Water property rights in rivers with large dams

Abstract

This thesis is concerned with the allocation of water in regulated rivers: rivers controlled by large dams. Water property right reforms undertaken in Australia and elsewhere have decentralised water allocation: both spatial (via market trade) and inter-temporal (via storage rights). While these reforms have proven successful, externalities and transaction costs always persist. This thesis asks: what is the preferred system of water rights - in terms of allocative efficiency - given that all proposals are in some way {u0300}second-best'? We test property right systems using a decentralised model of a regulated river, in which a large number of users make private trade and storage decisions. The model is unique in representing surface water rights in a multi-agent stochastic dynamic environment (a stochastic game). To solve the model we develop a novel computational method in the spirit of 'multi-agent systems', which combines reinforcement learning algorithms from computer science with learning concepts from game theory. Ultimately, this allows us to populate the model with near optimal selfish agents. We present three applications of the model. First, we consider the design of water storage rights - where users hold private storage reserves in public reservoirs. In particular, we compare 'capacity sharing' (Dudley and Musgrave 1998) with alternative approaches. Second, we reconsider the issue of priority water rights - where certain users receive water allocations before others - in the context of storage. Third, we reconsider all of these issues, in the context of in-stream users: in particular a large environmental water holder. To maintain generality we specify broad ranges for the parameters of our model - based on statistics for the Australian Murray-Darling Basin (MDB) - and present the results of a large number of model runs. We compliment our results with discussion of the water institutions of Australia and the western US. Show more