The effect of (strong) discrete absorption systems on the lyman alpha forest flux power spectrum

We demonstrate that the lyman alpha forest flux power spectrum of ``randomised'' QSO absorption spectra is comparable in shape and amplitude to the flux power spectrum of the original observed spectra. In the randomised spectra a random shift in wavelength has been added to the observed absorption lines as identified and fitted with VPFIT. At 0.03 s/km <k<0.1 s/km the ``3D'' power spectrum of the randomised flux agrees with that of observed spectra within the errors. At larger scales it is still greater than 50 % of that of the observed spectra. At smaller scales the flux power spectrum is dominated by metal lines. Lines of increasing column density contribute to the ``3D'' flux power spectrum at increasingly larger scales. Lines with 13< log (N_HI/cm^{-2})<15 dominate at the peak of the ``3D'' power spectrum while strong absorbers with log (N_HI/cm^{-2})>15 dominate at large scales, k< 0.03 s/km. We further show that a fraction of up to 20 % of the mean flux decrement is contributed by strong absorbers. Analysis of the flux power spectrum which use numerical simulations with too few strong absorption systems calibrated with the observed mean flux are likely to underestimate the inferred rms fluctuation amplitude and the slope of the initial dark matter matter power spectrum. In a combined analysis with other data which constrains the dark matter power spectrum on larger scales this can result in a spurious detection of a running spectral index.