Workshops > Accounting for the uncertainties, from the laboratory to the observations, through the model

Accounting for the uncertainties, from the laboratory to the observations, through the model

Organisers: Frédéric Galliano (frederic.galliano@cea.fr), Karine Demyk (karine.demyk@irap.omp.eu), Pierre Gratier (pierre.gratier@u-bordeaux.fr)

Wednesday, 26 October 8:30-12:00

Summary:

An accurate estimate of the uncertainties is essential in all empirical sciences. Yet, when using a model to interpret observations, there are numerous layers of uncertainties that affect the final estimate of the physical quantities, and that are difficult to track and account for. This is particularly important for ISM studies, as our models often contain microphysical quantities that have been measured in the laboratory (cross-sections, reaction rates, etc.) and are put into a radiative transfer framework. In addition, observations are frequently corrected using external models that have their own set of assumptions and errors (extinction correction, contamination subtraction, etc.). Overall, improperly estimated uncertainties could change the outcome of a study and jeopardize our assessment of the reality of a phenomenon. This will become more important in the future, as we will see an increase in the accuracy of measurements, in the amount of heterogeneous data to analyze and in the complexity of models.

The goal of this workshop is to initiate a dialogue between laboratory experimenters, observers, modelers and astrostatisticians about this question.  We will discuss a few scientific cases where the estimate of uncertainties is instrumental. We will highlight the different aspects that we need to take into account: (i) the microphysical quantities; (ii) their propagation through the model; (iii) the observations; (iv) their correction; and (v) the fit. We will emphasize the need to better characterize the errors by estimating their statistical and systematic components, their non-gaussianity, and their correlations. We will call for contributions from the audience. We are interested in discussing particular use cases where there is either a need for clarifying different sources of uncertainties or where original solutions to account for them have been found. A large fraction of this workshop will be devoted to discussions between participants. We hope, in the end, to delineate a few good practice suggestions and simple solutions to implement them. A summary of this workshop will be written by the organizers and will be available via the PCMI newsletter.

Presentations:

Video recordings of the presentations are available through the links below. The bundle of all the slides is available here.

Program:

08:30-08:55 INTRODUCTION: WHY UNCERTAINTIES ARE INSTRUMENTAL

08:30-08:40 Frédéric GALLIANO - Motivations and objectives of the workshop - An example of a nested uncertainty problem

08:40-08:50 Marie GUEGUEN - A philosopher's viewpoint on identifying, quantifying and communicating uncertainties

08:50-08:55 Everyone - Discussion

 

08:55-10:00 QUANTIFYING EXPERIMENTAL AND OBSERVATIONAL UNCERTAINTIES

08:55-09:05 Karine DEMYK - An overview of the challenges of estimating experimental uncertainties

09:05-09:15 Marco MINISSALE - Uncertainties in ice laboratory experiments

09:15-09:20 Everyone - Discussion

09:20-09:30 Lucas EINIG - Signal extraction from noisy line cubes: the problem of applying hyperspectral imaging methods

09:30-09:35 Everyone - Discussion

09:35-09:45 Nathalie YSARD - Uncertainties in dust models

09:45-10:00 Everyone - Discussion

 

10:30-11:20 PROPAGATING UNCERTAINTIES THROUGH DATA PROCESSING AND MODELING

10:30-10:40 Frédéric GALLIANO - Techniques to propagate uncertainties through data processing

10:40-10:45 Everyone - Discussion

10:45-10:55 Lise RAMAMBASON - Challenges for topological models of the interstellar medium

10:55-11:00 Everyone - Discussion

11:00-11:10 Erwan ALLYS - Evaluating uncertainties for components separation from observational data

11:10-11:20 Everyone - Discussion

 

11:20-12:00 HOW TO PUBLISH UNCERTAINTIES AND ALLOW FUTURE STUDIES TO USE THEM CONSISTENTLY?

11:20-11:25 Pierre GRATIER - Quoting and plotting errors, and accounting for their correlation with ancillary phenomena

11:25-11:35 Everyone - Discussion

11:35-11:40 Pierre GRATIER - How to store and distribute this information

11:40-11:50 Everyone - Discussion

11:50-12:00 Everyone - Conclusion: what to do next?

                                                          


 

 

Online user: 2 Privacy
Loading...