Use of spectral analysis to estimate short-term periodicities in growth rates of brook trout Salvelinus frontinalis
We used spectral analysis to investigate the periodic components of relative growth rates (on a mass and length basis) in two groups of six\b young-of-the-year brook trout (Salvelinus fontinalis). Fish were measured every 2 days for 25 days according to a repeated measures design in time. The predominant frequency components were estimated at the end of a stepwise procedure involving the analysis of the multifrequential periodogram. We found mainly two superimposed periodic components that resulted in short rhythms; these had a period of about 5 and 10 days for the growth rate in mass and of about 4 and 6 days for the growth rate in length (R2 = 91% for mass and 90% for length, both including an autocorrelation term). Repeatability of rhythms was assessed by ANOVA of the finite Fourier transform, which showed that the two groups of fish reared under the same experimental conditions exhibited growth rhythms that were not significantly different from each other. On that basis, we present a harmonic model which predicts short-term variation in growth rate of brook trout in mass (R2pred = 0.73, with an autocorrelation term) and in length (R2pred = 0.72).