Response and effect traits of arable weeds in agro‐ecosystems: a review of current knowledge

Integrating principles of ecological intensification into weed management strategies requires an understanding of the many relationships among weeds, crops and other organisms of agro‐ecosystems in a changing context. Extensively used during the last two decades in weed science, trait‐based approaches have provided general insights into weed community response to agricultural practices, and recently to understanding the effect of weeds on agro‐ecosystem functioning. In this review, we provide a holistic synthesis of the current knowledge on weed response and effect functional traits. Based on the literature and recent advances in weed science, we review current knowledge on (i) weed functional groups and ecological strategies, (ii) weed functional response traits to cropping systems and (iii) weed functional effect traits affecting agro‐ecosystem functioning. For each functional trait, we explicitly present the assumptions and evidence on the linkage between trait values and ecological functions, in response to either management practices, for example tillage, sowing and herbicides, or biotic interactions, for example crop–weed competition and pollination. Finally, we address and discuss major research avenues that may significantly improve the use of traits and the knowledge of functional diversity in weed science for the future, especially to design and implement more environmentally sustainable weed management strategies.     

Evaluating field‐scale sampling methods for the estimation of mean plant densities of weeds

The weed flora (comprising seven species) of a field continuously grown with soyabean was simulated for 4 years, using semivariograms established from previous field observations. Various sampling methods were applied and compared for accurately estimating mean plant densities, for differing weed species and years. The tested methods were based on (a) random selection wherein samples were chosen either entirely randomly, randomly with at least 10 or 20 m between samples, or randomly after stratifying the field; (b) systematic selection where samples were placed along diagonals or along zig‐zagged lines across the field; (c) predicted Setaria viridis (L.) P. Beauv seedling maps which were used to divide the field into low‐ and high‐density areas and to choose the largest sample proportion in the high‐density area. For each method, sampling was performed with 5–40 samples. Systematic methods generally resulted in the lowest estimation error, followed by the random methods and finally by the predicted‐map methods. In case of species over‐ or under‐represented along the diagonals or the zig‐zag sampling line, the systematic methods performed badly, especially with low sample numbers. In those instances, random methods were best, especially those imposing a minimal distance between samples. Even for S. viridis, the methods based on predicted S. viridis maps were not satisfactory, except with low sample numbers. The relationships between sampling error and species characteristics (mean density, variability, spatial structures) were also studied.     

The role of models for multicriteria evaluation and multiobjective design of cropping systems for managing weeds

Weeds are both harmful for crop production and important for biodiversity, while herbicides can pollute the environment. We thus need new cropping systems optimising all cultural techniques, reconciling agricultural production, herbicide reduction and biodiversity conservation. Here, we show how to (i) develop models quantifying the effects of cropping systems on weed dynamics, (ii) integrate interactions between weeds and other organisms, (iii) predict the impact on production and biodiversity and (iv) use the model for multicriteria evaluation and multiobjective design of cropping systems. Among the existing weed dynamics models, we chose the one closest to our requirements to illustrate these different steps, that is, FlorSys which predicts multispecific weed dynamics as a function of cultural techniques and pedoclimate. We have illustrated the development of interaction submodels with the example of a crop pathogen whose propagation is increased when infecting grass weeds. To evaluate the weed flora impact, predicted weed densities were translated into indicators of harmfulness (crop yield loss, technical harvest problems, harvest pollution, field infestation, crop disease increase) and biodiversity (weed species richness and equitability, trophic resources for birds, insects and pollinators). Simulations were run over several years and with different weather scenarios (i) to optimise cultural techniques to control harmful weeds, (ii) to analyse the impact of changing agricultural practices (e.g. simplified tillage and rotations, no‐till, temporary crops) on weed density, species and trait composition and (iii) to evaluate cropping systems for their ability to reconcile agricultural production and biodiversity, thus identifying levers for designing sustainable cropping systems.     

Crop and density effects on weed beet growth and reproduction

Weed beet populations growing in each crop of the arable rotation could be a relay for the gene flow from adjacent transgenic herbicide‐resistant sugarbeet. In this study, weed beet growth and reproduction were assessed under several conditions which could be found in the rotation: various weed beet densities (ranging from 1 to 120 plants m^?2) and various crops (winter wheat, spring barley, spring pea, sugarbeet, maize, ryegrass). Measurements were carried out both on life‐cycle dynamics (bolting time, time to flowering onset, dynamics of flower opening) and on other quantitative data (survival rate, bolting rate and pollen, flower and seed production). Increasing weed beet density resulted in decreases in bolting rate and flower and seed production per plant. In cereals, weed beet establishment and reproduction were strongly reduced, compared with bare ground as a control situation. In pea, there was no effect on establishment, but the early harvest limited seed set. In the other crops, flower and seed production were reduced to a lesser extent. Parameters of the fitted equations on the bolting and flowering progress were modified by the weed beet density and by the crop. Our data may be used in a model predicting weed beet demographic evolution according to cropping system, and in assessing gene flow.     

Influence of crop management on eyespot development and infection cycles of winter wheat

Wheat was assessed at four crop growth stages for eyespot (anamorph Pseudocercosporella herpotrichoides, teleomorph Tapesia yallundae) in a series of field trials that studied the effects on disease frequency of five wheat management techniques (sowing date and density, nitrogen fertiliser dose and form, removal/burial of cereal straw). An equation expressing disease level as a function of degree-days was fitted to the observed disease levels. This equation was based on eyespot epidemiology and depended on two parameters illustrating the importance of the primary and the secondary infection cycles respectively. Cultural practices were classified according to the importance of their effects on disease, and these effects could be related to infection cycles and host plant architecture. Sowing date had the earliest and strongest effect; early sowing always increased disease frequency through the primary infection cycle, and its influence on the secondary cycle was variable. Disease frequency was increased by high plant density and/or a low shoot number per plant through primary infection; the secondary cycle was, however, decreased by a low shoot number per plant, which reduced late disease development at high plant density. High nitrogen doses increased disease levels and the severity of both infection cycles, but this effect was partly hidden by a simultaneous stimulation of tillering and thus an indirect decrease of disease incidence. When significant, ammonium (vs ammonium nitrate) fertiliser decreased eyespot levels and infection cycles whereas straw treatment (burial vs removal of straw from the previous cereal crop) had no effect.     

How much of seed dormancy in weeds can be related to seed traits?

Seed dormancy contributes to species persistence in unpredictable environments and is a key process to be taken into account in weed dynamics models. As the level of seed dormancy, photosensitivity and the dates of dormancy induction and release are difficult to measure, our objective was to relate weed seed dormancy with morphological, chemical or physiological seed traits and with expert knowledge. Dormancy of four species was studied experimentally during a 2‐year seed burial experiment. Experiments were supplemented with data from the literature to increase the number of species analysed, resulting in a data set of 29 species. Proportions of non–dormant seeds were higher for elongated than spherical seeds, even when accounting for phylogenetic relatedness between species. Elongated seeds, which tend to remain on the soil surface in undisturbed habitats, may have been selected for lack of dormancy and immediate germination to limit mortality due to predation. Dormancy increased with seed coat thickness, which can act as a chemical and physical barrier to germination, while no relation was found with seed lipid or protein content. No correlation was found between photosensitivity parameters and any of the species traits analysed. Variations in dormancy dates (induction and release) were highly correlated with average field emergence season estimated from expert knowledge. The observed correlations suggest that the level of dormancy results both from direct and from indirect effects of traits being involved in trade‐offs together with seed mortality.     

Combining a weed traits database with a population dynamics model predicts shifts in weed communities

A functional approach to predicting shifts in weed floras in response to management or environmental change requires the combination of data on weed traits with analytical frameworks that capture the filtering effect of selection pressures on traits. A weed traits database (WTDB) was designed, populated and analysed, initially using data for 19 common European weeds, to begin to consolidate trait data in a single repository. The initial choice of traits was driven by the requirements of empirical models of weed population dynamics to identify correlations between traits and model parameters. These relationships were used to build a generic model, operating at the level of functional traits, to simulate the impact of increasing herbicide and fertiliser use on virtual weeds along gradients of seed weight and maximum height. The model generated ‘fitness contours’ (defined as population growth rates) within this trait space in different scenarios, onto which two sets of weed species, defined as common or declining in the UK, were mapped. The effect of increasing inputs on the weed flora was successfully simulated; 77% of common species were predicted to have stable or increasing populations under high fertiliser and herbicide use, in contrast with only 29% of the species that have declined. Future development of the WTDB will aim to increase the number of species covered, incorporate a wider range of traits and analyse intraspecific variability under contrasting management and environments.     

Designing crop management systems by simulation

To help agricultural advisors to propose innovative crop management systems, simulation models can be a complementary tool to field experiments and prototyping. Crop management systems can be modelled either by using a vector representing dates and quantities used as input parameters in crop models or by developing specific decision models linked with biophysical models. The general design process of crop management systems by simulation follows a four-step loop (GSEC): (i) generation; (ii) simulation; (iii) evaluation; (iv) comparison and choice. The Generation step can follow different approaches: from blind generation before simulation to optimization procedures using artificial intelligence algorithms during the loop process. Simulation is mainly an engineering problem. Evaluation process means assigning a vector of indicators to the simulated crop management systems. A three-point evaluation can be carried out on the simulated crop management systems: global, agronomic and analytical. Comparison and choice of different simulated crop management systems raise the question of “monetary” versus “non-monetary” comparison and how to aggregate different quantities such as drainage, nitrogen fertilisers, labour, etc. Different examples are given to illustrate the GSEC loop on the basis of research programs conducted in France. Methodological advances and challenges are then discussed.     

Assessing non-chemical weeding strategies through mechanistic modelling of blackgrass (Alopecurus myosuroides Huds.) dynamics

Because of environmental and health safety issues, it is necessary to develop strategies that do not rely on herbicides to manage weeds. Introducing temporary grassland into annual crop rotations and mechanical weeding are the two main features that are frequently used in integrated and organic cropping systems for this purpose. To evaluate the contribution of these two factors in interaction with other cropping system components and environmental conditions, the present study updated an existing biophysical model (i.e. AlomySys) that quantifies the effects of cropping system on weed dynamics. Based on previous experiments, new sub-models were built to describe the effects on plant survival and growth reduction of mechanical weeding resulting from weed seedling uprooting and covering by soil, and those of grassland mowing resulting from tiller destruction. Additional modifications described the effect of the multi-year crop canopy of grassland on weed survival, growth, development and seed return to the soil. The improved model was used to evaluate the weed dynamics over 27 years in the conventional herbicide-based cropping system most frequently observed in farm surveys (i.e. oilseed rape/winter wheat/winter barley rotation with superficial tillage) and then to test prospective non-chemical scenarios. Preliminary simulations tested a large range of mechanical weeding and mowing strategies, varying operation frequencies, dates and, in the case of mechanical weeding, characteristics (i.e. tool, working depth, tractor speed). For mechanical weeding soon after sowing, harrowing was better than hoeing for controlling weed seed production. The later the operation, the more efficient the hoeing and the less efficient the harrowing. Tractor speed had little influence. Increasing tilling depth increased plant mortality but increased weed seed production because of additional seed germination triggering by the weeding tool. Decreasing the interrow width for hoeing was nefarious for weed control. The best combinations were triple hoeing in oilseed rape and sextuple harrowing in cereals. The best mowing strategy was mowing thrice, every 4–6 weeks, starting in mid-May. The best individual options were combined, simulated over 27 years and compared to the herbicide-based reference system. If herbicide applications were replaced solely by mechanical weeding, blackgrass infestation could not be satisfactorily controlled. If a three-year lucerne was introduced into the rotation, weed infestations were divided by ten. Replacing chisel by mouldboard ploughing before winter wheat reduced weed infestations at short, medium and long term to a level comparable to the herbicide-based reference system.     

GeneSys-Beet: A model of the effects of cropping systems on gene flow between sugar beet and weed beet

A weedy form of the genus Beta, i.e. Beta vulgaris ssp. vulgaris (hence “weed beet”) frequently found in sugar beet is impossible to eliminate with herbicides because of its genetic proximity to the crop. It is presumed to be the progeny of accidental hybrids between sugar beet (ssp. vulgaris) and wild beet (ssp. maritima), or of sugar beet varieties sensitive to vernalization and sown early in years with late cold spells. In this context, genetically modified (GM) sugar beet varieties tolerant to non-selective herbicides would be interesting to manage weed beet. However, because of the proximity of the weed to the crop, it is highly probable that the herbicide-tolerance transgene would be transmitted to the weed. To evaluate the likelihood of gene flow from GM varieties to weed beet and to propose cropping systems that reduce this likelihood, a model of the effects of cropping systems on population dynamics and gene flow in weed beet was developed, based on the existing spatio-temporal framework GENESYS and on field experiments for parametrising the life-cycle of weed beet. The resulting GENESYS-Beet model consists in simulating every year the life-cycle of weed and crop beet in each field of a given region. During flowering, the various life-cycles connect, leading to pollen exchanges which depend on field areas, shapes and distances. The life-cycle consists of a succession of life-stages for which both densities and genotype proportions are calculated. The relationships between the various stages depend on the crop grown in the field, the stage and genotype of the modelled crop relative, as well as the cultivation techniques (tillage tools and dates, sowing date and density, herbicides, mechanical and manual weeding, harvest date) used to manage the crop. Simulations of GM spread in different farms and regions and of the effects of weed management on the advent of GM beet were carried out to illustrate the possible uses of the model and the consequences of co-existing GM and non-GM crops.