Assessing broomrape risk due to weeds in cropping systems with an indicator linked to a simulation model

EA SPE GESTAD INRA UB AGROSUP SUPDAT DOCTInternational audienceIntegrated crop protection tolerates residual weed floras if they are not harmful for crop production. These weeds can host harmful crop pests, among which parasitic plants such as branched broomrape (Phelipanche ramosa). This holoparasite is responsible for large yield losses in French crops such as oilseed rape. To date, there are no herbicides available to control it. To evaluate ex ante the impact of crop management practices on weed mediated parasite infection of crops, we developed an indicator calculated from outputs of the weed dynamics model FLORSYS. It consists of three components assessing weed impact on (1) stimulation of parasite germination during the whole cropping season, i.e. the potential risk reduction for future crops via a reduction of the parasite seed bank, (2) the stimulation of parasite germination in host crops, i.e. the potential risk increase for the current crop, (3) parasite reproduction on weed plants, i.e. the potential risk increase for future crops. This indicator was then used to predict weed-mediated broomrape risk in cropping systems from six regions from France and one from Spain. Antagonisms and synergies with other indicators of weed-harmfulness for crop production and weed contribution to plant and functional biodiversity were investigated with Pearson correlation analyses. For instance, cropping systems with a high parasite risk also had a high functional biodiversity (e.g. weed-based food offer for bees). Effects of crop management practices on the weed-mediated parasite risk indicator were identified with linear models; regression trees were used to identify the combinations of management practices that maximised or minimised weed-mediated broomrape risk. Parasite risk depended on crop rotation, sowing and harvest dates, tillage, herbicides and mechanical weeding. The lowest risk was observed in fields that were last tilled less than 21 days before sowing, with more than 0.6 herbicides per year (i.e. 3 applications in 5 years) with multiple entry modes into the weeds (e.g. leaves and roots) and the last herbicide sprayed no later than 127 days before harvest. RLQ analyses were used to identify correlations between weed species traits (Q matrix) and simulated parasite risk (R matrix), via simulated weed densities (L matrix). Early summer-emerging weed species increased parasite risk. No other notable correlations were found, indicating that parasite risk results from a weed community of interacting species, and not simply from individual weed species. An advice table was built to summarize and explain the effects of crop management practices on weed-mediated parasite risk