Towards strategies to manage weeds in turf without herbicides

Hahn, Daniel


The sportsturf industry in Europe is moving toward non-chemical broadleaf weed management. This thesis explored building a framework for alternative weed management strategies in turfgrass areas. Most importantly, I focused on the strategy of using suitable turfgrass species to maintain vigorous, dense turfgrass that is competitive against weeds. In light of climate change and stricter regulations regarding water, fertilizer and pesticide use, practitioners recently opted to use low input species such as Festuca spp. We therefore investigated the growth interfering capacity of 27 cultivars from five Festuca species (Chewings fescue [F. rubra L. ssp. fallax (Thuill.) Nyman], slender creeping red fescue [F. rubra L. ssp. littoralis (G.Mey.) Auquier] strong creeping red fescue [F. rubra L. ssp. rubra Gaudin], hard fescue [F. brevipila Tracey] and tall fescue [Schedonorus arundinaceus (Schreb.) Dumort., nom. cons.]) against three common broadleaf turfgrass weeds, namely clover (Trifolium repens L.), daisy (Bellis perennis L.) and yarrow (Achillea millefolium L.). In a climate chamber, 60 Festuca seeds were placed on water agar in a series of plastic containers for 30 days. Thirteen days after sowing, twenty weed seeds were introduced to each container, and germination and root length of those weeds was recorded. Interference by presence of Festuca species did not affect weed seed germination, but a pronounced negative effect on weed root growth was observed, with reductions of up to 85%. Clover was most severely affected in the presence of tall fescue, whereas all fescue species caused a similar reduction in root length for yarrow. Within most Festuca species we observed cultivar differences in growth interfering capacity. Weed species used in this experiment differed in their susceptibility to interference by fescue, with yarrow being more sensitive to growth interference by Festuca cultivars than clover. Daisy was most sensitive, and due to high mortality rates the species was removed from the experimental analysis. While we conducted the growth chamber screening, we also sowed a field trial with six cultivars representing each species used in the growth chamber experiment and four weed treatments including clover, daisy, yarrow and a mixture of these species in a randomized block design replicated by year. Weather conditions varied between years and caused different results, however cultivar Musica (Chewings fescue) and Barpearl (slender creeping red fescue) were least affected by weed growth over both years and resulted in acceptable visual sward quality. Manual counting of weeds with a 100-point quadrat in the above-described experiment was time consuming and limited the number of recordings. We therefore collected aerial multispectral images and applied random forest model (RF) machine learning algorithms to quantify vegetation cover using image analysis. Object-based classification using spectral features from a previous segmented orthoimage resulted in highest classification accuracy to detect weeds with 99% accuracy and high agreement to point quadrat measurements on the ground. Particularly weeds with distinctive shape features, such as daisy, were clearly detectable and had a good agreement with ground measurements. We believe that development of an automated weed recognition tool would greatly improve scalability and quality of turf research in the future and would also have applications in the early detection of weed cover in amenity turf. We conclude that weed control without traditional herbicides requires defining the purpose of turfgrass areas, establishing threshold levels for control, management strategies to maintain dense turf cover, early detection of weed species, and alternative control measures such a mechanical removal or development of bioherbicides.