A systematic review of the methodology of trade-off analysis in agriculture

Trade-off analysis (TOA) is central to policy and decision-making aimed at promoting sustainable agricultural landscapes. Yet, a generic methodological framework to assess trade-offs in agriculture is absent, largely due to the wide range of research disciplines and objectives for which TOA is used. In this study, we systematically reviewed 119 studies that have implemented TOAs in landscapes and regions dominated by agricultural systems around the world. Our results highlight that TOAs tend to be unbalanced, with a strong emphasis on productivity rather than environmental and socio-cultural services. TOAs have mostly been performed at farm or regional scales, rarely considering multiple spatial scales simultaneously. Mostly, TOAs fail to include stakeholders at study development stage, disregard recommendation uncertainty due to outcome variability and overlook risks associated with the TOA outcomes. Increased attention to these aspects is critical for TOAs to guide agricultural landscapes towards sustainability.

1 Supplementary tables With regards to the logging of criteria relevant to countries: Some corresponding authors listed affiliated institutes within multiple countries.In that case, the insitute listed as first affiliation and its respective location was registered.Furthermore, some articles considered a case-study that spanned multiple countries.In that case, all countries have been included.Global studies (n=1) have been excluded from all maps.
"The human development index (HDI) is composite index measuring average achievement in three basic dimensions of human development-a long and healthy life, knowledge and a decent standard of living" (UNDP, 2020).The HDI values for the year 2019 have been used, the classifications (low, medium, high, very high) are provided by the UNDP for each year.
Figure 3 shows that based on the countries that occure with a frequency > 1, the lead author's affiliation does not include countries in the lower two categories of the human development index (HDI) except for Kenya, Pakistan and Ethiopia (left-panel).However, countries where the case-study areas are located show a much larger share of countires in the lower two categories of the HDI (right-panel) as it also includes Zimbabwe, Senegal, Mali, Zambia, Tanzania, Ghana, Uganda, Rwanda, Mauritania, Guinea, Cameroon, Benin and Angola.

Cluster analysis
The cluster analysis (Fig. 1 in main manuscript) showed a clear distinction between articles that considered the TOA within an ecosystem services methodological framework (ESS).Results for this criterion were not reported in the main manuscript (see also the supplementary material included with the manuscript that contains the raw data).
Figure 5: The frequency of whether an ecosystem services (ESS) methdological framework has been applied within an article, binned by each cluster Fig. 6 shows the frequency of which system border was used to delineate the case-study area, binned by each cluster.Clusters that delineated the case-study area by administrative boundaries only (clusters 5 and 8) were characterized by economic, human health and agronomic indicators.Cluster 2 showed the highest share of biophysical delineation and was concerned with water quantity/quality, reflecting the use of watershed boundaries.Furthermore, 54 articles (43%) assessed biophysical indicators within their TOA but the case-study are was delineated by administrative boundaries only (not shown in graph).

Wordclouds
This section reports on wordclouds based on three qualitative criteria: farming system, farm management and reported knowledge gaps in the article's discussion section (see methodology main manuscript).Wordclouds have been generated by plotting words with a minimum frequency of three and a maximum of 100 words.Redundant words have been removed based on visual inspection.
One can infer from the wordcloud that 'small-holder' farms are the most prominent farming system studied, followed by 'livestock' and 'crop-livestock' production systems.This is not reflected within Fig. 3E in the main manuscript, where livestock took up a low share.This might show that the livestock component of the agricultural system under study is often disregarded or only accounted for implicitly.However, this wordlcoud is based on a smaller sample (n = 81) as not all articles explicitly listed the farming system under study.These words are followed by 'silviculture', 'semi-subsistence' and 'small-scale', mixed agriculture.The most prominent words are 'irrigated', 'maize' and 'wheat', followed by 'livestock', 'rainfed' and 'fertilizer'.'Rice' is the third most common crop, together with 'corn', re-emphasizing the high frequency of articles that study maize/corn cropping systems.The third level (pink) lists words such as 'intensive', 'rotation', 'tillage', 'manure' and 'conservation', together with adjectives such as 'high', 'low', 'dry' and 'agricultural' (showing the limitation of wordclouds as these adjectives cannot be traced back to their respective nouns).Words in the fourth category show that 'organic' and 'inorganic' are listed in the same frequency.Notable are the words 'bioenergy' and 'vegetables', indicating niche farming systems relative to the overall trend.The last category lists management types such as 'intercropping', 'zero-grazing' and 'extensive' (note: less common in general although this should be ideally viewed in the context of the case-study area) Figure 8: Word cloud for the "farm management" criterium (n=81) A total of 99 articles from the sample reported knowledge gaps which have been visualized by a word cloud.The most prominent words are 'water' (26%) and 'ecosystem services' (24%).The latter indicates the frequency in which a need to include other ecosystem services is expressed in articles that adopt an ESS framework (24% of articles adopted an ESS framework).The next two categories indicate the frequent mention that TOA should be performed for different crops (17-18%) (yellow category) and the words 'climate ', 'economic' and 'management' (15-16% occurrence, in green).Words such as 'land use change', 'policy' and 'data' indicate that these factors need 'additional' study or be accounted for ('account') (10-13% occurrence).The fourth category (purple) lists words relative to how the TOA framework is defined ('scale' (9%) and 'levels', 'multiple', 'objectives', 'inclusion' and 'stakeholders' occurring in 7% of the articles) as well as the study's robustness: 'uncertainty', 'constraints', 'required' (7-8%).

Figure 1 :
Figure 1: Cumulative distribution of TOA articles versus cumulative number of countries, both in terms of the corresponding author's affiliation and case-study area.Colors indicate the four classifications of the human development index (HDI) as provided by the UNDP for 2019.

Figure 2 :
Figure 2: The frequency of a country as the location for the corresponding author's affiliation or the case-study area.Inset shows the density distributions of the human development index (HDI) for the author's affiliation and case-study area together with the density curves for all countries included in the HDI (Global).

Figure 3 :
Figure 3: The relative frequency of a country in which the case-study area was located for each TOA indicator included within the article.The frequency of countries sums to 100% for each indicator panel.

Figure 4 :
Figure 4: The frequency of a country in which the case-study area was located, plotted as a time series of publication date.

Figure 6 :
Figure 6: The frequency in which the TOA case-study area defined by administrative or biophysical borders for each cluster

Table S1 -
Definitions of the criteria logged during the systematic review and their associated levels