given that in the results we were highlighting the local stability differed mainly due to the generative models, I am wondering if we are suggesting that we should use more than one generative model in the future e.g. predicting food web responses to environmental change? Or we just simply highlight that we need to understand the assumptions and limitations of each generative model?

I think we have to be careful here. Predictions of biomass equilibrum - I assume this means the network-level stability properties and local stability metrics. Because we currently don't have a PTA analysis with these stability metrics, we don't know whether they actually don't converge or how large the differece comparing to the PTA of realised network topologies.

I can add a PTA of these metrics later if you think this is needed.

we can discuss this at next meeting because the variation in reactivity is mainly because of the LTM model, other models seems to have similar reactivity.

functional??

Just realise that we don't need to use stability throughout the ms because I think it is weird to use wntwork-level stability metrics to describe persistence, biomass diversity, flux/interaction strngth distribution.

maybe we could use functional? which included the above metrics, resilience, and reactivity. Stability is one type of functional property. or ?

Yes there is no rewiring in the species interaction links. The feasible network is just a subset of initial interaction network.

I think we may demonstrate more like dynamic redistribution of realised interaction strengths (but not sure whether this is a good term). The initial network sets the initial interaction structure. the initial biomasses and biological rates set the initial energy fluxes and interaction strength distribution. During the dynamic simulation, changes in species biomasses alter energy fluxes and therefore realised interaction strengths. The system finally moves to a feasible steady state where the realised interaction strength distribution is redistributed into a configuration that supports species coexistence.

I understand what this title mean but it is a bit weird to read dynamics constrain structure. I feel we don't need to highlight ecological dynamics, as this may be misleading. Because we focus on food webs at steady state equilibrium, which means there is no temporal dynamics. Perhaps we can highlight that feasible food webs share topological structures independent of generative models??

does this mean these "failed" food webs are mainly due to the topological properties of richness and ChLen? accroding to Figure 1a.

a steady state

Do we need to mention somewhere that the food web cases analysed in Figure 1b are feasible food webs at steady state, corresponding to the equilibrium case in Figure 2

Then we know the convergence in structure is partly due to the pruning from the initial networks to feasible networks and partly due to excluding some fodo web cases, e.g. non-steady state, collapse.

explain S4?

I am thinking about some terminology in the text. 1. Equilibrium. I don't think this is a good term. Our simulation set-up actually pushes the simulation to steady state. The equilibrium case in Figure 2 actually means steady state. The non-equilibrium case in Figure 2 is just non-steady state, but they are at equilibrium; I am looking into whether they are all limit cycles. 2. How do we name the food web at equilibrium. Currently, we use terms like realised post-simulation network, post-dynamics networks. I am wondering whether we name it as feasible food web/networks at equilibrium (thereafter feasible food web/networks), considering the definition of a feasible food web is an ecological network model where the interaction strengths between species, their growth rates, and their mortality rates allow all species in the web to coexist with strictly positive biomasses at equilibrium.

This is generative model.

merge these two together? a bit redundant?

this makes me wonder whether other models allow cannibalism?

if there is a brief explanation of what the generative model mean would be great. Iam not sure whether generative models is a standardised term and widely used?

For me, this is not very clearly understandable. It is the realised networks across generative models that converge in structure. I am not sure whether the food web reconstructive frameworks include dynamic simulations.