We describe a way to parameterize power spectra extracted from fixed-frequency reflectometry data, with a view to systematic studies of turbulence properties in tokamak plasmas. Analysis of typical frequency spectra obtained from a new database suggests decomposition in a set of four key components: the direct current component, low-frequency fluctuations, broadband (BB) turbulence, and the noise level. For the decomposition in the identified components, different kinds of functions are tested and their fitting performance is analyzed to determine the optimal spectrum parametrization. In particular, for the BB turbulence, three models are compared qualitatively based on a number of representative spectrum test cases, notably the generalized Gaussian, the Voigt, and the Taylor model. In addition, quantitative performance testing is accomplished using the weighted residual sum of squares and the Bayesian information criterion in a large database including 350 000 spectra obtained in Tore Supra. Next, parametrization by the Taylor model is applied to Ohmically heated plasmas, and a BB energy basin is systematically observed in the core plasma region, which shrinks with decreasing radial position of the q = 1 surface. This basin might be explained by a drop of the density fluctuation level inside the q = 1 surface.