Probabilistic Commodity -Flow-Based Focusing of Monitoring Activities Commodity-Flow-Based to Facilitate Early Detection of Phytophthora ramorum Outbreaks Steven McKelvey Steven McKelvey, McKelvey, Associate Associate Professor, Professor, Department Department of of Mathematics, Mathematics, Statistics Statistics and and Computer Computer Science, Science, Saint Saint Olaf Olaf College College Frank Koch Frank Koch, Koch, Research Research Assistant Assistant Professor, Professor, Department Department of of Forestry Forestry and and Environmental Environmental Resources, Resources, North North Carolina Carolina State State University University National National Forest Forest Health Health Monitoring Monitoring Research Research Team, Team, Southern Southern Research Research Station, Station, USDA USDA Forest Forest Service Service Bill Smith Bill Smith, Smith, Quantitative Quantitative Research Research Ecologist, Ecologist, National National Forest Forest Health Health Monitoring Monitoring Research Research Team, Team, Southern Southern Research Research Station, Station, USDA USDA Forest Forest Service Service The The work work upon upon which which this this poster poster is is based based was was funded funded through through an an Evaluation Evaluation Monitoring Monitoring grant grant (Project (Project SO-R-08-01) SO-R-08-01) awarded awarded by by the the Southern Southern Region, Region, State State and and Private Private Forestry, Forestry, U. U. S. S. Forest Forest Service Service Project Project Objectives Objectives Products Products Through Through the the use use of of trace-forward trace-forward information information regarding regarding the the shipment shipment of of P. P. ramorum ramorum infected infected nursery nursery stock stock provided provided by by the the USDA USDA Animal Animal and and Plant Plant Health Health Inspection Inspection Service Service (APHIS), (APHIS), supplemented supplemented by by commodity commodity flow flow data data from from the the US US Departments Departments of of Commerce Commerce and and Transportation, Transportation, the the analytical analytical techniques techniques and and software software developed developed by by this this project project will will allow allow forest forest health health managers managers to to focus focus their their limited limited resources resources on on areas areas with with the the greatest greatest likelihood likelihood of of new new P. P. ramorum ramorum infestation infestation and and thus thus more more quickly quickly identify identify newly newly infested infested areas, areas, increasing increasing the the likelihood likelihood of of successful successful intervention intervention before before the the pathogen pathogen crosses crosses the the urban-forest urban-forest interface. interface. This This project project produced produced two two deliverables, deliverables, aa technical technical paper paper describing describing the the probabilistic probabilistic techniques techniques used used in in the the analysis analysis along along with with aa comprehensive comprehensive list list of of data data sources. sources. Examples Examples of of the the application application of of the the model model were were also also documented. documented. The The second second deliverable deliverable was was an an open open source source software software package, package, written written in in the the Java Java programming programming language language for for portability portability across across hardware hardware and and software software platforms, platforms, that that implements implements the the model. model. Full Full documentation documentation of of the the software software was was provided. provided. Sample Sample (Test) (Test) Program Program Output Output Parameter Parameter File File Name: Name: Node Information Node Information File File Name: Name: Link Information File Name: Link Information File Name: Mathematical Mathematical Methodology Methodology july30.sdp july30.sdp july30.sni july30.sni july30.sfl july30.sfl ----------------------------------------------------------------------------------------------------------------Newly Newly discovered discovered infected infected site site node node IDs: IDs: The The goal goal of of the the probabilistic probabilistic model model and and implementing implementing software software is is to to give give the the USDA USDA Forest Forest Service Service an an analytical analytical tool tool to to help help focus focus scarce scarce inspection inspection resources resources on on the the early early detection detection of of P. P. ramorum ramorum outbreaks outbreaks in in those those parts parts of of North North America America where where the the pathogen, pathogen, which which causes causes Sudden Sudden Oak Oak Death Death (SOD), (SOD), isis not not yet yet endemic. endemic. This This isis accomplished accomplished by by using using partial partial survey survey results, results, along along with with commodity commodity flow flow information, information, to to create create an an ordered ordered list list of of those those sites sites presently presently not not known known to to be be infected. infected. The The list list isis ordered ordered by by likelihood likelihood of of each each site site having having recently recently become become infected infected through through the the importation importation of of infectious infectious nursery nursery stock. stock. The The process process of of creating creating this this list list consists consists of of several several stages. stages. In In the the first first stage stage some some subset subset of of vulnerable vulnerable sites, sites, typically typically areas areas east east of of the the Rocky Rocky Mountains, Mountains, are are surveyed. surveyed. The The surveyed surveyed sites sites are are categorized categorized as as being being recently recently infected, infected, very very likely likely to to be be uninfected, uninfected, or or being being aa site site for for which which infection infection status status is is uncertain. uncertain. Sites Sites with with an an uncertain uncertain infection infection status status will will be be treated treated as as though though they they were were not not surveyed. surveyed. The The combination combination of of newly newly infected infected sites sites and and recently recently certified certified clean clean sites sites is is called called an an infection infection pattern. pattern. Once Once newly newly infected infected and and known known clean clean sites sites are are identified, identified, known known potential potential sources sources of of infectious infectious nursery nursery stock stock are are assigned assigned probabilities probabilities of of being being active active sources sources of of infectious infectious nursery nursery stock. stock. In In the the terminology terminology of of probability probability theory theory this this isis aa Bayesian Bayesian process process in in which which the the probability probability of of infectious infectious exports exports assigned assigned to to each each potential potential source source is is updated updated from from some some previous previous value value based based on on the the newly newly observed observed infection infection pattern. pattern. For For example, example, those those sources sources which which happen happen to to send send aa large large amount amount of of nursery nursery stock stock to to newly newly infected infected destinations destinations will will be be assigned assigned aa high high probability probability of of exporting exporting infectious infectious materials materials because because the the new new infections infections must must have have come come from from somewhere somewhere and and the the sources sources sending sending materials materials to to these these destinations destinations are are good good suspects. suspects. Similarly, Similarly, sources sources that that send send large large amounts amounts of of nursery nursery stock stock to to sites sites classified classified as as known known clean clean sites sites will will be be given given aa low low probability probability of of sending sending infectious infectious exports exports because because receiving receiving these these exports exports has has not not resulted resulted in in infection. infection. SODBuster Opening Window 66 88 Known Known clean clean site site node node IDs: IDs: 55 99 Source Source Infection Infection Probabilities Probabilities Node Node ID ID Node Node Name Name 11 22 33 44 2007 West-to-East Nursery Stock Shipments CA CA Los Los AA CA CA San San DD CA CA Sacra Sacra CA CA San San JJ AA priori priori Posterior Posterior 0.1000 0.1000 0.2000 0.2000 0.3000 0.3000 0.4000 0.4000 0.0817 0.0817 0.9629 0.9629 0.4952 0.4952 0.6346 0.6346 ------------------------------------------------------------------------------------------------------------------Destination --Destination Nodes Nodes Sorted Sorted by by node node ID-ID-(* (* indicates indicates known known infected infected or or known known clean clean node.) node.) ID ID 5* 5* 6* 6* 77 8* 8* 9* 9* Node Node Name Name AL AL Birmi Birmi FL FL Jacks Jacks FL Miami FL Miami MN MN Minne Minne NY New NY New YY P(Infected) P(Infected) 0.0000 0.0000 1.0000 1.0000 0.3738 0.3738 1.0000 1.0000 0.0000 0.0000 ------------------------------------------------------------------------------------------------------------------Destination --Destination Nodes Nodes Sorted Sorted by by node node names-names-(* (* indicates indicates known known infected infected or or known known clean clean node.) node.) Sample of a Threat Map (Model Test) 5* 5* 6* 6* 77 8* 8* 9* 9* Node Node Name Name AL AL Birmi Birmi FL Jacks FL Jacks FL FL Miami Miami MN MN Minne Minne NY New NY New YY P(Infected) P(Infected) 0.0000 0.0000 1.0000 1.0000 0.3738 0.3738 1.0000 1.0000 0.0000 0.0000 ------------------------------------------------------------------------------------------------------------------Destination --Destination Nodes Nodes Sorted Sorted by by probability probability of of infection-infection-(* (* indicates indicates known known infected infected or or known known clean clean node.) node.) ID ID Once Once risks risks have have been been assigned assigned to to the the unsurveyed unsurveyed destinations, destinations, inspection inspection resources resources can can be be mobilized mobilized to to high high risk risk destinations destinations with with the the aim aim of of identifying identifying those those sites sites that that are, are, in in fact, fact, infected infected and and taking taking actions actions to to eliminate eliminate the the threat threat of of introducing introducing P. P. ramorum ramorum into into forests forests currently currently free free of of the the pathogen. pathogen. AL AL Birmi Birmi NY NY New New YY ----------------------------------------------------------------------------------------------------------------- ID ID After After the the probabilities probabilities of of exporting exporting infectious infectious materials materials have have been been updated updated attention attention moves moves to to the the unsurveyed unsurveyed recipients recipients of of nursery nursery stock. stock. For For each each unsurveyed unsurveyed recipient recipient of of nursery nursery stock, stock, these these are are called called destinations destinations in in what what follows, follows, aa probability probability is is computed computed that that this this site site has has become become recently recently infected. infected. This This probability probability is is based based on on two two characteristics characteristics of of the the destination, destination, from from which which sources sources the the destination's destination's nursery nursery stock stock isis sent sent and and how how much much nursery nursery stock stock comes comes from from each each source. source. IfIf aa given given destination destination receives receives aa significant significant amount amount of of its its stock stock from from high high risk risk sources, sources, that that destination destination will will be be assigned assigned aa relatively relatively high high probability probability of of infection. infection. Conversely, Conversely, ifif aa destination destination receives receives very very little little stock stock from from high high risk risk sources, sources, itit will will be be assigned assigned aa low low risk risk of of infection. infection. FL FL Jacks Jacks MN MN Minne Minne 8* 8* 6* 6* 77 9* 9* 5* 5* Node Node Name Name MN MN Minne Minne FL Jacks FL Jacks FL FL Miami Miami NY New NY New YY AL AL Birmi Birmi P(Infected) P(Infected) 1.0000 1.0000 1.0000 1.0000 0.3738 0.3738 0.0000 0.0000 0.0000 0.0000