What is POME?
What is supported?
[Course list of University of Bergen]
[Course list of Dalhousie University]
POME exchange: Stay at Dalhousie University, Halifax
Name: Cara Nissen
Study program: University of Bergen, Joint Nordic Master’s Program in Marine Ecosystem and Climate
My master’s thesis will be supervised by Corinna Schrum (University of Bergen) and Katja Fennel (Dalhousie University) and will deal with ocean acidification in the North Sea and Baltic Sea. The goal is to investigate the impact of different biological processes such as production, mineralization, denitrification and sediment water exchange on the carbonate system. The model to be used for the thesis is the ECOSMO, a coupled biological-physical model described in Schrum et al. (2006)  .
To assess the importance of different processes and optimize the parametrization of for example alkalinity in the model, the so-called polynomial chaos expansion (further related to as PCE and described in Mattern et al. (2012)  ) will be applied. Jann Paul Mattern, currently postdoc in the Marine Environmental Modeling Group (MEMG) at Dalhousie University and former PhD student of Katja Fennel who is leading the above mentioned group, developed several Matlab tools to apply this method on biological-physical models.
During my first stay in Halifax – a second, three-months long stay is planed for fall 2013 - I was supposed to get introduced to the method and the tools to be able to apply them on the ECOSMO once the thesis work starts in summer 2013.
Generally, the PCE allows to quantify the uncertainty of a model output based on the uncertainty of the input parameter(s) it is relying on in a computationally efficient way. The idea of the PCE is to assume one/several of the input parameter(s) to be uncertain (thus, assign it/them a distribution with a range of values it/they can adopt), and to investigate the progression of this uncertainty into the model’s output. The distribution of the input parameter is sampled and the model is run at the so called quadrature points, the number of these is chosen individually according to the analysis. Having model output at the quadrature points, the output is then interpolated in parameter space according to the distribution chosen for the uncertain parameter. If one wants to do a sensitivity analysis, optimize the value of the uncertain parameter in the model or even add a time-dependency of the parameter into the model, one can easily compare the obtained output data to observations and find the best fit by doing a least distance analysis.
To get used to the method and the Matlab functions Paul has written to carry out this analysis I started applying the method on a simple 1D-NPZD Matlab model that was developed in the MEMG. With this model I could investigate the effect of different input parameters (e.g., maximum growth rate of phytoplankton at 1° C) on the phytoplankton abundance in the water column at different times during the year. I started assuming only one input parameter to be uncertain (max. growth rate) and added another one (max. grazing rate) after familiarizing with the method. I then proceeded to a test case of the ROMS (Regional Ocean Modeling System  ), the wind-driven upwelling/downwelling over a periodic channel test case. Paul helped me to setup the automatic functions to automatically do model runs at the quadrature points and interpolate the output. By the end, I managed to look at the influence of different input parameters on the phytoplankton abundance in ROMS.
Everybody I met at Dalhousie University was very welcoming and supportive when dealing with administrative and technical issues of my work. I was very involved in activities in the MEMG and the whole Department of Oceanography from the beginning. I participated in group meetings, went to seminars and to the weekly held ‘student coffee meeting’ which serves to put students at all academic levels in contact.
I'd also like to comment on a few general things of my stay:
Because it appeared to be impossible to organize private accommodation for such a short term
stay, I was staying in the Lord Nelson hotel for the two weeks of this first stay in Halifax. It is a
really nice hotel, which offers special prices in winter and for extended stays.
In a nutshell, I have to say that I really enjoyed my stay at Dalhousie University. I am definitely looking forward to coming back in the fall and can only encourage students at both universities to apply for an exchange within POME.
More reports to come!
Last updated: January 2013
Webmaster: Cara Nissen
Copyright © 2013 POME programme