How should we aggregate data? Methods accounting for the numerical distributions, with an assessment of aerosol optical depth