Comparison of statistical downscaling procedures for climate change impact assessment of water resources Henrik Madsen, Maria Sunyer, Keiko Yamagata DHI, Denmark HydroPredict 2010, 20-23 September 2010, Prague, Czech Republic
Downscaling Global climate model projections Downscaling • Dynamical • Statistical Local-scale impact assessment
Dynamical downscaling Regional climate model (RCM) Driven by GCM boundary • conditions Higher resolution (10-50 km) • Resolves sub-GCM grid scale • forcings in a physically-based way 7 Further statistical Observed 6 HIRHAM-ECHAM5 downscaling needed Mean [mm/day] 5 4 3 2 1 0 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
Statistical downscaling • Define relationship between large-scale model (GCM or RCM) and local climate • Methods based on Change Factor Methodology: Mean correction Mean and variance correction Weather Generators Statistics Statistics RCM Control RCM Future CF Statistics Statistics Future Observed
Statistical downscaling methods - Mean - Variance Change factors - Proportion of dry days - … Model fitting: Mean correction • -Mean (delta change) Model fitting: Mean and variance • -Mean correction -Variance Model fitting: Weather generators • Future Statistics -Mean -Variance Neyman-Scott -Skewness WG Rectangular Pulses -Dry-day prob. Markov Chain -Autocorrelation Fut sync TS -Transition prob. LARS WG
Case study – North Sealand Global Climate Model Regional Climate Model Dynamical downscaling Statistical [meter] downscaling 6220000 6210000 6200000 6190000 6180000 6170000 6160000 Asses climate change impacts 6150000 [meter] Above 100 90 - 100 6140000 80 - 90 70 - 80 on hydrology 60 - 70 6130000 50 - 60 40 - 50 30 - 40 6120000 20 - 30 10 - 20 0 - 10 North Sealand (3000 km 2 ) 6110000 -10 - 0 -20 - -10 -30 - -20 6100000 -40 - -30 Below -40 Undefined Value 620000 640000 660000 680000 700000 720000 [meter]
Methodology ENSEMBLES DATA OBSERVED DATA SET (1990-2008) A1B Scenario: • Precipitation • Temperature Driving GCM: ARPEGE ECHAM5 • Pot. evap. RCM: HIRHAM ALADIN HIRHAM REMO CHANGE FACTORS Mean Correction STATISTICAL Mean & Variance DOWNSCALING Correction Weather generator Future Time Series MIKE SHE Hydrological Model (Impact Assessment )
Methodology ENSEMBLES DATA OBSERVED DATA SET (1990-2008) A1B Scenario: • Precipitation • Temperature Driving GCM: ARPEGE ECHAM5 • Pot. evap. RCM: HIRHAM ALADIN HIRHAM REMO CHANGE FACTORS Mean Correction STATISTICAL Mean & Variance DOWNSCALING Correction Weather generator Future Time Series MIKE SHE Hydrological Model (Impact Assessment )
RCM compared to observations - precipitation Mean St.dev. Skewness Prop. dry days ALADIN-ARPEGE HIRHAM-ECHAM5 REMO-ECHAM5 HIRHAM-ARPEGE Observed
Methodology ENSEMBLES DATA OBSERVED DATA SET (1990-2008) A1B Scenario: • Precipitation • Temperature Driving GCM: ARPEGE ECHAM5 • Pot. evap. RCM: HIRHAM ALADIN HIRHAM REMO CHANGE FACTORS Mean Correction STATISTICAL Mean & Variance DOWNSCALING Correction Weather generator Future Time Series Mike SHE Hydrological Model (Impact Assessment )
Change factors precipitation (2070-2100) 1,50 1,50 Mean St.dev. 1,40 1,40 1,30 1,30 1,20 1,20 1,10 1,10 CF Mean CF StDev 1,00 1,00 0,90 0,90 0,80 0,80 0,70 0,70 0,60 0,60 0,50 0,50 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec 4,50 0,2 Skewness Prop. dry days 4,00 0,1 3,50 Absolute change Pdry ARPEGE 3,00 0,1 CF Skew ECHAM5 2,50 0,0 2,00 1,50 -0,1 1,00 0,50 -0,1 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
Methodology ENSEMBLES DATA OBSERVED DATA SET (1990-2008) A1B Scenario: • Precipitation • Temperature Driving GCM: ARPEGE ECHAM5 • Pot. evap. RCM: HIRHAM ALADIN HIRHAM REMO CHANGE FACTORS Mean Correction STATISTICAL Mean & Variance DOWNSCALING Correction Weather generator Future Time Series MIKE SHE Hydrological Model (Impact Assessment )
Statistical downscaling (2070-2100) Mean Corr. Observed Mean and Var Corr. HIRHAM-ECHAM5 SNSRP -WG 3,5 9 Mean St.dev. 8 3 Standard Deviation 7 Mean [mm/d] 2,5 6 [mm/d] 5 2 4 1,5 3 2 1 1 0,5 0 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec 16 0,9 Skewness Prop. dry days Probability Dry days [-] 14 0,8 12 Skewness [-] 10 0,7 8 0,6 6 4 0,5 2 0,4 0 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
Statistical downscaling (2070-2100) Mean St.dev. Skewness Prop. dry days HIRHAM- Mean correction ECHAM5 Mean and variance correction Markov chain WG LARS WG NSRP WG
Statistical downscaling - extreme events 110 SNSRP 90 > Precpitation [mm] Mean and Variance Correction 70 > Mean Correction 50 30 5 10 20 50 100 200 Return Period Obs HIRHAM- Mean correction ECHAM5 Mean and variance correction SNSRP CI-95%
Statistical downscaling - extreme events HIRHAM- Mean correction ECHAM5 Mean and variance correction Markov chain WG LARS WG NSRP WG
Methodology ENSEMBLES DATA OBSERVED DATA SET (1990-2008) A1B Scenario: • Precipitation • Temperature Driving GCM: ARPEGE ECHAM5 • Pot. evap. RCM: HIRHAM ALADIN HIRHAM REMO CHANGE FACTORS Mean Correction STATISTICAL Mean & Variance DOWNSCALING Correction Weather generator Future Time Series MIKE SHE Hydrological Model (Impact Assessment )
MIKE SHE model of NE Sealand Precipitation Temperature and Pot. Evap. [meter] 6220000 6210000 6200000 6190000 6180000 6170000 6160000 6150000 [meter] Above 100 90 - 100 6140000 80 - 90 70 - 80 60 - 70 6130000 50 - 60 40 - 50 30 - 40 6120000 20 - 30 10 - 20 0 - 10 6110000 -10 - 0 -20 - -10 -30 - -20 6100000 -40 - -30 Below -40 Undefined Value 620000 640000 660000 680000 700000 720000 [meter] Downscaling Precipitation: Mean correction • Mean and variance correction • SNSRP weather generator • Temperature and pot. evap. Mean correction •
MIKE SHE Results – water balance (2070-2100) 350 Baseflow & Recharge 300 250 CM Baseflow to river CMV Baseflow to river 200 SNSRP Baseflow to river CM GWRecharge 150 CMV GWRecharge SNSRP GWRecharge 100 50 0 Obs HIRHAM-ECHAM HIRHAM-ARPEGE REMO-ECHAM ALADIN-ARPEGE 1,2 Precipitation Prec 1 1 0,8 Obs HIRHAM-ECHAM HIRHAM-ARPEGE REMO-ECHAM ALADIN-ARPEGE
MIKE SHE Results – extremes (2070-2100) HIRHAM- Annual Maximum 40 ECHAM5 Discharge 35 30 Discharge [m3/s] 25 20 15 10 5 0 10 20 50 100 Return Period [years] Annual Minimum 0,43 Discharge 0,41 0,39 Discharge [m3/s] 0,37 0,35 0,33 0,31 0,29 0,27 0,25 10 20 50 100 Return Period [years]
Concluding remarks Statistical downscaling required for climate change impact • assessments Scaling of GCM/RCM to the appropriate spatial and temporal scales Statistical adjustments of GCM/RCM Choice of statistical downscaling procedure depends on • application Water balance studies -> Mean correction Extreme event analysis -> Stochastic weather generators Assessment of uncertainties important • Scenario uncertainty GCM/RCM uncertainty Statistical downscaling uncertainty Impact model uncertainty Probabilistic projections needed for climate change impact • assessments and decisions on adaptation.
Thank you for your attention Henrik Madsen hem@dhigroup.com HydroPredict 2010, 20-23 September 2010, Prague, Czech Republic
Recommend
More recommend
