1 of 16

Glaciers, climate, , and open science

Jordi Bolibar1, Facundo Sapienza2, Bert Wouters1, Fernando Pérez2, Redouane Lguensat3, Fabien Maussion4

1

2

3

© Tom Bernardo

4

2 of 16

LeConte glacier, Alaska

3 of 16

What is a glacier?

Accumulation

Mer de Glace, Mont-Blanc (France)

​

© JF Hagenmuller

​

Ablation

Climate

​

​

Mass changes

​

​

​

Ice flow

​

Glacier evolution model

4 of 16

Why are glaciers important?

Huss et al. (2017)

  • They store ~70% of the world’s freshwater

​

  • The equivalent of 70 meters or 230 feet of sea level rise

​

  • They regulate streamflow during dry and hot seasons

​

  • Provide essential nutrient fluxes and cold water for biodiversity

5 of 16

A new era of satellite observations

Malaspina glacier, Alaska

Baltoro glacier, Karakoram, Pakistan

6 of 16

Global glacier modelling

​

​

​

​

​

​

​

​

Model the 200,000 glaciers on Earth!

​

Randolph Glacier Inventory (Pfeffer et al., 2017)

7 of 16

Global glacier modelling

​

​

​

​

​

​

​

​

​

  • Traditional modelling approaches struggle
    • Too much data!
    • Need for collaborative models

​

​

  • Data science, machine learning and open source tools are crucial

8 of 16

​

​

​

​

​

​

​

​

​

​

​

​

​

​

9 of 16

Why ?

​

​

​

​

​

​

​

​

​

  • Solves the two language problem

​

  • Great ecosystem for scientific computing at the intersection of data science and differential equations

​

  • It’s elegant

10 of 16

Scientific machine learning (SciML)

​

​

​

​

​

​

​

​

Physics (domain knowledge)

Machine learning

(data-driven models)

Numerical methods

SciML

Rackauckas et al. (2020)

11 of 16

Good communication between programming languages

12 of 16

Working in the

  • Seamless portable workflow across continents and time zones

​

  • Different machines with computing power adapted to each problem

13 of 16

Open source development on

  • Open model development

​

  • Good project visibility

​

  • Easy collaboration with multiple researchers together with JupyterHub

14 of 16

A modern perspective on science

  • Modern Earth science strongly needs data scientists

​

  • Software as a complement to the scientific paper

​

  • Reproducibility through open repositories and notebooks

15 of 16

From a competitive

​

to a collaborative

​

approach to science

16 of 16

Thank you for your attention

@JordiBolibar @ODINN_SciML

© NASA

​

j.bolibar@uu.nl

@Jordi_bolibar

@ODINN_org

@

© Tom Bernardo