KULTIVAS: Feasability study of a variety- location model for apple cultivation

Abstract

Choosing the most suitable locations for a given apple variety is a difficult and wide-ranging decision for farmers and all other involved parties in apple production. The selected apple variety has substantial influence on the profitability of an orchard. In this perspective, with KULTIVAS we propose an approach for a data-driven decision support system to an ongoing problem in apple cultivation. In this project we use selected historical apple production and quality data, modern data management systems, algorithms, machine learning techniques, and a multidisciplinary approach involving experts from different disciplines, such as agronomy, physiology, climatology, and computer science, to assess the site-specific suitability for the cultivation of specific apple varieties. Our developed prediction model is based on spatially interpolated climate and topography data, as well as apple production and quality data from different cooperatives. It can estimate various crop parameters, such as apple size, color, and yield, using spatially available information on climate. The quality of the prediction of our statistical model depends on the coverage of the possible climatic and topographic variability in the areas by the training data sets. Predictions that fall outside this confidence range have limited predictive power. It is planned to improve the results of this study with data from additional locations and improved algorithms to improve geographical matching of sorting data, as this data set heavily influences the results.