Suitability of multi-agent simulations to study irrigated system viability: application to case studies in the Senegal River Valley

It is assumed that Agent-Based Modeling is a useful technique for water management issues. In particular, it may provide a suitable framework for representing irrigated systems. The objective of this paper is to demonstrate its potential for a specific use: research on irrigated systems' viability in the Senegal River Valley. The main assumption to be verified is that Multi-Agent Systems constitute a suitable architecture to study theoretically irrigated systems' viability using simulations. By using Multi-Agent Systems, virtual irrigated systems can be designed that might then be used as virtual laboratories. These virtual labs constitute an alternative when real labs cannot exist for some reason. In this paper we report on experiments we have conducted using such virtual labs for exploring an Agent-Based Model through the simulation of scenarios. A scenario is defined as a triplet: an environment, a set of individual rules, a set of collective rules. It is evaluated according to the longevity of the irrigated system. An index is defined, based on the ratio of long-enduring simulations among a set of repetitions of a given scenario. Even if simulation results display significant diversity for a given scenario due to random factors in the processes simulated, the ratio of long-enduring simulations is repeatable. This entails to explore the overall behavior of the virtual irrigated system and to build theories concerning the viability of Senegalese irrigated systems. An example is given showing the need for strong coherence for a given environment among individual rules and collective rules.