early warning and mitigation planning: epidemiological models add value to surveillance
DESCRIPTION
Dave Hodson and Chris GilliganTRANSCRIPT
Early warning and mitigation planning: Epidemiological models add value to surveillance
D.P. Hodson1 & C.A. Gilligan2 1CIMMYT-Ethiopia
2Department of Plant Sciences, University of Cambridge, UK
Overview: Partnerships adding value
1. Surveillance Component � Where are we now? � Starting to add value to surveillance � Foundation for epidemiological models
2. Epidemiological Modelling Component � How can epidemiological models help? ® Predicting pathogen arrival and spread ® ‘What if’ scenarios for management ® Sampling strategies
� Data/information needs
Global Wheat “Footprint” Rust Survey “Footprint” 2006 Rust Survey “Footprint” 2012
• 13,000+ survey records • 30+ countries • large % of developing world wheat
Information from Surveys: Stem Rust Hotspots
Ug99 races, Hotspots & Wheat
• Ug99 races detected in many hotspots (but not all)
• Current stem rust hotspots occupy a tiny fraction of wheat area
• What is the risk or hazard in those other wheat areas???
Information from Surveys: Yellow Rust Hotspots
• Different distribution • More widespread than stem rust
2009
2010
2011
2012 • Ethiopia: Yellow rust hotspots very dynamic! • Why??
Ethiopia: Less food for rusts? 2010
Yellow rust severity - surveys Susceptible vs resistant cultivars - surveys
Ethiopia: Less food for rusts? 2012
Yellow rust severity - surveys Susceptible vs resistant cultivars - surveys
Ethiopia: Estimated Wheat Area Susceptibility to Ug99 races
2005/06 2013/14
BGRI Cornell Screening Dbase CIMMYT Wheat Atlas
S
MR/MS
?
MR/MS
MR MS
S
?
Early warning – Ethiopia 2013
Action Steps: • Informal rust planning meeting: 12th June 2013 (CIMMYT, EIAR, FAO) • Comprehensive Belg season surveys (EIAR/CIMMYT) • Formal rust planning meeting , 6th August 2013 (CIMMYT, EIAR, MoA Extension Directorate, ATA, FAO, Animal & Plant Health Directorate) • MoA, Extension Directorate + EIAR: Early, main season surveys
Global Rust Monitoring System Assessment
CWANA – Yellow Rust Outbreaks (surveys)
Climatic Conditions – favourable for yellow rust? Regional Winds
Rust Caution – May 17th
Moving Forward: Value Addition from Epidemiological Models ● Good inputs = Good outputs � Surveillance platform providing critical foundation
layers: Host distribution, pathogen sources + environments, susceptibility distribution
● Despite an extensive surveillance network, many gaps remain e.g., where are the risks and hazards? Models have a key role here.
● Early warning. Some progress (e.g., Ethiopia 2013), but with model inputs can make substantial gains
Epidemiological toolbox ● Landscape-scale models for disease spread ● Stochastic models: allow for uncertainty and
variability ● Coupling meteorological with epidemiological
models to predict: � Risk – where might the pathogen arrive? � Hazard – likely rates of spread if pathogen arrives? � Control – ‘what if’ scenarios
Landscape scale models ● Chalara fraxinea � Ash dieback
● Detected in UK in 2012
Landscape scale models ● Chalara fraxinea � Ash dieback
● Meteorological model � risk of spore arrival
Landscape scale models ● Chalara fraxinea � Ash dieback
● Meteorological model � risk of spore arrival
● Consider all potential sources 2008-2011
Landscape scale models ● Chalara fraxinea � Ash dieback
● Meteorological model � risk of spore arrival
● Consider all potential sources 2008-2011
● Data supplied by UK Met Office ● Computational analysis based on NAME: also tested HYSPLIT
Landscape scale models ● Chalara fraxinea � Ash dieback
● Meteorological model � risk of spore arrival
● Identify principal sources that pose risk
2008
Landscape scale models ● Annual
variation
2009
Landscape scale models ● Annual
variation
2010
Landscape scale models ● Annual
variation
2011
Landscape scale models ● Annual
variation
2008 - 2011
Landscape scale models ● Cumulative
risk
2008 - 2011
Landscape scale models ● Model
predictions independent of disease observations
● Very strong agreement
● Good predictor of arrival
UK Spread Model: Infected Area
28
2013
● Epidemiological model � Transmission � Spread
® Wind dispersal ® Trade dispersal
● Host distribution � Density, connectedness
● Environmental conditions � Infection and sporulation
S I D R Susceptible Infected Detected Removed
UK Spread Model: Infected Area
29
2014
UK Spread Model: Infected Area
30
2015
UK Spread Model: Infected Area
31
2016
UK Spread Model: Infected Area
32
2017
UK Spread Model: Infected Area
33
2018
UK Spread Model: Infected Area
34
2019
UK Spread Model: Infected Area
35
2020
UK Spread Model: Infected Area
36
2021
UK Spread Model: Infected Area
37
2022 ● Risk maps Where is invasion
most likely?
● Hazard maps Where is impact of
spread most severe? ● Inform control
and sampling
Wheat stem rust: 1) Long distance spore dispersal
Meteorological dispersal model
Integrate multiple
sources of inoculum
Very low probability of long distance
dispersal
Generating risk and hazard maps
Wheat stem rust: 2) Density and connectivity of host
Generating risk and hazard maps
Wheat stem rust: 3) Environmental suitability
Coincidence: Temp X Leaf wetness X Light Infection Sporulation
Generating risk and hazard maps
UK Met Office data @3-6h intervals
Wheat stem rust: Generating risk and hazard maps
● Hazard maps Where is impact of
spread most severe?
● Risk maps Where is invasion
most likely?
Wheat stem rust: Input from BGRI community
● Environmental suitability � Infection � Sporulation
● Host � where when and how much? � Alternative hosts
● Pathogen dispersal � Data on dispersal � Snapshots of disease maps
Generating risk and hazard maps
Acknowledgements Dr Matt Castle Rich Stutt James Cox
Dr Nik Cunniffe Dr Stephen
Parnell Dr Alex Archibald
43
● Sampling method varies depending on question � First detection in new area � How much disease is present at time of first detection � Optimizing new detections after pathogen is introduced
Optimising Sampling
● Use of epidemiological models for sampling � Citrus greening in Florida � Chalara fraxinea in UK � Phytophthora ramorum in UK
Optimising Sampling Chalara fraxinea again
disease hazard map (potential outbreak size)
x distance to known outbreaks (probability of an outbreak)
= risk weighting
locations to sample
BBSRC
UK Research Councils
UK Government
& Industry
International sponsors
Acknowledgements