Talk:Complex socio-ecological systems/Networks

From Wikiversity
Jump to navigation Jump to search

I find the networks topic VERY interesting, but I had trouble to relate it to resilience and our previous conversations. What I have read and heard on social networks makes a bit more sense since it incorporates explicit social institutions and dynamics (like the Swedish paper). Enjoy the discussion and look forward to read the summary and comments, claudia128.227.83.224 23:12, 17 February 2011 (UTC)

Barabasi et al. (Science 1999)[edit source]

I started with Barabasi et al.'s Science paper, and I have to say I like it. There are obvious over-simplifications in their model, which the authors' point out in their concluding remarks, but the paper represents a succinct but major advancement in our understanding of networks. That growth and preferential attachment trigger the emergence of self-organization is, well, very cool. It makes sense that Poisson distributions were initially used, particularly when we consider the shapes of these distributions at low lambda, but I can see the limitation in the assumption that all the actors within the network are initially present; even in ecological systems, networks will evolve over time, with more components entering system with advancing time steps. The paper did make me think more theoretically about models, however. The authors created a model that I think had no choice but to form a power law distribution when they introduced the probability terms for vertex connectivity. It's a deterministic process, so it brings to my mind the question of whether a theory is (better) proved because a deterministic model can create a pattern as observed in nature, or whether a non-deterministic model that exhibits emergent behavior without rules is better proof of theory?

Bascompte and Jordano (Annu. Rec. Ecol. Ecol. Syst. 2007)[edit source]

I should start by stating I really enjoyed the review. It follows from the preferential attachments described by Barabasi’s Emergent of Scaling in Random Networks (Science 1999), so I’m very glad I accidentally read that one first. I find it very interesting that the power law distributions observed in many self-organizing systems are (may?) be scaled, in this case truncated, by a decay function. I wonder if all multi-modal networks have some sort of scaler. The decay function described here is an exponential function, and there is some critical number of links where if a network approaches that critical number (i.e., the exponential decay rate approaches 1), the probability of more connected species drops off. If I am interpreting this correctly, this means that the more connectivity there is among components, the less likely there is to be yet more species with high connectivity. This makes intuitive sense to me from niche occupation; there are only so many niches available to highly connected generalists that any system can contain. The authors lay out a number of very good arguments why interactions and species abundance cannot be random in section 4, and for why truncation occurs, where some connections are ‘forbidden’ because they simply cannot happen for phonological or morphological reasons. The power-law distribution need not be truncated only for reasons related to the preferential attachments; if a system starts out with a larger number of species to form a network, truncation also occurs.
To paraphrase the section on dependences and asymmetries, niches arising from the action of generalists are less susceptible to variation, and by the same token, generalists can more easily survive in abiotic (more highly variable) environments because of an inherent invariability due to several reasons. Essentially, generalists exhibit greater resilience. Yes! We knew it would come back to that. Further, mutualism confers greater nesting of networks than does consumerism/predation, which makes sense, since consumer/predation relationships are dead-end for at least one member of the interaction. However, I wish the author had expounded more on the “asymmetry may help interdependent groups of species coexist… This kind of downward loop is less common in uneven relationships because the plant could recover by relying on a generalist partner…” (pg 577). There are systems that exhibit the classic predator/prey oscillations (the classic timber wolves and snow hares) that, despite the amplitude of species abundances, have certain characteristics of stability. Both species may be generalists in some ways, but the prey definitely makes up the greatest part of the predator diet in these cases. Does the fact that the communities recover run contrary to the concept that asymmetry confers greater resilience in a network? Or, are the oscillation lags in those predator/prey relationships created by prey switching, which would mean asymmetry? We usually think of those lags being causes by differences in generation times between the two species. Or, are the networks wherein there is a predator/prey abundance oscillation like the timber wolf/snow hare example less complex, so that the usual generalist behavior of the wolf isn’t possible?
That phylogeny may play an important role is laid out, but not convincingly. An observation that may occur 30-50% of the time is not the same as discovering a rule that governs network behavior. That there might be secondary network controls created by phenology and other ecological factors makes sense, but the science hasn’t gotten to the point of really explaining network structure based on these factors. This brings up the question of if we don’t have any organizing rules for phenology and ecology, are species degree and abundance the correct measures for network behavior, or are the proximate measures for the true, ‘correct’ measure? Ecology is frequently hampered by figuring out the best measure for some system property.
The implication that generalists give networks resilience and the loss of generalists initiates a faster collapse of a network has huge implications for the way science talks about biodiversity. Policy and land management has continually focused on giving more conservation value to the rarest, most specialized species. The sole argument is that these species create biodiversity and their rarity (and general status as most-likely-to-go-extinct) is what gives them an intrinsic value above other species. The management of these species is in some (perhaps many) cases to the detriment of the stability of more generalist species and to the network they are all nested within (the management for the cape sable seaside sparrow being a clear example).

DWatts 21:58, 16 February 2011 (UTC)

I agree that conservation biology tends to manage for specialists, and that in systems terms these may not be critical to system functioning. If you look at the red list of endangered species, a large majority are from islands or otherwise have very restricted natural home ranges.

Liljeros et al. (2001) Interesting article. However, I did not understand some important concepts that the authors use. Specifically, I don't understand what is the "power law decay" and how is a scale free network different than other types of networks. flaleite 18:54, 17 February 2011 (UTC)

Power law decay is the mathematical shape of the curve that is characteristic of scale free networks, i.e. the straight line in Figure 2 a and b. One way to think of scale free networks is that a few nodes have many connections, while most nodes have fewer connections. In a random network, there would be an average number of connections and just as many nodes have above-average number of connections as have below-average number of connections, ie. a normal distribution. In scale-free networks, you have a few hubs with many connections.

General questions[edit source]

1. What is the significance of the observation that widely divergent types of networks (WWW, social, cellular) have similar structural properties?

2. Several of the papers look at the structure of networks at one point in time (static). The Palla et al paper looks at how network structure changes over time. Is this an important advance? What are the methodological challenges to such dynamic analysis, and how can they be overcome?

3. How do human and ecological networks interconnect?

4. What characteristics of their human and/or ecological networks makes systems more resilient?

Note about tomorrow's seminar[edit source]

We will talk about the above questions and points brought forth by Danielle and Flavia, and hopefully others if they should find the time to do so. I would also like to plumb the depths of question 3 by demonstrating the difference between whole networks and personal networks. All of the readings this week describe whole networks, but one difference between social and ecological networks is clearly the scope of the personal (egocentric) network. You cannot, at least with modern technology, ask a hummingbird how many flowers it pollinates, not to mention the specific attributes of said flower. You can do that with humans. Please also read the short primer on SNA that Allison Hopkins prepared.~Sam