In recent years, severe storms have impacted several large cities around the world, producing floods and significant property damage due to the lack of adequate flood warning systems. Such problems have occurred in major US cities, for example, the New York City metropolitan area was impacted by a Mesoscale Convective System (MCS) in August 2007, with rainfall exceeding three inches in less than two hours in some areas. The subway system was partially closed due to flooding, streets were impassable, and over two and a half million people, and numerous businesses were affected. To activate a warning system at the appropriate scale, a numerical urban flood forecasting model and its application to the New York area is presented. This model has two major integrated components: a hydrological model and a mesoscale meteorological model. The hydrological component employs unstructured two-dimensional grids to capture local terrain effects in a conservative approach to model the surface flow. The numerical weather prediction component is coupled with a real-time data feed from both NexRad radar and surface stations to simulate precipitation impact on public infrastructure. A case study of the flooding event that occurred on August 8, 2007 is used to illustrate the capabilities of this integrated system and validity of the approach. We will discuss simulated flood patterns using meteorological model outputs at a one to four km resolution, radar data oversampled at 100m resolution, and surface observations at sparsely located sites. The integrated model is designed to forecast flooding events with a lead time from six to 72 hours. We will discuss how flood modeling in urban areas can be simplified by employing one-dimensional grid elements with a reduction of computational effort and without any loss of generality. Future work to simulate a complete basin, namely coupling this model to a subsurface groundwater three-dimensional porous media simulator employing tetrahedral numerical meshes, will also be discussed.