Résumé
Agriculture since the 1950s has shown pronounced trends toward specialisation and intensification. Intensive measures have been taken for crop protection against pests through the widespread use of chemical pesticides in order to reduce the loss of agriculture yield. Although crop protection practices have reduced the overall potential losses of 50 % to actual losses of about 30 %, crop losses due to pests still vary from 14 to 35 % according to the country. Moreover, consequences of this intensive agriculture are now well known with an important increase of atmospheric CO
2
concentrations, water pollution and biodiversity loss. Current challenge is thus to design alternative sustainable cropping systems which maintain food production while reducing externalities. Application of ecological principles to agroecosystems has been proposed for that purpose. Nonetheless, it is difficult since crop systems are characterised by frequent and high disturbances, high nutrient input and high pressure of human activities. Here, we review the current knowledge in weed science and functional ecology and propose a conceptual framework to analyse weed community assembly in arable fields. Cropping systems are highly variable in their combination of agricultural techniques. We designed a trait-based approach of functional diversity (1) to establish a comparative description of the environmental gradients created by cropping systems and (2) to characterise the response of weeds to environmental gradients. We categorise the effects of cropping systems on the environment into disturbance and resource gradients. Disturbances induced by actual and previous agricultural practices are split into physical and chemical components, whose regime are defined by disturbance timing and frequency. Resource availability in arable fields is described by the value of effect traits of crops, such as plant height, that are related to their use of resources. Finally, we provide a list of relevant response traits of weeds to each component of the two gradients.