Title: P1252122133JHCMU
1MODELLING OVERVIEW CONTENT- V Modelling
Process VI Calibration and Assessing Model
Performance VII Data Collection Field
Measurement VIII Integrated Water Quality
Management IX Model Limitations
2V MODELLING PROCESS
3V MODELLING PROCESS Overview
- Problem specification
- Model Selection
- Preliminary Application
- Calibration Validation
- Management Applications
- Post-audit
4V MODELLING PROCESS 1) Problem Specification
- Customer ? objectives ? WQ engineer
- Management objectives
- Data relating to system description
- Physics, chemistry, biology
- Rough estimates of system behaviour
- Dilution, mass balance, salinity analysis
5V MODELLING PROCESS 2) Model Selection
- Processes to be modelled
- Temporal, spatial kinetic resolutions
- Choose existing software package
- OR
- May need to develop own model
6V MODELLING PROCESS 3) Preliminary Application
- Perform preliminary simulation
- Sensitivity analysis
- Enables
- Identification of data and/or theoretical gaps
- Identification of important parameters
- management of project resources
- Informing of field measurement campaign
7V MODELLING PROCESS 4) Calibration Validation
- Calibration
- tune the model to fit a data set
- Calibration parameters
- Calibration dataset should be similar to design
condition - Validation (Performance Assessment)
- Calibrated model run for a new set(s) of data
- Physical parameters forcing functions for new
conditions - Predictions match data gt accurate model
- No match gt analyse results to determine cause
recalibrate
8V MODELLING PROCESS 5) Management Applications
- Model used to ascertain effectiveness of
management decisions - Upgrading of waste treatment plants controls on
fertiliser use - Conservative modelling approach
- worst case scenario modelling
- 6) Post-Audit
- After implementation of actions ? check validity
of predictions
9VI MODEL CALIBRATION
10VI MODEL CALIBRATION Calibration
- tuning of model parameters to obtain best fit
11VI MODEL CALIBRATION Calibration
- tuning of model parameters to obtain best fit
- Calibrate modules in succession
- Hydrodynamics ? surface elevation, current
velocities - Solute transport ? dyes, salinity
- Water quality ? WQ constituents DO,
nutrients, Chlorophyll_a - Process
- Any measured parameters should be fixed
- Other parameters may be estimated from measured
data - Remaining parameters varied within recommended
ranges - Several approaches trial-by-error least squares
regression
12VI MODEL CALIBRATION Assessing Model Performance
- tuning of model parameters to obtain best fit
What is meant by best fit? Or How do we assess
the quality of a calibration?
- Two general approaches
- Subjective - based on visual comparison
- Objective - quantitative measure of quality of
fit ? error
13VI MODEL CALIBRATION Subjective Assessment
Model calibration output
14VI MODEL CALIBRATION Objective Assessment
Simulation error simulations Vs observations /
measurements NOTE always keep in mind inherent
uncertainty of measurements Errors - model
accuracy, bias, systematic error - model
precision, correlation
- Various Methods
- Average and relative error
- Least squares method
- Root-mean-square error
- Linear regression analysis
- Best assessment combines subjective and objective
15VI MODEL CALIBRATION Average Error, Ea
Oi , Si are observations and
matching simulations for i 1 ... N Ea gt 0 ?
model undersimulates Ea lt 0 ? model oversimulates
Relative Error, RE Oi , Si are
observations and matching simulations for i
1 ... N 0 RE 100 ? 100 poor
accuracy Use both with caution ? neither measure
imprecision
16VI MODEL CALIBRATION Least Squares Method
- residual difference between observed and
simulated data - The sum of the squared residuals, Sr
- Best fitis achieved when Sr is at a minimum
- Can be monitored as parameters are adjusted
during calibration
17VI MODEL CALIBRATION Root Mean Square, SE
- units are the same as those
- of the variable
- Best fitis achieved when SE is at a minimum
- Coefficient of variation, CV SE / Oa
dimensionless, expressed as - Oa is the average of observations Oi
- CV 0 ? high precision
- CV 100 ? low precision
18VI MODEL CALIBRATION Linear Regression Analysis
- Linear regression equation gives information on
the accuracy and precision of pairs of
observations, Oi, and simulations, Si - Regression equation takes the form
- Oi a ßSi e
- Where,
- a y-intercept and is a measure of accuracy
- ß slope and is a measure of accuracy
- e error in simulation
- Coefficient of correlation, r2, indicates degree
of correlation (precision) - Desired outcome is a 0, ß 1 and r2 ? 1
19VI MODEL CALIBRATION Linear Regression Analysis
EXCEL EXAMPLE
20VI MODEL CALIBRATION Sensitivity Analysis
- Allows greater understanding of the behaviour of
a WQ model - Usually carried out prior to or during
calibration - Number of different methods, two most common
- Parameter perturbation
- First-order sensitivity analysis
Example simple mass balance for well-mixed
waterbody gt c f ( Q, k, V, cm )
steady state
21VI MODEL CALIBRATION Sensitivity Analysis
a) Parameter Perturbation
b) First-Order Analysis
22VII DATA COLLECTION
23VII DATA COLLECTION Data Requirements
- Data collection and assimilation is crucial to
successful modelling - RUBBISH IN RUBBISH OUT
- Data is required for
- Boundary conditions
- Initial conditions
- Calibration and parameterisation
- Validation
24(No Transcript)
25VII DATA COLLECTION Measurement of Velocity and
Flow
- Floats / Drogues
- Visually / radar / satellite-tracked
- Current Meters
- mechanical
- acoustic
- electromagnetic
- Flow measurement at control structures
- Dams, gates, weirs, river gauges
- Remote Sensing
- Surface circulation patterns relative measures
but provide snapshots
5) Online data http//tidesandcurrents.noaa.gov/p
orts.html
26VII DATA COLLECTION Measurement of Surface
Elevation
- Tide pole / staff
- staff gauge must be tied to elevation datum
- Tide Gauges
- float guages
- pressure guages
- acoustic guages
- Online Data
- Admiralty Charts
27VII DATA COLLECTION Tracer Studies
- Excellent source of info for calibration/validatio
n of solute transport - Identification of dispersion and diffusion
coefficients - An effective tracer should be
- Water soluble
- Detectable at very low conc.
- Have minimal background interference
- Harmless in low concentrations
- Inexpensive
- Reasonably stable
- Common tracers
- commercial fluorescent dyes, e.g. rhodamine B /
WT - Natural tracers, e.g. salinity, temperature
28VII DATA COLLECTION Types of Dye Studies
- 1) Instantaneous / Slug Release
- Known quantity of dye released instantaneously
- Dye cloud movement monitored at sampling stations
- Useful for calibration and predicting spill
consequences - Data Analysis
- Concentration plots for stations
- Estimate travel time for
- leading edge, centroid, peak, trailing edge
- Time of travel Vs distance plot
29VII DATA COLLECTION Types of Dye Studies
- 1) Continuous Release
- Continuous dye release (constant flow conc)
over time interval - Monitoring stations upstream and downstream of
release - Data on flows, wind, tides also recorded
- Data Analysis
- Concentration plots for stations
- Contour maps of dye cloud
30VII DATA COLLECTION Planning Dyes Studies
- Acquisition of available data
- Reconnaissance study
- Estimation of mean velocities
- Estimate mixing and dye quantities
- Mixing Considerations
- Estimating quantity of dye release
- Determining locations of sampling stations
31VIII INTEGRATED WQ MANAGEMENT
32VIII INTEGRATED WQ MANAGEMENT
The integration of monitoring and modelling to
enable effective water quality management
Field Data
MONITORING
MODEL
Model Results
33(No Transcript)
34IX MODEL LIMITATIONS
35IX MODEL LIMITATIONS Model Limitations
36MODELLING OVERVIEW CONTENT- I Why
model? II Model Types III Model
Selection IV Industry Standard
Models V Modelling Process VI Model
Calibration VII Data Collection VIII Integrate
d WQ Management IX Model Limitations
37Estuarine Coastal Modelling Course
MODULE V MODELLING OVERVIEW
Stephen Nash, Environmental Change Institute, NUI
Galway