Modelling water allocation in community irrigation using multi-agent system : A thesis submitted in partial fulfilment of the requirements for the Degree of Doctor of Philosophy at Lincoln University

Insufficient water for irrigation is a common problem in New Zealand, particularly in the Canterbury region, where the use and demand have been steadily increasing over the past 20 years (PCE, 2004; The Parliamentary Commissioner for the Environment, 2004). As a limited resource, there are restrictions around its use. While farmers who need water for irrigation can apply for consent through Environment Canterbury, the process takes a long time and is expensive. As a result, only those with large farms or those who will be able to realise greater financial benefits and higher levels of productivity tend to apply. Instead, most farmers apply to a community irrigation scheme such as Central Plain Water Limited (CPWL) who sells water to individual farmers. As a farmer must pay for each unit of water that s/he uses, s/he needs to have a good irrigation plan in place to ensure they obtain the maximum profit from their investment. In New Zealand, most farmers use computer programmes to estimate their irrigation requirements. The two most common programmes in New Zealand are IrriCalc and OVERSEER. However, both have some limitations: they can only be used to calculate the water needs of an individual farm and neither can prioritise crop water needs during periods of water scarcity. To deal with this problem, we designed an agent-based irrigation management system that can be used to optimise water allocation around the farm which is particularly useful during periods of water scarcity by taking into account the crop types and prioritising them based on the crop utility value. As it calculates the water savings based on each crop’s growth stage and prioritises it in terms of its potential sales price, this agent-based system provides a way to increase farmers’ profitability and to enables them to thrive during periods of water scarcity. During the water reduction exercise, most farms suffer from water shortages. However, there are farmers (who may have overestimated their water needs) who will have excess water. Recognising this situation, we developed a multi-agent system to improve water allocation within a community of water users (where each individual agent represents a farm) and investigated the efficiency of water distribution mechanisms among farms. Farmers can use the proposed multi-agent water management system to negotiate with each other to buy and sell water among themselves. One of the most well-known and simplest methods to achieve this is by using an auction. The choice of an auction was deliberate as it allows agents to buy water at a price, they are comfortable with. An agent must consider how much they are willing to pay for a specific volume of water to ensure their farm remains profitable. This study considered three-auction types and compared the results of each auction in terms of fair water distribution, profit for the sellers and reductions in losses for bidders. We found that the pay-per-bid auctions (discriminatory and uniform) are the best strategies for water distribution that balance between water distribution and gaining profit in water community. In addition, we also investigated how varying behaviours of sellers and buyer affect the outcome of the auction