This paper reports on complex adaptive system (CAS) simulation of the River Murray Basin in Australia to compare capacity of institutional options to maintain functioning of key river system within a "bandwidth" that limits irreversible system state changes and highly adverse consequences. The modelling framework characterise diverse irrigation agents who profit from water diversion and cause external salinity impacts, water and salt process that form the link between irrigator actions and agricultural profits and external costs, and a river manager who sets institutional rules. Emphasis is on the CAS nature of the system and on institutional rules to accommodate choosing actions differently based on con dition of the system has been referred to as state contingent management (Wills, 2003) or threshold based management (Roe and Van Eeten, 2001). Key findings are that policy focus on the source of salinity by reducing drainage are much more cost effective than strategies to mitigate salinity once it occurs and that state contingent dilution provision when it has high benefit and low opportunity cost is also a cost effective way to manage salinity.