Regularised Inversion and Model Predictive Uncertainty Analysis.

Slides:



Advertisements
Similar presentations
Introduction Acoustic radiation forces on particles within standing waves are used in processes such as particle separation, fractionation and agglomeration.
Advertisements

Bridging the Gap Between Statistics and Engineering Statistical calibration of CFD simulations in Urban street canyons with Experimental data Liora Malki-Epshtein.
Introduction to Long-Throated Flumes and Broad-Crested Weirs
Getahun Wendmkun Adane March 13,2014 Groundwater Modeling and Optimization of Irrigation Water Use Efficiency to sustain Irrigation in Kobo Valley, Ethiopia.
Model calibration using. Pag. 5/3/20152 PEST program.
FIGURE 4-1 Some fate and transport processes in the subsurface and atmospheric environment.
Incorporation of Magnetic Resonance Sounding data into groundwater models through coupled and joint inversion 8th Annual Meeting of DWRIP 2014 JANUARY.
Landslide Susceptibility Mapping to Inform Land-use Management Decisions in an Altered Climate Muhammad Barik and Jennifer Adam Washington State University,
(Z&B) Steps in Transport Modeling Calibration step (calibrate flow model & transport model) Adjust parameter values.
The Calibration Process
BIOPLUME II Introduction to Solution Methods and Model Mechanics.
8. Permeability (Das, chapter 7)
Quantify prediction uncertainty (Book, p ) Prediction standard deviations (Book, p. 180): A measure of prediction uncertainty Calculated by translating.
Uses of Modeling A model is designed to represent reality in such a way that the modeler can do one of several things: –Quickly estimate certain aspects.
FSA Groundwater Flow and Solute Transport Model Presentation to Scottsdale Citizens Group March 19, 1999.
Hypothesis-Testing Model-Complexity. Hypothesis Testing …..
When Does Aquifer Heterogeneity Matter? Predicting the Influence of Alternative Conceptual Models on Contaminant Plume Migration A classroom exercise for.
Uncertainty Analysis and Model “Validation” or Confidence Building.
Feb 2003HEC-RAS Version 3.11 Slides adapted from HEC Unsteady Flow Course Unsteady Flow Course.
Well Tests to Characterize Idealized Lateral Heterogeneities by Vasi Passinos and Larry Murdoch Clemson University K 1,S 1 K 2,S 2.
III. Ground-Water Management Problem Used for the Exercises.
Normal Distributions Z Transformations Central Limit Theorem Standard Normal Distribution Z Distribution Table Confidence Intervals Levels of Significance.
(Zheng and Bennett) Steps in Transport Modeling Calibration step (calibrate flow model & transport model) Adjust parameter values Traditional approach.
Grid design/boundary conditions and parameter selection USGS publication (on course website): Guidelines for Evaluating Ground-Water Flow Models Scientific.
Modflow, GWVistas MODular three-dimensional finite- difference ground-water FLOW model
Scenarios 1.Tidal influence 2.Extreme storm surge (wave overtopping, max. limit 200 l/s/m, period 2 h) Outlook calibration and validation of 3D model transfer.
Aquifer Storage Properties CVEG 5243 Ground Water Hydrology T. Soerens.
VIII: Methods for Evaluating Model Predictions 1. Define predictive quantity and calculate sensitivities and standard deviations (Ex8.1a) 2. Assess data.
Sundermeyer MAR 550 Spring Laboratory in Oceanography: Data and Methods MAR550, Spring 2013 Miles A. Sundermeyer Observations vs. Models.
Center for Radiative Shock Hydrodynamics Fall 2011 Review Assessment of predictive capability Derek Bingham 1.
Final Project Summary of Results & Conclusions. Generally predicted ARM at targets > Calibrated ARM Generally, predicted ARM at pumping wells > Predicted.
Uniondale, NY At least two UST releases MTBE mass release estimated 7,000 lb Upper Glacial; Magothy Threatened PWS deeper in Magothy IRM design In-situ.
1 Overview of Unsteady Flow Modeling With HEC-RAS Gary W. Brunner, P.E.
DRAINMOD APPLICATION ABE 527 Computer Models in Environmental and Natural Resources.
Purpose Probe construction Field and laboratory measurements Data processing, applicational example Conclusions, outlook A NEW METHOD TO DETERMINE in.
Types of Models Marti Blad Northern Arizona University College of Engineering & Technology.
Top mass error predictions with variable JES for projected luminosities Joshua Qualls Centre College Mentor: Michael Wang.
CE 3354 Engineering Hydrology Lecture 21: Groundwater Hydrology Concepts – Part 1 1.
U.S. Department of the Interior U.S. Geological Survey Evaluating uncertainty in areas contributing recharge to wells for water-quality network design.
Model Predictive Uncertainty. Sensitivity analysis …..
Calibration & Sensitivity Analysis. Head measured in an observation well is known as a target. Baseflow measurements or other fluxes (e.g., ET) are also.
Metropolitan Groundwater Model for Water-Supply Availability Analysis Technical Advisory Group Meeting June 19, 2008.
The Islamic University of Gaza Faculty of Engineering Civil Engineering Department EENV 5326 Groundwater Modeling.
(Z&B) Steps in Transport Modeling Calibration step (calibrate flow & transport model) Adjust parameter values Design conceptual model Assess uncertainty.
Steady-State and Transient Models of Groundwater Flow and Advective Transport, Eastern Snake River Plain Aquifer, INL and Vicinity, Idaho Jason C. Fisher,
IX. Transient Forward Modeling. Ground-Water Management Issues Recall the ground-water management issues for the simple flow system considered in the.
Cheas 2006 Meeting Marek Uliasz: Estimation of regional fluxes of CO 2 … Cheas 2006 Meeting Marek Uliasz: Estimation of regional fluxes of CO 2 …
Hampton Bays At least one UST release MTBE, TAME TBA degradation product Upper Glacial; Magothy Sensitive shallow saline surface water discharge IRM rate.
Well Tests to Characterize Idealized Lateral Heterogeneities by Vasi Passinos K 1,S 1 K 2,S 2.
Building Transient MODFLOW Models
AUTOMATED PARAMETER ESTIMATION Model Parameterization, Inverse Modeling, PEST.
Soil wetting patterns under porous clay pipe subsurface irrigation A. A. Siyal 1 and T. H. Skaggs 2 1 Sindh Agriculture University, Tandojam, Sindh, Pakistan.
Simulation heat tracing experiment
Ground Water Modeling Concepts
Texas Groundwater 2004 Towards Sustainability
Looking for universality...
Modeling ASR Hydraulics and Plume Geometry
Statistical Methods for Model Evaluation – Moving Beyond the Comparison of Matched Observations and Output for Model Grid Cells Kristen M. Foley1, Jenise.
The Calibration Process
Uncertainty and Non-uniqueness
Chapter 6 Calibration and Application Process
Uses of Modeling A model is designed to represent reality in such a way that the modeler can do one of several things: Quickly estimate certain aspects.
Calibration.
Hydrologic River Routing
The application of an atmospheric boundary layer to evaluate truck aerodynamics in CFD “A solution for a real-world engineering problem” Ir. Niek van.
PARAMETERIZATION, UNCERTAINTY – AND “REAL WORLD” MODELING
Summary of Results & Conclusions
Laboratory in Oceanography: Data and Methods
Transient Models See Anderson and Woessner Chapter 7
Incorporating Initial Conditions Model Calibration Process
Presentation transcript:

Regularised Inversion and Model Predictive Uncertainty Analysis

PEST …

Model Input files Output files

Model Input files Output files PEST writes model input files reads model output files

Batch or Script File Input files Output files PEST writes model input files reads model output files

Model calibration conditions Input files PEST Input files Model predictive conditions Output files

Model calibration conditions Input files Model predictive conditions Output files Maximise or minimise key prediction while keeping model calibrated PEST

distance or time q1q1 q2q2 q3q3 etc value Model output Field or laboratory measurements and model output:- calibration datasetprediction

distance or time q1q1 q2q2 q3q3 etc value Model output Field or laboratory measurements and model output:- calibration dataset Lower predictive limit

distance or time q1q1 q2q2 q3q3 etc value Model output Field or laboratory measurements and model output:- calibration dataset Upper predictive limit

distance or time q1q1 q2q2 q3q3 etc value Model output Field or laboratory measurements and model output:- calibration dataset Confidence interval for prediction

distance or time q1q1 q2q2 q3q3 etc value Model output Field or laboratory measurements and model output:- calibration dataset Predictive uncertainty interval

Traditional Parameter Estimation Principal of parsimony Employ no more parameters than can be estimated Calibration complexity dictated by calibration dataset.

Regularised inversion…

Advantages of Regularised Inversion The inversion process is able to put the heterogeneity exactly where it is needed Maximum information content is extracted from the data Predictive error variance is thus minimised Parameterisation complexity determined by prediction Because complexity is retained in the system, we have the ability to realistic assess predictive uncertainty because we do not exclude the detail on which a prediction can depend.

Two Principal Types of Regularisatoin “Tikhonov” – constrained minimisation Subspace methods – principal component analysis

SVD-Assist

Advantages Highly stable numerically. Highly efficient in model run requirements. Can adapt to noise content of data.

Hydraulic conductivity

Specific Yield

Water levels

Snake River Inflow

Local Domain and Air Photo Recovery Well Source area

MTBE concentrations for an elevation of:- –35 ft-msl to –40 ft-msl

Pilot Points and Observations Pilot points – 58 per layer, L1-L7, for HHK, VHK, POR (crosses). Water level observations (circles); MTBE observations (stars) Calibrated ‘mean’ particle. Recovery Well Source area

Example Section Profile Profile across plume at IRM transect

Figure 4 Typical Concentration Profile

Observed MTBE Modelled MTBE -35 to -40 ft msl

Profile - data Source Area FLOW

Source Area FLOW Profile – data and modelled concentrations

Simulated and Observed MTBE at the Recovery Well

Calibrated Horizontal and Vertical Hydraulic Conductivities Ground Water Flow

The cost of uniqueness …..

Model grid Dimensions of model domain 500m by 800m

Boundary H = 0.0 Q = 50 m 3 /day

Particle release point

Reality

True time = days True exit point = easting of

12 head observations

Reality Exit time = 3256 Exit point = 206

Calibration to 12 observations (no noise) Exit time = 7122 [true=3256] Exit point = 241 [true=206]

This model (with its three parameters)…

Calibration to 12 observations Zone-based calibration Exit time = 6364 [true=3256] Exit point = 244 [true=206]

… does not even acknowledge the detail upon which a critical prediction will depend, whereas this model ….

Calibration to 12 observations (no noise) Exit time = 7122 [true=3256] Exit point = 241 [true=206]

Another important point… … does. The former model will grossly under-estimate predictive variance.

Calculation of Model Predictive Error Variance…..

Parameter space Increasing number of parameter combinations

Estimable parameter combinations Unestimable parameter combinations Increasing number of parameter combinations

Error variance calculable from measurement error C(h) Error variance supplied by hydrogeologists C(p) Increasing number of parameter combinations

Error variance calculable from measurement error C(h) Error variance supplied by hydrogeologists C(p) model prediction

σ 2 = y t (I-R) t C(p)(I-R)y + y t GC(h)Gy Therefore total “possible model error” depends on both C(h) and C(p)

Error variance calculable from measurement error C(h) Error variance supplied by hydrogeologists C(p) model prediction

Error variance calculable from measurement error C(h) Error variance supplied by hydrogeologists C(p) model prediction Where do we draw the line on what we try to estimate?

Number of singular values Predictive error variance “Null space” term “Measurement” term Total Predictive error variance vs dimensions of calibrated parameter space

Optimising Data Acquistion…..

Schematic block diagram illustrating model layers and boundary conditions

The prediction

Pumping from layer

Measurements

Observation wells Layer 1 Layer 2 Layer 3

Water levels

Parameters

Hydraulic conductivity – layer 1 Hydraulic conductivity – layer 2 Hydraulic conductivity – layer 3 VCONT – layer 2 VCONT – layer 3 Specific yield – layer 1 Specific yield – layer 2 Primary storage capacity – layer 2 Primary storage capacity – layer 3 Riverbed conductance Recharge Parameters included in analysis

Pre-calibration contribution to predictive error variance

Predictive error variance vs dimensions of calibrated parameter space Minimum = 418 ft 2 at 160 singular values

Contribution to pre- and post-calibration predictive variance by selected parameter types

Optimization of data acquisition:- How can I deepen the minimum in the predictive variance curve?

σ 2 = y t (I-R) t C(p)(I-R)y + y t GC(h)Gy

Reduction in predictive variance if VCONT 2 characterization at each point is reduced from 0.74 to 0.37 (maximum reduction = 112.7ft 2 )

Locations of proposed layer 2-3 differential head measurements (reduction in predictive error variance = 230 ft 2 )

Predictive error variance vs dimensions of calibrated parameter space Previous minimum = 418 ft 2 at 160 singular values New minimum = 188 ft 2 at 190 singular values

Error variance of an existing model…..

IBOUND array

Riverbed K parameters

Log of K (K ranges from 1e-4 to 500)

All lateral Inflow Zones (red cells are fixed head – except for zone 1)

Management zones

Head error variance Number of cells