Temporal delays and spatial randomness between ground-based data and satellite overpass involve important deviations between the empirical model output and real data; these are factors poorly considered in the model calibration. The inorganic matter-generated turbidity in Lake Chapala (Mexico) was taken as a study case to expose the influence of such factors. Ground-based data from this study and historical records were used as references. We take advantage of the at-surface reflectance from Landsat-8, sun-glint corrections, a reduced NIR-band range, and null organic matter incidence in these wavelengths to diminish the physical phenomena-related radiometric artifacts; leaving the spatio-temporal relationships as the principal factor inducing the model uncertainty. Non-linear correlations were assessed to calibrate the best empirical model; none of them presented a strong relationship (<73%), including that based on hourly delays. This last model had the best predictability only for the summer-fall season, explaining 71% of the turbidity variation in 2016, and 59% in 2017, with RMSEs < 24%. The instantaneous turbidity maps depicted the hydrodynamic complexity of the lake, highlighting a strong component of spatial randomness associated with the temporal delays. Reasonably, robust empirical models will be developed if several dates and sampling-sites are synchronized with more satellite overpasses.