Weed biodiversity and herbicide intensity as linked via a decision support system

Abstract

Background: Extensive herbicide use is one reason for the declining biodiversity of arable weeds. This study aimed to investigate (i) whether herbicide decisions recommended by a decision support system increase the weed species diversity compared to standard recommendations, and (ii) whether high weed species diversity reduces herbicide intensity, which in turn contributes to higher diversity. Data on weeds and herbicide applications in winter wheat fields in north-eastern Germany were collected in 15 field trials over 2 years. Five treatments differed in the way of decision-making for herbicide application, including two treatments according to recommendations of decision support systems.

Results: Along the Hill's series biodiversity metrics, the untreated control had the highest species richness (13.5 m^-2) per field but showed increasingly stronger dominance structures than the treated plots (equivalent species richness: 1.7-2.0 m^-2). The treatment frequency index as a metric for herbicide intensity was significantly lowest in the decision support system with low reliability (1.07). Path models, including weed diversity and density in autumn, weed diversity in summer, and herbicide intensity as a mediating variable showed a significant decreasing effect of Shannon diversity on herbicide intensity in all treatments. Only the decision support systems reacted to low weed densities with a significant reduction of the herbicide intensity.

Conclusion: Higher weed species diversity contributes to lower herbicide intensity, which is ecologically and economically valuable. Decision support systems for herbicide application should have other target functions than cost reduction for contributing to biodiversity. © 2025 The Author(s). Pest Management Science published by John Wiley & Sons Ltd on behalf of Society of Chemical Industry.

Keywords: Shannon index; herbicide decision; path model; species richness; weed density; winter wheat.

Figure 1

Field trial locations. top left: Germany with the north‐eastern region (grey); main map: north‐eastern region with trial locations in the years 2011–2012 (●) and 2012–2013 (Δ).

Figure 2

Graphical representation of a path model on the mutual influences of weed diversity and herbicide intensity, including weed density. Solid arrows a–e: regressions (explanatory variable at the start point explains response variable at the end point of the arrow); dashed arrow f: correlation (end point variables are related).

Figure 3

Weed species composition, richness, true diversity (= exp(Shannon index)) and density (m⁻²; means per field) on 15 winter wheat experimental fields in the cropping years 2011–2012 and 2012–2013. Weeds were counted in autumn before weed management at a wheat development stage of one to three leaves. Weed counting was block wise (four blocks) before the set‐up of treatment plots. Species densities are proportional to the segment areas in the star plots.

Figure 4

Diversity indices of weed communities along the Hill series. Indices are based on weed biomass (dry mass) of 15 winter wheat field trials, herbicides treated due to different decision making. DSSstd, decision support system standard; DSSred, decision support system with reduced target efficacy; AdvLoc, local advisor; AdvSer, advisory service; S, species richness; H′, Shannon entropy; D ₂, inverse Simpson index; BP⁻¹, inverse Berger–Parker index.

Figure 5

Weed diversity–herbicide intensity path model: direct effects are presented by path model coefficients (P(z)‐values). Data are based on 15 winter wheat field trials with different herbicide decision making. DSSstd, decision support system standard; DSSred, decision support system with reduced target efficacy; AdvLoc, local advisor; AdvSer, advisory service. Diversity in autumn is based on counts, diversity in summer after herbicide treatments is based on biomass. Herbicide intensity was determined as treatment frequency index (TFI). Paths a–f show direct effects from the start point to the end point of the arrows. Paths a–f refer to Table 3 and to the text.

Figure 6

Ordination biplots (redundancy analysis (RDA), left: axes 1 and 2, right: axes 1 and 3) showing the effects of different ways of herbicide decision‐making on the weed species composition. DSSstd, decision support system standard; DSSred, decision support system advising reduced efficacies; AdvLoc, local advisor; AdvSer, official advisory service. Species names: AGGRE, Elymus repens; APESV, Apera spica‐venti; BRSNN, Brassica napus; CAPBP, Capsella bursa‐pastoris; CHEAL, Chenopodium album; CENCY, Centaurea cyanus; CIRAR, Cirsium arvense; EQUAR, Equisetum arvense; GALAP, Galium aparine; HORVX, Hordeum vulgare; IUNSS, Iuncus spec.; MATSS, Matricaria spec.; MYOAR, Myosotis arvensis; PAPRH, Papaver rhoeas; POAAN, Poa annua; POLAV, Polygonum aviculare; STEME, Stellaria media; TTLWI, Triticale (winter); VERHE, Veronica hederifolia; VIOAR, Viola arvensis (EPPO codes ³⁴ ).