ABSTRACT

Groundwater in the Newark basin aquifer flows primarily through discrete water-bearing zones parallel to the strike and dip of bedding, whereas flow perpendicular to the strike is restricted, thereby imparting anisotropy to the groundwater flow field. The finite-element model SUTRA was used to represent bedrock structure in the aquifer by spatially varying the orientation of the hydraulic conductivity tensor to reflect variations in the strike and dip of the bedding. Directions of maximum and medium hydraulic conductivity were oriented parallel to the bedding, and the direction of minimum hydraulic conductivity was oriented perpendicular to the bedding. Groundwater flow models were prepared to simulate local flow in the vicinity of the Spring Valley well field and regional flow through the Newark basin aquifer.

The Newark basin contains sedimentary rocks deposited as alluvium during the Late Triassic and is one of a series of basins that developed when Mesozoic rifting of the super continent Pangea created the Atlantic Ocean. The westward-dipping basin is filled with interbedded facies of coarse-grained to fine-grained rocks that were intruded by diabase associated with Jurassic volcanism. The Newark basin aquifer is bounded to the north and east by the Palisades sill and to the west by the Ramapo Fault. Although the general dip of bedding is toward the fault, mapping of conglomerate beds indicates the rocks are folded into broad anticlines and synclines. An alternative, more uniform pattern of regional structure, based on interpolated strike and dip measurements from a number of sources, has also been proposed. Two groundwater flow models (A for the former type of bedrock structure and B for the latter type) were developed to represent these alternative depictions of bedrock structure.

Transient simulations were calibrated to reproduce measured water-level recoveries in a 9.3 mi² area surrounding the Spring Valley well field during a 5-day aquifer test in 1992. The models represented a 330-ft thick rock mass divided vertically into 10 equally spaced layers and were calibrated through nonlinear regression. Results of model B best matched the observed water-level recoveries with an estimated hydraulic conductivity of 9.5 ft/day, specific storage of 7.6 x 10^–6 ft^–1, and K_max: K_min anisotropy ratio (hydraulic conductivity parallel to bedding: perpendicular to bedding) of 72:1. Model error was 50 percent greater in model A because the assumed structure did not match the actual strike of bedding in this area.

Steady-state simulations of regional flow through the 85.4-mi² modeled extent of the Newark basin aquifer represented both the alluvial aquifer beneath the Mawah River and the fractured bedrock. The rock mass was divided into two aquifer units: an upper 500-ft thick unit divided into 10 equally spaced layers through which most groundwater is assumed to flow and a lower unit divided into 7 layers with increasing thickness. Models were calibrated through nonlinear regression to average water levels measured in 140 wells from August 2005 through April 2007. Water levels simulated using the two models were similar and generally matched those observed, and the average recharge rate estimated using both models was 19 inches/year for the simulated period. Estimated transmissivity parallel to the strike of bedding (1,100 ft²/d) was uniform in two transmissivity (T) zones in model A, but in model B the transmissivity of a high T zone (1,600 ft²/d), delineated on the basis of aquifer test data, was slightly greater than in a low T zone (1,300 ft²/d). The K_max: K_min anisotropy was estimated to be 58:1 in model A and 410:1 in model B, so the proportion of flow perpendicular to bedding is less in model B than in model A.

Distributions of groundwater age simulated with models A and B are similar and indicate that most shallow ground-water (225 ft below the bedrock surface) is 5 to 20 years old, with younger water (5 years or less) near upland recharge areas and older water (more than 100 years) in lowland discharge areas near the Hackensack River and Saddle River. The two simulated distributions differ in some areas where younger water is simulated with model A. Effective porosity (2 x 10^–2) was estimated by comparing simulated groundwater ages with those calculated previously on the basis of tritium/helium (³H/³He) dating. Well-field capture zones delineated for major well fields are generally elongated parallel to the assumed strike of bedding and are affected by the simulated potentiometric surface, the drainage network of stream channels, and capture zones of adjacent well fields. Sizes of capture zones range from 1.9 mi² (1,200 acres) for well fields with the largest average withdrawal rates (1,300 gal/min) to less than 0.04 mi² (25 acres) for well fields with the smallest rates (80 gal/min). Base flows in Pascack Brook and Nauraushaun Brook probably have been decreased by groundwater withdrawals because capture zones of several well fields overlap these watersheds.

The effects of annual and monthly changes in recharge and groundwater withdrawals on water levels and discharges to streams were assessed through transient flow simulations of two periods: a 47-year period from January 1960 through December 2006 and a 3-year period from January 2000 through December 2003. The simulations included an order-of-magnitude range in specific storage (S_s) and both saturated and variably saturated conditions. Monthly water levels simulated with the smallest S_s value (7.6 x 10^–6 ft^–1 from simulation of the Spring Valley water-level recovery test) fluctuated by about 15 ft in lowlands and 85 ft in uplands, and by about 3 ft and 20 ft using the largest S_s value, whereas measured water-levels fluctuated by 7 ft and 13 ft, respectively. The hydrographs of simulated annual groundwater levels indicate a trend of decreasing groundwater levels from 1960 through 2006 in upland areas where water levels decreased by 15 to 20 ft. Although the groundwater flow models described in this report are sufficiently accurate to estimate the water budget and delineate capture zones of pumped wells, smaller-scale higher-resolution models would be needed to more accurately simulate contaminant movement, interference between adjacent wells, and local effects of pumping on stream discharge.

First posted February 2011