Bottom water temperatures in (a) are presented as means±s.d. for each sampling event (2010: N=9–20, 2011: N=9–20, 2012: N=4–9 observations per sampling event) and related to date, in each year, with linear regression. Coloured lines in (b–d) were determined by logistic regression; dashed lines in b indicate 95% confidence intervals. Dates in (b–d) identify the date on which bottom water temperatures reached 10 °C in each year (determined from regressions conducted on data in (a)). Solid black lines in (b–d) indicate probabilities determined for 8, 10 and 12 °C each year. During (b) 2010 and (c) 2011, spawning had begun by the time the temperature reached 8 °C and was well under way by the time temperatures reached 10 °C. During (d) 2012, the year with the earliest spring, fish were not yet spawning when the temperature reached 8 °C, and were barely spawning at 10 °C. Instead, spawning was concentrated at warmer temperatures than during (a) 2010 and (b) 2011. The asterisks indicate that the probability of females being spent at 8, 10 and 12 °C differed in 2012 (P<0.05; 95% confidence intervals do not overlap) when compared with 2010 and 2011, which had similar probabilities across all temperatures (P>0.05; 95% confidence interval overlap).