Lack of consideration for end-users during the design of agronomic models. A review

Agriculture should now provide not only high yields but also sustainable development with a sound management of the diversity of ecosystems. Due to this increased complexity of objectives, models have recently become major tools that can integrate several parameters. Failure to apply models outside research is however a major issue. Here, to identify the precise grounds of this failure, we analyzed what models are intended for by scientists during their design. We performed a literature analysis on agronomic modelling practices. Specifically, we analyzed 518 scientific article abstracts reporting either new models or improved existing models. Articles were published in eight mainstream agronomy journals over a 10-year period. We also analyzed 25 full-text contents randomly selected from the initial dataset. In order to assess how models match the uses they are intended for, we first analyzed the design methodology used to build models. Second, we studied how authors defined the potential use of models by analyzing both the claimed objectives and references to model use and users. We then compared our findings on design methodology with our findings on intended use. Our results first show that the design methodology for modelling is presented as a segmented and standardized process. Each article refers to one or more of the following six steps to describe the design process for modelling: (1) description of the model structure, inputs, outputs and validity domain, (2) description of the data used to build the model, (3) model formalism, (4) calibration parameterisation, (5) validation, and (6) application. We found that information about the design process like iterations, errors, improvements is never emphasized in the abstracts, whereas this information is sometimes quoted in the full-text contents. This finding demonstrates that the design methodology for modelling is not addressed as a research topic. Second, we show that whereas 88.8% of authors claim in their abstracts that the major objective of their models is to improve understanding as opposed to support action, 19.5% of authors also quote a possible use of their models outside research. The initial objective of understanding is thus extended to use the models as tools for action. Overall, we conclude that the agricultural research community is not highly concerned by the effects of the design methodology on the suitability of the model structure and on potential applications. Moreover, although the six steps of the design process may be appropriate for designing models devoted to improve understanding, no specific methods are proposed to design models for action. We did not find evidence that the modellers connect the design of the model with its use by end-users. We suggest that this issue could be solved by developing participatory methodology design involving end-users in model design.