Modelling water operators' time varying risk aversion by an agent-based multi lateral negotiation protocol

Multipurpose water reservoirs are operated on a tradeoff point of multiple, often conflicting, operating objectives. Generally, this tradeoff has been selected to ensure efficient and balanced performance given the observed system boundary conditions. Extreme variability in the inputs might induce change in the operator’s risk aversion and a shift of tradeoff. In this work, we explore how the selection of different optimal tradeoffs among conflicting operating objectives is linked to the variations of the boundary conditions, such as, for example, drifting rainfall season or remarkable changes in crop prices. In particular, we argue that tradeoff dynamics is driven by recent, extreme variations in system performance: underperforming on one of the objectives should push the tradeoff toward disadvantaged objectives. To test this assumption, we develop an agent-based model to simulate the operators’ tradeoff selection process. We map the selection onto a multi lateral negotiation protocol, where different virtual agents optimize different operating objectives. The agents periodically negotiate a compromise on the operating policy. The agent’s rigidity in each negotiation round is determined by the recent system performance according to the specific objective it represents. The negotiation follows a set-based egocentric monotonic concession protocol: at each step an agent incrementally adds some options to the set of its acceptable compromises and (possibly) accepts lower and lower satisfying policies until an agreement is achieved. We apply this reiterated negotiation framework on the regulated Lake Como, Italy, simulating the lake dam operation and its recurrent updates over the last 50 years.