Climate Model Biases and Statistical Downscaling for ... A. Phunga, and Bohumil M. Svomac...
23
Sagar Gautam a , Christine Costello a , Claire Baffaut b , Quang A. Phung a , and Bohumil M. Svoma c a Department of Bioengineering, University of Missouri, Columbia, MO 65201 b USDA–ARS, Cropping Systems and Water Quality Research Unit, Columbia, MO 65211 c Department of Soil, Environmental and Atmospheric Sciences, University of Missouri, Columbia, MO 65201 University of Missouri Department of Biological Engineering Climate Model Biases and Statistical Downscaling for Application in Hydrologic Model 1
Transcript of Climate Model Biases and Statistical Downscaling for ... A. Phunga, and Bohumil M. Svomac...
Sagar Gautama, Christine Costelloa, Claire Baffautb,
Quang A. Phunga, and Bohumil M. Svomac
aDepartment of Bioengineering, University of Missouri, Columbia, MO 65201bUSDA–ARS, Cropping Systems and Water Quality Research Unit, Columbia, MO 65211
cDepartment of Soil, Environmental and Atmospheric Sciences, University of Missouri, Columbia, MO 65201
University of Missouri
Department of Biological Engineering
Climate Model Biases and Statistical
Downscaling for Application in
Hydrologic Model
1
• Climate change and hydrology
• Global Circulation Model (GCM)
• Coupled Model Inter-comparison Project Phase 5 (CMIP5)• Representative Concentration Pathways (RCPS)
• Climate model bias
• Statistical downscaling of precipitation and temperature for hydrological modeling
• Delta Method
• Quantile mapping
• Results
• Conclusions
Outline
2
Introduction
• Climate change is expected to have significant impacts on global temperature and precipitation change
• Output from the climate model are used to predict the climate change impact on hydrology
• Coupling climate and hydro-model
• Scale of operation of climate model and the hydro-model are different
• Need some sort of correction
Downscaling3
Global Circulation Model(GCM)
1990
1995
2001
2007
AR52013
source: IPCC Reports: FAR (IPCC, 1990), SAR (IPCC, 1996), TAR (IPCC, 2001a), AR4 (2007), and AR5 (2013)
4
Resolution
~500 km
~250 km
~180 km
~110 km
~87.5 km
Representative Concentration
pathways(RCM)
Figure IPCC report Reports: AR5 (2013)
5
Table adopted from Moss et.al. 2010. and Rogelj et.al. 2012
Name Radiative forcing Concentration
(ppm)
Temp anomaly
(°C)
SRES temp
anomaly equiv
RCP8.5 >8.5 W m-2 in 2100 >1,370 CO2-
equiv.
4.9 SRES A1F1
RCP4.5 ~4.5 W m-2 at stabilization
after 2100
~650 CO2-equiv. 2.4 SRES B1
Estimating Climate Change Impact on
Hydrology
2. Global Circulation Model
1. GHG Emissions Scenario
3. Large scale application
4. Impact Model
Ad
apte
d f
rom
Cay
anan
d K
no
wle
s, S
CR
IPP
S/U
SGS,
20
03
6
CMIP5 data BCCA CMIP5 data Statistical Downscaling
Study Area
• Drainage Area: 73.4 Km2
• Experimental watershed with long-term hydrologic dataset and well-calibrated SWAT model
• Daily Bias Correction Constructed Analogs (BCCA) downscaled datasets (1/8 degree) were downloaded from http://gdo-dcp.ucllnl.org/
• Downscaled data showed bias when compared to observed data
• This bias can influence impact assessment
7
Location of Goodwater creek Experimental watershed in the state of MO
. outlet
Precipitation – Historical and
CMIP5
0
200
400
600
800
1000
1200
1400
1600
1800
197
0
197
8
198
6
199
4
200
2
201
0
201
8
202
6
203
4
204
2
205
0
205
8
206
6
Pre
cip
ita
tion
(m
m)
Year
bcc-csm1-1.1.rcp4.5 observed
0
200
400
600
800
1000
1200
1400
1600
1800
197
0
197
8
198
6
199
4
200
2
201
0
201
8
202
6
203
4
204
2
205
0
205
8
206
6
bcc-csm1-1.1.rcp8.5 observed
8
Yearly Max. Temperature –
Historical and CMIP5
10
12
14
16
18
20
22
24
197
1
197
9
198
7
199
5
200
3
201
1
201
9
202
7
203
5
204
3
205
1
205
9
206
7
Te
mp
era
ture
(°C
)
Year
gfdl-esm2g.1.rcp4.5 observed
9
10
12
14
16
18
20
22
24
197
1
197
8
198
5
199
2
199
9
200
6
201
3
202
0
202
7
203
4
204
1
204
8
205
5
206
2
206
9
gfdl-esm2g.1.rcp8.5 observed
Yearly Min. Temperature –
Historical and CMIP5
10
0
2
4
6
8
10
12
197
1
197
9
198
7
199
5
200
3
201
1
201
9
202
7
203
5
204
3
205
1
205
9
206
7
Te
mp
era
ture
(°C
)
Year
gfdl-esm2g.1.rcp4.5 observed
0
2
4
6
8
10
12
197
1
197
8
198
5
199
2
199
9
200
6
201
3
202
0
202
7
203
4
204
1
204
8
205
5
206
2
206
9
gfdl-esm2g.1.rcp8.5 observed
Downscaling
11
Simple Downscaling
• Applying coarse scale climate changes to the observed data
Dynamic Downscaling• Involves using numerical meteorological modeling to reflect
how global patterns affect local weather conditions
• Able to simulate local conditions in greater detail
• Still has some bias
• Need more computational power
Statistical Downscaling
• Based on stationary transfer function
eg. Delta method, Quantile mapping
Statistical Downscaling
12
Delta method:
• Linearly shifting every model data by scaling factor
• Monthly scale factor is calculated
Quantile Mapping:
• Calculate the quantiles (median, quartiles, etc.) for observed data and GCM historical data
• Apply linear transformation to each quantile of GCM data to adjust its range to match the quantile in observed data
• Apply same correction to the quantiles of the future data
Delta Method-Precipitation
13
0
200
400
600
800
1000
1200
1400
1600
1800
197
0
197
8
198
6
199
4
200
2
201
0
201
8
202
6
203
4
204
2
205
0
205
8
206
6
Pre
cip
ita
tion
(m
m)
Year
bcc-csm1-1.1.rcp4.5
observed
Delta_ bcc-csm1-1.1.rcp4.5
0
200
400
600
800
1000
1200
1400
1600
1800
197
0
197
8
198
6
199
4
200
2
201
0
201
8
202
6
203
4
204
2
205
0
205
8
206
6
bcc-csm1-1.1.rcp8.5
obsereved
Delta_ bcc-csm1-1.1.rcp8.5
Delta Method-Max. Temperature
14
10
12
14
16
18
20
22
24
197
1
197
8
198
5
199
2
199
9
200
6
201
3
202
0
202
7
203
4
204
1
204
8
205
5
206
2
206
9
Te
mp
era
ture
(°C
)
Year
gfdl-esm2g.1.rcp4.5
observed
Delta_gfdl-esm2g.1.rcp4.5
10
12
14
16
18
20
22
24
197
1
197
7
198
3
198
9
199
5
200
1
200
7
201
3
201
9
202
5
203
1
203
7
204
3
204
9
205
5
206
1
206
7
gfdl-esm2g.1.rcp8.5
observed
Delta_gfdl-esm2g.1.rcp8.5
• gfdl-esm2g.1rcp
Delta Method-Max. Temperature
15
10
12
14
16
18
20
22
24
197
11
97
71
98
31
98
91
99
52
00
12
00
72
01
32
01
92
02
52
03
12
03
72
04
32
04
92
05
52
06
12
06
7
Te
mp
era
ture
(°C
)
Year
gfdl-esm2g.1.rcp8.5
observed
Delta_gfdl-esm2g.1.rcp8.5
gfdl-esm2g.1rcp8.5
Delta_gfdl-esm2g.1rcp8.5
observed
Delta Method-Min. Temperature
16
0
2
4
6
8
10
12
197
1
197
8
198
5
199
2
199
9
200
6
201
3
202
0
202
7
203
4
204
1
204
8
205
5
206
2
206
9
gfdl-esm2g.1.rcp8.5
observed
Delta- gfdl-esm2m.1.rcp8.5
0
2
4
6
8
10
12
197
1
197
8
198
5
199
2
199
9
200
6
201
3
202
0
202
7
203
4
204
1
204
8
205
5
206
2
206
9
Te
mp
era
ture
(°C
)
gfdl-esm2g.1.rcp4.5
observed
Delta- gfdl-esm2m.1.rcp4.5
Quantile Mapping
17
Distribution of the precipitation data follows the gamma distribution.
observed gfdl-esm2g.1.rcp8.5 corrected CMIP5
Pre
cip
itation (
mm
)
Gamma Quantiles
Alpha=0.602
Scale=14.611
Gamma Quantiles
Alpha=0.356
Scale=8.726
Gamma Quantiles
Alpha=0.603
Scale=14.457
uncorrected CMIP5
Results-Quantile Mapping
18
0
200
400
600
800
1000
1200
1400
1600
1800
197
0
197
6
198
2
198
8
199
4
200
0
200
6
201
2
201
8
202
4
203
0
203
6
204
2
204
8
205
4
206
0
206
6
Pre
cip
ita
tion
(m
m)
Year
gfdl-esm2g.1.rcp4.5
observed
QM_gfdl-esm2g.1.rcp4.5
gfdl-esm2g.1rcp4.5
QM_gfdl-esm2g.1rcp4.5
observed
Results-Quantile Mapping
19
gfdl-esm2g.1rcp8.5
QM_gfdl-esm2g.1rcp8.5
observed
0
200
400
600
800
1000
1200
1400
1600
1800
197
0
197
6
198
2
198
8
199
4
200
0
200
6
201
2
201
8
202
4
203
0
203
6
204
2
204
8
205
4
206
0
206
6
Pre
cip
ita
tion
(m
m)
Year
gfdl-esm2g.1.rcp8.5
observed
QM_gfdl-esm2g.1.rcp8.5
Results
20
• Comparison of the daily
precipitation amount
shows that quantile
mapping work better
than delta method
gfdl-esm2g.1rcp8.5
Delta_gfdl-esm2g.1rcp8.5
QM_gfdl-esm2g.1rcp8.5
observed
Conclusions
21
• Even the best available downscaled BCCA CMIP5 data
has bias for its application for impact assessment in
smaller scale
• Delta method was able to correct the bias for daily
temperature data
• Quantile mapping turn out to be a better bias correction
method for daily precipitation data
• Daily max, yearly max and min value were well
represented
Funding of this project was provided by USDA-ARS (2014-2017)
• Dr. Christine Costello
• Dr. Claire Baffaut
• Dr. Bohumil M. Svoma
• Dr. Allen Thompson
• Dr. John Sadler
• Dr. Fessehaie Ghidey
• Mr. Quang Phung
Acknowledgement
22
