6533b831fe1ef96bd12995aa

RESEARCH PRODUCT

Dynamique et assemblage des communautés adventices : Approche par modélisation statistique

subject

description

To develop solutions for a productive and sustainable agriculture, principles, theories, andmethods of ecology may contribute to understand the biological processes governing the agroecosystem.The present case study was based on data collected by a network of observatories of weeds covering the whole of France (‘Biovigilance Flore’) and aimed at establishing forrules governing the assemblage and dynamics of weed communities in fields grown with annual crops. We particularly studied the possible relationships between species within acommunity, as well as the relationships between communities and their environment. Analyses were based on species abundances to take account of their effect on community assemblages. The main results of this PhD thesis are: Abundances of weed species are not randomly distributed within a field, and this distribution as well as the community pattern result into a homogenization of observed weed abundances at the field scale. The community pattern depends on the total weed abundance in the field and on whether the species is dominant in the community. We concluded that weed abundances responded to weed-weed competition in certain conditions though the sampling methodology was not adapted to prove this assumption. The choice of the null model for comparisons in situations without a control greatlyinfluences the procedure for analysing assemblages. This problem arises particularly at the semi-quantitative scale since no null model has yet been developed for this particular case. Hence, a generic null model was developed for our particular situation. It considerably improved the study of co-occurrences and trait dispersal in case of semi-quantitative datasets and thus our ability to identify factors and processes determining flora composition.Modelling the weed dynamics of a field over the crop succession needs to take into account the dynamics of the weed seed bank in the soil. As it was unknown, the seedbank wasconsidered as a hidden variable and modelled using a Hidden Markov Model (HMM). This approach was well adapted for predicting the life cycle of annual species. We thus identified effects of cultural practices on the dynamics of the seedbank and of the emerged flora. Thisstudy allowed us to identify life-history traits and weed management strategies adapted to aspecific weed flora.We successfully applied various ecological theories and improved methods to take into account the specificities of weed species, the variations in their environment as well as the specificities of our data-set. The quality and reliability of the data collected in the observatory network must be increased if we want to correctly identify management levers likely to contribute to sustainable weed management