Design

The study used a sequential mixed-methods design with three stages: a survey, rapid evidence synthesis, and stakeholder focus groups (Fig. 1). Ethical approval was granted by Lancaster University Faculty of Science and Technology Research Ethics Committee (FST19070), and all methods were performed in accordance with the relevant guidelines and regulations. Informed consent was obtained from all participants prior to data collection.

Participants for the survey and focus groups were purposively sampled from three pre-defined stakeholder groups:

• Farm-level

  1. o

     Farmers, farmer associations or cooperative representatives

  2. o

     Farm- or agri-environment-focused non-governmental organisation representatives

  3. o

     Farm- or agriculture-focused policy makers

• Agri-businesses

  1. o

     Plant breeding representatives

  2. o

     Food producer/processor association representatives

  3. o

     Other agri-business stakeholders (survey only)

• Consumer-level

  1. o

     Consumer group representatives

  2. o

     Consumer experts

  3. o

     Consumers (survey only)

To present stakeholders with the discussion topics, we developed crop improvement “option cards”, which displayed 15 potential crop improvement options organised into three categories: yield, nutrition and sustainability. These were used in both the survey and focus groups to appraise different crop improvement strategies in quantitative and qualitative terms. One side offered a simple explanation of the improvement (e.g. improving plant water use) and the other an example of that improvement being applied through research (see [Supplementary 5]). A blank card labelled “Option Card #16” was provided to allow for opinions missing from other cards.

A brief description of the survey and focus group methods follows, however detailed reporting on both can be found in Supplementary 1.

Priorities survey

Participants

Participants of the survey were volunteer stakeholders identified through the professional networks of the CropBooster-P consortium and through snowball sampling. The Cropbooster-P consortium consisted of a variety of stakeholders based in the EU and the UK including academics and representatives of farmer, plant breeder and seed producer groups. Participants self-identified as belonging to each one of the predefined groups. A total of 324 participants took part in the online survey (288 in English, 22 in French, and 14 in German). For more information, see Supplementary 1 and 3.

Survey instrument

The survey was designed to identify which of the CropBooster-P crop improvement options were prioritised among a wide group of European food system stakeholders. The survey consisted primarily of closed questions, with some open-ended qualitative questions included to elicit more complex responses to key questions. It was programmed and administered using Qualtrics (www.qualtrics.com). Reporting follows CHERRIES guidelines²³ (See Supplementary 6 for a copy of the survey in English). To access as many participants as feasible, it was translated and piloted in German and French using a modified TRAPD method²⁴.

Preferences for crop improvement goals (e.g. sustainability) were elicited on a 1–3 scale, with 1 being the most preferred, using a forced ranking. Preferences for crop improvement options (e.g. increasing plant water use) was assessed on a single item Likert scale labelled 1:‘Very important’; 2: ‘Important’; 3: ‘Neither important nor unimportant’; 4: ‘Unimportant’; 5: ‘Very unimportant’ 6:‘Don’t know’. Rating was selected over forced ranking as this allows participants to indicate ties, and to rate as many options high or low as they prefer.

Analysis

For each stakeholder group, the total number of valid responses was used to analyse: (1) goal prioritisation and (2) the option prioritisation questions.

The percentage of each stakeholder group ranking a given goal (yield, nutrition, or sustainability) as one (top priority), two (medium priority) and three (lowest priority) was recorded, and the most frequently selected priority goal highlighted. Data from the free text questions was exported to Nvivo 12 and thematically analysed to identify key issues. The 15 Likert-style items relating to the 15 option cards were treated as individual responses. Each of the choices was tallied and the percentage of participants choosing each statement calculated. Differences were reviewed for: top goal priority, and between stakeholder groups.

Impacts: rapid evidence synthesis

A rapid evidence synthesis (RES, sometimes called rapid evidence assessments or rapid reviews) made up of three strands was used to explore the empirical impact of different crop improvement strategies. Rapid evidence syntheses provide relatively quick, tactical answers to key questions and are increasingly favoured by policymakers^25,26. Given the range of possible combinations of crop types, location and types of impacts, the three highest-ranked options were selected from the survey priorities: (1) improving plant water use (sustainability), (2) improving photosynthesis (yield) and (3) improving protein content and quality (nutrition). Due to a lack of relevant peer-reviewed papers assessing the impact of photosynthesis on sustainability indicators, this category was broadened to focus on yield impacts more generally.

A common research question framed the evidence synthesis: “What are the social, economic and environmental impacts of improving [plant water use/yield (photosynthesis)/protein content and quality]?”. Slight adjustments were made depending on specifics of the priorities.

A query combining several multi keyword concept operationalisation was similarly created for the three synthesis categories (see Table 1). Search strings were adjusted based the specific needs of the priority in question: for example, the targeted improvement of photosynthetic pathways is only a recent field with relatively few impact studies²⁷, so more generic improvement in yield—and the social, economic or environmental impacts this has—was targeted.

Identified papers were abstract and title screened on relevance and contents, and subsequently methodologically screened by the researcher leading that part of the synthesis within the project team on the basis of methodological norms in the relevant field of research. In cases of doubt experts in relevant fields were asked for advice on the quality assessment. When a given paper did not meet basic methodological criteria (e.g. because of issues with field trial design, model validation, or statistical analysis) the paper was removed and no further analysis of it undertaken. Papers were identified in the scientific literature using Scopus and Web of Science.

Initial searches yielded 1398 papers, 390 relating to water use, 491 relating to protein, and 515 relating to yield. After content and quality screening 10 papers remained for water use, 6 for protein and 3 for yield. Given the lack of peer-reviewed publications assessing impacts of yield, two additional papers in the grey literature were identified through Google-Scholar, bringing the total number of yield-relevant papers up to 5. Included papers were coded on relevant impact indicators.

Emerging issues: stakeholder focus groups

Between April and June 2020, 10 participatory focus groups were held to identify emerging issues with the CropBooster crop improvement options, whilst also probing how plant breeding targets can be determined and what the challenges are for European agriculture.

Focus groups complement the survey and rapid evidence synthesis as they permit the generation of new ideas, the assessment of potential ideas and insights into the differences in opinion that exist between members of particular groups^28,29. Face to face focus group protocols were adjusted to an online format to deal with coronavirus restrictions in Europe in 2020. A detailed description of these adjustments is described in Menary et al.³⁰ and incorporates insights on online-specific focus group difficulties such as those reported by Tuttas³¹. Reporting follows COREQ guidelines³². A total of thirty five participants participated in one of ten focus groups (five with farm-level, two with agri-business, and three with consumer-level participants).

Key questions and prompts

A detailed semi-structured focus group protocol was developed to guide the moderator and ensure consistency and comparability between the data from each stakeholder group (for the full protocol, see [Supplementary 7]). The protocol was piloted at Lancaster University and Wageningen University (n = 16). Primary topics were:

• The biggest challenges for the European agri-food sector over the next 30 years

• The most important CropBooster option

• The least important CropBooster option

• The social, environmental or economic impacts of a particular option

• The relevance of the options for the challenges facing the European agri-food sector

• What other things should be included in the CropBooster options?

Topics were discussed around the 15 option cards also used in the survey. Participants were asked to fill in a blank option card (#16) with a crop improvement they thought was missing from the 15 option cards and this additional input was discussed at the end of the focus groups.

Participants were encouraged to discuss the relative merits of their suggestions and agree on the most important. Prompts were used to probe participant choices—or why certain options had not been mentioned.

Focus group analysis

Adopting a Framework Analysis approach^33,34 an initial coding framework was developed by open coding of transcripts associated with each stakeholder group by the moderator responsible for that group. After these were agreed through consultation with at least one other member of the research team, the transcripts were fully coded and analysed using NVivo software. Emergent themes were cross-referenced by the moderators of the focus groups (AN, JM and SS) and an overview of themes was discussed within the wider research team. Mutual language was agreed upon for the purposes of illustrating shared themes for integrative analyses based on agreement between stakeholder specific coding trees and code books; which include non-identifying coded data and show the underlying quotes for each theme.