Upload
phamdung
View
220
Download
0
Embed Size (px)
Citation preview
Dan Cayan SIO and USGS David Rupp and Phil Mote Oregon State University, NWCSC Mary Tyree SIO Suraj Polade, Sasha Gershunov, Dave Pierce, Mike De=nger
In Seek of Climate change scenarios for California from CMIP5
20
DTR.SONDTR.JJA
DTR.MAMDTR.DJF
SpaceSD.SON−PSpaceSD.JJA−P
SpaceSD.MAM−PSpaceSD.DJF−PSpaceSD.SON−TSpaceSD.JJA−T
SpaceSD.MAM−TSpaceSD.DJF−TSpaceCor.SON−PSpaceCor.JJA−P
SpaceCor.MAM−PSpaceCor.DJF−PSpaceCor.SON−TSpaceCor.JJA−T
SpaceCor.MAM−TSpaceCor.DJF−T
TimeCV.8yr−PTimeVar.8yr−TTimeCV.1yr−PTimeVar.1yr−T
Hurst−PHurst−TENSO−PENSO−TTrend−PTrend−T
SeasonAmp−PSeasonAmp−T
Mean−PMean−T
CESM
1−FAST
CHEM
MIRO
C5CC
SM4
CNRM
−CM5
CanESM
2CE
SM1−BG
CCE
SM1−CA
M5IPSL−C
M5A−MR
HadG
EM2−ES
NorESM
1−M
GFDL
−ESM
2MFIO−
ESM
HadG
EM2−AO
HadC
M3Ha
dGEM
2−CC
CMCC
−CM
MPI−E
SM−M
RCS
IRO−
Mk3−6−0
GFDL
−CM3
MPI−E
SM−LR
IPSL−C
M5A−LR
EC−EAR
THGF
DL−C
M2p1
bcc−csm1
−1−m
NorESM
1−ME
CESM
1−WAC
CMGF
DL−ESM
2GMIRO
C−ES
MMR
I−CGC
M3MIRO
C−ES
M−CH
EMGISS
−E2−H
GISS
−E2−R
BNU−
ESM
inmcm
4bcc−csm1
−1CM
CC−C
ESM
FGOA
LS−s2
0.0
0.2
0.4
0.6
0.8
1.0
Figure 19. Relative error of the ensemble mean of each metric for each CMIP5 GCM. Models are ordered from least (left) to most (right) total relative error, where total relative error is the sum of relative errors from all metrics, excluding the diurnal temperature range (DTR) metrics. For 7 GCMs, the diurnal temperature range (DTR) metrics were not calculated (white squares).
Figure 19. Rela*ve error of the ensemble mean of each metric for each CMIP5 GCM. Models are ordered from least (le=) to most (right) total rela*ve error, where total rela*ve error is the sum of rela*ve errors from all metrics, excluding the diurnal temperature range (DTR) metrics. For 7 GCMs, the diurnal temperature range (DTR) metrics were not calculated (white squares).
Ranking GCM’s by Rupp and Mote using a mul*-‐factor historical climate evalua*on scheme, SW US
8
Figure 8. Mean precipitation for DJF (a and b) and JJA (c and d) from NCEP (a and c) and from the CMIP5 multi-model mean (b and d). re 8. Mean precipita*on for DJF (a and b) and JJA (c and d) from NCEP (a and c) and from the CMIP5 mul*-‐model mean (b and d).
2
Figure 2. Mean seasonal cycle of temperature (upper panel) and precipitation rate (lower panel) averaged over the SWUS. Monthly means calculated from gridded observation datasets (NCEP, CRU, PRISM) and from all ensemble members from 37 CMIP5 GCMs.
Figure 2. Mean seasonal cycle of temperature (upper panel) and precipita*on rate (lower panel) averaged over the SWUS. Monthly means calculated from gridded observa*on datasets (NCEP, CRU, PRISM) and from all ensemble members from 37 CMIP5 GCMs.
3
Figure 3. SWUS mean seasonal cycle amplitude in temperature and precipitation by model. For each CMIP5 model, blue-filled circles show the ensemble average, yellow-filled circles show the first ensemble member, and the open circles show the remaining ensemble members. For the observations (Obs), the red-filled circle shows the average, the yellow-filled circle shows NCEP, and the open circle shows CRU and PRISM.
Figure 3. SWUS mean seasonal cycle amplitude in temperature and precipita*on by model. For each CMIP5 model, blue-‐filled circles show the ensemble average, yellow-‐filled circles show the first ensemble member, and the open circles show the remaining ensemble members. For the observa*ons (Obs), the red-‐filled circle shows the average, the yellow-‐filled circle shows NCEP, and the open circle shows CRU and PRISM.
18
Figure 16. Response of SWUS winter (JFM) temperature and precipitation to the Niño3.4 index averaged over ASOND of the previous year. For each CMIP5 model, blue-filled circles show the ensemble average, yellow-filled circles show the first ensemble member, and the open circles show the remaining ensemble members. CRU was used for the observations (Obs). Figure 16. Response of SWUS winter (JFM) temperature and precipita*on to the
Niño3.4 index averaged over ASOND of the previous year. For each CMIP5 model, blue-‐filled circles show the ensemble average, yellow-‐filled circles show the first ensemble member, and the open circles show the remaining ensemble members. CRU was used for the observa*ons (Obs).
11
Varia
nce o C2
1 2 3 4 5 6 8 10
0.01
0.1
1
CRUPRISMCMIP5
Temperature
Resolution (years)
CV
1 2 3 4 5 6 8 10
0.02
0.05
0.1
0.2
0.3
CRUPRISMCMIP5
Precipitation
Figure 11. Variance of temperature anomalies (upper panel) and coefficient of variation of precipitation (lower panel) against temporal resolution for the SWUS-averaged time series.
Figure 11. Variance of temperature anomalies (upper panel) and coefficient of varia*on of precipita*on (lower panel) against temporal resolu*on for the SWUS-‐averaged *me series.
22
Figure 21. Models ranked according to normalized error score from EOF analysis of 18 (upper panel) and 16 (lower panel) performance metrics. The upper panel includes winter and summer mean diurnal temperature range (DTR), but 7 fewer GCMs. Ranking is based on the first 6 principle components (filled blue circles). The open symbols show the models’ error scores using the first 4, 5, and all principle components (PCs). The best scoring model has a normalized error score of 0.
Figure 21. Models ranked according to normalized error score from EOF analysis of 18 (upper panel) and 16 (lower panel) performance metrics. The upper panel includes winter and summer mean diurnal temperature range (DTR), but 7 fewer GCMs. Ranking is based on the first 6 principle components (filled blue circles). The open symbols show the models’ error scores using the first 4, 5, and all principle components (PCs). The best scoring model has a normalized error score of 0.
Mid century median change 2040-‐2069 16 cmip5 models
rcp26
rcp45
rcp85
Models: • Noresm1m • Canesm2 • Cnrmcm5 • Csiromk3.6 • Gfdlcm3 • Gfdlesm2g • Gisse2r • Inmcm4 • Mirocesm • Mpiesmlr • Mpiesmmr • Ccsm4 • Cesm1cam5 • Hadgem2ao • Mricgcm3 • Ipslcm5alr
Mid century median change 2037-‐2066 28 cmip5 models
rcp26
rcp45
rcp85
7
Figure 7. Mean temperature for DJF (a and b) and JJA (c and d) from NCEP (a and c) and from the CMIP5 multi-model mean (b and d).
Figure 7. Mean temperature for DJF (a and b) and JJA (c and d) from NCEP (a and c) and from the CMIP5 mul*-‐model mean (b and d).
4
Figure 4. Mean seasonal amplitude for temperature (a and b) and precipitation (c and d) from NCEP (a and c) and from the CMIP5 multi-model mean (b and d). Note that the legends have been reversed between the upper and lower plots so that red implies a more extreme temperature cycle and blue implies a more extreme precipitation cycle.
Figure 4. Mean seasonal amplitude for temperature (a and b) and precipita*on (c and d) from NCEP (a and c) and from the CMIP5 mul*-‐model mean (b and d). Note that the legends have been reversed between the upper and lower plots so that red implies a more extreme temperature cycle and blue implies a more extreme precipita*on cycle.
13
Figure 13. Coefficient of variation of precipitation at temporal resolutions of 1 year and 8 years for all CMIP5 simulations and observations. Values were averaged over the SWUS domain. For each model, blue-filled circles show the ensemble average, yellow-filled circles show the first ensemble member, and the open circles show the remaining ensemble members. For the observations (Obs), the yellow-filled circle shows PRISM, the open circle shows CRU, and the red-filled circle gives their average.
Figure 13. Coefficient of varia*on of precipita*on at temporal resolu*ons of 1 year and 8 years for all CMIP5 simula*ons and observa*ons. Values were averaged over the SWUS domain. For each model, blue-‐filled circles show the ensemble average, yellow-‐filled circles show the first ensemble member, and the open circles show the remaining ensemble members. For the observa*ons (Obs), the yellow-‐filled circle shows PRISM, the open circle shows CRU, and the red-‐filled circle gives their average.
1
Figure 1. SWUS mean annual temperature and precipitation bias by model. For each CMIP5 model, blue-filled circles show the ensemble average, yellow-filled circles show the first ensemble member, and the open circles show the remaining ensemble members. For the observations (Obs), the red-filled circle shows the average, the yellow-filled circle shows NCEP, and the open circles show CRU and PRISM.
SWUS mean annual temperature and precipita*on bias by model. For each CMIP5 model, blue-‐filled circles show the ensemble average, yellow-‐filled circles show the first ensemble member, and the open circles show the remaining ensemble members. For the observa*ons (Obs), the red-‐filled circle shows the average, the yellow-‐filled circle shows NCEP, and the open circles show CRU and PRISM.
21
−4 −2 0 2 4−4
−20
24
PC1
PC2
12
456
78
9
10
11
12
13
14
15
16
18
19
20
2324
252627
3031
32
3334
35
36
Obs
−4 −2 0 2 4
−4−2
02
4
PC3
PC4
1
2
4
56 78
9
10
1112
13
14
15
16
18 1920
23
2425 26
27
30
31
32
3334
35
36Obs
1. bcc−csm1−12. bcc−csm1−1−m4. CanESM25. CCSM46. CESM1−BGC7. CESM1−CAM58. CESM1−FASTCHEM9. CESM1−WACCM10. CMCC−CESM11. CMCC−CM12. CNRM−CM513. CSIRO−Mk3−6−014. EC−EARTH15. FGOALS−s216. FIO−ESM18. GFDL−CM319. GFDL−ESM2G20. GFDL−ESM2M23. HadCM324. HadGEM2−AO25. HadGEM2−CC26. HadGEM2−ES27. inmcm430. MIROC−ESM31. MIROC−ESM−CHEM32. MIROC533. MPI−ESM−LR34. MPI−ESM−MR35. MRI−CGCM336. NorESM1−M
Figure 20. Loadings of the first four principle components (PC1, PC2, PC3, PC4) from EOF analysis of 18 evaluation metrics (includes diurnal temperature range) and 30 of the full 37 GCMs. “Obs” indicates the observation dataset.
Figure 20. Loadings of the first four principle components (PC1, PC2, PC3, PC4) from EOF analysis of 18 evalua*on metrics (includes diurnal temperature range) and 30 of the full 37 GCMs. “Obs” indicates the observa*on dataset.
Late century median change 2070-‐2099 28 cmip5 models
rcp26
rcp45
rcp85