Impact of farming on water resources: Assessing uncertainty with Monte Carlo simulations in a global change context

Most of the hydro-economic models used for assessing the environmental impact of agricultural policies are deterministic and can only reflect uncertainties through analyses of scenarios. In this article, we propose a methodology to assess uncertainty using Monte Carlo simulations. Taking three different global change scenarios, we vary economic parameters liable to influence the future of agriculture within each scenario. The simulations are based on farming models developed for two French regions (Midi-Pyrenees and Alsace), using linear programming (LP). These are used to simulate the impacts of a "business as usual", "liberal" and "interventionist" scenario, on water abstraction for irrigation (Neste basin) and on nitrate leaching into groundwater (Alsace). The simulations all predict a drop in farm income in both regions, with a stronger effect in the liberal scenario. Water consumed in the Neste basin increases a little (+0.3 to +3.7% in the interventionist scenario). A slight decrease of agricultural nitrate leaching is observed in Alsace, with nearly no difference between the averages for the three scenarios. Considering all Monte Carlo simulations the nitrate leaching should decrease between -28% and -43%, so uncertainty is not very important from the water planning and management point of view. However, the uncertainty on incomes is greater. A comparison between the Monte Carlo results and those from the deterministic approach demonstrates the value of taking uncertainties into account in foresight modelling exercises: and suggest that Monte Carlo associated to LP is a partial response to classical criticism addressed towards basic LP