A modern, modular and extensible framework
for large scale glacier modeling

Fabien Maussion


IARPC Glaciers & Sea-level Collaboration Team April 2020 Meeting
These slides are available online at oggm.org/framework_talk

Use space, shift+space and arrow keys to navigate between slides.
Creative Commons License
Unless specified otherwise, the content of this site is licensed under a CC BY 4.0 license.

Goals of today's talk

  • (Re-)brand OGGM as a modelling framework in which you can fit your own model and ideas
  • Showcase some of OGGM's features, including modularity, online tutorials and reproducible science
  • List some challenges the project is facing ...
  • ... and discuss ways forward
    (which may involve your participation)

Today's Non-goals

  • Sell OGGM as "the model which has it all":
    other models out there are objectively "better", and model diversity is the motto that drives our project
  • Getting lost in project specific details: the ideas and tools presented here are applicable to other projects as well

Motivation

Many glaciers, few data

Uncertainty assessment is hard

GlacierMIP model spread shows that the individual models' uncertainty estimates are way too low
Share of each source to the total uncertainty:

GlacierMIP consortium (in press)

Uncertain and uncoordinated boundary conditions

Difficulties to agree on experimental protocols within GlacierMIP leads to apples and oranges comparisons.

More physics is always better, right?

Right?
Little reflection (so far) in the community about
which processes are adding value,
and which are adding noise.

Poor data availability and model intercomparison issues make it harder.

A common misconception

Open source = free (as in "free beer")

Free ∈ Open source

Science relies on

  • peer review
  • skepticism
  • transparency
  • attribution
  • accountability
  • collaboration
  • impact


Since always, academic science has been perfecting these tenents.
Open source software is now excellent at all of them.
© original slide by Katy Huff

Collaborative development workflow

The OGGM workflow

"Glacier centric" approach

Local grid and DEM

Centerlines

Flowlines and widths

Mass-balance & ice thickness

Glacier evolution

Workflow in practice

Glacier directories


Persistent data containers on disk

Entity tasks


Actions to perform on a glacier directory
CC BY-SA 3.0 Wikimedia

Global tasks


Actions to perform on many glaciers

In code


# Go - initialize glacier directories
gdirs = workflow.init_glacier_directories(rgidf)

task_list = [
    tasks.define_glacier_region,
    tasks.glacier_masks,
    tasks.compute_centerlines,
    tasks.catchment_area,
    tasks.catchment_width_geom,
]
for task in task_list:
    execute_entity_task(task, gdirs)

Extensible and Modular

© Lego


(alpha version)

Example: the RGI-TOPO dataset


See the rgitools documentation

Example: ITS_LIVE 

gdir = init_glacier_directories(['RGI60-01.10689'])[0]

from oggm.shop import its_live, dems
dems.select_dem(gdir, dem_source='COPDEM')
tasks.glacier_masks(gdir)
its_live.vel_to_gdir(gdir)

Online documentation and reproducible science


See our online tutorials

Challenges

  • Technical debt
  • Entry level for new contributors is high
  • Maintenance work relying on too few scientists
  • Funding incremental improvements is difficult

Take home

  • OGGM is a modelling framework:
    cherry picking is encouraged!
  • your own fantastic model or game changing idea might fit into the framework: get in touch with us!
  • check out oggm.org and edu.oggm.org,
    and help us to make it even better!