Introduction

Insects account for 80% of the animal species in the world and, therefore, the recently reported unprecedented rate of insect decline is cause for alarm about the state of biodiversity on Earth (Dirzo et al., 2014; Hallmann et al., 2017; van Klink et al., 2020). An array of biodiversity metrics provide strong evidence for declines of especially terrestrial insects across all continents (van Klink et al., 2020). This includes declines in abundance of both common and rare species (Hallmann et al., 2017; Seibold et al., 2019; Pilotto et al., 2020; van Klink et al., 2023) and in total species richness and changes in assembly composition (Wagner et al., 2021; Blowes et al., 2022). In some areas, terrestrial insect biomass decreased by 75% in the last 30 years (Hallmann et al., 2017), and a further 40% of all species is estimated to face extinction by 2100 (IPBES, 2019). Insects are essential for crop production through their role in decomposition, pollination, pest control and sustaining food webs. Therefore, erosion of insect communities can have potentially devastating effects on ecosystem functioning, provision of ecosystem services and ultimately on human civilization (IPBES, 2019). The main drivers of insect biodiversity decline are habitat loss due to conversion to agriculture, pollution, invasive species and climate change (Díaz et al., 2019; Wagner et al., 2021; Müller et al., 2023).

Strategies for biodiversity conservation in conjunction with adequate food production require understanding of how biodiversity responds to agricultural management (Saunders, 2020). The idea of land sparing seeks to conserve natural habitats by separating high-yielding agriculture from protected natural habitats, while land sharing combines food production with biodiversity conservation on agricultural land (Green et al., 2005; Loconto et al., 2020). Combining both pathways is vital (Fischer et al., 2008; Scariot, 2013) as land sparing counteracts land-use change and land sharing makes agricultural landscapes better suited to support biodiversity and associated ecosystem services that support crop production (Dainese et al., 2019). Ideally, sustainable agricultural practices retain yield and enhance biodiversity, preventing the need to convert natural habitats to agricultural land to maintain food production (Tscharntke et al., 2021).

Increasing crop heterogeneity can facilitate biodiversity conservation in highly productive agricultural landscapes without compromising yield (Martin-Guay et al., 2018; Sirami et al., 2019). Crop diversification can enhance niche complementarity by creating heterogeneous habitats and increasing availability and diversity of resources (Lichtenberg et al., 2017; Tamburini et al., 2020). A promising crop diversification strategy is strip cropping, where crops are grown in alternating strips, wide enough for using standard agricultural machines yet narrow enough to facilitate ecological interactions among crops (Ditzler et al., 2021). Crops that are grown in strips may benefit from increased resource use efficiency, and the suppression of pests and diseases (Tajmiri et al., 2017b; Ditzler et al., 2021; Juventia et al., 2021; Alarcón-Segura et al., 2022; Cuperus et al., 2023; Karssemeijer et al., 2023; Rakotomalala et al., 2023; Croijmans et al., 2024) without major yield compromises (Tajmiri et al., 2017a, b; Mollaei et al., 2020; van Oort et al., 2020; Juventia et al., 2021; Campanelli et al., 2023). Growing multiple crops on a field may foster a larger diversity of organisms than monocultures through greater plant species richness that cascades into richer herbivore and predator communities (Crutsinger et al., 2006; Cuperus et al., 2024). Moreover, the expected increase in available and potentially complementary niches within the agricultural field due to higher spatial diversity in crops can result in admixture of communities related to individual crops (Hummel et al., 2012), or the creation of completely new communities by enhanced richness of agriculture-related species, and/or the occurrence of species rarely found in agricultural fields (Rischen et al., 2021). So far, it is not well understood if and how insect communities respond to strip cropping across distinct crops.

Here, we present data of four years of pitfall trapping of ground beetles at several moments during the growing season in 14 crops at four organic experimental farms across the Netherlands where strip cropping was compared to monocultures. Ground beetle community dynamics are sensititve to changes in farming practices and are frequently used to examine agricultural sustainability (Holland et al., 2002; Makwela et al., 2023). Furthermore, ground beetle species are important for maintaining ecolocal functions as they comprise scavengers and predators of (weed) seeds, detritivores (e.g., collembolas and earthworms) and herbivores (e.g., aphids and caterpillars). As such, they are often used as indicator group for wider insect diversity in agroecosystems (Gerlach et al., 2013). We first examine whether strip cropping fields have greater ground beetle richness and activity density than monocultural fields. Next, we present biodiversity metric changes for individual crops to assess the benefits of strip cropping for ground beetle biodiversity. We then evaluate whether ground beetle community changes are caused by admixture of communities, whether these assemblages promote species associated with agriculture or with natural ecosystems, and whether they contain rare species.

Results

A total of 49,199 ground beetles belonging to 71 species were caught using pitfall traps over four years at four different organically managed experimental locations in The Netherlands: 40,159 at Almere; 3,777 at Lelystad; 1,126 at Valthermond; and 4,137 at Wageningen (Table S1, S2).

Strip cropping enhances ground beetle richness

Strip cropping fields had on average 15% higher ground beetle taxonomic richness than monoculture fields after rarefaction to the number of samples of the least-sampled crop configuration (β₀ = 0.151, SE = 0.044, p < 0.001; Fig. 1c,d). However, strip cropping fields did not harbour more species than monocultural fields with the highest ground beetle richness (β₀ = - 0.008, SE = 0.037, p = 0.821; Fig. 1d). The difference in field-level taxonomic richness could not be explained by an increase in the number of ground beetle species per crop in strip cropping compared to monocultures. At crop-level, the 5% increase in ground beetle taxonomic richness in strip cropping was not statistically significant (Fig. 1f). Similarly, crop-level absolute evenness, inverse Simpson index and Shannon entropy did not differ significantly among crop configurations (Fig. 1h-j, S1c,d, S2). The effect of crop configuration on crop-level taxonomic richness was variable and was not associated with location or crop species. The effect ranged from 56% more species in potato in monoculture at Wageningen in 2022, to 136% more species in barley in strip cropping at Wageningen in 2020 (Fig. 1k).

Fig. 1.

Strip cropping enhances ground beetle activity density

Ground beetle activity density was on average 30% higher in strip cropping fields than in monoculture fields (β₀ = 0.303, SE = 0.121, p = 0.012; Fig. 1e), based on rarefaction. However, there was no significant difference in activity density between the strip cropping fields and monocultures with crops that harboured the richest beetle communities (β₀ = -0.110, SE = 0.088, p = 0.215). Crop-level activity density of ground beetles was not affected by crop configuration (Fig. 1g, S1a,b).

Crop configuration alters ground beetle community composition

Ground beetle communities were significantly influenced by crop configuration, but the effects of location, year and crop dominated those of crop configuration (Fig. S3-S5, Table 1, S3). The context-dependency of configuration effects on ground beetle communities is illustrated in redundancy analyses of ground beetle assemblages per crop combination at Wageningen in 2021 and 2022 (Fig. S6). Here, we found distinct ground beetle communities among crop configurations for pumpkin, barley and potato in 2021, and for cabbage and oat in 2022. In the other cases the difference between crop configurations was not significant (Fig. S6, Table S4). Moreover, in all crop combinations except for potato-grass in 2021, the difference in ground beetle communities between the monocultures of the consitutuent crops were significant, while this was never the case for ground beetle communities of crops in strip cropping (Table S4). This indicates that strip cropping leads to overlapping crop-related communities. However, these results could be spatially biased as samples from different crops were in closer proximity of each other in strip cropping than among monocultures (Fig S7d).

Table 1.

Rare and common species indicators for strip cropping

Eleven ground beetle species were significantly associated with crop configuration: four species associated with strip cropping and seven with monocultures (Fig. 2, Table S5). Specifically, the rare Harpalus griseus (p=0.001, specificity=0.900) and the more common Anchomenus dorsalis (p=0.001, specificity=0.806) were strongly associated with strip cropping (Fig. 2, Table S5).

Fig. 2.

Discussion

It is well established that different crop types have distinct ground beetle communities (Holland and Luff, 2000; Eyre et al., 2009) and that increasing habitat diversity by including multiple crops in a field can enhance ground beetle diversity (Puliga et al., 2023; Cuperus et al., 2024). Our study shows that strip cropping increased field-level ground beetle richness by 15%. However, ground beetle richness in crops grown in strip cropping and in monocultures were mostly similar, and there were relatively few ground beetle species that were unique in strip cropping. In fact, our indicator species analysis showed that more species were associated with monocultures than with strip cropping. This indicates that the 15% increase in richness at the field level can be mostly attributed by the higher number of crops in strip cropped fields that harboured crop-specific ground beetle communities, and that there was only a limited mixture of ground beetle species among crops in strip crops. This is in line with findings from Anderson et al. (2024) who found that ground beetle movement is reduced by crop edges. Further research on movement behaviours of ground beetles at crop edges might help explain how ground beetles distribute themselves within a strip cropping field, and whether they utilize the different resources provided by a more diverse cropping system.

The 30% higher ground beetle activity density in strip cropping fields compared to monoculture fields may be explained by a more stable and diverse habitat with refuges and alternative resources in strip crops (Ratnadass et al., 2012). Crop diversification enhances prey biomass for ground beetles (Lichtenberg et al., 2017) and increases weed seed biomass as compared to monocultures (Carbonne et al., 2022; Ditzler et al., 2023), both of which reduce bottom-up control by increased food provision. In addition, insectivorous farmland birds regularly prefer more diverse fields for foraging, consequently increasing top-down control (Josefsson et al., 2017). Crop diversification might thus influence bottom-up and top-down control of ground beetles in opposing ways and the relative strength of each might explain the variation in responses of ground beetle activity density among locations and crops. Multi-taxa evaluation of future strip cropping studies will be a valuable approach to increase understanding of biomass flows and trophic interactions within agricultural systems.

The ground beetles H. griseus and A. dorsalis were specifically associated with strip cropping. The granivorous H. griseus may have benefitted from the relatively high weed abundance and diversity in the strip crops (Carbonne et al., 2022; Ditzler et al., 2023). A. dorsalis particularly associated with strip cropping at Almere, which could have been caused by the closeness of preferred hibernation locations such as grass-clover strips in the strip cropping configuration, compared to monoculture crops (Eyre et al., 2009; Marrec et al., 2017). Seven out of the eleven crop configuration indicator species in our study are typical eurytopic species, adapted to a wide range of environments and common to the Netherlands (Turin, 1991; Turin, 2000). Amara aenea, Calathus melanocephalus, Loricera pilicornis and Trechus quadristriatus were associated with monocultures. These polyphagous, generalist species are strongly adapted to crop monocultures characterised by large-scale physical disturbance through tillage and harvest (Kromp, 1999). Overall, our data indicate that multiple ground beetle species have a preference for a specific crop configuration.

Biodiversity gains in our study did in most cases not coincide with productivity losses (Table S6). Earlier studies on crop yield in the strip cropping fields in Lelystad and Wageningen during the years of pitfall trapping show a yield decrease when strip cropping cabbage (Carrillo-Reche et al., 2023) and wheat (Ditzler et al., 2023), whereas potato yield was unaffected by crop configuration (Ditzler et al., 2023). In Almere bean and parsnip yields were higher in strip cropping, whereas oat and onion yield were lower (Juventia et al., accepted). In Valthermond crop productivity was similar for monocultures and strip cropping (Table S6). Therefore, strip cropping has the potential to reconcile farmland biodiversity conservation and agricultural production.

We show that changing crop configuration from monoculture to strip cropping enhances ground beetle richness by 15% and activity density by 30% within agricultural fields. These results show that strip cropping leads to increases in biodiversity that approach those achieved by shifting from conventional to organic farming practices (+19% richness, Lichtenberg et al., 2017; +23% richness, Gong et al., 2022) and by other in-field diversification measures (+23% richness, Lichtenberg et al., 2017; +24% richness, Beillouin et al. (2021)). While organic management or in-field diversification measures generally lead to lower productivity (Gong et al., 2022), strip cropping can maintain crop productivity without taking land out of production (Tajmiri et al., 2017a, b; Mollaei et al., 2020; van Oort et al., 2020; Juventia et al., 2021; Campanelli et al., 2023), although occasional yield reductions have been reported (Carillo-Reche et al., 2023; Ditzler et al., 2023). The biodiversity gain in our study was achieved without additional crop-level diversification strategies designed for biodiversity conservation, such as cover cropping or (flowering) companion plants. This increase in biodiversity stacks on top of already higher biodiversity achieved through organic management (Lichtenberg et al., 2017; Gong et al., 2022). Furthermore, the estimated richness effects of strip cropping, i.e., no change at crop level and 15% increase at field level, are conservative because they only consider the pairwise comparison of biodiversity of two or three crops in a strip cropping configuration to monocultures, whereas the inclusion of more crops within strip cropping could further enhance ground beetle biodiversity. Also, measurements were conducted in relatively young strip cropping fields, which were not older than five years (less than a full rotation). Ground beetle communities might require more time to adapt to the new crop configuration (Rusch et al., 2013). The ground beetle communities in our study were dominated by farmland species and contained only few rare species. Nevertheless, even after only four years, specific species strongly associated with strip cropping were found. For Dutch farming systems, which are dominated by non-flowering crops, increased use of flowering crops or inclusion of semi-natural habitats with flowering plants might be necessary on top of strip cropping to increase biodiversity of groups such as pollinators. Strip cropping in combination with other in-field diversification measures, such as the establishment of more perennial semi-natural habitats (Sirami et al., 2019; Rischen et al., 2021) or uptake at larger spatial extents (Tscharntke et al., 2021) could further enrich agricultural landscape diversity and help bending the curve of biodiversity loss.

Materials and methods

Study area

A multi-location study was conducted on four organic farms across the Netherlands (Fig. 1a). Three experimental farms were managed by Wageningen University & Research (Lelystad, Valthermond, Wageningen) and one commercial farm was managed by Exploitatie Reservegronden Flevoland (ERF B.V.) located in Almere. All four locations contained both strip cropping and monocultural crop fields, but differed in soil type, establishment year of the strip cropping experiment, number of crops grown, length of the crop rotation, number of sampled crops and sampling years, and farm and landscape characteristics such as percentage of on-farm semi-natural habitat (SNH), mean field size, and landscape configuration (Fig. 1b, S7). The locations Almere and Lelystad were located in a homogeneous, open polder landscape characterized by intensive arable crop production and non-crop habitats consisting of grass margins, tree lines and watercourses. Valthermond was located in an open, reclaimed peat landscape with intensive arable crop production characterized by long and narrow fields separated by grassy margins and ditches and limited areas of woody elements. The site at Wageningen was located in a more complex landscape with smaller field sizes, and non-crop habitat consisting of woodlots, hedgerows, tree lines, ditches and farmyards.

Experimental lay-out

At Almere and Valthermond the crops in strip cropping were all grown alongside each other, whereas at Lelystad and Wageningen two alternating crops (crop pairs) were grown alongside each other (Fig. 1b, S7). At each location, strip cropping and monoculture fields were always paired on the same experimental field (Fig. S7). At Almere eight different crops were grown in alternating strips of 6 m width, including celeriac (Apium graveolens var. rapaceum), broccoli (Brassica oleracea var. italic), oat (Avena sativa), onion (Allium cepa), parsnip (Pastinaca sativa), faba bean (Vicia faba), potato (Solanum tuberosum) and a mix of ryegrass and white clover referred to as grass-clover (Lolium perenne/Trifolium repens) (Fig. S7a). At Lelystad four different crop pairs were grown in alternating strips of 3 m width, including carrot (Daucus carota subsp. sativus) and onion, white cabbage (Brassica oleracea var. capitata) and wheat (Triticum aestivum), sugar beet (Beta vulgaris) and barley (Hordeum vulgare), and potato and ryegrass (Fig. S7b). At Valthermond eight different crops were grown in alternating strips of 6 m width, including potato, barley, barley mixed with broad bean (Vicia faba) in 2020, barley mixed with pea (Pisum sativum) in 2021, sweet corn (Zea mays convar. saccharata var. rugosa), sugar beet, common bean (Phaseolus vulgaris), and grass-clover (Fig. S7c). At Wageningen three different crop pairs were grown in alternating strips of 3 m width, including white cabbage and wheat (2019 – 2021) or oat (2022), barley and pumpkin (Cucurbita maxima, 2020 - 2022) or bare soil due to crop failure (2019), and potato and ryegrass (Fig. S7d). The crop combinations and neighbors were selected based on literature, expert knowledge and experience of functionality in terms of expected advantages for yield and pest and disease control. Large scale monoculture plots (0.25 ha to 2.30 ha) served as reference, hereafter referred to as monoculture (Fig. S7). At Lelystad, Valthermond and Wageningen not each crop grown in strips was present as monoculture in each year, but only those crops for which a monoculture was present were sampled (Fig. 1k, Table S7). All fields were managed according to organic regulations, yet at each location fertilization and weed management reflected regional practices and were adjusted to local soil conditions. Flower strips were sown within the experimental fields at Almere and Valthermond (Fig. S7, Table S8).

Sampling

The ground beetle community was sampled using pitfall traps in all crops for which both a monoculture and strip cropping field were present, at each location and in multiple rounds per year between March and September. The sampled crops, number of rounds and number of pitfalls differed per year and per location (Table S7). Pitfall traps consisted of a plastic cup (8.5 cm diameter) placed in the soil so the top of the cup was level with the soil surface. Pitfalls were filled with water mixed with non-perfumed soap and covered with a black roof (12.5 cm diameter). Pitfall trap type was similar in all years and locations and traps were placed at the same specific location per year. For all analyses, year series were made, in which all ground beetle catches from the same pitfall trap were pooled per year.

Statistical analyses

We used R, version 4.2.2 for all statistical analyses.

Effect of crop configuration on field-level richness and activity density

To analyse the difference in species richness and activity density between monocultures and strip cropping configurations at field level, we used rarefaction of samples within the same field. To rarify to an equal sampling intensity, we calculated the average cumulative number of species or individuals within x year series, where x is the largest number of year series available for the crop configuration comparison. Next, we calculated the relative change due to strip cropping by subtracting the number of species or individuals found in the monoculture field from the number found in the strip cropping field, and then dividing the result by the number of species or individuals in the monoculture. This gave the relative change centered around zero, where negative values indicated higher richness in monocultures and positive values higher richness in strip cropping. We then analysed this data using Generalized Linear Mixed Models (GLMM) with a Gaussian distribution, and assessed whether the intercept deviated significantly from zero. As random variables, we used location and year, with year nested in location. We ran these analyses using a dataset that included all comparisons among monocultural fields and strip cropping fields of all locations. To test whether strip cropping was more beneficial for beetle diversity than the most beetle rich monoculture, we separately analysed a dataset that only included the monocultural fields with the highest taxonomic richness or activity density among the constitutive crops of the strip cropping field. Generalized linear models were run using the glmmTMB package and tested for model fit using the DHARMa package.

Effect of crop configuration on crop-level biodiversity

To quantify biodiversity we used five variables: (1) activity density, the total number of ground beetles found per year series; (2) taxonomic richness, the total number of species or genera (lowest taxonomic level available) found per year series; (3) the inverse Simpson index, the inverse of the sum of proportions of different species over the total abundance (Simpson, 1949); (4) absolute evenness, the number of effective species calculated by dividing the inverse Simpson index by the taxonomic richness (Williams, 1964); and (5) Shannon entropy (Shannon and Weaver, 1949). We chose absolute evenness as our measure for evenness, as this method removes the richness component from the inverse Simpson index and adheres to all requirements for an evenness index (Smith and Wilson, 1996; Tuomisto, 2012). We included both the inverse Simpson index and Shannon entropy as the former is more sensitive to changes in eveness and the latter to species richness (DeJong, 1975).

To analyse the effect of crop configuration on ground beetle activity density, taxonomic richness, evenness, inverse Simpson index and Shannon entropy we used GLMM. We constructed models for each response variable, using data from all four locations. In these models we included crop configuration (monoculture or strip cropping) as a fixed factor. We included location, year and crop as nested random variables in these models. To quantify and visualize the variation in responses between locations, years and crops, we ran generalized linear models (GLM) with a variable that combined these three variables into one, which was also included as a fixed factor.

Here, we also included the interaction between crop configuration and the combined variable for crop, location and year. For the model on activity density we used negative binomial distribution (log link function); for richness, evenness and Shannon entropy we used Gaussian distribution; and for inverse Simpson index we used gamma distribution (inverse link function). All models were tested for model fit using the DHARMa package.

Community composition of crop configurations and crops

To assess whether crop configurations have distinct ground beetle communities, we used permanova with a Hellinger transformation (with 999 permutations). We only used data from locations and years where pitfall catches had been identified to species level (Almere 2021/22, Lelystad and Wageningen). We considered four models, one for all locations combined and one for each location separately. In all models we included crop configuration and the nested variables of location (whenever applicable), year and crop as fixed factors. We also analysed the interaction between crop configuration and this nested structure of location, year and crop. To visualize crop-configuration-dependent ground beetle communities, we used redundancy analysis (RDA) on Hellinger transformed data. Here, we again conducted four analyses, one for all locations combined and one for each of the three considered locations. We only used crop configuration as a predictor, to force RDA to show any change in ground beetle community associated with crop configuration. As such, only one RDA axis was created per model, which was plotted against the first principal component describing the residual variation.

Due to the large influence of location and year on ground beetle communities, visualizing any general effects of strip cropping on these communities using all available data was challenging. To address this, we conducted RDA and visualized the effect of crop configuration on a subset of the data from one location and one year. We chose the data from Wageningen in 2021 and 2022 for this analysis because it provided a set-up where strip cropping of two crops could be compared with their constituent monocultures within the same experimental fields. Here, we used both crop configuration, crop and their interaction term as explanatory variables for each crop pair separately. To analyse whether ground beetle communities significantly differed among combinations of crops and crop configurations, we ran pairwised permanova on all three fields and two years separately, using the “pairwiseAdonis” package.

Crop configuration specific species

To assess whether there are crop-configuration-specific species we used indicator species analyses (ISA) (indicspecies package). Here, we excluded data from Almere in 2020, as they only contained genus-level identification. As this method does not allow for conditioning the data on location, we analysed the data for the different locations separately. To test whether species mainly occur in a monoculture or in strip cropping, we ran an ISA over the whole dataset per location. ISA uses two indicators: specificity and sensitivity of species to a certain crop configuration. Specificity refers to the predictive value of the species as indicator of the specific crop configuration, whereas sensitivity refers to the percentage of samples of the respective crop configuration that include the species. Both indicators do not take the abundance of species into account (de Cáceres, 2023).

Statements

Funding

This study has received funding from the European Union’s Horizon 2020 research and innovation program under grant agreements No 727482 (DiverIMPACTS) and No 727672 (LegValue), from the Dutch Public Private Partnership research program under grant agreement No LWV19129 (Crop diversification), from regional funds provided by provinces Groningen and Drenthe, and through internal Wageningen University and Research funds financed by the Dutch Ministry of Agriculture, Nature and Food Quality under grant agreement No KB36003003 (Nature Based Solutions in Field Crops). This publication is part of the project CropMix (with project number NWA.1389.20.160) of the Dutch Research Agenda (NWA-ORC) which is (partly) financed by the Dutch Research Council (NWO).

Author contributions

Luuk Croijmans: Conceptualization, Methodology, Formal analysis, Investigation, Data curation, Writing – original draft, Writing – review & editing, Visualization, Supervision. Fogelina Cuperus: Conceptualization, Methodology, Formal analysis, Investigation, Data curation, Writing – original draft, Writing – review & editing, Visualization, Supervision. Dirk F. van Apeldoorn: Conceptualization, Writing – review & editing, Visualization, Supervision, Project administration, Funding acquisition. Felix J.J.A. Bianchi: Conceptualization, Writing – original draft, Writing – review & editing, Project administration, Funding acquisition. Walter A.H. Rossing: Conceptualization, Writing – original draft, Writing – review & editing, Project administration, Funding acquisition. Erik H. Poelman: Conceptualization, Methodology, Writing – original draft, Writing – review & editing, Supervision, Project administration, Funding acquisition.

Ethics approval

This research was not reviewed by an institutional or governmental regulatory body as the work was performed on invertebrates.

Availability of data and materials

Data will be made publicly available upon acceptance in the 4TU repository.

Supplementary figures

Fig. S1.

Fig. S2.

Fig. S3.

Fig. S4.

Fig. S5.

Fig. S6.

Fig. S7.

Supplementary tables

Table S1.

Table S2.

Table S3.

Table S4.

Table S5.

Table S6.

Table S7.

Table S8.

Acknowledgements

First, we are grateful to Ron Anbergen, Martine Arkema, Bart Burger, Jonas Driessen, Michiel van de Glind, Angelo Grievink, Roelof Gruppen, Rolinde de Haan, Willem Hendriks, Nashita Maniran, Marina Martino, Sara Michellin, Ciska Nienhuis, Ralph Rustom, Simone Verdonschot, Pim Vrehen, Rik Waenink, and Xiaoshen Wang for their contribution to data collection. This work could not have been completed without the occasional help of many colleagues and students, for this we thank Zhaoqi Bin, Lenora Ditzler, Hilde Faber, Merel Hofmeijer, Gabriel Joachim, Stella Juventia, Peter Karssemeijer, Nelson Ríos Hernández, and any other people that helped. Thanks go to Roy Michielsen and Dirk van de Weert, for maintaining the Almere fields; the team at Wageningen Field crops, and in particular Joost Rijk and Laurens van Run, for maintaining the Lelystad fields; Gerard Hoekzema for maintaining the Valthermond fields; Olivia Elsenpeter, Esther Hofkamp, Titouan le Noc and the Unifarm staff, and in particular Andries Siepel and Peter van der Zee, for maintaining the Wageningen fields. Lastly, we thank Daan Mertens, Marcel Dicke, Liesje Mommer and Thijs Fijen for constructive feedback on this manuscript.