Computer models are widely used tools that aid engineers and scientists in answering water resources research and management questions. To date, many of these questions have been isolated to parts of the overall system, and as a result, so have the majority of models. Current management policies and scientific questions require a more holistic view of water resource systems, which requires models to be coupled so that they are able to simulate interactions and feedbacks across disciplinary boundaries. One challenge facing the community is how to effectively and efficiently couple models to achieve a more holistic view of the environment. The overarching goal of this work was to advance one approach for coupling water resource systems in which models act as loosely integrated components within a larger modeling system. This is achieved through three research studies, presented in manuscript form, that address specific challenges to the wide spread adoption of loosely integrated modeling of water resource systems.

The first study quantified the computational overhead of the loose integration modeling approach. The results of this study show that a loosely integrated model is able to reproduce the predictions of a more traditional modeling system, and that both approaches exhibit similar run time performance. Run time performance was evaluated by varying the number and size of data communications between components. The second study presents a methodology for establishing interoperability between observational data standards and loosely integrated model component standards. This methodology is leveraged to design an integrated modeling tool, named HydroModeler. It is then used within a larger software framework to perform an end-to-end simulation of a hydrologic system where data is discovered, downloaded, and supplied to models. The third study presents a technique for exposing water resources models as web services. The results of this study show how the challenges encountered when integrating existing models, such as operating system, computer architecture, and programming language incompatibilities, can be overcome by developing models able to operate within a service oriented architecture.