Salt-front movement in the Hudson River estuary, New York — Simulations by one-dimensional flow and solute-transport models

The Hudson River is being considered for use as a supplemental source of water supply for New York City during droughts. One proposal entails withdrawal of Hudson River water from locations near Newburgh, Chelsea, or Kingston, but the extent to which this could cause the salt front to advance upstream to points where it could adversely affect community water supplies is unknown. The U.S. Geological Survey (USGS) one-dimensional Branch-Network Dynamic Flow model (BRANCH) was used in conjunction with the USGS one-dimensional Branched Lagrangian Solute-Transport Model (BLTM) to simulate the effect of five water-withdrawal scenarios on the salt-front location.

The modeled reach contains 132 miles of the lower Hudson River between the Federal Dam at Troy and Hastings-on-Hudson (near New York City). The BRANCH model was calibrated and verified to 19 tidal-cycle discharge measurements made at 11 locations by conventional and acoustic Doppler current-profiler methods. Maximum measured instantaneous tidal flow ranged from 20,000 ft³/s (cubic feet per second) at Albany to 368,000 ft³/s at Tellers Point; daily-mean flow at Green Island near Troy ranged from 3,030 ft³/s to 45,000 ft³/s during the flow measurements. Successive ebb- and flood-flow volumes were measured and compared with computed volumes; daily-mean bias was -1.6 percent (range from -21.0 to +23.7 percent; 13.5 percent mean absolute error). Daily-mean deviation between simulated and measured stage at eight locations (from Bowline Point to Albany) over the 19 tidal-cycle measurements averaged +0.06 ft (range from -0.31 to +0.40 ft; 0.21 ft root mean square error, RMSE). These results indicate that the model can accurately simulate flow in the Hudson River under a wide range of flow, tide, and meteorological conditions.

The BLTM was used to simulate chloride transport in the 61-mi reach from Turkey Point to Bowline Point under two seasonal conditions in 1990.one representing spring conditions of high inflow and low salinity (April-June), the other representing typical summer conditions of low inflow and high salinity (July-August). Measured chloride concentrations at Bowline Point were used to drive the BLTM simulations, and data collected at West Point were used for calibration. Mean bias in simulated chloride concentration for the April-June 1990 (high flow) data (observed range from 12 to 201 mg/L [milligrams per liter]; 30 mg/L RMSE) was .16 mg/L, and mean bias for the July-August 1990 (low flow) data (observed range from 31 to 2,000 mg/L; 535 mg/ L RMSE) was +126 mg/L. The salt front (saltwater/ freshwater interface) on the Hudson River was defined as the furthest upstream location where the chloride concentration exceeded 100 mg/L. Data from August 1991 were used to evaluate solute transport between West Point and Poughkeepsie because a chloride concentration of 100 mg/L was not observed at Clinton Point in 1990. The BLTM then was used to simulate chloride concentrations at Chelsea Pump Station and Clinton Point. Regression equations, based on daily mean values of specific conductance measured at West Point, were used to estimate daily mean chloride concentrations at Chelsea Pump Station and Clinton Point for model analysis. Mean biases in BLTM-simulated daily mean chloride concentrations for August 1991 were .38 mg/L at Chelsea Pump Station (range from 189 to 551 mg/L; 103 mg/L RMSE) and .9 mg/L at Clinton Point (range from 53 to 264 mg/L; 62 mg/L RMSE).

Hypothetical withdrawals at (1) Newburgh, (2) Chelsea, (3) Chelsea and Newburgh, (4) Chelsea and Kingston, and (5) Kingston and Newburgh, were simulated to compute the effects of withdrawals on salt-front movement. Withdrawals of 300 Mgal/d from any combination of Chelsea or Newburgh could result in upstream movement of the salt front of as much as 1.0 mi, given an initial salt-front location between West Point and Rogers Point. Scenarios that included withdrawals at Kingston caused the greatest upstream salt-front movement. Simulation of a 90-day April-June high-flow period during which discharges at Green Island averaged 25,200 ft³/s indicated that withdrawals of 1,939 Mgal/d (million gallons per day) at Chelsea Pump Station would not measureably increase chloride concentrations at Chelsea Pump Station under normal tidal and meteorological conditions, but withdrawals at twice that rate (3,878 Mgal/d) could increase the chloride concentration at Chelsea Pump Station to 250 mg/L.