1. BACKGROUND Bias Correction Method For Climate … · validated at the Pampanga, Angat and Kaliwa...
1
1. BACKGROUND Global Climate Models (GCM) precipitation are often characterized by biases and coarse resolution that limit direct application for basin sacle hydrological modeling. The biases in GCM are a low number of dry days which are compensated by too much drizzle, a bias in the mean and the inability to reproduce the observed heavy rainfall event. To address the bias of GCM, most of the climate studies couple bias correction and downscaling scheme. This study describes how to select appropriate GCM over the target region and a complete Cho Thanda Nyunt 1 , Toshio Koike 1 , Pactricia Ann Jaranilla Sanchez 1 , Akio Yamamoto 2 , Toshihoro Nemoto 2 , Masaru Kitsuregawa 2 1 Department of Civil Engineering, River and Environmental Laboratory, University of Tokyo (Bunkyo-ku, Tokyo 113-8656, Japan) Paper No : H21F-1237 Bias Correction Method For Climate Change Impact Assessment at a Basin Scale describes how to select appropriate GCM over the target region and a complete, efficient and comprehensive statistical bias correction method covering extreme rainfall, normal rainfall and frequency of dry days. The method is applied and validated at the Pampanga, Angat and Kaliwa basins in Philippines. Residents of the Manila metropolitan area rely on Angat Dam for 97% of their water resources and detailed analysis of future climate change impacts on these basins are urgently required for irrigation efficiency and future water supply balancing in Metro Manila. Department of Civil Engineering, River and Environmental Laboratory, University of Tokyo (Bunkyo ku, Tokyo 113 8656, Japan) 2 Institute of Industrial Science, University of Tokyo (Meguro-ku, Tokyo 153-8505, Japan) [email protected] 4. RESULTS Validate extreme value in sorting (1981-2000) Validate extreme frequency analysis (1981-2000) Validate long term climatology inter annual variability (1981 2000) 121°0'0"E 120°0'0"E '0"N '0"N ± Location of Pampanga, Angat and Kaliwa Basins 3. STATISTICAL BIAS CORRECTION Sorting data decreasing order Assign equal total rain days as observed station in GCM Define zero correction threshold ! ( 8 12 6 4 Miroc32_hires August spatial distribution (1981-2000) ! ( 6 12 4 2 28 18 18 Observed August spatial distribution (1981-2000) Validate long term climatology inter-annual variability (1981-2000) Analyze recurrence level of extreme precipitation (2046-2065) Analyze total number of dry days and continuous dry spell (2046-2065) !( ! ( !( ! ( !( ! ( h Balungao ClsuMunoz BalerRadar BaguioCity DagupanCity Pantabangan Dam 17°0 17°0 16°0'0"N 16°0'0"N ± Pampanga Define zero correction threshold Define extreme threshold from top rank Extract extremes events Check sensitivity ξ (be high enough to converge to the GPD fitting and low enough for determination of κ and α) Lowest of AMS of observed is defined as ξ Analyze same number of extremes in GCM gridded series Fit GCM to GPD by itself Make correction by inverse function observed GPD Extract monthly normal rainfall between extreme and zero correction Before correction After correction ! ( ! ( ! ( ! ( ! ( ! ( ! ( ! ( ! ( ! ( ! ( ! ( ! ( ! ( ! ( ! ( ! ( ! ( 14 16 2 10 18 20 22 28 26 24 30 30 10 6 14 16 12 14 14 10 18 12 12 18 ! ( ! ( ! ( ! ( ! ( ! ( ! ( ! ( ! ( ! ( ! ( ! ( ! ( ! ( ! ( ! ( ! ( ! ( 8 14 16 18 10 22 24 20 26 0 28 30 12 12 16 14 30 16 12 18 2 8 6 10 GCM biases 0 50 100 150 200 1 11 21 31 41 51 rainfall mm/day Rank Extreme rainfall underestimation CLSU csiromk3_0 miroc32_medres gfdl_cm2_0 gfdl_cm2_1 giss_aom ingv_echam4 GCM seasonal rainfall pattern 200 300 mm/day Extreme rainfall before correction CLSU csiromk3_0 miroc32_medres gfdl_cm2_0 gfdl_cm2_1 ingv echam4 200 300 mm/day Extreme rainfall after correction CLSU csiromk3_0 miroc32_medres gfdl_cm2_0 gfdl_cm2_1 ingv echam4 ! ( !( ! ( ! ( !( ! ( ! ( ! ( ! ( !( ! ( ! ( ! ( ! ( h Iba Tabak Maputi Matulid Infanta Talaguio Angat Dam Cabanatuan SibulSpring SanFernando BaiMagalang ScienceGarden CuyambayTanay HaciendaLuisita 15°0'0"N 15°0'0"N Pampanga Angat Kaliwa h Dam threshold Gamma CDF mapping for correction Same transfer function applied future projection Extreme Rainfall Bias Correction by AMS Define lowest of AMS as threshold of extreme Fitting log-normal or Gumbel by plotting position method Find the inverse function of in-situ station as corrected extreme C td l hi it t ti ’ t l ! ( ! ( 10 8 18 Rainfall mm/day High : 31.7533 Low : 0.466246 ! ( ! ( 20 14 4 4 1 2 6 Rainfall mm/day High : 33.0755 Low : -3.28925 0 5 10 15 20 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec rainfall mm/day CLSU csiromk3_0 miroc32_medres gfdl_cm2_0 gfdl_cm2_1 giss_aom ingv_echam4 5 10 15 all mm/day Low intensity long drizzel rain day CLSU csiromk3_0 miroc32_medres gfdl_cm2_0 fdl 21 10 15 20 all mm/day GCM Raw monthly distribution CLSU csiromk3_0 miroc32_medres gfdl_cm2_0 gfdl_cm2_1 10 15 20 m/day Biascorrected monthly distribution CLSU csiromk3_0 miroc32_medres gfdl_cm2_0 gfdl_cm2_1 ingv echam4 0 100 1 11 21 31 41 51 rainfall m Rank _ ipsl_cm4 0 100 1 11 21 31 41 51 rainfall m Rank ingv_echam4 ipsl_cm4 In the study region, we have a sufficient number of rain gauges within or near the basin, so the bias correction can be applied to each station. This method is a t pe of statistical bias correction and spatial disaggregation It has 2. GCM SELECTION 21 in-situ stations data provided by Philippines Atmospheric, Geophysical d At i l S i Ad iit ti (PAGASA) 5. SUMMARY Six GCMs were selected According to their performances over Luzon Island in the Philippines Reduce the uncertainty of single GCM use. Address various bias components associated with GCM output ! ( ! ( Tayabas Ambulong 121°0'0"E 120°0'0"E 14°0'0"N 14°0'0"N 0 30 60 15 Kilometers !( Rain gauges h Dam Corrected value reach in-situ stations’ extreme values Big discrepancy in frequencies 1981-2000 2046-2065 0 5 1500 1800 2100 2400 2700 3000 rainfa Rank gfdl_cm2_1 ingv_echam4 ipsl_cm4 300 400 500 m/day Frequency analysis 2046-2065 CLSU csiromk3_0 miroc32_medres gfdl_cm2_0 0 5 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec rainfa ingv_echam4 ipsl_cm4 0 5 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec rainfall mm ingv_echam4 ipsl_cm4 0 50 100 150 200 250 1 21 41 61 81 101 121 141 161 181 rainfall mm/day Rank Extremes corrected by AMS CLSU csiromk3_0 miroc32_medres gfdl_cm2_0 gfdl_cm2_1 ingv_echam4 ipsl_cm4 0 50 100 150 200 250 1 21 41 61 81 101 121 141 161 181 rainfall mm/day Rank Raw Extreme by AMS CLSU csiromk3_0 miroc32_medres gfdl_cm2_0 gfdl_cm2_1 ingv_echam4 ipsl_cm4 method is a type of statistical bias correction and spatial disaggregation. It has been developed and validated for both intensity at a point and basin scale for climate change impact study. 400 500 m/day Frequency analysis 1981-2000 CLSU csiromk3_0 miroc32_medres gfdl_cm2_0 Rain Bias Correction Method Extreme Generalized Pareto Distribution Non every year statistics Long or short tailed fitting Peak over threshold series and Astronomical Services Administration (PAGASA) Coupled Model Inter-comparison Project 3 (CMIP3) GCM gridded daily data 1981-2000( the control period) 2046-2065 (SRESA1B.) Spatial correlation and root mean square error (Reference JRA-25 reanalysis data and GPCP ) Address various bias components associated with GCM output (extreme rainfall, normal rainfall and total frequency of dry days) Biases remove well by fitting the GPD. Validate extreme rainfall frequency analysis against observation Estimate the maximum intesity of different return periods in near future. Solve frequency of dry days bias by sorting and matching rank Two-parameter gamma distribution mapping monthly normal rainfall An effective, simple and comprehensive bias correction method was developed for climate change impact assessment of basin scale hydrology, and 20°N 20 o N 0 ) ( - 1 - 1 ) ( 1 ≠ ⎥ ⎦ ⎤ ⎢ ⎣ ⎡ − = ⎞ ⎛ κ α ξ κ ξ κ x x x F 0 100 200 300 1 10 100 1000 rainfall mm Return years gfdl_cm2_1 ingv_echam4 ipsl_cm4 200 300 of days Number of continuous no rain day 1986-2000 2051-2065 6000 er of days Total number of dry days 1981-2000 2045-2065 0 100 200 300 1 10 100 1000 rainfall mm Return years gfdl_cm2_1 ingv_echam4 ipsl_cm4 No rain day Ranking order statistics frequency of no rain day in GCM is same as station developed for climate change impact assessment of basin scale hydrology, and for other impacts. Moreover, the method can be analyzed by discharge flow in further hydrological impact studies. Finally, the bias correction method should be proven by applying to different climate regions. Selected GCMs No GCMs Country Scores 1 fdl 2 1 USA 7 0°N 160°E 80°E Large Scale area considered 115 o E 130 o E 10 o N Local Scale PDS) of limit (lower parameter location parameter scale ) 2 1 2 1 (- parameter shape 0 1 ) ( = = < < = = = ⎟ ⎠ ⎞ ⎜ ⎝ ⎛ − − ξ α κ κ κ α α ξ x e x F 5 10 15 ll mm/day Low intensity long drizzel rain day CLSU csiromk3_0 miroc32_medres gfdl_cm2_0 6. REFERENCES 1) Nyunt, C.T. et al: Bias correction method for climate change impact assessments in the Philippines, JSCE, revised, 2013. 2) C Piani J O Haerter E Coppola Statistical bias correction for daily precipitation in regional PARAMETER Precipitation 0 100 CLSUcsiromk3_0 miroc32_medres gfdl_cm2_0 gfdl_cm2_1 ingv_echam4 ipsl_cm4 number o 2000 4000 CLSUcsiromk3_0 miroc32_medres gfdl_cm2_0 gfdl_cm2_1 ingv_echam4 ipsl_cm4 numbe 100 150 200 change % Change in precipitation % w.r.t 1981‐2000 csiromk3_0 miroc32_medres gfdl_cm2_0 gfdl cm2 1 40 60 80 100 ion change (%) Change in standard deviation % w.r.t 1981‐2000 csiromk3_0 miroc32_medres gfdl_cm2_0 gfdl cm2 1 less than no rain day threshold change zero rainfall. Normal Gamma Distribution monthly CDF Gamma mapping 1 gfdl_cm2_1 USA 7 2 gfdl_cm2_0 USA 6 3 ipsl_cm4 France 5 4 ingv_echam4 Italy 4 5 csiro_mk3_0 Australia 3 6 miroc3_2_medres Japan 2 0 5 1500 1800 2100 2400 2700 3000 rainfa Rank gfdl_cm2_1 ingv_echam4 ipsl_cm4 GCM=0 No rain day Sorting ACKNOWLEDMENT 2) C. Piani.J.O.Haerter.E. Coppola. Statistical bias correction for daily precipitation in regional climate models over Europe. Theor Appl Climatol 99: 187-192, 2010. 3) IPCC: Contribution of Working Group I to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change, edited by: Solomon, S., Qin, D., Manning, M., Chen, Z., Miller H. L., Cambridge University Press, Cambridge, United Kingdom and New York, 2007a. 4) M.E.Elahamy, I.A. Seierstad and A. Sorteber. Impacts of climate change on Blue Nile flows using bias-corrected GCM scenarios. Hydrol. Earth Syst. Sci., 13,551-565, 2009. 5) IPCC : The regional impacts of Climate Change: An Assessment of Vulnerability, edited by Watson R.T., Moss R.H., Zinyowera M. C., Dokken D. J., A Special report of IPCC Working Group II, 1997. 6) Gupta VK, Duckstein L.A stochastic analysis of extreme droughts. Water Resour. Res., 11:221-228. 1975. The change in seasonal rainfall (%) for three seasons with respect to 1981-2000, 2046-2065 minus 1981-2000. DJF is December-January-February, MAM is March-April-May and JJASON is wet season from June to November. For CLSU station, winter show maximum increase in precipitation and it may be due to global warming trend in winter There is a small decrease in the monsoon Precipitation Sea level pressure Sea surface temperature Meridonial wind Zonal wind OLR Air Temperature ‐100 ‐50 0 50 DJF MAM JJASON Precipitation gfdl_cm2_1 ingv_echam4 ipsl_cm4 Average ‐60 ‐40 ‐20 0 20 DJF MAM JJASON standard deviati _ _ ingv_echam4 ipsl_cm4 Average to monthly CDF of station inverse of Gamma CDF in each month is corrected rain GCM No rain day threshold Extreme threshold Normal rain ACKNOWLEDMENT This research was implemented as a part of a project of the Japan International Cooperation Agency (JICA), "The Study of Water Security Master Plan for Metro Manila and its Adjoining Area" by using the Data Integration and Analysis System (DIAS) funded by the Ministry of Education, Culture, Sports, Science and Technology (MEXT) of Japan.. We are grateful to the EDITORIA science team for permission to use the GCM time series dataset of the Data Integration and Analysis System (DIAS) project. 7) Lana S, Burgueno A. Spatial and temporal characterization of annual extreme droughts in Catalonia (NE Spain). International Journal of Climatology 18: 93-110. 1998b. 8) Bobee B, Rasmussen P. Recent advances in flood frequency analysis. US National Report of IUGG, 1991-1994, 1995. 9) Hosking, J.R. M., and J.R. Wallis. Parameter and quantile estimation for the generalized Pareto distribution. Techno metrics 29(3): 339-349, 1987. 10) Amor V.M. Ines, James W. Hansen. Bias correction of daily GCM rainfall for crop simulation studies. Agricultural and Forest Meteorology y138: 44-53, 2006. due to global warming trend in winter . There is a small decrease in the monsoon season and small increase in rainy season by averaging all GCM as in Figure below. Another figure compares the change signal by the standard deviation change (%) of each season with respect to 1981-2000. It is also prominent increases in DJF and same trend as in change in precipitation. In the rainy season, all GCM agree in change of precipitation increase and standard deviation change also increase means seasonal variation also robust.
Transcript of 1. BACKGROUND Bias Correction Method For Climate … · validated at the Pampanga, Angat and Kaliwa...
1. BACKGROUNDGlobal Climate Models (GCM) precipitation are often characterized by
biases and coarse resolution that limit direct application for basin saclehydrological modeling. The biases in GCM are a low number of dry days whichare compensated by too much drizzle, a bias in the mean and the inability toreproduce the observed heavy rainfall event. To address the bias of GCM, most ofthe climate studies couple bias correction and downscaling scheme. This studydescribes how to select appropriate GCM over the target region and a complete
Cho Thanda Nyunt1, Toshio Koike1, Pactricia Ann Jaranilla Sanchez1, Akio Yamamoto2, Toshihoro Nemoto2, Masaru Kitsuregawa2
1 Department of Civil Engineering, River and Environmental Laboratory, University of Tokyo (Bunkyo-ku, Tokyo 113-8656, Japan)
Paper No : H21F-1237
Bias Correction Method For Climate Change Impact Assessment at a Basin Scale
describes how to select appropriate GCM over the target region and a complete,efficient and comprehensive statistical bias correction method covering extremerainfall, normal rainfall and frequency of dry days. The method is applied andvalidated at the Pampanga, Angat and Kaliwa basins in Philippines. Residents ofthe Manila metropolitan area rely on Angat Dam for 97% of their water resourcesand detailed analysis of future climate change impacts on these basins areurgently required for irrigation efficiency and future water supply balancing inMetro Manila.
Department of Civil Engineering, River and Environmental Laboratory, University of Tokyo (Bunkyo ku, Tokyo 113 8656, Japan)2 Institute of Industrial Science, University of Tokyo (Meguro-ku, Tokyo 153-8505, Japan)
4. RESULTSValidate extreme value in sorting (1981-2000)Validate extreme frequency analysis (1981-2000)Validate long term climatology inter annual variability (1981 2000)
121°0'0"E120°0'0"E
'0"N
'0"N
±Location of Pampanga, Angat and Kaliwa Basins
3. STATISTICAL BIAS CORRECTIONSorting data decreasing orderAssign equal total rain days as observed station in GCMDefine zero correction threshold
!(
8
12
6 4
Miroc32_hires August spatial distribution (1981-2000)
!( 612 4 228 1818
Observed August spatial distribution (1981-2000)
Validate long term climatology inter-annual variability (1981-2000)Analyze recurrence level of extreme precipitation (2046-2065)Analyze total number of dry days and continuous dry spell (2046-2065)
!(
!(
!(
!(
!(
!(h
Balungao
ClsuMunoz BalerRadar
BaguioCity
DagupanCity
Pantabangan Dam
17°0
17°0
16°0
'0"N
16°0
'0"N
±Pampanga
Define zero correction threshold Define extreme threshold from top rankExtract extremes eventsCheck sensitivity ξ (be high enough to converge to the GPD fitting and low enough for determination of κ and α)Lowest of AMS of observed is defined as ξAnalyze same number of extremes in GCM gridded seriesFit GCM to GPD by itself Make correction by inverse function observed GPDExtract monthly normal rainfall between extreme and zero correction
Before correction After correction
!(
!(
!(
!(!(
!(
!(
!(
!(
!(
!(
!(
!(
!(
!(
!(
!(
!(
1416
2
10
1820
22
2826
24
30
30
10
6
14
16
12
14
14
10
18
12
12
18
!(
!(
!(
!(!(
!(
!(
!(
!(
!(
!(
!(
!(
!(
!(
!(
!(
!(
8
1416
18
10
2224
20
26
0
2830
12
12
16
14
30
16
12
18
2
8
6
10
GCM biases
0
50
100
150
200
1 11 21 31 41 51
rainfall mm/day
Rank
Extreme rainfall underestimation CLSUcsiromk3_0miroc32_medresgfdl_cm2_0gfdl_cm2_1giss_aomingv_echam4
GCM seasonal rainfall pattern
200
300
mm
/day
Extreme rainfall before correction CLSUcsiromk3_0miroc32_medresgfdl_cm2_0gfdl_cm2_1ingv echam4
200
300
mm
/day
Extreme rainfall after correctionCLSUcsiromk3_0miroc32_medresgfdl_cm2_0gfdl_cm2_1ingv echam4
!(
!(
!(
!(
!(
!(
!(
!(
!(
!(
!(
!(
!(
!(
h
Iba
Tabak
Maputi
Matulid
Infanta
Talaguio
Angat Dam
Cabanatuan
SibulSpring
SanFernando
BaiMagalang
ScienceGarden
CuyambayTanay
HaciendaLuisita
15°0
'0"N
15°0
'0"N
Pampanga
Angat
Kaliwa
h Dam
ythresholdGamma CDF mapping for correctionSame transfer function applied future projection
Extreme Rainfall Bias Correction by AMSDefine lowest of AMS as threshold of extremeFitting log-normal or Gumbel by plotting position methodFind the inverse function of in-situ station as corrected extremeC t d l h i it t ti ’ t l
!(
!(
10
8
18
Rainfall mm/dayHigh : 31.7533
Low : 0.466246
!(
!(
20
14
44
1
2
6
Rainfall mm/dayHigh : 33.0755
Low : -3.28925
0
5
10
15
20
Jan
Feb
Mar
Apr
May Jun
Jul
Aug Sep
Oct
Nov Dec
rainfall mm/day
p
CLSUcsiromk3_0miroc32_medresgfdl_cm2_0gfdl_cm2_1giss_aomingv_echam4
5
10
15
all m
m/day
Low intensity long drizzel rain dayCLSUcsiromk3_0miroc32_medresgfdl_cm2_0fdl 2 1
10
15
20
all m
m/d
ay
GCM Raw monthly distribution CLSU
csiromk3_0
miroc32_medres
gfdl_cm2_0
gfdl_cm2_110
15
20
m/d
ay
Biascorrected monthly distribution CLSU
csiromk3_0
miroc32_medres
gfdl_cm2_0
gfdl_cm2_1
ingv echam4
0
100
1 11 21 31 41 51
rain
fall
m
Rank
g _ipsl_cm4
0
100
1 11 21 31 41 51
rain
fall
m
Rank
ingv_echam4ipsl_cm4
In the study region, we have a sufficient number of rain gauges withinor near the basin, so the bias correction can be applied to each station. Thismethod is a t pe of statistical bias correction and spatial disaggregation It has
2. GCM SELECTION21 in-situ stations data provided by Philippines Atmospheric, Geophysicald A t i l S i Ad i i t ti (PAGASA)
5. SUMMARYSix GCMs were selected According to their performances over Luzon Island in the PhilippinesReduce the uncertainty of single GCM use.Address various bias components associated with GCM output
!(
!(
TayabasAmbulong
121°0'0"E120°0'0"E
14°0
'0"N
14°0
'0"N
0 30 6015 Kilometers
!( Rain gaugesh Dam Corrected value reach in-situ stations’ extreme values
Big discrepancy in frequencies
1981-2000 2046-2065
0
5
1500 1800 2100 2400 2700 3000
rainfa
Rank
gfdl_cm2_1ingv_echam4ipsl_cm4
300
400
500
m/d
ay
Frequency analysis 2046-2065CLSUcsiromk3_0miroc32_medresgfdl_cm2_0
0
5
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
rain
fa ingv_echam4
ipsl_cm4
0
5
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
rain
fall
mm ingv_echam4
ipsl_cm4
0
50
100
150
200
250
1 21 41 61 81 101 121 141 161 181
rainfall mm/day
Rank
Extremes corrected by AMS
CLSU
csiromk3_0
miroc32_medres
gfdl_cm2_0
gfdl_cm2_1
ingv_echam4
ipsl_cm4
0
50
100
150
200
250
1 21 41 61 81 101 121 141 161 181
rainfall mm/day
Rank
Raw Extreme by AMS
CLSU
csiromk3_0
miroc32_medres
gfdl_cm2_0
gfdl_cm2_1
ingv_echam4
ipsl_cm4
method is a type of statistical bias correction and spatial disaggregation. It hasbeen developed and validated for both intensity at a point and basin scale forclimate change impact study.
400
500
m/d
ay
Frequency analysis 1981-2000CLSUcsiromk3_0miroc32_medresgfdl_cm2_0
Rain Bias Correction MethodExtreme Generalized Pareto Distribution
Non every year statisticsLong or short tailed fittingPeak over threshold series
and Astronomical Services Administration (PAGASA)Coupled Model Inter-comparison Project 3 (CMIP3) GCM gridded daily data1981-2000( the control period)2046-2065 (SRESA1B.)Spatial correlation and root mean square error
(Reference JRA-25 reanalysis data and GPCP )
Address various bias components associated with GCM output(extreme rainfall, normal rainfall and total frequency of dry days)
Biases remove well by fitting the GPD.Validate extreme rainfall frequency analysis against observationEstimate the maximum intesity of different return periods in near future. Solve frequency of dry days bias by sorting and matching rankTwo-parameter gamma distribution mapping monthly normal rainfall
An effective, simple and comprehensive bias correction method was developed for climate change impact assessment of basin scale hydrology, and20°N
20oN0 )(-1-1)(
1
≠⎥⎦⎤
⎢⎣⎡ −
=
⎞⎛
καξκ
ξ
κ
x
xxF
0
100
200
300
1 10 100 1000
rain
fall
mm
Return years
gfdl_cm2_1ingv_echam4ipsl_cm4
200
300
of d
ays
Number of continuous no rain day
1986-20002051-2065
6000
er o
f day
s
Total number of dry days
1981-2000
2045-2065
0
100
200
300
1 10 100 1000
rain
fall
mm
Return years
gfdl_cm2_1ingv_echam4ipsl_cm4
No rain day
Ranking order statisticsfrequency of no rain day
in GCM is same as station
developed for climate change impact assessment of basin scale hydrology, and for other impacts. Moreover, the method can be analyzed by discharge flow in further hydrological impact studies. Finally, the bias correction method should be proven by applying to different climate regions.
Selected GCMsNo GCMs Country Scores
1 fdl 2 1 USA 7
0°N160°E80°E
Large Scale area considered
115oE 130oE
10oN
Local Scale PDS) oflimit (lower parameter location parameter scale
)21
21(- parameter shape
0 1 )(
==
<<=
==⎟⎠⎞
⎜⎝⎛ −
−
ξα
κκ
κα
αξx
exF
5
10
15
ll mm/day
Low intensity long drizzel rain dayCLSUcsiromk3_0miroc32_medresgfdl_cm2_0
6. REFERENCES1) Nyunt, C.T. et al: Bias correction method for climate change impact assessments in the
Philippines, JSCE, revised, 2013.2) C Piani J O Haerter E Coppola Statistical bias correction for daily precipitation in regional
PARAMETERPrecipitation
0
100
CLSUcsiromk3_0miroc32_medresgfdl_cm2_0gfdl_cm2_1ingv_echam4ipsl_cm4
num
ber o
2000
4000
CLSUcsiromk3_0miroc32_medresgfdl_cm2_0gfdl_cm2_1ingv_echam4ipsl_cm4
num
be
100
150
200change
%Change in precipitation %
w.r.t 1981‐2000
csiromk3_0
miroc32_medresgfdl_cm2_0gfdl cm2 1 40
60
80
100
ion change
(%)
Change in standard deviation % w.r.t 1981‐2000
csiromk3_0
miroc32_medres
gfdl_cm2_0
gfdl cm2 1
less than no rain day threshold change zero rainfall.
Normal Gamma Distributionmonthly CDF Gamma mapping
1 gfdl_cm2_1 USA 72 gfdl_cm2_0 USA 63 ipsl_cm4 France 54 ingv_echam4 Italy 45 csiro_mk3_0 Australia 36 miroc3_2_medres Japan 2
0
5
1500 1800 2100 2400 2700 3000
rainfa
Rank
gfdl_cm2_1ingv_echam4ipsl_cm4
GCM=0
No
rain
day
Sorting
ACKNOWLEDMENT
2) C. Piani.J.O.Haerter.E. Coppola. Statistical bias correction for daily precipitation in regional climate models over Europe. Theor Appl Climatol 99: 187-192, 2010.
3) IPCC: Contribution of Working Group I to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change, edited by: Solomon, S., Qin, D., Manning, M., Chen, Z., Miller H. L., Cambridge University Press, Cambridge, United Kingdom and New York, 2007a.
4) M.E.Elahamy, I.A. Seierstad and A. Sorteber. Impacts of climate change on Blue Nile flows using bias-corrected GCM scenarios. Hydrol. Earth Syst. Sci., 13,551-565, 2009.
5) IPCC : The regional impacts of Climate Change: An Assessment of Vulnerability, edited by Watson R.T., Moss R.H., Zinyowera M. C., Dokken D. J., A Special report of IPCC Working Group II, 1997.
6) Gupta VK, Duckstein L.A stochastic analysis of extreme droughts. Water Resour. Res., 11:221-228. 1975.
The change in seasonal rainfall (%) for three seasons with respect to1981-2000, 2046-2065 minus 1981-2000. DJF is December-January-February,MAM is March-April-May and JJASON is wet season from June to November.For CLSU station, winter show maximum increase in precipitation and it may bedue to global warming trend in winter There is a small decrease in the monsoon
PrecipitationSea level pressureSea surface temperatureMeridonial windZonal windOLRAir Temperature
‐100
‐50
0
50
DJF MAM JJASONPrecipitation gfdl_cm2_1
ingv_echam4ipsl_cm4Average
‐60
‐40
‐20
0
20
DJF MAM JJASON
standard deviati g _ _
ingv_echam4
ipsl_cm4
Average
y pp gto monthly CDF of station
inverse of Gamma CDF in each month is corrected rain
GCM
No
rain
day
thre
shol
d
Ext
rem
eth
resh
old
Normal rain
ACKNOWLEDMENTThis research was implemented as a part of a project of the Japan International
Cooperation Agency (JICA), "The Study of Water Security Master Plan for Metro Manila and itsAdjoining Area" by using the Data Integration and Analysis System (DIAS) funded by theMinistry of Education, Culture, Sports, Science and Technology (MEXT) of Japan.. We aregrateful to the EDITORIA science team for permission to use the GCM time series dataset of theData Integration and Analysis System (DIAS) project.
: . 97 .7) Lana S, Burgueno A. Spatial and temporal characterization of annual extreme droughts in
Catalonia (NE Spain). International Journal of Climatology 18: 93-110. 1998b.8) Bobee B, Rasmussen P. Recent advances in flood frequency analysis. US National Report
of IUGG, 1991-1994, 1995.9) Hosking, J.R. M., and J.R. Wallis. Parameter and quantile estimation for the generalized
Pareto distribution. Techno metrics 29(3): 339-349, 1987.10) Amor V.M. Ines, James W. Hansen. Bias correction of daily GCM rainfall for crop
simulation studies. Agricultural and Forest Meteorology y138: 44-53, 2006.
due to global warming trend in winter. There is a small decrease in the monsoonseason and small increase in rainy season by averaging all GCM as in Figurebelow. Another figure compares the change signal by the standard deviationchange (%) of each season with respect to 1981-2000. It is also prominentincreases in DJF and same trend as in change in precipitation. In the rainyseason, all GCM agree in change of precipitation increase and standarddeviation change also increase means seasonal variation also robust.
