Impact of Fine Sediment on Egg-To-Fry Survival of Pacific Salmon: A Meta-Analysis of Published Studies

Abstract

Egg-to-fry survival of salmonids is tempered by habitat degradation, including increased sediment in streams. To best manage multiple salmon species and prioritize scarce habitat restoration funds for the benefit of fish recovery, many studies have described and predicted the relationship between fine sediment deposited in spawning gravels and salmonid egg-to-fry survival. In this article, we used published studies, agency reports, and university theses (N= 14) to create predictive relationships between percent fine sediment and egg-to-fry survival of Chinook (Oncorhynchus tshawytscha), coho (O. kisutch) and chum (O. keta) salmon, and steelhead trout (O. mykiss). In our analysis, coho survival tended to decline more rapidly per unit sediment increase and chum survival least rapidly. Threshold effects were observed, with survival dropping rapidly when percent fines less than 0.85 mm was greater than 10%. For other size classes of fines, a threshold was primarily observed only for eyed egg survival when fines exceeded 25–30%. Our predictive models combine both field and laboratory data and take into account a variety of conditions; they include estimates of uncertainty in the impact of sediment on egg-to-fry survival. These models can be used to forecast effects of watershed management practices on salmonids and to make comparisons between predicted salmonid survival rates under alternative management strategies for conditions where fine sediment is the limiting factor for survival.

Keywords:

• salmon incubation
• salmon spawning
• sediment deposition

ACKNOWLEDGMENTS

We thank Will Holden for his help in the literature search. Martin Liermann, Naomi Yoder, and Phil Roni provided constructive comments on earlier drafts of the manuscript. The editor and an anonymous reviewer also provided helpful suggestions that improved the manuscript.

This article is not subject to U.S. copyright law.

Notes

^a Number of redds (or artificial equivalent) is the same for each experiment unless indicated otherwise.

^b All relationships shown are statistically significant (Wald test) at the 0.05 level. A – indicates the slope is not significant, + indicates the individual relationship is not statistically significant, but when combined with other observations within the same species and egg stage, a significant relationship emerges.

^c Development stage not identified.

^a The same estimate within a group signifies that studies were not significantly different from one another (p> 0.05), and different estimates indicate significant differences within the group (p< 0.05). Standard error is from logistic regression if bootstrapping was unnecessary; otherwise, it was from bootstrap. Excluded studies were those with non-significant slopes (Table 1).

^b p-value refers to the test of whether the parameter is the same for all group members.

^c Intercept fit using logistic regression. Standard error was included if no bootstrapping was necessary.

^d Intercept fit during bootstrap procedure with bootstrap standard error.

^a The same estimate within a group signifies that studies were not significantly different from one another (p> 0.05), different estimates indicate significant differences within a group (p⩽ 0.05). Standard error is from logistic regression if bootstrapping was unnecessary; otherwise, it was from bootstrap.

^b p-value refers to the test of whether the parameter is the same for all group members.

^c Intercept fit with logistic regression. Standard error was included if no bootstrap was necessary.

^d Intercept fit during bootstrap procedure with bootstrap standard error.

^a The same estimate within a group signifies that studies were not significantly different from one another (p> 0.05), different estimates indicate significant differences within a group (p< 0.05). Standard error was from logistic regression if bootstrapping was unnecessary; otherwise, it was from bootstrap.

^b p-value refers to the test of whether the parameter is the same for all group members.

^c Intercept fit with logistic regression. Standard error was included if no bootstrapping was necessary.

^d Intercept fit during bootstrap procedure with bootstrap standard error.