In addition to the obvious direct consequences and importance of the global water cycle, understanding the movement and storage of water at the land surface has crucial consequences for climate prediction and society. The flux of freshwater from the continents, especially from northern rivers, controls the thermohaline circulation of the world ocean. Furthermore, evaporation is a common term in the surface energy and water balance equations, therefore changes in the water cycle - both natural variability and change related to anthropogenic causes (land use and cover change, and/or greenhouse gas emissions) are inextricably linked. We evaluate the mean seasonal and interannual variability for the major terms in the surface water balance over the extra-polar continents, including precipitation, evaporation, runoff, and storage change in the subsurface, in snow and glaciers, and in lakes and wetlands. Our main source of information is a 21-year global simulation of the Variable Infiltration Capacity (VIC) model, forced with observed precipitation and temperature, and derived surface radiation. For many of the terms, we find that the mean interannual variability (mean absolute value of the differences in annual total from one year to the next) is comparable to the mean. Furthermore, there are large spatial differences in the storage and movement of water globally .We evaluate the magnitude of the simulated variability as compared with observations where available, and with simulations of the NCAR/DOE Parallel Climate Model (PCM). Neither VIC nor PCM simulates the storage of water in river channels, and in lakes and wetlands, so these terms are inferred from observations. We also make rough estimates of the effects of global water management on the mean and variability of the various water cycle fluxes and storage change terms. At the continental scale, PCM simulations of the means of the major fluxes match observations and/or VIC simulations reasonably well, but the estimates of interannual and interseasonal variability of the major fluxes, and storage change terms, are considerably smaller for PCM as compared with VIC and observations, to the extent they are available. Using the VIC data and atmospheric divergence and storage data from the NCEP/NCAR Reanalysis we also make comparisons of the land and atmospheric water budgets at various spatial and temporal scales. We discuss the reasons for the differences and relate systematic biases to known weather and climate features. The differences may also give some insight into inadequacies in observational networks and methods and how these may be improved. With a view to this, the worth of the reanalysis data is assessed by comparing with TOVS and SSMI atmospheric moisture data.