A report on the modelling of multiple pests

Agricultural commercial fields are submitted to a wide range of pests (plant pathogens, weeds, animal pests). In order to limit damages caused by injury profiles, farmers need tools to design IPM strategies that integrate various control methods (vertical integration) and to manage simultaneously various pests (horizontal integration).

This deliverable presents advances in terms of modelling of i) injury profiles; ii) damages caused by an injury profile on a crop in a given production situation.

The developed tools aims at helping farmers and advisors to design IPM-based cropping system to limit damages caused by biotic stress and to help researchers and engineers involved in crop protection to develop their own models for research purposes. An agro-ecosystem embeds many pests that interact with the physical, biological and chemical environment. The modelling of injury profiles is an ambitious goal since they are parts of complex and open systems. In order to cope with this high level of complexity, an innovative qualitative modelling approach was proposed (IPSIM4; Aubertot and Robin, 2013). It consisted in developing a platform that permits users to easily and quickly develop hierarchical deterministic bayesian networks using a user friendly software (Bohanec, 2003).

One key advantage of the approach is that it permits to combine various sources of knowledge: i) experimental data; ii) data from diagnoses of commercial fields; iii) expert knowledge; iv) existing quantitative models; and v) information available in the scientific and technical literature. A proof of concept was proposed for several pests on wheat and several research program are now using the platform to develop modelling research programs. Methodological developments were achieved in order to assess the quality of prediction of the IPSIM models. They proved to have fair predictive qualities as measured on independent datasets (e.g. for eayspot on wheat, efficiency = 0.51; Bias= 5.0%; Root Mean Squared Error of Prediction = 24%; Robin et al, 2013), especially when considering that no adjustment procedure was applied. Complementary to this modelling approach, another modelling platform, named X-PEST, was developed in order to help design models that simulate yield losses caused by an injury profile in a given production situation. The platform is composed of 4 sections: a home section where the user can get information on the purposes and functionalities of the platform, the forge section where the user can easily develop his own damage model, and a modelling section where multi-simulation can be performed and stored in a database. X-PEST was implemented under the Virtual Laboratory Environment.

However, the implementation of the internet interface is still under progress. As an alternative, the platform was implemented under the UNISIM platform (Holst, 2013), with wheat as an example. In addition, a specific work on vineyard was developed under R, in collaboration with WP6. In all, this activity in PURE permitted significant methodological breakthroughs that will be extended beyond the end of the project.

The research leading to these results has received funding from the European Union Seventh Framework Programme (FP7/ 2007-2013) under the grant agreement n°265865- PURE