The prediction of seasonal and inter-annual climate - in the eastern - PowerPoint PPT Presentation

The prediction of seasonal and inter-annual climate - in the eastern seaboard of Thailand - Department of Geohydraulics and Engineering Hydrology, University of Kassel

Introduction : background / overview Short-prediction : seasonal prediction Long-prediction : inter-annual prediction Impacts : climate change impacts Summary : results summary 2 Introduction Short-prediction Long-prediction Impacts Summary

Sample impacts : Climate change / traditional management 1.Water crisis during drought (Rayong, East of TH, 2005) no supply pattern of climate fluctuation changes, season shift and extreme weather no more water supply Jan Dec 2.Water crisis during Flood (North and Central of TH, 2011) Spillway overflow flood Jan Dec 3 Introduction Short-prediction Long-prediction Impacts Summary

Objectives : prediction of climate Development • Climate prediction tool for local area by employing statistical/stochastical framework Prediction • Local weather toward the impact of climate change on water/climate short and long term Water resources planning Ultimate goal To investigate of the climate change impact toward water resources 4 Introduction Short-prediction Long-prediction Impacts Summary

Study area / Pilot area : economic-heart of TH Thailand N A.Phan Thong Khlong Yai basin \ & A.Phanat Ni Khom & \ Chonburi 1560 km 2 A.Ban Bung W ater Supply Area \ & \ & Bang Phra Res. A.Bo Thong A.Si Racha A.Nong Yai Nong Khoa Res. & \ & \ " ! \ & 7 ! " ! " 9 4 i Chang " ! 5 Khlong Yai Res. ! " 6 A.Pluak Daeng Nong Pla Lai Res. A.Bang Lamung \ & K.A.K & \ ! " 4 \ & Sea A.Wang Chan " ! Dog Krai Res. 3 (Thai Gulf) A.Ban Khai Rayong " ! A.K drought crisis 1 & \ Irrigation Area A.Ban Chang \ & in 2005 ! " 2 \ & Indus trial Area A.Sattahip Coastal basin with 3 reservoirs Thai Gulf 5 Introduction Short-prediction Long-prediction Impacts Summary

• seasonal prediction • up to 1 year • using Teleconnection • development of AR models 6 Introduction Short-prediction Long-prediction Impacts Summary

Teleconnection : connection between local & ocean climate Correlation coef. Thailand Nino 3.4 WTIO Nino 4 Nino 3 Equator -0.5 SETIO Nino1+2 SOI SWIO Correlation of +0.7 minimum temp Autocorrelation coeff (within -/+11 months) and Nino1+2 Mean min. temp. vs NINA 1.2 -1.0 +0.1 -0.9 +0.2 Ocean state index 7 ± -0.8 +0.3 -0.7 +0.4 ENSO Niño 1+2 -0.6 +0.5 0.0 +0.6 -0.5 -0.4 +0.7 with 3 month lag -0.3 +0.8 -0.2 +0.9 -0.1 +1.0 Thailand borderline 0 37.5 75 150 225 300 375 Kilometers 7 Introduction Short-prediction Long-prediction Impacts Summary

short-prediction enhancement : by teleconnection reduce prediction residuals 37 36 temperature (deg C) 35 34 33 32 31 obs Tmax 30 prediction : HiRes+SSTs (with teleconnection) prediction : HiRes (no teleconnection) 29 seasonal developing in reduce RMSE prd annual dry premonsoon monsoon1 monsoon2 Tmax 13% 3% 54% 18% 35% Tmin 3% 2% 4% 17% 0% PCP 5% 5% 5% 11% 6% 8 8 Introduction Short-prediction Long-prediction Impacts Summary

short-term climate prediction • p order of autoregressive terms • d order of integrated term (non-seasonal differences; linear, quardatic, etc.) • q order of moving average (forecast errors) 9 9 Introduction Short-prediction Long-prediction Impacts Summary

Model performance : short-term (Nash,1970) 1.0 Tmax Tmin PCP average Nash-Sutcliffe coefficience 0.8 0.6 0.4 0.2 0.0 HiRes GCMs GCMs+HiRes SSTs ECHO-G + SST GCMs+SSTs GCMs+HiRes+SSTs HiRes+SSTs ARIMAex-GCMs ARIMAex-HiRes ARIMAex-SST -0.2 -0.4 -0.6 -0.8 -1.0 GCMs GCMs+teleconnection ARIMA 1 10 Introduction Short-prediction Long-prediction Impacts Summary 0

• climate downscaling • GCMs • conventional tools • development of new models 11 Introduction Short-prediction Long-prediction Impacts Summary

Conventional downscaling tools SDSM LARS-WG statistical downscaling model stochastic downscaling model (Wilby, 1999) (Racsko, 1991; Semenov, 2002 ) 12 Introduction Short-prediction Long-prediction Impacts Summary

Developing of downscaling model y = x 1 𝛾 1 + ⋯ + x 𝑞 𝛾 𝑞 + 𝜁 •y dependent climate variable vector (local climate) •x independent GCM predictor vector •𝛾 regression coefficients of GCM predictor 13 Introduction Short-prediction Long-prediction Impacts Summary

multi-domain GCMs MLR experiments number experiments domain resolution model sub- MLR model predictor 1. Single-domain MLR 2.5°x2.5° ECHO-G 1 21 single domain 2. Hi-Res MLR 0.5°x0.5° High-Resolution grid 5 5x5 single domain 3. Multi-domain MLR 2.5°x2.5° 1 1340 multi ECHO-G,BCCR, domain ECHAM5,GISS,PCM 5 1340+5x5 4. Multi-domain multi 2.5°x2.5°+ ECHO-G,BCCR, +HiRes MLR domain 0.5°x0.5° ECHAM5,GISS,PCM,Hi-Res 24 stations : precipitation 4 stations : max & min temperature : humidity 2 stations : solar radiation 14 Introduction Short-prediction Long-prediction Impacts Summary

multi-season : optimal seasonal schemes seasonal schemes Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec no variation single period 2 season 1) Dry 2) Wet 1) 3 season 1) Winter 2) Summer 3) Rainy 1) 4 season 2) pre-monsoon 3) monsoon_1 4) monsoon_2 1) Dry optimal number of predictors max_temp[season 1] = 0.345*giss_model_e_r.r1.ta.20000 + 0.1549*giss_model_e_r.r1.ta.15000 max_temp[season 2] = 1.4e8*ncar_pcm1.r1.tro3.92500 + + 0.0959*mpi_echam5.r2.rsutcs + 0.254*ncar_pcm1.r1.ta.40000 - 149.23 0.609*mpi_echam5.r2.rsutcs - 12.46 AIC R 2 R 2 AIC season 1 season 2 max_temp[season 3] = 0.438*ncar_pcm1.r1.ta.30000 -0.109*ncar_pcm1.r1.ta.25000 max_temp = -1.2e8*ncar_pcm1.r1.tro3.92500 + 1.95e8*ncar_pcm1.r1.tro3.70000 -0.217*mpi_echam5.r1.ta.1000 +0.765 -0.012*giss_model_e_r.r1.rsus + ... - 18.98 AIC R 2 R 2 AIC season 3 no season number of predictors 15 Introduction Short-prediction Long-prediction Impacts Summary

Model performance : Long-term (Nash,1970) 1.0 0.88 0.87 Nash – Sutcliffe model efficiency coefficient Tmax 0.79 0.8 0.76 0.73 0.71 0.68 Tmin 0.62 0.59 0.59 0.57 0.6 0.55 0.53 PCP 0.44 0.41 0.38 0.4 0.37 0.37 0.30 0.27 0.22 0.2 0.13 0.00 0.0 SDSM LARS-WG MLR MLR MLR MLR MLR AR ARIMA ARIMAex-GCM -0.2 -0.13 -0.15 -0.27 -0.4 -0.31 -0.40 -0.6 -0.8 -0.69 -1.0 ECHO-G ECHO-GECHO-G HiRes CMIP3 CMIP3 - - CMIP3 +HiRes +HiRes daily ensemble daily monthly ensemble monthly ensemble monthly ensemble ensemble conventional single-domain MLR multi-domain MLR autoregressive 16 Introduction Short-prediction Long-prediction Impacts Summary

Stochastic generation of daily climate precipitation generation temperature generation time series of daily time series of daily time series of downscaled time series of downscaled precipitation max/min temperature monthly precipitation monthly precipitation prob. amount 2006-2096 2006-2096 1971-2006 1971-2006 monthly precipitation parameters temperature parameters precipitation estimation estimation amount monthly stochastic generation of stochastic generation of precipitation daily precipitation daily max/min probability occurrence temperature stochastic generation of daily wet/dry temperature on daily precipitation state wet and dry days amount precipitation amount on wet days daily maximum minimum precipitation temperature temperature 17 Introduction Short-prediction Long-prediction Impacts Summary

• impact studies • application of predicted results • hydrological impacts 18 Introduction Short-prediction Long-prediction Impacts Summary

climate change 18 obs PCP sim PCP 20c3m 16 sim PCP A1B monthly preicpitation (mm/day) sim PCP A2 14 sim PCP B1 12 per. Mov. Avg. (sim PCP A1B) 12 12 per. Mov. Avg. (sim PCP B1) 10 8 6 A2 4 sim PCP A1B B1 obs PCP 2 20c3m 0 Year 43 obs Temp sim A1B sim A2 sim B1 sim 20c3m 38 12-month moving avg. A1B linear trend A2 temperarure (deg C) Tmax B1 33 28 A1B Tmin A2 B1 23 20c3m 18 Year 19 Introduction Short-prediction Long-prediction Impacts Summary

Hydrological study (Arnold, 1998) runoff 1971 2096 24 30 rlz sim sb4 22 avg 30 rlz sim sb4 obs Z4 20 obs climate sim sb4 calibration verification 18 monthly runoff (cms) 16 14 12 10 8 9 8 6 2 4 1 2 6 0 5 7 4 3 1971 2096 hydrological components 10 11 2000 Soil+Surface ET 20c3m SRES 12 1800 PERC PCP.obs.sim amount of water (mm/year) 1600 ET.obs.sim PERC.obs.sim 1400 precipitation 1200 1000 800 evapotranspiration 600 400 200 Meteorological data percolation 0 HMD, SLR year 20 Introduction Short-prediction Long-prediction Impacts Summary

Impact on Hydrology 80% ET PERC Soil+Surface 72.0% 69.3% 69.2% 69.1% 68.8% 70% 67.7% ratio to amount of precipitation 61.8% 60% 57.3% 50% 40% 30% 24.5% 22.7% 20.2% 19.6% 19.3% 19.0% 18.7% 18.2% 20% 17.9% 15.5% 12.1% 12.1% 11.9% 11.6% 11.4% 10.1% 10% 0% obs.sim 20c3m A1B A2 B1 A1B A2 B1 1980-1999 2000-2049 2000-2096 average hydrological component along 21 st century +12 to 15 % ET (evaportranspiration) - 5 to 7 % PERC (groundwater recharge) - 6 to 8 % Soil+Surface (surface water) 21 Introduction Short-prediction Long-prediction Impacts Summary