Transcript
Quantitative Metrics to Gage the Effects of an Enhanced Biodegradation Program Iowa Army Ammunitions Plant, Middletown, Iowa
Melisa Geraghty, Brian Caldwell, PG; Dr. Tiffany Downey, Dr. Ronnie Britto, and Dr. Rick Arnseth
Site History Iowa Army Ammunition Plant (IAAAP) - located
in Des Moines County, Iowa Munitions production and testing beginning in
1941 Resulted in contamination of the soil,
groundwater, and surface water with explosives – Extensive offsite Royal Demolition Explosive
(RDX) groundwater plume sourced by RDX in surface water runoff.
Off-Site Plume map here
Remediation Plan To enhance biodegradation and expedite the
natural attenuation of the RDX plume using an enhanced degradation process (sodium acetate injection) – 11 designed-injection wells upgradient of the
highest concentration plume core. – Analytically modeled injection rates and locations – Five injection events between October 2007
and April 2009• initial event employed all 11 injection wells. • Subsequent events customized injected mass
and number of injection points based on analytical data
Development of Evaluation Metrics Remedial progress metrics developed to
account for temporal and spatial comparisons in order to maintain optimal reducing conditions– Quantitative analysis of plume configuration
– Statistical analysis of the RDX concentrations from individual sampling events
Evaluation Metrics Include Point by point comparisons
– Statistical trend analysis – First-order kinetics concentration change
analysis • degradation rates and times to achieve
remediation goals. Plume wide analysis
– Representative population using differential analysis
– Central Tendency Analysis– Change in overall plume core mass.
Mann-Kendall Trend Analysis Non-parametric statistical test (the data are not
required to be normally distributed) Assesses point changes in a data set over time for an
increasing or decreasing trend at a given statistical confidence level. – Designed to assess four to eighteen rounds of data at
an 80% confidence level– Determines if the data can be used to estimate a first
order degradation or augmentation rate – To avoid biasing the MK test, the same value for all ND
results was used (one half of the detection limit from the round with the lowest detection limit for that well).
– For wells that did not exhibit a increasing or decreasing trend at an 80% confidence level the coefficient of variation was used to determine if the well was stable or unstable
Mann-Kendall Trend Analysis Con’t For wells that did not have a minimum of four
sampling events or did not exhibit an overall trend at 80% confidence, the last three measurements were used to approximate the direction of the concentration change.
The point-by-point analysis results indicate an overall decreasing trend in a majority of the wells. – 18 wells trended
• 8 = decreasing• 1 = increasing • 9 = stable
Quantifying a Decreasing Trend Plotting of the log-transformed concentrations versus a linear
unit of time (days were used). Finite source degradation is described by an exponential degradation curve when plotted on a linear scale. In order to use linear regression to calculate the slope (the degradation rate), the data were log-transformed and plotted on a linear scale.
Performing linear regression calculations on Step 1 by plotting a trend line (least-squares fit trend line) that minimizes the variance (squared deviations) of all of the data points from the line.
Using the slope of the equation represented by the least-squares trend line as the fractional change per day. This was then used to calculate a half-life expressed in days, and to develop a degradation curve that predicts the time at which the concentrations at that sampling point will reach 2 ppb.
Example Output
DATE RDX (ug/L) Ln Conc. (ug/L)Degradation Rate
(% per day)09/28/2007 104 4.644390899 1.02176599812/19/2007 95.8 4.562262685 1.00369779102/13/2008 76.9 4.342505877 0.95535129312/09/2008 41.3 3.7208625 0.81858975
First-order Degradation Rate (day-1) = 0.0022Half Life (days) = 315.05Mann Kendall Statistic (S) = -6.0Number of Rounds (n) = 4Average = 79.50Standard Deviation = 27.880Coefficient of Variation(CV)= 0.351Trend ≥ 80% Confidence Level DECREASINGTrend ≥ 90% Confidence Level DECREASING
Example Output Graphs - Decreasing
3930
0
3940
0
3950
0
3960
0
3970
0
3980
0
3990
0
0
20
40
60
80
100
120
f(x) = − 0.144037624628416 x + 5771.56772215897R² = 0.973120132443412
EMW-06
Date
RD
X ug
/L
3930
0
3940
0
3950
0
3960
0
3970
0
3980
0
3990
0
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
5
f(x) = − 0.00217427496108336 x + 90.2403354244016R² = 0.987992540092612
EMW-06
Date
Ln R
DX
ug/L
0 200 400 600 800 1000 1200 1400 16000
5
10
15
20
25
30
35
40
45
Degradation Curve
Time (days)
Conc
entra
tion
(ug/
L)
Example Output Graphs – Stable or Non-Stable
11/0
9/20
04
02/1
7/20
05
05/2
8/20
05
09/0
5/20
05
12/1
4/20
05
03/2
4/20
06
07/0
2/20
06
0
10
20
30
40
50
60
70
f(x) = 0.014359634334823 x − 506.668356346557R² = 0.0768510632675288
IW-05
Date
RD
X ug
/L
04/0
8/20
05
05/2
8/20
05
07/1
7/20
05
09/0
5/20
05
10/2
5/20
05
12/1
4/20
05
02/0
2/20
06
03/2
4/20
06
05/1
3/20
06
0
10
20
30
40
50
60
f(x) = 0.0523149703166883 x − 1978.16565832776R² = 0.924461506594045
IW-05
Date
RD
X ug
/L
Example Output Graphs - Increasing
08/0
6/20
07
11/1
4/20
07
02/2
2/20
08
06/0
1/20
08
09/0
9/20
08
12/1
8/20
08
03/2
8/20
09
07/0
6/20
09
10/1
4/20
09
01/2
2/20
10
0
0.5
1
1.5
2
2.5
3
3.5
f(x) = 0.00350760288795276 x − 138.069184305981R² = 0.740959616062104
EMW-08
Date
RD
X ug
/L
Plume Wide Analysis
Plume Core Mass
Plume Event
Number of wells sampled
RDX mass(g)
Change in mass from Previous events
(g)Percent Change
1 16 3,693,903.46 N/A N/A
2 24 61,897,939.34 58,204,035.88 1576%
3 24 61,815,101.54 -82,837.80 0%
4 26 57,380,772.09 -4,434,329.45 -7%
Insert example of plume core mass wksheet
Statistical Mean The plume-wide statistical mean decreased
from 101 ppb at baseline in October 2007 to 42.38 ppb in December 2008
Between December 2008 and December 2009, there was a 70.0% decrease in UCL means to 12.73 ppb.
However, based on the differential test (Kolmogorov-Smirnov), the change in populations is not statistically significant at a 95% confidence level. All data sets result from the same population
top related