Plant disease symptom segmentation in chlorophyll fluorescence imaging with a synthetic dataset.

Despite the wide use of computer vision methods in plant health monitoring, little attention is paid to segmenting the diseased leaf area at its early stages. It can be explained by the lack of datasets of plant images with annotated disease lesions. We propose a novel methodology to generate fluorescent images of diseased plants with an automated lesion annotation. We demonstrate that a U-Net model aiming to segment disease lesions on fluorescent images of plant leaves can be efficiently trained purely by a synthetically generated dataset. The trained model showed 0.793% recall and 0.723% average precision against an empirical fluorescent test dataset. Creating and using such synthetic data can be a powerful technique to facilitate the application of deep learning methods in precision crop protection. Moreover, our method of generating synthetic fluorescent images is a way to improve the generalization ability of deep learning models.

Sustainable deployment of QTLs conferring quantitative resistance to crops: first lessons from a stochastic model.

Quantitative plant disease resistance is believed to be more durable than qualitative resistance, since it exerts less selective pressure on the pathogens. However, the process of progressive pathogen adaptation to quantitative resistance is poorly understood, which makes it difficult to predict its durability or to derive principles for its sustainable deployment. Here, we study the dynamics of pathogen adaptation in response to quantitative plant resistance affecting pathogen reproduction rate and its colonizing capacity. We developed a stochastic model for the continuous evolution of a pathogen population within a quantitatively resistant host. We assumed that pathogen can adapt to a host by the progressive restoration of reproduction rate or of colonizing capacity, or of both. Our model suggests that a combination of quantitative trait loci (QTLs) affecting distinct pathogen traits was more durable if the evolution of repressed traits was antagonistic. Otherwise, quantitative resistance that depressed only pathogen reproduction was more durable. In order to decelerate the progressive pathogen adaptation, QTLs that decrease the pathogen's maximum capacity to colonize must be combined with QTLs that decrease the spore production per lesion or the infection efficiency or that increase the latent period. Our theoretical framework can help breeders to develop principles for sustainable deployment of QTLs.

Timing of pathogen adaptation to a multicomponent treatment.

The sustainable use of multicomponent treatments such as combination therapies, combination vaccines/chemicals, and plants carrying multigenic resistance requires an understanding of how their population-wide deployment affects the speed of the pathogen adaptation. Here, we develop a stochastic model describing the emergence of a mutant pathogen and its dynamics in a heterogeneous host population split into various types by the management strategy. Based on a multi-type Markov birth and death process, the model can be used to provide a basic understanding of how the life-cycle parameters of the pathogen population, and the controllable parameters of a management strategy affect the speed at which a pathogen adapts to a multicomponent treatment. Our results reveal the importance of coupling stochastic mutation and migration processes, and illustrate how their stochasticity can alter our view of the principles of managing pathogen adaptive dynamics at the population level. In particular, we identify the growth and migration rates that allow pathogens to adapt to a multicomponent treatment even if it is deployed on only small proportions of the host. In contrast to the accepted view, our model suggests that treatment durability should not systematically be identified with mutation cost. We show also that associating a multicomponent treatment with defeated monocomponent treatments can be more durable than associating it with intermediate treatments including only some of the components. We conclude that the explicit modelling of stochastic processes underlying evolutionary dynamics could help to elucidate the principles of the sustainable use of multicomponent treatments in population-wide management strategies intended to impede the evolution of harmful populations.

Quantitative plant resistance in cultivar mixtures: wheat yellow rust as a modeling case study.

Unlike qualitative plant resistance, which confers immunity to disease, quantitative resistance confers only a reduction in disease severity and this can be nonspecific. Consequently, the outcome of its deployment in cultivar mixtures is not easy to predict, as on the one hand it may reduce the heterogeneity of the mixture, but on the other it may induce competition between nonspecialized strains of the pathogen. To clarify the principles for the successful use of quantitative plant resistance in disease management, we built a parsimonious model describing the dynamics of competing pathogen strains spreading through a mixture of cultivars carrying nonspecific quantitative resistance. Using the parameterized model for a wheat-yellow rust system, we demonstrate that a more effective use of quantitative resistance in mixtures involves reinforcing the effect of the highly resistant cultivars rather than replacing them. We highlight the fact that the judicious deployment of the quantitative resistance in two- or three-component mixtures makes it possible to reduce disease severity using only small proportions of the highly resistant cultivar. Our results provide insights into the effects on pathogen dynamics of deploying quantitative plant resistance, and can provide guidance for choosing appropriate associations of cultivars and optimizing diversification strategies.

The role of prey taxis in biological control: a spatial theoretical model.

We study a reaction-diffusion-advection model for the dynamics of populations under biological control. A control agent is assumed to be a predator species that has the ability to perceive the heterogeneity of pest distribution. The advection term represents the predator density movement according to a basic prey taxis assumption: acceleration of predators is proportional to the prey density gradient. The prey population reproduces logistically, and the local population interactions follow the Holling Type II trophic function. On the scale of the population, our spatially explicit approach subdivides the predation process into random movement represented by diffusion, directed movement described by prey taxis, local prey encounters, and consumption modeled by the trophic function. Thus, our model allows studying the effects of large-scale predator spatial activity on population dynamics. We show under which conditions spatial patterns are generated by prey taxis and how this affects the predator ability to maintain the pest population below some economic threshold. In particular, intermediate taxis activity can stabilize predator-pest populations at a very low level of pest density, ensuring successful biological control. However, very intensive prey taxis destroys the stability, leading to chaotic dynamics with pronounced outbreaks of pest density.