The Retsof salt mine in upstate New York was flooded from 1994 to 1996 after two roof collapses created rubble chimneys in overlying bedrock that intersected a confined aquifer in glacial sediments. The mine now contains about 60 billion liters of saturated halite brine that is slowly being displaced as the weight of overlying sediments causes the mine cavity to close, a process that could last several hundred years. Saline water was detected in the confined aquifer in 2002, and a brine-mitigation project that includes pumping followed by onsite desalination was implemented in 2006 to prevent further migration of saline water from the collapse area. A study was conducted by the U.S. Geological Survey using geochemical and variable-density flow modeling to determine sources of salinity in the confined aquifer and to assess (1) processes that control movement and mixing of waters in the collapse area, (2) the effect of pumping on salinity, and (3) the potential for anhydrite dissolution and subsequent land subsidence resulting from mixing of waters induced by pumping.

The primary source of salinity in the collapse area is halite brine that was displaced from the flooded mine and transported upward by advection and dispersion through the rubble chimneys and surrounding deformation zone. Geochemical and variable-density modeling indicate that salinity in the upper part of the collapse area is partly derived from inflow of saline water from bedrock fracture zones during water-level recovery (January 1996 through August 2006). The lateral diversion of brine into bedrock fracture zones promoted the upward migration of mine water through mixing with lower density waters. The relative contributions of mine water, bedrock water, and aquifer water to the observed salinity profile within the collapse area are controlled by the rates of flow to and from bedrock fracture zones. Variable-density simulations of water-level recovery indicate that saline water has probably not migrated beyond the collapse area, while simulations of pumping indicate that further upward migration of brine and saline water is now prevented by groundwater withdrawals under the brine-mitigation project. Geochemical modeling indicates that additional land subsidence as a result of anhydrite dissolution in the collapse area is not a concern, as long as the rate of brine pumping is less than the rate of upward flow of brine from the flooded mine.

The collapse area above the flooded salt mine is within a glacially scoured bedrock valley that is filled with more than 150 meters of glacial drift. A confined aquifer at the bottom of the glacial sediments (referred to as the lower confined aquifer, or LCA) was the source of most of the water that flooded the mine. Two rubble chimneys that formed above the roof collapses in 1994 hydraulically connect the flooded mine to the LCA through 180 meters of sedimentary rock. From 1996 through 2006, water levels in the aquifer system recovered and the brine-displacement rate ranged from 4.4 to 1.6 liters per second, as estimated from land-surface subsidence above the mine. A zone of fracturing within the bedrock (the deformation zone) formed around the rubble chimneys as rock layers sagged toward the mine cavity after the roof collapses. Borehole geophysical surveys have identified three saline-water-bearing fracture zones in the bedrock: at stratigraphic contacts between the Onondaga and Bertie Limestones (O/B-FZ) and the Bertie Limestone and the Camillus Shale (B/C-FZ), and in the Syracuse Formation (Syr-FZ). The only outlets for brine displaced from the mine are through the rubble chimneys, but some of the brine could be diverted laterally into fracture zones in the rocks that lie between the mine and the LCA.

Inverse geochemical models developed using PHREEQC indicate that halite brine in the flooded mine is derived from a mixture of freshwater from the LCA (81 percent), saline water from bedrock fracture zones (16 percent), and an hypothesized bromide-rich brine (3 percent) assumed to originate from salt-bearing rocks above the flooded mine. Geochemical modeling results also indicate that halite brine entering the rubble chimneys is diluted by both bedrock water and aquifer water, and that water from the mine has not reached the bedrock surface. Forward geochemical models indicate that additional land subsidence could occur if pumping from the brine-mitigation project were to introduce either freshwater or bedrock water that is undersaturated with respect to anhydrite into the lower part of the rubble chimneys. In this unlikely scenario, the maximum subsidence rates are predicted to range from 0.6 to 1.1 centimeters per year—subsidence rates would be lower (0.1 to 0.6 centimeters per year) if ion-exchange reactions affect the water chemistry.

Variable-density, transient groundwater-flow models were constructed using SEAWAT to simulate the movement of saline water, aquifer water, bedrock water, and brine within the rubble chimneys and surrounding deformation zone during the 10.7-year period following flooding of the salt mine. Two three-dimensional models reproduced the profile of halite saturation with depth measured in September 2006 reasonably well, and neither model indicated that saline water had migrated beyond the collapse area. The models differed in the number of fracture zones represented: one zone in model A (O/B-FZ) and three zones in model B (O/B-FZ, B/C-FZ, and Syr-FZ). It is unknown whether model A or model B better represents current conditions because the lateral extents of the B/C-FZ and Syr-FZ have not been delineated beyond the collapse area.

In model A, the salinity of water in the upper part of the rubble chimneys is derived mainly from the inflow of bedrock water from the O/B-FZ, as indicated by geochemical models. Bedrock water that was pushed upward by brine during the 10.7-year simulation period formed a diffuse front above a nearly horizontal brine level in both chimneys. In model B, some of the salinity in the upper part of the rubble chimneys is derived from mine water. The rate of bedrock-water inflow from the O/B-FZ was lower in model B than in model A, and mixing with waters from the Syr-FZ and B/C-FZ transported mine water higher in the water column than in model A. Simulated brine levels in both chimneys sloped northward, reflecting lateral diversion of brine into the B/C-FZ, and less aquifer water was displaced from the collapse area than in model A.

Models A and B were used to simulate changes in water levels and salinity produced by pumping for the brine-mitigation project from September 2006 through February 2008. Both simulations indicated that current pumping rates are sufficient to offset upward migration of brine and saline water through the collapse area and, therefore, to further prevent contamination of the LCA. A greater decrease in salinity was simulated in model B, however, because the porosity of the rubble chimneys was lower (6 percent compared to 10 percent in model A), and some brine and saline waters were diverted through the B/C-FZ. Model B better simulates the influent saturation to the desalination plant, the amount of halite produced, and the observed declines in saturations than model A, which is more consistent with results of geochemical modeling. Sensitivity analyses indicate that the actual brine-displacement rate could be lower than estimated because simulated declines in saturations underpredict the observed decline from September 2006 through February 2008.

Although halite saturations within the upper part of the collapse area are predicted to decrease with continued pumping, brine displacement from the flooded mine is expected to continue for hundreds of years. Simulations of a shutdown of the brine-mitigation project indicate southward migration of saline water through the LCA, extending 700 meters to the model boundary within 10 years. Continued migration of saline water would eventually form a pool in the LCA in a bedrock depression 8 kilometers south of the collapse area near Sonyea, but the large relative density of the saline water would likely prevent it from reaching overlying aquifers. Simulations also indicate that brine will migrate through bedrock fracture zones—some brine could possibly emerge updip to the north where the subcrop area of the Bertie Limestone intersects the bedrock surface near Avon, but the projected time of travel is unknown.