kakearney
commented
8 months ago 1) If I understand correctly, you should be able to generate the ensemble of Ecopath models following the standard ecopathmodel workflow (see html/ecopathmodel_overview.html in this repo for an example). Whether iteratively widening the uncertainties (by increasing the pedigree values associated with each) will give you the sensitivities you're looking for will be highly dependent on your specific ecosystem data, and how much wiggle room each parameter has before meeting other constraints that affect the overall mass balance.
2) I'm not sure I understand this question. In the current code, the value+pedigree combo is translated into either a uniform, lognormal, or triangular distribution of parameters based on the 'pdfname' input chosen via the |createensemble| method. The actual distribution of balanced parameters will be unique to your particular dataset. Once you have an ensemble, you can calculate uncertainty statistics (e.g. standard deviation, percentile ranges, etc.) across those parameters, or any downstream calculations (like network indices). Is that what you're looking for? You could also look at patterns within the resulting ensemble (e.g., PB of critter A vs a network index); in that case it would likely make sense to run targeted sensitivity studies where you only applied pedigree values to one or two parameters at a time.
Johnme24
I get confused about utilizing a Monte Carlo simulation method by conducting "Ecopath_matlab" . I want examine the robustness of the indicators against parameters (B, P/B,Q/B,and EE) uncertainty and each parameter was simulated with a sequentially increasing lognormal uncertainty for corresponding parameters from 0% to 100% at intervals of 10%. The idea was derived paper I read recently, as: Evaluating the impacts of El Niño events on a marine bay ecosystem based on selected ecological network indicators (https://www.sciencedirect.com/science/article/abs/pii/S0048969720377366). Would you show some suggestions on : 1)how to excute the process above； 2)how to quantify the uncertainty level of the parameters sampled, particularly with respect to their pedigree? Much thanks!