Beruflich Dokumente
Kultur Dokumente
“Programme with a difference”
applied to
problems of aquatic environment
Modelling
Computer-based model is
a simplified description of reality
an encapsulation of knowledge about a particular physical
or social process in electronic form
Hydroinformatics integrates
data,
models,
people
1. Computers used as
calculation devices of analytical
expressions – 1950s
no friendly interfaces
User interface
Judgement engines
Decision support systems for management
Real world
Modelling
Instrumentation
Systems Environment
sciences,
Optimisation
Management
D.P. Solomatine. Introduction to Hydroinformatics 12
Encapsulation of knowledge
related to water
Tacit (implicit) knowledge embedded within a person
Words, texts, images
printed
stored in electronic media
Mathematical models
formulas
algorithms
algorithms encapsulated in computer programs
(software)
Integrated systems encapsulating all of above –
Hydroinformatics systems
Connector to
Hydrologic/ Decision
Trigger
Hydraulic Tree
Model Module
GIS DSS
Module MAIN Database
INTERFACE Module
Solar
WATER SURFACE Radiation
2 3 5
6 16
Sobek Rural Sobek Rural 1 Water Volume 15
1D2D DELWAQ 5 13
Norg Porg
7 9 10 Water
Velocity 14
Hydro Water 4 NH4 Hyacinth
dynamic Water Depth 11
Quality PO4 12
Flow 6 NO3
8 9
Ecosystem SEDIMENT
Organic Matter Settled
PROCESSES
Water Hyacinth 1. Input / Output 5. Input / Output 9. Resuspension 13. Photosynthesis
Model (coded 2. Rainfall 6. Input / Output 10. Hydrolysis 14. Respiration
using SOBEK 3. Evapotranspiration 7. Sedimentation 11. Oxidation 15. Mortality
RURAL Open 4. Advection/Dispersion 8. Resuspension 12. Uptake/Growth 16. Losses
D.P. Solomatine. Introduction to Hydroinformatics 29
Process Library)
Importance of modelling
reduces complexity
encapsulates knowledge
provides a laboratory experience
refines tacit knowledge
enables reasoned intervention by humans
facilitates communication
assists education and training
Modelling
Archiving Decision
and analysis making
“Side effect”:
Knowledge
Acquisition Feedback
discovery
of and control
information
Application
Objectives and
Evaluation
World of water
D.P. Solomatine. Introduction to Hydroinformatics 31
Beyond physically-based models:
Computational intelligence
Optimisation and integration
Internet-based computing
mountaneous
catchment in Southern
Europe
area of 822 sq. km
variables for building a decision tree model were selected on the basis of
cross-correlation analysis and average mutual information:
inputs: rainfalls REt, REt-1, REt-2, REt-3, flows Qt, Qt-1
outputs: flows Qt+1 or D.P.QSolomatine.
t+3 Introduction to Hydroinformatics 34
Process (physically-based) modelling of flow:
river modelling context
Qtup
Available data:
rainfalls Rt lateral inflows QL
catchment and river physical properties (soil,
geometry, roughnesses) Rt Qt
initial and boundary conditions for flows Q 0(x,t)
Inputs: QL(x,t), Qup(t), Q 0(x,t) , system
properties ∂Q ∂h
+b = 0
Output: flow Q (x, t) ∂x ∂t
∂Q ∂ ⎛ Q2 ⎞ ∂ QQ
Model: + ⎜⎜ β ⎟⎟ + gA (h + H ) +gA 2 =0
∂t ∂x ⎝ A ⎠ ∂x K
Q (x, t)=F (QL(x,t), Qup(t), Q 0(x,t) , system properties)
Questions:
are the physical properties of the catchment
known?
is F good enough ?
D.P. Solomatine. Introduction to Hydroinformatics 35
Using data-driven methods in
rainfall-runoff modelling
Qtup
Available data:
rainfalls Rt
runoffs (flows) Qt
Rt
Inputs: lagged rainfalls Rt Rt-1 Rt-2 … Qt
Output to predict: Qt+T
Questions:
how to find the appropriate lags? (lags embody the
physical properties of the catchment)
how to build non-linear regression function F ?
weights weights
x1 a ij b jk y1 ⎛ N hid ⎞
x2
u 1x
y2
yk = F ⎜⎜ bok +
⎝
∑ b jk u j ⎟⎟
⎠
i =1
x3 y3
k=1,..., N out
xn us ym
Inputs Hidden layer Outputs
f(x)
⎛ N inp ⎞ 1
uj = F ⎜⎜ aoj + ∑ aij xi ⎟⎟
⎝ i =1 ⎠
0 x
j=1,..., N hid
Binary Sigmoid : F(x) = 1/ (1 + e-x)
200
MT verification
RMSE=12.548 150
NRMSE=0.258 100
COE=0.9331
50
0
0 20 40 60 80 100 120 140 160 180
t [hrs]
D.P. Solomatine.
Hydroinformatics. 40
Urban drainage system rehabilitation:
use of multi-objective optimization
rehabilitation: changing pipes, creating additional storages
optimization by multi-objective genetic algorithm:
find a compromise btw. min. cost and min. damage due to flooding
Compromise
Flood Damage
optimal solutions
Wastewater System Pipe
Network Model (MOUSE)
Data Processor Data Processor
Optimization Procedure Costs
(GLOBE, NSGA-II)
D.P. Solomatine. Introduction to Hydroinformatics 41
Using Random Search Global Optimization methods
in water distribution network rehabilitation
D.P. Solomatine.
Hydroinformatics. 44
Role of uncertainty
in water management
80
Upper bound
Lower bound
60
Deterministic
50 forecast
Prediction interval
Di schar ge
Forecasted river
40
(uncertainty)
30
20
10
0
1 11 21 31 41 51
Ti me
D.P. Solomatine.
Hydroinformatics. 45
Monte Carlo simulation of parametric uncertainty
y = M(x, s, θ) + εs + εθ + εx + εy
D.P. Solomatine.
Hydroinformatics. 46
Estimated prediction bounds: verification
(Bagmati river basin, Nepal)
Rainfall-Discharge plot
6000 0
50
5000
100
Jan-88
M ay-88
Sep-88
Feb-89
Jun-89
Oct-89
M ar-90
Jul-90
Nov-90
Apr-91
Aug-91
Jan-92
M ay-92
Sep-92
Feb-93
Jun-93
Oct-93
M ar-94
Jul-94
Dec-94
Apr-95
Aug-95
Time [days]
4000
90% prediction limits
Observed flow (m /s)
Observed flow
3000
3
SF – Snow
RF – Rain
EA – Evapotranspiration
SP – Snow cover
SF
RF IN – Infiltration 2000
EA R – Recharge
SM – Soil moisture
CFLUX – Capillary transport
SP UZ – Storage in upper reservoir
IN
PERC – Percolation 1000
SM
LZ – Storage in lower reservoir
R CFLUX Qo – Fast runoff component
Q0 Q1 – Slow runoff component
UZ Q – Total runoff
0
PERC D.P. Solomatine.
Q1 Q=Q0+Q1 750 775 800 825 850
LZ Hydroinformatics. Transform
function 47 Time(day)
Internet-based computing
and knowledge management
Databases
Document
bases
Software
Database and (modelling
Web Application Server
transaction server systems)
Access to data-,
Forum Computer
Messaging knowledge- and Distance learning
(bulleting board) conferencing
document base
Internet
simulation modelling
• RIBASIM
• Delft 3D
• HEC-
HEC-RAS
• MIKE 21
with applications to:
and tools
• SWAT • MIKE SHE - River basin management
• EPANET • RIBASIM
• MOUSE • WEST++ - Flood management
Data-driven modelling • Aquarius • MODFLOW
- Urban systems
and computational • NeuroSolutions - Coastal systems
• NeuralMachine
intelligence tools • AFUZ - Groundwater and
• WEKA
catchment hydrology
Systems analysis, • LINGO
- Environmental systems
decision support, • GLOBE
• BSCW (options)
optimization • AquaVoice
Links:
http://www.ihe.nl/Education/Short-courses/Regular-short-courses/Urban-Flood-Modelling-and-Disaster-Risk-Management
http://www.ihe.nl/Education/Short-courses/Regular-short-courses/Urban-Water-Systems-Modelling
http://www.ihe.nl/Education/Short-courses/Online-courses/Flood-Modelling-for-Management
http://www.ihe.nl/Education/Short-courses/Regular-short-courses/Flood-Risk-Management
http://www.ihe.nl/Education/Short-courses/Regular-short-courses/Introduction-to-River-Flood-Modelling
http://www.ihe.nl/Education/Short-courses/Regular-short-courses/Environmental-Systems-Modelling
http://www.ihe.nl/Education/Short-courses/Regular-short-courses/River-Basin-Modelling
http://www.ihe.nl/Education/Short-courses/Online-courses/Decision-Support-Systems-in-River-Basin-Management
http://www.unesco-ihe.org/Education/Short-courses/Regular-short-courses/New-data-sources-to-support-flood-modelling
Conclusion