Every day, important decisions am made regarding activities affected by variations in water levels and flows on the Great Lakes. These involve large-scale issues, such as lake-level control or land-use regulation, as well as local issues, such as siting and design of structures and protective works. Such decisions can and should make use of statistical models that quantify the variability of levels and flows, To date, the only widespread applications of statistical models have been to estimate the probability distributions of high lake levels for use in shoreline zoning and of waves for use in the design of shoreline facilities and protective works. New statistical models of Great Lakes levels should be able to correctly account for serial correlation in hydrologic levels, provide estimates of the marginal and joint distribution of hydrologic levels and storm surge, provide estimates of the joint distribution of various wave parameters and storm surge, and be readily applied to specific coastal locations, The alternative modeling strategies explored address some of the deficiencies of existing models. To improve Great Lakes water level statistics, a comprehensive, coherent, and unified strategy for modeling Great Lakes hydrology is required. Key elements of such a strategy include user community accessibility, linkage between dcterministic and stochastic elements, and validity over a wide range of temporal and spatial scales. With the development of improved hydrologic models, statistics that reflect the level of model sophistication would be derived. These statistics would be conditioned on present levels and existing climate regimes, and incorporate the concept of planning horizon, correctly compute the joint probability of the combined effects of mean levels, surges, and waves, and correct for physical trends such as crustal movement.