Workpackage 5

This task will use simulations of pest population dynamics on virtual landscapes, coupled with optimization algorithms, to design spatio-temporal patterns for pest control.

First, we intend to adapt and develop population dynamics models capable of reproducing how the landscape composition and its structure influence the population dynamics of some specific pests and of their natural enemies, and their inter-trophic level interactions. The modeled ecological processes will depend on knowledge about ecological requirements and behaviors of the considered species and will benefit from results of Task 1, 2 and 4. The models will include, for example, niche saturation, dispersal barriers or spatial habitat dilution in order to represent the effects of the spatial configuration of crops and agronomical practices within the landscape. In order to understand these ecological processes, which are naturally hierarchical, we need a set of multi-scale analysis tools, i.e. a multi-scale toolkit of models. Already available models in this subtask concern (i) aphids (Sitobion avenae, Rhopalosiphum padi) and their natural enemies (Syrph, Hymenopetra parasitoids) and (ii) Meligethes (Meligethes aeneus) and its parasitoid (Tersilochus heterocerus). These models will be improved or adapted according to the results of Task 1, 2 and 4. We will complete this multi-scale toolkit by adding reaction-diffusion models for the population dynamics of weeds (Viola arvensis) associated to a generalist predator carabid (Poecilus cupreus) and the population dynamics of aphids and their natural enemies (Aphidius ervi, A. matricariae, Episyrphus balteatus).

In order to assess landscape properties that could mitigate the impacts of pests and weeds on crop production, we need to generate realistic landscapes to be used as input data in our models. This landscape generation step will be based on available geomatic and agronomical data collected from Task 1/ Task 2 and already available in historical datasets of the same experimental sites. As we want to investigate new integrated pest management (IPM) solutions derived from landscape features, these generated landscapes need to be realistic but also to explore spatial organizations that differ from real existing landscapes. We propose to adapt available landscape simulation frameworks to our needs to create landscape controlling for crop plots density and shapes with agronomical practices spatially distributed according to user-defined rules.

Finaly, we will explore the system formed by the population dynamics of pests and their natural enemies, on one hand, and the landscape dynamic, on the other hand, as a complex system. Indeed, this system is (i) nonlinear (both in time and space) (ii) susceptible to initial conditions and (iii) showing many feedback loops. While more complex to assess, the spatio-temporal patterns that could be extracted from such study would be invaluable in the context of IPM. These patterns could explore the importance of crop rotation; pesticides use reduction, and many other time-dependent, spatially heterogeneous agronomical practices. Both 'static' and 'dynamic' landscape suppressiveness will be explored using models from subtask 5.1 coupled with multiobjective optimization algorithms e.g. evolutionary algorithm and particle swarm. Such algorithms aim at minimizing or maximizing simultaneously several criteria (for example minimizing the amount of pest damage on crops) with respect to a set of constraints on the decision variables (for example by respecting realistic landscape with a given utilized agricultural area). Those algorithms are known to be suitable to a wide range of model types as long as they can provide numerical output, which is the case of all our models.