2025MNRAS.543.1133R

Light curves of young stars exhibit photometric variability over hours to decades and across a wide range of amplitudes. On time-scales beyond a few rotation periods, these light curves are typically stochastic. The variability arises from a combination of accretion rate changes, line-of-sight extinction variations, and evolving spotted stellar surfaces. We aim to develop a methodology to quantitatively compare the full variability statistics of these inhomogeneously sampled light curves with model calculations. To achieve this, we converted the light curves into variability fingerprints. They map the probability of variation by a given amount over a given time-scale. Applying principal component analysis to these fingerprints produces a stable distribution of the first two principal components. We show that this distribution is a continuum without clusters. Adding a model-generated fingerprint to an observational sample does not significantly alter the distribution of the sample, allowing a robust comparison between the model and observed light curves to assess statistical realism. We show that photometric uncertainties, timing, and observing cadence have a minimal impact on model placement within the observational distribution. The main source of variance among highly variable light curves of young stars is the time-scale of the onset of significant variability (above 0.3 mag), with time-scales of one to three months being the most critical. The secondary cause of variance is long-term (above 1.5 yr) dimming or rising trends.