Effects of mechanical weeding with lightweight autonomous field robots on soil biological indicators

Regulatory constraints on herbicide use and the spread of herbicide-resistant weeds have increased the interest in mechanical weed control in European agriculture. In this context, autonomous field robots, for which mechanical weeding is currently the dominant application, are receiving growing attention in research and practice.

Mechanical weeding generally affects the upper soil layers compared to conventional tillage. Nevertheless, its higher frequency and timing may impose additional pressures on soil biodiversity through regular habitat disruption, direct damage to soil fauna, or interactions with soil water. While the effects of conventional mechanical weeding on soil biology are sparsely studied, even less is known about the effects of mechanical weeding with autonomous field robots on soil biological parameters. Robotic weed control may affect soils differently from traditional mechanical weed control due to variations in driving speed, working width, and operational frequency.

To assess potential effects on soil fauna, we conducted several field experiments in 2024 and 2025, comparing mechanical weeding by different robots (NaioOZ, FarmDroid FD20, and FarmingGT) with chemical weed control or mechanical weeding using conventional machinery. The experiments were conducted at three sites in Germany and were cropped either with sugar beet (Beta vulgaris) or maize (Zea mays). The first site in Eastern Germany (landscape laboratory patchCROP) is sand dominated (Loamy sand), whereas the soils at the second site in Bavaria and the third site in central Germany have finer textured soils (Silty loams).

Several biological soil indicators were assessed depending on the experimental site, including feeding activity using Von Törne bait lamina sticks placed in consecutive periods starting directly after the last of several weeding operations, carabid beetle and spider abundance collected via pitfall traps in consecutive sampling intervals during and after weeding, and earthworm abundance determined by hand sorting in the autumn following robotic activity in summer. In addition, chemical and physical soil parameters were determined before and after weeding, including pH, soil organic carbon content, bulk density, and aggregate stability indices.

Preliminary results indicate trends towards reduced feeding activities, decreased earthworm biomass, and lower carabid abundance under mechanical weeding with autonomous field robots, highlighting the need for systematic assessment of biological soil responses to robotic field management. We will discuss implications for soil-smart robot implementation with respect to the frequency and intensity of robotic interventions and outline future research directions.