CORIE is a pilot environmental observation and forecasting system (EOFS) for the Columbia River. The goal of CORIE is to charac- terize and predict complex circulation and mixing processes in a system encompassing the lower river, the estuary, and the near-ocean using a multi-scale data assimilation model. The challenge for scientists is to maintain the accuracy of their modeling system while minimizing resource usage. In this paper, we first propose a metric for characterizing the error in the CORIE data assimilation model and study the impact of the number of sensors on the error reduction. Second, we propose a genetic algorithm to compute the optimal config- uration of sensors that reduces the number of sensors to the minimum required while maintaining a similar level of error in the data assimila- tion model. We verify the results of our algorithm with 30 runs of the data assimilation model. Each run uses data collected and estimated over a two-day period. We can reduce the sensing resource usage by 26.5% while achieving comparable error in data assimilation. As a result, we can potentially save 40 thousand dollars in initial expenses and 10 thou- sand dollars in maintenance expense per year. This algorithm can be used to guide operation of the existing observa- tion network, as well as to guide deployment of future sensor stations. The novelty of our approach is that our problem formulation of network configuration is influenced by the data assimilation framework which is more meaningful to domain scientists, rather than using abstract sensing models.