by Judith Curry
A novel method for causal inference has been motivated by . . . sardines.
Scripps has a press release entitled Predictions of climate impacts on fisheries can be a mirage, with subtitle New mathematical tool developed by a Scripps scientist can help avoid misleading conclusions for species management.
Cause and effect
This press release references a previous press release entitled Scripps-led Team Takes on Centuries-old Cause-and-effect Problem, with subtitle Novel approach distinguishes cause-and-effect from misleading correlation. The press release refers to the following paper:
Detecting Causality in Complex Ecosystems
George Sugihara, Robert May, Hao Ye, Chih-Hao Hsieh, Ethan Deyle, Michael Fogarty, Stephan Munch
Identifying causal networks is important for effective policy and management recommendations on climate, epidemiology, financial regulation, and much else. We introduce a method, based on nonlinear state space reconstruction, that can distinguish causality from correlation. It extends to nonseparable weakly connected dynamic systems (cases not covered by the current Granger causality paradigm). The approach is illustrated both by simple models (where, in contrast to the real world, we know the underlying equations/relations and so can check the validity of our method) and by application to real ecological systems, including the controversial sardine-anchovy-temperature problem.
Sugihara G, et al. (2012) Detecting causality in complex ecosystems. Science 338(6106):496–500. [link to full manuscript]
Excerpts from the press release:
Until recently, scientists had a limited toolbox for detecting causation. Often they would correlate two variables and suggest it implied causation. Yet as long ago as 1710, Irish philosopher Bishop Berkeley remarked “correlation does not imply causation.”
George Sugihara of Scripps Institution of Oceanography at UC San Diego and colleagues from around the world have developed a new approach to help ecologists distinguish true causal interactions from misleading correlations.
Published in the most recent online issue of the journal Science, the method described in the paper, “Detecting Causality in Complex Ecosystems,” extracts the “signature” left by causes embedded in ecological observations—-historical records known as time series. The new mathematical approach deduces causes from the affects.
“The major novelty of this method is that it is based on a dynamic and interconnected view of nature,” said Sugihara, the McQuown Chair Distinguished Professor of Natural Science at Scripps. “Ice cream sales and rates of violent crime might rise and fall at the same time, but our method is able to determine whether this is due to cause-and-effect, or whether both are simply more common during hot summer months.”
The new tool is distinct from methods developed by UC San Diego economists Clive Granger and Robert Engle for financial and economic data, which earned them the Nobel Prize in Economic Sciences. Granger’s technique is aimed at purely random systems rather than those having rules governing how the parts move. Sugihara and his colleagues developed their tool specifically for complex ecosystem analysis, yet its applications could have far-ranging implications across multiple areas of science. For example, “one could imagine using it with epidemiological data to see if different diseases interact with each other or have environmental causes,” said Sugihara.
Excerpts from the paper’s Summary:
Apparent relationships among variables can switch spontaneously in nonlinear systems as a result of mirage correlations or a threshold change in regime, and correlation can lead to incorrect and contradictory hypotheses. Growing recognition of the prevalence and importance of nonlinear behavior calls for a better criterion for evaluating causation where experimental manipulation is not possible.
In resource management, as elsewhere, accurate knowledge of the causal network can be essential for avoiding unforeseen consequences of regulatory actions.
JC comment: About 15 years ago, I was very intrigued by possible applications of Granger causality to the climate feedback problem. My naive attempts at applying GC to climate time series were inconclusive, it is now more clear to me why this didn’t work. This new method looks very exciting to me, particularly since it has demonstrated application to climate related problems. I would appreciate hearing from those of you who are knowledgable about statistics and data analysis as to your take on this method and how we might apply this more broadly to climate impact and attribution analyses.
Back to sardines
Back to the original press release on sardines. The press release refers to the following paper:
Predicting climate effects on Pacific sardine
Ethan R. Deyle, Michael Fogarty, Chih-hao Hsieh, Les Kaufman, Alec D. MacCall, Stephan B. Munch, Charles T. Perretti, Hao Yea, and George Sugihara
Abstract. For many marine species and habitats, climate change and overfishing present a double threat. To manage marine resources effectively, it is necessary to adapt management to changes in the physical environment. Simple relationships between environmental conditions and fish abundance have long been used in both fisheries and fishery management. In many cases, however, physical, biological, and human variables feed back on each other. For these systems, associations between variables can change as the system evolves in time. This can obscure relationships between population dynamics and environmental variability, undermining our ability to forecast changes in populations tied to physical pro-cesses. Here we present a methodology for identifying physical forcing variables based on nonlinear forecasting and show how the method provides a predictive understanding of the influence of physical forcing on Pacific sardine.
Published online before print March 27, 2013, doi:10.1073/pnas.1215506110
PNAS March 27, 2013 [link to full manuscript]
Excerpts from the press release:
In the early 1940s, California fishermen hauled in a historic bounty of sardine at a time that set the backdrop for John Steinbeck’s “Cannery Row” novel. But by the end of the decade the nets came up empty and the fishery collapsed. Where did they all go? According to a new study led by scientists at Scripps Institution of Oceanography at UC San Diego, the forces behind the sardine mystery are a dynamic and interconnected moving target.
What is the impact of climate on sardines? What is the effect of overfishing on sardines? Focusing on single variables in isolation can lead to misguided conclusions, the researchers say.
“Studying ecosystems in this piecemeal way makes it hard to find quantitative relationships, the kind that are useful for management and stand the test of time,” said Deyle.
Instead, using novel mathematical methods developed last year at Scripps, the researchers argue that climate, human actions, and ecosystem fluctuations combine to influence sardine and other species populations, and therefore such factors should not be evaluated independently.
For example, based on data from the Scripps Institution of Oceanography Pier, studies in the 1990s showed that higher temperatures are beneficial for sardine production. By 2010 new studies proved that the temperature correlation was instead a misleading, or “mirage,” determination.
“Mirages are associations among variables that spontaneously come and go or even switch sign, positive or negative,” said Sugihara. “Ecosystems are particularly perverse on this issue. The problem is that this kind of system is prone to producing mirages and conceptual sand traps, continually causing us to rethink relationships we thought we understood.”
By contrast, convergent cross mapping avoids the mirage issue by seeking evidence from dynamic linkages between factors, rather than one-to-one statistical correlations.
“Sustainable sardine fishing based on ecosystem-based management should adapt to dynamic changes in the ocean environment, and future policies should incorporate these effects to avoid another ‘cannery row,'” said Deyle.
From the Discussion:
Scenario exploration with multivariate embeddings motivates the development of new adaptive management schemes based on short-term forecasting. If time series of a human control variable (e.g., fishing effort or mortality) are available, management could be based on scenario exploration with multivariate embeddings that include abundance, temperature, and the control variable. Simultaneously exploring different temperature and harvesting scenarios could then reveal how temperature affects the relationship between fishing and future biomass. This type of scenario exploration would tell managers the permissible level of fishing given a target biomass and recent ocean conditions.
More immediately, these methods offer a data-motivated perspective on the dynamic relationship between sardine and environmental conditions in the CCE. We find that including either SIO SST or the PDO in the multivariate embedding improves forecast skill (ρ) by ∼30%. These indicators reflect environmental forcing of sardine population dynamics that is not captured by more traditional methods. We conclude that environmental considerations should in fact play an important role in Pacific sardine management.
Our results further suggest that good management must reflect the complexity underlying the interaction between environmental changes and population dynamics. . . This corroborates the previous results used in the management plan that warm water is better for sardine. However, our empirical analysis shows that the effect of temperature depends on the specific state of the population. For example, changes in temperature seem to have little or opposite effect in the early years of the time series, when the population is at a much lower abundance, and in later years with very large abundance (e.g., 1999–2001). This suggests any temperature-sensitive control rule for sardine should be different at low, intermediate, and high sardine abundances.
With regards to natural modes of climate variability:
Several other variables show positive ΔF, including the Pacific Decadal Oscillation (PDO), North Pacific Gyre Oscillation (NPGO), and the Southern California Bight (SCB) satellite SST, suggesting these are also relevant to Pacific sardine population dynamics. Of these, only the PDO is significant (P < 0.05). The PDO and SIO SST are highly correlated, so this is not surprising. The method suggests three variables that are least likely important to sardine dynamics: Newport Pier SST, North Pacific Index (NPI), and Southern Oscillation Index (SOI); each of these has a strong negative effect on forecasting.
American Fisheries Society
The American Fisheries Society is a professional society whose mission is to improve the conservation and sustainability of fishery resources and aquatic ecosystems by advancing fisheries and aquatic science and promoting the development of fisheries professionals. They publish Transactions of the American Fisheries Society and the North American Journal of Fisheries Management.
Under their Policy and Media button, they have an article entitled Defining and Implementing Best Available Science for Fisheries and Environmental Science, Policy and Management. From the Introduction:
The report examines how scientists and nonscientists perceive science, what factors affect the quality and use of science, and how changing technology influences the availability of science. Because the issues surrounding the definition of best available science surface when managers and policymakers interpret and use science, this report also will consider the interface between science and policy and explore what scientists, policymakers, and managers should consider when implementing science through decision making.
This paper will
- provide a practical description of the concept of best available science;
- identify the limits to creating, distinguishing, and using the best available science; and
- suggest ways to improve the use of science in policy and management.
To accomplish these objectives, we break the concept of best available science down into the cumulative components of science, best science, and best available science. Throughout the discussion, we highlight the factors that influence best available science, including (a) the changing nature of science, (b) the increasing role of uncertainty in scientific decision making, (c) the influence that the values and ethics of scientists have on the scientific process, (d) the changing availability of information via peer-reviewed journals, gray literature, expert opinion, and anecdotal evidence, and (e) the bridges that need to be forged and maintained between science, policy, and management.
This is an interesting essay and I encourage you to read it; it arguably deserves its own thread.
From the AFS Advocacy Guidelines:
The guidelines that follow were designed to ensure
- that the Society’s external advocacy will be ethically and professionally sound;
- that advocacy will not degrade the Society’s reputation as the most reliable source of scientific information on aquatic resources;
- that fisheries-related scientific information will be used appropriately when members address aquatic resource issues; and
- that AFS advocacy positions will be widely supported within the Society because they will be technically correct, respectful of alternative views, and consistent with AFS policies and the Code of Practices. Members and subunits planning to influence an issue external to the Society and to invoke the credibility of the AFS or its members shall adhere to the Society’s policy.
The AFS has Policy Statement on Climate Change that relates to fisheries, it is fairly extensive (43 pages). It does make the statement: In the interest of sustaining marine fisheries and habitats, the American Fisheries Society encourages immediate reductions in greenhouse gas emissions and implementation of adaptation policies described above for fisheries communities and habitats. And The uncertainty of impacts of climate change on communities and habitats necessitates that work should be carried out within an adaptive management framework where evaluation of policies and management are strong components.
A recent letter to President Obama sent by the AFS on Jan 13, includes 9 recommendations that it states are derived from their policy analyses, the first one stating: Do not delay emissions reductions. Encourage reductions in anthropogenic sources of carbon dioxide and other greenhouse gases.
Further, the AFS has posted a survey for its members about Climate Change and Freshwater Fisheries, that includes an implicit assumption that anthropogenic climate change is generally a threat to fisheries.
JC questions: two points for discussion.
1) Particularly in light of the Scripps research, do you think that the AFS has a strong case that anthropogenic global warming is having an overall adverse impact on fisheries?
2) In the letter to President Obama, do you think the strongly prescriptive recommendation about limiting greenhouse gas emissions from the AFS (a group that does not have much expertise on the attribution of climate change or energy policy) helps or hinders their other recommendations, which are arguably closer to the purview of AFS expertise?
I realize that this is a lot of material for a single post, but the Scripps press releases and the AFS letter landed in my in box at about the same time, and my brain rapidly connected them together.
I think that the two papers from Scripps provide a superb example of use-inspired research related to fisheries whereby a new causal inference model is developed that potentially has much broader applications.
The sardine paper provides a new paradigm for climate impact assessments for complex systems influenced by both physical and human actions. Lead authors of IPCC WGII, please take note. Apart from the new causal inference model, the method includes scenario generation strategies, integration with adaptive management approaches, and consideration of modes of natural climate variability.
Back to the AFS. In light of the Scripps papers, some rethinking of their premises and conclusions regarding the impact of anthropogenic (emissions induced) climate change on fisheries seems to be in order. But this raises a broader issue. Except for the area of climate change, the AFS seems to be quite careful and thorough in its policy assessments. Why has the AFS, and other renascent subfields related to climate change impacts, so wholeheartedly and unquestioningly adopted the assumption that anthropogenic climate change is having, or will have, adverse impacts, often showing more confidence than the IPCC in terms of attribution of the adverse impacts? A corollary assumption is that reducing greenhouse gas emissions will act to eliminate or reduce the adverse impacts. The combination of flawed causal reasoning (mirages), combined with the very substantial uncertainties in climate change and climate impact attribution, makes this a very risky management strategy.
In trying to understand the AFS policy recommendation on greenhouse gas emissions, it seems that the AFS strategy is to rely on the ‘best available science’ on climate change. The IPCC is arguably the best available assessment of climate science for purposes of decision making (at least the WG I Report; the AR4 WG II Report is probably not). I make this statement in spite of the many criticisms that I have made about the IPCC process and many of their arguments. The bigger issue is an assessment of the confidence we should place in the ‘best available science’ for purposes of decision making, in view of uncertainties, ambiguities, disagreement, and acknowledged areas of ignorance; it is in this regard that I think the IPCC falls quite short. The complexities and uncertainties in climate change attribution and future projections are magnified when the focus is regional impact assessments.
To clarify, we have two competing views on how science supports decision making:
- ‘Best available science’ is used to drive an optimal decision making model
- Decision making under deep uncertainty focus on a broad range of possible future scenarios
With regards to the climate change problem, I have made it very clear in previous posts (e.g. see this previous post) why I regard the climate change issue as being characterized by ‘deep uncertainty’ for decision making purposes.
The AFS emphasis on adaptive management strategies for fisheries is a sensible one, but this seems inconsistent with their emphatic “Do not delay emissions reductions.” The U.S. emissions are now at 1994 levels owing to the natural gas boom, something that happened pretty much independently of any federal policy on emissions reductions. The scenario approach mentioned by the sardines paper and that has been discussed many times at Climate Etc. would produce much more robust fisheries management outcomes particularly at regional/local levels and on decadal time scales.
Moderation note: this is a technical thread and comments will be moderated for relevance.