Federal report fails to stanch science debate

12/01/2011

Last Wednesday, at the height of rush hour traffic and just over 24 hours after the release of a federal scientific report, which in part supported the claim by Point Reyes National Seashore scientists that a positive correlation exists between mariculture output and harbor seal declines in Drakes Estero, Seashore spokeswoman Melanie Gunn announced on Bay

Area public radio station KQED that, often in the field of ecology, “correlation is as good as it gets.”

Earlier that day, in response to a question about the park service’s correlative—but not causal—conclusions, Gunn explained that the theory of climate change is essentially the same as the theory that oystering in the estero has displaced seals in recent years—people who believe the earth is warming through anthropogenic means do so without any causal evidence.

But Gunn’s comparison is contested by at least one climate expert. Dr. Peter H. Gleick, a member of the National Academy of Sciences, a MacArthur fellow and the director of the Pacific Institute, explained that the Seashore’s findings are, in fact, quite the opposite from those of climate change.

“There are natural factors that affect the climate and there are human factors, and in the climate debate we see very clearly that natural factors alone do not explain the results we are seeing,” he said, adding that there have been thousands of studies—examining thousands of variables—conducted on climate change, compared to the Seashore’s single study, which examined just one of 11 known potential sources of seal displacement. “The [Seashore’s] statistics are absolutely unable to show that the human factors have played a role.”

Meanwhile, even the correlative finding remains in dispute. In a letter sent this week to commission director Timothy Ragen, independent biologist Dr. Corey Goodman and David Lewis, of the University of California Cooperative Extension Marin, claimed the Marine Mammal Commission’s report erred in accepting the Seashore’s correlation, which they alleged was and remains innately flawed.

Specifically, they alleged that Ragen had incorrectly dismissed their 140-page analysis of the Seashore paper (commonly referred to as Becker 2011) on the grounds that it was confounded by statistical errors.

“Was our analysis fatally flawed?” they wrote. “The answer is no.”

Goodman and Lewis explained that their model could be easily modified to address Ragen’s technical concerns, and that the modified version “leads to the same conclusions”—that a catastrophic event in 2003 at Double Point, a nearby harbor seal habitat, and the natural ebb and flow of regional seal populations are the likeliest proxy for fluctuating counts in Drakes Estero.

They also stated that, despite previously agreeing to, Ragen never raised his questions or concerns to them regarding their analysis. “Between August 29 and November 22, neither Goodman nor Lewis were asked by [Ragen] a single statistical question, asked for a single clarification, or asked for additional information,” they wrote. “The process became closed.”

Goodman and Lewis’ analysis had asserted that Becker 2011’s findings were leveraged by the 2003 event—when a marauding elephant seal killed dozens of harbor seals and drove hundreds more into the estero for temporary refuge—and that when that year’s data is eliminated from the paper’s overall analysis the statistical significance of a correlation between oystering and seal counts becomes null and void.

But Ragen, writing in the commission’s report, said he did not request that Becker 2011’s authors re-run their models without 2003 because no one had provided “legitimate justification for doing so.”

“You can go through any statistically significant relationship and pick out certain data points that, if removed, would reduce the significance of the relationship,” he wrote.  “However, if there is not a specific reason for doing so, then that is cherry picking.”

Ragen, relying on the advice of marine scientist Dr. John Harwood, of the University of St. Andrews, wrote in the report that he requested Becker 2011’s authors to adjust for the 2003 event, rather than eliminating it altogether, and that a positive correlation remained nonetheless.

But rather than subtracting the 2003 event from their correlation, the park service scientists added another independent variable that was leveraged by the 2003 event, Goodman and Lewis wrote. “Instead of effectively testing 2 – 2 = 0, NPS in essence tested 2 + 2 = 4.  MMC accepted an incorrect test.”

They added that eliminating individual data points to test for leveraging outliers—elements that unduly influence an outcome—is basic statistical protocol, and that the adjusted Becker 2011 compounded, rather than lessened, the impacts of the 2003 event.

Prior to the publication of Becker 2011 this spring, the commission’s six-member panel of marine experts expressed concern about the study’s very parameters. In his individual report, marine scientist Brian Kingzett questioned why Seashore scientists had only examined mariculture as a source of disturbance in the estero. He noted that in a 2010 presentation, Becker 2011 co-author Dr. Sarah Allen claimed that disturbances from oyster boats accounted for just four percent of total observed disturbances in the estero.

“The use of a priori hypothesis to test for correlation between aquaculture activities and seal population without also incorporating alternate hypotheses seems irresponsible,” Kingzett wrote. “Especially in light of other potential factors such as the much larger number of observed anthropogenic disturbances listed by Sarah Allen in her presentation and the influence of larger oceanographic factors in the California nearshore environment...”

Even Gleick, after examining Becker 2011’s analysis, alleged that it stretches so hard to show a correlation that it’s unlikely one actually exists. “It’s so far from causality…,” he said, failing out of frustration to finish the sentence. “It’s just bad science to argue that correlation is causality, and it’s not even based on good correlation.”