By Todd Seamons
Millions of insects, birds, fish and other animals are released into the wild around the globe every year. These animals are released for many different reasons. In fisheries, these releases are often intended to support declining populations, direct fishing effort away from wild populations, mitigate habitat losses, or provide harvest opportunities. This is true whether you are in the Pacific Northwest of the United States or in Europe. The potential negative ecological and genetic effects of these releases have been identified and studied for several decades and are now more widely accepted and understood. Currently, scientists and managers are looking for ways to minimize or eliminate these negative effects.
The Hatchery Scientific Review Group, an independent scientific review panel organized as part of the federally funded Hatchery Reform Project, undertook an extensive review of hatchery programs in Washington State and the Columbia River. As a result of their assessment, they proposed two management strategies specifically designed to minimize the negative genetic effects of artificially propagated individuals on wild populations. The basic idea was to either treat the wild and hatchery populations as one population, intentionally interbreeding wild and hatchery populations to minimize domestication of wild stocks (integration strategy), or to treat the wild and hatchery populations as separate populations, intentionally preventing or minimizing interbreeding between domesticated hatchery stocks and their wild counterparts (segregated strategy).
Nearly all of the winter steelhead (Oncorhynchus mykiss) hatchery programs in Washington are currently managed using a strategy similar to the proposed segregated strategy. These programs use the domesticated “Chambers Creek” stock as hatchery broodstock. The Chambers Creek stock was intentionally selected to migrate home and spawn months earlier (December-January) than most wild populations (March-April). (The original purpose of this selection was to minimize the time needed to grow a fish to smolt size; most wild Washington winter steelhead spend two years in freshwater before migrating to the sea as smolts, but Chambers Creek stock are grown to smolt size in a little over one year.) The winter steelhead hatchery strategy is similar to the proposed segregated strategy in that interbreeding between wild and hatchery fish is prevented in the hatchery by using only adults produced in the hatchery as broodstock. However, the proposed segregated strategy also calls for minimizing interbreeding outside of the hatchery on spawning grounds used by wild fish. Currently, most winter steelhead hatchery programs do not actively limit the number of hatchery fish allowed on wild spawning grounds. This may be partially due to the belief that a) the divergent life history of the wild and hatchery steelhead naturally limits interbreeding between wild and hatchery fish, and b) early spawn timing of hatchery fish is maladaptive, that is, spawning at the wrong time of year produces few surviving offspring. This belief appears to have led to the recent suggestion that future marine aquaculture programs can purposely select for divergent life history in hatchery broodstock in order to minimize interbreeding between hatchery and wild stocks. However, to our knowledge, no one had conclusively tested this aspect of the steelhead hatchery management strategy, upon which this recommendation rested. Thus, using winter steelhead as a model system, our objective was to determine if the difference in return and spawn timing effectively prevented interbreeding of hatchery and wild fish, and if not, to see if we could identify some factors that may affect the levels of interbreeding.
We performed our research at Forks Creek Hatchery, located on Forks Creek, a tributary to the Willapa River in southwest Washington. Forks Creek was a great place for this research because we were able to sample fish from the start of the current program, which began spawning fish in 1996; that is, we were able to obtain a pre-hatchery program baseline of wild fish data. Like other winter steelhead programs, hatchery produced fish were all marked by removal of the adipose fin. After the first three years of the program, hatchery fish began to return to Forks Creek. Some of those fish spawned in the wild and likely produced offspring. However, all naturally produced fish, whether their parents were hatchery fish or wild fish (or one of each) were unmarked. Thus, we used our baseline genetic data and several statistical tests to estimate the ancestry (hatchery or wild) of unmarked naturally produced smolts and adults. In total, we had 12 years of smolt collections and 10 years of adult collections that would have been affected by naturally spawning hatchery fish. For each of these collections, we added up the number of fish identified as wild ancestry and divided by the total number in that collection to get a proportion or percent of a collection that was wild. We then looked for trends across time in the wild proportions.
All statistical methods produced the same pattern; the proportion of collections genetically identified as wild ancestry declined over time by 10% to 20% after three generations, and the proportion identified as hatchery ancestry increased over the same time period. So wild fish were declining and naturally produced hatchery ancestry fish were increasing. However, the statistical tests produced a third group of fish. These were fish that we could not confidently categorize as hatchery or wild. They could have truly been hatchery or wild fish, but because of relatively small genetic differences between hatchery fish and wild fish (reflective of the overall relatively low genetic diversity in the species), the 8 genetic loci used were not sufficiently powerful or informative enough to allow a confident assignment. Another possibility was that they were hatchery-wild hybrid fish. Hybrids were even more difficult to distinguish from purely hatchery and purely wild fish given these same genetic data, so we simulated some hybrid individuals using hatchery and wild genotypes (from the earliest years before hatchery fish were returning to Forks Creek) and ran those through the statistical assignment test. Then, using the assignment rates of the statistical test and linear algebra (and viable assumptions), we “corrected” our estimates of proportions to include hatchery-wild hybrids. That is, we could not identify individual hatchery-wild hybrids, but we could say that a proportion of an annual collection was made up of hatchery-wild hybrids.
The difference in migration and spawn timing was not enough to prevent successful reproduction of hatchery fish or interbreeding between hatchery and wild fish. We found hatchery ancestry fish in almost all collections and we found hatchery-wild hybrids in all smolt and adult collections. An average of 10% of smolts and 5% of adults were naturally produced hatchery ancestry fish and an average of 56% of outmigrating smolts and 43% of returning adults were hatchery-wild hybrids. Hatchery and hybrid estimates varied annually, for example, the percent of hybrids in smolt collections ranged from 37 to 76, and the percent of hybrids in adult collections ranged from 17 to 85, and we used the annual variation to evaluate three factors that might influence hatchery and hybrid proportions.
First, one might expect to find naturally produced hatchery ancestry fish or interbreeding between hatchery and wild fish if a large number of hatchery fish spawn in the wild, both in absolute terms and relative to the number of spawning wild fish. We knew that some hatchery fish spawned outside of the hatchery in Forks Creek, the Willapa River or maybe in another Willapa River tributary, but we did not know how many. The exact number of naturally spawning hatchery-produced fish was less important than capturing the annual variation in this number, therefore the number of hatchery-produced fish returning to the hatchery in each year was used as an estimate of the number of naturally spawning hatchery fish. Likewise, the number of wild fish spawning in the wild was not known. To capture the annual variation in wild fish number (the second factor we evaluated), we used the estimates of wild fish numbers estimated by the Washington Department of Fish and Wildlife (WDFW) for the entire Willapa River. WDFW estimates the number of wild spawners by counting the number of steelhead redds in index reaches of the Willapa River and its tributaries. These numbers are entered into a statistical formula that estimates the total adult spawner abundance.
Third, we evaluated the effect of stream flow during the time of year when hatchery steelhead migrate and spawn. Remember, hatchery fish return and spawn months earlier than wild fish, and the stream flow at that time is thought to be poor quality for steelhead spawning and incubation. For this factor, we used data from the USGS stream gauge on the Willapa River. To analyze these factors, we used a method of regression analysis that, in theory, would allow us to say if one factor had more of an influence than another on the proportion of hatchery-wild hybrids.
We found that we did not have enough data to conclusively select one regression model over another; all three factors were plausible drivers of variation in hatchery-wild hybrid proportions. However, the estimated trends fit our expectations. For example, the proportion of hatchery ancestry smolts was lower in years when stream flow during hatchery spawning was high and higher in years when winter stream flow was low. This matches our expectation given the hypothesis that high winter stream flow is not conducive to successful spawning and egg incubation. We also found that hybrid smolt and adult proportions were larger in years when more hatchery-produced fish were spawning in the wild. If you release more hatchery-produced fish into the wild you get more hybrid fish. The number of spawning adult wild fish did not seem to have much of an effect. This could be a limitation of the data, that is, the estimates of wild fish abundance were not very accurate, or it could be that the number of wild adults does not affect the proportion of hatchery or hybrid fish that end up in the population.
As stated earlier, successful implementation of the proposed segregated strategy depends on controlling what happens in the hatchery and, to some degree, what happens outside the hatchery. Unfortunately, there are no easy solutions, it is extremely difficult to control what happens outside of the hatchery. Weirs, in place in many river systems (including Forks Creek), may allow for selectively removing hatchery fish and releasing wild fish. However, high stream flows simultaneously induce upstream migration of fish and incapacitate weirs and fences used to exclude hatchery fish. Hatchery fish are also removed by selective harvest allowed during the migration and spawn timing of hatchery fish. In the Willapa River system, a recreational fishery allows for harvest of hatchery fish downstream of the hatchery, intercepting fish on their way home to Forks Creek. However, this selective harvest is unreliable and does not remove enough hatchery fish from the naturally spawning population as evidenced by the number of hatchery fish that make it through the fishery to arrive at the hatchery and by our analysis. The easiest solution may be to end hatchery programs. However ending hatchery programs would be unpopular and politically difficult to implement. Other solutions may exist, but they are not obvious.
A common misconception of our analysis is that our research shows that the wild population is doomed to low fitness due to interbreeding with hatchery fish. This is not what our data and analysis demonstrate. Our research simply shows that you cannot rely on divergent life history alone to minimize genetic interactions between hatchery and wild fish.