Limitations on the recovery of the true AGN variability parameters using Damped Random Walk modeling

Context: The damped random walk (DRW) stochastic process is nowadays frequently used to model aperiodic light curves of AGNs. A number of correlations between the DRW model parameters, the signal decorrelation timescale and amplitude, and the physical AGN parameters such as the black hole mass or luminosity have been reported. Aims: We are interested in whether it is plausible to correctly measure the DRW parameters from a typical ground-based survey, in particular how accurate the recovered DRW parameters are compared to the input ones. Methods: By means of Monte Carlo simulations of AGN light curves, we study the impact of the light curve length, the source magnitude, cadence, and additional light on the DRW model parameters. Results: The most significant finding is that currently existing surveys are going to return unconstrained DRW decorrelation timescales, because typical rest-frame data do not probe long enough timescales or the white noise part of PSD for DRW. The experiment length must be at least ten times longer than the true DRW timescale, being presumably in the vicinity of one year, meaning a minimum 10-years-long AGN light curves (rest-frame). The DRW timescales for sufficiently long light curves are typically weakly biased, and the exact bias depends on the fitting method and used priors. The DRW amplitude is mostly affected by the photometric noise (so the source magnitude or the signal-to-noise ratio), cadence, and the AGN host light. Conclusions: Because the DRW parameters appear to be incorrectly determined from typically existing data, the reported correlations of the DRW variability and physical AGN parameters from other works seem unlikely to be correct. In particular, the anti-correlation of the DRW timescale with redshift is a manifestation of the survey length being too short. Application of DRW to modeling typical AGN optical light curves is questioned.