This paper addresses the synthesis problem of non-isothermal interplant water networks by using a mathematical programming approach based on superstructure optimisation. A recently proposed compact superstructure (Ibrić et al., 2015) was used and the mixed integer nonlinear programming (MINLP) model modified in order to identify the existence of process water using units, wastewater treatment units, and hot/cold streams within different plants, as well as optimal water flow and heat transfer between the plants. The superstructure includes direct and indirect heat exchange opportunities, with a manageable number of hot and cold streams enabling the control of heat exchanger network (HEN) complexity. In addition, a compact superstructure reduction strategy is employed in order to reduce the model size. By using those parameters having values of 1 or 0, the existence of superstructure elements of the different plants can be addressed simply, without introducing additional variables. The proposed model is solved by using a two‒step iterative solution strategy (Ibrić et al., 2015). The first step in the proposed strategy provides initialisation and rigorous bounds on water and utilities consumption. In the second step, an MINLP model is solved, simultaneously minimising the total annual cost of the network. Different case scenarios were considered, analysing water and heat integration within and between the plants. The solutions obtained show that the model can be successfully used for the synthesis of interplant non-isothermal water networks, thus minimising the total annual cost of the overall network. The results show that additional saving in total annual costs can be achieved by enabling direct water and heat integration between plants.