A smart decision framework for the prediction of thrips incidence in organic banana crops

Various pests which diminish the quality of the fruit have a big influence on the organic banana production in the Piura region of Peru (and not only) and prevent it from being sold on the international market. In this study, a framework for facilitating the prediction of the pest incidence in organic banana crops is developed. To achieve this, a data acquisition system with smart sensors is implemented to monitor the meteorological variables that influence the growth of the pests. The proposed framework is utilised for the assessment of various mathematical representations of the pest incidence. These models are adapted from population growth functions and built in such way as to predict the behaviour of the insects at non-regular time intervals. A hybrid approach, combining mechanistic and data-based methods is utilised for the development of the models. Both linear and nonlinear dynamic relationships with the temperature are assumed. The results show that nonlinear model representations have greater accuracy (a fit index of more than 70%), which provides a basis for improving pest management actions on the organic banana farms.