An Evaluation of Sand Lens Connectivity in a Glacial Till at the Niagara Falls Storage Site

 

By William T. Frederick, CPG, U.S. Army Corps of Engineers, Buffalo District; Don DeMarco, HydroGeoLogic, Inc.; and Eric Evans, HydroGeoLogic, Inc.

 

Introduction to Site Problem

 

The Niagara Falls Storage Site (NFSS) in Lewiston, NY, is a 191-acre Formerly Utilized Sites Remedial Action Program (FUSRAP) facility where low-level radioactive residues including radium, thorium and uranium are stored in an interim waste containment structure (IWCS).  Currently the U.S. Army Corps of Engineers (USACE) are focused on evaluating remedial actions for the site, and as part of this effort a groundwater flow and transport model is being developed to help with the evaluation of remedies.  During IWCS construction in 1986, residues were slurried into the reinforced basement foundations of decommissioned site buildings and separated from the hydrologic environment by a multi-layer clay cap and circumferential clay dike keyed into a laterally continuous confining unit designated the Glacio-Lacustrine Clay (GLC).  The IWCS is situated above the GLC and entirely within a local stratigraphic unit commonly designated the Brown Clay Till (BCT).  The BCT  is approximately 15’ thick and is the uppermost unit of a 40’ multi-layer glacial complex blanketing the surface of the Queenston Formation shale.  The BCT is characterized as a clayey till with intermittent and discontinuous sand lenses that average 2’ in thickness; these sand lenses are most prevalent in the lower third of the unit.

 

Three-dimensional structure maps of the glacial sediments and fractured rock interface were developed using boring logs from the NFSS and adjacent properties.  Based on a preliminary evaluation of these maps, a hypothesis was developed that the sand lenses within the BCT may be spatially continuous across the site.  Due to the potential for correlated sand lenses to provide preferential pathways for groundwater flow and contaminant migration, a more thorough analysis was deemed necessary.  Results of this analysis were also needed to evaluate the adequacy of the site conceptual model, which forms the basis for ongoing groundwater flow and solute transport modeling.  The initial groundwater flow and solute transport model was developed based on the assumption that the BCT can be represented, in bulk, as a single hydrostratigraphic unit and represented by a single model layer with localized lateral heterogeneity.  The existence of laterally continuous sand lenses within the BCT infer that there are thin zones of preferential groundwater flow, which may necessitate the need to adjust the groundwater modeling approach.

 

Technical Approach

 

Given the implications for solute movement through connected sand lenses, a hydrogeologic and geostatistical analysis was performed to assess sand lens connectivity within the BCT.  The effort was organized into three phases, including:  1) an analysis of lithologic characteristics described in boring logs; 2) a geostatistical analysis of sand-lens occurrences; and 3) an evaluation of hydrographs from wells in the brown till.  A comprehensive environmental database was used to support the analyses.

 

Lithologic Analysis

 

Initially, a qualitative evaluation of lithologic logs collected from borings in the vicinity of the IWCS was conducted to determine whether the physical characteristics of the sand lenses, such as color, thickness, or elevation, could be used to correlate the sand lenses across boreholes.  Geologic descriptions were evaluated to identify stratigraphic correlations, using lithologic data primarily from proximal borings.  The results of this analysis indicated that the lithologic characteristics of sand lenses observed in proximal boreholes were occasionally similar; however, in many of these cases there was no correlation in the elevation and/or thickness of the sand lenses.  Conversely, several proximal borings were found to have sand lenses that are present at similar elevations but are not consistent in terms of the lithologic characteristics.   To supplement the lithologic data collected from boreholes, geotechnical logs recorded during the excavation of a vertical cut-off wall surrounding the IWCS were evaluated.  This cut-off wall completely surrounds the IWCS on four sides.  The lithologic descriptions of the BCT within the excavations indicate that the sand lenses are highly heterogeneous and they vary in color, texture, elevation, and thickness.    Moreover, sand lenses do not appear similar between the east and west or north and south cut-off excavation walls; nor do sand lenses appear similar between adjacent excavation walls.  These combined geologic data strongly suggest that the sand lenses within the BCT are discontinuous features.

 

Geostatistical Analysis

 

In addition to the qualitative evaluation of lithologic information, a semivariogram analysis was conducted to quantify the vertical and horizontal continuity of sand lenses.  This geostatistical technique can be used to characterize and describe the spatial correlation of phenomena that are spatially distributed and in the case of the BCT sand lenses, determine whether potential preferential flow or contaminant migration pathways may exist.

 

The semivariogram analysis was conducted by first developing experimental semivariograms from a set of spatially distributed data and secondly by attempting to fit mathematical semivariogram models to the experimental semivariograms.  Experimental semivariograms are developed using the following equation:

 

where:    g              =              semivariogram;

                h              =              separation distance between data points;

                n              =              number of data pairs separated by separation distance, h;

                Z             =              value of the spatially distributed phenomena; and

                x              =              location of an individual data point.

 

Spatially distributed points separated by a given distance are grouped into data pairs that are partitioned into a range of separation distances commonly referred to as lag intervals.  For each lag interval, one semivariogram value is calculated.               

 

Constructing a semivariogram requires a sufficient number of data that also produce a meaningful semivariogram with a structure. If there is no structure in the semivariogram, then the data are considered to be spatially uncorrelated and geostatistical interpolation is not warranted.  The experimental semivariograms are calculated as a function of the spatial separation vector (the distance and direction between each pair of estimation data).  Spatial correlation that is independent of direction indicate the data are isotropic, whereas directional semivariograms describe the relationship between data pairs oriented in a specified direction .  Directional semivariograms are used to determine whether the spatially distributed data are anisotropic. 

 

Prior to initiating the semivariogram analysis, lithologic data from 100 boreholes on the NFSS were queried from the database and compiled with borehole coordinates, elevation, and binary representation of sand lens occurrence.  This binary conversion of the lithologic data, where a 1 represents the presence of a sand lens and 0 represents its absence, is referred to as indicator transformation.  Semivariogram analyses were conducted using the indicator transformed data.  Experimental semivariograms were then prepared for the vertical and horizontal directions. 

 

The semivariogram analysis conducted to evaluate spatial continuity in the vertical direction indicated that the vertical correlation length of sand lenses within the BCT is approximately 4 to 5 feet.  The vertical variogram, shown below, illustrates the correlation structure of the data and indicates that the sand lenses are highly correlated at short separation distances and are poorly correlated at larger separation distances.  In addition, it appears that the experimental semivariogram would cross the y-axis near the origin of the plot.  This indicates that there is little “nugget effect”, which often represents variability that occurs at a scale smaller than the distance separating samples and/or measurement uncertainties. 

                     

 

An aerial analysis was performed using an iterative method that initially assumed an isotropic spatial distribution of sand lenses.  The resulting isotropic aerial variogram is characterized by a high degree of variability and a large nugget effect.  The large nugget effect indicates that the lithologic data are not strongly correlated at small separation distances.  This poor correlation suggests that the sand lenses are not spatially correlated over the distances that separate the most closely spaced borings.  Based on the relatively poor isotropic semivariogram, it appears that the aerial correlation length is between 15 to 20 feet.  Anisotropic semivariograms were also conducted to evaluate whether the sand lenses are more continuous in specific directions.  A spatial correlation structure was not observed in any of the directional semivariograms.  Consequently, the directional semivariogram analyses did not indicate that the sand lenses are more continuous in specific directions. 

 

Hydrographic Analysis

 

An additional analysis measure evaluated the hydraulic response in wells screened across sand lenses in comparison to wells that are not screened across sand lenses.  The premise behind this analysis was that hydraulic responses and water-level elevations in spatially continuous sand lenses may differ considerably from the surrounding finer grained, poorly sorted sediments.  Well hydrographs and water-level summary statistics were prepared and evaluated to support this analysis

 

A comparison of hydrographs from wells screened across sand lenses indicates that there are marked differences in hydraulic head between proximal wells and in the variability of hydraulic head at proximal locations; this suggests a lack of hydraulic connection between sand lenses.  Potentiometric contours generated using hydraulic head data from sand lens wells are generally coincident with contours generated from wells without sand lenses in their screened interval.  The general agreement between the contour sets indicate the sand lenses do not comprise an independent flow zone but a component in the BCT that are governed by the ambient flow in the BCT.  Finally, the average standard deviation of hydraulic head for wells without sand lenses along the screened interval was calculated to be 3.2 feet, whereas it was 3.0 feet for wells with sand lenses.  This indicates that the transient water level responses are similar in wells screened within sand lenses and screened in finer grained sediments.

 

Results of Sand Lens Analyses

 

The three-phase analysis conducted on the hydrogeologic data from the NFSS provided the following results.

 

 

 

 

These data evaluations together prove the sand lenses are random and discontinuous but individually may act as small-scale preferential flow pathways limited to 20 feet in length.

 

Implications for Groundwater Modeling

 

The potential correlation of the sand lenses in the BCT would have required additional characterization to defensibly model their extent and interconnection.  Additional site data would include confirmatory drilling to define statistically derived sand-lens extents and an optimized groundwater monitoring well array to define preferential transport in sand lenses near the IWCS.  The site numerical model would have been greatly complicated with finely discretized areas where sand lenses exhibited both horizontal and vertical continuity, which would have increased both costs and model uncertainty.         

 

Since the geostatistical and hydrogeologic analyses indicate that the sand lenses are not interconnected, additional discretization of the BCT is not required.  The complexity of the NFSS-area groundwater model will reflect the site-scale hydrogeology and will defensibly evaluate the remedial alternatives in terms of effectiveness and risk reduction.  A MODFLOW-SURFACT model will simulate the BCT as a single vertically- integrated layer with lateral variations in hydraulic conductivity that account for localized areas of disconnected sand lenses.