Autumn foliage color is an important phenological characteristic associated with climate and appeals to populations as a cultural ecosystem service (CES). Land surface phenology (LSP) analyzed using time-series remotely sensed imagery can facilitate the monitoring of autumn leaf color change (ALCC); however, the monitoring of autumn foliage by LSP approaches is still challenging because of complex spatio-temporal ALCC patterns and observational uncertainty associated with remote sensing sensors. Here, we evaluated the performance of several LSP analysis approaches in estimation of LSP-based ALCCs against the ground-level autumn foliage information obtained from 758 sightseeing (high CES) sites across Japan. The ground information uniquely collected by citizens represented ALCC stages of greening, early, peak, late, and defoliation collected on a daily basis. The ALCC was estimated using a second derivative approach, in which normalized difference vegetation index (NDVI), kernel normalized difference vegetation index (kNDVI), enhanced vegetation index (EVI), two-band enhanced vegetation index (EVI2), and green red vegetation index (GRVI) were applied based on MODerate resolution Imaging Spectroradiometer (MODIS) MOD09A1 with four (Beck, Elmore, Gu, and Zhang) double logistic smoothing methods in 2020. The results revealed inconsistency in the estimates obtained using different analytical methods; those obtained using EVI with the Beck model estimated the peak stage of the ALCC relatively well, while the estimates obtained using other indices and models had high discrepancies along with uncertainty. Our study provided insights on how the LSP approach can be improved toward mapping the CESs offered by autumn foliage.