Transcript Change - PIRCS - Iowa State University

PIRCS:
Approach and Lessons
Learned
William Gutowski
Iowa State University
With thanks to
R.Arritt, G. Takle,
Z. Pan, J. Christensen,
R. Wilby,L. Hay, M. Clark,
PIRCS modelers
http://rcmlab.agron.iastate.edu
PIRCS:
Approach and Lessons
Learned
1. History - PIRCS 1a & 1b
2. PIRCS 1c
3. Spinoff: 10-yr “ensemble”
4. Transferability
5. Impacts
6. Summary
Project to Intercompare Regional
Climate Simulations (PIRCS)
• Systematically examine regional climate model
simulations to identify common successes and
errors
– "Regional"  "limited area"
– Different models, parameterizations, computer
hardware
– Same domain and period of simulation
– Consistent analysis procedures and software
• Provide a starting point for other community
efforts (e.g., NARCCAP)
PIRCS Experiments
Expt. 1a: 15 May - 15 July 1988 (Drought)
Expt. 1b: 1 June - 30 July 1993 (Flood)
Expt. 1c: July 1986 - Dec 1993 …
(reanalysis boundary conditions)
Spin-off: 1979-1988 & Scenarios
(reanalysis & GCM boundary conditions)
PIRCS Participants
 Danish Met. Inst. (HIRHAM4; J.H. Christensen, O.B. Christensen)
 Université du Québec à Montréal (D. Caya, S. Biner)
 Scripps Institution of Oceanography (RSM; J. Roads, S. Chen)
 NCEP (RSM; S.-Y. Hong)
 NASA - Marshall (MM5/BATS; W. Lapenta)
 CSIRO (DARLAM; J. McGregor, J. Katzfey)
 Colorado State University (ClimRAMS; G. Liston)
 Iowa State University (RegCM2; Z. Pan)
 Iowa State University (MM5/LSM; D. Flory)
 Univ. of Maryland / NASA-GSFC (GEOS; M. Fox-Rabinovitz)
 SMHI / Rossby Centre (RCA; M. Rummukainen, C. Jones)
 NOAA (RUC2; G. Grell)
 ETH (D. Luethi)
 Universidad Complutense Madrid (PROMES; M.Gaertner)
 Université Catholique du Louvain (P. Marbaix)
 Argnonne / Lawrence Livermore National Labs (MM5 V3; J. Taylor, J. Larson)
 St. Louis University (Z. Pan)
Z(500 hPa) Differences. Period = PIRCS 1b - PIRCS 1a
(a)
Reanalysis
(b)
PIRCS Ensemble
PIRCS Ensemble - VEMAP
June
1988
July
1993
-3
0
+3
[mm/d]
Area-averaged precipitation
in the north-central U.S.
45
40
30
25
20
15
450
10
400
5
0
01.Jun
01.Jul
Multi-Model (PIRCS 1B)
Precipitation (mm)
Precipitation (cm)
35
Mixed Physics
Ensemble Mean
run1
run2
run3
run4
run5
run6
run7
run8
run9
run10
350
300
250
ClimRAM S
DARLAM
M M 5-ANL
PROM ES
CRCM
HIRHAM
M M 5-BATS
RegCM 2
RSM -NCEP
SweCLIM -ECM WF
M odel
Average
31.Jul
OBS-VEM AP
RSM -Scripps
SweCLIM -NCEP
OBS-HIGGINS
200
150
100
50
0
1-Jun
11-Jun
21-Jun
1-Jul
11-Jul
21-Jul
31-Jul
PIRCS 1a & 1b: Conclusions
• Ensembles are important
–Reveal common & unique problems
–No model is “best”
• Distinction between problems of
–Lateral forcing/dyamics (“common”)
–Surface processes (“unique”)
• Interannual climate variation
–Simulated in large-scale dynamics
–Muted in precipitation response
PIRCS:
Approach and Lessons
Learned
1. History - PIRCS 1a & 1b
2. PIRCS 1c
3. Spinoff: 10-yr “ensemble”
4. Transferability
5. Impacts
6. Summary
PIRCS 1c: Participants
Model
Lead Investigator
MM5-ISU
Chris Anderson
MM5-ANL/LLNL John Taylor
RSM-Scripps
John Roads
SweCLIM
Colin Jones
CRCM
Sebastian Biner
Ensemble spread:
Upper Ms. River
lagged
ensemble
Upper Mississippi River
base
50
lag1
lag2
1991
1992
25
0
• Shown: % variations of precip.
For each member about the
mean for that ensemble
• Internal variability is less than
variability due to physics
• Large year-to-year variations
in spread due to physics
• The types of variability do not
appear to be correlated
-25
-50
1987
1988
1989
1990
1993
Upper Mississippi River
base
50
FC
FC150
25
0
physics
ensemble
(RW Arritt, 2004)
-25
-50
1987
1988
1989
1990
1991
1992
1993
Ensemble spread:
Pacific Northwest
lagged
ensemble
Pacific Northwest
base
50
lag1
lag2
1991
1992
25
0
-25
• Internal variability is
extremely small because
most precipitation occurs in
the winter, when large-scale
control is strong
• Physics variability also is
smaller than for central U.S.,
even in summer
physics
ensemble
-50
1987
1988
1989
1990
1993
Pacific Northwest
base
50
FC
FC150
25
0
-25
(RW Arritt, 2004)
-50
1987
1988
1989
1990
1991
1992
1993
Current Status
• Runs and analysis for PIRCS 1C are
presently at an early stage
• Potential coordination with other projects:
– perform complementary simulations
– suggest diagnostics
Details: http://rcmlab.agron.iastate.edu
PIRCS:
Approach and Lessons
Learned
1. History - PIRCS 1a & 1b
2. PIRCS 1c
3. Spinoff: 10-yr “ensemble”
4. Transferability
5. Impacts
6. Summary
Simulations
Model
Observed
RegCM2
NCEP
Hadley
Reanalysis Centre
(1979-1988) (~1990’s)
HIRHAM
(DMI)
“
GCM-control GCMScenario
“
Hadley
Centre
(2040-2050)
“
Possible Comparisons?
Reanalysis
HadCM
Cont/Scen
Driving
RegCM2
OBS
HIRHAM
HadCM
Cont/Scen
Differences
Definition of Biases
Reanalysis
RegCM2
OBS
RCM (performance) bias
Definition of Biases
Reanalysis
RegCM2
Inter-model
bias
HIRHAM
Definition of Biases
Reanalysis
RegCM2
Forcing
bias
HadCM
RegCM2
Definition of Biases
RegCM2
HadCM
G-R
nesting
bias
HadCM
Climate Change
HadCM
control
RegCM2
Change
HadCM
scenario
RegCM2
Climate Change
Change
Control
Scenario
(Pan et al., JGR, 2001)
Climate Change
Max Bias
Control
Change
Scenario
(Pan et al., JGR, 2001)
Climate Change
Max Bias
Control
Change
Scenario
Rchng = Change / Max-Bias
(Pan et al., JGR, 2001)
0
1
2
0
1
2
0
1
2
PIRCS:
Approach and Lessons
Learned
1. History - PIRCS 1a & 1b
2. PIRCS 1c
3. Spinoff: 10-yr “ensemble”
4. Transferability
5. Impacts
6. Summary
Transferability Working Group
(proposed)
GEWEX Hydrometeorology Panel
World Climate Research Programme
Objective: Improved understanding and predictive capability
through systematic intercomparisons of regional climate
simulations on several continents with observations and analyses
• Build on coordinated observations from GEWEX continental scale
experiments
• Provide a framework for evaluating regional model simulations of climate
processes of different climatic regions.
• Evaluate transferability of regional climate models, for example a model
developed to study one region as applied to other, “non-native”, regions
• Examine individual and ensemble performance between domains and on
individual domains
Proposal coordinated by
E. S. Takle, W. J. Gutowski, Jr., and R. W. Arritt
Iowa State University
Relevance to California?
 “When climate changes, will your model be
ready?”
 How do models perform elsewhere?
RegCM3 Simulations - Various Regions
RegCM3 Simulations - Various Regions
Analysis Regions
RegCM2
7
Rchng
6
winter
spring
summer
autumn
5
4
3
2
1
0
PNW
CA
MW
NE
Region
Rchng
Pchng

Max ( PRCM , Pforc, Pitmd )
NS
HIRHAM
7
6
winter
spring
summer
autumn
Rchng
5
4
3
2
1
0
PNW
CA
MW
Region
NE
SE
Relevance to California?
 “When climate changes, will your model be
ready?”
 How do models perform elsewhere?
 Results suggest using large enough area to
encompass other climatic regions.
PIRCS:
Approach and Lessons
Learned
1. History - PIRCS 1a & 1b
2. PIRCS 1c
3. Spinoff: 10-yr “ensemble”
4. Transferability
5. Impacts
6. Summary
BASINS
San Juan
Obs. Stations
37
Animas
Obs. Stations
3
Model Points
3
Model Points
16
Snowpack - Anim as
40
SIMULATED
RegCM
StatDS
35
Depth [cm]
30
25
20
15
10
5
0
1980
1981
1982
1983
Year
1984
1985
1986
1987
Precipitation by Intensity Category
- Animas - cold -
500
OBS
RegCM
StatDS (ens-sdev)
StatDS (ens+sdev)
Cumulative P [mm]
400
300
200
100
0
0
5
10
15
20
25
Category [mm/d]
30
35
40
Cold Season - Tmax
- Animas -
350
300
OBS
RegCM
StatDS (ens)
250
Count
200
150
100
50
0
-50
220
240
260
280
Temperature [K]
300
320
Comparison of Simulated Stream Flow under
Climate Change with Various Model Biases
Relation of Runoff to Precipitation
for Various Climates
Yield Summary
(all in kg/ha)
Observed Yields
Simulated by CERES with
Observed weather
RegCM2/NCEP
HIRHAM/NCEP
RegCM2/HadCM2 current
HIRHAM/HadCM2 current
Mean St. Dev.
8381 1214
8259 4494
5487 3796
3446 2716
5002 1777
6264 3110
Yield Summary
• Deficiencies in RCMs and GCMs for
driving crop models likely is due to poor
timing and amounts of precipitation
• Crop models expose and amplify
vegetation-sensitive climate features of
a GCM or RCM
PIRCS:
Approach and Lessons
Learned
1. History - PIRCS 1a & 1b
2. PIRCS 1c
3. Spinoff: 10-yr “ensemble”
4. Transferability
5. Impacts
6. Summary
PIRCS:
Lessons Learned
1. Ensembles are important
2. Models have common precipitation biases
(daily and interannual)
3. Must understand model behavior in a variety
of climates
4. Two-way interaction with impacts groups is
vital
5. Require common data formatting
Acknowledgements
 Primary Funding:
Electric Power Research Institute (EPRI)
NOAA
 Guidance/Support:
Andrew Staniforth, Eugenia Kalnay,
Filippo Giorgi, Roger Pielke, AMIP group
 Special Thanks:
Participating Modelers
http://rcmlab.agron.iastate.edu
Without sufficient resolution,
it just doesn’t look right.
EST&LM