Preprints
https://doi.org/10.5194/bg-2021-231
https://doi.org/10.5194/bg-2021-231

  06 Sep 2021

06 Sep 2021

Review status: this preprint is currently under review for the journal BG.

Comment on "Estimating causal networks in biosphere–atmosphere interaction with the PCMCI approach"

Jarmo Mäkelä1, Laila Melkas1, Ivan Mammarella2, Tuomo Nieminen2,3, Suyog Chandramouli1, Rafael Savvides1, and Kai Puolamäki1,2 Jarmo Mäkelä et al.
  • 1Department of Computer Science, P.O. Box 68, FI-00014 University of Helsinki, Helsinki, Finland
  • 2Institute for Atmospheric and Earth System Research/Physics, P.O. Box 64, FI-00014 University of Helsinki, Helsinki, Finland
  • 3Institute for Atmospheric and Earth System Research/Forest Sciences, P.O. Box 27, FI-00014 University of Helsinki, Helsinki, Finland

Abstract. This is a comment on "Estimating causal networks in biosphere–atmosphere interaction with the PCMCI approach" by Krich et al., Biogeosciences, 17, 1033–1061, 2020, which gives a good introduction to causal discovery, but confines the scope by investigating the outcome of a single algorithm. In this comment, we argue that the outputs of causal discovery algorithms should not usually be considered as end results but starting points and hypothesis for further study. We illustrate how not only different algorithms, but also different initial states and prior information of possible causal model structures, affect the outcome. We demonstrate how to incorporate expert domain knowledge with causal structure discovery and how to detect and take into account overfitting and concept drift.

Jarmo Mäkelä et al.

Status: open (until 22 Oct 2021)

Comment types: AC – author | RC – referee | CC – community | EC – editor | CEC – chief editor | : Report abuse

Jarmo Mäkelä et al.

Jarmo Mäkelä et al.

Viewed

Total article views: 327 (including HTML, PDF, and XML)
HTML PDF XML Total BibTeX EndNote
283 40 4 327 3 4
  • HTML: 283
  • PDF: 40
  • XML: 4
  • Total: 327
  • BibTeX: 3
  • EndNote: 4
Views and downloads (calculated since 06 Sep 2021)
Cumulative views and downloads (calculated since 06 Sep 2021)

Viewed (geographical distribution)

Total article views: 305 (including HTML, PDF, and XML) Thereof 305 with geography defined and 0 with unknown origin.
Country # Views %
  • 1
1
 
 
 
 
Latest update: 23 Sep 2021
Short summary
Causal structure discovery algorithms have been making headway into Earth system sciences and they can be used to e.g. increase our understanding on biosphere-atmosphere interactions. In this paper we present a procedure on how to utilise prior knowledge of the domain experts together with these algorithms in order to find more robust causal structure models. We also demonstrate how to avoid such pitfalls as overfitting and concept drift during this process.
Altmetrics