Selective weed management strategies and their impacts on crop yield and biodiversity

Abstract

Balancing crop productivity with biodiversity conservation in agricultural systems remains a critical challenge of our time. The emergence of autonomous and AI-driven weeding technologies, such as field robots, offers promising solutions to address this issue. To fully exploit their potential, it is essential to develop targeted weed management strategies that can be effectively implemented within the field.<br /> The aim of this thesis is to investigate on which type of input information (e.g., weed cover, weed species) a weed management strategy must be based on and which level of technological precision for weed removal is needed to reconcile biodiversity and crop production. To assess the effects of various weed management strategies on crop yield and biodiversity parameters, different strategies were modelled using three datasets from organically farmed fields. The results demonstrated that sustaining long-term biodiversity within fields without compromising crop yields necessitates intervention technologies capable of operating at the plant level. Furthermore, these technologies need to be capable of identifying and quantifying weed species and processing this information in real time. In the course of evaluating weed management strategies research gaps regarding species-specific effects of weeds, both competitive and beneficial, were identified.<br /> To address these data gaps, two field experiments were conducted. To evaluate the competitive effects of weeds, different weed species were sown as monocultures and as mixtures beside different crops. The results of the experiment demonstrated that decisions regarding weed management strategies must consider not only the weed species or composition of a weed community but also their current relationships with the crop plant. General assumptions about the occurrence of the competitive effects of a specific weed species or a community are not reliable if they are not considered in the specific context.<br /> To investigate the beneficial impacts of different weed species on associated biodiversity, a video-based monitoring system was tested to measure species-specific flower visitation rates on various weed species within organically farmed fields. This method offers significant potential to accelerate data collection on the interactions between weeds and their associated biodiversity. However, for data such as flower visitation rates to serve as effective input for decision-making in weed management, it is essential to define specific objectives – such as promoting particular ecosystem services – in advance. This is crucial given the complexity and context-dependent nature of interactions between weed species and their associated fauna.<br /> Further interdisciplinary research is needed to advance selective weed management strategies. To anticipate long-term effects, the selection pressure exerted by the implementation of selective weed management strategies on weed populations and their associated species needs to be investigated. The use of multi-sensor systems for data collection, combined with modelling the gathered data across various environments, holds significant potential to accelerate research in this field.