Fig 1.
Location of the study and place names.
Blue outline represents bottom depths greater than 50 m in the central Strait of Georgia, brown outline represents bottom depths greater than 50 m in the northern Strait of Georgia. Expanding symbols indicate the number of observations at each location (maximum number of observations at a location was 131). Black squares indicate the locations of the oceanographic or atmospheric data, and “X” marks locations of the salmon data used in this study. Map was created using code and bathymetric data in the R package PBSmapping [19].
Table 1.
Zooplankton taxonomic groups and physical time series considered in the analyses.
Table 2.
Definitions of symbols used in Eq 1 to describe the calculations of zooplankton biomass anomalies.
Table 3.
Number of zooplankton samples for the deep central and northern regions of the Strait of Georgia by season, available within the selection parameters (see Methods).
Fig 2.
Total zooplankton abundance, biomass, distribution of seasonal biomass, and annual biomass anomaly.
A) Grey symbols: abundance (number m-2) of all selected samples in each year. Black dots and line represent the annual mean derived from the selected samples each year, and the vertical lines represent 1 standard error about these mean annual estimates. High outlier points (representing 0.5% of the selected data) have been removed to more clearly show the spread along the Y-axis; B) Grey symbols: biomass (g m-2) of all selected samples in each year. Black dots and line represent the annual mean derived from the selected samples each year, and the vertical lines represent 1 standard error about these mean annual estimates. High outlier points (representing 0.5% of the selected data) have been removed to more clearly show the spread along the Y-axis; C) box and whisker plots of total biomass (g m-2) by season; D) Anomalies of log10 annual total biomass (based on the annual average of seasonal anomalies). Data range is from 1996 to 2018, with the climatology period for the anomaly calculations from 1996 to 2010.
Fig 3.
Total zooplankton biomass (g m-2) by taxonomic group.
A) Total biomass of each group; B) percent biomass of each group. Abbreviations of taxonomic group names are defined in Table 1.
Fig 4.
Percent biomass (g m-2) by year for the two main genera of large and medium calanoid copepods.
Large calanoid copepod taxonomic groups: Neocalanus spp., Eucalanus spp. Medium calanoid copepod taxonomic groups: Metridia spp., Calanus spp.
Fig 5.
Annual anomalies of physical variables, 1996–2018.
A) sea surface temperature (SST) at Entrance Island, b) sea surface temperature (SST) at Chrome Island, c) sea surface salinity (SSS) at Entrance Island, D) sea surface salinity (SSS) at Chrome Island, E) vertically-averaged temperature measured off of Nanoose, F) wind stress measured at Sand Heads light station at the mouth of the Fraser River, G) annual fresh water flow volume from the Fraser River (measured at Hope), H) day of the peak flow from the Fraser River (measured at Hope), I) modelled date for the peak of the spring phytoplankton bloom in the central Strait of Georgia, J) Pacific Decadal Oscillation index. Abbreviations of physical variable names (Y-axis labels) are defined in Table 1.
Fig 6.
Latent (underlying) trends derived from the 12 zooplankton taxonomic groups, and the 10 physical variables.
(A, B) Zooplankton groups; (C, D) physical variables. Black dots represent the trends derived for each year; blue line and shading represent 95% confidence bands about a loess smoother applied to these annual values (derived from a general additive model with year as the independent variable) to better show the general patterns. Dashed line between 2006 and 2007 represents the major two-cluster divisions identified in the chronological cluster analysis, and the dotted lines represent the other significant separations between groups of years in the four-cluster model (see Fig 8, below).
Fig 7.
Loadings of the variables on latent trends 1 and 2.
A) The 12 zooplankton taxonomic groups, B) the 10 physical variables. Abbreviated variable names are defined in Table 1.
Table 4.
Dynamic factor analyses of the 12 zooplankton taxonomic groups (A), and of the 10 physical variables (B).
Each combination of variance-covariance matrix configurations, and number of latent (underlying) trends, was ranked according to their best fit statistics (log-likelihood and Akaike Information Criterion with the correction for small sample sizes (AICc)). Only the top 10 of 44 trials are shown for (A), and the top 10 of 36 trials for (B). The top-ranked model is in the first row in each table.
Fig 8.
Constrained (chronological) clustering based on the two zooplankton and the two physical latent trends.
A) Broken stick model to derive the number of significant clusters (black values on or above the red line are significant). B) Cluster dendrogram. Y-axis represents the Euclidean distance (magnitude of the difference) between successive comparisons of years. Blue boxes represent the four statistically-significant clusters.
Table 5.
Best multiple regression models of the two zooplankton latent trends (Z1, Z2) and annual anomalies of total zooplankton biomass (TB), against the 10 physical variables.
Fig 9.
Statistical model fits against original early marine survival rates for Chinook salmon stocks which enter the Strait of Georgia as smolts.
Black dots and line: original marine survival data; open black circles: estimated survivals from incomplete returns; blue crosses: model fits; red triangles and vertical lines: predicted marine survivals and their 95% confidence intervals using the model with its explanatory variables. A) Cowichan River Chinook, B) Puntledge River Chinook, C) Harrison River Chinook stocks. Note that Y-axis scales differ. Model statistics are provided in Table 6.
Table 6.
Best statistical models describing early marine survival rates for selected stocks of Chinook and Coho salmon entering the Strait of Georgia as juveniles, against the available zooplankton and physical variables.
Fig 10.
Statistical model fits against original marine survival rates for the Big Qualicum Coho stock, which enter the Strait of Georgia as smolts.
Black dots and line: original marine survival data; blue crosses: model fits; red triangles and vertical lines: predicted marine survival and its 95% confidence interval using the model with its explanatory variables. Model statistics are provided in Table 6.