research inst. for humanity and nature chikyu.ac.jp (also at iis, university of tokyo)
DESCRIPTION
Water Resources Application Project (WRAP). Research Inst. for Humanity and Nature http://www.chikyu.ac.jp (also at IIS, University of Tokyo) Taikan Oki. Made by RID. (the most critical problem). Sirikit. Bhumipol. Almost dry up !!!. 8 Sub River Basin 6 =Ping 7 =Wang 8 =Yom 9 =Nan - PowerPoint PPT PresentationTRANSCRIPT
http://hydro.iis.u-tokyo.ac.jp/GAME-T/ http://game-t.nrct.go.th/GAME-T/
Research Inst. for Humanity and Naturehttp://www.chikyu.ac.jp
(also at IIS, University of Tokyo)
Taikan Oki
Water Resources Application Project
(WRAP)
http://hydro.iis.u-tokyo.ac.jp/GAME-T/ http://game-t.nrct.go.th/GAME-T/
Criteria: the most critical problemCriteria: the most critical problem
Recovery Cost
Duration of problem
Frequency of occurrence
Future perspective
(the most critical problem)
Drought
Water storage elevation in Bhumipol Dam in drought years (1992-1994)
210
220
230
240
250
260
270
1 2 3 4 5 6 7 8 9 10 11 12
Month
Wa
ter
Sto
rag
e e
lev
ati
on
(m,m
sl)
Dead Storage
Upper Rule Curve
1992
Lower Rule Curve
19941993
Data Source: RID, 2001; PAL and Panya ,1999.
Almost dry up !!!Almost dry up !!!
1. Water Resources and Water Problems in the Study Area
8 Sub River Basin
6 =Ping
7 =Wang
8 =Yom
9 =Nan
10 =Main Chao Phraya
11 =Sakakrang
12 =Pasak
13 =Thachin
Made by RID
Study Area Water-related Problems
Flooding
Drought
Water Pollution
Excessive Groundwater Extraction
Bhumipol
Sirikit
http://hydro.iis.u-tokyo.ac.jp/GAME-T/ http://game-t.nrct.go.th/GAME-T/
5,357
10,034
5,964
8,326
4,371
15,607
8,500
4,821
2,000
5,615
10,531
7,163
-
4,000
8,000
12,000
16,000
20,000
NormalYear
Drought(1992)
Drought(1993)
Drought(1994)
MC
M
Stored Water on 1st Jan.Released WaterInflow
Water Situation in Drought 1992-1994
Data Source: EGAT (for whole year)
Planning ProblemPlanning Problem
Possible Solution: Reliable Long-term Hydroclimatic PredictionPossible Solution: Reliable Long-term Hydroclimatic Prediction
http://hydro.iis.u-tokyo.ac.jp/GAME-T/ http://game-t.nrct.go.th/GAME-T/
• Rainfall (RF) and StreamflowRainfall (RF) and Streamflow
• Southern Oscillation Index (SOI)Southern Oscillation Index (SOI)
• Sea Surface Temperature (SST)Sea Surface Temperature (SST)
2.DATA AND METHODOLOGY USED IN HYDROCLIMATIC PREDICTION
Sea Surface Temperature (SST)
The British Atmospheric Data Centre (BADC), UK
-GISST_2.3B Dataset (1 degree x 1 degree)
SOI
-35
-25
-15
-5
5
15
25
35
1960 1965 1970 1975 1980 1985 1990 1995 2000
SOI
SOI = (Standardized Tahiti - Standardized Darwin) / MSD
Bureau of Meteorology (BOM) Australia, AU
99 100 101
14
15
16
17
18
19C hiangm ai
Lam pangPhrae
N an
U ttaradit
Tak
M ae Sot
Bhum ibol Dam
Phitsanulok
Phetchabun
Nakhon Saw an
Lopburi
D on M uang
Suphanburi
Kanchanaburi
Bangkok
P.1
C.2
R aingauge station (by TM D )
Stream flow station (by R ID )
Sub-river basin
R iver
100 105
10
15
20
Thai Meteorological Department (TMD), TH
Royal Irrigation Department (RID), TH
Global Energy and Water Cycle Experiment (GEWEX), Asian Monsoon Experiment - Tropics (GAME-T)
Data UsedData Used (Monthly 1960-2000)
aP.1
98°0'E
98°0'E
98°30'E
98°30'E
99°0'E
99°0'E
99°30'E
99°30'E
100°0'E
100°0'E
18°30'N 18°30'N
19°0'N 19°0'N
19°30'N 19°30'N
Mae NgatReservoir
:
2
5
9
aP.1
98°0'E
98°0'E
98°30'E
98°30'E
99°0'E
99°0'E
99°30'E
99°30'E
100°0'E
100°0'E
18°30'N 18°30'N
19°0'N 19°0'N
19°30'N 19°30'N
Mae NgatReservoir
:
2
5
9
Upper Ping River Basin
http://hydro.iis.u-tokyo.ac.jp/GAME-T/ http://game-t.nrct.go.th/GAME-T/
Model Used for Long-term Hydroclimatic PredictionModel Used for Long-term Hydroclimatic Prediction
PURPOSE TYPE MODEL
Physically Based Model
Understand the physical mechanisms
Climate Model (GCM, AGCM, etc.)
Statistically Based Model
More Accurate Predicted Result
(for real application)
Linear Regression ModelGeneralized Additive ModelArtificial Neural Networks,
etc.
Artificial Neural Network (ANN)
• Use limited data• Computational skill in complex problem• No assumption needed as other statistical models• Updating parameters process
By Manusthiparom (2003)Manusthiparom (2003)
http://hydro.iis.u-tokyo.ac.jp/GAME-T/ http://game-t.nrct.go.th/GAME-T/
Methodology Used for Hydroclimate Prediction
Influence of ENSOOn rainfall and streamflow
1.1 El Niño/La Niña composites1.2 Categorical contingency analysis 11
Prediction process 2.1 Prediction by ANN modeling22
Improvement and extension3.1 Prediction using additional predictors3.2 Input Sensitivity Analysis3.3 Spatial rainfall prediction
33
Potential use of prediction for improved WRM system
4.1 Irrigation Water Demand Forecasting44
By Manusthiparom (2003)Manusthiparom (2003)
http://hydro.iis.u-tokyo.ac.jp/GAME-T/ http://game-t.nrct.go.th/GAME-T/3. LONG-TERM RAINFALL PREDICTION
BY ANN MODELING APPROACH
Input layer Hidden layer Output layer
I1
Ini
I2
Bias
H1
Hnh
H2
Bias
O2
O1
Onmax
Feed-Forward Direct Multi-step Network
Difficulties in Using ANN ModelingDifficulties in Using ANN Modeling
Physical considerations,Correlation analysis
Determination of Input Nodes1 SST&RF, SOI& RF,
RF& RF
Trial and error processDetermination of
No.of Hidden Layers and Hidden Nodes
2 Hidden layer=1No. of hidden node=5-10
Adaptive process with changing initial
weighting parameters
Training Process(Weighting factors)3 Good pattern=97 %
(target error=15%)
http://hydro.iis.u-tokyo.ac.jp/GAME-T/ http://game-t.nrct.go.th/GAME-T/
Predicted Rainfall Anomaly (12 months-ahead)
-200
-100
0
100
200
300
400
500
396 408 420 432 444 456
month
rain
fall
an
om
aly
(m
m/m
on
th)
CASE 1, Bias=3.3 mm/ EI=55.3%CASE 2, Bias=1.8 mm/ EI=76.4 %CASE 3, Bias=1.7 mm/ EI=91.5 %Observed
1997 19981996
Rainfall Anomaly Prediction 12 months ahead12 months ahead
Observed and Predicted Rainfall (12 months ahead)
-
100
200
300
400
500
600
- 100 200 300 400 500 600
Predicted Rainfall (mm/month)
Ob
se
rve
d R
ain
fall
(mm
/mo
nth
)Testing
Predicted Rainfall (12 months ahead)
-
100
200
300
400
500
600
1 25 49 73 97 121 145 169 193 217 241 265 289 313 337 361 385
month
Ra
infa
ll (m
m/m
on
th)
Predicted
ObservedTraining
Case 1Case 1: Train (1962-1979, 18 yrs), Test (1980-1999, 20 yrs)
Case 2: Train (1962-1989, 28 yrs), Test (1990-1999, 10 yrs)
Case 3: Train (1962-1994, 33 yrs), Test (1995-1999, 5 yrs)
SSTs: 3 areasSSTs: 3 areas
http://hydro.iis.u-tokyo.ac.jp/GAME-T/ http://game-t.nrct.go.th/GAME-T/
-
100
200
300
400
500
600
- 100 200 300 400 500 600
Predicted Rainfall (mm/month)
Ob
se
rve
d R
ain
fall
(mm
/mo
nth
)
Predicted rainfall for 12 months ahead (Testing)
0
100
200
300
400
500
600
1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000
Year
Ra
infa
ll (m
m/m
on
th)
Predicted
Observed
Case 3:
Train (1962-1994, 33 yrs), Test (1995-1999, 5 yrs)
Target
Error=
15%
Predicted rainfall for 12 months ahead (Testing)
0
100
200
300
400
500
600
1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000
Year
Ra
infa
ll (m
m/m
on
th)
Predicted Rainfall
Observed Rainfall
-
100
200
300
400
500
600
- 100 200 300 400 500 600
Predicted Rainfall (mm/month)
Ob
se
rve
d R
ain
fall
(mm
/mo
nth
)Case 2:
Train (1962-1989, 28 yrs), Test (1990-1999, 10 yrs) Large Error
Drought
Target
Error=
15%
Smaller Error
Bad Pattern
Bad Pattern
Rainfall Prediction 12 months ahead12 months ahead SSTs: 3 areasSSTs: 3 areas
http://hydro.iis.u-tokyo.ac.jp/GAME-T/ http://game-t.nrct.go.th/GAME-T/
O bserved ra in fa ll anom aly (August 1997) (a)
O bserved rain fa ll anom a ly (August 1998) (b )
P red ic ted ra in fall anom a ly (A ugust 1997) (d)
P red ic ted ra in fa ll anom a ly (A ugus t 1998) (e )
O bserved rainfall anom a ly (August 1999) (c)
P red ic ted rain fall anom a ly (August 1999) (f)
9 8 9 9 1 0 0 1 0 1
1 4
1 6
1 8
2 0
9 8 9 9 1 0 0 1 0 1
1 4
1 6
1 8
2 0
9 8 9 9 1 0 0 1 0 1
1 4
1 6
1 8
2 0
9 8 9 9 1 0 0 1 0 1
1 4
1 6
1 8
2 0
9 8 9 9 1 0 0 1 0 1
1 4
1 6
1 8
2 0
9 8 9 9 1 0 0 1 0 1
1 4
1 6
1 8
2 0
Rainfall (m m )
-200
-150
-100
-50
0
50
100
150
200
250
300
Spatial Rainfall PredictionSpatial Rainfall Prediction
Rainfall Anomaly in August 1997-1999Rainfall Anomaly in August 1997-1999
ObservationObservation
• Software: Surfer 8
• Total: 16 stations
• Gridding method: Kriging
• Variogram model: Linear
• Slope =1.0, Aniso= 1,0
• Kriging type: point
• Drift type: None
• No search: Use all data (16)
PredictionPrediction
One MonthOne Month AheadAhead
1997 1998 1999
By Manusthiparom (2003)Manusthiparom (2003)
http://hydro.iis.u-tokyo.ac.jp/GAME-T/ http://game-t.nrct.go.th/GAME-T/
Date: 2 November 2002Project: Krasieo Operation and Maintenance Project, Royal Irrigation DepartmentLocation: Suphanburi, Thailand
Learning the existing system of WRM Period: 4-15 November 2002Tutor: Mr. Sombat Sontisri (Irrigation Eng.)Chief: Mr. Pongsak ArunwichitsakulWater Allocation GroupOffice of Hydrology & Water ManagementRoyal Irrigation Department (RID), Thailand
Water Manager
Irrigation Eng.
Learning Existing Learning Existing System of WRM fromSystem of WRM from Water ManagerWater Manager
Meeting and InterviewMeeting and Interview Water Water UsersUsers
Water Users (Agriculture)Water Users (Agriculture)
• DroughtDrought is the most serious problem is the most serious problem
• They want to know how much water They want to know how much water will be will be availableavailable for them in next growing season for them in next growing season
Water Manager (RID)Water Manager (RID)
• They want to know how much They want to know how much water will be water will be availableavailable in next season in next season
• They want to improve the existing system if it is They want to improve the existing system if it is easy to understandeasy to understand and and easy to doeasy to do in practice) in practice)
5. POTENTIAL USE OF PREDICTION IN IMPROVING WATER RESOURCES MANAGEMENT SYSTEM
http://hydro.iis.u-tokyo.ac.jp/GAME-T/ http://game-t.nrct.go.th/GAME-T/
Mean Prediction1991 Drought 84.27 79.79 -4.47 1992 Drought 73.06 94.81 21.751993 Drought 66.25 95.39 29.141994 Transit 75.92 96.43 20.501995 Flood 86.06 94.59 8.52
YearEI (%)
YearDifference
(%)
1993
0
50
100
150
200
250
300
350
400
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
Rai
nfa
ll (
mm
)
Long-term Mean Observation Prediction
1993
-200
-150
-100
-50
0
50
100
150
200
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
Rai
nfa
ll A
no
mal
y (m
m)
Long-term Mean Observation Prediction
1993
0
5
10
15
20
25
30
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
Irri
gat
ion
Wat
er D
eman
d (
mcm
)
Observation Long-term Mean Prediction
1993
-10
-5
0
5
10
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
Irri
gat
ion
Wat
er D
eman
d A
no
mal
y (m
cm)
Long-term Mean Observation Prediction
Rainfall
Irrigation Water Demand (IWD)
An
omal
y V
alu
e
Ab
solu
te V
alu
eDrought:1993
Forecasted Irrigation Water Demand
Using long-term mean: 120.61 mcm
Using observation: 150.25 mcm
Using prediction: 147.61 mcm
Mae Ngat Irrigation Project, Chiang Mai
Using predicted rainfall is better
Using predicted rainfall is worse
12-month ahead forecasted IWD12-month ahead forecasted IWD
By Manusthiparom (2003)Manusthiparom (2003)
http://hydro.iis.u-tokyo.ac.jp/GAME-T/ http://game-t.nrct.go.th/GAME-T/
5,357
10,034
5,964
8,326
4,371
15,607
8,500
4,821
2,000
5,615
10,531
7,163
-
4,000
8,000
12,000
16,000
20,000
NormalYear
Drought(1992)
Drought(1993)
Drought(1994)
MC
M
Stored Water on 1st Jan.Released WaterInflow
Water Situation in Drought 1992-1994
Forecasted Irrigation Water Demand
Using long-term mean: 120.61 mcm
Using observation: 150.25 mcm
Using prediction: 147.61 mcm
1993
0
5
10
15
20
25
30
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
Irri
gati
on
Wate
r D
em
an
d (
mcm
)
Observation Long-term Mean Prediction
IWD
Adjustment of Irrigation Area for 120.61 mcm
Based on long-term mean: 30,000 rai (120.61/120.61*30,000)
Based on observation: 24,081 rai (120.61/150.25*30,000)
Based on prediction: 24,502 rai (120.61/147.61*30,000)
1 rai =1,600 m2 1 km2 = 625 rai
Drought:1993Potential Use of Forecasted IWD to improve Planning System
Water scarcity situation in 1994
should have been improved.
5,953 6,2024,373
By Manusthiparom (2003)Manusthiparom (2003)
http://hydro.iis.u-tokyo.ac.jp/GAME-T/ http://game-t.nrct.go.th/GAME-T/
Summary ANN can predict monthly rainfall a year
ahead with fairly good accuracy based on SST, SOI, and preceding rainfall.
Seasonal prediction of rainfall will substantially contribute for better water resources/reservoir operations.
GAME-T Database is there:
http://game-t.nrct.go.th/GAME-T/ New research opportunities under GAME-
Tropics/Phase II and WRAP for everybody!
http://hydro.iis.u-tokyo.ac.jp/GAME-T/ http://game-t.nrct.go.th/GAME-T/
Thank you!