Week 6 Discussion Questions — 7 Comments

  1. Predictability of catastrophic events: Materialrupture, earthquakes, turbulence, financial crashes, and human birth
    Didier Sornette† PNAS 2006

    On p 2522, Sornette attempts to extend the idea of ruptures to social and personal events: “Political crises and revolutions shape the long-term geopolitical landscape. Even our personal lives are shaped in the long run by a few key decisions or events.” He then states that in order to predict catastrophic patterns, “the rules for the interactions are presumed identifiable and known.” But can this really be true of human ideas and aspirations (which presumably play a part in revolutions)? What are the ethical and social implications for this kind of “historical determinism”?
    What exactly is meant by “a discrete hierarchical structure with a preferred scaling ratio”? (2526, and Figure 2). More broadly, I am confused about how, in practice, one goes about combining statistical physics and dynamic systems theory (Conclusion).

    Early-warning signals for critical transitions Marten Scheffer1, Jordi Bascompte2, William A. Brock3, Victor Brovkin5, Stephen R. Carpenter4, Vasilis Dakos1, Hermann Held6, Egbert H. van Nes1, Max Rietkerk7 & George Sugihara8
    Complex dynamical systems, ranging from ecosystems to financial markets and the climate, can have tipping points at which a sudden shift to a contrasting dynamical regime may occur. Although predicting such critical points before they are reached is extremely difficult, work in different scientific fields is now suggesting the existence of generic early-warning signals that may indicate for a wide class of systems if a critical threshold is approaching.

    In Box 1, what is the difference between plots c and d? That is, what is the difference between catastrophic shifts and multiple stable states, and how could that difference be detected in data?
    “Flickering” and “slowdown” are advanced as two detectable signals for imminent bifurcation. In real (contemporary) environmental systems, where the data record is often short, how can we determine whether a particular regime does represent deviation from a longer term stable state? (They discuss some of these difficulties in terms of data analysis, but not from a data collection perspective).
    Cross-scale interactions, nonlinearities, and forecasting catastrophic events
    Debra P. C. Peters*†, Roger A. Pielke, Sr.‡, Brandon T. Bestelmeyer*, Craig D. Allen§, Stuart Munson-McGee¶, and Kris M. Havstad*
    The real-world fire example in this paper complicates the cellular model we saw last week, by introducing regional-scale feedbacks such as pyrocumulus cloud formation. In what cases do these large scale feedbacks trouble the findings from the modeling experiments?

  2. Sornette contends that rare catastrophic events control long-term behavior. Do you think this is the case? What are some mechanisms that would support/refute this claim?
    How is analysis of flickering toward prediction of catastrophic shifts useful in a scenario (such as climate) where there is no historical precedent for the forcings? Does the catastrophic shift analysis still apply?
    Scheffer et al. state that shifts due to “fast and permanent change of external conditions cannot be detected from early-warning signals.” In the context of historically unprecedented anthropogenic climate forcings, can catastrophic shifts within broader climate change still be predicted?
    What are the implications of false positives and false negatives? How does this impact the practice of catastrophic shift prediction?
    What are some examples of how early-warning signals can be useful for indicating proximity to different kinds of thresholds (not just catastrophic shifts)?
    How does an understanding of a system as a SOC or HOT system influence prediction of catastrophic shifts?

  3. Which is the probability that anormal climatic events (e.g. extremely warm season) is “quiescent state of crisis” or an “early warning” signal of climate change?

    Sornette. Clarification: Could you explain what does mean “the failure of compositive system may often be viewd as a result of correlated percolation process”? p2523.

    Which would be a realistic approach of connectivity in ecosystem in order to avoid spread of contagious events, and promote the natural dynamics of population? For example avoid the spread of fire events and in the same time avoid the isolation (genetic process) of biota.

    Which will be the early warning signals of world food production in modern agriculture within a system that is controlled by global markets, agroecosystems (highly dependent agrochemicals), and climate change?

    Which is the risk of the hydraulic fracturing methods to extract natural gas in from a shale to cause a collapse of the whole watershed? Could hydraulic fracturing methods have an early warning signal as a human caused event?

  4. 1. How far have simulations gone in the prediction attempts, and what are the main constraints? Coming from astronomy, it’s generally computational power that limits simulations, and the complexity of the simulations seems like it would be similar to those in astronomy.

    2. What is it that drives the rupture instability that Sornette mentions? I can conceptualize stress causing the ruptures he mentions in physical systems (weakness increasing over time in materials), but are such stressors universal? For instance, how would this apply to forest fire prediction?

    3. Regarding forest fire predictions, is Scheffer’s discussion of critical transitions and predictive observations of spatial organization comparable to the patterns we saw in the HOT model that we discussed last week?

    4. Although we can observe “flickering”, how well can we constrain the timescale of catastrophic events? It seems that we can foresee that they are coming using these predictive techniques, but how difficult is it to know when this transition occurs?

    5. It seems as if the Peters paper is purely a suggestion that we need to look at multiple factors when accessing system stability. How is this paper actually useful in changing the analytical methods utilized in studying complex environmental systems (it seems to me that it is self-obvious that you need to take into account more complexities if possible, and that we reduce systems only when systems appear too difficult to understand at more complex levels).

  5. Predicting catastrophic shifts (Week 6–2/26/13)

    1. Sornette (2002) suggests a new approach to prediction using log-periodicity, a series of oscillations that become more and more rapid before a catastrophe. During the last two centuries all major financial indices have advanced at a faster-than-exponential trend rate, as have population and GDP growth. The log-periodic model implies that this trajectory is unsustainable and that the world economy, may cease to grow or even slump heavily in the middle of the next century. Should we be worried?

    2. Given the risks and remaining difficulties involved in predicting earthquakes and stock market crashes, is Sornette’s (2002) predictive approach useful when it is “probably inaccurate in details”? Even Sornette does not recommend actually using his earthquake prediction model.

    3. Scheffer at al. (2009) discuss spatial patterns as early warning signals. They present the classic example that ecosystems may undergo a predictable sequence of self-organized spatial patterns as they approach a critical transition in the case of modeling the response of semi-arid vegetation to increasing dryness of the climate. Before a critical point, the patches are spotted, not maze-like. How do we know the pattern is more influential for prediction than the decreasing area of vegetation cover itself?

    4. If scientists propose an early warning signal for a particular type of catastrophe (e.g. earthquakes), how might people react to false negatives, situations in which a quake occurred but no early-warning signal was detected? How might they react to false positives, in which the signal of impending earthquake is wrong? Which error is worse?

  6. 1. In Rietkerk et al 2007, are the distinct spatial patterns of self-organized patchiness in arid ecosystems a signal that the system has strong and dominant drivers, water and evapotranspiration? Why do some systems exhibit strong spatial patterns versus time-series patterns? How much of these differences reflect our ability to measure/detect patterns rather than distinct characteristics of these systems? For instance, Sornette’s paper is time-focused, but both climate and earthquake phenomena seem spatially oriented. His treatment of space doesn’t seem as well-developed as time. Are our spatial detection tools/methods less well-developed in general? What are they: spatial statistics, landscape ecology, network topology?

    2. In Scheffer et al 2009, does the critical slowing down, increased ‘memory’ and autocorrelation result in a system that seems more stable prior to the catastrophic shift? Does this imply that the depth of attractor basin changes as the system approaches a threshold (from deep to shallow)? Could this ‘critical slow down’ stage then be easily mistaken for a new stable state without sufficient time-series data — except that it has low resilience, so there’s a high risk for a sudden, major shift? Are there any examples of this? For example is the relatively stable climate of our recent past an indicator?

    3. In Scheffer et al 2009’s discussion of desertification in Kefi et al 2007: how do vegetation patch size distributions “lose their scale-free structures and become characterized by truncated power laws?” How can patch sizes be scale-free or truncated power laws — I’d like to visualize somehow. This seems like a good example of spatial analysis that Sornette lacked.

    4. In Peters et al 2004, should I infer that prior to each threshold, there’s probably some combination of autocorrelation, log-oscillation, flickering, recovery slowing and/or increased spatial homogeneity that may be detectable and can help predict the catastrophe. The argument for forest thinning or controlled burns seemed contradictory to this model (local processes can be overcome by more broad-scale drivers), but the cross-scale interactions were well-described with examples. It made non-linearity more accessible, but is it overly simplified given the discussions in our other readings? It seems like predicting between stages 3 and 4 is relatively useless (and less well-defined in this model) — even if you could detect, there’s not enough time to do anything about it. But the cross-scale interactions do seem critically important, is there more to this or a way to more explicitly integrate this perspective into models/behaviors described in other papers?

Leave a Reply