Basically seasonality prior scale is setting the standard deviation for the normal distribution . Higher seasonality prior scale means we are setting a wider normal distribution when sampling for the 2*n terms for our fourier. At the end, the yearly components will be simply the fourier generated dot beta.