2011 Annual Report
1a.Objectives (from AD-416)
Develop stochastic models to estimate the temporal increase and spatial spread of Huanglongbing (HLB). The models and statistical methods will be used to test alternative disease control strategies and select those strategies which appear to have the best chance of success prior to deployment of control methods in the field. The ultimate goal is to improve disease control and thus improve citrus productivity with the minimal horticultural disruption. Such new methodologies are in direct support of future USDA, APHIS and Florida Department of Agriculture and Consumer Services regulatory actions as well as the citrus industries of Florida, California, Texas and Arizona.
1b.Approach (from AD-416)
Existing epidemiology data sets will be used from sites in Florida and other countries infected with HLB. Parameter estimation will focus on the use of Markov-Chain Monte-Carlo (MCMC) simulation methods of SIR Susceptibles/ Infectious/Recovered) models for spatio-temporal data, together with maximum likelihood estimation of mean-field temporal models in order to derive estimates for epidemiological parameters including rates of primary and secondary infection and, where appropriate, parameters for dispersal kernels. Simple methods will be used to quantify anisotropy to analyse and predict directionality of spread.
This project is related to inhouse objective 3: Develop or improve comprehensive integrated disease management strategies.
The objectives of this study are: .
1)To develop a series of flexible stochastic models to predict the temporal increase and spatial spread of citrus huanglongbing (HLB) and canker. They can be used in a number of ways: to predict spread and to analyze the effectiveness of control strategies both in plantations and State-wide. .
2)Test various control methods under field conditions to evaluate effects and collect data to parameterize models. Using Markov-chain Monte Carlo methods, and extensive data from infected areas of South Florida for successive snapshots of the occurrence of symptomatic detected trees in known populations of susceptible trees, we have been able to estimate the transmission rates and dispersal kernel for HLB. We now have a working model that focuses on the differential effects of host age on epidemiological parameters and variability across the plantation. This allows for uncertainty in the parameters as well as variability over time and through space. We used Baysian methods to infer posterior densities on the model parameters. The uncertainty is then incorporated in models to predict spread and to allow for uncertainty in the efficiency and comparison of control methods. A front-end (a web based version of the model) was developed for non-researcher users has been nearly finalized and is in validation testing. Both residential and commercial citrus scenarios are being tested and a wide variety of epidemiological and climate/weather variables have been included and are user selectable and changeable via sliding controls. Currently we are examining various disease control/mitigation parameters and are visualizing the effect of these various control strategies. This web tool runs simulations one at a time and is highly instructive to growers and regulators. It is based on a more formal analytical model that can run thousands of simulations based on the same parameters and provide more statistically valid predictions for regulatory intervention strategy building and regulatory/industry decision making. Via this model we have been able to examine the effects of various controls such as using insecticide applications or not, removing infected trees or not, and the effect of HLB infection in young versus older trees. Model output suggests that at least for older trees controlling secondary infections by diseased tree removal and insecticide applications plus controlling primary infection from new insect immigrations by areawide control strategies, can reduce disease increase to a manageable 2 to 5% increase per year, which appears to be economically sustainable. The model continues to be improved upon to capture more and more of the true features of the data and the disease. The model is being extended to estimate spread, should HLB be introduced into new areas such as TX, CA or AZ.
Progress was monitored via through direct involvement in lab and field activities, research meetings, telephone calls and email communications.