1865–1877
Carslaw K, Lee L, Reddington C, Pringle K, Rap A, Forster P, Mann G, Spracklen D, Woodhouse M, Regayre L et al (2013) Large contribution of natural aerosols to uncertainty in indirect forcing. Nature 503(7474):67–71
Cess RD, Potter G, Blanchet J, Boer G, Del Genio A, Deque M, Dymnikov V, Galin V, Gates W, Ghan S (1990) Intercomparison and interpretation of climate feedback processes in 19 atmospheric general circulation models. J Geophys Res Atmos 95(D10):16,601–16,615
Collins M (2007) Ensembles and probabilities: a new era in the prediction of climate change. Phil Trans R Soc A 365:1957–1970. https://doi.org/10.1098/rsta.2007.2068
Collins M, Booth BB, Harris GR, Murphy JM, Sexton DM, Webb MJ (2006) Towards quantifying uncertainty in transient climate change. Clim Dyn 27(2–3):127–147
Collins M, Brierley C, MacVean M, Booth B, Harris G (2007) The sensitivity of the rate of transient climate change to ocean physics perturbations. J Clim 20(10):2315–2320
Collins M, Booth BB, Bhaskaran B, Harris GR, Murphy JM, Sexton DM, Webb MJ (2011) Climate model errors, feedbacks and forcings: a comparison of perturbed physics and multi-model ensembles. Clim Dyn 36(9–10):1737–1766
Compo GP, Whitaker JS, Sardeshmukh PD, Matsui N, Allan RJ, Yin X, Gleason BE, Vose RS, Rutledge G, Bessemoulin P (2011) The twentieth century reanalysis project. Q J R Meteorol Soc 137(654):1–28
Covey C, Lucas DD, Tannahill J, Garaizar X, Klein R (2013) Efficient screening of climate model sensitivity to a large number of perturbed input parameters. J Adv Model Earth Syst 5(3):598–610
Dee DP, Uppala S, Simmons A, Berrisford P, Poli P, Kobayashi S, Andrae U, Balmaseda M, Balsamo G, Bauer P (2011) The ERA-Interim reanalysis: Configuration and performance of the data assimilation system. Q J R Meteorol Soc 137(656):553–597
Edwards NR, Cameron D, Rougier J (2011) Precalibrating an intermediate complexity climate model. Clim Dyn 37(7–8):1469–1482
Flato G, Marotzke J, Abiodun B, Braconnot P, Chou SC, Collins WJ, Cox P, Driouech F, Emori S, Eyring V, et al (2013) Evaluation of climate models. In: Climate change 2013: the physical science basis. contribution of Working Group I to the fifth assessment report of the intergovernmental panel on climate change. Climate Change 2013 5:741–866
Gates WL, Boyle JS, Covey C, Dease CG, Doutriaux CM, Drach RS, Fiorino M, Gleckler PJ, Hnilo JJ, Marlais SM (1999) An overview of the results of the atmospheric model intercomparison project (AMIP I). Bull Am Meteorol Soc 80(1):29–55
Gleckler PJ, Taylor KE, Doutriaux C (2008) Performance metrics for climate models. J Geophys Res Atmos 113(D6):D06104. https://doi.org/10.1029/2007JD008972
Golaz JC, Larson VE, Hansen JA, Schanen DP, Griffin BM (2007) Elucidating model inadequacies in a cloud parameterization by use of an ensemble-based calibration framework. Mon Weather Rev 135(12):4077–4096
Gong W, Duan Q, Li J, Wang C, Di Z, Dai Y, Ye A, Miao C (2015) Multi-objective parameter optimization of common land model using adaptive surrogate modeling. Hydrol Earth Syst Sci 19(5):2409–2425
Gregoire LJ, Valdes PJ, Payne AJ, Kahana R (2011) Optimal tuning of a GCM using modern and glacial constraints. Clim Dyn 37(3–4):705–719
Gregory J, Webb M (2008) Tropospheric adjustment induces a cloud component in CO2 forcing. J Clim 21(1):58–71
Harris G, Collins M, Sexton D, Murphy J, Booth B (2010) Probabilistic projections for 21st century european climate. Nat Hazards Earth Syst Sci 10(9):2009–2020
Harris GR, Sexton DM, Booth BB, Collins M, Murphy JM (2013) Probabilistic projections of transient climate change. Clim Dyn 40(11–12):2937–2972
Hazeleger W, Van den Hurk B, Min E, Van Oldenborgh G, Petersen A, Stainforth D, Vasileiadou E, Smith L (2015) Tales of future weather. Nat Clim Change 5(2):107–113
Hewitt H, Copsey D, Culverwell I, Harris C, Hill R, Keen A, McLaren A, Hunke E (2011) Design and implementation of the infrastructure of HadGEM3: The next-generation Met Office climate modelling system. Geosci Model Dev 4(2):223–253
Hourdin F, Mauritsen T, Gettelman A, Golaz JC, Balaji V, Duan Q, Folini D, Ji D, Klocke D, Qian Y (2017) The art and science of climate model tuning. Bull Am Meteorol Soc 98(3):589–602
Huffman GJ, Bolvin DT, Nelkin EJ, Wolff DB, Adler RF, Gu G, Hong Y, Bowman KP, Stocker EF (2007) The TRMM multisatellite precipitation analysis (TMPA): Quasi-global, multiyear, combined-sensor precipitation estimates at fine scales. J Hydrometeorol 8(1):38–55
Jackson CS, Sen MK, Huerta G, Deng Y, Bowman KP (2008) Error reduction and convergence in climate prediction. J Clim 21(24):6698–6709
Jackson LC, Vellinga M, Harris GR (2012) The sensitivity of the meridional overturning circulation to modelling uncertainty in a perturbed physics ensemble without flux adjustment. Clim Dyn 39(1–2):277–285
Klocke D, Rodwell M (2014) A comparison of two numerical weather prediction methods for diagnosing fast-physics errors in climate models. Q J R Meteorol Soc 140(679):517–524
Klocke D, Pincus R, Quaas J (2011) On constraining estimates of climate sensitivity with present-day observations through model weighting. J Clim 24(23):6092–6099
Knutti R, Furrer R, Tebaldi C, Cermak J, Meehl GA (2010) Challenges in combining projections from multiple climate models. J Clim 23(10):2739–2758
Lambert FH, Harris GR, Collins M, Murphy JM, Sexton DM, Booth BB (2013) Interactions between perturbations to different earth system components simulated by a fully-coupled climate model. Clim Dyn 41(11–12):3055–3072
Lee L, Carslaw K, Pringle K, Mann G, Spracklen D (2011) Emulation of a complex global aerosol model to quantify sensitivity to uncertain parameters. Atmos Chem Phys 11(23):12,253–12,273
Linz M, Tziperman E, MacMartin DG (2014) Process-based analysis of climate model ENSO simulations: intermodel consistency and compensating errors. J Geophys Res Atmos 119(12):7396–7409
Loeb NG, Wielicki BA, Doelling DR, Smith GL, Keyes DF, Kato S, Manalo-Smith N, Wong T (2009) Toward optimal closure of the Earth’s top-of-atmosphere radiation budget. J Clim 22(3):748–766
Loeppky JL, Sacks J, Welch WJ (2009) Choosing the sample size of a computer experiment: a practical guide. Technometrics 51(4):366–376
Lucas D, Klein R, Tannahill J, Ivanova D, Brandon S, Domyancic D, Zhang Y (2013) Failure analysis of parameter-induced simulation crashes in climate models. Geosci Model Dev 6(4):1157–1171
Ma HY, Xie S, Klein S, Williams K, Boyle J, Bony S, Douville H, Fermepin S, Medeiros B, Tyteca S (2014) On the correspondence between mean forecast errors and climate errors in CMIP5 models. J Clim 27(4):1781–1798
Martin G, Milton S, Senior C, Brooks M, Ineson S, Reichler T, Kim J (2010) Analysis and reduction of systematic errors through a seamless approach to modeling weather and climate. J Clim 23(22):5933–5957
Mauritsen T, Stevens B, Roeckner E, Crueger T, Esch M, Giorgetta M, Haak H, Jungclaus J, Klocke D, Matei D et al (2012) Tuning the climate of a global model. J Adv Model Earth Syst 4:M00A01. https://doi.org/10.1029/2012MS000154
McKay MD, Beckman RJ, Conover WJ (1979) Comparison of three methods for selecting values of input variables in the analysis of output from a computer code. Technometrics 21(2):239–245
McNeall D, Williams J, Booth B, Betts R, Challenor P, Wiltshire A, Sexton D (2016) The impact of structural error on parameter constraint in a climate model. Earth Syst Dyn 7:917–935. https://doi.org/10.5194/esd-7-917-2016
Meehl GA, Covey C, Taylor KE, Delworth T, Stouffer RJ, Latif M, McAvaney B, Mitchell JF (2007) The WCRP CMIP3 multimodel dataset: a new era in climate change research. Bull Am Meteorol Soc 88(9):1383–1394
Min SK, Simonis D, Hense A (2007) Probabilistic climate change predictions applying bayesian model averaging. Philos Trans R Soc Lond A 365(1857):2103–2116
Murphy JM, Sexton DM, Barnett DN, Jones GS, Webb MJ, Collins M, Stainforth DA (2004) Quantification of modelling uncertainties in a large ensemble of climate change simulations. Nature 430(7001):768–772
Murphy JM, Booth BB, Collins M, Harris GR, Sexton DM, Webb MJ (2007) A methodology for probabilistic predictions of regional climate change from perturbed physics ensembles. Philos Trans R Soc Lond A 365(1857):1993–2028
Murphy J, Sexton D, Jenkins G, Boorman P, Booth B, Brown C, Clark R, Collins M, Harris G, Kendon E (2009) UK climate projections science report: UKCP09. Met Office Hadley Centre, Exeter
Murphy JM, Booth BB, Boulton CA, Clark RT, Harris GR, Lowe JA, Sexton DM (2014) Transient climate changes in a perturbed parameter ensemble of emissions-driven earth system model simulations. Clim Dyn 43(9–10):2855–2885
Neelin JD, Bracco A, Luo H, McWilliams JC, Meyerson JE (2010) Considerations for parameter optimization and sensitivity in climate models. Proc Natl Acad Sci 107(50):21,349–21,354
Olson R, Sriver R, Chang W, Haran M, Urban N, Keller K (2013) What is the effect of unresolved internal climate variability on climate sensitivity estimates? J Geophys Res Atmos 118(10):4348–4358
Phillips TJ, Potter GL, Williamson DL, Cederwall RT, Boyle JS, Fiorino M, Hnilo JJ, Olson JG, Xie S, Yio JJ (2004) Evaluating parameterizations in general circulation models: climate simulation meets weather prediction. Bull Am Meteorol Soc 85(12):1903–1915
Piani C, Frame D, Stainforth D, Allen M (2005) Constraints on climate change from a multi-thousand member ensemble of simulations. Geophys Res Lett 32(23):L23825. https://doi.org/10.1029/2005GL024452
Räisänen J, Palmer T (2001) A probability and decision-model analysis of a multimodel ensemble of climate change simulations. J Clim 14(15):3212–3226
Reichler T, Kim J (2008) How well do coupled models simulate today’s climate? Bull Am Meteorol Soc 89(3):303–311
Rienecker MM, Suarez MJ, Gelaro R, Todling R, Bacmeister J, Liu E, Bosilovich MG, Schubert SD, Takacs L, Kim GK (2011) MERRA: NASA’s modern-era retrospective analysis for research and applications. J Clim 24(14):3624–3648
Ringer MA, Andrews T, Webb MJ (2014) Global-mean radiative feedbacks and forcing in atmosphere-only and coupled atmosphere-ocean climate change experiments. Geophys Res Lett 41(11):4035–4042
Rodwell M, Palmer T (2007) Using numerical weather prediction to assess climate models. Q J R Meteorol Soc 133(622):129–146
Ross SM (2003) Peirce’s criterion for the elimination of suspect experimental data. J Eng Technol 20(2):38–41
Rossow WB, Schiffer RA (1999) Advances in understanding clouds from ISCCP. Bull Am Meteorol Soc 80(11):2261–2287
Rougier J (2007) Probabilistic inference for future climate using an ensemble of climate model evaluations. Clim Change 81(3):247–264
Rougier JC, Sexton DMH (2007) Inference in ensemble experiments. Phil Trans R Soc A 365:2133–2143. https://doi.org/10.1098/rsta.2007.2071
Rougier J, Sexton DM, Murphy JM, Stainforth D (2009) Analyzing the climate sensitivity of the hadsm3 climate model using ensembles from different but related experiments. J Clim 22(13):3540–3557
Rousseeuw PJ (1987) Silhouettes: a graphical aid to the interpretation and validation of cluster analysis. J Comput Appl Math 20:53–65
Roustant O, Ginsbourger D, Deville Y (2012) Dicekriging, Diceoptim: Two R packages for the analysis of computer experiments by kriging-based metamodelling and optimization. J Stat Softw 51(1):54p
Rowell DP (1998) Assessing potential seasonal predictability with an ensemble of multidecadal GCM simulations. J Clim 11(2):109–120
Rowlands DJ, Frame DJ, Ackerley D, Aina T, Booth BB, Christensen C, Collins M, Faull N, Forest CE, Grandey BS (2012) Broad range of 2050 warming from an observationally constrained large climate model ensemble. Nat Geosci 5(4):256–260
Saltelli A, Chan K, Scott EM et al (2000) Sensitivity analysis, vol 1. Wiley, New York
Scaife AA, Copsey D, Gordon C, Harris C, Hinton T, Keeley S, O’Neill A, Roberts M, Williams K (2011) Improved atlantic winter blocking in a climate model. Geophys Res Lett 38(23):L23703. https://doi.org/10.1029/2011GL049573
Schmidt GA, Bader D, Donner LJ, Elsaesser GS, Golaz JC, Hannay C, Molod A, Neale RB, Saha S (2017) Practice and philosophy of climate model tuning across six US modeling centers. Geosci Model Dev 10(9):3207
Sexton DM, Murphy JM (2012) Multivariate probabilistic projections using imperfect climate models. Part ii: robustness of methodological choices and consequences for climate sensitivity. Clim Dyn 38(11–12):2543–2558
Sexton DM, Murphy JM, Collins M, Webb MJ (2012) Multivariate probabilistic projections using imperfect climate models part i: outline of methodology. Clim Dyn 38(11–12):2513–2542
Sexton DMH, Karmalkar AV, Murphy JM, Williams KD, Boutle IA, Morcrette CJ, Stirling AJ, Vosper SB (2019) Finding plausible and diverse variants of a climate model. Part I: establishing the relationship between errors at weather and climate time scales. Clim Dyn. https://doi.org/10.1007/s00382-019-04625-3
Shindell DT, Lamarque JF, Schulz M, Flanner M, Jiao C, Chin M, Young P, Lee YH, Rotstayn L, Mahowald N (2013) Radiative forcing in the accmip historical and future climate simulations. Atmos Chem Phys 13(6):2939–2974
Shiogama H, Watanabe M, Yoshimori M, Yokohata T, Ogura T, Annan JD, Hargreaves JC, Abe M, Kamae Y, Oishi R (2012) Perturbed physics ensemble using the miroc5 coupled atmosphere-ocean gcm without flux corrections: experimental design and results. Clim Dyn 39(12):3041–3056
Stainforth DA, Aina T, Christensen C, Collins M, Faull N, Frame DJ, Kettleborough JA, Knight S, Martin A, Murphy J et al (2005) Uncertainty in predictions of the climate response to rising levels of greenhouse gases. Nature 433(7024):403–406
Suzuki K, Golaz JC, Stephens GL (2013) Evaluating cloud tuning in a climate model with satellite observations. Geophys Res Lett 40(16):4464–4468
Taylor KE, Stouffer RJ, Meehl GA (2012) An overview of CMIP5 and the experiment design. Bull Am Meteorol Soc 93(4):485–498
Tebaldi C, Knutti R (2007) The use of the multi-model ensemble in probabilistic climate projections. Philos Trans R Soc Lond A 365(1857):2053–2075
Tett SF, Mitchell JF, Parker DE, Allen MR (1996) Human influence on the atmospheric vertical temperature structure: Detection and observations. Science 274(5290):1170–1173
Tokmakian R, Challenor P (2014) Uncertainty in modeled upper ocean heat content change. Clim Dyn 42(3–4):823–842
Walters D, Williams K, Boutle I, Bushell A, Edwards J, Field P, Lock A, Morcrette C, Stratton R, Wilkinson J et al (2014) The Met Office Unified Model global atmosphere 4.0 and JULES global land 4.0 configurations. Geosci Model Dev 7(1):361–386
Webb MJ, Lock AP (2013) Coupling between subtropical cloud feedback and the local hydrological cycle in a climate model. Clim Dyn 41(7–8):1923–1939
Webb MJ, Senior C, Sexton D, Ingram W, Williams K, Ringer M, McAvaney B, Colman R, Soden B, Gudgel R (2006) On the contribution of local feedback mechanisms to the range of climate sensitivity in two GCM ensembles. Clim Dyn 27(1):17–38
Webb MJ, Lambert FH, Gregory JM (2013) Origins of differences in climate sensitivity, forcing and feedback in climate models. Clim Dyn 40(3–4):677–707
Williams K, Bodas-Salcedo A, Déqué M, Fermepin S, Medeiros B, Watanabe M, Jakob C, Klein S, Senior C, Williamson D (2013) The Transpose-AMIP II experiment and its application to the understanding of southern ocean cloud biases in climate models. J Clim 26(10):3258–3274
Williamson D, Goldstein M, Allison L, Blaker A, Challenor P, Jackson L, Yamazaki K (2013) History matching for exploring and reducing climate model parameter space using observations and a large perturbed physics ensemble. Clim Dyn 41(7–8):1703–1729
Williamson D, Blaker AT, Hampton C, Salter J (2015) Identifying and removing structural biases in climate models with history matching. Clim Dyn 45(5–6):1299–1324
Williamson DB, Blaker AT, Sinha B (2017) Tuning without over-tuning: parametric uncertainty quantification for the NEMO ocean model. Geosci Model Dev 10(4):1789
Willmott CJ, Matsuura K (2005) Advantages of the mean absolute error (MAE) over the root mean square error (RMSE) in assessing average model performance. Clim Res 30(1):79–82
Xie P, Arkin PA (1996) Analyses of global monthly precipitation using gauge observations, satellite estimates, and numerical model predictions. J Clim 9(4):840–858
Xie S, Ma HY, Boyle JS, Klein SA, Zhang Y (2012) On the correspondence between short-and long-time-scale systematic errors in CAM4/CAM5 for the year of tropical convection. J Clim 25(22):7937–7955
Yamazaki K, Rowlands DJ, Aina T, Blaker AT, Bowery A, Massey N, Miller J, Rye C, Tett SF, Williamson D (2013) Obtaining diverse behaviors in a climate model without the use of flux adjustments. J Geophys Res Atmos 118(7):2781–2793
Yokohata T, Webb MJ, Collins M, Williams KD, Yoshimori M, Hargreaves JC, Annan JD (2010) Structural similarities and differences in climate responses to CO2 increase between two perturbed physics ensembles. J Clim 23(6):1392–1410
Zelinka MD, Andrews T, Forster PM, Taylor KE (2014) Quantifying components of aerosol-cloud-radiation interactions in climate models. J Geophys Res Atmos 119(12):7599–7615
Zhang X, Zhang S, Liu Z, Wu X, Han G (2015) Parameter optimization in an intermediate coupled climate model with biased physics. J Clim 28(3):1227–1247
Acknowledgements
This research was supported by the Joint UK BEIS/Defra Met Office Hadley Centre Climate Programme (GA01101). John Rostron was supported by the UK-China Research & Innovation Partnership Fund through the Met Office Climate Science for Service Partnership (CSSP) China as part of the Newton Fund. We would like to thank Mark Webb, Mark Ringer, and Alejandro Bodas-Salcedo for their help with designing and setting up the experiments.
Author information
Authors and Affiliations
Corresponding author
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations'.
Appendices
Appendix 1: experimental design details
-
(i)
Transpose-AMIP (TAMIP)
Our design for this experiment is based on the second phase of Transpose-Atmospheric Model Intercomparison Project (TAMIP-II) activity as part the fifth phase of Coupled Model Intercomparison Project (CMIP5)(http://www.transpose-amip.info; Williams et al. 2013). The TAMIP-II simulations are 5 days in length and the state variables are initialized from 16 Met Office Analyses (global analyses of observations using the Met Office Numerical Weather Prediction Model). The 16 analyses were sampled by picking four dates from each season between October 2008 and August 2009, and exploring the diurnal cycle within each season by sampling start times of 000Z, 0600Z, 1200Z, and 1800Z once. This approach allows a relatively inexpensive examination of error growth from observed initial conditions on relatively short time scales (see Sect. 2.1), and facilitates evaluation of the processes responsible for the origin of model biases in our perturbed climate model variants. Since the atmospheric initial states did not include aerosol information (see Sect. 3), aerosols were represented in the TAMIP experiments by the use of a simple prescribed climatology from HadGEM2-ES (Collins et al. 2011) whereas the other experiments all use an interactive scheme predicting aerosol concentrations from precursor emissions.
-
(ii)
ATMOS
ATMOS runs are 10 years long (1999–2008) and are used to assess model performance on climate time scales. We chose this recent period to provide maximum overlap with observational datasets. Since the vegetation response to changing climate conditions is relatively slow (at least 20–30 years), we decided not to use a dynamic vegetation scheme in these simulations, instead prescribing the same observed vegetation distribution used in previous HadGEM2 simulations lacking an interactive representation of the earths carbon cycle. The statistical method used to quantify model error in this study requires an estimate of the component of uncertainty due to internal variability (Sect. 5). This was obtained by running 16 ATMOS simulations using the standard version of HadGEM3-A, from 16 quasi-independent initial conditions obtained from different model variants in our ensemble.
-
(iii)
ATMOS + \(4\times \hbox {CO}_2\)
This experiment has the same set-up as the ATMOS experiment above but is run for one year with instantaneous quadrupling of \(\hbox {CO}_2\). It is designed to estimate \(\hbox {CO}_2\) forcing in individual model variants, and its uncertainties across the PPE. The \(4 \times \hbox {CO}_2\) forcing is the change in the net TOA radiation in this run relative to the first year of the ATMOS run. This experimental setup is the same as that used in CFMIP simulations included in CMIP5 (Expt 6.5, see http://cmip-pcmdi.llnl.gov/cmip5/docs/Taylor_CMIP5_design.pdf). Our results can therefore be compared with those from these other models, with the caveat that our forcing estimates for individual PPE variants are subject to larger uncertainty from internal variability than those for individual CMIP5 models, since the latter use the longer 30 year simulation period recommended in the CFMIP/CMIP5 experimental design. However, this sampling issue is expected to play a smaller role in determining the overall range of PPE estimates, facilitating comparison against the CMIP5 range (see discussion of Fig. 5 in Sect. 4.2).
-
(iv)
ATMOS + patterned + 4K SST (amipFuture)
This experiment has the same set-up as the ATMOS experiment above but is run for 5 years with a prescribed + 4K SST pattern (Expt 6.6 in http://cmip-pcmdi.llnl.gov/cmip5/docs/Taylor_CMIP5_design.pdf, referred to as `amipFuture'). This experiment is used to examine the strength of climate feedbacks and associated uncertainties, estimated by prescribing climate change through changes in SST (see Webb and Lock 2013). The imposed SST pattern is obtained from coupled model simulations of 13 CMIP3 AOGCMs at the time of CO2 quadrupling from simulations driven by a 1% per year increase in CO2 concentration. Sea-ice extents are not perturbed in this setup. Feedbacks are quantified as the difference between net TOA radiation in the + 4K SST experiment and the corresponding ATMOS experiment, divided by the surface temperature response. The corresponding CMIP5 simulations were 30 years in length, so as in (iii) above. Internal variability is expected to contribute a larger sampling uncertainty to feedback estimates in individual PPE simulations. However, the impact on the cross-ensemble range, in comparison to that of CMIP5 (see Fig. 5 and Sect. 4.2), is expected to be smaller.
-
(v)
ATMOS + preindustrial aerosols
The ATMOS experiments in (ii)–(iv) above include the CLASSIC aerosol module of Bellouin et al. (2011), which predicts aerosol concentrations of certain species (sulphate, soot, biomass, organic carbon) interactively, from emissions of their natural and anthropogenic precursors. In addition, we ran 1-year pre-industrial aerosol simulations including only natural (i.e. 1860) emissions, with an experimental setup otherwise identical to the simulation using the standard ATMOS protocol of (ii). This experiment allowed us to obtain a rough estimate of anthropogenic aerosol forcing and associated uncertainties by computing changes in net TOA radiation in the run with pre-industrial emissions, relative to the first year of the corresponding ATMOS run from (ii).
Appendix 2: variables and verification datasets
Verification datasets used to assess TAMIP and ATMOS performance for the variables listed include gridded station and satellite observations and reanalysis data and are tabulated in Table 3. In each case, we selected one verification dataset (Obs 1) to calculate model performance scores, and used the second one (Obs 2; if available) to determine an estimate of observational error.
Appendix 3: leave-one-out (LOO) validation for TAMIP and ATMOS emulators
The leave-one-out (LOO) cross validation scores for TAMIP and ATMOS emulators for the six assessment variables are shown in Figs. 15 and 16, respectively. We use Gaussian Process emulators (see Part I for more information) as described in Lee et al. (2011) but the emulators are built for global MSEs of individual variables for TAMIP and ATMOS separately instead of emulating individual grid points. The \(\hbox {R}^2\)-values of the LOO estimates plotted against the values of the left-out member are at least 0.7 for all TAMIP MSEs, and at least 0.6 for the ATMOS MSEs. Overall, high number of model failures in the ATMOS case due to instabilities in HadGEM3-GA4.0 meant limited training data (80) available for building ATMOS emulators. As a result, the ATMOS emulators are clearly less reliable than TAMIP. The main issue is a poor fit for high MSEs for pr, ta850, and ua250, perhaps due to a lack of training data at the high end. Also note that one of the outliers in these plots is from the same member of the ensemble which had low values for all the parameters perturbed and high MSEs. Another possibility is that we define the Gaussian Process to have a linear term and correlation length scale for each parameter, and this means a large number of parameters relative to the ensemble size. Originally, we used these emulators for both parts I and II to select 50 parameter combinations and run the validation ensemble. Since then, we improved the method to build the emulator by reducing the number of linear terms using a stepwise algorithm prior to fitting the Gaussian Process; potentially a similar reduction could be used for the correlation terms. In part I we have used the improved emulators but have retained the original emulators here for consistency with the validation ensemble.
Whilst the fit for these high MSEs for pr, ta850, and ua250 is poor in absolute terms, it does not affect the ranking too much. For example, five of the six members with the worst precipitation MSEs in the ATMOS experiment are predicted to have high MSEs by the emulator. These five would still be ruled out by the filtering. The other member of these six is predicted to have a moderate MSE value and may well have been accepted by the filtering. It may then have been selected to be run as an ATMOS experiment, then evaluated as a poor performer, and so not made it to the stage where it would be run in coupled mode. This is not ideal because it would have wasted a run of the ATMOS experiment. If we had been intending to run a coupled ensemble, rather than simply demonstrate the selection algorithm, then we would have had to have run a larger ATMOS ensemble to build better emulators and reduce the risk of this happening.
Rights and permissions
About this article
Cite this article
Karmalkar, A.V., Sexton, D.M.H., Murphy, J.M. et al. Finding plausible and diverse variants of a climate model. Part II: development and validation of methodology. Clim Dyn 53, 847–877 (2019). https://doi.org/10.1007/s00382-019-04617-3
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00382-019-04617-3