Epistics: A dynamic model to generate nitrogen fertilisation and irrigation schedules in apple orchards, with special attention to qualitative evaluation of the model

Epistics is a model combining a biophysical and a decisional model designed to generate irrigation and N fertilisation schedules in apple orchards. These techniques were chosen since they are key elements in the management of fruit tree cropping systems. The biophysical model representing water and N dynamics in orchards was based on the water and N dynamics of Stics and was completed using a crop water and N requirement estimation method adapted to orchards. It was linked to an agronomic decision rule in a combined model able to generate N fertilisation and irrigation schedules. The Epistics evaluation process dealt with numerical evaluation of state variables (water and N soil content) and qualitative evaluation of model-generated schedules. The numerical evaluation, which concerned the biophysical model of Epistics, was performed on the basis of (i) soil nitrate and water content at the end of winters 2002 and 2003, and on (ii) nitrate and water dynamics during spring and summer 2003. The mean Root Mean Squared Error (RMSE) between observed and simulated values at the end of winter was 3.3% water per horizon and 56 kg N/ha, which is relatively good owing to the high spatial and temporal variability of soil water and nitrate content. The qualitative evaluation of generated schedules was performed during interviews with farmers. Farmers were asked to evaluate the model with reference to their own practices. A sharp difference between farmers and the model concerned the beginning of the irrigation period. This suggested that the model should take into account the constraints imposed by scab and codling moth control practices and irrigation rounds. The difference between model-generated and farmers’ fertilisation practices suggested that the model may take plot vigour into account in the fertilisation decision rule. Such a study is a first step towards the design of models linking sound agronomic decision rules to crop modelling and representing interactions between practices.