2011/03/03 at the ICSS-Asia 2011 conference Multi-model approach for projecting future climate change conditions in Central Vietnam Thanh NGO-DUC, Van Tan PHAN, Trung NGUYEN QUANG Department of Meteorology Hanoi University of Science, Vietnam National University ngoducthanh@vnu.edu.vn
Outline Outline 1 Motivations of the study 2 Models and numerical experiments 3 Preliminary comparison with observations for the baseline period (1980-1999) 4 Future projections (until 2050) 2
I. Motivations 2007: IPCC Fourth Assessment Report (AR4) The National Target Program to response to climate change: Decision158/2008/QĐ-TTg 2009-2010 2011-2015 Phase I Kick-off post-2015 Phase III Phase II Development Implementation Objectives: to assess climate change s impacts & develop feasible action plan to effective respond to CC, take over opportunities to develop towards a low-carbon economy, and joint international community s effort to CC impacts and protect global climatic system 3
Climate Change, Sea level rise scenarios for Vietnam (MONRE, 2009) MAGICC/SCENGEN 5.3 software and Statistical Downscaling Method Question about the range of uncertainty? The Scenarios will be updated by using PRECIS & MRI 4
Statistical downscaling: MAGICC/SCENGEN Tools for dynamical downscaling: I. Motivations HMO faculty Hanoi University of Science RegCM REMO CCAM MM5CL Institute of Meteorology Hydrology and Environment (IMHEN) MRI-20km PRECIS 5
FC5 Q9450 192.168.1.0/24 I. Motivations Computing system (Faculty of Hydrology, Meteorology and Oceanography, Hanoi University of Science) Head Nodes Computing Network ~120 CPUs ~ 80TB storage Internet U P S Computing Nodes Data & Man. & Pub. Net 10.8.52.0/24 Webmeteo Login Node Sun HTTP FTP PCs NAS 6
Computing system - LINUX Cluster 7
I. Motivations Region of study: Central Vietnam Vulnerability to natural disaster & climate change Heavy rainfall and flood event occurred in 9 provinces in Central Vietnam in Nov 1999: 592 deaths, 421 injured, 30 people were missing. Damage ~220 million USD. 1841mm/2days in Nov 2 nd & 3 rd, 1999. 60% of the country population poor living-standard compared to other regions 8
II. Models and experiments II. Models and experiments Domain for experiments: Vietnam, Thailand, Laos, Cambodia, Bangladesh, Myanmar, Malaysia, Singapore, part of Indonesia Intercomparison? 14 observation stations: daily data Period: 1980-1999 (baseline), 2000-2050 (projection) This study can be expanded for the whole Vietnam 9
II. Models and experiments Scenario choices: A1B (average emission scenario) A2 (high emission scenario) (source IPCC, 2007) Currently, numerical experiments is set only to 2050 due to limited computational resources (computing speed and storage limitations). 10
Models and experiments II. Models and experiments CCAM CCSM3.0 ECHAM GCM boundary CCAM (26km) RegCM (36 km) MM5 (36 km) REMO (36 km) CCAM: Conformal Cubic Atmospheric Model, CSIRO, Australia CCSM: Community Climate System Model, US ECHAM: European Centre Hamburg Model, Germany PRECIS outputs (IMHEN)? MRI output? 11
III. Preliminaray comparison 2m-Temperature, 1980-1999 average RegCM MM5 REMO CCAM Similar spatial patterns among the models MM5 lowest temperature 12
Average 2m-temperature ( o C) of the 14 stations III. Preliminaray comparison RegCM & MM5: underestimation, similar behavior due to using the same CCSM3 boundary condition Overestimation of REMO CCAM is good in term of amplitude 13
Average 2m-temperature ( o C) of the 14 stations RegCM MM5 REMO CCAM MAM JJA SON DJF III. Preliminaray comparison Seasonal variations of Temperature can be well represented RegCM & MM5: behave similarly, cold bias larger in Winter, smaller in Summer REMO: over estimate, largest in Winter and Spring CCAM can well match the amplitude of observation 14
2m-temperature ( o C) of the 14 stations, 1980-1999 average for DJF, MAM, JJA, SON III. Preliminaray comparison Cold bias for most stations RegCM overestimates temperature for only 3 stations Largest cold bias in winter in the northern part. 15
2m-temperature ( o C) of the 14 stations, 1980-1999 average for DJF, MAM, JJA, SON MM5 Cold bias for most stations, except 2 stations Largest cold bias in winter in the northern part. 16
2m-temperature ( o C) of the 14 stations, 1980-1999 average for DJF, MAM, JJA, SON REMO Warm bias for most stations, except BaTo Systematic bias characteristics 17
2m-temperature ( o C) of the 14 stations, 1980-1999 average for DJF, MAM, JJA, SON CCAM well represents the obs. 18
III. Preliminaray comparison Precipitation Validation Annual Precipitation (mm/month) - 1980-1999 average RegCM3 REMO CCAM OBS Precipitation patterns are very different among models OBS: APHRODITE data (Yatagai et al., 2007) 19
14 station Average Precip: 1980-1999 RegCM REMO CCAM MAM JJA SON DJF III. Preliminaray comparison SON is the rainfall season in Central Vietnam. RegCM3 overestimates rainfall in winter. REMO largely underestimates rainfall CCAM underestimates rainfall during SON 20
RegCM: 1980-1999 Average Precip for MAM, JJA, SON, DJF III. Preliminaray comparison Average rainfall for the baseline period is well simulated at each stations, particularly in the rainy season. 21
REMO: 1980-1999 Average Precip for MAM, JJA, SON, DJF III. Preliminaray comparison REMO largely underestimates Precipitation 22
CCAM: 1980-1999 Average Precip for MAM, JJA, SON, DJF III. Preliminaray comparison CCAM underestimates Precip in the rainy season 23
Future Temperature -RegCM IV. Future Projections A1B JJA A1B DJF rising temperature for both A1B & A2 for 2041-2050, JJA increase > DJF increase A2 JJA A2 DJF 24
Future Precipitation - RegCM Difference (%) between 2041-2050 & baseline period IV. Future Projections Rainfall varies spatially & temporally A1B-MAM A1B-JJA A1B-SON A1B-DJF A1B DJF A2 DJF A2-MAM A2-JJA A2-SON A2-DJF 25
Temperature IV. Future Projections Increasing CCAM-A2 >> CCAM-A1B Linear trends seem to be similar 26
Increasing clear trend in JJA large variability in DJF
Increasing clear trend in JJA CCAM A2 increase remarkably in DJF
Precipitation IV. Future Projections non-clear trend for A2 increasing trend for A1B RegCM has more rainfall than the baseline time for the whole future 50-yr period CCAM-A2: less precipitation than the baseline period 29
Increasing trends (particularly CCAM) in SON: rainy season Large variability in DJF RegCM shows big increase of rainfall in MAM
no clear trend large difference among models
Summary 4 models were used: RegCM, MM5, REMO, CCAM Baseline period: 1980-1999 Temperature shows consistency among models and with obs. Large differences for simulated precipitation Future projection: 2000-2050 Increasing temperature No clear trend for precipitation Large variability among models 32
Future Challenges How to obtain the final projected scenarios (weighted average, arithmetic average, etc.?) Expand the simulations to 2100 Add more models, more scenarios, more GCM boundary inputs? (MRI-20km proposal accepted) Improve the computing system Possible Collaboration? Intercomparison? Thank you for your attention! 33