American glass eels recruitment begins in Florida around December and progresses northward to Newfoundland–Labrador until June/July (Helfman et al. 1987). The first waves of glass eels at each location were sampled in 2008 following this latitudinal trend at 17 sites evenly distributed along eastern North America up to the St. Lawrence estuary. For each location, 50 individuals were measured and preserved in 95% ethanol. Yellow eels were also collected between May and September 2008 at 15 locations ranging from the upper St. Lawrence River to the Atlantic coast of Canada. The emphasis on yellow eel sampling in this region was motivated by the occurrence of strikingly different recruitment trends reported between Atlantic Canada versus the upper St. Lawrence River and Lake Ontario (Cairns et al. 2008). Sample size varied from 69 to 100 yellow eels per location. Fin clips were preserved in 95% ethanol for DNA extraction and genotyping, and heads were kept for otolith extraction. Yellow Eels collected in 2007 were finally rejected since they were frozen before being preserved in ethanol, wich caused a bias in microsattelits amplification. AGE DETERMINATION: Age was determined for 946 yellow eels, which allowed us to subdivide them into annual cohorts. Sagittal otoliths were extracted, stored in glass vials in a 95% ethanol: glycerine solution (1 : 1 ratio), and cleaned with successive baths of bleach, water, and 95% ethanol. Once dried, otoliths were embedded in a mix of epoxy resin and hardener (4:1 ratio) inside gelatine capsules for 24 h, ground to the core on the sagittal plane, and polished with alumina powder on a polishing disc. Sections were etched, decalcified in 5% EDTA for annuli enhancing, stained in 0.01% toluidine blue solution, and digitally photographed (e.g., Tremblay 2009). The first annulus after the dark central nucleus was considered as the elver check (metamorphosis from leptochephalus larva to glass eel) and subsequent annuli as winter checks (e.g., ICES 2011). Eels were considered by convention to be of age 0 + in their year of arrival in continental waters, and their “cohort year” was thus defined as so. Each otolith was aged twice by two eel experts to confirm aging. A total of 946 yellow eel otoliths were readable, representing 17 different cohorts, each comprising 1 to 127 individuals.

We tested for statistical associations among several combinations of the NAO time series and temporal fluctuations of three variables used as proxies of relative abundance. Variation in Nb was used as a proxy for relative variation in the number of breeders. Allelic richness (Ar), measured for each cohort, was used as a proxy for the relative abundance of recruits since it has previously been proposed to correlate with offspring recruitment (McCusker & Bentzen 2010). YCSI was used as a second proxy of recruit abundance. We first tested for pairwise correlations between Nb, Ar, and YCSI, and time series of these three parameters were then compared with the monthly normalized NAO (http://www.cgd.ucar.edu). The “corresponding year” between time series represented the year when glass eels reached the continent for the Ar, Nb, and YCSI time series. To test environmental influence on previous life stages, +2 to -2 year lags were also tested. To assess the statistical significance of climate influence on eel abundance, multivariate models were run where the explanatory variables considered were the NAO time series. Stepwise regressions of the three relative abundance variables (Nb, Ar, and YCSI) were fitted to the explanatory variables to determine which ones were significant. The Akaike Information Criterion (AIC) was used to select models. Cross-validation R2 was computed to determine the prediction strength of the selected model and semi-partial R2 were computed to assess the relative importance of each selected variable. Analyses were performed using SAS 9.2 software.