Simulated Effects of Salt-Mine Collapse on Ground-Water Flow and Land Subsidence in a Glacial Aquifer System, Livingston County, New York

By Richard M. Yager, Todd S. Miller, and William M. Kappel
Professional Paper 1611

Prepared in cooperation with the Livingston County Department of Health

We have an html version of the abstract and table of contents below.  The full report is available in Portable Document Format (pdf).  Links to the pdf.



The bedrock ceiling in parts of the Retsof salt mine in the the Genesee Valley, Livingston County, N.Y. collapsed on March 12, 1994 and water from overlying aquifers began to flow into the mine at a rate of 5,500 gal/min (gallons per minute), and the rate increased to as much as 20,000 gal/min after a second collapse 3 weeks later in April 1994. Efforts to save the mine were abandoned by the end of 1994, and the mine was completely flooded by January 1996.

The Genesee Valley (including the tributary Canaseraga Valley) is about 40 miles long and contains as much as 750 feet of glacial sediment and recent alluvium, which rest on mostly carbonate bedrock along the valley axis near and north of the collapse site, and on shale in areas farther south. The mined salt bed lies within Silurian shale and is overlain by 600 feet of Devonian shale and limestone. Rock-rubble zones and fractures that formed within units overlying the collapse resulted in about 45 feet of subsidence at the bedrock surface and the formation of sinkholes as much as 70 feet deep at land surface.

The glacial aquifer system in the Genesee Valley generally consists of an unconfined alluvial aquifer (upper) and two confined glacial aquifers (middle and lower) separated by two confining layers. The underlying carbonate bedrock also contains water-bearing fracture zones in which water flowed northward before the collapse. The salinity of ground water generally increases with depth; specific conductance ranges from 430 to 1,390 µS/cm (microsiemens per centimeter) in the middle aquifer and from 2,980 to 39,300 µS/cm in the lower aquifer. Concentrations of tritium in the lower aquifer indicate the water to be generally more than 45 years old.

Subsidence at land surface ranges from 15 feet over some parts of the mine (through shrinkage of the mine cavity in response to deformation of salt beds) to more than 70 feet over two areas within the mine, and is as much as 0.8 ft outside the mined area as a result of compaction of fine-grained sediments in the confining layers. By January 1996, when the mine was completely flooded, drawdowns in the lower aquifer were as much as 400 feet near the collapse area, as much as 50 feet in an area 7 miles to the north, and as much as 135 feet in an area 8 miles south of the collapse area. The northward direction of ground-water flow in the confined aquifers north of the collapse was reversed, and five domestic wells 7 miles north of the collapse began to produce saline water from the lower aquifer, apparently because the drawdown either induced the inflow of saline water from bedrock, or allowed saline water from further north to flow southward. Ratios of chloride to bromide concentrations in ground-water samples indicate two sources of salinity in the lower aquifer: (1) saline water from carbonate bedrock and (2) halite-bearing water from an unknown source. Both thermogenic and biogenic gases (methane and sulfide) were produced by wells screened in confined aquifers as a result of exsolution as water-level declines lowered the pore pressure.

A five-layer ground-water flow model based on generalized geologic sections was used to simulate flow conditions within the unconsolidated sediments before and after the mine collapse. The model represents a 61.3 square-mile section of the Genesee Valley that extends from the Village of Avon in the north to the Village of Dansville, 40 miles in the south. Nonlinear regression was used to calibrate the model to (1) assumed water-table altitudes and estimated base flows in steady-state simulations representing precollapse conditions, and (2) measured water levels in wells and estimated ground-water discharges to the mine in transient-state simulations that represent the flooding of the mine and water-level recovery thereafter.

The computed water budget indicated that ground water released from storage in the lower aquifer accounted for about 11 percent of the total inflow to the model, and most of the water discharged to the mine. Values of specific storage estimated by regression for the two confined aquifers (2.9 x 10-4 and 6.9 x 10-5 ft-1) were much larger than values typically estimated for other sand and gravel aquifers affected by land subsidence (1 x 10-6 to 3 x 10-6 ft-1); this could be explained by the exsolution of natural gas (methane) from ground water during the period of drawdown; exsolution also could account for part of a bias in the head distribution computed in model simulations. A gas-partitioning equation developed to compute specific storage in the presence of a gas phase indicates that specific storage was not constant, as assumed in the flow model, but varied temporally and spatially as a function of pressure. Computations of specific storage in the lower aquifer near the collapse area, where the maximum drawdown occurred, included the effects of gas exsolution and ranged from 2.4 x 10-5 to 1.8 x 10-4 ft-1; the maximum pore volume occupied by gas was estimated to be 7 percent. Including (1) gas exsolution and (2) aquifer compressibility typically associated with aquifer-system compaction, decreased model error by about 15 percent and slightly reduced model bias.

Hydraulic heads computed in 12.4-year transient-state simulations indicate that water levels will return to precollapse conditions by about the year 2006, but the recovery would be completed by 1999 if the specific storage were smaller (2.3 x 10-6 ft-1). Ground-water flow paths computed from model results indicate that the reversal of the northward hydraulic gradient in the northern part of the valley allows saline water to migrate southward toward the collapse at a velocity of 1 to 2 ft/d (feet per day). The time between the collapse and the onset of salinity in two domestic wells indicates a greater velocity (8 ft/d), however, and suggests that the saline water could have traveled along discrete flowpaths with a larger hydraulic conductivity (2,000 ft/d) than the 300 ft/d estimated by regression for the lower aquifer. The present southward gradient (1.2 x 10-3 ft/ft) is about 3 times larger than the northward gradient (3.9 x 10-4 ft/ft) that will develop once the natural hydraulic gradient is reestablished. The calculated northward velocity indicates that, if saline water migrates southward for 10 years after the mine collapse, about 30 years will be required for freshwater to flush the area into which the saline water has intruded.

Transient-state simulations with one- and three-dimensional models were conducted to estimate the extent of land subsidence due to dewatering of fine-grained sediments outside the mined area, where subsidence measurements were not made. The one-dimensional models represent vertical flow and compression in the upper and lower confining layers at two locations--one near a borehole with pressure transducers screened in the upper confining layer near the collapse area, and the second at a survey monument 2,800 ft to the south. Vertical hydraulic conductivity of the upper confining layer, and specific storage of both confining layers, were estimated through nonlinear regression from observations of pore pressure and subsidence. Both elastic and inelastic storage were represented in both confining layers. Elastic storage was assumed, from an observed relation between compressibility and depth computed from consolidation tests of confining-layer sediments, to be inversely proportional to effective stress (or depth); inelastic storage was assumed, from published values, to be 30 times greater than elastic storage. Computed pressure changes in the upper confining layer near the collapse were in close agreement with measured pressures changes, and the maximum computed subsidence was about 94 percent of the observed subsidence (0.78 feet). About 92 percent of the computed subsidence was the result of drainage and compression in the lower confining layer; 90 percent of this compression was estimated to be inelastic, nonrecoverable compaction. Representing drainage of confining-layer sediments by elastic storage gave pressure and maximum-subsidence values that matched observed values as closely as those based on inelastic storage; elastic storage also simulated land-surface rebound after water-level recovery, which had not occurred by 1999. Hydraulic properties of the confining-layer sediments estimated by the one-dimensional models that represented only elastic compression were incorporated in the three-dimensional model and resulted in close agreement between computed subsidence and subsidence measured along a survey line. Simulated subsidence closely matched the actual subsidence in Mt. Morris, 3 mi south of the collapse, where as much as 0.3 ft of subsidence was measured. Model results indicate that at least 0.1 ft of subsidence occurred by February 1996 over the 16 square-mile area that extended 5 miles north of the collapse area and 7 miles south of the collapse.





Purpose and scope

Previous studies

Methods of investigation

Seismic surveys

Water-level and streamflow measurements

Ground-water quality


Geologic setting

Bedrock units

Silurian units

Devonian units

Glacial and postglacial geology

Erosion of bedrock by ice

Unconsolidated deposits

Subglaciofluvial sand and gravel

Englacial and glaciolacustrine sediments

Supraglacial sand and gravel

Glaciolacustrine and deltaic sediments and till

Alluvial sediments

Hyrdologic setting

Aquifers and confining units

Upper aquifer

Upper confining layer and deltaic deposits

Middle aquifer

Lower confining layer

Lower aquifer

Bedrock aquifers

Ground-water flow patterns before the collapse

Upper aquifer

Middle and lower aquifers

Bedrock aquifers

Ground-water quality


Isotopic composition

Natural gas

Effects of mine collapse and flooding

Collapse of mine ceiling

Land subsidence

Surface water

Ground-water hydrology

Ground-water levels

Ground-water quality

Natural gas exsolution

Simulation of ground-water flow

Model design

Model layers and grid

Boundary conditions

Model layer 1

Model layers 3 and 5

Model layers 2 through 5

Model calibration

Calibration procedure

Steady-state simulations

Transient-state simulations

Precollapse conditions

Head distribution

Recharge and discharge

Aquifer properties

Postcollapse conditions (flooding and water-level recovery)


Water budget

Estimates of aquifer properties

Model bias

Model results

Effect of natural gas on specific storage

Water-level recovery

Effect of mine collapse on saline water intrusion

Simulation of land subsidence

One-dimensional simulations

Model design

Model calibration

Three-dimensional simulation


Hydrogeologic setting

Effects of mine collapse

Simulation of ground-water flow

Simulation of land subsidence

References cited





This report is presented in Portable Document Format (PDF). To view and print report you need to use Adobe Acrobat Reader, available as freeware.


Users with visual disabilities can visit this site for conversion tools and information to help make PDF files accessible.


Report in PDF format  (8.9MB)
      Companion Professional Paper 1767
      Companion Open-File Report 2013-1174


A printed bound copy of this report is available from:
USGS Information Services (Map and Book Sales)
Box 25286
Denver Federal Center
Denver, CO 80225


FirstGov button  Take Pride in America button