Using locally distributed computing to aid water quality modelling

Watershed managers and other experts frequently use water quality models to manage and study watershed-waterbody systems. Many modern water quality models have become increasingly complex and require significant amount of computational resources to execute. This situation is further exacerbated by linked models, which require sequential execution of multiple models to simulate the entire system, along with data conversion to transfer data between the linked models. Many tasks, such as evaluations of the robustness of the modelling system through sensitivity analysis and applications of water quality models as a tool to aid decision-making, require multiple executions of water quality models and are problematic to undertake with standard computational abilities. The Locally Distributed Simultaneous Model Execution (LDSME) framework and the tool (based on the framework) creates a server cluster from disparate machines to increase the computation resources available for model executions, thereby enabling simultaneous execution of several water quality model instances. Moreover, the LDSME framework has the ability to execute many modelling software with ease, which allows for easy comparison of various modelling software, and the extension of the existing modelling system by incorporating additional modules/modelling software. An implementation of the LDSME framework, based on Java Remote Method Invocation (RMI), for the Occoquan Reservoir and its tributary watershed modelling system in Northern Virginia was designed. This system comprised of computers with varying capabilities that were available on the local network, and did not require any new hardware.