Describir: Adapting the QUEFTS model to predict attainable yields when training data are characterized by imperfect management