• Udig-‐JGrasstools installa0on
• JAMI, temperature interpola0on applica0on
• NewAge Rainfall-‐Runoff model applica0on
• NewAge Rainfall-‐Runoff model PSO calibra0on
• NewAge Rainfall-‐Runoff model LUCA calibra0on
PART 1
Objectives: 1. Run JAMI OMS3 NewAge model component for air temperature
interpolation;
2. Plot the interpolated variables and compare them with the measures (scatter)
How JAMI works…
JAMI (Just Another Meteo Interpolator)
• Split in altimetric band
How JAMI works…
JAMI (Just Another Meteo Interpolator)
How JAMI works…
• Split in altimetric band
• Look for the closer stations
JAMI (Just Another Meteo Interpolator)
How JAMI works…
• Split in altimetric band
• Look for the closer stations
• Interpolate according to the variable and number of stations
JAMI (Just Another Meteo Interpolator)
1
2
3
4
JAMI (Just Another Meteo Interpolator)
OMS3 Component
Areas Reader
Altmetry Reader
Stations Reader
Basin Reader
Data Reader
JAMI (Just Another Meteo Interpolator)
What energy and altimetry files are?
Area file Altimetry file
JAMI (Just Another Meteo Interpolator)
PART 2
Objectives: 1. Run NewAge rainfall runoff model
2. Plot the simulated variables (timeseries)
NewAge runoff prodiction and routing components
PART 3
Objectives: 1. NewAge rainfall runoff model calibration by using PSO
2. NewAge rainfall runoff model calibration by using LUCA
What is PSO?
Mono and Mul0 Objec0ve Calibra0on
Ispra - 24 June 2011
Example 2
Mono and Mul0 Objec0ve Calibra0on
2) Optimization Algorithm
1) Objective Functions to optimize
1) Objectives Functions to optimize:
• Nash-Sutcliffe
• Pbias
• RMSE
• KGE
• FHF
• FLF
Mono and Mul0 Objec0ve Calibra0on
Mono and Mul0 Objec0ve Calibra0on 2) Optimization Algorithms:
Par0cle Swarm Op0miza0on
Amalgam
SCE
Mono and Mul0 Objec0ve Calibra0on 2) Optimization Algorithms:
Par0cle Swarm Op0miza0on
Amalgam
SCE
2 1
Coopera0on example Adapted
from
Maurice.Clerc@
WriteM
e.com
2 1
Coopera0on example
Parameter space
Par0cles
Veloci0es Objec0ve func0on
Adapted
from
Maurice.Clerc@
WriteM
e.com
2 1
Coopera0on example
Parameter space
Par0cles
Veloci0es Objec0ve func0on
Adapted
from
Maurice.Clerc@
WriteM
e.com
We love animals, is just an example
PSO Algorithm
Personal influence
Social influence
Iner0a
Start
Initialize particles with random position and velocity vectors.
For each particle’s position (xik)
evaluate fitness
If fitness f(xik) is better than
fitness f(pik-1) then pi
k-1= xik
Set best of pik as pg
k
Loop
unt
il st
oppi
ng
crite
ria is
sat
isfie
d
Stop: giving pgk, optimal solution.
PSO Algorithm
Update particles velocity and position
PSO Algorithm
HOW?
Uniform distribu0on LHS
Start
Initialize particles with random position and velocity vectors.
For each particle’s position (xik)
evaluate fitness
If fitness f(xik) is better than
fitness f(pik-1) then pi
k-1= xik
Set best of pik as pg
k
Loop
unt
il st
oppi
ng
crite
ria is
sat
isfie
d
Stop: giving pgk, optimal solution.
Update particles velocity and position
PSO Algorithm
HOW?
TOPOLOGY?
Uniform distribu0on LHS
Start
Initialize particles with random position and velocity vectors.
For each particle’s position (xik)
evaluate fitness
If fitness f(xik) is better than
fitness f(pik-1) then pi
k-1= xik
Set best of pik as pg
k
Loop
unt
il st
oppi
ng
crite
ria is
sat
isfie
d
Stop: giving pgk, optimal solution.
Update particles velocity and position
The circular neighbourhood
1
5
7
6 4
3
8 2
Adapted
from
Maurice.Clerc@
WriteM
e.com
The circular neighbourhood
1
5
7
6 4
3
8 2
Adapted
from
Maurice.Clerc@
WriteM
e.com
The circular neighbourhood
1
5
7
6 4
3
8 2
Adapted
from
Maurice.Clerc@
WriteM
e.com
LUCA, Let Us CAlibrate
Hay, L.E., Umemoto, M., (2006) Mul$ple-‐objec$ve stepwise calibra$on using Luca: U.S. Geological Survey Open-‐File Report 2006-‐1323, 25p. Hay, L.E., Leavesley, G.H., Clark, M.P., Markstrom, S.L., Viger, R.J., and Umemoto, M. (2006). Step-‐wise, mul$ple-‐objec$ve calibra$on of a hydrological model for a snowmelt-‐dominated basin. Journal of the American Water Resources Associa0on.
one or more steps execu0on(s)
selec0on of parameters from a given distribu0on
shuffled complex evelu0on SCE
KEY-WORDS
STEP(S)
ROUND(S)
LUCA, Let Us CAlibrate
1) The calibration proceeds one step at a time.
LUCA, Let Us CAlibrate
LUCA, Let Us CAlibrate
1) The calibration proceeds one step at a time.
2) After completing a step, the calibrated values of the parameters passed into the next step.
1) The calibration proceeds one step at a time.
2) After completing a step, the calibrated values of the parameters passed into the next step. 3) This is repeated until all steps are executed
LUCA, Let Us CAlibrate
1) The calibration proceeds one step at a time.
2) After completing a step, the calibrated values of the parameters passed into the next step. 3) This is repeated until all steps are executed 4) All the n steps are repeated #R
LUCA, Let Us CAlibrate