Spatially and spectrally resolved models were used to explore the observational sensitivity to changes in atmospheric and surface properties and the detectability of surface biosignatures in the globally averaged spectra and light-curves of the Earth. Compared with previous efforts to characterize the Earth using disk-averaged models, a more comprehensive and realistic treatment of the surface and atmosphere was taken into account here. Our results are presented as a function of viewing geometry and phases at both visible/near-infrared (0.5-1.7 microm) and mid-infrared (5-25 microm) wavelength ranges, applicable to the proposed NASA-Terrestrial Planet Finder visible coronagraph and mid-infrared interferometer and to the ESADarwin mission architectures. Clouds can change the thermal emission by as much as 50% compared with the cloud-free case and increase the visible albedo by up to 500% for completely overcast cases at the dichotomy phase. Depending on the observed phase and their distribution and type, clouds can also significantly alter the spectral shape. Moreover, clouds impact the detectability of surface biosignatures in the visible wavelength range. Modeling the disk-averaged sensitivity to the "red-edge," a distinctive spectral signature of vegetation, showed that Earth's land vegetation could be seen in disk-averaged spectra, even with cloud cover, when the signal was averaged over the daily time scale. We found that vegetation is more readily discriminated from clouds at dichotomy (50% illumination) rather than at full phase. The detectability of phytoplankton was also explored, but was found to be more difficult to detect in the disk-average than land vegetation.