Marie Gosme

Agroecosystem management and biotic interactions: a review

Increasing the use of synthetic fertilisers and pesticides in agroecosystems has led to higher crop yields, accompanied by a decline in biodiversity at the levels of field, cropping system and farm. Biodiversity decline has been favoured by changes at landscape level such as regional farm specialisation, increases in field size, and the removal of hedgerows and woodlots. The loss of biodiversity in agroecosystems has increased the need for external inputs because beneficial functions are no longer provided by beneficial species as natural enemies of crop pests and ecosystem engineers. This trend has led to a strong reliance on petrochemicals in agroecosystems. However, many scientists have been arguing for more than two decades that this reliance on petrochemicals could be considerably reduced by a better use of biotic interactions. This article reviews options to increase beneficial biotic interactions in agroecosystems and to improve pest management and crop nutrition whilst decreasing petrochemical use. Four agronomic options are presented. First, it has been shown that the choice of cultivar, the sowing date and nitrogen fertilisation practices can be manipulated to prevent interactions between pests and crop, in either time or space. Nevertheless, the efficacy of these manipulations may be limited by pest adaptation. Second, beneficial biotic interactions may result from appropriate changes to the habitats of natural enemies and ecosystem engineers, mediated by soil and weed management. Here, knowledge is scarce, and indirect and complex effects are poorly understood. Third, changes achieved by crop diversification and, fourth, by landscape adaptation are promising. However, these practices also present drawbacks that may not necessarily be outweighed by beneficial effects. Overall, these four management approaches provide a powerful framework to develop sustainable agronomic practices.

Explaining host–parasitoid interactions at the landscape scale: a new approach for calibration and sensitivity analysis of complex spatio-temporal models

Linking spatial pattern and process is a difficult task in landscape ecology because spatial patterns of populations result from complex factors such as individual traits, the spatio-temporal variation of the habitat, and the relationships between the target species and other species. Mechanistic models provide tools to bridge this gap but they are seldom used to study the influence of landscape patterns on biological processes. In this paper, we develop a methodological approach based on sensitivity and multivariate analyses to investigate the relationship between the biological parameters of species and landscape characteristics. As a case study, we used a tritrophic system that includes a host plant (oilseed rape, Brassica napus L.), a pest of the host plant (the pollen beetle, Meligethes aeneus F.), and the main parasitoid of the pest (Tersilochus heterocerus). This tritrophic system was recently represented by a model (Mosaic-Pest) that is spatially explicit at the landscape scale and that includes 32 biological parameters. In the current study, model simulations were compared with observed data from 35 landscapes differing in configuration. Sensitivity analysis using the Morris method identified those biological parameters that were highly sensitive to landscape configuration. Then, multivariate analyses revealed how a parameter’s influence on model output could be affected by landscape composition. Comparison of simulated and observed data helped us decrease the uncertainty surrounding the estimated values of the literature-derived parameters describing beetle dispersal and stage transition of the parasitoid at emergence. The advantages of using multivariate sensitivity analyses to disentangle the links between patterns and processes in landscape-scale spatially explicit models are discussed.

Disease Spread Across Multiple Scales in a Spatial Hierarchy: Effect of Host Spatial Structure and of Inoculum Quantity and Distribution

Spatial patterns of both the host and the disease influence disease spread and crop losses. Therefore, the manipulation of these patterns might help improve control strategies. Considering disease spread across multiple scales in a spatial hierarchy allows one to capture important features of epidemics developing in space without using explicitly spatialized variables. Thus, if the system under study is composed of roots, plants, and planting hills, the effect of host spatial pattern can be studied by varying the number of plants per planting hill. A simulation model based on hierarchy theory was used to simulate the effects of large versus small planting hills, low versus high level of initial infections, and aggregated versus uniform distribution of initial infections. The results showed that aggregating the initially infected plants always resulted in slower epidemics than spreading out the initial infections uniformly. Simulation results also showed that, in most cases, disease epidemics were slower in the case of large host aggregates (100 plants/hill) than with smaller aggregates (25 plants/hill), except when the initially infected plants were both numerous and spread out uniformly. The optimal strategy for disease control depends on several factors, including initial conditions. More importantly, the model offers a framework to account for the interplay between the spatial characteristics of the system, rates of infection, and aggregation of the disease.

Cascade: An Epidemiological Model to Simulate Disease Spread and Aggregation Across Multiple Scales in a Spatial Hierarchy

Disease spread occurs at several spatial scales, e.g., from field to field, plant to plant, and leaf to leaf. So far, epidemiological models have largely overlooked the multiscale nature of epidemics. Here, we propose a model that simulates disease spread across multiple scales in a nested hierarchy. The model is based on the central ideas of hierarchy theory, i.e., (i) the system is decomposed vertically into levels and horizontally into holons (elements at one level, which are complete systems when seen from the lower level), and (ii) higher levels are characterized by slower processes than lower levels. The model is individual-based, the individuals being the holons, which are either susceptible or infected. At each level, infections within one holon (i.e., infections between holons of the level below) occur independently from the other holons: infections between holons happen at the higher level. The self-similarity of the model structure and processes across all levels allows implementing the model with a simple recursive algorithm. The behavior of the model was studied using methods commonly applied to field data. Aggregation of the disease was characterized through the incidence-incidence relationship and the binomial power law, in order to study the effect of infectiousness at each level on disease aggregation. Sensitivity analyses showed that disease incidences at all levels were influenced by the infectiousness at any level, but infectiousness at higher levels had more effect than infectiousness at lower levels. It was also shown that increasing the probability of infection at a given level increased aggregation at higher level(s) and decreased aggregation at lower level(s). The results were consistent between incidence-incidence relationship and power law analysis, but the incidence-incidence relationship was more sensitive in detecting the differences in aggregation between treatments.