UPPLEMENT CMIP5 CLIMATE MODEL ANALYSES Climate Extremes in the United States by Donald Wuebbles, Gerald Meehl, K atharine Hayhoe, Thomas R. K arl, Kenneth Kunkel, Benjamin Santer, Michael Wehner, Brian Colle, Erich M. Fischer, Rong Fu, Alex Goodman, Emily Janssen, Viatcheslav Kharin, Huikyo Lee, Wenhong Li, Lindsey N. Long, Seth C. Olsen, Z aitao Pan, Anji Seth, Justin Sheffield, and Liqiang Sun
This document is a supplement to “CMIP5 Climate Model Analyses: Climate Extremes in the United States,” by Donald Wuebbles, Gerald Meehl, Katharine Hayhoe, Thomas R. Karl, Kenneth Kunkel, Benjamin Santer, Michael Wehner, Brian Colle, Erich M. Fischer, Rong Fu, Alex Goodman, Emily Janssen, Viatcheslav Kharin, Huikyo Lee, Wenhong Li, Lindsey N. Long, Seth C. Olsen, Zaitao Pan, Anji Seth, Justin Sheffield, and Liqiang Sun (Bull. Amer. Meteor. Soc., 95, 571–583) • ©2014 American Meteorological Society • Corresponding author: Donald J.Wuebbles, Department of Atmospheric Sciences, University of Illinois, Urbana, IL 61801 • E-mail:
[email protected] • DOI:10.1175/BAMS-D-12-00172.2
MODELS USED IN THE ANALYSES. Individual studies were done by different coauthors and use different choices of which models were included in the analysis [at least in part because of availability at the time the analysis was done; descriptions of the phase 5 of the Coupled Model Intercomparison Project (CMIP5) models can be found at http://cmip-pcmdi .llnl.gov/cmip5/availability.html]. The particular models used for the analyses (and full model expansions) and associated figures in the paper are shown in Table ES1 and, to the degree possible, summarized below: Figure 2: See Table ES1 (because of the complexity of this figure, the models are only shown in the table). Figures 3 and 7: The number of model simulations available for each model and experiment. Model output is generally available for years 1850–2005 for the historical experiment and for years 2006–2100 (or 2099) for the RCP experiments, unless explicitly specified in brackets. Shown in Table ES2 is the number of ensembles used from each model. Figure 4: The first and second numbers in parenthesis are the numbers of ensemble members in the historical run and RCP4.5 run respectively: ACCESS1.0 (1,1), BCC_CSM1.1 (3,1), BNU-ESM (3,1), CanESM2 (5,1), CCSM4 (6,5), CNRM-CM5 (8,1), CSIRO Mk3.6.0 (10,10), FGOALS-S2.0 (3,0), AMERICAN METEOROLOGICAL SOCIETY
GFDL-CM3 (4,0), GFDL-CM3 (1,0), GFDL-ESM2M (1,1), GISS-E2H (15,0), GISS-E2-R (16,15), HadCM3 (1,0), HadGEM2-AO (1,1), HadGEM2-CC (1,1), HadGEM2-ES (1,4), INM-CM4.0 (0,1), IPSL-CM5ALR (5,4), IPSL-CM5A-MR (2,1), IPSL-CM5b-LR (0,1), MIROC5 (4,3), MIROC-ESM (1,1), MIROC-ESMCHEM (1,1), MPI-ESM-LR (3,3), MPI-ESM-P (2,0), MRI-CGCM3 (4,1), and NorESM1-ME (1,1). Figure 5: (top) Historical simulations: IPSLCM5A-LR, GISS-E2-R, IPSL-CM5A-MR, CESM1FASTCHEM, MPI-ESM-MR, MIROC5, MR ICGCM3, CMCC-CM, INM-CM4.0, BCC_CSM1.1, CSIRO Mk3.6.0, MPI-ESM-LR, IPSL-CM5BLR, NorESM1-M, MPI-ESM-P, CNR M-CM5, ACCESS1.0, CCSM4, MIROC-ESM-CHEM, GFDLESM2M, FGOALS-s2, MIROC-ESM, CESM1-BGC, GFDL-ESM2G, HadCM3, and CanESM2. (bottom) RCP8.5: ACCESS1.0, IPSL-CM5B-LR, MIROC-ESM, CanESM2, FGOALS-s2, MPI-ESM-MR, CSIRO Mk3.6.0, CESM1-BGC, INM-CM4.0, CCSM4, GFDL-E SM 2G, CMCC- CM, GFDL-E SM 2M, NorESM1-M, IPSL-CM5A-MR, IPSL-CM5A-LR, MIROC-ESM-CHEM, MRI-CGCM3, MPI-ESM-LR RCP4.5: ACCESS1.0, IPSL-CM5B-LR, MIROC-ESM, CanESM2, FGOALS-s2, MPI-ESM-MR, CSIRO Mk3.6.0, CESM1-BGC, INM-CM4.0, CCSM4, APRIL 2014
| ES75
ES76 |
APRIL 2014
X X X X
X
X
X X X X X X X X X
Fourth Generation Canadian Coupled Global Climate Model
Second Generation Canadian Earth System Model
Community Climate System Model, version 4
Community Earth System Model, version 1 (biogeochemical)
Community Earth System Model version 1 - Community Atmospheric Model version 5.1
Community Earth System Model, version 1 (Community Atmosphere Model, version 5)
Community Earth System Model, version 1 (with FASTCHEM)
Centro Euro-Mediterraneo sui Cambiamenti Climatici Climate Model
Centre National de Recherches Météorologiques Coupled Global Climate Model, version 5
Commonwealth Scientific and Industrial Research Organisation Mark, version 3.6.0 EC-Earth Consortium Flexible Global Ocean–Atmosphere–Land System Model gridpoint, version 2.0 Flexible Global Ocean–Atmosphere–Land System Model gridpoint, second spectral version First Institute of Oceanography (FIO) Earth System Model (ESM) Geophysical Fluid Dynamics Laboratory Climate Model, version 2.1 Geophysical Fluid Dynamics Laboratory Climate Model, version 3 Geophysical Fluid Dynamics Laboratory Earth System Model with Generalized Ocean Layer Dynamics (GOLD) component (ESM2G) Geophysical Fluid Dynamics Laboratory Earth System Model with Modular Ocean Model 4 (MOM4) component (ESM2M) Goddard Institute for Space Studies Model E, coupled with the HYCOM ocean model Goddard Institute for Space Studies Model E, coupled with the Russell ocean model Goddard Institute for Space Studies Model E, coupled with the Russell ocean model Goddard Institute for Space Studies Model E, coupled with the Russell ocean model Hadley Centre Coupled Model, version 3 Hadley Centre Global Environment Model, version 2—Atmosphere and Ocean Hadley Centre Global Environment Model, version 2—Carbon Cycle
CanCM4
CanESM2
CCSM4
CESM1-BGC
CESM1-CAM5.1.FV2
CESM1-CAM5
CESM1-FASTCHEM
CMCC-CM
CNRM-CM5
CSIRO Mk3.6.0
HadGEM2-CC
HadGEM2-AO
HadCM3
GISS-E2-R_p3
GISS-E2-R_p2
GISS-E2-R_p1
GISS-E2H
GFDL-ESM2M
GFDL-ESM2G
GFDL-CM3
GFDL-CM2.1
FIO-ESM
FGOALS-s2
FGOALS-g2
EC-EARTH
X X X X
X
Beijing Normal University Earth System Model
BNU-ESM
X X
X X
X X X
X
X
X X
X
X
X X
X
X X X X
X
X
X
X
X
X X
X
X
X X
X
X X X X
X
X
X
X
X
X
X
X X X X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
4.5 8.5
X
X
X
X
X
X
X
X X X
X
X
X
X
X
X
X
9 10
X X X
X
X
X
X
X
X
8
X X
X X
X
X
X X
X
X
X
X
X
X
X
X
RCP
6
X X X
X
X X
X
X X X X
X
X
X
X
X
X
X
X
Historical
5
X
X
X
X
X
X
X
X
4
X
X
X
X
X
X
X
X X
Beijing Climate Center, Climate System Model, version 1.1
BCC-CSM1.1
X
Australian Community Climate and Earth-System Simulator, version 1.3
X
RCP 2.6 4.5 6.0 8.5
ACCESS1.3
X
Historical
Australian Community Climate and Earth-System Simulator, version 1.0
Expansion
2
ACCESS1.0
Figure
Table ES1. CMIP5 models used in the analyses and associated figures used in this paper.
20 19 21 16 18 15 26 27 32 18 32 31 42
X X X X X X X X NorESM1-M
NorESM1-ME
MRI-CGCM3_p2
MRI-CGCM3_p1
MPI-ESM-P
MPI-ESM-MR
MPI-ESM-LR
MIROC-ESM
MIROC-ESM-CHEM
MIROC5
IPSL-CM5B-LR
IPSL-CM5A-MR
IPSL-CM5A-LR
Total number of models used
X X X X X X X X X
X
X
X
X
X
X
X X
X
X
X X X X X
X
X X X X X X X X X X X X X X
X X X X X X X X X X X X X X X X X X
X X X X X X X X X X X
X X
X X
X X X X X
X
X
X X X X X X X X X X X
X X X
X X X X X X X X
X
X
X
X
X X X X X X X X X X X X X X X X X X X X X X X
X X X X
X X X INM-CM4.0
HadGEM2-ES
Hadley Centre Global Environment Model, version 2—Earth System Institute of Numerical Mathematics Coupled Model, version 4.0 L’Institut Pierre-Simon Laplace Coupled Model, version 5, coupled with NEMO, low resolution L’Institut Pierre-Simon Laplace Coupled Model, version 5, coupled with NEMO, mid resolution Institut Pierre Simon Laplace, Paris, France), version 5B (low resolution) Model for Interdisciplinary Research on Climate, version 5 Model for Interdisciplinary Research on Climate, Earth System Model, Chemistry Coupled Model for Interdisciplinary Research on Climate, Earth System Model Max Planck Institute Earth System Model, low resolution Max Planck Institute Earth System Model, medium resolution Max Planck Institute Earth System Model, paleo Meteorological Research Institute Coupled Atmosphere–Ocean General Circulation Model, version 3 Meteorological Research Institute Coupled Atmosphere–Ocean General Circulation Model, version 3 Norwegian Earth System Model, version 1 (medium resolution) Norwegian Earth System Model, version 1 (intermediate resolution) AMERICAN METEOROLOGICAL SOCIETY
GFDL-ESM2G, CMCC-CM, GFDL-ESM2M, NorESM1-M, IPSL-CM5A-MR, IPSLCM5A-LR, CNRM-CM5, MIROC-ESMCHEM, MRI-CGCM3, and MPI-ESM-LR. Figure 6: ACCESS1.0, CCSM4, CESM1BGC, CanESM2, GFDL-ESM2G, IPSLCM5A-LR, IPSL-CM5A-MR, IPSL-CM5BLR, MIROC-ESM, MIROC-ESM-CHEM, MIROC5, MPI-ESM-LR, MPI-ESM-MR, MR I- CGCM3, GFDL-ESM 2M, BCC _ CSM1.1, CMCC-CM, and FGOALS-s2. Figure 7: As in Fig. 2 Figure 8: Top two panels: CanESM2, CCSM4, CSIRO Mk3.6.0, GFDL-ESM2M, GISS-E2H, GISS-E2-R, HadGEM2-CC, HadGEM2-ES, INM-CM4.0, IPSL-CM5ALR, IPSL-CM5A-MR, MIROC5, MIROCESM, MIROC-ESM-CHEM, MPI-ESM-LR, MRI-CGCM3, NorESM1-M, NorESM1-ME. Bottom four panels: CNRM-CM5, CSIRO Mk3, CanESM2, FIO-ESM, GFDL-CM3, CFDL-ESM2M, GISS-E2, HadGEM2-ES, IPSL-CM5A-LR, IPSL-CM5A-MR, MIROCESM, MIROC5, MPI-ESM-LR, MRI-2005 historical; CNR M-CM5, CSIRO Mk3, CanESM2, FIO-ESM, GFDL-CM3, CFDLESM2M, GISS-E2, HadGEM2-ES, IPSLCM5A-LR, IPSL-CM5A-MR, MIROC-ESM, MIROC5, MPI-ESM-LR, MRI-CGCM3, NorESM1-M, and INM-CM4.0. Figure 9: CanESM2, CCSM4, GFDLESM2M, HadGEM2-CC, HadGEM2-ES, I NM- CM4 .0, IPSL- CM5A-LR , IPSLCM5A-MR, MIROC-ESM, MIROC-ESMCHEM, MPI-ESM-LR, MRI-CGCM3, and NorESM1-M. Figure 10: CESM, EC-EARTH, MRICGCM3, CNR M- CM5, MIROC5, HadGEM 2-ES, HadGEM 2-CC , I NMCM4.0, IPSL-CM5A-MR, MPI-ESM-LR, NorESM1, GFDL-ESM2M, IPSL-CM5A-LR, BCC_CSM1.1, and MIROC-ESM-C. Analysis of Southeast United States summer precipitation variability: CSIRO; C C SM4 ; GF DL -E SM 2 G a nd GF DL ESM2M; HadCM3; HadGEM2-ES; INM-CM4.0, MIROC4h and MIROC5; MIROC-ESM; MPI-ESM-LR; and MPIESM-MR-g2: ACCESS1.0, BCC_CSM1.1, CanESM2, CNR M-CM5, FGOALS-g2, FGOALS-s2, GFDL-CM3, HadGEM2CC, IPSL-CM5A, IPSL-CM5A-MR, MRICGCM3, and NorESM. APRIL 2014
| ES77
year for the RCP8.5 scenario. For the contiguous United States, cold spell temperature increases range from around 3°C in Florida to more than 8°C in the north-central United States for 2071–99 compared to 1971–2000. Hot spell temperature increases range from around 5°C in far southern areas and along the west coast to more than 7°C in parts of the Midwest and northern Rockies. Temperature increases in Alaska (Hawaii) are similar (slightly lower) for the hottest month and greater (lower) for Alaska (Hawaii) for the coldest month. Table ES2. The number of ensembles used from each Figure 4: The effect of warming trend is model. computed from Wergen and Krug (2010) Model Historical RCP2.6 RCP8.5 where the probability of record breaking (pn) ACCESS1.0 1 — 1 is expressed as NOTES ABOUT INDIVIDUAL FIGURES. Figure 2: An ensemble average for each model based on all available realizations was calculated prior to the calculation of the equally weighted multimodel averages. Figure 3: An alternative analysis is shown in Fig.ES1, as discussed in the paper. Figure ES1 shows projected multimodel mean changes in the temperatures of the hottest and coldest months of the
ACCESS1.3
1
—
1
BCC_CSM1.1
3
1
1
BNU-ESM
1
1
1
CanESM2
5
5
5
CCSM4
2
5
5
CESM1-BGC
1
—
1
CMCC-CM
1
—
1
CNRM-CM5
5
1
1
CSIRO Mk3.6.0
5
5
5
EC-EARTH
5
2
5
FGOALS-g2
2
1
1
FGOALS-s2
3
1
3
GFDL-CM3
5
1
1
GFDL-ESM2G
1
1
1
GFDL-ESM2M
1
1
1
GISS-E2-R
1
—
—
HadCM3
1
—
—
HadGEM2-ES
4
4
4
HadGEM2-CC
3
—
3
INM-CM4.0
1
—
1
IPSL-CM5A-LR
5
3
4
IPSL-CM5A-MR
1
1
1
IPSL-CM5B-LR
1
—
1
Model for Interdisciplinary Research on Climate, version 4 (high resolution) (MIROC4h)
3
—
—
MIROC5
4
3
3
MIROC-ESM
3
1
1
MIROC-ESM-CHEM
1
1
1
MPI-ESM-LR
3
3
3
MPI-ESM-MR
3
1
1
MRI-CGCM3
5
1
1
NorESM1-M
3
1
1
Total number of models
32
22
29
ES78 |
APRIL 2014
Here, υ is the mean temperature trend in yr–1, and σ is the standard deviation. For an independently and identically distributed (iid) sequence (without linear trend, υ = 0), the second term vanishes. More scattered distribution of observed record temperatures is because of their smaller sample size compared with those in the multimodel ensemble. Figure 5: The observed data are from the United States Cooperative Observer Network as included in the Global Historical Climate Network Daily dataset from the National Climatic Data Center (NCDC). The extreme precipitation index (EPI) was determined in a method similar to that used in Kunkel et al. (1999). For example, for 2-day duration, 1- in-5-yr return period, N is the number of days and Y the number of years in each time series. For a 5-yr return there would be Y/5 events in each time series. Then total precipitation (P) is computed for days 1–2 and set as the maximum precipitation value (Pmax). Total precipitation is then computed for days 2–3 and set as P. If P is greater than Pmax, P is set as the new Pmax. If any data are missing from the days then this last step is skipped. This process is repeated until days (N – 1) – N are reached. The amount and year/ days of Pmax are then stored in some array and the days of Pmax are set to “missing.” This whole process is repeated until the top Y/5, 2-day duration events have been found in the time series.
Once the number of extreme events (day –1) for each station time series were f lagged in the data the EPI for the observations were calculated using a gridded spatial average. Stations were broken into g r id poi nt s across t he United States. Then an arithmetic mean was done for each year using the number of extreme events over the number of days for each grid point. Next, the average time series were avFig. ES1. Projected change (°C) in the temperature of the (left) coldest month eraged over the grid points and (right) hottest month in 2071–99 compared to 1971–2000 for the RCP8.5 to get one EPI time series scenario. In both periods, the single hottest and coldest months of the entire period were identified at each grid point. for the contiguous United States. The number of models analyzed for the mid–low RCP4.5 (20 models) and higher RCP8.5 (19 models) scenarios differ because historical data had to be available for both the historical simulations and the projections for each model. Figure ES2 shows the correlation coefficient of observed and modeled decadal average EPI values for the CONUS for each of the 26 CMIP5 models used for the time period (1906–2005). Many models show a correlation coefficient greater than 0.50, with BCC_CSM1.1 having the best correlation, approaching 1.00 for a 10-yr return. A negative correlation is shown for seven of Fig. ES2. Correlation coefficients for each of the 26 CMIP5 models used to calculate EPI model median values are shown. Both observed the models, thus demonstrating the and individual model EPI values were averaged by decade from 1906 large spread among how well the to 2005 for all three return periods. The correlation coefficient for models capture observed extreme models and observations was then calculated for each model over precipitation events. the entire time period. There is large variability in the ensemble runs for the HadCM3 model (Fig. ES3) as well as for the ensembles from 1) For each model: At each grid point, calculate other models (not shown). For the HadCM3 model the 99th percentile for the baseline period there is a large amount of variability among the 1901–60 and take the sum of data for days for ensembles’ runs but there is better agreement among every year above that threshold (precipitation all the runs for the last decade of a positive EPI of less than 1 mm day–1 is not counted in the percent anomaly. A similar result was found in the percentile calculations). Express that sum as a ensembles from other models (examining models percent of the total precipitation for that year. with five or more simulations). Average these percentage values for the continenFigure 6: Time series analysis methodology: tal United States. AMERICAN METEOROLOGICAL SOCIETY
APRIL 2014
| ES79
In Fig. 6, the mean value of the 99th percentile for CONUS was 34.6 mm day–1 for the models and 83.8 mm day–1 for the observations. This difference is likely related to the point values of the observations versus the area-averaged weighting of the model grid differences. Figure 8: All models were regridded to T42 resolution prior to computing the multimodel ensemble mean for the present-day and future periods. Figure 9: Drought extent is defined as the area in drought over a region at any one time. A drought is defined as when the soil moisture is below the 20th percentile. Figure 10: In Fig. 10a, the difference in cyclone track density is shaded every Fig. ES3. EPI percent anomaly for the various ensemble runs from 0.2 and the percent change is contoured the CMIP5 HadCM3 model. The EPI was calculated for 1901–2005 every 10% with negative dashed. In and then the percent anomalies were calculated by decade for the period 1906–2005. There are 10 different ensemble runs for this Fig. 10b, shaded areas are the number particular model. of cyclone tracks per 5 cool seasons per 50,000 km2 and the percentage change is 2) Calculate the 1901–60 average of the data calculated contoured every 10% with negative dashed. In Figs. 10c above for each model. Then for each model, calculate and 10d, the mean and spread of the CMIP5 pressure difthe percent change at each year from this value. ferences are centered for each 10-hPa bin. 3) Plot a central projections line by taking the running means of the percent changes for all the models at REFERENCES each year. The actual line that is plotted is the average of these values across all the models. Each year is Kunkel, K. E., K. Andsager, and D. R. Easterling, 1999: calculated as the center of the running average, so the Long-term trends in extreme precipitation events over first and final 4 yr are cut off. the conterminous United States and Canada. J. Climate, 4) Calculate the standard deviation of the percent 12, 2515–2527. changes for the models at each year, then add and Wergen, Q., and J. Krug, 2010: Record-breaking temperatures subtract this from the running average above to plot reveal a warming climate. Europhys. Lett., 92, 30008, the ranges (shaded areas). doi:10.1209/0295-5075/92/30008.
ES80 |
APRIL 2014
