Issue comments, feedback, suggestions
NW Fishletter #384, August 1, 2018

[7] NOAA Fisheries Turns To Newer Modeling To Assess Harvest Impacts On Chinook Abundance

When it comes to studying salmon and making predictions about their future, there's much room for error.

First, a multitude of variables change from year to year: fluctuating river temperatures, changing ocean conditions, toxins, predators and dam operations, plus impacts from an early or late, a high or low, or a fast or slow runoff. Then there's the human variable--whether biologists are looking in the right places, or at the right time, or during the right conditions to accurately assess population numbers.

A newly released technical memorandum from NOAA's Northwest Fisheries Science Center uses a different kind of modeling to assess the impact of harvest rates on future abundance of 29 spring/summer Chinook salmon populations in the Columbia and Snake rivers, and on the likelihood numbers would drop so low they may be unable to recover--called quasi-extinction.

The lead author, Eric Buhle, said this study attempts to bring cutting-edge population modeling to salmon in the Columbia River Basin, which has lagged behind in the scientific world of life-cycle models. An ecologist, consultant for Biomark Applied Biological Services and affiliate with NOAA's Northwest Fisheries Science Center, Buhle said this is his first formal effort to apply a more accurate form of modeling he's been working on for years. He said the integrated population model approach avoids some flaws in traditional modeling that have been identified in scientific literature for decades. And, while this study looks at the impact of harvest rates on future abundance, the same methods can be used in other life-cycle population assessments in the Columbia River Basin, Buhle said.

Modeling challenges are well known and regularly pointed out. For example, last fall, the Independent Scientific Advisory Board noted the problems of modeling in its review of NOAA Fisheries' ongoing effort to use life-cycle modeling in its adaptive management strategy for Columbia basin salmon and steelhead studies. "Life-cycle modeling remains a significant challenge because of the complexity of the wide-ranging life histories of these fish and the many locations where fish are affected by human activities and the changing environment," ISAB wrote, adding, "Models are always a tradeoff between realism and simplicity."

NOAA's new technical memo notes in its introduction that limited data, errors in population surveys or imprecise age estimates in a population can bias a study's outcome. "In turn, any management decisions based on such models could be misguided, potentially hindering recovery of the population. Therefore, proper consideration of all sources of uncertainty in the data is necessary to design robust conservation strategies," it states.

The outcome of the study itself--which allows fish managers to plug in a harvest rate and check the forecast spring or summer Chinook abundance in, say, 25 or 50 years--was no surprise. Abundance declines and extinction risks go up across a range of fixed harvest rates. "Large-scale environmental fluctuations (e.g., ocean conditions and hydrosystem operations represented by the shared process error) were at least as important as harvest in determining long-term population viability. If future environmental conditions are relatively poor, and especially if they are assumed to have undergone a persistent state shift at some point in the last 60 years, then quasi-extinction risks are dramatically elevated even in the absence of harvest," the study concludes.

Buhle said these offer another piece of consistent evidence in Chinook salmon abundance predictions, especially if assuming a dramatic environmental shift in the 1970s, and future conditions look more like they have in the last few decades. This study's difference, Buhle said, is the model used to reduce uncertainties.

First, Buhle's method separates errors that occur in the environment, or process errors, from errors that occur in the gathering of information, or observation errors. Separating the two is important, he said, because an impact from the environment one year, like poor ocean conditions, will play out over several generations of salmon, while an observation error will have no impact in future years. "The fish don't care if we count them or not," he noted. "So it's really important to try to tease that out."

Buhle's modeling allows for these two kinds of errors as separate variables, and also uses a method known as partial pooling, or the Robin Hood Principle. With it, he said, "We're not just analyzing one population out of 29 at a time, we're pooling information across all of them." Buhle said some populations have been highly monitored, while others have significant holes in the information that's been gathered. Scientists can make key assumptions for populations with less data by assuming the population experienced certain similar impacts as populations with lots of data--borrowing data from the rich and giving it to the poor. It works because some things are shared--like ocean conditions, or flow in the main stem--and will have a similar impact on all populations, Buhle said. "All of the data is brought to bear, but weighted according to how much they contribute. The data-poor populations are leaning on the data-rich," he said.

One of the region's best known life-cycle studies--the Comparative Survival Study put out by the Fish Passage Center every year--does not use the classical spawner-recruitment modeling that is fraught with flaws, Buhle said, but it also does not separate the two kinds of errors, nor does it use partial pooling.

"It's fair to say the IPM approach we used is more cutting-edge, and the best available methodology," he said, adding that the CSS model "includes a lot of things we haven't done yet." And while the two approaches are different, Buhle said other differences may be even more consequential. He noted his study looked at 29 populations of spring/summer Chinook while the CSS examined six, but also included smolt abundance data and other components of hydro system survival that haven't been included in Buhle's study.

Although integrated population modeling has not been widely used to study salmon in the Pacific Northwest, it has been used in stock assessments in the ocean, and bird studies, he said. And scientists in Northern California have been using similar methods since the early 2000s, leading the way in advancing salmon population modeling, he said.

Buhle said in his application of the integrated population model, he used raw data that has been gathered since the 1950s on the number of Chinook spawners, their age structure, the fraction born in the wild versus in a hatchery, and harvest rates.

It's all basic and widely available information. Using more cutting-edge modeling to interpret the data is adding value to work already done, he said. "This is giving us the ability to make some inferences that haven't been possible, or certainly emphasized, in the past," he said. "To me, as a scientist, this is just a really exciting advance in our ability to model these populations that so many people care about, and that are super important ecologically and culturally." -K.C. Mehaffey

Subscriptions and Feedback
Subscribe to the Fishletter notification e-mail list.
Send e-mail comments to the editor.

THE ARCHIVE :: Previous NW Fishletter issues and supporting documents.

NW Fishletter is produced by NewsData LLC.
Publisher/Editor-in-Chief: Mark Ohrenschall, Editor: K.C. Mehaffey
Phone: (206) 285-4848 Fax: (206) 281-8035

Energy Jobs Portal
Energy Jobs Portal
Check out the fastest growing database of energy jobs in the market today.
What's New

NW Wholesale Power Markets Conference: Transition in Progress--Wednesday, May 15, 2019, Portland, Ore.

Substation Northwest Podcast, Episode 4
Taste editorial excellence.