Using sensitivity analysis to identify key factors for the propagation of a plant epidemic

Bayesian inference for spatio-temporal stochastic transmission of plant disease in the presence of roguing: A case study to characterise the dispersal of Flavescence dorée

Estimating the distance at which pathogens disperse from one season to the next is crucial for designing efficient control strategies for invasive plant pathogens and a major milestone in the reduction of pesticide use in agriculture. However, we still lack such estimates for many diseases, especially for insect-vectored pathogens, such as Flavescence dorée (FD). FD is a quarantine disease threatening European vineyards. Its management is based on mandatory insecticide treatments and the removal of infected plants identified during annual surveys. This paper introduces a general statistical framework to model the epidemiological dynamics of FD in a mechanistic manner that can take into account missing hosts in surveyed fields (resulting from infected plant removals). We parameterized the model using Markov chain Monte Carlo (MCMC) and data augmentation from surveillance data gathered in Bordeaux vineyards. The data mainly consist of two snapshot maps of the infectious status of all the plants in three adjacent fields during two consecutive years. We demonstrate that heavy-tailed dispersal kernels best fit the spread of FD and that on average, 50% (resp. 80%) of new infection occurs within 10.5 m (resp. 22.2 m) of the source plant. These values are in agreement with estimates of the flying capacity of Scaphoideus titanus, the leafhopper vector of FD, reported in the literature using mark-capture techniques. Simulations of simple removal scenarios using the fitted model suggest that cryptic infection hampered FD management. Future efforts should explore whether strategies relying on reactive host removal can improve FD management.

Coupling machine learning and epidemiological modelling to characterise optimal fungicide doses when fungicide resistance is partial or quantitative

Increasing fungicide dose tends to lead to better short-term control of plant diseases. However, high doses select more rapidly for fungicide resistant strains, reducing long-term disease control. When resistance is qualitative and complete-i.e. resistant strains are unaffected by the chemical and resistance requires only a single genetic change-using the lowest possible dose ensuring sufficient control is well known as the optimal resistance management strategy. However, partial resistance (where resistant strains are still partially suppressed by the fungicide) and quantitative resistance (where a range of resistant strains are present) remain ill-understood. Here, we use a model of quantitative fungicide resistance (parametrized for the economically important fungal pathogen Zymoseptoria tritici) which handles qualitative partial resistance as a special case. Although low doses are optimal for resistance management, we show that for some model parametrizations the resistance management benefit does not outweigh the improvement in control from increasing doses. This holds for both qualitative partial resistance and quantitative resistance. Via a machine learning approach (a gradient-boosted trees model combined with Shapley values to facilitate interpretability), we interpret the effect of parameters controlling pathogen mutation and characterising the fungicide, in addition to the time scale of interest.

Global sensitivity analysis in epidemiological modeling

Operations researchers worldwide rely extensively on quantitative simulations to model alternative aspects of the COVID-19 pandemic. Proper uncertainty quantification and sensitivity analysis are fundamental to enrich the modeling process and communicate correctly informed insights to decision-makers. We develop a methodology to obtain insights on key uncertainty drivers, trend analysis and interaction quantification through an innovative combination of probabilistic sensitivity techniques and machine learning tools. We illustrate the approach by applying it to a representative of the family of susceptible-infectious-recovered (SIR) models recently used in the context of the COVID-19 pandemic. We focus on data of the early pandemic progression in Italy and the United States (the U.S.). We perform the analysis for both cases of correlated and uncorrelated inputs. Results show that quarantine rate and intervention time are the key uncertainty drivers, have opposite effects on the number of total infected individuals and are involved in the most relevant interactions.

Field and Landscape Risk Factors Impacting Flavescence Dorée Infection: Insights from Spatial Bayesian Modeling in the Bordeaux Vineyards

Flavescence dorée (FD) is a quarantine disease threatening European vineyards. Its management is based on mandatory insecticide treatments and the uprooting of infected plants identified during annual surveys. Field surveys are currently not optimized because the drivers affecting FD spread in vineyard landscapes remain poorly understood. We collated a georeferenced dataset of FD detection, collected from 34,581 vineyard plots over 5 years in the South West France wine region. Spatial models fitted with integrated nested Laplace approximation were used to identify local and landscape factors affecting FD detection and infection. Our analysis highlights the importance of sampling period on FD detection and of local practices and landscape context on FD infection. At field scale, altitude and cultivar choice were the main factors affecting FD infection. In particular, the odds ratio of FD infection in fields planted with the susceptible Cabernet Sauvignon, Cabernet Franc, or Muscadelle varieties were approximately twice those in fields planted with the less susceptible Merlot. Field infection was also affected by the field's immediate surroundings (within a circle with a radius of 150 to 200 m), corresponding to landscapes of 7 to 12 ha. In particular, the probability of FD infection increased with the proportions of forest and urban land and with the proportion of susceptible cultivars, demonstrating that the cultivar composition impacts FD epidemiology at landscape scale. The satisfactory predictive performance of the model for identifying districts with a prevalence of FD detection >10% of the fields suggests that it could be used to target areas in which future surveys would be most valuable.

Emerging strains of watermelon mosaic virus in Southeastern France: model-based estimation of the dates and places of introduction

Where and when alien organisms are successfully introduced are central questions to elucidate biotic and abiotic conditions favorable to the introduction, establishment and spread of invasive species. We propose a modelling framework to analyze multiple introductions by several invasive genotypes or genetic variants, in competition with a resident population, when observations provide knowledge on the relative proportions of each variant at some dates and places. This framework is based on a mechanistic-statistical model coupling a reaction–diffusion model with a probabilistic observation model. We apply it to a spatio-temporal dataset reporting the relative proportions of five genetic variants of watermelon mosaic virus (WMV, genus Potyvirus, family Potyviridae) in infections of commercial cucurbit fields. Despite the parsimonious nature of the model, it succeeds in fitting the data well and provides an estimation of the dates and places of successful introduction of each emerging variant as well as a reconstruction of the dynamics of each variant since its introduction.

Composite Monte Carlo decision making under high uncertainty of novel coronavirus epidemic using hybridized deep learning and fuzzy rule induction

In the advent of the novel coronavirus epidemic since December 2019, governments and authorities have been struggling to make critical decisions under high uncertainty at their best efforts. In computer science, this represents a typical problem of machine learning over incomplete or limited data in early epidemic Composite Monte-Carlo (CMC) simulation is a forecasting method which extrapolates available data which are broken down from multiple correlated/casual micro-data sources into many possible future outcomes by drawing random samples from some probability distributions. For instance, the overall trend and propagation of the infested cases in China are influenced by the temporal–spatial data of the nearby cities around the Wuhan city (where the virus is originated from), in terms of the population density, travel mobility, medical resources such as hospital beds and the timeliness of quarantine control in each city etc. Hence a CMC is reliable only up to the closeness of the underlying statistical distribution of a CMC, that is supposed to represent the behaviour of the future events, and the correctness of the composite data relationships. In this paper, a case study of using CMC that is enhanced by deep learning network and fuzzy rule induction for gaining better stochastic insights about the epidemic development is experimented. Instead of applying simplistic and uniform assumptions for a MC which is a common practice, a deep learning-based CMC is used in conjunction of fuzzy rule induction techniques. As a result, decision makers are benefited from a better fitted MC outputs complemented by min–max rules that foretell about the extreme ranges of future possibilities with respect to the epidemic.

•

Composite Monte-Carlo (CMC) simulation is a forecasting method.

•

A case study of using CMC through deep learning network is developed.

•

Decision makers are benefited from a better fitted Monte Carlo outputs.

•

Novel Coronavirus Epidemic is studied.

Improving Management Strategies of Plant Diseases Using Sequential Sensitivity Analyses

Improvement of management strategies of epidemics is often hampered by constraints on experiments at large spatiotemporal scales. A promising approach consists of modeling the biological epidemic process and human interventions, which both impact disease spread. However, few methods enable the simultaneous optimization of the numerous parameters of sophisticated control strategies. To do so, we propose a heuristic approach (i.e., a practical improvement method approximating an optimal solution) based on sequential sensitivity analyses. In addition, we use an economic improvement criterion based on the net present value, accounting for both the cost of the different control measures and the benefit generated by disease suppression. This work is motivated by sharka (caused by Plum pox virus), a vector-borne disease of prunus trees (especially apricot, peach, and plum), the management of which in orchards is mainly based on surveillance and tree removal. We identified the key parameters of a spatiotemporal model simulating sharka spread and control and approximated optimal values for these parameters. The results indicate that the current French management of sharka efficiently controls the disease, but it can be economically improved using alternative strategies that are identified and discussed. The general approach should help policy makers to design sustainable and cost-effective strategies for disease management.

Analyzing the Influence of Landscape Aggregation on Disease Spread to Improve Management Strategies

Epidemiological models are increasingly used to predict epidemics and improve management strategies. However, they rarely consider landscape characteristics although such characteristics can influence the epidemic dynamics and, thus, the effectiveness of disease management strategies. Here, we present a generic in silico approach which assesses the influence of landscape aggregation on the costs associated with an epidemic and on improved management strategies. We apply this approach to sharka, one of the most damaging diseases of Prunus trees, for which a management strategy is already applied in France. Epidemic simulations were carried out with a spatiotemporal stochastic model under various management strategies in landscapes differing in patch aggregation. Using sensitivity analyses, we highlight the impact of management parameters on the economic output of the model. We also show that the sensitivity analysis can be exploited to identify several strategies that are, according to the model, more profitable than the current French strategy. Some of these strategies are specific to a given aggregation level, which shows that management strategies should generally be tailored to each specific landscape. However, we also identified a strategy that is efficient for all levels of landscape aggregation. This one-size-fits-all strategy has important practical implications because of its simple applicability at a large scale.