Home >> Forums >> How to assess the stability of my configuration ?

How to assess the stability of my configuration ?

Dear Sir/Ma'am,
Given that the calibration is done for the year 1980 and let's say I ran the OSMOSE model for 30 years (time-step for the model = 1 year). Let's consider that the model attained stability from time-step=25 onwards. Could you help me in clarifying the following queries?
Query 1:
Is it posssible to see if OSMOSE has stabilized (in real-time) when it's actually running the simulations? Or do we have to wait and finish the simulation and check the output in order to identify the time-step at which the model has become stable?
Query 2:
Let's consider that I selected the OSMOSE model simulated values for biomass/catch corresponding to time-step=25. In order to verify if the stabilized biomass/catch values are well within the expected biomass/catch values, should I compare the simulated values (corresponding to time-step=25) with the observed biomass/catch values for the year 2005 (since the starting year is 1980 and time-step= 25)?
Or does it mean that the simulated biomass/catch values corresponding to time-step at which the model attains stability should be compared with the observed biomass/catch  values of the year for which the model was calibrated (in this case 1980)?
Please do correct me if I'm wrong.
Kind regards,
Tarun Joseph

Hello Tarun,

There is no other way to check whether the model has stabilized than to wait and plot the outputs at the end of the simulation. You could always write your own script, in R for instance, that plot in real time the species biomass and catches since the Osmose output CSV files are updated on the fly, but I am not sure it is worth doing it. Stability of the Osmose model should be evaluated over a few years, even decades : you want to make sure that the biomass remain between acceptable intervals over the years, not just that it matches a certain value at a given time step or a given year. The trend may be more important than the absolute value at first.

Osmose needs a numerical spinup : it is the time (let's say 20 years or 25 years, but it really depends on your configuration, at least the time of the longuest lived species to get old) that the species need to populate the simulation domain. This numerical spinup does not mean anything biologically speaking. Once this spinup period is over, Osmose must run for a few more years or decade and if stable the biomasses and catches should stay within acceptable limits.

If you calibrated Osmose with year 1980, you should then run Osmose for 40 years. Let's say that the first 25 years correspond to the numerical spinup, then the last 15 years should have biomass and catches matching on average the year 1980. That is what we usually do for the steady-state calibration. Interannual calibration is not ready yet for public release, it is work in progress.

When you are running such simulations, what do you use as low trophic level compartments ? Is is annual time series or do you loop over the same year over and over ?

Cheers, Philippe