Federal Department of Home Affairs FDHA Federal Office

Federal Department of Home Affairs FDHA Federal Office of Meteorology and Climatology Meteo. Swiss Climate Change Projections for Switzerland: A Bayesian Multi-Model Combination using ENSEMBLES Regional Climate Models Andreas Fischer, Andreas Weigel, Mark Liniger, Christoph Buser, Christof Appenzeller 11 th International Meeting on Statistical Climatology, 12 July 2010, Edinburgh

ENSEMBLES R 2 TB 8 AOGCMs / 21 Model Chains AOGCMs BCM Standard sens. Had. CM 3 High sens. SRES A 1 B Low sens. ECHAM 5 ARPEGE CGCM 3 IPSL 6 AOGCMs / 15 Model Chains RCMs@25 km HIRHAM (DMI) HIRHAM (Met. No) RCA (SMHI) CLM (ETHZ) PROMES (UCLM) RRCM (VMGO) Had. RM 3 (Met Office) HIRHAM (Met. No) RCA 3 (C 4 I) Had. RM 3 (Met Office) RCA (SMHI) REMO (MPI) HIRHAM (DMI) RACMO (KNMI) RCA (SMHI) REGCM 3(ICTP) ALADIN v 1 (CNRM) ALADIN v 2 (CNRM) HIRHAM (DMI) CRCM (OURANOS) CLM (GKSS) 1950 - 2050 Climate Services, IMSC Edinburgh | 12 July 2010 andreas. fischer@meteoswiss. ch Final Report (2009) AOGCMs BCM Standard sens. Had. CM 3 RCMs@25 km HIRHAM (DMI) RCA (SMHI) CLM (ETHZ) Had. RM 3 (Met Office) High sens. RCA 3 (C 4 I) Had. RM 3 (Met Office) Low sens. Had. RM 3 (Met Office) RCA (SMHI) ECHAM 5 REMO (MPI) HIRHAM (DMI) RACMO (KNMI) RCA (SMHI) REGCM 3(ICTP) ARPEGE ALADIN v 2 (CNRM) HIRHAM (DMI) 2050 - 2100 2

Derivation of Probablistic Scenarios Modelled Climate Change Signals Bayes Algorithm PDF ? (Buser et al. , 2009) Assumptions transparent Climate Services, IMSC Edinburgh | 12 July 2010 andreas. fischer@meteoswiss. ch 3

Bayesian Multi-Model Combination (Buser et al. , 2009) Obs NOW Models NOW „Obs“ FUTURE Models FUTURE Seasonally averaged 30 yr time periods Climate Services, IMSC Edinburgh | 12 July 2010 andreas. fischer@meteoswiss. ch 4

Bayesian Multi-Model Combination (Buser et al. , 2009) Obs NOW Models NOW „Obs“ FUTURE Models FUTURE Mean Climate Shift NOW Climate Services, IMSC Edinburgh | 12 July 2010 andreas. fischer@meteoswiss. ch Model Projection Errors FUTURE 5

Bayesian Multi-Model Combination (Buser et al. , 2009) Obs NOW Models NOW „Obs“ FUTURE Models FUTURE Mean Climate Shift Model Projection Errors • μ and βi non identifiable • Assumption has to be taken about projection error Δβi ~ N(0; σ2 β) NOW Climate Services, IMSC Edinburgh | 12 July 2010 andreas. fischer@meteoswiss. ch FUTURE 6

Bayesian Multi-Model Combination (Buser et al. , 2009) Obs NOW Models NOW „Obs“ FUTURE Prior p(x) Models FUTURE Posterior p(x|data) P(x|data) p(x) * p(data|x) Gibbs Sampler Likelihood p(data|x) Climate Services, IMSC Edinburgh | 12 July 2010 andreas. fischer@meteoswiss. ch 7

Sensitivity Experiments: Effect of Likelihood All prior distributions set to be uninformative Likelihood affects variance and location of posterior distribution Climate Change Signal Climate Services, IMSC Edinburgh | 12 July 2010 andreas. fischer@meteoswiss. ch 8

Bayesian Multi-Model Combination (Buser et al. , 2009) Obs NOW Models NOW „Obs“ FUTURE Prior p(x) Models FUTURE Posterior p(x|data) P(x|data) p(x) * p(data|x) Gibbs Sampler Likelihood p(data|x) Climate Services, IMSC Edinburgh | 12 July 2010 andreas. fischer@meteoswiss. ch 9

Mean Climate Shift Sensitivity Experiments: Effect of Prior The uncertainty in Δμ is increased with a wider priorsetting for Δβi Climate Services, IMSC Edinburgh | 12 July 2010 andreas. fischer@meteoswiss. ch Projection Uncertainty 10

Sensitivity Experiments: Effect of Prior Outlier CC Signal Informative Prior Δβi CC Signal Central tendency of posterior Nondistributions also affected by Informative Prior Δβi prior Climate Services, IMSC Edinburgh | 12 July 2010 andreas. fischer@meteoswiss. ch CC Signal 11

Application of Algorithm using ENSEMBLES data Different considerations: 1. Estimation of Projection Uncertainty (σ2 β) 2. Role of Internal Variability 3. Independent Model Data Climate Services, IMSC Edinburgh | 12 July 2010 andreas. fischer@meteoswiss. ch 12

1. Estimating Projection Uncertainty Assumption: Projection Uncertainty is fully sampled by range of available model simulations (1) GCM Uncertainty 8 different GCMs (2) RCM Uncertainty Had. CM 3 Q 0 ECHAM Smoothing of timeseries by polynomial fit (Hawkins & Sutton, 2009) Climate Services, IMSC Edinburgh | 12 July 2010 andreas. fischer@meteoswiss. ch 13

2. Internal Variability (1) As a pre-processing step we remove internal variability from time-series (2) Calculate posterior distributions with Bayes Algorithm (3) Add internal variability to posterior distribution of μ Climate Services, IMSC Edinburgh | 12 July 2010 andreas. fischer@meteoswiss. ch 14

2. Internal Variability Summer Temperature over CHNE (Model: ETHZ – Had. CM 3 Q 0) 4 th order polynomial fit 30 -yr Running Mean (Hawkins and Sutton, 2009) Climate Services, IMSC Edinburgh | 12 July 2010 andreas. fischer@meteoswiss. ch 15

2. Internal Variability Summer Temperature over CHNE (Model: ETHZ – Had. CM 3 Q 0) 4 th order polynomial fit 30 -yr Running Mean (Hawkins and Sutton, 2009) Climate Services, IMSC Edinburgh | 12 July 2010 andreas. fischer@meteoswiss. ch 16

3. Independent Model Data DJF Temperature 1980 -2009 (AL) ECHAM Had. Q 0 Had. Q 3 HQ 16 ARP. BCM Average all RCMs driven by the same GCM ECHAM Had. CM 3 Q 0 Climate Services, IMSC Edinburgh | 12 July 2010 andreas. fischer@meteoswiss. ch 17

Probabilistic Climate Change Scenarios Northeastern Switzerland Reference Period 1980 - 2009 Orography of Switzerland Climate Services, IMSC Edinburgh | 12 July 2010 andreas. fischer@meteoswiss. ch 18

Swiss Climate Scenario (A 1 B) Temperature (K) 2035 2060 2084 Internal Variability Climate Services, IMSC Edinburgh | 12 July 2010 andreas. fischer@meteoswiss. ch GCM groups GCM-RCM chains 19

Swiss Climate Scenario (A 1 B) Relative Precipitation 2035 2060 2084 GCM groups GCM-RCM chains Internal Variability Climate Services, IMSC Edinburgh | 12 July 2010 andreas. fischer@meteoswiss. ch 20

Conclusions The Bayes Algorithm by Buser et al. (2009) is a transparent tool for generating probabilistic climate change scenarios. The uncertainty range in the climate change signal is highly dependent on the prior-settings of the projection uncertainty. The Buser Algorithm does not account for internal variability. To circumvent this problem a pragmatic solution has been proposed. The probabilistic climate change scenarios for Northeastern Switzerland show a continous increase in temperature over the 21 st century. For precipitation only in summer a signal in the second half of the century is detectable. Climate Services, IMSC Edinburgh | 12 July 2010 andreas. fischer@meteoswiss. ch 21

Swiss Climate Scenarios: Precipitation 2035 2060 2084 DJF Precipitation Change [%] 2035 2060 2084 JJA Precipitation Change [%] Climate Services, IMSC Edinburgh | 12 July 2010 andreas. fischer@meteoswiss. ch 22

Effect correlated models Delta Mu JJA T 2 CHN ECHAM av. / Had. Q 0 av. / Had. Q 3 av. / Had. Q 16 av. / CNRM-ARPEGE / BCM av. / OURANOS Average of 1 GCM group / Rest as standard Climate Services, IMSC Edinburgh | 12 July 2010 andreas. fischer@meteoswiss. ch 23 KNMI-ECHAM / ETHZ-Had. Q 0 / SMHI-Had. Q 3 / C 4 I-Had. Q 16 / CNRM-ARPEGE / SMHI-BCM / OURANOS

Climate Scenarios Global Mean Temperature wrt 1980 -2009 ? A 2 A 1 B [K] B 1 comm 2035 Climate Services, IMSC Edinburgh | 12 July 2010 andreas. fischer@meteoswiss. ch 2060 2084 24

Pattern Scaling with CMIP A 1 B Bayes Estimate 2035 Scaled from 2060 Scaled from 2084 Temperature Climate Services, IMSC Edinburgh | 12 July 2010 andreas. fischer@meteoswiss. ch Relative Precipitation 25

Swiss Climate Scenarios A 1 B B 1 Climate Services, IMSC Edinburgh | 12 July 2010 andreas. fischer@meteoswiss. ch A 2 26

Aim: Update of Probabilistic Scenarios for Northern and Southern Switzerland based on PRUDENCE RCM simulations Temperature Relative Precipitation Oc. CC (2007) 2030 2050 2070 Climate Services, IMSC Edinburgh | 12 July 2010 andreas. fischer@meteoswiss. ch 27

Temperature (°C) Model validation (1980 – 2009) EOBS v 3 Orography CHNE CHW Precipiation (mm/mt) CHS Temperature (°C) EOBS v 3 Climate Services, IMSC Edinburgh | 12 July 2010 andreas. fischer@meteoswiss. ch 28