Nitrate in a Small South-Central Pennsylvania Basin

Analysis of Nitrate Concentrations from Piezometers and Monitoring Wells in Big Spring Run, Pennsylvania, 2012-16

Introduction

Big Spring Run is a small basin in south-central Pennsylvania that encompasses roughly 1.7 square miles in the western portion of Muddy Run-Mill Creek watershed in Lancaster County (see maps below).

Location of Big Spring Run, Pennsylvania (U.S. Geological Survey, 2019).

Explore the Big Spring Run study area using the map.

Streams in Big Spring Run flow north into Mill Creek, which flows west and joins the Conestoga River. The Conestoga River flows southwest into the Susquehanna River, which flows southeast into the Chesapeake Bay. The map below illustrates how Mill Creek eventually reaches the Chesapeake Bay.

The  U.S. Geological Survey's Streamer website  was used to create a map showing how Mill Creek eventually reaches the Chesapeake Bay (follow red line and orange arrows on map).

Sediments deposited across the landscape as a result of widespread mill pond dam (see image below) construction in the Piedmont region of southeast Pennsylvania in the 1800’s have been shown to contribute to modern day sediment and nutrient loads in tributaries that eventually drain into the Chesapeake Bay (Walter and Merritts, 2008).

Example of a mill pond dam in Chiques Creek, Lancaster County, Pennsylvania. From the  Business for Water Stewardship, 2021 .

Lancaster County is an area widely impacted by mill pond dams and resultant legacy sediments (see image below), and the county has thus, been identified as a contributor to run-off affecting water-quality in the Chesapeake Bay.

Lancaster County is an area widely impacted by mill pond dams and resultant legacy sediments (see image below), and the county has thus, been identified as a contributor to run-off affecting water-quality in the Chesapeake Bay.

Locations of historic mill pond dams in Chester, Lancaster, and York counties, Pennsylvania. From Walter and Merritts, 2008.

The county’s small, rural streams, such as Big Spring Run, are not an exception, as this basin has been the subject of numerous studies regarding legacy sediments and nutrients in surface water and groundwater due to a mill pond dam that had been erected in the mid-1700's, causing stream incision (see image below), bank formation, and bank retreat that result in high sediment yields (Galeone and others, 2006; Walter and Merritts, 2008; Hartranft and others, 2011; Weitzman and Kaye, 2017; Langland and others, 2020).

Legacy sediment deposits in Big Spring Run prior to sediment removal and stream restoration efforts. From Hartranft and others, 2011.

Background

With the purpose of exposing previously buried wetlands and reconnecting hydrologic features to the original floodplain, legacy sediments were removed from a portion of Big Spring Run in 2011 (see image below).

Approximately six months post-restoration image of Big Spring Run basin, Pennsylvania. Taken by Dan Galeone on 3/20/2012.

To date, Big Spring Run has been the subject of numerous studies regarding legacy sediments and nutrients in surface water and groundwater (Galeone and others, 2006; Walter and Merritts, 2008; Hartranft and others, 2011; Weitzman and Kaye, 2017). The most recent study by Langland and others (2020) of the U.S. Geological Survey (USGS) involved a concentrated monitoring approach at three surface-water sites within the basin to quantify nutrient contributions in pre- and post- restoration periods, and the study found that nitrate concentrations at the surface-water sites were not significantly different despite the changes that had occurred in the basin (see images below to view changes over time in Big Spring Run). 

Time lapse of changes in Big Spring Run from 2008-19 using Google Earth.

Up to the present time, post-restoration trends in nitrate concentrations in groundwater samples collected seasonally (2 to 4 times a year from September 2012 to April 2016) from 18 piezometers and 28 shallow monitoring wells have not yet been explored or quantified.

Approximately two years and three months post-restoration image of Big Spring Run basin, Pennsylvania. Taken by Eliza Gross on 1/14/2014.

Purpose

The goal of this analysis to explore and analyze nitrate concentrations in water-quality samples collected seasonally for approximately 3 years (September 2012 to April 2016) from 18 piezometers and 28 shallow monitoring wells after legacy sediment removal in 2011 from Big Spring Run, Lancaster County, Pennsylvania.

A shallow monitoring well in Big Spring Run basin, Pennsylvania. Taken by Eliza Gross on 1/14/2014.

Data Description

The USGS, in cooperation with the Pennsylvania Department of Environmental Protection and in collaboration with Franklin and Marshall College and the U.S. Environmental Protection Agency (EPA), collected 958 water-quality samples from 18 piezometers and 28 monitoring wells in Big Spring Run between September 2012 and April 2016 as part of a post-restoration monitoring effort. These sites are located at the northern, most upstream part of the basin (see map below).

Explore Big Spring Run sampling site locations using the map (basin outline from U.S. Geological Survey, 2019).

Nitrate samples were analyzed by the EPA’s Groundwater Characterization and Remediation Division at its Robert S. Kerr Environmental Research Center in Ada, Oklahoma. Results were provided to the USGS as memorandum laboratory data reports with each report in a separate digital data table (U.S. Environmental Protection Agency, various dates). Results are tied to USGS groundwater sites that were established in the National Water Information System (NWIS; U.S. Geological Survey, 2021), but results provided by the EPA are not available from NWIS.

Nitrate concentrations in samples ranged from less than 0.008 to 25.5 milligrams per liter (mg/L) and had a median of 6.285 mg/L (see figure below). Sample results included 27 nitrate concentrations that were reported as less than the method detection limit (MDL), which ranged from 0.008 to 0.022 mg/L for these samples. Because samples had nitrate concentrations below several MDLs (0.008, 0.009, 0.016, and 0.022 mg/L), the highest MDL of 0.022 mg/L was selected for censoring. As a result, 5 reported nitrate concentrations ranging from 0.009 to 0.018 mg/L and the 27 nitrate concentrations reported as less than the associated MDL were censored to the highest MDL of 0.022 mg/L. For this analysis, 3.3 percent of the nitrate data (32 of 958 sample results) were censored to the highest MRL of 0.022 mg/L.

Distribution of nitrate concentrations.

Of the 958 samples collected, 31.2 percent (299 sample results) resulted in nitrate concentrations greater than the EPA’s maximum contaminant level (MCL) for nitrate in drinking water of 10 mg/L (see table below). Also, 57.6 percent (552 sample results) of samples had nitrate concentrations greater than 5 mg/L, which represents half the MCL. The EPA established an MCL for nitrate in drinking water to protect against blue baby syndrome, which is caused by oxygen deprivation and can occur in infants who drink water with nitrate exceeding the MCL (U.S. Environmental Protection Agency, 2009). Although sites used in this analysis are not used for drinking water supplies, the MCL was used as a reference to give the nitrate concentration data context.

Information about the EPA's MCL for nitrate in drinking water ( U.S. Environmental Protection Agency, 200  9 ).

Data Exploration

Median concentrations at each of the 46 sites ranged from 0.0315 to 20.1 mg/L (see map below). Site LN2142, located in the central portion of Big Spring Run, had the lowest median concentration. Site LN2159, in the northeast part of the study area, had the highest median nitrate concentration of 20.1 mg/L and also had the highest maximum concentration of 25.5 mg/L.

Median nitrate concentrations at 46 sites in Big Spring Run.

Monitoring wells and piezometers from which water-quality samples were collected were relatively shallow in depth. Monitoring well depths ranged from 2.6 to 24.21 feet below the land-surface, and piezometer depths ranged from 2 to 6.65 feet below the land-surface. Two of the 18 piezometers had undetermined depths. Site LN2142, which had the lowest median nitrate concentration, was also the deepest monitoring well. Site LN2159, which had the highest median concentration, had a depth of 3.75 feet. The three-dimensional scene below can be used to explore well and piezometer depths (depths are represented above the land-surface for display purposes).

Explore Big Spring Run monitoring well and piezometer depths using the interactive scene (depths are represented in the scene as above the land-surface for display purposes).

Piezometers and monitoring wells were sampled between 16 and 22 times between September 2012 and April 2016 (see data clock below), with a majority (13 of 18 piezometers and 11 of 28 wells) sampled 22 times. An attempt was made to collect samples every other month, but this was not always possible due to adverse weather conditions; therefore, samples were collected the following month in several instances to make up for missed sampling. In addition, samples were not always able to be collected at every site during a sampling event due to water availability or inability to find or access a site, which is why piezometers and monitoring wells had differing sampling frequencies. Samples collected at Big Spring Run were typically collected over two consecutive days, and this was considered a single sampling event. The sampling event that took place in March 2014 spanned March 31-April 1, which is why the sample counts for those months are lower than those for the other months.

Data clock showing the number of samples collected by month.

For the purposes of consistency in data analysis, sample collection dates were reclassified by month so that sample collection dates reflect a standard bimonthly pattern. The data clock below shows median nitrate concentrations by (reclassified) month for the years 2012-16. Samples collected in Big Spring Run in March 2015 had the lowest median concentration of 4.57 mg/L, while samples collected in May 2014 had the highest median nitrate concentration of 8.87 mg/L. Overall, samples collected in 2013-14 appear to have higher median nitrate concentrations by month than the other years.

Data clock showing median nitrate concentrations by (reclassified) month.

The matrix below shows nitrate concentrations according to each sample collection site and the reclassified date of collection. It is noticeable that several sites (LN2158, LN2159, LN35, and LN36) had consistent nitrate concentrations above the MCL of 10 mg/L across most sampling events. Conversely, some sites (LN2142, LN2144, LN2145, LN2148, LN2153, LN2165, LN2166, LN2168, LN27, LN38, LN39, and LN40) had nitrate concentrations below 5 mg/L across most sampling events.

Matrix showing nitrate concentrations by sample collection site and date.

Analysis and Results

Because samples were collected from fixed locations representing geo-referenced spatiotemporal data, these defined locations were used to structure the data into a new data format through the creation of space-time bins based on reclassified sample dates. Unique locations for monitoring wells and piezometers were defined, nitrate concentrations were brought into the space-time cube, and the results were mapped (see map below). Since 22 of the 46 defined sampling locations were sampled less than 22 times, empty bins for these sites were filled using an interpolated univariate spline algorithm. Of the 22 sites sampled less than 22 times, 8 of them needed to be excluded from the final space-time cube because their empty bins could not be estimated, resulting in 38 total sites used. The mapped data, which depict results of the Mann-Kendall statistic being run on space-time hot spot analysis z-scores of the nitrate concentrations, indicate a majority of sites (63.2 percent; 24 out of 38 sites) had no statistically significant trend in nitrate concentrations from 2012-16. Statistically significant downward trends in nitrate concentrations were identified at 12 of the sites that appeared to have a random spatial distribution. Statistically significant upward trends were identified at 2 sites (LN2142 and LN2169) located near each other in the central portion of Big Spring Run. Overall, there was no statistically significant data trend found in post-restoration nitrate concentrations in samples collected from the sites.

Space-time cube trends in nitrate concentrations at 38 sites for 2012-16.

Values representing nitrate concentrations for 10 time-steps were forecast for each location of the space-time cube using linear, parabolic, exponential, and s-shaped (Gompertz) curve fitting forecasts to perform space-time pattern mining. This was done by building a forecast model to forecast future time-step values by fitting a selected curve type to the values and extrapolating them to the future. A validation model was also built to validate forecasted values by comparing forecasted values with a portion of data that were withheld for validation. Both forecast and validation models were evaluated using their root mean square error (RMSE), which is a measure of how much compared values differ. RMSE values can be compared across curve types but not compared between forecast and validation models, with the smallest values considered ideal, as they represent smaller differences between raw and fitted values. Although each of the curve types had comparable results (see table below), the parabolic model appeared to have fit the data slightly better. The parabolic model resulted in the smallest overall maximum (4.34) and mean (2.00) values when compared to the other curve types. 

Table comparing forecast results for different curve types.

In addition, values representing nitrate concentrations for 10 time-steps were forecast using an exponential curve fitting forecast for each location of the space-time cube. Season lengths of 2, 4, 6, and 12 months were experimented with (see table below). The models appear to fit the data similarly since their resulting forecast RMSE values are similar, and a season length of 12 months appears to fit the data slightly better than the others. Using a season length of 12 months resulted in the smallest overall minimum (0.21), mean (1.78), and median (1.62) values than those values resulting from the other season lengths.

Table comparing forecast results for different season lengths.

The four curve forecasts and three seasonal forecasts previously described were evaluated to select the most accurate forecasting of nitrate concentrations among seven different models for each location of the space-time cube. The output map (see below) shows an evaluation of forecasts by location for nitrate concentrations with the resulting final forecasted time-step (November 2017) for the selected forecast method shown for each of the 38 sites. The exponential curve fitting forecast using a season length of 12 months was most commonly (11 of 38) selected for the sites. The map shows forecasted nitrate concentrations below 1 mg/L for 12 sites mainly located in the south and west portions of Big Spring Run. Nitrate concentrations between 1 to 5 mg/L are forecasted for 9 sites located in the central and northern parts of the area of interest. Nitrate concentrations more than 5 mg/L and less than 10 mg/L are forecasted for 8 sites located in the central part of Big Spring Run. Nitrate concentrations greater than 10 mg/L are forecasted for 9 sites with 3 clustered in the southeast part of the basin and the remaining 5 in the western and northern portion.

Resulting final forecasted time step for the selected forecast method for each of the 38 sites

LN2155 had the highest forecasted nitrate concentration of 15.15 mg/L for November 2017, and a figure showing forecasted nitrate concentrations for LN2155 can be viewed below.

Forecasted nitrate concentrations for LN2155.

An emerging hot spot analysis was performed by calculating a statistic to compare a local statistic calculated from neighbors for each space-time bin to a global statistical value. The global window was defined so that each neighborhood was analyzed in relation to the entire space-time cube. Analysis results (see map below) show patterns detected in the context of nitrate concentrations in samples collected from wells and piezometers in Big Spring Run from 2012-16. Results for 24 sites showed no patterns in hot or cold spot detection. Eight sites in the central portion of the basin showed a sporadic cold spot, or on-again then off-again, pattern. Results for 4 sites in the southwest corner of the basin were identified as persistent cold spots. Two sites (LN2155 and LN2159) in the eastern part of Big Spring Run resulted in historical hot spot determinations, meaning that they are not recently hot but show that at least 90 percent of time-step intervals were statistically significant hot spots.

Emerging hot spot analysis results for 38 sites.

Summary

This analysis focuses on exploration and analysis of nitrate concentrations in 958 water-quality samples collected from 18 piezometers and 28 monitoring wells in Big Spring Run between September 2012 and April 2016 as part of a post-restoration monitoring effort.

Exploration of nitrate concentrations in 958 samples collected from 46 sites found that:

  • nitrate concentrations ranged from less than 0.008 to 25.5 mg/L;
  • median nitrate concentrations at sites ranged from 0.0315 to 20.1 mg/L;
  • 28 monitoring well depths ranged from 2.6 to 24.21 feet below the land-surface, 16 piezometer depths ranged from 2 to 6.65 feet below the land-surface, and 2 piezometers had undetermined depths;
  • site LN2142 had the lowest median nitrate concentration and was also the deepest monitoring well; and
  • site LN2159 had the highest median nitrate concentration and had a shallow well depth of 3.75 feet.

Analysis of nitrate concentrations for 38 sites that were able to be included in a space-time cube found that:

  • statistically significant trends in nitrate concentrations were identified at 12 sites as a downward trend and at 2 sites as an upward trend; 
  • there was no overall statistically significant data trend found in post-restoration nitrate concentrations in collected samples;
  • nitrate concentration forecasting resulted in nitrate concentrations greater than 10 mg/L forecasted for November 2017 for 9 sites in the southeast and western portions of Big Spring Run; and
  • emerging hot spot analysis identified 8 sites as sporadic cold spots, 4 sites as persistent cold spots, and 2 sites as historical hot spots.

Approximately two years and three months post-restoration image of Big Spring Run basin, Pennsylvania. Taken by Eliza Gross on 1/14/2014.

References

Galeone, D.G., Brightbill, R.A., Low, D.J., and O’Brien, D.L., 2006, Effects of streambank fencing of pastureland on benthic macroinvertebrates and the quality of surface water and shallow groundwater in the Big Spring Run basin of Mill Creek watershed, Lancaster County, Pennsylvania, 1993–2001: U.S. Geological Survey Scientific Investigations Report 2006–5141, 183 p.

Hartranft, J.L., Merritts, D.J., Walter, R.C., and Rahnis, M., 2011, The Big Spring Run restoration experiment – Policy, geomorphology, and aquatic ecosystems in the Big Spring Run watershed, Lancaster County, PA: International Journal of Environmental, Cultural, Economic and Social Sustainability, v. 24, p. 24–30.

Langland, M.J., Duris, J.W., Zimmerman, T.M., and Chaplin, J.J., 2020, Effects of legacy sediment removal and effects on nutrients and sediment in Big Spring Run, Lancaster County, Pennsylvania, 2009–15: U.S. Geological Survey Scientific Investigations Report 2020-5031, 28 p., https://doi.org/10.3133/sir20205031.

U.S. Environmental Protection Agency, various dates, Memorandum laboratory data report Big Spring Run stream restoration 2012-16: U.S. Environmental Protection Agency digital data tables, e-mail correspondence.

U.S. Environmental Protection Agency, various dates, Memorandum laboratory data report Big Spring Run stream restoration 2012-16: U.S. Environmental Protection Agency digital data tables, e-mail correspondence.

U.S. Geological Survey, 2019, USGS National Hydrography Dataset Best Resolution (NHD) for Hydrologic Unit (HU) 4 - 0205 (published 20190702): U.S. Geological Survey, National Geospatial Program, digital data, accessed July 10, 2019 at https://www.sciencebase.gov/catalog/item/5a58a3bde4b00b291cd681ed.

U.S. Geological Survey, 2019, USGS National Hydrography Dataset Best Resolution (NHD) for Hydrologic Unit (HU) 4 - 0205 (published 20190702): U.S. Geological Survey, National Geospatial Program, digital data, accessed July 10, 2019 at https://www.sciencebase.gov/catalog/item/5a58a3bde4b00b291cd681ed.

U.S. Geological Survey, 2021, USGS water data for Pennsylvania, accessed March 16, 2021, at http://waterdata.usgs.gov/pa/nwis/nwis.

Walter, R.C. and Merritts, D.J., 2008, Natural streams and the legacy of water-powered mills: Science, 319(5861), p. 299–304, accessed March 19, 2021, at  http://dx.doi.org/10.1126/science.1151716 .

Weitzman, J.N. and Kaye, J.P., 2017, Nitrate retention capacity of milldam-impacted legacy sediments and relict A horizon soils: SOIL, 3, p. 95-112.

Questions? Please contact Eliza Gross at egross16@jh.edu

This story map is part of a final project completed by Eliza Gross and due on May 15, 2021 in partial fulfillment of requirements for AS.430.633.81.SP21 Advanced Spatio-Temporal Statistics taught by Dr. Orhun Aydin the Spring 2021 semester at Johns Hopkins University.

Location of Big Spring Run, Pennsylvania (U.S. Geological Survey, 2019).

The  U.S. Geological Survey's Streamer website  was used to create a map showing how Mill Creek eventually reaches the Chesapeake Bay (follow red line and orange arrows on map).

Example of a mill pond dam in Chiques Creek, Lancaster County, Pennsylvania. From the  Business for Water Stewardship, 2021 .

Locations of historic mill pond dams in Chester, Lancaster, and York counties, Pennsylvania. From Walter and Merritts, 2008.

Legacy sediment deposits in Big Spring Run prior to sediment removal and stream restoration efforts. From Hartranft and others, 2011.

Approximately six months post-restoration image of Big Spring Run basin, Pennsylvania. Taken by Dan Galeone on 3/20/2012.

Time lapse of changes in Big Spring Run from 2008-19 using Google Earth.

Approximately two years and three months post-restoration image of Big Spring Run basin, Pennsylvania. Taken by Eliza Gross on 1/14/2014.

A shallow monitoring well in Big Spring Run basin, Pennsylvania. Taken by Eliza Gross on 1/14/2014.

Distribution of nitrate concentrations.

Information about the EPA's MCL for nitrate in drinking water ( U.S. Environmental Protection Agency, 200  9 ).

Median nitrate concentrations at 46 sites in Big Spring Run.

Data clock showing the number of samples collected by month.

Data clock showing median nitrate concentrations by (reclassified) month.

Matrix showing nitrate concentrations by sample collection site and date.

Space-time cube trends in nitrate concentrations at 38 sites for 2012-16.

Table comparing forecast results for different curve types.

Table comparing forecast results for different season lengths.

Resulting final forecasted time step for the selected forecast method for each of the 38 sites

Forecasted nitrate concentrations for LN2155.

Emerging hot spot analysis results for 38 sites.

Approximately two years and three months post-restoration image of Big Spring Run basin, Pennsylvania. Taken by Eliza Gross on 1/14/2014.