Regional_downscaling_Park_2011

advertisement
UM collaboration meeting, 21-22 November 2011, KMA
Task: (ECSK06) Regional downscaling
Regional modelling with HadGEM3-RA
driven by HadGEM2-AO projections
National Institute of Meteorological Research (NIMR)/KMA
Outline
 Introduction
 50km-res CORDEX-East Asia experiment
 Evaluation of current climate simulation
 Projection of future climate change
 12.5km-res Korea experiment
 Evaluation of current climate simulation
 Projection of future climate change
 Summary and future plan
Introduction
 Task: (ECSK06) Regional downscaling
Objective:
•Build UM-regional model over the East Asian region and perform
experimental runs for simulation of regional climate.
Deliverables:
•Report on the installation of the UM-regional model for the East Asian
region and its evaluation using perfect boundary conditions on seasonal
simulations of East Asian monsoon activity (2008-2010)
• Report on evaluation of HadGEM3-RA with a focus on climate variability
in long-term integrations using ECMWF interim reanalysis data,
associated with CORDEX participation (Dec 2011)
•Report on assessment of East Asian climate downscaled by HadGEM3RA using global climate change projections, associated with CORDEX
participation (Dec 2011)
Introduction
 Strategy for generating high resolution climate change
scenarios under IPCC AR5
New IPCC Scenarios
RCP 4.5/8.5/2.6/6.0
Anthropogenic forcing
GCM projection
HadGEM2-AO : ~135 km
CMIP5
Dynamical downscaling
RCM projection
HadGEM3-RA : ~12.5/50 km
Period (years)
Scenario
Model
Grid spacing
Global Projection
1850 ~ 2300
RCP 4.5/8.5/2.6/6.0
HadGEM2-AO
~135km (1.875°x1.25°)
CORDEX
Regional Projection
1950 ~ 2100
RCP 4.5/8.5/2.6/6.0
HadGEM3-RA
~12.5/50km (0.11/0.44°)
• HadGEM2-AO: Atmosphere-Ocean coupled model of Hadley Centre Global Environment Model version 2
• HadGEM3-RA: Atmospheric regional climate model of Hadley Centre Global Environment Model version 3
Plan of generating regional climate change scenarios
 Experiments and progress (GA3.0 version)
 Simulations of Current Climate (to evaluate the performance of RCMs)
- Experiments using reanalysis boundary conditions (1989-2008) - done
* Forcing: ERA-Interim atmospheric field & Daily Reynolds SST
- Experiments using GCM boundary conditions (1950-2005) - done
* Forcing: HadGEM2-AO atmospheric field & daily SST
 Simulations of Climate Change (to project future climate)
- Experiment using GCM RCP 8.5/4.5 runs (2006-2100) - done
* Forcing: HadGEM2-AO atmospheric field & daily SST
CORDEX 50km domain
Korea 12.5km domain
Evaluation of current climate simulation
in GCM forcing run (50km-res)
- surface climate
Climatology (1971-200): Precipitation
Observation
Winter
Summer
Annual
CRU
GCM
RCM
GCM
RCM
 RCM could resolve
small-scale features
related with topography
and coastlines.
Climatology (1971-200): Temperature
Observation
Winter
Summer
Annual
CRU
GCM
RCM
GCM
RCM
 RCM could resolve
small-scale features
related with topography
and coastlines.
Bias: Precipitation and temperature
Summer
RCM
Winter
GCM
 Temperature
Annual
 Precipitation
GCM
RCM
Statistics: Precipitation & Temperature (Land)
 Mean, bias, Root-mean-squared error (RMSE) and pattern correlation
coefficient of precipitation and temperature. (Ref.CRU)
Variables
Precip.
(mm/day)
Temp.
(°C)
Mean
(CRU)
ANN
2.70
JJA
4.37
DJF
1.59
ANN
10.87
JJA
20.50
DJF
-0.11
Mean
(Model)
Bias
RMSE
Pattern
Corr.
GCM
2.87
0.17
0.99
0.837
RCM
2.76
0.06
0.92
0.821
GCM
4.45
0.08
1.67
0.754
RCM
4.51
0.15
1.64
0.752
GCM
1.84
0.25
0.66
0.907
RCM
1.53
-0.06
0.63
0.894
GCM
9.73
-1.15
2.11
0.977
RCM
8.91
-1.96
2.13
0.984
GCM
21.14
0.64
1.80
0.942
RCM
20.36
-0.14
1.31
0.970
GCM
-3.12
-3.01
3.77
0.980
RCM
-3.65
-3.54
3.74
0.980
 Overall, both GCM and RCM show similar performance and wet/cold biases.
Annual cycle of Precipitation and temperature
 30-yr mean annual cycle of area-averaged precipitation and surface
air temperature (1951~1980): East Asia monsoon region(100E-150E,20N-50N)
Precip.
Temp
 Black: Observation (CRU)
 Red: GCM
 Blue: RCM
Climate change projection (50km-res)
•Change in surface air temperature and precipitation
Climate change Projection: Temperature
 Time series of annual mean surface air temperature averaged over
model domain
OBS (CRU)
GCM-Historical
GCM –RCP4.5
GCM –RCP8.5
RCM -Historical
RCM –RCP4.5
RCM – RCP8.5
Difference-Historical
Difference –RCP4.5
Difference –RCP8.5
 RCM tends to underestimate warming trend
Time series of CO2 concentration in RCP scenarios
1000
900
800
700
hist
600
MiniCAM - RCP 4.5
MESSAGE - RCP 8.5
500
400
300
200
1850
1900
1950
2000
2050
2100
 RCM are using constant value of CO2 concentration with concentration for 1985
 Underestimation of warming trend is seems to be due to lack of increase of green house gases.
Climate change Projection: Precipitation
 Time series of annual mean precipitation averaged over model
domain
OBS (CRU)
GCM-Historical
GCM –RCP4.5
GCM –RCP8.5
RCM -Historical
RCM –RCP4.5
RCM – RCP8.5
Difference-Historical
Difference –RCP4.5
Difference –RCP8.5
 Inter-annual variability of both GCM RCM is weak.
Climate change Projection: Anomalies
 Reference period: 1971-2000
OBS (CRU)
GCM-Historical
GCM –RCP4.5
GCM –RCP8.5
RCM -Historical
RCM –RCP4.5
RCM – RCP8.5
 It is clear that RCM tends to underestimate warming trend.
Climate change projection: Temperature
Change (RCP4.5)
Change (RCP8.5)
RCM
GCM
Current
Change (2070-2099)
Current climate
(1971-2000)
RCP4.5
RCP8.5
GCM
18.51 ℃
2.80 ℃
4.87 ℃
RCM
18.13 ℃
2.69 ℃
4.62 ℃
Climate change projection: Precipitation
Change (RCP4.5)
Change (RCP8.5)
RCM
GCM
Current
Change (2070-2099)
Current climate
(1970-2000)
RCP4.5
RCP8.5
GCM
4.84 mm/day
8.29 %
9.27 %
RCM
5.24 mm/day
6.24 %
7.43 %
Summary 1
Overall, performance of HadGEM3-RA on current climate
simulation is similar to HadGEM2-AO.
 However, HadGEM3-RA could resolve small-scale features
related with topography and coastline.
General patterns of regional climate change projection by
HadGEM3-RA is similar to projection by HadGEM2-AO.
But, HadGEM3-RA tends to underestimate warming trend due to
lack of increase of green house gases.
Evaluation of current climate simulation
in GCM forcing run (12.5km res)
- surface climate
Climatology (1971-200): Precipitation
Annual
Observation
GCM
RCM
 RCM could resolve
small-scale features
related with topography
and coastlines.
Winter
Summer
 RCM of 12.5km-res is
better than not only
GCM but also RCM of
50km-res.
Climatology (1971-200): Temperature
Annual
Observation
GCM
RCM
 RCM could resolve
small-scale features
related with topography
and coastlines.
Winter
Summer
 RCM of 12.5km-res is
better than not only
GCM but also RCM of
50km-res.
Bias: Precipitation and Temperature
Summer
RCM
Winter
GCM
 Temperature
Annual
 Precipitation
GCM
RCM
Statistics: Precipitation & Temperature (Land)
 Mean, bias, Root-mean-squared error (RMSE) and pattern correlation
coefficient of precipitation and temperature. (Ref. APHRO and CRU)
Variables
Precip.
(mm/day)
Temp.
(°C)
Mean
(OBS)
ANN
2.28
JJA
4.50
DJF
0.83
ANN
8.68
JJA
21.93
DJF
-5.90
Mean
(Model)
Bias
RMSE
Pattern
Corr.
GCM
2.48
0.20
0.42
0.936
RCM
2.58
0.30
0.46
0.938
GCM
4.27
-0.23
0.84
0.830
RCM
4.78
0.28
1.12
0.723
GCM
1.23
0.40
0.45
0.894
RCM
1.13
0.30
0.36
0.916
GCM
8.04
-0.64
1.38
0.954
RCM
7.95
-0.73
0.91
0.986
GCM
22.87
0.94
1.40
0.904
RCM
22.73
0.80
1.20
0.935
GCM
-8.93
-3.03
3.80
0.948
RCM
-8.93
-3.03
3.04
0.986
 Overall, RCM show better performance than GCM. But, RCM shows wet/cold biases.
Annual cycle of Precipitation and temperature
 30-yr mean annual cycle of area-averaged precipitation and surface
air temperature (1971~2000)
Precip.
Temp
 Black: OBS
 Red: GCM
 Blue: RCM
Corr.
Precip
Temp
GCM
0.951
0.999
RCM
0.963
0.999
Probability of daily precipitation
 The probability of daily precipitation with thresholds up to 50 mm/day
100
Observation
Probability (%)
GCM
RCM
10
1
0.1
(0.1-10)
(10-20)
(20-30)
(30-50)
(50~)
Thresholds (mm/day)
 RCM simulated probability is much more realistic than GCM simulation.
 RCM projections of changes in extremes in the future are likely to be very different
to, and much more credible than, those from GCMs.
Climate change projection (12.5km)
•Change in surface air temperature and precipitation
Climate change projection: Temperature
 Time series of annual mean surface air temperature averaged over
model domain
OBS (CRU)
GCM-Historical
GCM –RCP4.5
GCM –RCP8.5
RCM -Historical
RCM –RCP4.5
RCM – RCP8.5
Difference-Historical
Difference –RCP4.5
Difference –RCP8.5
 RCM tends to underestimate warming trend
 Underestimation of warming trend is seems to be due to lack of increase of green house gases.
Climate change projection: Precipitation
 Time series of annual mean precipitation averaged over model
domain
OBS (CRU)
GCM-Historical
GCM –RCP4.5
GCM –RCP8.5
RCM -Historical
RCM –RCP4.5
RCM – RCP8.5
Difference-Historical
Difference –RCP4.5
Difference –RCP8.5
 Inter-annual variability of RCM is similar to observation.
Climate change Projection: Anomalies
 Reference period: 1971-2000
OBS (CRU)
GCM-Historical
GCM –RCP4.5
GCM –RCP8.5
RCM -Historical
RCM –RCP4.5
RCM – RCP8.5
 It is clear that RCM tends to underestimate warming trend.
Climate change projection: Temperature
Change (RCP4.5)
Change (RCP8.5)
RCM
GCM
Current
Change (2070-2099)
Current climate
(1971-2000)
RCP4.5
RCP8.5
GCM
11.19
3.51
6.13
RCM
11.14
3.33
5.79
Climate change projection: Precipation
Change (RCP4.5)
Change (RCP8.5)
RCM
GCM
Current
Change
Current climate
(1970-2000)
RCP4.5
RCP8.5
GCM
3.53
11.70
14.96
RCM
3.17
11.83
17.90
Summary 2
Overall, performance of HadGEM3-RA on current climate
simulation is better than HadGEM2-AO.
 HadGEM3-RA could resolve small-scale features related with
topography and coastline.
And, HadGEM3-RA reproduced climate extreme better than
HadGEM2-AO.
General patterns of regional climate change projection by
HadGEM3-RA is similar to projection by HadGEM2-AO.
But, HadGEM3-RA tends to underestimate warming trend due to
lack of increase of green house gases.
Future plan
New downscaling experiments will be performed with all RCP
scenarios (RCP2.6/4.5/6.0/8.5) including prescribed green house gases.
Task:
Regional downscaling
ECSK06
Key milestones:
HS Kang, S Park
R Jones
Jul 2008: agree a visit (3~6 months) of a KMA expert to set up and test the HadGEM regional model for using
ERA-40
Oct 2008: initiate a visit to work on the regional model
Dec 2008: joint report on implementation of UM-regional model for downscaling over the East Asian region
(0.2FTE)
Jun 2009: report on evaluation of the UM-regional model using perfect boundary conditions on seasonal
simulations of East Asian monsoon activity (0.3FTE)
Jun 2010: progress report on evaluation of HadGEM3-RA over the East Asian region in long-term simulations
Dec 2011: report on evaluation of HadGEM3-RA with a focus on climate variability in long-term integrations
using ECMWF interim reanalysis data, associated with CORDEX participation (0.3FTE)
Dec 2011: report on assessment of East Asian climate downscaled by HadGEM3-RA using global climate
change projections, associated with CORDEX participation (0.3FTE)
Dec 2012: report on assessment of regional climate over Korea peninsular downscaled by HadGEM3-RA with
using global climate change projections of RCP 2.6/4.5/6.0/8.5 (0.3FTE)
Thank you very much!
Precipitation: Annual mean climatology
 Climatology of annual precipitation
Observation
GCM bias
GCM
RCM effect
RCM
RCM bias
Precipitation: JJA mean climatology
 Climatology of summer precipitation
Observation
GCM bias
GCM
RCM effect
RCM
RCM bias
Large-scale field: 500-hPa height (JJA)
Observation
GCM bias
GCM
RCM effect
 Both GCM and RCM enhanced upper trough.
RCM
RCM bias
Low level circulation: SLP, 850-hPa wind/humidity
Observation
GCM bias
GCM
RCM effect
 Both GCM has cyclonic anomalies over East Asian monsoon region.
 And, RCM enhanced cyclonic anomalies .
RCM
RCM bias
Download