Best management practices of tillage and nitrogen fertilization in Mediterranean rainfed conditions: Combining field and modelling approaches

Abstract

In this work, appropriate management practices for crop production under the variable climate conditions of the Mediterranean region, in particular rainfall, were tested with the use of a modelling system applied to long-term (i.e. 18 years) field data. The calibration of the CropSyst model was performed using data collected from 1996 to 1999 at three different Mediterranean locations (i.e., HYP-Guissona, MYP-Agramunt and LYP-Candasnos, i.e. high, medium and low yield potential, respectively) within a degree of yield potential. The model simulated reasonably well barley growth and yield to different tillage and N fertilization strategies.

Simulations of barley performance over 50 years with generated weather data showed that yields were often greater and never smaller under no-tillage compared to conventional tillage with a mean increase of 36%, 63% and 18% for HYP-Guissona, MYP-Agramunt and LYP-Candasnos. In MYP-Agramunt, the long-term data showed a 40% increase in grain yields when using no-tillage compared to conventional tillage, as an average of 18 years.

The model also predicted that greater N applications in no-tillage were appropriate to take advantage of additional water supply. Taking into account the limited amount of soil water available, overall N fertilizer applications could be reduced to about half of the traditional rate applied by the farmers without yield loss. The 50-yr simulation, confirmed by the long-term experimental data, identified no-tillage as the most appropriate tillage practice for the rainfed Mediterranean areas. Also, N fertilization must be reduced significantly when tillage is used or when increasing aridity. Our work demonstrates the usefulness of the combination of long-term field experimentation and modelling as a tool to identify the best agricultural management practices. It also highlights the importance of posterior analysis with long-term observed field data to determine the performance of simulation results.