Early pandemic outbreak detection in cities is a crucial but challenging task. Complementary to the costly massive individual testing, urban sewage surveillance offers a rare, cost-effective solution for large-scale monitoring of pandemic spread in cities with minimal interference to people’s lives. One emerging question is how to derive a cost-effective sensor placement plan in city-scale sewage networks having complicated topologies. Inspired by remote sensing, we first provide a general multi-objective formulation of the optimal sensor placement problem on directed networks. Then, we introduce a connectivity-based objective evaluation approach and embed it into an NSGA-II algorithm to enable efficient optimization on large-scale directed graphs. The effectiveness of the proposed method is verified on a real-world sewage network in Hong Kong serving more than 500,000 urban residents. Results show that the proposed method efficiently generated optimal sensor placement plans on city-scale networks. Optimized sensor placement plans outperformed human placement heuristics by a significant margin of 102%, highlighting the necessity for data-driven decision support for large-scale urban sensing. Methodologically, this study provides a benchmark problem and datasets for network-based spatial optimization studies. Codes and datasets developed in this study are open-sourced to support future research in a real-world scenario.