| Article ID: | iaor20114237 |
| Volume: | 11 |
| Issue: | 1 |
| Start Page Number: | 19 |
| End Page Number: | 39 |
| Publication Date: | May 2011 |
| Journal: | Operational Research |
| Authors: | Crespo Olivier, Bergez E, Garcia F |
| Keywords: | simulation: applications, optimization, stochastic processes |
Optimization by simulation of agricultural practices can help to improve irrigation water use efficiency. This work introduces an efficient hierarchical decomposition method to design irrigation management strategies that is modeled as a continuous stochastic problem. Various combinations of selection (greedy, Pareto‐based), division (middle, pivot, maximization) and evaluation techniques (global, standard deviation) were tested. We present results of an 8‐continuous‐parameter irrigation strategies design. Two criteria were chosen to evaluate the different combinations: the achieved direct margin, and the number of simulation runs that were needed to reach it. Selection techniques impacted the resolution time, while the evaluation techniques impacted the direct margin efficiency. Based on the two former criteria, the trade‐off combination of greedy selection, pivot partition and average value evaluation appeared to be the most efficient to design irrigation strategies.