NextGen Water Resources Modeling Framework: An Overview of the Model Agnostic Approach to Integrating and Coupling Model Runtime Applications.

Five years of operating the National Water Model for hydrologic prediction across the Continental United States and beyond has taught many lessons in both hydrologic prediction as well as the technologies used to make those predictions. The Next Generation Water Resources Modeling Framework (NextGen) under active development by the Office of Water Prediction, aims to bring new technologies and predictive capabilities to hydrologic prediction from single catchment scale to continental scale.

NextGen applies community standards to facilitate model interoperability, allowing each feature in the domain to be modeled with a potentially unique model formulation or combination of formulations, without the need to re-compile and deploy the model application. This key feature promotes the scientific concept of uniqueness of place, and allows for the exploration of dominant processes as well as parsimony, both critical concepts for operational modeling.

We implement and utilize several key standards, including the Basic Model Interface (BMI) to provide model integration and coupling, the Open Geospatial Consortium (OGC) Hy Features Model and associated hydrofabric for describing the hydrologic features of the landscape, and other technical standards for quality software development and deployment. NextGen integrates models with BMI in a dynamic runtime environment. The framework is capable of dynamically loading computational models written in Python, C, C++ or Fortran, allowing model developers flexibility in programming and development, from language selection to tool chains. This also eases the research to operations transition, as various model components can be developed, tested, and validated independently. The integration of the components into an operational configuration requires no code modifications, and instead focuses on the predictive capabilities and analysis of the model results.

We will present details of the framework model engine’s capabilities, as well discuss current and ongoing work to develop this model agnostic framework and the utilities required to operate and integrate a diverse set of computational models into an operational model serving continental predictions.