A computer model, the Finite Element Storm Hydrograph Model (FESHM), was developed to integrate spatiotemporal variability in climatic and watershed descriptors into a model structure to determine stormwater runoff from small watersheds. The model consisted of two major components: a rainfall excess generator and a flood routing algorithm. The Holtan infiltration equation was used as the basis of a soil moisture routine to estimate rainfall excess from a given rainfall distribution. The finite element technique, with Galerkin's residual method, was used to provide a numerical solution of the equations of continuity and momentum for one-dimensional transient flow for routing overland and channel flow through a drainage system. A spatially responsive modeling concept was assumed. To implement this concept the watershed was discretized into hydrologic response units (HRU's) based on soil mapping units and landuse to improve estimates of rainfall excess. The watershed's topography and drainage patterns were used to define finite elements to improve the accuracy of flow routing. These two discretization schemes resulted in poor geometric registration. This problem was resolved by a linear weighting of the rainfall excess from each HPU located within a given element as a function of the area occupied by the HRU to the total element area. Spatial uniqueness was maintained on a relatively small scale. Sufficient flexibility was incorporated into the model to allow varying levels of discretization without rebuilding the model structure. Procedures were developed for defining model parameters for application of the model in an ungaged context. The effect of landuse changes on the hydrologic response of a watershed can be, at least, conceptually evaluated with this model structure. A modeling framework also is provided for the prediction of water quality. A hypothetical watershed was used te conduct a sensitivity analysis of model parameters. Six experimental watersheds located in different physiographic regions of Virginia were used to evaluate the capability of the model to predict stormwater flow for flood-producing events in an ungaged context. Comparison of simulated and recorded hydrographs ranged from good to excellent / Ph. D.