ADASSX

Quasar accretion disks, predicted by the standard thin disk theory to emit optical continuum from regions spanning light‐hours, remain challenging to resolve via traditional reverberation mapping (RM) techniques limited by: 1. Long, daily to monthly observational cadence, and 2. Oversmoothing in damped random walk (DRW) time-series models. We present a 6-month, high-cadence (180s exposure, 3-5 hours per night), cost-effective photometric RM campaign using the 24-inch Keeler Telescope at the Allegheny Observatory, monitoring 3 circumpolar quasars in g′ and r′ bands. Preprocessing pipeline incorporated differential photometry and 10-min binned spline interpolation, yielding <0.01 mag precision. By developing an adaptive Bayesian RM model, we achieved hourly-scale lag sensitivity through a truncated DRW kernel, a mixture density network, and Markov chain Monte Carlo sampling. On 15,000 simulated optical light curves with known, short time lags, our model correctly predicted 71.17% of lags within 0.2 days, outperforming the JAVELIN RM model by 17.92%. For quasar PG0804+761, we revealed inter-band time lags of 0.61 ± 0.24 days, corresponding to disk radii scaling with the thin disk theory. Applied to 13 publicly available daily-cadence quasar light curves, our model recovered time lags from 0.02 to 1.23 days, consistent with thin disk estimations, and achieved reduced uncertainties compared to existing studies. In essence, we place promising constraints on light-hour-scale accretion disk radii predicted by the standard thin disk model. Our sub-meter telescope’s photometric quality is comparable to space-based surveys, and our model enables robust hour-scale reverberation mapping critical for next-generation active galactic nuclei research.