Salish Sea Nowcast: A Real-time High-Resolution Model for Forecasts and Research Support
Salish Sea Nowcast: A Real-time High-Resolution Model for Forecasts and Research Support
Abstract:
The Salish Sea real-time model system produces two forecasts and a nowcast daily, providing storm surge forecasts for several municipality stakeholders.
Without human intervention, the automation system collects the required forcing data from various web services, runs the model and publishes results in the form of plots on several web pages.
Here we will present the automation framework that enables a research model to be run operationally.
The automation runs across two computer systems with a cloud computing facility running the numerical model (NEMO in our case), and a local Linux server doing everything else.
The system has a modular, asynchronous architecture that is coordinated by a messaging framework.
A manager process coordinates the sequencing and operation of a collection of worker processes each responsible for a specific task in the preparation for a model run, execution of the run, analysis, visualization, and publication to the web of the run results.
Techniques that make the system reasonably fault tolerant will be discussed.
The modular design easily allows researchers with a variety of skill sets to contribute to the framework to the benefit of the project and its knowledge transfer to stakeholders.
We will discuss the performance of the system during the 2014-2016 storm surge seasons,
and routine evaluation against sea surface height observation data streams.
Daily model runs with best available weather and river runoff forcing facilitate continuous evaluation against cabled observatory data streams.
We will show how those evaluations provide important insights that help to drive
research that improves the model.
Without human intervention, the automation system collects the required forcing data from various web services, runs the model and publishes results in the form of plots on several web pages.
Here we will present the automation framework that enables a research model to be run operationally.
The automation runs across two computer systems with a cloud computing facility running the numerical model (NEMO in our case), and a local Linux server doing everything else.
The system has a modular, asynchronous architecture that is coordinated by a messaging framework.
A manager process coordinates the sequencing and operation of a collection of worker processes each responsible for a specific task in the preparation for a model run, execution of the run, analysis, visualization, and publication to the web of the run results.
Techniques that make the system reasonably fault tolerant will be discussed.
The modular design easily allows researchers with a variety of skill sets to contribute to the framework to the benefit of the project and its knowledge transfer to stakeholders.
We will discuss the performance of the system during the 2014-2016 storm surge seasons,
and routine evaluation against sea surface height observation data streams.
Daily model runs with best available weather and river runoff forcing facilitate continuous evaluation against cabled observatory data streams.
We will show how those evaluations provide important insights that help to drive
research that improves the model.