Home >> Forums >> Estimation of OSMOSE model parameters

Estimation of OSMOSE model parameters

Dear Sir/Ma'am,
I have the following queries regarding estimation of certain parameters before we initiate the model run.
Query 1
Are there alternate methods (other than calibration) to estimate "population.seeding.biomass.sp#" parameter which is used to initialize the model? If so, could you kindly redirect me to resources (scientific literature, user guides, weblinks, etc) that deal with estimation of "population.seeding.biomass.sp#" parameter?
Query 2
From the documentation, I have understood the role played by predation acessibility coefficients in OSMOSE model run. I would like to know as to how the default values (such as 0,0.8,1, etc) were estimated for predation acessibility coefficients. Could you provide insight as to what should the end-user be absolutely aware of while estimating values for predation accessibility coefficients?
Query 3
Given that we have information(regarding the size range) about the species along with their diet, how should the end-user go about estimating the minimum and maximum values for predator to prey size ratio (since a single predator can have multiple prey species and vice-versa)?
Query 4
Why do we have the default value for the parameter "mortality.subdt" set to 10 in OSMOSE v3 update 2 configuration?  While configuring the model for a given set of species found within a particular study domain, could you provide insight as to how should we go about while deciding what value should be assigned for "mortality.subdt" parameter?
Query 5
How do we go about estimating the  value for   "mortality.starvation.rate.max.sp#" parameter? Could you share some information (eg. scientific literature) that can shed more insight with respect to the estimation of this parameter?
We deeply appreciate the time and effort that OSMOSE team has spent to go through our queries and clarify our doubts so far. And we look forward to your input with respect to the following queries mentioned above.
Thanking you,
Kind regards,
Tarun Joseph

Dear Tarun Joseph,
I am an OSMOSE user, and hope my answers could help you.
This parameter refers to the initial biomass of the species in the system. As far as I know, there were no available formulas for calculating it. In my opinion, you could set it according to your survey data at first, and then, you could also estimate it by OSMOSE’s calibration within a reasonable range.
The predation accessibility coefficients is the percentage which the prey is accessible to the predator, and is not used to define diet preferences. It is up to the characteristic of your study area to set this parameter. In my project, my principle of setting this parameter is that all the preys are accessible to their predators by 0.8.
As we know that, OSMOSE model assumes opportunistic predation based on spatial co-occurrence and size adequacy. And therefore, I think the predation.predPrey.sizeRatio is the key to the progress of predation in OSMOSE, and the main function of predation accessibility is to take into account for a difference of positions in the water column.
This parameter is set according to the actual data which were from your survey or other studies.
It is a technical issue, and I think Philippe is able to answer it.
I set all species’ mortality.starvation.rate.max to 0.1 in my project. I think that this parameter is not the key to OSMOSE, and there was no need for estimating it. It is just the max mortality rate of starvation.
Happy modeling.
Best regards,
Lei Xing
Ocean University of China

Thank you Lei Xing for your answers that I'll try to complement.

Q1. I provide some detailed information about the seeding approach in the Osmose 3 Update 2 release note, section 6 "Population initialisation". The parameter seeding.biomass does not have any real ecological or biological meaning. It reprensents the guaranteed Spawning Stock Biomass (SSB) during the spin-up time for every species, i.e. if there is no mature adult at a given time step, Osmose will consider that they are virtually $(seeding.biomass) tonnes of adults to ensure the reproduction process. That way the population grows from scratch and the dynamics should stabilize after a few years.

I doubt you will find any litterature as it is an Osmose intrinsic parametre and concept. But given this meaning of "guaranteed SSB", the target biomass of the species could be an acceptable estimation. If you have a little time ahead of you, it could be very usefull to do a kind of sentivity analysis of this aparameter, to see whether the system reaches an equilibrium no matter what levels of seeding biomass, how long it takes, etc.

Q2. As explained in the documentation, the predation accessibility accounts for the overlapping of the species (at different stages if needed) in the water column. Some species may be in the same cell but always at different depth levels so accessibility is zero. On the contrary two species may share the same habitat in the water column so accessibility is one. Some mesopelagic species may spend half of the night at surface and the day deep down, so the accessibility for pelagic species will be 0.5. Accessibility 0.8 means that 20% of the schools of a prey are not available to a given predator because they do not share exactly the same distribution on the water column for instance, or for any other reason. This is the kind of consideration you should take into account when setting up these coefficients. And as Lei Xing underlined, it must not be used as diet preference.

Q3. For setting up the predator/prey size ratios I reckon I would merge/average the information you have for every prey species, chooses the smallest preys and the largets preys indistincly of the species and calculate the ratios. Osmose does not allow any finer parametrisation, at prey species level. Nonetheless the information on diets will be very usefull for checking out whether Osmose diet outputs are consistent with observations. If not then you should consider to alter some of the ratios and predation accessibilities.

Q4. Indeed the mortality.subdt is an other intrinsic Osmose parameter, related to the mortality algorithm. We set it to 10 because we noticed empirically that it provides enough stochasticity to avoid any trend with the succession of the mortality sub-processes (predation, starvation, fishing, additional) and with the succession of the predation events (predators are drawn randomly at every sub time step). Secondly the computational cost is acceptable. An in-depth sensitivity analysis is missing but I invite you to play around with the parameter and check what kind of effect it has on mortality rates and diet outputs ? At some point increasing the subdt should have no impact on the results, then you know that you found the appropriate value.

Q5. I just realized the parameter is missing in the documentation... sorry. I have no idea how to estimte this parameter. I must ask Yunne, she set it up and we used it since then as a default value. As LeiXing said though, this is not a crucial parameter but you'd better be sure whether the default value is suitable in you case. I'll ask Yunne and let you know.




Thank you Lei Xing and Philippe for your input.
Kind regards,
Tarun Joseph