Evaluating the drivers of banana flowering cycle duration using a stochastic model and on farm production data

International audienceThe use of data produced by farmers to generate knowledge and to inform production decisions is one of the objectives of precision agriculture (PA). Frameworks to analyse and represent those data are now widely available for many crops but are not relevant to banana cropping systems because of its asynchronicity. The average period between two flowering events on the same plant, called CD, is variable, as opposed to most crops whose phenology is synchronized by the seasons. Therefore, CD is a property of interest for understanding yields but it cannot be easily measured in plantations. In this study, a method was proposed to generate information on CD using the producer’s commercially collected production data on flowering from 724 fields over 15 years in a Cameroonian plantation. A new stochastic model based on the development of the banana plant and using CD as a parameter was formulated. This model was then adjusted using temporal flowering data to estimate CD at the field level. Results showed that CD was variable between fields (median of 209 days and standard deviation of 24 days) and spatially structured (autocorrelation ratio of 0.87). The effect of management (irrigation, cultivar) and environmental conditions (temperatures, elevation) on CD was studied. There was no positive effect of thermal time (degree days) vs. Julian days on model fitting and the CD estimation. In contrast, elevation, cultivar and irrigation had an effect on CD. For growers, this highlights the importance of taking into account CD when analysing yields, and not to rely only on the weight of the bunches at harvest. It also shows that production data can be effectively used and mapped to inform production decisions