Orchestrating end-to-end reproducible NextGen and NextGen adjacent workflows: Part 2 Orchestrating parallel model calibration
Day 1 Session 1 (0:00AM)
Presenters:
Raymond Spiteri, University of Saskatchewan
Darri Eythorsson, University of Calgary
Jordan Read, CUASHI
James Halgren, Brigham Young University
Martyn Clark, University of Calgary
In the second session of this workshop, we will discuss the technical challenges associated with establishing and maintaining the computational environment required to use HPC resources to calibrate NextGen models for both individual watershed studies and large-sample and continental-domain experiments. We will perform a parallel calibration exercise using the model developed in the first workshop session, demonstrating the utility of massively parallel computing resources for large-domain hydrological modelling.
In both sessions, we will discuss best practices for reproducibility and provenance management throughout the workflow. We will consider key model decision points, discuss alternative model design structures, and assess their impact on model performance. We will discuss the sources of uncertainty in the model workflow and methods for characterizing and estimating them.
Learning Outcomes:
- Configure and execute a NextGen calibration workflow.
- Understand the flow of data through a general NextGen workflow
- Understand the key decision points in a NextGen model setup
- Understand approaches for scaling hydrological calibration workflows using parallel computing resources
Prerequisites:
- Familiarity with Python and Jupyter notebook environments
- Familiarity with NextGen principles and hydrological modelling concepts
- CIROH-2i2c account
- Git clone of Symfluence repository
Additional Details:
- Complements existing NGIAB workshops by addressing the pathway from single-basin calibration to large-sample and continental-scale inference using HPC resources
- Directly supports NOAA-OWP’s stated objectives of “model interoperability, intercomparison, testing of research hypotheses, and deploying into operations science-driven, evidence-based models”
- Demonstrates reproducible end-to-end workflow provenance tracking—a critical gap for operational adoption and community contribution
- Symfluence extends beyond single-model calibration to orchestrate workflows across NextGen and NextGen-adjacent models, enabling systematic model intercomparison within a unified framework.