The role of climate forecast in water resources management - Case - PowerPoint PPT Presentation

The role of climate forecast in water resources management - Case studies in Southeast Asia - Shinjiro KANAE Research Institute for Humanity and Nature, Kyoto IIS, The University of Tokyo

Acknowledgement to � Staff and students of IIS, University of Tokyo (= Taikan Oki and his group) � Staff and students of University of Yamanashi (= Yukiko Hirabayashi and her group) � Staff of JMA (Mr. Yamada, Mr. Maeda, ….) � Staff of RIHN � Dr. Jun Matsumoto, Dr. Akiyo Yatagai. � Many colleagues in agencies in Southeast Asia (e.g. TMD, RID, …)

Contents � 2006 Floods in the Chao Phraya River � One-month climate forecast for Mekong flood. � A new gridded precipitation dataset for validation of climate forecast.

Chao Phraya

Introduction: many floods in 2006 Aug. ～ the beginning of Oct. Nan river Ping river Yom river Floods 5 times � Chiang Mai The death 104 � Bhumipol Sirikit The completely destroyed � dam Dam 51 The partially destroyed � 8779 The amount of damage � Bung Boraphet 17 billion B Dam The damaged farmland � Nakhon Sawan 3856 km 2 Chao Phraya river Pa Sak Compensation to farmer � Ang Thong (Mr. Chatchai) 2000 B/rai Dam (Mr. Panya) 500 B/rai Pa Sak river Ayutthaya http://www.business-i.jp/news/world- page/news/200610250026a.nwc 1rai = 1600 m 2 Bangkok

18 th October 2006

Rule curve of a dam-reservoir Bhumipol Dam-reservoir Bhumipol ダム 15000 Maximum ( 総貯水量 1,346,200 万 m 3 ) 2006 Cross-section （計画貯水量－上限） middle of Sep. ダム貯水量 ( × 100 万 m 3 ) Plan 12000 (=1994) 2006 (=2003) 9000 計画貯水量－下限） (=2004) (=2005) 6000 ( 堆砂容量 380,000 万 m 3 ) 3000 2/20 3/1 3/11 3/21 3/31 4/20 4/30 5/30 6/9 7/9 8/8 9/7 9/17 9/27 10/7 10/17 10/27 11/6 11/16 11/26 12/6 12/16 12/26 1/1 1/11 1/21 1/31 2/10 4/10 5/10 5/20 6/19 6/29 7/19 7/29 8/18 8/28

Flood in Ayutthaya and Ang Thong Yom river Nan river Ping river Typhoon came! Several weeks Reservoir of dam before dry-season Bhumipol Sirikit dam Dam 99% 100% actual Plan Bung Boraphet Dam Nakhon Sawan Chao Phraya river The rainy The dry Pa Sak Ang Thong season season Dam Pa Sak river King’s land Ayutthaya Oct. Bangkok

Interview in Ang Thong -1 � Worse than 1995, 2001 floods � 20-30 cm/day increase � Information by TV (3 wks prior to flood) Sex distinction Female � Temporary walls made about 3 days before Estimated age 40-50s flood (concrete) Occupation Shop owner � Most trouble : toilet Lives in area along the main road of Ang Thong

Typhoon “ Xangsane ” arrived … 2 nd Oct., in early dawn, Xangsane landed Thailand and changed to tropical depression

Typhoon Track

Introduction Reservoir of dam Few weeks before Ping river Yom river Nan river 100% Chiang Mai Bhumipol dam The rainy The dry season season Nakhon Sawan Oct. Ang Thong One month forecast. Pa Sak river Few days forecast. Ayutthaya both are important ! Bangkok

Summary up to here � Numerical climate forecast data is probably not used for actual purpose in many countries. � Information is useful, both for “few weeks beforehand” ( � climate) and “few days beforehand” ( � weather). � We should study: � Current accuracy of few weeks climate forecast. � How to use “non-perfect” climate forecast for water management.

Current accuracy and uncertainty of one-month rainfall forecast � Target basin is the Mekong. � JMA one-month forecast (hindcast) for 1992- 2001. � Spatial resolution is 2.5 degree. � Evaluation is based on one-month rainfall. (basically 10day-forecast*3 =30day is used.) (sometimes 30day-forecast is used.) � Observed precipitation data is collected mostly as a part of GAME/MAHASRI.

Major Flood disasters in the Mekong 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 VNM 10/9 7/7 8/28 11/18 11/4 KHM 7/31 9/30 8/26 10/9 THA 9/8 8/15 9/18 5/4 8/11 8/31 LAO 10/5 8/15 10/18 10/20 MMR

○ 2.5-degree grid and observation stations 上流域・・・格子 A 、 B 、 C 、 D 中流域・・・格子 E 、 F 、 G 、 H 、 I 、 J 、 K 、 L 、 O 下流域・・・格子 M 、 N Observation stations Flooded points

Correlation between observation and forecast (Precipitation) August September

Flood of 2000 July 7th: Forecast error and Ensemble uncertainty ％ ％ Ensemble Uncertainty Forecast Error (anomaly is compared)

Flood of 2000 July 7th: Forecast error based on 10-day forecast and 30-day f ％ ％ ％ Error by 10-day*3 Error by 30-day

Flood of 2000 July 7th: Forecast error based on 10-day forecast and 30-day f ％ ％ Ensemble uncertainty Ensemble uncertainty by 10-day*3 by 30-day

For the validation of forecast, observed precipitation dataset is very important!!

( A sian P recipitation － H ighly- R esolved O bservational D ata I ntegration T owards E valuation of the Water Resources ) Global Environmental Research Fund by the Ministry of Environment, Japan (Project B062 Approved as a three year project; May 2006 – March 2009) Principal Investigator: Dr. Akiyo Yatagai (RIHN) Research Institute for Humanity and Nature (RIHN) Co-PI: Dr. Akio Kitoh, Meteorological Research Institute (MRI)/ Japan Meteorological Agency (JMA) Members: Akiyo YATAGAI, Shinjiro KANAE, Tsugihiro WATANABE, Jumpei KUBOTA, Itsuki HANDOH (RIHN) Akio KITOH, Kenji KAMIGUCHI, Osamu ARAKAWA (MRI/JMA) Project Home Page: http://www.chikyu.ac.jp/precip/aphrodite.htm

East Asia Gauge-Based Daily Precipitation Analysis Climatology (Xie et al. 2004) Yatagai et al. (2005) (mm/day)

Xie et al. (2004) In each 0.5 deg. box

Cross Validation Tests Cross validation tests are conducted for 365 days of 1997; � Each time, daily precipitation observations at 10% randomly � selected stations are withdrawn and data at the remaining 90% stations are used to define the daily analyses at 0.05 o lat/lon grid resolution; This is repeated for 10 time so that each station is dropped � once; Withdrawn station observations are compared to analyzed � values at the gauge location to examine the accuracy of the daily analyses;

Accuracy – Gauge Network Density Analyses were made � without 10% stations Observation at (10 times) ； each 100-km is preferable. Correlation between the � withdrawn station observations and the analysis is calculated for each station; Scatter plots between � Horizontal distribution of correlation and the CCs will be used for distance to the closest Developing observational station are examined; Network in future 0 100 200 300 400 500 Distance (km) Xie et al. (2004,2007)

Conclusion My analysis is preliminary. Your help is necessary. � Let’s share climate-forecast dataset. (database for everyone is preferable.) � Let’s make a good observation dataset of precipitation over Asia. (Gridded data is easy to use!) � Let’s investigate from the viewpoint of - current accuracy - required accuracy for application - remove bias, and consider ensemble-uncertainty - importance of time-scale for application (several weeks? several days? hours?)

Thank you very much

○アンサンブル予報のばらつき 2000 年 7 月 7 日の洪水 ・・・ （ 2000/6/8-2000/7/7 の 30 日積算降水量の比較） 観測 予測 約 10 日先の予測 予測降水量は観測降水量の半分程度と過小評価気味

2000 年 7 月 7 日の洪水 ・・・ （ 2000/6/8-2000/7/7 の 30 日積算降水量の比較） ％ 予測スコア アンサンブルのばらつき ％ 30 日先の予測 30 日先の予測 予測スコアは概ね－ 20.0 ～ 20.0 の範囲 予測スコアの増大に偏りはない 変動係数は一様にとても大きい ⇒ 降水量予測のしにくかった状態