Effects of climate on salmonid productivity: A global meta-analysis across freshwater ecosystems

Abstract

Salmonids are of immense socio-economic importance in much of the world, but are threatened by climate change. This has generated a substantial literature documenting the effects of climate variation on salmonid productivity in freshwater ecosystems, but there has been no global quantitative synthesis across studies. We conducted a systematic review and meta-analysis to gain quantitative insight into key factors shaping the effects of climate on salmonid productivity, ultimately collecting 1321 correlations from 156 studies, representing 23 species across 24 countries. Fisher's Z was used as the standardized effect size, and a series of weighted mixed-effects models were compared to identify covariates that best explained variation in effects. Patterns in climate effects were complex and were driven by spatial (latitude, elevation), temporal (time-period, age-class), and biological (range, habitat type, anadromy) variation within and among study populations. These trends were often consistent with predictions based on salmonid thermal tolerances. Namely, warming and decreased precipitation tended to reduce productivity when high temperatures challenged upper thermal limits, while opposite patterns were common when cold temperatures limited productivity. Overall, variable climate impacts on salmonids suggest that future declines in some locations may be counterbalanced by gains in others. In particular, we suggest that future warming should (1) increase salmonid productivity at high latitudes and elevations (especially >60° and >1500 m), (2) reduce productivity in populations experiencing hotter and dryer growing season conditions, (3) favor non-native over native salmonids, and (4) impact lentic populations less negatively than lotic ones. These patterns should help conservation and management organizations identify populations most vulnerable to climate change, which can then be prioritized for protective measures. Our framework enables broad inferences about future productivity that can inform decision-making under climate change for salmonids and other taxa, but more widespread, standardized, and hypothesis-driven research is needed to expand current knowledge.

Keywords: abundance; data synthesis; freshwater fish; growth; population dynamics; systematic review.

FIGURE 1

Summary of predicted patterns in effects of temperature (a, c, e) and precipitation (b, d, f) on salmonid productivity. Predictions are structured according to spatial (a, b), temporal (c, d), and biological (e, f) patterns that were of most interest, and stages 1–3 (boxes and arrows) correspond to the order variables were inputted into models during the stepwise model selection process (see Section 2). All panels have a shaded background to highlight expected climate effects when temperatures exceed upper thermal limits (red shading), or when low temperatures limit productivity (blue shading; see Introduction). Note that predicted effects on productivity were expected to be largely similar for measures of abundance and growth.

FIGURE 2

Best‐fit model for the Abundance–Precipitation data set, showing categorical coefficients and 95% confidence intervals plotted by season for spatial (silver) or temporal (gold) study designs (see Table S1). Total sample sizes for each level of season are shown for reference.

FIGURE 3

Best‐fit model for the Abundance–Temperature data set. Predicted values are plotted by latitude (a) and elevation (b), with fitted slope and intercepts corresponding to a reference level (c; arrow). Intercepts in (a) and (b) were adjusted to reflect the mean elevation and latitude, respectively, while points were sized according to the inverse of their sampling variance. Categorical coefficients and 95% confidence intervals (c) are plotted by age‐class for native (silver) or non‐native (gold) range portions, and spatial (circles) or temporal (triangles) study designs. Coefficients in (c) were estimated as contrasts relative to a reference level (bottom; see text) while controlling for latitude and elevation (see Table S1). Total sample sizes for each level of age‐class are shown in (c) for reference.

FIGURE 4

Best‐fit model for the Growth–Precipitation data set, showing categorical coefficients and 95% confidence intervals plotted by life‐stage for anadromous (silver) or freshwater resident (gold) populations (see Table S1). Total sample sizes for each level of life‐stage are shown for reference.

FIGURE 5

Best‐fit model for the Growth–Temperature data set. Predicted values are plotted by latitude (a) and elevation (b), with fitted slope and intercept corresponding to a reference level (c; arrow). The relationship with latitude in (a) was not signficant, so the fitted line is not shown. Points in (a) and (b) are sized according to the inverse of their sampling variance. Categorical coefficients and 95% confidence intervals (c) are plotted by life‐stage*age for lentic (silver) or lotic (gold) habitat types (see Table S1). Coefficients in (c) were estimated as contrasts relative to a reference level (bottom; see text) while controlling for latitude and elevation. Total sample sizes for each level of life‐stage*age are shown in (c) for reference.