Designing underwater sonar networks is a challenging task due to the complex marine environment and combinatorial structure of the optimization problem. In recent years, heterogeneous multistatic sonar networks have gained increased attention due to improved performance, spatial coverage and robustness to different scenarios. However, many existing approaches rely on simplified acoustic models that neglect key environmental factors such as temperature gradients, salinity variations, and bathymetry. These limitations are particularly important in regions such as the Baltic Sea, where seasonal stratification leads to substantial variability in acoustic propagation. In this work, we consider the design of a static and predominantly passive sonar network, augmented with active sensors, which can be cued for improved localization and tracking. This hybrid sensing paradigm balances persistent monitoring with high-accuracy localization when needed. We propose a scalable framework for optimal network design that integrates mixed-integer linear programming with high-fidelity acoustic simulations. To address the environmental uncertainty, we formulate the problem as a robust optimization problem. The resulting approach enables the network to achieve reliable performance across a wide range of uncertain and dynamic operating conditions despite the computational complexity of the problem.