Probabilistic methods for aggregate and cumulative exposure to pesticides Marc Kennedy Risk and Numerical Sciences team

Slides:



Advertisements
Similar presentations
1Scientific Advisory Board meeting | 31 maart 2011 Milan WP2: Data organisation and electronic platform Polly Boon.
Advertisements

ACROPOLIS WP5 – Integrated Risk Model Hilko van der Voet Biometris, DLO, Wageningen University and Research Centre ACROPOLIS kick-off meeting 7-8 June.
Bayesian tools for analysing and reducing uncertainty Tony OHagan University of Sheffield.
Insert Date HereSlide 1 Using Derivative and Integral Information in the Statistical Analysis of Computer Models Gemma Stephenson March 2007.
Running a model's adjoint to obtain derivatives, while more efficient and accurate than other methods, such as the finite difference method, is a computationally.
SE503 Advanced Project Management Dr. Ahmed Sameh, Ph.D. Professor, CS & IS Project Uncertainty Management.
Materials for Lecture 11 Chapters 3 and 6 Chapter 16 Section 4.0 and 5.0 Lecture 11 Pseudo Random LHC.xls Lecture 11 Validation Tests.xls Next 4 slides.
Sensitivity Analysis In deterministic analysis, single fixed values (typically, mean values) of representative samples or strength parameters or slope.
1 SSS II Lecture 1: Correlation and Regression Graduate School 2008/2009 Social Science Statistics II Gwilym Pryce
Modelling cumulative risk Hilko van der Voet Biometris, DLO, Wageningen University and Research Centre Third ACROPOLIS consortium meeting 31 March 2011,
PROBABILISTIC DIETARY EXPOSURE ASSESSMENT TO PESTICIDE RESIDUES.
Training Manual Aug Probabilistic Design: Bringing FEA closer to REALITY! 2.5 Probabilistic Design Exploring randomness and scatter.
Sensitivity and Scenario Analysis
Approaches to Data Acquisition The LCA depends upon data acquisition Qualitative vs. Quantitative –While some quantitative analysis is appropriate, inappropriate.
©GoldSim Technology Group LLC., 2004 Probabilistic Simulation “Uncertainty is a sign of humility, and humility is just the ability or the willingness to.
W. McNair Bostick, Oumarou Badini, James W. Jones, Russell S. Yost, Claudio O. Stockle, and Amadou Kodio Ensemble Kalman Filter Estimation of Soil Carbon.
Michael H. Dong MPH, DrPA, PhD readings Human Exposure Assessment II (8th of 10 Lectures on Toxicologic Epidemiology)
Sensitivity Analysis for Complex Models Jeremy Oakley & Anthony O’Hagan University of Sheffield, UK.
A Two Level Monte Carlo Approach To Calculating
Lecture 10 Comparison and Evaluation of Alternative System Designs.
Using ranking and DCE data to value health states on the QALY scale using conventional and Bayesian methods Theresa Cain.
Value of Information for Complex Economic Models Jeremy Oakley Department of Probability and Statistics, University of Sheffield. Paper available from.
The Calibration Process
Lecture II-2: Probability Review
Lecture 11 Implementation Issues – Part 2. Monte Carlo Simulation An alternative approach to valuing embedded options is simulation Underlying model “simulates”
© Harry Campbell & Richard Brown School of Economics The University of Queensland BENEFIT-COST ANALYSIS Financial and Economic Appraisal using Spreadsheets.
Gaussian process modelling
Calibration of Computer Simulators using Emulators.
Applications of Bayesian sensitivity and uncertainty analysis to the statistical analysis of computer simulators for carbon dynamics Marc Kennedy Clive.
Monte Carlo Simulation CWR 6536 Stochastic Subsurface Hydrology.
Chapter 14 Monte Carlo Simulation Introduction Find several parameters Parameter follow the specific probability distribution Generate parameter.
Number of Blocks per Pole Diego Arbelaez. Option – Number of Blocks per Pole Required magnetic field tolerance of ~10 -4 For a single gap this can be.
Probabilistic Mechanism Analysis. Outline Uncertainty in mechanisms Why consider uncertainty Basics of uncertainty Probabilistic mechanism analysis Examples.
Lecture 10: Mean Field theory with fluctuations and correlations Reference: A Lerchner et al, Response Variability in Balanced Cortical Networks, q-bio.NC/ ,
12a - 1 © 2000 Prentice-Hall, Inc. Statistics Multiple Regression and Model Building Chapter 12 part I.
Two Approaches to Calculating Correlated Reserve Indications Across Multiple Lines of Business Gerald Kirschner Classic Solutions Casualty Loss Reserve.
Assimilation of HF Radar Data into Coastal Wave Models NERC-funded PhD work also supervised by Clive W Anderson (University of Sheffield) Judith Wolf (Proudman.
Geo597 Geostatistics Ch9 Random Function Models.
Monte Carlo analysis of the Copano Bay fecal coliform model Prepared by, Ernest To.
17 May 2007RSS Kent Local Group1 Quantifying uncertainty in the UK carbon flux Tony O’Hagan CTCD, Sheffield.
Center for Radiative Shock Hydrodynamics Fall 2011 Review Assessment of predictive capability Derek Bingham 1.
Dutch plan for finalising Hair software package Alterra – Wageningen University and Research Centre Roel Kruijne Working Group Meeting on Pesticide Statistics,
Limits to Statistical Theory Bootstrap analysis ESM April 2006.
Experiences in assessing deposition model uncertainty and the consequences for policy application Rognvald I Smith Centre for Ecology and Hydrology, Edinburgh.
Random processes. Matlab What is a random process?
Machine Design Under Uncertainty. Outline Uncertainty in mechanical components Why consider uncertainty Basics of uncertainty Uncertainty analysis for.
Additional Topics in Prediction Methodology. Introduction Predictive distribution for random variable Y 0 is meant to capture all the information about.
WP2: Cumulative dietary exposure and hazard assessment Bernadette Ossendorp en Polly Boon.
Reducing MCMC Computational Cost With a Two Layered Bayesian Approach
Probabilistic Design Systems (PDS) Chapter Seven.
Portfolio wide Catastrophe Modelling Practical Issues.
Tutorial I: Missing Value Analysis
Uncertainty and Reliability Analysis D Nagesh Kumar, IISc Water Resources Planning and Management: M6L2 Stochastic Optimization.
CORRELATION-REGULATION ANALYSIS Томский политехнический университет.
Probabilistic Slope Stability Analysis with the
Introduction to emulators Tony O’Hagan University of Sheffield.
Development of improved approaches for exposure estimations of operators, workers, bystanders and residents Rianda Gerritsen-Ebben, representing BROWSE.
Monte Carlo Methods CEE 6410 – Water Resources Systems Analysis Nov. 12, 2015.
8 Sept 2006, DEMA2006Slide 1 An Introduction to Computer Experiments and their Design Problems Tony O’Hagan University of Sheffield.
Dealing with Uncertainty: A Survey of Theories and Practice Yiping Li, Jianwen Chen and Ling Feng IEEE Transactions on Knowledge and Data Engineering,
Marc Kennedy, Tony O’Hagan, Clive Anderson,
Types of risk Market risk
M. Kuhn, P. Hopchev, M. Ferro-Luzzi
The Calibration Process
Date of download: 12/16/2017 Copyright © ASME. All rights reserved.
Materials for Lecture 18 Chapters 3 and 6
Monte Carlo Simulation Managing uncertainty in complex environments.
Types of risk Market risk
Statistical Thinking and Applications
Uncertainty Propagation
Presentation transcript:

Probabilistic methods for aggregate and cumulative exposure to pesticides Marc Kennedy Risk and Numerical Sciences team Willem Roelofs, Vicki Roelofs, Hilko van der Voet

Potential human exposure to pesticides

POPULATION Individual Food consumption Activities Operator activities Worker activities Resident activities Bystander activities Consumer Product use Operator exposure Worker exposure Resident exposure Bystander exposure Consumer Product exposure Non-dietary exposure contributions (internal and external) Pesticides: EUROPOEM database or BROWSE model Pesticides: BREAM or BROWSE model Pesticides: EUROPOEM database BREAM or BROWSE model Pesticides & biocides: CONSEXPO model BROWSE/EFSA surveys or user-specified Specified by user MCRA Consumption Database MCRA Recipes Database MCRA Pesticide Residues Database Residues in food MCRA model Absorption factors Dietary exposure contribution (external) TOTAL AGGREGATE EXPOSURE (INTERNAL or EXTERNAL) Biocides: BEAT/ART models Vet meds: case by case Pesticides: EUROPOEM or BROWSE model Biocides: BEAT/ART models Vet meds: case by case Repeat calculation for large sample of individuals Conceptual model Cumulative Aggregated

Cumulative dietary exposure: Data sources, uncertainties Food consumption survey data  Per country of consumption (target population) Pesticide monitoring data  Multiple residues per sample  Vast majority of measurements ND (<LOR) impossible to estimate correlations reliably Pesticide usage survey data  Use and co-use of pesticides on a single crop

Estimate correlations from monitoring data? 5355 possible pair-wise correlations to estimate, based on UK data (119 foods, 10 triazoles measured):  1.1% have at least 2 samples with positive residues in both  0.1% have at least 10 samples with positive residues in both Conclusion: very little information available on correlation between residues  Information will have to be obtained from elsewhere Our model combines monitoring and usage data

Illustration: Triazole residues in carrots DifenoconazoleTebuconazoleNumber of samples <LOR <LOR <LOR0.01 (10), 0.02 (9), 0.03 (2), 0.04 (4), 0.06 (1)26 Limited residue monitoring data (2009, Pesticide Residue Committee)

Pesticide Usage Survey data Field treatment type (triazoles) Percent of total GB Carrot, parsnips and celery crop (2007) none46.08 Azoxystrobin/difenoconazole0 Azoxystrobin/difenoconazole+8.63 Difenoconazole2.21 Difenoconazole+0.43 Tebuconazole3.62 Tebuconazole Azoxystrobin/difenoconazole, Tebuconazole+0.72 Field treatment type (triazoles) Percent of total GB Parsnip crop (2007) none26.40 Difenoconazole1.71 Tebuconazole47.49 Difenoconazole, Tebuconazole24.40 Field treatment type (triazoles) Percent of total GB Carrot crop (2007) none46.08 Difenoconazole5.27 Tebuconazole33.39 Difenoconazole, Tebuconazole15.26 Combined treatments over the year (field-year level) Individual combinations applied (field-treatment level) Relatively few tank mixes, so we’ll assume uncorrelated amounts

Uncertain field size distributions Model accounts for: Dependence between treatment and field size distribution Limited survey data – does it matter?

Independent LN distributions for amounts of D, T Using PUS increases precision Residue data alone (2 points) provide some information Tebuconazole Difenoconazole Residues only PUS + residues Point estimate for p0

Operators Many uncertain/ variable inputs, little data

Code Emulator: 2 code runs Gaussian process response surface (meta-model) Emulator estimate interpolates data Emulator uncertainty grows between data points

3 code runs Adding another point changes estimate and reduces uncertainty

5 code runs And so on…with enough runs, emulator becomes surrogate for original model, and we can derive uncertainty/sensitivity measures

Sensitivity analysis for one of the outputs (preliminary bystander model) Main & joint effects for adult spray output Uncertainty due to emulation is small, for estimating these ‘average effects’

Partitioning the output variation Main input contribution (or additional interaction pair)% output variance Boom height23.0 Crop height13.0 Wind angle9.2 Forward speed8.5 Wind speed8.2 Boom height, Crop height3.8 Boom height, Forward speed3.2 Boom height, Wind angle2.7 Wind speed, crop height2.1 Boom height, wind speed1.8 Wind angle, crop height1.7 Number of nozzles1.5 Main effect contributions Joint effect contributions

Simulated spray outputs distributed along the x-axis extra variability around the ‘mean bystander deposit’ for a given spray level Monte Carlo estimate, 10,000 runs of the emulator with independent distributions for Wind Speed, Wind angle, Boom height (variability in real conditions during a single spray event)

Integrating Acropolis and Browse using 2D Monte Carlo Variability CDF: exposure for all individuals in population Uncertainty Repeat many times with different model parameter values (e.g. sampled from posterior), or via bootstrap variability uncertainty Integration Acropolis & Browse exposures both coded as 2DMC simulation matrices Scale Browse matrix to correspond to exposure on same scale as dietary (internal dose) Add dietary and non-dietary matrices