By taking information about the past and using it to make predictions about the future, "temporal modeling"—whether of long- or short-term trends such as changes in the size of a population—can alert managers to declines of vulnerable species and help them identify which conservation actions will be the most effective. Similarly, assessing the population trends for wild species that people harvest can help managers set quotas that allow for hunting and fishing those species without putting them at risk.
One example is the use of models by fisheries managers to set sustainable limits for salmon fishing. Predictions drawn from models inform their decisions about how many fish can be harvested in the present while also ensuring a healthy salmon population in the future. Currently, the model used to predict future salmon abundance is based on counts of "jacks"—two-year-old male fish that return to rivers in smaller numbers than the more typical three- or four-year-olds. Because a known percentage of the population is made up of jacks, managers know that when jack numbers are high the overall salmon population is large—and that more salmon can be fished the following year.
|Cassin's Auklet. Photo by Peter LaTourrette.|
We recently evaluated the possibility that seabirds too can be used to predict salmon abundance. If so, this would potentially add a valuable tool to the resource manager's kit. The connection between seabirds and salmon centers on the oceanic part of the fishes' life cycle. Salmon return to spawn in their natal rivers and streams only after spending several years at sea. While they are feeding and growing in the ocean, these fish rely on prey populations that may undergo declines or increases, affecting salmon survival. Seabirds depend upon many of the same prey species as salmon and should respond similarly to fluctuations in the marine environment.
We hypothesized that we could use seabirds to assess the effects of changes in populations of small forage fish—a major prey resource but difficult to measure directly and rarely incorporated into fisheries models. Seabirds can be especially useful indicators of conditions in the marine ecosystem, including the status of forage fish populations. Highly visible members of the marine community, seabirds are are relatively easy to study at their breeding colonies such as the Farallon Islands, where PRBO has conducted research for more than 40 years.
We tested our idea by modeling the relationship, over time, between seabird breeding success on the Farallon Islands and salmon abundance in Central California (measured as the ocean harvest south of Point Arena plus the number of fish returning to spawn in Central Valley rivers). We focused on the period 1990–2004, a subset of our long-term data, because it corresponds to the time period used by fisheries managers to predict salmon abundance. This enabled us to compare our results to theirs.
|Figure 1. In years after Cassin's Auklet breeding success is high, more Chinook salmon return to spawn. The data for each year (red circles) were used to build a model describing the relationship (red line). Click on red type for larger view of Figure 1.|
Of the five seabird species we included in our analysis, one in particular—the Cassin's Auklet—revealed a strong correlation between breeding success in one year and salmon abundance the following year. More salmon return to spawn in Central Valley rivers following years of high auklet breeding success (Figure 1). In fact, this correlation between seabirds and salmon is as strong at the correlation between salmon abundance and jack returns, described earlier.
An important task when creating a model is to assess the reliability of the predictions. One way to do this is to exclude each year in turn from the analysis, re-create the model using the remaining data, predict the value for the excluded year, and compare the predicted and actual values. Using this technique, we found that the auklet model was as reliable as the jack model. The average difference between actual salmon numbers and what we predicted based on seabirds was 7%. This difference was 5% with the jack model (Figure 2).
|Figure 2. To validate this model, PRBO compared actual numbers of returning salmon to the predictions made by both models. The Cassin's Auklet model was closer in some years, the "jack" returns model in others. Click on red type for larger view of Figure 2.|
While the average performance of the models across all years was similar, there were differences in how well the models performed from one year to the next. The auklet model was a better predictor of salmon abundance in some years, the jack model in other years. We're currently working to understand why this occurs— and potentially identify which model will provide the most accurate predictions in a given year.
Chinook salmon have suffered sharp declines in recent years due to degraded river systems and unfavorable ocean conditions. This led to a complete closure of the Chinook salmon fishery in Central California in 2008. The situation bodes unfavorably for salmon, for the ecosystems that support them, and also for California's fishermen—and makes the need for accurate predictions of salmon abundance greater today than ever.
PRBO models indicate that seabird data may yield new insights into the health of prey populations and the marine ecosystem—information of value to fisheries managers. Ultimately, discovering ways to use seabird data in fisheries models may provide us with new tools that will help to protect both salmon stocks and commercial fisheries.