Anabelle Laurent

Coordonnées Mails: alaurent@iastate.edu anabelle.laurent@inrae.fr

Key words: on-farm research network, yield analysis, data visualization, crop production

The analysis of data from on-farm research network: Statistical approaches to test the efficacy of new management practices and data visualization

Supervisors: Fernando Miguez (Iowa State University) & David Makowski

Thèse en cotutelle ISU et ABIES

Abstract

On-farm research can be defined as research carried out on farmers’ fields and farmers’ environment (FAO, 2017). Experiments are conducted by the farmers themselves, under the guidance of the research team. Working under the farmers’ environment means that not all the variables can be controlled, such as seeding conditions or soil texture, for example. Therefore, it can be challenging to analyze on-farm data among other reasons because explanatory variables are unbalanced. In order to test a treatment, the experimental design and the environment have to be controlled to minimize the possibility to affect the results and the conclusion. For example, working under greenhouse conditions or with small plot research, both typically performed in a university setting, are two ways to control the environment as much as possible. The main shortcomings of these experiments are that they do not represent the range of conditions observed in growers’ fields and yield is often overestimated. This is why on-farm research has become more common in recent years. An extension of this is an on-farm research network which is a way to share useful information and knowledge across a group of farmers and also with industry and academia.

The statistical analysis of on-farm trials raises two important issues. First, on-farm networks may vary in terms of numbers of trials, numbers of years where data are collected and include different numbers of replicates. Second, trials from an on-farm research network usually covers contrasting environments corresponding to different soils, climates, and cropping systems (crop sequence and managements technique over time for a trial). There is a pressing need to develop approaches that can deal with data from an on-farm research network far and beyond those available from individual reports. In fact, individual reports cannot provide an estimate of the probability that a new management practice will or will not perform for many farmers. As several trials across a state or a region are included in an on-farm network, a more general analysis of the efficacy of a practice or product should be provided.

This project will demonstrate the importance of identifying appropriate statistical methods for analyzing on-farm research network data. It will also provide recommendations to create protocols for standardizing data, formatting datasets, conducting statistical analyses, and producing visual representations of the results (through a web application tool).

Education

M.S., Agroecology, Ecole Supérieure d'Agriculture (Angers, France), 2012

Links

Researchgate

Linkedin

Google Scholar