Watersheds of three streams, the Mousam River, Branch Brook, and Merriland River in southeastern Maine were investigated from 2010 through 2013 under a cooperative project between the U.S. Geological Survey and the Maine Geological Survey. The Branch Brook watershed previously had been deemed at risk by the Maine Geological Survey because of the proportionally large water withdrawals compared to estimates of the in-stream flow requirements for habitat protection. The primary groundwater withdrawals in the study area include a water-supply well in the headwaters of the system and three water-supply wells in the coastal plain near the downstream end of the system. A steady-state groundwater flow model was used to understand the movement of water within the system, to evaluate the water budget and the effect of groundwater withdrawals on streamflows, and to understand streamflow depletion in relation to the State of Maine's requirements to maintain in-stream flows for habitat protection. Delineation of the simulated groundwater divides compared to the surface-water divides suggests that the groundwater divides in the headwater areas do not exactly correspond to the surface-water divides. Under both pumping and non-pumping conditions, groundwater flows from the headwaters of the Branch Brook watershed into the Mousam River watershed. Pumping in the Mousam River watershed captures a small amount of groundwater from the Branch Brook basin. The cumulative effect of groundwater withdrawals on base flows in two rivers in the study area (Branch Brook and the Merriland River) was evaluated using the groundwater flow model. Streamflow depletion in the headwaters of Branch Brook was 0.12 cubic feet per second (ft3/s) for the steady-state simulation, or about 10 percent of the average base flow at that location. Downstream on Branch Brook, the total streamflow depletion from all the wells was 0.59 ft3/s, or 3 percent of the average base flow at that location. In the Merriland River downstream from the Merriland River well, the total amount of streamflow depletion was 0.6 ft3/s, or about 7 percent of the average base flow. The groundwater model was used to evaluate several different scenarios that could affect streamflow and groundwater discharging to the rivers and streams in the study area. The scenarios were (1) no pumping from the water-supply wells; (2) current pumping from the water-supply wells, but simulated drought conditions (25 percent reduction in recharge); (3) current recharge, but with increased pumping from the large water-supply wells; and (4) drought conditions and increased pumping combined. Simulations of increased pumping in the water-supply wells resulted in streamflow depletion in the headwaters of Branch Brook increasing to 16 percent of the headwater base flow. Simulated increases in the pumping in the coastal plain wells increased the amount of streamflow depletion to 6 percent of the flow in Branch Brook and to 8 percent of the flow in the Merriland River. The additional stress of a drought imposed on the model (25 percent less recharge) had a substantial impact on streamflows, as expected. If the simulated drought occurred simultaneously with an increase in pumping, the base flows would be reduced 48 percent in the headwaters of Branch Brook, compared to the no-pumping scenario. Downstream in Branch Brook, the total reduction in flow would be 29 percent of the simulated base flows in the no-pumping scenario, and in the Merriland River, the reduction would be 33 percent of the base flows in the no-pumping scenario. The study evaluated two different methods of calculating in-stream flow requirements for Branch Brook and the Merriland Rivera set of statewide equations used to calculate monthly median flows and the MOVE.1 record-extension technique used on site-specific streamflow measurements. The August median in-stream flow requirement in the Merriland River was calculated as 7.18 ft3/s using the statewide equations but was 3.07 ft3/s using the MOVE.1 analysis. In Branch Brook, the August median in-stream flow requirements were calculated as 20.3 ft3/s using the statewide equations and 11.8 ft3/s using the MOVE.1 analysis. In each case, using site-specific data yields an estimate of in-stream flow that is much lower than an estimate the statewide equations provide.