Data-Driven Methods for Modelling Drainage Water Outflow
This paper presents predictive models for identification of drainage periods and amount of drained water from tile-drained agricultural fields. The study utilizes data mining methodology over data collected in Western France. The learning of the models encompasses data preprocessing and the building itself, as well as model evaluation using unseen data. The results show that by considering the data mining methodology, the process of drainage water outflow can be accurately described and predicted using descriptive climatic, soil and crop data. The discussion encompasses possible applicability of the generated models for the purpose of ground-water protection from plant protection products used in agriculture.
Index terms- Tile-drained water, ground water protection, outflow prediction, data mining.