During droughts and heat waves, thermal power plants that discharge heated effluent into rivers are often granted thermal variances permitting them to exceed the temperature restrictions imposed on effluent for protection of local aquatic ecosystems. These thermal variances are often justified as necessary for maintaining electricity reliability, particularly as heat waves typically cause an increase in electricity demand. However, current practice lacks tools for the development of grid-scale operational policies that specify the minimal thermal variances required to ensure reliable electricity supply. Creating these policies requires consideration of characteristics of individual power plants, topology and characteristics of the electricity grid, and locations of power plants within the river basin. We develop a methodology for creating such policies that considers these necessary factors. Conceptually, the operational policies developed are similar to the widely used rule curves of reservoir management, as we develop optimal rules for different hydrological and meteorological conditions. The rules are conditioned on leading modes of the ambient hydrological and meteorological conditions at the different power plant locations, leveraging the statistical correlation that exists between these conditions due to geographical proximity and hydrological connectedness. Heat dissipation in rivers and cooling ponds is modeled using the equilibrium temperature concept. Optimal rules are determined through a linear optimization with stochastic costs. We illustrate the methodology with a representative electricity grid model of eight power plants in Illinois that were granted thermal variances in the summer of 2012. Our methodology can facilitate cooperative decision making between environmental agencies, power grid operators, and power plant operators. (C) 2017 Elsevier Ltd. All rights reserved.