Future ground- and space-based telescopes will enable the characterization of rocky exoplanets in reflected light, allowing for the observation of their albedo, which depends on surface, cloud, and atmospheric properties. Identifying key atmospheric, cloud, and surface features is essential for assessing the potential habitability of these exoplanets. We present reference spectra and phase curves for a spatially unresolved Earth-like exoplanet in reflected and polarized light, highlighting how wavelength-dependent and phase-angle-dependent reflectance reveals key planetary properties. Performing simulations with the 3D Monte Carlo radiative transfer code MYSTIC, we improve surface and cloud modeling by introducing validated wavelength-dependent albedo maps of Earth’s seasonal and spectral features, as well as a novel treatment of subgrid cloud variability and inhomogeneities based on reanalysis data from ERA5. Our models incorporate high-resolution 3D cloud structures, demonstrating that subgrid cloud variability significantly affects both intensity and polarization. It reduces total reflectance and increases phase curve variability, especially at large phase angles where ocean glint dominates. Additionally, we show that neglecting realistic wavelength-dependent albedo maps leads to a significant overestimation of the vegetation red edge feature in reflected light spectra. Comparing an ocean planet to an Earth-like planet with seasonal cloud variability, we find that polarization is far more sensitive than intensity alone to identify the two scenarios. Moreover, polarization captures richer information on surface properties, making it a critical tool for breaking degeneracies in retrieval frameworks. We present detailed model simulations that provide a ground-truth reference for observing Earth as an exoplanet and that serve as critical benchmarks for developing observational strategies and retrieval frameworks for future telescopes targeting small rocky exoplanets. Furthermore, this study informs model requirements and establishes a framework to optimize strategies for characterizing rocky exoplanets, emphasizing the pivotal role of polarization in breaking retrieval degeneracies across different models.