Send message to Authors

Short Abstract

This study introduces an approach for profiling the contributors of agri-food citizen science projects. The methodology relied only on the contributions’ characteristics. 773 contributors from a crowdsourcing application in viticulture were analysed.

Abstract

To monitor the dynamics of grapevine water stress at the regional level, Apex-Vigne, a free smartphone application, has been offered to wine and viticulture professionals for over five years. This application has enabled the collection of 33,140 observations, with over 773 contributors for nearly 13,855 farm plots monitored in France. It constitutes a unique crowdsourced dataset in agriculture and confirms the value of citizen science for monitoring complex phenomena at large-scale (e.g., drought, pests).

The main challenge in this type of project is to identify, within the large volume of collected data, the observations that accurately describe the phenomenon of interest. One classical approach is to assign a confidence score to each observation based on the profile of the contributor. In the Apex-Vigne project, contributors are supposed to be farmers, seasonal workers, advisors, or researchers with different motivations and expertise. However, the exact profiles of these contributors and their respective proportions remain unknown. The objective of this paper is to propose an automatic approach to profile contributors, considering only the characteristics of their contributions.

Contributor profiling was performed using hierarchical clustering based on 6 descriptors, like the number of contributions, the number of plots, and years of app use. Clusters were characterised through the descriptors and a user survey.

Four profiles were identified, aligning with a priori knowledge and survey results. This foundation could support future work on assigning confidence scores to contributions, thereby facilitating the characterisation of data quality in agri-food citizen science projects.