reproducible research an historical oceanographic perspective - - PowerPoint PPT Presentation

reproducible research

an historical oceanographic perspective dipl.-oz. felix morsdorf, felix.morsdorf@geo.uzh.ch

reproducibility is not “nice to have”

“Reproducibility is one of the main principles of the scientific method.”

sea-surface salinity simulation

ctd probe conductivity, temperature and depth

• physical or experimental
• ceanographers probe the
• cean using ctds
• ships are used to navigate

along transects

• every 10-30 nm a probe is

lowered

world ocean circulation experiment (woce) transects

different nations, different scientists,different ships, different instruments, different calibration schemes

student’s question

• how can single ship-based measurements in such

a complex spatio-temporal system be reproducible?

• e.g. how do we know if a measurement is right or

wrong?

• supervisor: “it will come out eventually…”

physical models

Study of a Karman vortex using the FFD solver – “Design Systems, Ecology and Time” – ACADIA 2012

inverting hydrographic box models

• horizontal layers to incorporate the stratification
• stationary subsets of governing equations, constrain

the model with data (and priors…)

• find one (of the many…) inverse solutions

http://www.seos-project.eu/modules/

• ceancurrents/oceancurrents-c06-s02-

p01.html

CLIVAR Conference, Kiel, 2000

• A. Ganachaud (PhD of inverse modelling “guru” Carl

Wunsch) showed his inversion results using the WOCE transects

60°W 0°W 60°W 120°W 180°W 120°W 50°S 25°S 0° 25°N 50°N

15± 2 14± 2 12± 2 3± 1 16± 2 13± 2 4± 2 (Entire Atlantic) 10± 2.5 16± 3 6± 1.3 23± 3 140± 6 21± 6 8± 9 157± 10 9± 3 7± 2 19± 5 5.5± 2 2.5± 4 0.5± 4 1 ± 3 1.5± 1 16± 5 8± 4 3± 5 27± 6 14± 6 11± 5

28.11<γn 27.72<γn<28.11 γn<27.72 Global circulation summary standard solution (Sv)

(Entire Southern)

Ganachaud, A. and Wunsch, C. (2000). Improved estimates of global ocean circulation, heat transport and mixing from hydrographic data. Nature, 408(6811):453–457.

Ganachaud, A. and Wunsch, C. (2000). Improved estimates of global ocean circulation, heat transport and mixing from hydrographic data. Nature, 408(6811):453–457.

How to synthesize? Estimation/optimal control problem: Use a model (MITgcm) and its adjoint:

Argo T/P, Jason GRACE WOCE

slide from C. Wunsch, NCAR, 2008

Conclusions (i)

• Data needs to fit abstracted view (e.g. models) of

the world system

• If data does not fit -
• i) model is wrong or we are outside model

domain

• ii) data might have “problems”, i.e. measurement

errors that hinder reproduction of results

Conclusions (ii)

• Sampling is always an issue.
• Model development helps to test “reproducibility”
• f field measurements.
• State of the art for complex systems with well-

defined principles: model based data assimilation.

• Atmosphere (++), Oceans (+), Biosphere (-)