International PhD Programme in Agrobiodiversity
Academic Year
2016 / 2019
Highlighting the role of diversity in
driving weed community dynamics
and weed:crop interactions
Author
Guillaume ADEUX
Supervisor
SANT’ANNA SCHOOL OF ADVANCED STUDIES, Pisa, Italy Institute of Life Sciences – Group of Agro-ecology
FRENCH NATIONAL RESEARCH INSTITUTE FOR AGRICULTURE, FOOD AND ENVIRONMENT (INRAE), Dijon, France
Joint Research Unit (UMR) 1347 for Agroecology – Sustainable Weed Management Department (GestAd)
PhD
In Agrobiodiversity
Highlighting the role of diversity in driving weed
community dynamics and weed:crop interactions
By
Guillaume ADEUX
Defended on the 24th of April, 2020
Thesis director: Paolo BÀRBERI
Thesis co-director: Stéphane CORDEAU
Thesis supervisor: Nicolas MUNIER-JOLAIN
JURY
Paolo BÀRBERI Professor, Sant’Anna School of Advanced Studies, IT Director
Stéphane CORDEAU Researcher, INRAE Dijon, FR Co-Director
Bärbel GEROWITT Professor, University of Rostock, DE Referee
Jonathan STORKEY Plant Ecologist, Rothamsted Research Station, UK Referee Anna-Camilla MOONEN Assistant Professor, Sant’Anna School of Advanced Studies, IT Examiner Francesco Xavier SANS SERRA Professor, University of Barcelona, ES Examiner
Acknowlegements
A special thanks to the PhD programme in Agrobiodiversity, delivered by the Scuola Superiore Sant’Anna in Pisa, IT, which gave me a unique opportunity to carry out a thesis on a subject dear to my heart and to the Centre INRAE Dijon (UMR Agroécologie) which accepted to host me.
Paolo BÀRBERI, thank you for having directed this thesis. Thank you for leaving me the time I felt was necessary to familiarize myself with different techniques or to refine analyses. Thank you for backing up my different periods abroad. Thank you for the autonomy and trust you gave me, which result today in a great deal of personal satisfaction. Thank you and all the Group of Agroecology (Stefano CARLESI, Anna-Camilla MOONEN, Gionnata BOCCI, Cian BLAIX, Fernando PELLEGRINI, Mariateresa LAZZARO, Marzia RANALDO, Simone MARINI, Hailie SHIFERAW WOLLE, Antsa RAFENOMANJATO, Dylan WAREN RAFFA, Elisa LORENZETTI, Federico LEONI) for the warm welcome within the lab. Always nice to be greeted with a handmade Ethiopian scarf, some homemade beer and some authentic Italian pasta and wine. I particularly appreciated the effort made to speak English (just for me!) at lunchtime.
Stéphane CORDEAU, I will be eternally grateful for all the help you provided. As I tend to say to new PhD students, the director is as important as the subject and I was truly lucky to have you as co-director. Despite your more-than-booked agenda, you always found the time for me (I don’t know how you do it), whether for a 5mn discussion – cigarette break, a skype call across countries to draft the outline of a paper, or for a whole afternoon of whiteboard brainstorming. You always happily accepted to review my work, even for a simple team meeting, relieving me of any additional source of stress. You never criticized my tendency to double check everything or to take considerable time familiarizing myself with new statistical techniques (or taking days off for very serious botanical reasons). Thank you for adjusting to my personality and always caring about my well-being. In hindsight, thank you for pushing me to find my own answers, even if you initially had your ideas. You gave me numerous opportunities to exchange with farmers, experimenters and researchers and I believe this gave me a wider perspective of agricultural research than what is present in this thesis. I am grateful to have shared your motivations during these three years, i.e. to vulgarize research, bring farmers and researchers together, and find sustainable solutions, always with great humility. I believe what started out as a student:director relationship became sincere friendship. Our vivid, zero-filter discussions (which scared others) would never have been possible otherwise. I can also honestly say you’re the only person with whom I’ve eaten fennel ham (for breakfast!) or played paddle tennis. Thank you for absolutely everything. Don’t change. Except for a little more sleep and little less Ricard – Rocquefort dinners!
Nicolas MUNIER-JOLAIN, a huge part of this work would not have been possible without your innovative thinking. Thank you for the experiments you initiated (even though I was still in elementary school) and which represented the core of my thesis. Your ability to provide quick, clear-cut answers in moments of intense questioning and uncertainty proved to be a real relief. You never hesitated to gather and pool different sources of funding to allow me to explore new horizons, and I thank you for that. Thank you for the numerous tennis games (but not for being a fearsome opponent!). Thank you for the great satisfaction you gave me the times I was able to make you laugh (or – more humbly – squeeze out a smile)!
I consider myself more than lucky to have been able to carry out this thesis within the « pole GestAd » of the UMR Agroécologie in Dijon, FR, a building full of people with remarkable skills in diverse fields of study. What a chance to have answers only doors away! Enumerating the reasons why a « thank you » is required would make the following chapters look small so I will limit myself to names: Nathalie COLBACH, Delphine MOREAU, Fabrice DESSAINT, Bruno CHAUVEL, Bernard NICOLARDOT, Catherine HENAULT, Benoit RICCI, Matthieu SIOL, Adam VANBERGEN, Guillaume FRIED, Lucile MUNERET, Bérenger BOURGEOIS, Hugues BUSSET, Annick MATEJICEK, Carole REIBEL,
Severine MICHEL, Emeline FELTEN, Luc BIJU-DUVAL, and Claude SARRASIN. A special thanks to Dominique MEUNIER who collected weed data in a very conscientious manner over 15 years, whom trained me to weed identification at the seedling stage and whom trained himself to « Access » days before his retirement to extract the data for me. A special thanks also to Eric VIEREN, my botanical buddy, with whom I spent numerous days in the field, always in good spirit. Thanks to my deskmate, Benjamin CARBONNE, which was always ready to lend a hand and provide some valuable tips. Thanks also to the staff of the experimental station in Bretenière (Pascal FARCY, Philippe CHAMOY, Laurent FALCHETTO, Pascal MARGET, Violaine DEYTIEUX, Guillaume POUSSOU, Alain BERTHIER, Rodolphe HUGARD, Benjamin POUILLY, Vincent CELLIER), which carried out the experiments with dedication and without whom the « archeological » part of my work would never have been possible. Thank you also to all the interns whom lent a hand for field work: Auxence BAUDRON, Alexandre LACHMANN, Maïwen ABGRALL, Léa GRALL, Justine DEGENMANN, Marion SCHWARTZ, Noémie LECOMTE. Auxence BAUDRON, it was a pleasure to co-supervise your master’s thesis, you made it too easy for me!
Thank you to the adorable organizers (Jan LEPŠ (aka “Suspa”), Petr ŠMILAUER, Francesco DE BELLO, Lars GÖTZENBERGER) of the Quantitative Ecology Module at the University of South Bohemia, CZ, with whom I had the chance to spend three months. Whom would have thought that botany, statistics, music and beer drinking could blend together so well! Thank you Suspa for the botanical excursions over the weekend and the invitations to various local events. Thank you also to Thomas GALLANDT which so successfully coordinated a group project that we were able to valorize it as a publication. Thanks to my funny brazilian roomate Jhonny CAPICHONI MASSANTE for all those mean caipirinhas.
Thank you to all the people whom lent a hand with statistics over the web: Henrik SINGMANN, Patrick BREHENY, Russell LENTH, Benjamin BOLKER, Mollie BROOKS, Torsten HAUFFE, Michael COLLYER, Geng CHEN, John FOX, Maarten JUNG, John MAINDONALD, John SORKIN, Daniel LÜDECKE and “Amoeba”.
Employment contracts, hosting agreements, mission orders, applications for periods abroad … A great deal of bureaucracy I would have never able to deal with alone. Thank you to the dedicated administration staff in both France and Italy for their incredible help: Laura BEVACQUA, Samuel VAUCHET, Francine VIEREN, Catherine MOREL, Sylvie BELOTTI and Clodine CHOTEL. Thank you to Jean-Philippe GUILLEMIN for joining the battle and making sure I would receive my “additional” sources of funding, one way or another.
Thank you to Daniele ANTICHI for his « archeological » work and his prompt responses.
Thank you to Simon GIULIANO and Maurizio SATTIN for their scientific input during the steering committees.
Thank you to Bärbel GEROWITT, Jonathan STORKEY, Francesco Xavier SANS SERRA and Anna – Camilla MOONEN for accepting to review this thesis.
Thank you to my botanical friends, Johann LALLEMAND and Cécile FRELIN, for sharing their knowledge on local flora, always in good spirit!
Thank you to my close friends from Dijon, Jérémie ZERBIB, Antonin DOUILLET, Alice CHARALIBIDIS and Thomas RIGALDO, which, on top of being a fun crowd, were always ready to help out in times of need.
Thank you to my love, Maé GUINET, for her precious help all throughout.
Publications in peer-reviewed journals
Adeux, G., Vieren, E., Carlesi, S., Bàrberi, P., Munier-Jolain, N., Cordeau, S., 2019. Mitigating crop yield losses through weed diversity. Nature Sustainability 2, 1018-1026.
Adeux, G., Munier-Jolain, N., Meunier, D., Farcy, P., Carlesi, S., Barberi, P., Cordeau, S., 2019. Diversified grain-based cropping systems provide long term weed control while limiting herbicide use and yield losses. Agron. Sustainable Dev. 39, 42.
Galland, T., Adeux, G., Dvořáková, H., E‐Vojtkó, A., Orbán, I., Lussu, M., Puy, J., Blažek, P., Lanta, V., Lepš, J., 2019. Colonization resistance and establishment success along gradients of functional and phylogenetic diversity in experimental plant communities. Journal of Ecology 107, 2090-2104.
Adeux, G., Giuliano, S., Cordeau, S., Savoie, J.-M., Alletto, L., 2017. Low-input maize-based cropping systems implementing IWM match conventional maize monoculture productivity and weed control. Agriculture 7, 74.
Adeux, G., Baudron, A., Cordeau, S., Strategic tillage in conservation agriculture systems: consequences on weed communities and winter wheat productivity. [submitted to Weed Research]
Adeux, G., Cordeau, S., Antichi, S., Carlesi, S., Mazzoncini, M., Munier-Jolain, N.M., Barberi, P., Cover crops promote crop productivity but do not enhance weed management in tillage-based cropping systems. [accepted in European Journal of Agronomy]
Publications in extension journals
Baudron, A., Adeux, G., Cordeau, S., 2019. Et si une impasse de désherbage en Agriculture de Conservation vous poussait à retravailler le sol ? Quelle intervention choisiriez-vous ? Techniques Culturales Simplifiées 103, 7-11.
Cordeau, S., Adeux, G., Chamoy, P., Farcy, P., Munier, J., N.M., 2019. On a les adventices qu’on mérite, mais ce n’est pas toujours mauvais signe ! Retour sur 17 ans d’essai INRA sur la réduction des herbicides. Techniques Culturales Simplifiées 101, 11-16.
Cordeau, S., Adeux, G., 2018. Régulation des plantes adventices par la compétition : effet des couverts et de leur conduite sur la gestion des adventices. In: INRA, Nouvelle-Aquitaine, C.r.d.a., ACTA (Eds.), Couverts végétaux : des opportunités à saisir. Rencontres régionales de la recherche, du développement et de la formation. INRA, Angoulème, France.
Munier-Jolain, N., Abgrall, M., Adeux, G., Alletto, L., Bonnet, C., Cordeau, S., Darras, S., Deswarte, C., Farcy, P., Gavaland, A., Justes, E., Giuliano, S., Meunier, D., Pernel, J., Raffaillac, D., Gleizes, B., Tison, G., Ubertosi, M., 2018. Projet SYSTEM-ECO4 : Evaluation de systèmes de grandes cultures à faible usage de pesticides. Innovations Agronomiques 70, 257-271. Munier-Jolain, N., Farcy, P., Adeux, G., Cordeau, S., 2018. Intégration des leviers de gestion de la flore
adventices à l’échelle du système de culture : 17 ans d’enseignements sur l’essai système PIC Adventices - Encadré. In: Chauvel, B., Darmency, H., Munier-Jolain, N., Rodriguez, A., (coord.) (Eds.), Gestion durable de la flore adventice des cultures. Éditions Quæ, Versailles (France), pp. 122-124.
Communications (oral and poster) in conference
Cordeau, S., Baudron, A., Adeux, G., 2019. Et si une impasse de désherbage en agriculture de conservation des sols vous poussait a re-travailler le sol ? Quelle intervention choisiriez vous ? In: AFPP (Ed.), 24e Conférence du COLUMA - journées internationales sur la lutte contre les mauvaises herbes - 3 au 5 décembre 2019, Orléans, France.
Adeux, G., Cordeau, S., Meunier, D., Farcy, P., Chamoy, P., Munier-Jolain, N., 2018. Systèmes de culture à faible usage d'herbicides : évaluation malherbologique de l'essai PIC-adventices. Séminaire DEPHY grandes cultures / polycultures élevage - 29-30 mai 2018, Dijon, France.
Adeux, G., Meunier, D., Farcy, P., Chamoy, P., Cordeau, S., Munier-Jolain, N., 2018. Evaluation malherbologique de l'essai "Protection Intégrée des Cultures" PIC-adventices. Rencontres avec les Jeunes Agriculteurs de Bourgogne Franche-Comté (JA BFC) - 20 mars 2018, Bretenières, France.
Baudron, A., Adeux, G., Cordeau, S., 2018. Systèmes de culture à faible usage d'herbicides : impact de 17 années de systèmes de culture contrastés sur la flore levée. Séminaire DEPHY grandes cultures / polycultures élevage, Dijon, France.
Baudron, A., Adeux, G., Felten, E., Vieren, E., Chamoy, P., Farcy, P., Cordeau, S., 2018. Quelles sont les conséquences d’un travail du sol dans des parcelles menés en agriculture de conservation durant 17 ans : étude de la flore adventice. In: INRA, APAD (Eds.), Gestion de l'enherbement en agriculture de conservation des sols - 31 mai 2018, Vendôme, France. Baudron, A., Adeux, G., Vieren, E., Felten, E., Farcy, P., Chamoy, P., Cordeau, S., 2018. Révélation de l'effet de 17 ans de systèmes de culture contrastés sur la flore adventice d'un blé. Visite d'essai avec les agriculteurs JA BFC - 20 mars 2018, Bretenières, France.
Métais, P., Labreuche, J., Adeux, G., Cordeau, S., 2018. Weed response to tillage and cover crop tactics in contrasted crop sequences within long-term experiments. 21st International Soil Tillage Research Organization Conference, Paris, France.
Adeux, G., Cordeau, S., Munier-Jolain, N., Barberi, P., 2017. Highlighting the role of diversity in driving weed dynamics and weed-crop interactions. 6ème Journée des Doctorants de l’UMR 1347 Agroécologie, Dijon.
Giuliano, S., Adeux, G., Cordeau, S., Savoie, J.-M., Alletto, L., 2016. Maize-based low-input cropping systems can provide effective weed control while ensuring crop productivity. In: AFPP - Association Française de Protection des Plantes (Ed.), 23. Conférence du COLUMA - Journées Internationales sur la Lutte contre les Mauvaises Herbes, (06 - 2016-12-08), Dijon, FRA.
Guiliano, S., Adeux, G., Cordeau, S., Savoie, J.-M., Alletto, L., 2016. Contrasted maize based cropping systems influence weed dynamics and impact maize productivity. In: ESA (Ed.), 14th Congress of the European Society of Agronomy Edinburgh (UK).
Courses delivered by the Scuola Superiore Sant’Anna:
- Principles of Agrobiodiversity (Paolo Bàrberi, Anna-Camilla Moonen, 20 hours) - Advanced Statistical Methods for Field Experiments (Nevio Dubbini, 24 hours) - Multivariate Data Analysis (Anna-Camilla Moonen, Elisa Pellegrino, 14 hours) - Introduction to Statistical Analysis for Agrobiosciences (Paolo Bàrberi, 20 hours)
Courses delivered by the Quantitative ecology in Ceske Budejovice, Czech Republic (3
months full time)
- Design and analysis of ecological experiments (Jan Lepš, Petr Šmilauer, 39 hours) - Community ecology (Jan Lepš, Vojtěch Novotný, 39 hours + 2 day field trip in Ceske
Krumlov)
- Functional traits in ecology (Francesco de Bello, 21 hours) - Modern regression methods (Petr Šmilauer, 39 hours)
- Creative publishing in community ecology (Jan Lepš, Francesco de Bello, Lars Goetzenberger, Jan Hrcek, 20 hours)
- Ecology seminars (8 speakers, 16 hours)
Reviewing activities
- review of one article for Nature Plants
Teaching activities
2 days of teaching (13-14/03/2018) on sampling and statistical methods for weed science at the Agronomic Engineering School of Purpan in Toulouse (reference teacher: Simon Giuliano).
Mentoring activities
Auxence Baudron’s master’s thesis (Ecole Supérieure d’Agriculture d’Angers) which focused on revealing the legacy effect of contrasted cropping systems and investigating the effect of strategic tillage.
Naturalist activities
floristic inventory of the CA-SYS platform (Experimental unit of “Domaine d’Epoisses” in Bretenière, France, 120 ha, >400 plant species).
Table of contents
General introduction ... 1
I. General context: the challenges of 21st century agriculture ... 3
II. Community assembly rules and trait-based approaches ... 6
1. Plant communities and community ecology ... 6
2. The importance of dispersal, environmental, and internal constraints ... 7
3. Coupling weed community assembly and trait-based approach... 8
III. Weed communities as a subject of study... 10
1. Definition of a weed ... 10
2. Assessment of weed communities ... 12
3. Focus on weed diversity ... 19
IV. Managing weeds while reducing herbicide use ... 23
1. Non-chemical weed management ... 24
2. Integrated weed management and cropping system approach ... 31
3. Cropping system experiments ... 32
V. Weed:crop interactions ... 34
1. Interactions among plants ... 34
2. Prerequisites for competition ... 35
3. Drivers of weed:crop competitive outcomes ... 35
4. Quantification of competitive effects ... 37
5. Mechanisms suspected to alleviate competition ... 40
VI. Presentation of the thesis ... 42
1. Objective, research questions, hypotheses ... 42
2. Overview of the experiments mobilized ... 46
3. Organization of the thesis ... 51
Chapter 1: Do cover crops contribute to long-term weed management and crop productivity? 53 Chapter Introduction ... 54
Article I: Cover crops promote crop productivity but do not enhance weed management in tillage-based cropping systems ... 55
1. Introduction ... 57
2. Material and Methods ... 59
3. Results ... 62
4. Discussion ... 68
5. Conclusion ... 70
Chapter 2: Does cropping system diversification allow reduced herbicide use and long-term
weed management? ... 77
Chapter Introduction ... 78
Article II: Diversified grain-based cropping systems provide long-term weed control while limiting herbicide use and yield losses……….79
Chapter 3: How does cropping system diversification influence weed communities? ... 93
Chapter Introduction ... 94
Article III: Weed community response to contrasted pathways of cropping system diversification 95 1. Introduction ... 96
2. Materials and Methods ... 99
3. Results ... 103
4. Discussion ... 108
5. Conclusion ... 112
References ... 113
Chapter 4: Can weed diversity mitigate crop yield losses? ... 119
Chapter Introduction ... 120
Article IV: Mitigating crop yield losses through weed diversity……….……….121
General discussion ... 131
I. Thesis highlights ... 133
II. Outlook on cropping systems experiments ... 136
1. Comparability of cropping systems ... 136
2. Environmental variability ... 138
3. Disentangling the effects of individual management factors ... 141
III. Weed diversity and crop productivity ... 142
IV. Weed diversity and yield losses ... 142
V. Focus on segetal plants ... 144
VI. Barriers to the adoption of IWM systems ... 145
VII. ... Conclusion ... 146
References ... 149
Appendices ... 171
Appendix 1: Low-input maize-based cropping systems implementing IWM match conventional maize monoculture productivity and weed control ... 172
Appendix 2: Colonization resistance and establishment success along gradients of functional and phylogenetic diversity in experimental plant community ... 189 Appendix 3: Strategic tillage in conservation agriculture: consequences on weed communities and
winter wheat productivity... 204 Appendix 4: Supplementary Methods ‘Cover crops promote crop productivity but do not enhance
weed management in tillage-based cropping systems’ – Chapter 1 ... 215 Appendix 5: Supplementary Methods ‘Weed community response to contrasted pathways of
cropping system diversification’ – Chapter 3... 227 Appendix 6: Supplementary Methods ‘Weed diversity can mitigate crop yield losses’ – Chapter 4
1
General introduction
It all has to start sometime… seedling of common verbena on winter wheat stubble.
Verbena officinalis L., 1753
Observed on the 27/06/2018 at the INRAE experimental farm in Bretenière, Côte-d’Or, France Canon EOS 100D, Obj. 55 mm, 1/640 s, F/5.6, ISO-100
3
I. General context: the challenges of 21
stcentury agriculture
The increased use of synthetic fertilizers and pesticides (7-fold increase between 1960 and 2000 for nitrogen and herbicides), new crop strains, irrigation (2-fold increase in irrigated land over the past 50 years) and heavy machinery allowed a phenomenal increase in crop productivity (Figure 1), thereby reducing hunger and improving nutrition (Tilman et al., 2002a). Global cereal production doubled from 1960 to 2000 (Tilman et al., 2002a). Between 1985 and 2005, crop yields per hectare increased considerably: 34% for cereal crops, 57% for oil crops and 11% for fodder crops (Faostat, 2020). However, agricultural expansion and intensification have generated a wide array of negative externalities (Stoate et al., 2001).
Figure 1: Borrowed from Tilman et al. (2002). Agricultural trends over the past 40 years. a, Total global cereal production; b, total global use of nitrogen and phosphorus fertilizer (except former USSR not included) and area of global irrigated land; and c, total global pesticide production and global pesticide imports (summed across all countries).
4 Each factor responsible for increased crop productivity is now questioned for environmental reasons (Stoate et al., 2001; Tilman et al., 2002a). Agricultural expansion is responsible for the loss of natural ecosystems (e.g. 70% of grasslands, 50% of savannas, 45% of temperate deciduous forests and 27% of tropical forests have been converted to agriculture) and associated ecosystem services such as climate regulation and water infiltration (Ramankutty et al., 2008; Ramankutty and Foley, 1999). Leaching, run-off and volatilization of agricultural inputs (i.e. nitrogen and phosphorous fertilizers, pesticides) has led to water and air pollution (Stoate et al., 2001). Eutrophication of waterbodies due to excessive nitrogen and phosphorus fertilization has increased purification costs and provoked low-oxygen conditions that endanger aquatic biodiversity, fisheries and recreational sites (Canfield et al., 2010; Diaz and Rosenberg, 2008). In France, certain hydrogeographic sectors have already exceeded the maximal pesticide concentration authorized for water purification (Dubois and Lacouture, 2011). Agriculture, through tropical forest clearings, livestock breeding, and nitrogen volatilization, is now responsible for ~25% of total anthropogenic greenhouse gas emissions and therefore, an important contributor to global warming (Paustian et al., 2016). Pesticide drifts have generated intense conflicts between farmers and surrounding communities (Harrison, 2011), particularly due to the growing recognition of their impact on human health (Kim et al., 2017), the extension of rural area onto agricultural land, and counterurbanization. Agricultural pests are quickly developing resistance to plant protection products, e.g. 262 herbicide resistant weeds have been documented to date (Heap, 2020). Availability of water resources is becoming more and more scarce, e.g. 20% of irrigated land in the United States is supplied by groundwater pumped in excess of recharge (Postel, 2014). The combined effect of these factors, along with simplification of landscapes, has also led to a dramatic decline in farmland biodiversity (Donald et al., 2001a; Fried et al., 2009; Sánchez-Bayo and Wyckhuys, 2019), responsible for numerous ecosystem services (see following sections for a special focus on weed diversity).
Moreover, human population is projected to increase by 45% by the end of the century (Roser, 2020) and global grain demand to double by 2050 (Tilman et al., 2011), unless there are drastic changes in agricultural consumption patterns (Stoate et al., 2001)or in human demographic trend. Such challenges are colossal considering the continuously decreasing number of farmers, the increasing prevalence of malnutrition (from lack of access to food quantity or quality), the stagnation of crop yields, market speculations, the expansion of bioenergy crops, increasing climatic disturbances etc. Therefore, agriculture is facing one of the greatest challenges of the twentieth century: provisioning enough food of good quality while reducing environmental impacts (Stoate et al., 2001) and ensuring a suitable revenue for farmers.
Sustainable weed management appears as a key point for ecological intensification in agriculture because weeds can generate severe yield losses, contribute to farmland functional biodiversity and are strongly associated to the overall use of pesticides (Petit et al., 2015). Furthermore, a growing body of literature indicates that higher weed diversity could actually mitigate crop:weed competition (Cierjacks
5 et al., 2016; Ferrero et al., 2017; Gonzalez-Andujar et al., 2019; Storkey and Neve, 2018). However, cropping systems which benefit from high weed diversity and low weed:crop competition while resorting to low herbicide use remain to be identified. Hence, the introduction of this thesis will review (i) the general framework of how weed communities are structured and assessed, (ii) how agricultural practices may be combined to achieve long-term weed management, low herbicide use, and high weed diversity, and (iii) how increased weed diversity may mitigate crop:weed competition.
********************************************************************************** Elements to remember:
Agriculture is facing multiple challenges.
Solutions are needed to promote weed diversity, maintain crop productivity and drastically reduce agricultural inputs (i.e. herbicides and nitrogen fertilizers).
6
II. Community assembly rules and trait-based approaches
1. Plant communities and community ecology
A plant community can be defined as an assembly or association of populations of two or more plant species which occupy the same space at the same time (Begon et al., 1986). Consequently, plant community ecology can be defined as the study of how plant species interact between one another, or with their environment, at different spatial and temporal scales, to determine their composition and relative abundance (Morin, 2009). More precisely, plant community ecology focuses on i) assembly rules, i.e. how plant communities are shaped from a given regional species pool, and ii) response rules,
i.e. how plant communities respond to variations in environmental conditions (Keddy, 1992). When
studying community assembly, one must distinguish between phylogeographic assembly and ecological assembly (Figure 2). Phylogeographic assembly relates to processes at large spatio-temporal scales (i.e. speciation, extinction, and migration) whereas ecological assembly refers to dispersal, biotic and abiotic processes (Götzenberger et al., 2012). Only ecological assembly will be presented here (Figure 2).
Figure 2: Borrowed from Götzenberger et al. (2012). The different processes that might produce assembly rules and the relative scales at which they are most influential. At any point in time there is a global species pool that defines a regional species pool through the speciation, extinction and migration of speceis (phylogeographic assembly). At a given local site the species pool constitutes species from the regional species pool that are able to disperse there (dispersal assembly). At the local site, habitat filtering and biotic interactions define the actual assemblage of plant species (ecological assembly).
7 Plant communities are dynamic, they constantly respond to changes in biotic and abiotic constraints that act at multiple scales (Booth, 2002). In addition, factors which contribute to the maintenance of community structure (e.g. herbivory) may differ from the factors which determined the original direction of the trajectory (e.g. competitive effect of early invaders)(Abrams et al., 1985; Petraitis and Latham, 1999; Pimm and Pimm, 1991). A community can even take different trajectories depending on the timing of the disturbance (Abrams et al., 1985), the order/rate/frequency of arrival of invaders (Abrams et al., 1985), or even converge to a state of predictable cycling (Samuels and Drake, 1997). Occasional or rare events (e.g. drought) can also have a disproportionate role in shaping communities (Strange et al., 1999).
2. The importance of dispersal, environmental, and internal constraints
A fundamental concept to the study of plant community assembly is the concept of species pool (Zobel, 2016). In its simplest form, a species pool can be defined as a set of species. One may identify two types of species pool relevant to ecological assembly: the regional species pool and the local species pool (Figure 2). The regional species pool comprises all species capable of arriving at a given site, and is thus determined by dispersal constraints (Belyea and Lancaster, 1999). The local species pool comprises all species capable of persisting under the local abiotic conditions, and is thus determined by environmental constraints (Belyea and Lancaster, 1999). However, being able to disperse and grow under the local abiotic conditions is not sufficient to become a member of the realized community. Species capable of dispersing and growing under the local abiotic conditions will be further filtered by biotic interactions (i.e. internal constraints such as herbivory, predation, facilitation and competition) to determine the actual community (Belyea and Lancaster, 1999). A species may be able to disseminate to a given site and grow under the local abiotic conditions but not persist due to its incapacity to tolerate competition from other plants. In reality, these filters interact constantly at different scales (Catford et al., 2009; Kraft et al., 2015a), rather than operating as the dispersal – environmental – biotic interaction sequence usually presented (Figure 2).
Several considerations need to be taken into account when studying weed communities (Gaba et al., 2017; Smith and Mortensen, 2017). Farmers may contribute to seed dispersal beyond species natural ability, either through contaminated seed lots or agricultural machinery. Farming practices (e.g. fertilizers and irrigation) can directly modify internal constraints such as competition by creating abiotic conditions that are more favorable to a given species or indirectly, by modifying local abiotic conditions and allowing the establishment of species that would have otherwise not been adapted. Disturbances imposed by farming practices (e.g. tillage, herbicides) are also expected to have a direct effect on weed community structure. Nevertheless, the general idea remains the same: a weed species needs to be able to disperse (either naturally or with human help), and grow under the local (abiotic and biotic) conditions determined by the interaction between farming practices and local abiotic conditions.
8
3. Coupling weed community assembly and trait-based approach
Community assembly gains considerable predictive power and comprehension when combined to a trait-based approach (Booth and Swanton, 2002; Keddy, 1992). The trait-based approach advocates for a focus on plant traits rather than species taxonomic status (Keddy, 1992; Weiher and Keddy, 1999). Violle et al. (2007) defined a trait as “any morphological, physiological or phenological feature measurable at the individual level, from the cell to the whole‐organism level, without reference to the environment or any other level of organization”. However, any measure carried out at the scale of an individual does not necessarily reflect an organism’s fitness in a given ecological habitat. This led Violle et al. (2007) to distinguish functional traits, i.e. “any trait which impacts fitness indirectly via its effects on growth, reproduction and survival”. Hence, the description of weed species through functional traits goes beyond species identity, allowing greater generalization across sites which do not share any common weed species, and to focus on the mechanistic relationship between weed species’ functional traits and response to environmental conditions (Figure 3).
Figure 3: Adapted from Weiher et al. (1999). Coupling weed community assembly and a trait-based approach. Farming practices and local abiotic conditions interact to determine local constraints which will filter species with poorly adapted traits from species with well adapted traits.
As such, local dispersion capacity will depend on intrinsic functional traits associated to species’ propagules (seed mass, shape, specific adaptations). Even if a weed species is capable of dispersing in a given field, it will have to present favorable functional traits to tolerate the new abiotic environment (e.g. adaptations to limestone). Environmental filtering can result in trait convergence (Götzenberger et al., 2012), i.e. specific trait values confer the weed species an advantage in the local abiotic conditions. Dissimilar weed species are excluded due to their incapacity to tolerate local abiotic conditions and associated stress (Grime, 1973; Mayfield and Levine, 2010). Nevertheless, biotic filtering can also lead
9 to trait convergence when a single phenotype is best adapted (Grime, 2006). Weed species which have successfully arrived in a site where they are locally adapted to the abiotic conditions will have to display advantageous trait values to cope with constraints induced by farming practices (e.g. herbicides, tillage, Figure 4). Theories (limiting similarity and trait hierarchies) and traits underlying competitive outcomes will be discussed in detail further on. All throughout, we will adopt a functional perspective to focus on how farming practices constrain arable plant communities (response traits) and how the characteristics of the selected communities may affect agroecosystem services or disservices (effect traits) (Gaba et al., 2017).
Figure 4: Borrowed from Gaba et al. (2017). Synthesis of the most significant weed functinal traits over a plant life cycle. This cycle is represented by main life stages from germination and emergence, the developement of a standing plant, a flowering plant to the production of seeds. Biotic interactions affecting the development of the plant are represented in purple, whereas the two main management practices affecting the plant performance are presented in red.
********************************************************************************** Elements to remember:
Farming practices are expected to act as a set of filters on weed species traits.
Hence, contrasted agronomic pathways should result in weed communities with contrasted set of adapted traits.
10
III. Weed communities as a subject of study
1. Definition of a weed
1.1. A human perspective
A diversity of definitions has been proposed for «weeds», highlighting the lack of consensus among stakeholders to define these organisms (Zimdahl, 2018). In the broadest sense, «any plant which grows where it is not desired» could be considered a weed (Buchholtz, 1967). The European Weed Research Society proposed a similar definition: “any plant or vegetation interfering with human objectives”. From a purely agricultural standpoint, weeds are usually defined as all species found in a cultivated field except the sown crop (Gaba et al., 2016). Hence, volunteers of a preceding crop in a subsequent crop can also be considered as weeds.
1.2. An ecological perspective
Weeds are also difficult to define from an ecological standpoint. They represent an extremely diversified group of plant species, belonging to the phytosociological class of Ruderali-Secalietea (Jauzein, 1997). According to Grime (1977), plant strategies can be resumed by tradeoffs between three components (Figure 5): competition (C), tolerance to stress (S) and tolerance to disturbance (R). Competition relates to interactions among plants, stress relates to factors which limit plant growth and disturbance relates to factors which partially or completely remove plant biomass. Considering that farming, in its traditional sense, essentially relied on repeated tillage operations for weed control, most weed species can be associated to the R strategy (Jauzein, 2001a). The R strategy includes plants defined as « ruderals », i.e. plants that prosper in situations of high intensity of disturbance and low intensity of stress. Such species usually display high growth rate, fast establishment of flowering phase, continuous and high seed production, and tolerance to a wide range of environmental conditions (Baker, 1974; Grime, 1979; Harper, 1960). Through the comparison of weed surveys in agricultural fields and grasslands, Bourgeois et al. (2019) highlighted that weeds had a more restrained ecological niche than grassland species (in terms of specific leaf area/height/seed mass or resources). Weeds associated to arable fields were characterized by an increased ability to acquire resources while tolerating competition from crops (higher specific leaf area and low Ellenberg index for light, i.e. sciophytic species), adaptations to nutrient rich environments (high Ellenberg index for nitrogen, i.e. nitrophilic species), an annual life cycle, and abilities to escape weed control and competition (earlier and longer flowering period). Nevertheless, functional niches of arable weeds and non-weeds largely overlapped, indicating that weeds also resort to a diversity of ecological strategies (Bourgeois et al., 2019).
11 Figure 5: Borrowed from (Pierce et al., 2017). Species names represent examples of the seven secondary CSR strategy classes suggested by Grime (2001). Weed species are essentially found in the R part of the triangle. C: competitors; S: stress tolerant; R: ruderals.
1.3. Origin of weeds
Agricultural fields represent a secondary habitat for the vast majority of weeds (Jauzein, 1997; Jauzein, 2001a). Indeed, the history of agriculture (~10 000 years) is very recent compared to the existence of vascular plants (~425 million years ago). Certain pre-adapted species were present at the beginning of agriculture and simply transgressed from their primary marginal habitats (e.g. dry grasslands on sand/gravel, loose stones, screes, temporary ponds, exposed river banks…) to agricultural fields (Jauzein, 1997; Jauzein, 2001a). However, this does not imply that weeds have not evolved ever since or that their primary habitat was not severely degraded (references in Albrecht et al. (2016)). Other species were on the limit of their distribution range (mainly Mediterranean species for the French case) and benefited from the favorable conditions created by the beginning of agriculture (Jauzein, 1997; Jauzein, 2001a). The presence of close relatives of certain weed species (e.g. Nigella arvensis, Papaver
rhoeas) in the Middle-East could also point out to an ancient introduction through diverse commercial
routes (Spain, Italy, and Central Europe) (Youssef et al., 2020). Former crops (which have lost commercial interest since), or plants which have evolved from domesticated ancestors, have also continuously increased the floral diversity of agricultural fields (Ellstrand et al., 2010). More recently,
12 numerous spring and summer weed species have been introduced since the « discovery » of the Americas (i.e. neophytes).
1.4. Focus on segetal plants
A more restricted group of weeds (~102 species in the French national list), called « messicoles » in French and “messicole” or “segetali” in Italian (from the latin « messis » and « colere » which means « harvest of winter cereals » and « dwell ») or « segetal plants » in English, is usually considered separately from general weed flora, for conservation purposes (Cambecèdes et al., 2012; Jauzein, 1997). Segetal plants are winter annuals of the phytosociological order Secalietalia which dwell in winter cereals (Jauzein, 1997). Segetal plants can be further distinguished into archeophytes (« messicole »
sensu stricto) and crop mimics (« messicole » sensu strictissimo). Archeophytes (in opposition to
neophytes) are plants introduced before the Middle Age which have persisted in winter cereals ever since (Jauzein, 1997; Jauzein, 2001a). Crop mimics relate to species which have evolved in such a phenomenal way to cope with farming practices that they are now considered as distinct taxa (e.g.
Lolium temulentum subsp. temulentum or linicolum, Vaccaria hispanica subsp. liniflora, Agrostemma githago, Camelina sativa, Avena sativa subsp. fatua, Bromus bromoideus, Sinapis alba subsp. dissecta…).
2. Assessment of weed communities
Weed communities can be assessed through different methods (direct counts of seeds in the seedbank, seedbank germination assay, multiple sampling strategies of emerged flora in the field) and at different spatial (quadrat, plot, landscape…) and temporal scales (before and/or after weeding, different phenological stages of the crop, during the intercropping period, between each farming practice…) depending on the research objective (Hanzlik and Gerowitt, 2016; Nkoa et al., 2015). In this thesis, we have focused on weed communities which have emerged during crop or cover crop growth at different spatio-temporal scales. Emerged weed communities directly represent the species that were – at least partially – unfiltered by the agronomic practices considered (Keddy, 1992). Indeed, the presence of a species in the seedbank does not necessarily imply that the species was well suited to pass through the filters imposed by management practices (e.g. persistent seeds which indicate a legacy effect). Similarly, the seedbank does not necessarily reflect past weed communities to a similar extent (e.g. dominance of a species with transient seeds in a given year will not be proportionally represented years after). Weed communities can be described in terms of abundance or structure (Adey and Loveland, 2007). Abundance relates to the absolute quantity of biomass produced, of space occupied or of number of plants present whereas structure relates to diversity and relative composition (Adey and Loveland, 2007). In a general sense, diversity can be considered as an instance of being composed of differing elements. Furthermore, weed community structure (diversity and composition) can be assessed from a taxonomic or functional perspective (Louca et al., 2016). Under a taxonomic framework, the focus is on
13 species: How many species are there? What is their relative proportion? Under a trait-based framework, the focus is on functional traits: What extent of trait space does the community occupy? What are the average trait values? Both community abundance and structure ought to be considered in weed science as a current objective is to promote weed diversity while preventing high levels of weed biomass, which could jeopardize crop productivity.
2.1. Weed community abundance
Common measurements of weed abundance include biomass, cover, or density (Nkoa et al., 2015). Weed biomass can be assessed per species by weighing the amount of dry matter (usually aboveground) produced by each weed species on a given surface. Weed cover can be assessed per species by visually estimating the percentage of a given surface covered by each weed species. Weed density can be assessed per species by counting or estimating (Barralis, 1976) the number of individuals of each weed species on a given surface. Of course, total weed community abundance can be assessed as the sum of the measures obtained for each of the species on a given surface.
Weed biomass and cover are particularly relevant measures in weed science. Higher weed biomass at flowering can be interpreted as greater resource uptake when crop demands’ are highest (Malhi et al., 2006) and therefore, as an indicator of weed:crop competition. Indeed, the effect of weeds on crop productivity has often been highlighted by regressing total weed biomass against crop biomass (Milberg and Hallgren, 2004). Biomass of clonal plants is usually considered more biologically meaningful than density as it directly reflects abundance (Aschehoug et al., 2016). However, measurement of weed biomass by species is time consuming and therefore, often replaced by assessment of weed cover by species. This is a simple non-destructive measure, borrowed from phytosociology, which attempts to reflect how space is occupied by species (Mueller-Dombois and Ellenberg, 1974). Total weed cover is usually interpreted in a similar fashion as total weed biomass, even though total weed cover is a two dimensional measure (Guo and Rundel, 1997), i.e. it does not take height into account (otherwise it would be a volume). High weed cover could result in either low or high weed biomass depending on the growth form of the dominant plants (prostrate: low weed biomass; erect: high weed biomass) and phenological stage (early: low weed biomass; late: high weed biomass).
Biomass or cover measured late in the season (when it is most meaningful) do not represent valuable monitoring tools for weed management, which is usually carried out early in the season, before weeds and crops begin to interfere (Knezevic et al., 2002). Indeed, predictive models of yield loss have usually been based on the density of a particular weed species and its relative time of emergence with respect to the crop (Kropff and Spitters, 1991; O'Donovan et al., 1985). Weed density before weeding can provide valuable information on the long-term cumulative effect of farming practices. Assessment of weed communities after weeding usually provides more information on the effect of current weed management practices than on the long-term cumulative effect of farming practices (Doucet et al., 1999).
14 2.2. Weed community structure
Assessment of weed community abundance per species is essential to compute abundance weighted indices describing community structure. Weed community structure is determined by species relative abundance (e.g. %), rather than species absolute abundance (e.g. g/m²). Two weed communities drastically differing in terms of total abundance can present identical structure if the weed species are found in identical proportions.
2.2.1. Taxonomic framework
Taxonomic diversity
Under a taxonomic framework, weed community structure relates to the diversity of species present (i.e. taxonomic diversity) and how their relative proportion (whether weighed by biomass, cover, or density) varies (i.e. taxonomic composition). The simplest diversity index to characterize a plant community is species richness, i.e. the number of species found in the community. Communities harboring more species than others are usually considered as more diverse. However, species richness does not account for the relative proportion of each species within the community. Indeed, a community which shows 10 individuals (or grams, or % cover) of both species A and B could be considered as more diverse than a community which shows 19 individuals of species A and 1 individual of species B. Therefore, authors have developed indices to characterize how total abundance is shared among species, i.e. Simpson index (Simpson, 1949), Pielou’s evenness index (Pielou, 1966) ect. A community is qualified as “even” (low dominance) if total abundance is equally shared among species and “uneven” (high dominance) if one species represents a large proportion of total abundance. Nevertheless, evenness does not account for differences in species richness: a community with 10 individuals of species A and B will be considered as diverse as a community with 10 individuals of species A, B, and C. The Shannon-Wiener diversity index, first developed to measure the amount of information transmitted in communication, combines these two properties, i.e. species richness and evenness (Shannon, 1948). Hence, species richness and evenness allow to identify whether variations in the Shannon-Wiener diversity index are due to variations in species richness or relative abundance between species. Finally, Jost (2006) argued that the Shannon-Wiener diversity index was an index of diversity, not diversity itself. It does not behave as we expect of a diversity, i.e. it does not double between a community of 8 and 16 equally-common species. Therefore, the author proposed to exponentiate the Shannon-Wiener diversity index (2.71828Shannon-Wiener diversity index) so as to reflect the effective number of species, i.e. the number of equally-common species
required to give a particular value of the index (this is also true for other diversity indices but the transformation function can change).
Relative composition
Unimodal multivariate ordination techniques, such as correspondence analysis (CA, or its constrained form: canonical correspondence analysis (CCA)), have proven to be particularly helpful to investigate
15 how relative abundances vary between different communities (Borcard et al., 2018). Such methods allow to quickly detect associations between sites and species. However, one must bear in mind that absolute abundance is not considered. If site A presents 50 g/m² of species X and Y and site B presents only 1 g/m² of species X, then, species X will be associated to site B (relative abundance of species X is 100% in site B whereas it is 50% in site A). If the focus is placed on species abundance, rather than relative abundance, linear methods (i.e. principal component analysis (PCA) and its constrained form: redundancy analysis (RDA)) are more appropriate. However, the choice between linear and unimodal methods will be determined by whether differences between sites are driven by changes in species abundance (PCA/RDA) or species turnover (CA/CCA).
2.2.2. Functional framework
Under a functional framework, community structure relates to the extent of functional trait dissimilarity between species (i.e. functional diversity) and dominant trait values within the community (i.e. functional identity). Functional diversity relates to Tilman’s “diversity hypothesis”, i.e. that trait diversity within a community can affect ecosystem processes (Tilman et al., 1997), while functional identity relates to Grime’s “mass ratio hypothesis”, i.e. that the functioning of ecosystem is determined to a large extent by the trait values of the dominant species (Grime, 1998).
Functional diversity
Functional diversity may be computed on a single trait or across multiple traits (Mason et al., 2005; Villéger et al., 2008). A scientist wishing to investigate weed community structure must dispose of a trait by species matrix. This matrix comprises all the species found across all samples and allocates each species a value for a given trait. This matrix is supported by a site by species presence/absence matrix or, better, a site-by-species abundance matrix. Then, the basic idea behind the computation of functional diversity indices is to represent the functional distance between species. Historically, functional distances were computed based on the length of the branches of hierarchical classification trees (Petchey and Gaston, 2002). A more recent and robust approach is to represent pairwise distances between species in a multivariate space (Laliberté and Legendre, 2010; Villéger et al., 2008). This can be performed by computing an Euclidean (only continuous traits), Gower (missing data and/or presence of factors), or any other distance matrix on the species by trait matrix and then, using it as input in a principal coordinate analysis (i.e. PCoA). (Villéger et al., 2008) proposed three complementary indices of functional diversity, which were then generalized by Laliberté and Legendre (2010):
- Functional richness (Figure 6): This corresponds to the difference between the maximum and minimum (i.e. range) trait values when only one trait is considered (Mason et al., 2005). For multiple traits, this corresponds to the volume of functional space occupied by the community (i.e. convex hull volume (Cornwell et al., 2006)). This index does not consider species relative abundance.
16 - Functional evenness (Figure 6): For a single trait, this index measures the regularity of spacing between species along a functional trait gradient and evenness in the distribution of abundance across species (Mouillot et al., 2005). For multiple traits, the idea is identical except that multivariate space is linearized on a single axis based on a minimum spanning tree. Functional evenness decreases either when abundance is less evenly distributed among species or when functional distances among species are less regular.
- Functional divergence (Figure 7): For a single trait, this reflects how abundance is spread along a functional trait axis, within the range occupied by the community (Mason et al., 2005). Functional divergence is low when abundance is clumped around the center of the functional trait range and high when abundance is clumped around the extreme functional trait values. For multiple traits, functional divergence reflects how abundance is distributed within the convex hull volume.
Figure 6: Borrowed from Mason et al. (2005). Functional richness and functional evenness. The vertical axes represent abundance (e.g. biomass). The bell-shaped curves indicate the distribution of the abundance of individual species in niche space. The histograms indicate the summed abundance of the species occuring in each functional character category (i.e. equal-width sections of the functional character range). The vertical dotted lines indicate the amount of the niche space filled by the species together. Functional richness can decrease without a change in functional evenness if the evennes of abundance within the niche space is unchanged (going from B to A1). Similarly, functional evenness can decrease without a change in functional richness if the amount of niche space filled is unchanged (going from B to C).
17 Figure 7: Borrowed from Mason et al. (2005). Functional divergence. As in Figure 6 the bell-shaped curves show the distribution of species abundance (e.g. biomass) in niche space (e.g. leaf nitrogen content) while the histograms show the summed abundance of the species present in each category. The vertical dotted lines indicate the amount of niche space filled by the species. (A) A community with relatively high functional divergence, with the most abundant species occuring at the extremities of the functional character range. (B) A community with relatively low functional divergence, with the most abundant species occuring towards the centre of the functional character range. Functional divergence can change without a change in either functional richness or functional evenness (going between A and B).
One of the main limits of the framework proposed by Villéger et al. (2008) is that none of the indices proposed allow to assess the extent of multivariate trait space occupied while accounting for species relative abundance (Laliberté and Legendre, 2010). This led Laliberté and Legendre (2010) to propose an index of multivariate dispersion (FDis), which corresponds to “the mean distance in multivariate trait space of individual species to the centroid of all species”. It can account for differences in abundance by “shifting the position of the centroid towards the more abundant species and weighting distances of individual species by their relative abundances” (Laliberté and Legendre, 2010). This index is conceptually similar to Rao’s quadratic entropy, which expresses the mean distance between two randomly selected individuals (Botta-Dukát, 2005; Rao, 1982; Ricotta, 2005).
An essential element when computing functional diversity resides in the choice of traits. The choice of traits must relate to the question of interest. Focusing on functional diversity of single traits allows to investigate specific ecological questions: Do diversified crop sequences promote weed communities with contrasted germination periods? Focusing on functional diversity of multiple traits may allow to investigate more complex questions: Do diversified crop sequences promote weed communities with diversified resource acquisition strategies? Within this context, one may choose to describe functional trait space based on Ellenberg indices for nitrogen, light and moisture (Bourgeois et al., 2019). Nevertheless, one may wish to characterize functional diversity in a general sense and choose traits which reflect plant strategies (e.g. specific leaf area, plant height, and seed mass (Westoby, 1998)).
18 Trait values may either be retrieved from existing databases or measured in the field. Trait measures allow to account for intra-specific variability (de Bello et al., 2013) whereas database values do not. This is of importance because different traits may be more or less plastic (e.g. seed mass tends to be more stable than specific leaf area). Finally, it is important to note that the indices presented here represent the most commonly used, not the wide range of possibilities one may choose from (Petchey and Gaston, 2006).
Functional identity
As specified previously, functional diversity is often coupled with functional identity to provide a complete view of community structure (Li et al., 2017). Functional identity is reflected through community weighted means (CWM), i.e. the average trait value of the species forming the community, weighted by species relative abundance (Lavorel et al., 2008), either biomass, density, or cover. Hence, community weighted means reflect the dominant trait value expressed by the community.
2.3. The matter of scale
Weed abundance and composition can be assessed at different scales. Three common scales include the quadrat scale, the plot scale or the crop sequence scale. The plot scale can be considered as the sum of all quadrats sampled in a given field in a given year, and the crop sequence scale as the sum of all quadrats sampled in a given field across years. Weed:crop competition is governed by processes (e.g. preemption of light or soil resources) which occur at a local scale throughout the season, and therefore best assessed through late quadrat measures (Lindquist et al., 1994; Pollnac et al., 2009), albeit quadrat size is representative of the neighborhood of competitive interactions. On top of local abiotic conditions, observed yield within a quadrat will be determined by the characteristics of the weed community present in this exact same quadrat. However, certain weed species (e.g. Cirsium arvense, Avena fatua) are well known to present a patchy distribution, due to dispersal type, life form, variations in soil type, interactions between the latter, or stochastic events (Andreasen et al., 1991; Wiles and Brodahl, 2004). Therefore, pooling all quadrats of a given plot:year allows to encompass all such species and reflect the diversity of species (or traits) that were not filtered by management practices in that given crop, irrespectively of the factors that might explain their patchy distribution. Nevertheless, one must be careful that soil types show similar variability within each of the statistical units, as higher soil variability could allow the expression of a greater diversity of weed species (Andreasen et al., 1991; Fried et al., 2008). Similarly, pooling all quadrats over the crop sequence allows to encompass all the species which may have been associated to specific crops (due to contrasted germination requirements for example) and reflect the diversity of species (or traits) that were not filtered by management practices at the crop sequence scale.
Comparisons between the quadrat and plot scale can provide information as to how weeds are distributed. Similar weed diversity at the quadrat and plot scale implies that weed communities are fairly
19 homogeneous at the field scale (all species are found in the different quadrats). In contrast, higher diversity at the field scale than at the quadrat scale implies a certain degree of species turnover between quadrats. Similar weed diversity at the plot and crop sequence scale implies that weed communities were fairly uniform across all crops. In contrast, higher diversity at the crop sequence scale than at the plot scale implies a certain degree of species turnover between plot:years.
3. Focus on weed diversity
3.1. State of weed diversity
Agricultural intensification after World War II had a dramatic impact on farmland diversity (Kleijn et al., 2009). The current state of weed diversity is troubling (not to say disastrous) all across Europe (see references for all countries in Albrecht et al. (2016)). (Richner et al., 2015) conducted a meta-analysis on 53 studies and concluded that weed species richness decreased by 20% in average across Europe after the end of World War II (vs. 50% decline in the cumulative review by Albrecht and Bachthaler (1990) between 1930 and 1980). In Germany, 35% of weeds are considered threatened (Korneck and Sukopp, 1988). In Britain, weeds are considered as the most threatened group of plants (Still and Byfield, 2007). In the Netherlands, 88% of weed species associated to calcareous soils are red-listed (Sparrius et al., 2014). In France, Fried et al. (2009) observed a 42% decline in weed species richness and a 67% decline in total weed density at the field level between 1970 and 2000. A French National Action Plan drafted a list of 102 weed species: 52 are considered in a « precarious state », 30 require « close monitoring », and 7 are extinct (Cambecèdes et al., 2012). Of the 52 species considered as in a precarious state, 12 have regressed by more than 70% between 1970 and 1990 (Cambecèdes et al., 2012). Historically present in seven departments of France, Nigella nigellastrum subsists in one last station (Mérindol, FR) thanks to specific measures (CBN de Porquerolles). According to phytosociologists, the minimum area required to encompass the characteristic species of the class Secalinetea (i.e. weed communities on calcareous soils) has been multiplied by a thousand (Jauzein, 2001b). It is important to note that decline in weed diversity has not only concerned segetal plants. Drastic reductions in European weed seed-banks have been imputed to reduced abundance of common species (Jensen and Kjellsson, 1995; Lutman et al., 2009; Robinson, 1997; Robinson and Sutherland, 2002; Squire et al., 2003). Jauzein (1995) inventoried approximately 1200 plant species which could be found in French agricultural fields. However, only 220 are considered as frequent, harmful for crop production or of concern for diverse reasons in France and Europe, respectively (Mamarot and Rodriguez, 2014).
3.2. Benefits associated to weed diversity
The importance of weed diversity in supporting biological diversity in agricultural fields is well documented. First of all, a basic but crucial element is that weeds are primary producers. Therefore, they represent the base of agricultural trophic networks. Weed leaves, flowers and seeds provide resources
20 for a diversity of organisms (e.g. pollinators, earthworms, granivorous and omnivorous arthropods such as carabid beetles, farmland birds, mammals…) which will then be consumed by higher trophic levels (references inMarshall et al. (2003)and Petit et al. (2011)). For example, the Phytophagous Insect Data Base compiles over 70 records of different insect species feeding on Stellaria media (L) Vill., whose seeds represent an important part of farmland bird diet (Marshall et al., 2003). As a matter of fact, specialist insects depend on the presence of one specific weed species to complete their life cycle. Weed patches can also represent important dwelling and reproduction sites for a wide set of invertebrates (Bohan et al., 2007).
Weeds may also contribute to the provision of regulating ecosystem services, such as increasing beneficial insects for pest control and pollination (Alignier et al., 2014; Petit et al., 2011; Sutter et al., 2017). Through an extensive literature search, Blaix et al. (2018) concluded that the main regulating service which could be attributed to weeds was pest control (71% of articles) and that the prevailing mechanism was the provision of suitable habitats for natural enemies. Less frequent services associated to weeds included improved nutrient cycling (18% of articles), soil physical properties (5%) and crop pollination (4%). However, the yield gain due to increased pest control or crop pollination is rarely measured. Studies mainly highlight higher abundance/diversity of natural enemies or pollinators in the presence of higher weed abundance/diversity. Unfortunately, an increase in natural enemies’ diversity/abundance does not necessarily imply a reduction of pests and eventually, an increase in crop yield. Moreover, a wide body of literature has highlighted the negative effects of weeds on crop productivity due to competition for light, water and soil nutrients (Zimdahl, 2007). Hence, further studies which identify trade-offs between increased weed:crop competition and increased pest control/pollination (see Afun et al. (1999)) at different spatial scales are clearly needed (Bretagnolle and Gaba, 2015). More recently, increasing attention has been paid to the effect of weed diversity in mitigating crop yield losses due to weeds (Ferrero et al., 2017; Gonzalez-Andujar et al., 2019; Navas, 2012; Pollnac et al., 2009; Storkey and Neve, 2018). Theories which could support such an interest will be detailed in a further section.
3.3. Factors responsible for the decline in weed diversity
A wide range of factors can be held accountable for the decline in weed diversity. Improved seed cleaning at the beginning of the twentieth century greatly impaired the dispersal of « crop mimics » (e.g.
Lolium temulentum, Agrostemma githago, Bromus secalinus agg.), i.e. species whose seeds’ size, shape
and time of maturity had evolved to resemble that of the crop to ensure being harvested and sown along with the crop (Kornaś, 1988; Storkey et al., 2010). Moreover, these speirochorous species were strictly associated to agricultural fields (segetal plants sensu strictissimo) and were not able to find refuge in secondary habitats (Jauzein, 2001b). Such effects were particularly reinforced with the arrival of commercially certified seeds which guaranteed low levels of impurities. Similarly, the transition from crop-livestock farming to a few productive crops led to the rarefaction of fallows (in its traditional sense,