FPC Memo Defends
by Bill Rudolph
An Oct. 7 memo from the Fish Passage Center to Ed Bowles, fish division administrator at the Oregon Department of Fish and Wildlife, contains comments that are critical of the draft BiOp and NOAA's decision to nix a proposal to boost spill levels being pushed by the state of Oregon, along with fishing and conservation groups.
The feds based their decision in part on a BPA-funded report by University of Washington researcher and statistician John Skalski and two co-authors that raised serious questions about the analysis used by spill advocates to press their case, which was presented at the September meeting of the Northwest Power and Conservation Council.
The CSS spill analysis, led by U.S. Fish and Wildlife Service scientist Steve Haeseker, was developed by members of the Comparative Survival Study, an ongoing project that analyzes survival of PIT-tagged salmonids. It claims that smolt-to-adult return rates could rise to recovery levels of around 4 percent if spill levels at federal dams were bumped up enough to produce total dissolved gas levels of 125 percent in tailraces. Such an increase would be 5 percent higher than the 120-percent level now allowed by state waivers to the Clean Water Act, and would push the spill percentage into the 50-percent range.
The Skalski paper investigated the likelihood of spurious correlations that could have affected the CSS analysis, a possibility the center simply cannot accept. The FPC memo took the Skalski paper head on. "Although these statistical treatments may be impressive to some, the basic approach suffers from logical errors and does not provide a convincing argument against the Haeseker et al. (2012) analyses that documents the benefits of spill at multiple life stages," said the memo.
Skalski suggested the correlation between higher spill levels and better adult returns was highly suspect, since the spill percentage also correlated with increased adult returns of transported fish, which obviously didn't benefit from it, and which should not show a positive correlation if Haeseker's hypothesis of delayed mortality was correct.
The FPC likened Skalski's reasoning to old arguments used by the tobacco industry. "Skalski et al. (2013) selects a variable that the subject population is not exposed to, in order to illustrate that correlation is not causation. This is a fundamental princip[le], but good princip[le] can also be used to buttress bad arguments (Gould, 1991). This is similar to historic arguments regarding smoking and the occurrence of lung cancer in which Ronald A. Fisher (preeminent statistician and paid consultant for the Tobacco Standing Committee) argued that the considerable body of decades of data that showed a significant correlation between smoking and lung cancer did not establish causation. On that basis, legally required warnings and recognition that smoking caused lung cancer was delayed for decades at obvious costs."
But the Skalski paper explained how the relationship between spill percentage and fish survival may be misconstrued because of other fresh water variables that correlate with better ocean survival. It pointed to a better indicator known as the Pacific Northwest Index [PNI], which contains air temperature, precipitation, and snow pack data.
"This index was significantly correlated (P < 0.10) with 3 of the 5 inriver covariates and the PNI," the paper noted. "Flows at Bonneville Dam were correlated with water temperature, turbidity, and [water transit time], but not percent spill. The reason for the latter is that spill targets are set pre-season, based on policy and altered only during flow extremes."
The Skalski paper also found little evidence of delayed mortality in returns of transported fish. But they did find fish length (as a surrogate for fish condition) correlated well for SARs of transported fish, and once it was included in their own analysis, no freshwater covariates were significant.
"If large scale climate variation affects fish condition during both inriver and ocean stages," said the review, "then the apparent effect of freshwater environment on SARs may actually be due to the correlation between fish condition and climate indices."
The FPC memo points to Skalski's analysis of the dataset of length for wild and hatchery smolts marked and transported at LGR that found a correlation between smolt length and offshore upwelling.
"The authors seem to be suggesting that the lengths of wild and hatchery fish are somehow related to nearshore ocean processes, which defies logic and conflicts with Skalski et al.'s (2013) earlier point of causation and correlation."
But what the FPC doesn't acknowledge is that when ocean conditions are generally favorable -- which includes increased upwelling -- freshwater conditions are generally more favorable to juvenile fish as well, which includes growth of fish during the juvenile stage.
The Skalski report also noted that the Haeseker analysis did not include a variable for total dissolved gas, or possible fallback of adults from increased spill. The FPC memo did not mention this issue.
The FPC also criticized the Skalski review because it did not use the data from Haeseker's analysis to "examine whether or not their concerns regarding multicollinearity were valid," but constructed a new data set of in-river, oceanic, and terrestrial variables.
The FPC pointed to recent returns of fall Chinook as an example of improved SARs from the high spill levels of 2011. NOAA had stated in its draft BiOp that spring Chinook returns had been sub-par despite the high levels of flow and spill in 2011. Also, steelhead SARs have declined significantly the past two years from outmigrants in the high-flow, high-spill year of 2011. The FPC acknowledged the B-run steelhead were returning below expectations, but added somewhat hopefully, that they "are still migrating and counting continues." The FPC failed to note that the B-run is returning at only about 30 percent of basin harvest managers' preseason expectations.
But the FPC said the feds failed to take into account adverse hydro operations that occurred in the spring of 2011, when turbine outages and high flows led to debris buildup at turbine screens. Some screens had to be removed, which routed more fish through turbines. TDG gas levels were sometimes above 130 percent, the FPC also said.
The Skalski review pointed out that other peer-reviewed studies found that juvenile survival through the hydro system is as high as it was prior to dam construction on the lower Snake (Williams, 2001) and is similar to survival in the Fraser River with no dams (Welch, 2008).
"Should these conclusions be correct, then there may [be] no expectation of addressing the remaining sources of variation (e.g., those in the ocean) in an experimental framework," the review noted. "General purpose, observational data will often be the only source of information, and correlation studies will usually be the only means of interpreting salmonid survival relationships in the ocean."
Proponents of the spill proposal call it a "test," but they call for 10 years of the 125-percent TDG regime without any alternating spill as a control. Without such a control, it is nothing like a test, critics say.
Skalski's review said by alternating spill percentages annually between Lower Granite and McNary dams, it would take more than five decades (25,000 PIT tags annually) before researchers would have an 80-percent probability of detecting a 10-percent difference in survival.
Meanwhile, some BPA customers would like NOAA to present its own position on the controversial spill proposal and explain Skalski's analysis in front of the Power Council, since the state of Oregon and some other regional stakeholders are recommending it be added as an amendment to the region's F&W program. They made a pitch for it in September, when Haeseker presented his analysis before the full Council.
But a National Marine Fisheries Service presentation doesn't seem likely at this point, according to Ritchie Graves, NOAA Fisheries' Federal Columbia River Power System branch chief in Portland. He told NW Fishletter that his agency has already explained in the draft BiOp why it has issues with the Haeseker analysis, and will address comments on the BiOp from the FPC and others as part of the final BiOp's rollout by Jan. 1.
Graves also pointed out that hydro actions in the BiOp are designed to make sure that dam operations don't jeopardize or impede recovery of ESA-listed stocks, which is different from the type of long-term actions developed for recovery plans.
He added that his agency's own research says increasing spill would reduce overall survival of listed stocks by reducing the number of fish barged through the hydro system, and the draft BiOp is clear that NOAA wants more fish transported because of demonstrated benefits, especially for steelhead, whose returns have lagged the past two years.
The following links were mentioned in this story:
Oct. 7 memo from the Fish Passage Center
Agencies, Tribes Release Chinook Smolt Survival Study by Bill Crampton, Columbia Basin Bulletin, 12/5/3
learn more on topics covered in the film
see the video
read the script
learn the songs