Research Collection

Journal Article

Identifying indicator species for post-release monitoring of genetically modified, herbicide resistant crops

Author(s): Hilbeck, Angelika; Meier, Matthias; Benzler, Armin

Publication Date: 2008

Permanent Link: https://doi.org/10.3929/ethz-b-000013202

Originally published in: Euphytica 164(3), http://doi.org/10.1007/s10681-008-9666-9

Rights / License: In Copyright - Non-Commercial Use Permitted

This page was generated automatically upon download from the ETH Zurich Research Collection. For more information please consult the Terms of use.

ETH Library Euphytica (2008) 164:903–912 DOI 10.1007/s10681-008-9666-9

Identifying indicator species for post-release monitoring of genetically modiWed, herbicide resistant crops

Angelika Hilbeck · Matthias Meier · Armin Benzler

Received: 30 March 2007 / Accepted: 6 February 2008 / Published online: 28 February 2008 © Springer Science+Business Media B.V. 2008

Abstract In Europe, regulations for release and conjunction with the cultivation of GM HR crops. placing-on-the-market of genetically modiWed (GM) Since the ‘Farm-scale Evaluations’, it is clear that crops require post-release monitoring of their impact consequences for farmland biodiversity can be on the environment. Monitoring potential adverse expected. The objective of this project was to identify eVects of GM crops includes direct eVects as well as relevant indicator species for the long-term impact of indirect eVects, e.g. GM crop speciWc changes in land GM HR maize cultivation and the application of their and pest management. Currently, there is a gap in the corresponding non-selective herbicides, glyphosate pre-release risk assessments conducted for regulatory and glufosinate. In this article, we describe the out- approval of GM herbicide resistant (HR) crops. Since come of a modiWed Event Tree Analysis, essentially a the relevant non-selective herbicides have been regis- funnel-like procedure allowing to reduce the large tered many years ago, in current dossiers requesting number of potentially aVected non-target species to regulatory approval of GM HR crops, the environ- those with greatest ecological relevance and highest mental impacts of the corresponding non-selective risk to be adversely aVected based on a number of herbicides are either entirely omitted or the applicant ecological criteria. This procedure allowed us to iden- simply refers to the eco-toxicological safety assess- tify a total of 21 weed- associations that ments conducted for its original pesticide approval we proposed for post release monitoring of GM HR that do not address environmental issues arising in maize in Germany.

Keywords Indicator species · Non-selective herbicides · Non-target species · Transgenic crops A. Hilbeck (&) Institute of Integrative Biology, ETH Zurich, Universitätstrasse 16, 8092 Zurich and EcoStrat GmbH, Hottingerstrasse 32, 8032 Zurich, Switzerland Introduction e-mail: [email protected] In Europe, regulations for release and placing-on-the- M. Meier Kompetenzzentrum für Sicherheit und Risikoprävention market of genetically modiWed (GM) crops require (KSR), ZHAW Zurich University of Applied Sciences, post-release monitoring of their impact on human Jägerstrasse 2, 8401 Winterthur, Switzerland health and the environment. A monitoring plan under the Directive 2001/18/EC (Annex VII) foresees ‘case- A. Benzler I 1.3 Monitoring, Federal Agency for Nature Conservation, speciWc’ monitoring and ‘general surveillance’. Case- Konstantinstrasse 110, 53179 Bonn, Germany speciWc monitoring aims to refute or conWrm risks 123 904 Euphytica (2008) 164:903–912 identiWed in the required pre-release environmental GM HR crops. Since the relevant non-selective herbi- risk assessment. General surveillance aims to detect cides have been registered many years ago, in current unanticipated adverse eVects and long-term cumula- dossiers requesting regulatory approval of GM HR tive eVects that could not be detected in pre-release crops, the environmental impacts of the correspond- testing and escaped the pre-release risk assessment. ing non-selective herbicides are either entirely omit- Monitoring possible adverse eVects of GM crops ted or the applicant simply refers to the eco- includes direct eVects as well as indirect eVects, e.g. toxicological safety assessments conducted for its GM crop speciWc changes in land management (see original pesticide approval that, for one, can be a long also Council Decision 2002/811/EC, which supple- time ago but, more importantly, do not address and ments Annex VII to Directive 2001/18/EC by detailed investigate environmental issues arising in conjunc- guidance notes). However, while both are meant to be tion with the cultivation of GM HR crops. Before the complementary, in practice, they are diYcult to sepa- advent of GM HR crops, non-selective herbicides rate and the discussion is controversial as reporting were not routinely used within arable Welds during the responsibilities among diVerent authorities and Wnan- cultivation period of the crops as they would kill the cial consequences are tied to it. crops as well. The possible impact of the cultivation In contrast to most existing environmental moni- of GM HR crops through the application of the corre- toring programs that were typically invoked by docu- sponding non-selective herbicides on farmland biodi- mented damage (e.g., loss of certain species) as a re- versity has long been recognized and, in fact, sparked active instrument, the monitoring of GM crops in the largest Weld trials ever conducted with herbicide Europe is largely a pro-active, precautionary measure. resistant GM crops, the ‘Farm-scale Evaluations’ Since there is no large scale GM crop production in (FSE). The FSE largely conWrmed previous predic- Europe yet, little experience exists to date with tions that at least for oilseed rape and sugar beet an regional monitoring and long-term ecosystem impacts additional loss of farmland biodiversity can be of GM crops. Reports on environmental eVects from expected beyond and above current conventional elsewhere are contradicting and certainly controver- practices. Fields were ‘cleaner’ due to more eVective sial (Brookes and Barfoot 2006; Garcia and Altieri weed control, hence, less weeds left less food for their 2005; Friends of the Earth 2007). In fact, when look- associated wildlife. For maize there did not seem to ing closer into this issue, one Wnds that reliable and be an additional loss and some species even occurred independent data on long-term and larger scale envi- at higher densities than in conventionally treated Weld, ronmental impacts of the cultivation of GM crops are at least as long as the chosen herbicide was atrazine scarce to non-existent globally. Over 95% of all GM (Hawes et al. 2003). That herbicide, however, is crops are grown on a signiWcant scale only in six banned in Europe today and with other herbicides it countries in the world. Of these six countries, 53.5% was challenged whether the same Wndings would hold are grown in the US, 17.6% in Argentina, 11.3% Bra- (Burke 2005; Perry et al. 2004). zil, 6.0% Canada, 3.7% India and 3.4% in China While recognizing, for one, the lack of data and the (James 2006). In Argentina, almost all soybeans pro- current controversial nature of the discussion on how duced today are GM HR soybean involving the appli- to close this signiWcant gap in pre-release risk assess- cation of enormous amounts of glyphosate. None of ments for GM HR crops and, secondly, recognizing the above listed countries, however, has regional the fact that the few data and reports existing to date monitoring programs in place that systematically sur- give reason to believe that long-term eVects on farm- vey and collect data on the impact of these non-selec- land biodiversity and ecological functions must be tive herbicides for instance on farmland biodiversity expected with increasing cultivation of GM HR crops, and ecosystem functioning, nor the development of the German Agency for Nature Conservation as part resistant weeds (Heap 2007). of the national regulatory body had to take action in Hence, only few data exist to date that stem from order to fulWl its legal mandate and responsibility to coordinated scientiWc research and monitoring on the protect the biodiversity in Germany. The Agency long-term environmental impact of GM HR crops. launched a research and development project and This is largely due to a gap in the current pre-release tasked us to identify indicator species that would risk assessments conducted for regulatory approval of allow to systematically monitor the long-term, 123 Euphytica (2008) 164:903–912 905 regional impact of the cultivation of GM HR crops Fault trees are ‘top-down’ risk analysis tools where arising through the application of their corresponding the analyst speciWes a failure event (i.e. ‘top-event’) non-selective herbicides in the agro-ecosystem. The and then by combining logical functions such as ‘and’ focus was on indicator species. This choice was and ‘or’, identiWes all events that can or must contrib- also based on the FSE Wndings that conWrmed that ute to the speciWed failure (Hayes et al. 2004). An larger such as birds and small mammals Event Tree is the complementary ‘bottom-up’ while often being a target for conservation eVorts, are approach where an analyst speciWes an ‘initiating quite diYcult to monitor due to their habitat require- event’ and lays out the logical chain of events that can ments largely exceeded crop Weld sizes. But they all occur and lead to a number of possible consequences. are reliant either on certain farmland plants or arthro- Both tools yield more or less complex tree-like charts pods associated with these farmland plants for food. where each event chain forms one branch of the tree. Following the argument of the FSE, if indicator spe- They do graphically model all of the parallel and cies at that level can be identiWed, it is expected that sequential combinations of events that can lead to a this can be extrapolated to higher trophic level organ- particular ‘top event’ or arise from a particular ‘initi- isms and indicate potential eVects at the top end ating event’. This structured, logical approach allows (Burke 2003). to rigorously evaluate the potential of these events to occur. It is based on scientiWc data and expert knowl- edge and identiWes what data and information is nec- Methodology essary to determine reliably the outcome and the gaps of knowledge associated with the possible events in a As risk identiWcation and assessment of the repeated transparent manner. Both tools provide a fairly good and large scale application of non-selective herbicides understanding of the reliability of the analysis and the during the crop’s cultivation period are currently involved uncertainties and identify research priorities missing in the risk assessment parts of the dossiers for closing the most critical data gaps. submitted for regulatory approval which could serve In this paper, we report about the use and out- as starting point for the development of a monitoring come of a modiWed Event Tree Analysis that program, we had to Wrstly identify the potential risks allowed us to model the potential risks arising from and their impact pathways. This subsequently allowed the application of non-selective herbicides and iden- us to determine suitable candidate species for moni- tify plant and insect species that most likely will be toring that would indicate whether or not a particular aVected and are suitable to indicate within a moni- potential risk actually occurs. Starting point are toring program whether these potential risks occur cause-and-eVect chains of the potential environmental in the Weld. A GM HR maize variety expressing impacts of GM HR crops triggered by the application resistance to either one of the two non-selective her- of their corresponding non-selective herbicides. To do bicides ‘Glyphosate’ and ‘Glufosinate’ served as the that, we adopted and modiWed the well-known risk model GM crop. analysis tools called ‘Event Tree Analysis’ and ‘Fault Tree Analysis’. Fault and Event Tree Analyses are complementary tools used in risk assessment that Results were originally developed by engineers identifying critical steps in complex engineering processes, e.g. We Wrstly identiWed the potential risks and described aviation or large scale industrial production facilities. the pathways how they could realize. This was fol- For environmental purposes and ecological systems, lowed by the actual analysis. The Event Tree is before us, also Hayes (1998) used these tools for depicted in Fig. 1 and the results are summarized in marine ecosystems but also to identify hazards of GM Fig. 2. In essence, it serves as funnel-like procedure HR oilseed rape in Australia (Hayes et al. 2004). Both that reduces the number of possible non-target species were also proposed as valuable tools for risk assess- systematically and in a transparent fashion to a ment of GM crops by the US National Research number that is in practical terms feasible and that are Council (National Research Council 2002). ecologically meaningful.

123 906 Euphytica (2008) 164:903–912

Initiating event Causalchain of events Endpoint risk Step 1Step 2 Step 3Step 4 Application of non- Weed x is present at Weed x damaged by Habitatofweed xis Associated non- Associated non-target selectiveherbicide AND time t and location s AND non-selective AND arable field /field AND target lepidoptera AND lepidoptera species k attimet and herbicide margin species k feeds loses host plant location s monophagously on weed x weed x is not OR OR Reduction in present attimet and OR OR Weed x not Associated non- population size location s areexposed damaged by non- Weed xoccurs target lepidoptera Loss of species to non-selective selective herbicide (also) outsideof species k feeds herbicides attimet arable field /field oligophagously on and location s margins weed x OR Associated non- Insects are exposed target lepidoptera to Bt-toxins in species k feeds pollen deposited on polyphagously on weed x attimetand weed x location s.

Fig. 1 Event Tree for analysing the potential risks arising from the application of non-selective herbicides when growing genetically modiWed, herbicide resistant maize on farmland biodiversity

in arable Welds, non-selective herbicides were only List of weed species known to occur in 257 maize in Germany applied on bare soils prior to seedling emergence. Thus, also damage from spray drift to neighboring Weed species that are closely nontarget vegetation will likely be higher for GM HR associated with the arable field and are 55 sensitive to glyphosate and/or crop production. Therefore, as the primary location of glufosinate impact, the arable Weld and the habitats in the near surrounding were identiWed. Lepidoptera species feeding mono- or 21 oligophagous on these weeds The primary concerns with GM HR crops and the application of their corresponding non-selective herbi- Number of weed species serving the cides are the reduction in density and diversity of weed lepidoptera listed above as host plants 11 species, in excess to those induced by the regular herbi-

Proposed indicator weed-insect associations for post-release monitoring cide regimes in conventional crops, and a shift in weed of GM HR maize in Germany species composition towards more robust and less sen- Fig. 2 Summary of the outcome of an Event-Tree Analysis for sitive weed species that carry a high risk of developing identiWcation of weed-Lepidoptera associations as indicators for resistance. This would constitute a further undesired W a monitoring program of genetically modi ed, herbicide resis- W tant maize intensi cation in agricultural habitats. Weed shifts and the development of resistant weeds has been described in GM HR crops for North America (van Gessel 2001). Potential risk identiWcation and impact pathways In fact, since the introduction of GM HR crops in the mid nineties, reports of glyphosate-resistant weeds The ‘Initiating Event’ is the application of a non- stem almost exclusively from countries that grow gly- selective herbicide such as Glyphosate or Glufosinate. phosate resistant GM crops (Heap 2007). From no The application of such a non-selective herbicide reports of glyphosate resistant weeds in 1995, this intends to reduce or eliminate any plant other than the number has now risen to 12, including some quite crop in the Weld. Hence, most if not all plant species problematic weed species such as horseweed (Conyza of a maize Weld and its Weld margins will be aVected canadiensis) in soybean production in the US. This pre- certainly to some degree. Spray drift of herbicides dicted and troublesome development has very recently also varies depending on the height of the spray motivated Monsanto company, the largest marketing beams. Since with GM HR crops, post-emergence company of glyphosate and its resistant GM crops application during the growing period is possible, the worldwide, to respond by establishing a web-based, heights of spray beams can be expected to be higher questionaire-driven ‘weed resistance risk assessment than in conventional Welds. If at all used conventionally program’ (www.weedtool.com) for growers who use 123 Euphytica (2008) 164:903–912 907 glyphosate resistant GM crops. Along with the reduc- species list good or medium control of Glufosinate tion in density and diversity of weed species a decline was reported. For glyphosate, we found information of the populations of nontarget species asso- on 119 tested weed species. Glyphosate was reported ciated with these weeds was predicted and in part con- to have low to some to good eYcacy against 113 Wrmed by the FSE (Hawes et al. 2003). Both, the weeds weed species and no eYcacy against 6 species. One of and their associated fauna are an important—if not the the six unaVected species was Trifolium repens which only—food source for many farmland birds and small can be controlled by Glufosinate. Against 70 of the mammals. Hence, their abundance is likely to be sig- 113 weed species that are well controlled by Glyphos- niWcantly positively correlated to the density and diver- ate also Glufosinate works well. For the remaining 43 sity of their primary food sources, weeds and insects. species no information on susceptibility to Glufosi- But they are much more diYcult to monitor due to their nate was found. large habitat requirements, high mobility and long gen- eration times. Therefore, our focus was on Wnding indi- Association with biotope types (=likelihood of cator species at lower taxonomic orders that can be exposure/experiencing the identiWed adverse eVect) monitored with a reasonable amount of time and fund- (Step 3) ing. Similarly, also a longer term weed shift can be indicated by the presence or absence of their associated In principle, all 257 weed species listed for maize insect fauna. The most likely aVected organisms will be Welds in Germany can be exposed to either one or those weed and insect species occurring in the arable both of the two non-selective herbicides used for most Weld and their surrounding habitats. of todays GM HR crops. However, the overall impact on their populations on a regional scale will diVer IdentiWcation of indicator species depending on their degree of association with the ara- ble Weld and surrounding habitats. The closer associ- Following from the above identiWed primary areas of ated a weed species is with the arable Weld (i.e. large expected impact, in the Wrst step of the Event Tree proportion of a particular weed species population analysis (Fig. 1), we identiWed to the best of our knowl- occurs only within the boundaries of the arable Weld), edge and data available the weed species that are the closer correlated will the abundance and density known to occur within maize crop Welds during the of the whole population of that weed species be with time of cultivation and application of herbicides (Step the frequency and spatial scale of the application of 1, see Fig. 1). These were compiled into a comprehen- the particular non-selective herbicide. In order to Wlter sive list containing 257 weed species from 40 plant out those weed species that are most at risk because a families (full list in Meier and Hilbeck 2005). In order signiWcant proportion of their population or all of it to select those weed species whose populations would occur only in the cropped Weld or in adjacent non-cul- be at greatest risk of being locally eliminated and, thus, tivated habitats, all weed species were ranked accord- whose regional abundance and population size would ing to their association with the two critical habitat be directly linked to the spatio-temporal scale of GM types ‘agroecosystem’ and the ‘arable Weld’ (see HR maize cultivation, the listed species were subjected Box 1 and Table 1). For information about the biolog- to a ranking procedure. In Steps 2 and 3, the weed spe- ical attributes of the weed species and the association cies were ranked according to their known sensitivity with certain habitat types see Haeupler and Muer towards the non-selective herbicides (=adverse eVect) 2000. and the strength of their association with certain bio- Further, all weed species were grouped in one of tope types (=likelihood of exposure/experiencing the three categories combining the two classes ‘agroeco- identiWed adverse eVect) (Fig. 1). system’ and ‘agricultural Weld’ to get a rough estimate of their risk to be adversely aVected through the appli- EYcacy of non-selective herbicides (adverse eVect) cation of the non-selective herbicides (Table 1). (Step 2) Of the total weed species list comprising 257 spe- cies, 81 weed species were found to be only weakly Based on producers information for Glufosinate, for associated with the agroecosystem and the agricul- 88 tested weed species of the above described weed tural Weld and therefore be at lower risk to experience 123 908 Euphytica (2008) 164:903–912

Box 1: Association with ‘agroecosystems’ and ‘arable fields’

A) Association with ‘agroecosystem’: this includes various habitat types belonging to the agricultural environment: different types of arable fields, pastures, grassland, meadows, unmanaged strips or lanes along small ditches, shrub thickets, but also orchards (extensively or intensively managed), abandoned areas, fallows, or unpaved field ways.

Ranking:

Close association (rank 1) – Weed species occurs only in habitat types within agroecosystems

Medium association (rank 2) – Weed species occurs at least in 1 habitat type outside of agroecosystems but in less habitat types outside than inside agroecosystems

Weak association (rank 3) – Weed species occurs in more habitat types outside than inside the agroecosystem

B) Association with ‘arable field’: this includes the cropped area and adjacent structures , e.g. field margins (incl. managed hedges) or unpaved field ways.

Ranking:

Close association (rank 1) – Weed species occurs only in cropped area (arable field) and potentially in unpaved field roads (similar disturbed biotope type)

Medium association (rank 2) – Weed species occurs in cropped area but also other structures listed above or exclusively in unpaved field roads

Weak association (rank 3) – Weed species hardly occurs in the cropped area but mainly in other habitat types along cropped areas (e.g. field margins, hedges, forest margins,)

Table 1 Combining and grouping species according to their to the scale of the application of non-selective herbi- association with the agroecosystem and the arable Welds cides during the growing period. Agricultural Agroecosystem In a synthesis step of the procedure, we determined Weld the cross section (logically connecting two causal events: presence at location s and sensitive to the her- Rank 1 2 3 bicide applied at time t) of both the association with 1 High risk High risk Medium risk habitat type and sensitivity to either one or both non- 2 High risk Medium risk Low risk selective herbicides most commonly used in conjunc- 3 Medium risk Low risk Low risk tion with GM HR crops, glyphosate and glufosinate (Table 2). This allowed us to create risk categories a signiWcant decline in their populations. This is and list the relevant species that must be considered because many populations also occur and survive out- for the next steps in the analysis. side of the most aVected areas. For another 118 weed This procedure allowed us to identify a total of species, a medium association (=medium risk) was 55 ‘high risk’ weed species (see Table 3) whose concluded and only 55 weed species were found to be population’s fate will be closely correlated with the closely associated with both the arable Weld and the spatio-temporal scale of application of non-selective agroecosystem. Therefore, their population density herbicides in maize Welds in Germany and, thus, and abundance is driven by the agricultural measures can indicate the intensity and eVectiveness of a applied in the arable Weld (Fig. 2). These include the region-wide cultivation of GM HR crops employing weed species whose populations will be at high risk either Glyphosate or Glufosinate-based non-selective of experiencing a regional decline directly correlated herbicides.

Table 2 Risk categories Herbicide Association group when combining association with habitat and sensitivity Close (55 species) Medium (118 species) Weak (81 species) to herbicides glyphosate and glufosinate Sensitive 25 glyphosate; 50 glyphosate; 36 glyphosate; 21 glufosinate; 48 glufosinate; 19 glufosinate; 17 both (high risk) 37 both (medium risk) 16 both (low risk) Not sensitive 0 1 glyphosate 5 glyphosate Unknown 26 57 42

123 Euphytica (2008) 164:903–912 909

Table 3 Remaining high risk weed species (high sensitivity to Table 3 continued non-selective herbicides, occur predominantly within arable Welds or in their close surroundings, likely aVected by spray Latin name Family drift) Cuscuta epilinum Weihe Convolvulaceae Latin name Family Diplotaxis muralis (L.) DC. Brassicaceae crispus L. Polygonaceae Euphorbia exigua L. Euphorbiaceae Viola arvensis Murray Violaceae Geranium rotundifolium L. Geraniaceae Fallopia convolvulus (L.) A. Löwe Polygonaceae Iberis amara L. Brassicaceae syn. convolvulus L. Legousia hybrida (L.) Delarbe Campanulaceae L. ssp. arvensis Legousia speculum-veneris (L.) Chaix Campanulaceae Vicia angustifolia L. ssp. segetalis Fabaceae Linaria arvensis (L.) Desf. Scrophulariaceae (Thuill.) Corb. Lithospermum arvense L. ssp. arvense syn Boraginaceae Vicia tetrasperma (L.) Schreb. Fabaceae Buglossoides arvensis (L.) I.M.Johnston Vicia villosa Roth ssp. villosa Fabaceae Myosotis discolor Pers. Boraginaceae Myosotis arvensis (L.) Hill Boraginaceae Neslia paniculata (L.) Desv. Brassicaceae L. Polygonaceae ssp. paniculata Alopecurus myosuroides Huds. Poaceae Ornithopus perpusillus L. Fabaceae Anthemis arvensis L. Asteraceae Pastinaca sativa L. ssp. sativa var Apiaceae pratensis & ssp. urens Apera spica-venti (L.) P. Beauv. Poaceae (Req ex Godr.) Celak. syn Agrostis spica-venti (L.) P.B. Rapistrum rugosum (L.) All. Brassicaceae Atriplex patula L. Chenopodiaceae ssp. orientale Avena fatua L. Poaceae Rorippa palustris (L.) Besser Brassicaceae Digitaria sanguinalis (L.) Scop. Poaceae syn Rorippa islandica (Oeder) Borbás Senecio vulgaris L. Asteraceae Scleranthus annuus L. Caryophyllaceae Setaria pumila (Poir.) Roem. & Poaceae Silene dichotoma Ehrh. Caryophyllaceae Schult. syn Setaria glauca Silene noctiXora L. Caryophyllaceae Anthemis austriaca Jacq. Asteraceae Spergula arvensis L. ssp. arvensis Caryophyllaceae Tripleurospermum perforatum (Merat) Asteraceae W Lainz syn Matricaria perforata, Identi cation of associated nontarget Lepidoptera Matricaria inodora species potentially at risk (Step 4) Vicia sativa L. Fabaceae Mentha arvensis L. Lamiaceae For the remaining medium (see list in Meier and Hil- tuberosus L. Fabaceae beck 2005) and high risk (Table 3) weed species, the Ranunculus arvensis L. Ranunculaceae associated Lepidoptera fauna reported to utilize them W W Arabidopsis thaliana (L.) Heynh. Brassicaceae as host plants was identi ed and classi ed according Chrysanthemum segetum L. Asteraceae to their feeding preference (Table 4). Monophagous Fumaria parviXora Lam. Fumariaceae associated species are considered to be at high risk for V Fumaria schleicheri Soy.-Will. Fumariaceae adverse e ects if they rely exclusively on a high or medium risk host plant. Similarly, also an oligopha- Papaver rhoeas L. Papaveraceae gous associated species is considered at high risk if it Adonis aestivalis L. Ranunculaceae feeds at a high risk host plant since its choice for Wnd- Adonis Xammea Jacq. Ranunculaceae ing the other one or two alternative host plants are Althaea hirsuta L. Malvaceae rather slim in particular as these also must be Anthriscus sylvestris (L.) HoVm. Apiaceae expected to be at least locally eliminated. On the other Brassica rapa L. Brassicaceae hand, a polyphagous species feeding on a high or Calendula arvensis L. Asteraceae medium risk host plant can be expected to experience Consolida regalis Gray Ranunculaceae itself only a medium to low risk, respectively, as it Crepis sectosa Asteraceae has alternatives around. Likewise an oligophagous

123 910 Euphytica (2008) 164:903–912

Table 4 Weed-Lepidoptera associations proposed for region-wide monitoring programs of GM HR maize (combining high and medium risk weeds with feeding preference of associated fauna (here: Lepidoptera)) Risk category Feeding preference weeds Monophagous Oligophagous Polyphagous

High risk (3 species on 3 weeds) High risk (14 species on 9 weeds) Medium risk Lepidoptera Weed sp. Lepidoptera Weed sp. A total of 3 species on High Issoria lathonia Viola arvensis Argynnis adippe Viola arvensis 3 weeds (see Meier and Hilbeck (2005) Lycaena hippothoe Rumex crispus chamomillae Anthemis arvensis for details) Tripleurospermum perforatum purpuraria Polygonum aviculare Cucullia lucifuga Sonchus arvensis ssp. arvensis Cucullia umbratica Sonchus arvensis ssp. arvensis scabriuscula Polygonum convolvulus Polygonum aviculare Rumex crispus Eupithecia virgaureata Senecio vulgaris Hecatera bicolorata Sonchus arvensis ssp. arvensis Lycaena dispar Rumex crispus Lycaena phlaeas Rumex crispus Pyropteron chrysidiformis Rumex crispus Tyria jacobaeae Senecio vulgaris Leptidea sinapis Lathyrus tuberosus viciae Lathyrus tuberosus Zygaena lonicerae Lathyrus tuberosus Medium High risk (4 species on 2 weeds) Medium Low risk Lepidoptera Weed A total of 45 species on 8 weeds A total of 3 species funesta (see Meier and Hilbeck (2005) on 13 weeds (see for details) Meier and Hilbeck trabealis Convolvulus arvensis (2005) for details) Tyta luctuosa Convolvulus arvensis Lycaena hippotoe Rumex acetosa associated species feeding at a medium risk host plant today under conventional weed control regimes, must is considered to experience a medium risk for its be expected. However, as little to no data from long- regional abundance. term surveillance programs is available on the region- This Wnal step yielded a total of 21 Lepidoptera wide impacts of the large scale and repeated applica- species whose population dynamics will be strongly tion of non-selective herbicides and, as currently inXuenced by the spatio-temporal scale of cultivation practised pre-approval risk assessment omits these of GM HR crops and the application of their corre- adverse eVects (see Introduction), we propose to mon- sponding herbicides. The identiWed 21 high risk lepi- itor selected weed-Lepidoptera associations rather doptera species are all species that are highly than their individual components (i.e. either weed dependent on 11 weed species. For all of these weed species or associated lepidopteran species). Certainly species a negative correlation of their abundances and until suYcient evidence and experience has been densities locally and regionally with the spatio-tem- gained that would allow to possibly monitor only the poral scale of cultivation of GM HR crops, beyond host plant (i.e. weed) or the corresponding associated and above of what these weed species experience species by itself. 123 Euphytica (2008) 164:903–912 911

Summarizing conclusions example, Issoria lathonia is reported to feed mono- phagously on Viola arvensis. While V. arvensis does It is widely known that agricultural habitats presently occur in maize Welds, it has its peak density and is contribute signiWcantly to biodiversity in terms of most problematic as a weed in oilseed rape where species richness at the European level. At the same most herbicides used today have a gap of eYcacy. time farming practices have intensiWed rapidly, lead- Hence, today, they occur in conventional oilseed rape ing to a dramatic loss of biodiversity in agricultural in high and very high densities. Both non-selective areas (Hoogeveen et al. 2002). Today, many charac- herbicides, glyphosate and glufosinate, control this teristic species of agricultural systems feature on weed well. Therefore, in the case of GM HR oilseed Europe’s red lists of endangered species. At the EU rape cultivation, I. lathonia will almost inevitably level, agri-environment policies have been intro- experience a serious if not locally complete loss of duced as a promising policy tool to reverse the nega- host plants which can likely aVect its regional popula- tive biodiversity trend. Preserving low intensity tion densities and abundances. This particular species farming and preventing further intensiWcation should will be under great pressure when both HR maize and have top priority (Hoogeveen et al. 2002). Therefore, HR oilseed rape will be grown. An additional threat in their report commissioned by the Council of will also constitute the Bt-containing pollen from the Europe, the French Government and the United stacked GM HR/Bt maize varities that are being Nations Environment Program, Hoogeveen et al. increasingly used (James 2006). Therefore, the Viola (2002) concluded that special conservation eVorts for arvensis-Issoria lathonia association was scrutinized preserving agricultural biodiversity are justiWed. This further using the complementary Fault Tree Analysis and relevant European environmental legislation where the loss of both species are the identiWed ‘top forces the competent authority on nature conserva- event’ (Meier and Hilbeck 2005). tion in Germany to act on the issue of GM HR crops All of the above identiWed 21 weed-Lepidoptera in German agriculture. In this article, we described associations are in principle suited for post-release mon- the outcome of a modiWed Event Tree Analysis itoring programs. However, it might not be necessary to essentially a funnel-like procedure allowing to monitor all of them. Further reduction eVorts could for reduce the large number of potentially aVected non- example focus on locally most common weed-Lepidop- target species to those with greatest ecological rele- tera associations. Monitoring monophagous species has vance and highest risk to be adversely aVected based the advantage that these species will represent quite sen- on a number of ecological criteria (Fig. 2). A total of sitive indicator species as they are dependent on one 21 weed-Lepidoptera associations were identiWed host plant and will reXect their density and abundance and proposed for monitoring. without much delay. On the other hand, they might However, the outcome of the selection procedure over-proportionally reXect heterogenous distribution could still be developed further by including other patterns of their host plants that are independent of the ecological aspects that might modulate the Wnal out- herbicide use intensity. Oligophagous species could come. Some associated species considered to be at buVer this better but might be less sensitive to intensity lower risk might experience a higher degree risk of of the usage of non-selective herbicides. Therefore, we local extinction than thought depending on their propose to include a variety of locally existing monoph- mobility and location of hatch. For instance, even if agous and oligophagous weed-Lepidoptera associations highly polyphagous, caterpillars in a large maize Weld in a regional monitoring program. sprayed with glyphosate will have to walk long dis- Risk analysis driven identiWcation of indicator spe- tances before Wnding suitable alternative host plants cies has proven useful in particular in face of lacking outside of the cropped Weld again, possibly too long, pre-release risk assessments and post-release Weld data. depending where they hatch and have to start their It allows to select indicator species that are ecologically journey. However, their overall population might be meaningful and sensitive to the anticipated impacts in a better position to compensate for such losses, if caused by the introduction of GM HR crops in a trans- these losses remain localized. parent and scientiWcally logical manner. While we con- Another issue is if certain weed-Lepidoptera ducted this procedure using GM HR maize as the case associations occur also in other GM HR crops. For example, this could be done similarly with any GM 123 912 Euphytica (2008) 164:903–912 crop. This project was a signiWcant step forward Garcia MA, Altieri MA (2005) Transgenic crops: implications towards the identiWcation of a reduced list of sensitive for biodiversity and sustainable agriculture. Bull Sci Tech- nol Soc 25:335–353 indicator species for biodiversity impact of GM HR Haeupler H, Muer T (2000) Bildatlas der Farn- und BlütenpXan- crop plant production but more eVorts need to go into zen Deutschlands. Verlag Eugen Ulmer, Stuttgart their Wne-tuning to regional conditions. This could Hawes C, Haughton AW, Osborne JL et al (2003) Responses of potentially allow to further reduce the investments nec- plants and invertebrate trophic groups to contrasting herbi- cide regimes in the Farm Scale Evaluations of genetically essary for the monitoring but yet maintain its indicative modiWed herbicide-tolerant crops. Philos Trans R Soc power. Further, sampling strategies and test runs must Lond B Biol Sci 358:1899–1913 now be carried out to have the necessary methodolo- Hayes KR (1998) Ecological risk assessment for ballast water gies ready when large scale GM HR crop production introductions: a suggested approach. ICES J Mar Sci 55:201–212 begins in Germany. Hayes KR, Gregg PC, Gupta VVSR, Jessop R, Losndale M, Sin- del B, Stanley J, Williams C (2004) Identifying hazards in Acknowledgements This work was supported by the German complex ecological systems. Part 3: Hierarchical Holo- Federal Agency for Nature Conservation with funds of the Ger- graphic Model for herbicide tolerant oilseed rape. Environ man Ministry for Environment, Nature Conservation and Nucle- Biosafety Res 3:1–20 ar Safety. Heap I (2007) International survey of herbicide resistant weeds. www.weedscience.org/in.asp Hoogeveen YR, Petersen J-E, Gabrielsen P (2002) Agriculture References and biodiversity in Europe. Council of Europe, Strasbourg James C (2006) Global status of commercialized biotech/GM crops: 2006. ISAAA Brief 35 W Brookes G, Barfoot P (2006) GM crops: the rst 10 years—Glo- Meier M, Hilbeck A (2005) Faunistische Indikatoren für das bal socio-economic and environmental impacts. ISAAA Monitoring der Umweltwirkungen gentechnisch veränder- Brief No. 36 www.pgeconomic.co.uk./pdf/global_impact- ter Organismen (GVO). Naturschutz und Biologische Vi- study_2006_v1_finalPGEconomics.pdf (last checked: elfalt 29:137 February 2008) National Research Council (2002) Risk assessment models: V Burke M (2005) Managing GM crops with herbicides: E ects environmental eVects of transgenic plants—the scope and on farmland wildlife. Farm Scale Evaluation Research adequacy of regulation. National Academy Press, Wash- Consortium www.defra.gov.uk/environment/gm/fse/results/ ington DC fse-summary-05.pdf Perry JN, Firbank LG, Champion GT, Clark SJ, Heard MS, May V Burke M (2003) GM crops: e ects on farmland wildlife. MJ, Hawes C, Squire GR, Rothery P, Woiwod IP, Pidgeon Farmscale Evaluation Research Team and the Steering JD (2004) Ban on triazine herbicides likely to reduce but Committee. www.defra.gov.uk/environment/gm/fse/results/ not negate relative beneWts of GM HT maize cropping. Nat fse-summary-03.pdf Biotechnol 428:313–316 W Friends of the Earth International Report (2007) Who bene ts Van Gessel MJ (2001) Glyphosate-resistant horseweed from from GM crops? An analysis of the global performance of Delaware. Weed Sci 49:703–705 GM crops (1996–2006). www.foe.co.uk/resource/reports/ who_benefits_from_gm_crops.pdf

123