[PhD’s Corner] Léo Pichon: Development of a crowdsourcing approach adapted to agriculture specificities: the case of monitoring vine water status with the shoot growth approach

 Léo Pichon is one of the #DigitAg labelled PhDs

Léo will defend his thesis on 13th July 2021 at 10 AM at the Institut Agro, Montpellier (Amphithéâtre Philippe Lamour – Building 9, Coeur d’école, Floor 2).
The link to attend the defence through visioconference will be given later on.

Development of a crowdsourcing approach adapted to agriculture specificities: the case of monitoring vine water status with the shoot growth approach

  • Start Date: November 2017
  • University: MUSE Montpellier University of Excellence / Institut Agro – Montpellier SupAgro
  • PhD School: GAIA, Montpellier
  • Field(s): ICT for Agriculture, Agronomie
  • Doctoral Thesis Advisor: Véronique Bellon-Maurel (Inrae, UMR Itap)
  • Co-supervisors : Bruno Tisseyre (Institut Agro, Itap), James Taylor (Newcastle University, UK / Inrae, Itap)
  • Funding: Institut Agro – Montpellier SupAgro
  • #DigitAg: Labelled PhD – Axis 3, 4 & 6 – Challenge 1 & 6

Keywords: Precision Viticulture, Decision making, Data quality, Crowdsourcing, High resolution data

Abstract: Crowdsourcing is an approach consisting in answering a question defined by an organisation (research laboratory, company, etc.) by relying on the collective intelligence of a community of contributors. To date, crowdsourcing is not widely spread in agriculture, but it has great potential for collecting georeferenced observations to monitor phenomena at regional scale (e.g. diseases, pests or abiotic stresses monitoring). These crowdsourcing projects in agriculture have specificities in terms of participants (professional contributors, importance of the role of advisors), studied phenomena (with strong spatial and temporal covariances) and datasets collected (asynchronous and heterotopic) that have led some authors to coin the concept of farmsourcing to describe them. These specificities of farmsourcing projects influence the design of the projects and the involvement of the different stakeholders. They also influence the criteria and indicators for evaluating the success of such projects. Finally, they influence the methods for identifying outliers and surprising observations in corresponding datasets. To date, there is no existing approach taking into account the specificities of farmsourcing projects. The objective of this thesis is to propose tools and methods to develop a farmsourcing approach in both the design and the evaluation of the project (How to foster the contribution of participants? How to evaluate the success of a project?) and then in the characterisation of the quality of the resulting observations (How to identify outliers and surprising observations? How can these approaches be automated?) The thesis is based on a systemic approach with the implementation of a case study. This case study is the monitoring of the vine water status at regional scale using i) an indicator (iG-Apex) based on observations of vine shoot growth and ii) the development of a dedicated farmsourcing application (ApeX-Vigne).
Firstly, the work demonstrated the value of a simple but noisy approach, such as the one based iG-Apex, for characterising an agronomic variable of interest (in this case, the vine water status) at the field and intra-field levels in a decision support context. They demonstrated how an approach like this could be used to promote participation in farmsourcing projects. The work carried out explored the technological and methodological choices for designing and deploying on a large scale a mobile application promoting the gathering of georeferenced farmsourcing observations. It also proposed a simple approach based on the study of spatial structure to assess the capacity of these projects to provide relevant information at the regional scale. Finally, the work carried out explored an approach for automatically identifying outliers and surprising observations in farmsourcing datasets. This approach is based on density-based clustering methods taking into account spatial, temporal and attribute characteristics of observations.
In the coming years, this work should enable the development of farmsourcing tools and projects giving access to new sources of information for decision support at different spatial scales.

Contact:  leo.pichon [AT] supagro.fr

Networks: ResearchGate  LinkedIn


Papers in international journals

Pichon, L., Brunel, G., Payan, J.C. et al. (2021). ApeX-Vigne: experiences in monitoring vine water status from within-field to regional scales using crowdsourcing data from a free mobile phone application, Precision Agriculture, 22, 608–626  https://doi.org/10.1007/s11119-021-09797-9