Soft Sensors to Measure Energy Consumption of Topographic Uncertainty Propagation in Mass Flow Simulations

Authors: Eren Edgü, Alan Correa

Motivation

Benchmarking methods to propagate topographic uncertainties in surface flow simulations involves comparing their accuracy, resource usage, and energy consumption. Currently, compute time is used as a proxy for energy consumption; however, it can be misleading due to the increasing heterogeneity in modern computational workflows.

Goal

Estimate consumption and resource usage of surface flow simulators for topographic uncertainty propagation using soft sensors.

Tasks

• >> Develop a workflow to propagate topographic uncertainty using a surface flow simulator with the Monte-Carlo method
• >> Use a Soft Sensor (e.g. Alumet) to estimate energy consumption and resource usage of the workflow

Getting started

Ensure you have Micromamba installed on your system.

We use a unified setup script to build all Micromamba environments and compile Alumet from source:

bash setup.sh

To execute the monte carlo and the analysis pipeline:

bash run_pipeline.sh

The resulting plots will be saved in the /plots directory.

References

[1] Zhao, H. and Kowalski, J.: Topographic uncertainty quantification for flow-like landslide models via stochastic simulations, Nat. Hazards Earth Syst. Sci., 20, 1441–1461, https://doi.org/10.5194/nhess-20-1441-2020, 2020.