Watershed characteristics affect the stream flow of the watershed. Those effects vary in different temporal and spatial scales. The long-term stream flow can be statistically described by magnitude, frequency, and duration; those hydrological indicators are important metrics that can be used to reflect the effects on flow by watershed characteristics. However, the responses of hydrological indicators to watershed characteristics are still a controversial topic throughout the world. Our goal is to address the following questions based on data analyses of a large number of hydrological monitoring sites and their corresponding watersheds: (1) what and how watershed characteristics affect stream flow? (2) Which hydrological indicators have significant responses to watershed characteristics and what are the differences of their responses among different physiographic provinces? We randomly selected 150 small watersheds (<282km²) in three physiographic provinces in Chesapeake Bay region (37 in Coastal Plain area, 56 in Piedmont area, and 57 in Highland area). Using long-term daily stream flow data from the U.S. Geological Survey (USGS), we calculated 34 hydrological indicators for all watersheds. Seventeen hydrological indicators were selected using principle component analysis for further analyses. For all watersheds, we used ArcGIS to develop proportions of agriculture, developed land, grassland, and forest from the Mid-Atlantic RESAC 2000 land use map. We derived the values of impervious surface coefficients for all watersheds from the Mid-Atlantic RESAC impervious surface map. The proportions of sand, rock depth, hydrologic group and permeability rate were collected from the STATSGO soil database of the United States. We then used stepwise regression to quantify the effects of watershed land use, impervious surface, and soil physical properties on the selected 17 hydrological indicators in three physiographic provinces of the Chesapeake Bay drainage area. Hydrological data during 2005-2009 was used to validate our predictive regression model. We found that the most important factors affected stream flow were the proportions of grassland, developed land, forest, and impervious surface and soil hydrologic group. Land use types had more strongly influence on stream flow than soil physical properties. Flow volume and flow variability decreased where proportion of grassland or forest increased, but increased where proportion of developed land or impervious surface increased. Flow was lower at higher category of soil hydrologic group. Among the 17 hydrological indicators, we found that the number and duration of the flow pulses with high frequency were the best indicators that were correlated to the watershed characteristics at the whole Chesapeake watershed scale; the peak flow, and the number and duration of high flow pulses were the best indicators that were correlated to the watershed characteristics in the Highland area; the number and duration of high flow pulses were the best indicators that were correlated to the watershed characteristics in the Piedmont area; the number and duration of high flow pulses, and the rate and frequency of flow variability were the best indicators that were correlated to the watershed characteristics in the Coastal plain area. Watershed characteristics can be used to predict number of high flow pulses in the whole Chesapeake Bay watershed, Piedmont and Coastal Plain regions.