Using PCA & Repeated ANOVA to evaluate the In Situ Bioremediation performance of sites contaminated by trichloroethylene

(1)

Master Thesis

HALMSTAD

Master's Programme in Applied Environmental Science, 15 credits

Using PCA & Repeated ANOVA to evaluate the In Situ Bioremediation performance of sites contaminated by trichloroethylene

Applied Environmental Science, 15 credits

Halmstad 2018-5-31

Xinyao Chen

(2)

Date 2018-05-31

Halmstad University, Applied Environmental Science Master thesis, 15 credits

Using PCA & Repeated ANOVA to evaluate the In Situ Bioremediation performance of sites contaminated by

trichloroethylene

Name: Chen Xinyao

Supervisor: Sylvia Waara Examiner: Marie Mattsson

(3)

HALMSTAD UNIVERSITY MASTER THESIS 2018 – AER

Abstract

Currently, one of the most common techniques to remediate contaminated sites of TCE is in situ bioremediation (ISB). In this study, PCA and repeated ANOVA were used to statistically analyze the trends of variables over time to aid in the interpretation of the performance of the in situ bioremediation (ISB) technique. cDCE, Mn²⁺, chloride and alkalinity have appeared a significant trend over time suggested they have relative stronger indicating power to the performance of ISB. The variables that most effectively describe the bioremediation performance are Fe²⁺, DOC, Mn²⁺, methane and alkalinity.

Their dramatic changes with time indicate the active functioning of dechlorinating bacteria to remediate the contamination. Three group of indicators can be identified according to their trends over time having a certain consistent character. The first group is methane and ethane, the second group consists of chloride, sulfate and alkalinity and the third group consists of cDCE and tDCE. Definitely, PCA can be an effective tool to analyze the overall trends and transformation pattern of variables over time and at different sampling points within the site. However, the fragmented data set reduce the possibilities for a complete understanding of the remediation process at the site.

(4)

1 Introduction

1.1 Chlorinated solvents pollution

Chlorinated solvents are a large family of chemical compounds that contain chlorine called halogenated volatile organic compounds (VOCs) (Sheu et al., 2018, Pant and Pant, 2010), which are widely used in various industrial processes, mainly as cleaning solvents in dry-cleaning and semiconductor industries (Aulenta et al., 2006), typically such as trichloroethylene (TCE), tetrachloroethylene (PCE). Their careless storage, handling and disposal, as well as high chemical stability, resulted in chlorinated hydrocarbons becoming one of the most frequently encountered subsurface (groundwater) contaminants and have contributed to a large number of contaminated sites (Nemecek et al., 2016, Aulenta et al., 2006, Kouznetsova et al., 2010). Reports are also available documenting contamination of surface water as a result of TCE release from groundwater sources (Pant and Pant, 2010).

In a solvent Dense non-aqueous phase liquid (DNAPL) source zone, aqueous phase chlorinated ethenes continuously dissolve into groundwater, resulting in an aqueous phase plume emanating down gradient. Such plumes are of concern due to the carcinogenic and mutagenic potential of TCE to different target organs (Kouznetsova et al., 2010, Aulenta et al., 2006). Exposure to TCE will occur if the contaminated groundwater is used as potable water, used in cooking, bathing and watering gardens.

TCE is considered a primary pollutant with 5 μg/L maximum contamination level allowed in drinking water (World Health Organization, 2010). In addition, if sufficient levels are present and groundwater is shallow, TCE vapors can also move upwards from the water and penetrate though the soil, building foundations and underground service infrastructure and contaminate indoor air (Government of South Australia, 2017).

People who work or live in adjacent areas can be exposed to TCE through the absorption of solvents, through inhalation and skin contact. Inhalation is the most common form of workplace exposure, because the solvents can readily evaporate. Skin contact is another important route of exposure in the workplace.

Hence the presence of TCE in the soil and groundwater environment poses

(7)

important health risks and has prompted investigations concerning their remediation of the contaminated sites so as to reduce the potential risk to human beings (Aulenta et al., 2006, Rusyn et al., 2014).

1.2 Degradation characteristics

In the environment, chlorinated solvents break down rapidly in air and surface water but much slower in soil and groundwater resulting in more serious groundwater and soil chlorinated solvents pollution (Government of South Australia, 2017). In the soil and groundwater environment, it has been shown that chlorinated hydrocarbons can be biodegraded by added or activated indigenous microorganisms in a variety of ways lowering the concentrations of chlorinated ethenes. Successful remediation might result in the formation of the nontoxic product ethene with the supplement of primary substrates and where chlorinated solvent is used as electron acceptors (Majone et al., 2015; David et al., 2015; Sheu et al., 2018). The mechanisms of degradation include aerobic and anaerobic biodegradation, reductive dechlorination and co-metabolism (Sheu et al., 2018). Generally, the process of reductive dechlorination of TCE relies on sequential reductive dechlorination with each step in the process removing one chlorine atom and replacing it with a hydrogen atom under anaerobic conditions, for example, TCE is indeed transformed to ethene through step-by-step reactions where 1,2-cis- dichloroethene (cis-DCE) and vinyl chloride (VC) are formed as intermediate products (Nemecek et al., 2018).

There are evidences, indicating that microbes have diverse and ample potential for adaptation to use the dechlorination of chlorinated hydrocarbons as an energy source (Govender et al., 2011). The only known genus of bacteria that can perform the complete transformation from chlorinated hydrocarbons to ethene is Dehalococcoides mccartyi (Kao et al., 2016, Matturro et al., 2016, Dolinova et al., 2017, Ni et al., 2018, Nemecek et al., 2018).

1.3 In situ bioremediation (ISB)

Several methods are available for remediation of TCE pollution. Physical

(8)

processes, such as soil excavation and venting, and ground water extraction techniques, all of which have been used in remediation of TCE pollution in the subsurface environments, have been found relatively slow, costly, and inefficient, along with being environmentally disruptive in nature (Pant and Pant, 2010) leading to the development of bioremediation.

Since the mid-1990s, in situ bioremediation (ISB) using reductive dechlorination has been a widely used remedial technology for the treatment of groundwater contaminated by TCE (Nemecek et al., 2018). It is evident that ISB is a highly promising and cost-effective technology for remediation of contaminated soil, groundwater and sediments, and the wide metabolic diversity of microorganisms makes it applicable to an ever-increasing number of contaminants and contamination scenarios (Majone et al., 2015). Hence, ISB technology can meet the requirements of the green remediation, and it can be applied for source treatment and plume migration control for contaminated groundwater (Sheu et al., 2018). TCE-contaminated sites are very worthwhile for us to remediate by ISB methods now.

However, in situ anaerobic bioremediations (ISABs) are not perfectly executed processes without any problems. During the transformation of TCE there might be an accumulation of toxic, lower chlorinated compounds such as cis-DCE and VC that are more slowly degradable under anaerobic conditions than their mother compounds (Scherr et al., 2011, Ferguson and Pietari, 2000). High concentrations of chloroethenes in source areas can also inhibit microbial degradation due to their toxic effects (Aulenta et al., 2006). The accumulation of VC is also considered to be a major problem due to VC being a proven carcinogen for humans (He et al., 2003). Other complications that can occur, often associated with the use of external electron donor are methane formation which may lead to the risk of explosion, clogging of aquifer due to organic material, and byproducts of fermentation such as acetate and propionate. These negative side effects can lead to a deterioration of groundwater quality (Aulenta et al., 2007).

There are some developments about using ISB to enhance the remediation of TCE- contaminated sites. Sheu et al. (2018) showed that the gamma poly-glutamic acid as the

(9)

efficiency of reductive dechlorination of chlorinated compounds in groundwater, including TCE. Nemecek et al. (2018) studied the effectiveness of thermally enhanced ISB of groundwater contaminated by chlorinated solvents by conducting a field test, the results indicated that heating of the treated aquifer can significantly accelerated the ISB treatment process but only in the case of an abundant substrate. Although the ISB technology have its unreplaceable advantages it is still under development, there are a lot of research conducted enhancing remediation efficiency and improving remediation performance as well as reducing negative effects.

For the potential adverse effects on public and environmental, the monitor of TCE and its metabolites and relevant chlorinated solvents have been very important to indicate the pollution level so as to assess the ecological risk and health risk and put forward recommendations to recovery this contamination effectively in the later time.

1.4 Background of studied remediation project 1.4.1 Site

There is a place which was polluted by chlorinated solvents and which is now treated by ongoing bioremediation with the help of micro-organisms. It is located in southern Sweden and it is the property Rampen 36 in the city of Falkenberg, Halland County. On site there was a factory which up until 2009 made lighting and fixtures.

The area consists of an industrial area with ongoing activities. The pollution originates from the activities during the 1930-1970's.

1.4.2 Remediation technique- 3-D Microemulsion (3DMe)

The company WSP has been performing ISB in this contaminated soil by injecting electron donor since the fall 2010 and has monitored the site and gathered data during the remediation process. The injected electron donor 3-D Microemulsion (3DMe) is a slightly viscous injectable liquid that incorporates a molecular structure composed of tetramers of lactic acid and fatty acids esterified to a carbon backbone molecule of glycerin, and the liquid is designed for ISB at sites where the process of enhanced reductive dechlorination is possible.

(10)

3-D Microemulsion (3DMe)^TM (3-D Microemulsion (3DMe) – Regenesis, 2006), a form of Hydrogen Release Compound (HRC) Advanced, is the new paradigm in time- release electron donors for groundwater and soil remediation. 3DMe is based on a new molecular structure (patent applied for) designed specifically to optimize anaerobic degradation of contaminants in subsurface environments. The structure incorporates esterified lactic acid and esterified long chain fatty acids. This structure provides an advantage as it allows for the controlled-release of fatty acids (which is the most efficient electron donors) and the controlled-release of fatty acids (a very cost-effective source of slow release hydrogen). Upon injection, the controlled-release of lactic acid dominants serving to initiate and stimulate anaerobic dechlorination. Over time the controlled-release of fatty acids will dominate, acting to continue microbial stimulation.

The expected single-injection longevity of this product is 1-2 years and in excess of 4 years under optimal conditions, e.g. concentrated application in low permeability, low sample consumption environments.

1.4.3 Indicators of remediation performance

The data for analysis comes from groundwater pipes installed between 2008 and 2009. The flow direction of the groundwater was estimated by the installment surveying and leveling. Groundwater sampling and analysis for chlorinated hydrocarbons (including TCE) was performed in a total of 31 groundwater pipes at several occasions up until 2015. During these sampling occasion field measurements of pH, temperature, redox potential and conductivity were performed in some pipes. Samplings were carried out with a peristaltic pump after pumping two to three pipe volumes to ensure that samples were not compromised with outside sources. When groundwater chemistry is combined with chlorinated hydrocarbons and with dechlorinating microorganisms that interacts with both the water and the pollutant the complexity and number of possible chemical reactions increases multifold, thus creating a large number of variables making it difficult to get a complete picture of remediation success with univariate statistical analysis at each sampling point.

(11)

from the groundwater samples before and during remediation, principal component analysis (PCA), a type of multivariate analysis, were chosen as the statistical approach.

The advantage of this approach is the dimension reduction ability of PCA, meaning that PCA reduces the number of variables without losing valuable information.

PCA is a commonly used statistical method for evaluation soil- and water related environmental issues.

1.4.4 Current remediation results

Obtained data suggested a breakdown of TCE to ethene occurs at all contaminated points on Rampen 36, but VC which has higher level toxicity were observed at some points after the injection of 3Dme during the degradation process. It also indicated the degradation has been in a phase of complete degradation for approximately three years (roughly since the fall of 2012). Before the microbial injection in April, 2010 there was 11 contaminated sampling points in the area and one year after, in 2011, there was 8 such sampling points. Measurements in 2012 showed that there were 11 sampling points of concern again, and in 2013 there were only six sampling points that could be classified as polluted. In 2014 and 2015 they were down to two sampling points.

1.5 Aims and issues

The aims of this project are mainly,

I Analyze how the ISB performance at TCE-contaminated sites varies over time.

II Statistically study how different environmental factors varies over time.

III Using multivariate methods for data reduction (for example PCA) identifying which variables that are crucial for affecting the performance of ISB technique used in this and similar cases.

The questions this thesis prepared to answer,

I Is the remediation occurring at the same pace and with the same efficiency in all sampling points?

II Which variables can be used for effectively describing the bioremediation performance of the contaminated sites?

(12)

III Is there any interaction between measured variables indicating the remediation performance at different sampling sites?

2 Materials and Methods

2.1 Data collection

Data for this thesis were supplied by WSP Halmstad, Soil and Water Department.

Totally, we have 40 different sampling points, and 28 variables, including time passed since injection date, the concentration of chlorinated solvents and chemicals parameters of groundwater, and field physico-chemical properties over time since the 3DMe was injected. The variables are dichloromethane, 1,1-dichloroethane, 1,2-dichloroethane, trans-1,2-dichloroethene (tDCE), cis-1,2-dichloroethene (cDCE), 1,2-dichloropropane, trichloromethane, tetrachloromethane, 1,1,1-trichloroethane, 1,1,2-trichloroethane, trichloroethene (TCE), tetrachloroethylene (PCE), vinyl chloride (VC), methane, ethene, ethane; iron (Fe²⁺), manganese (Mn²⁺), chloride, nitrate, sulfate, dissolved organic carbon (DOC), alkalinity; pH, conductivity, redox, temp, respectively.

2.2 Datasets

The sampling sites and analyzed parameters are presented in Table 1. (For all screened parameters analyzes were made in Apr 2008, Dec 2009, Apr 2010 (samples before injection), Jan 2011, Oct 2011, Oct 2012, Nov 2013, Nov 2014, Sep 2015. The data set is extremely fragmented and consequently only some parameters at some sampling points have been measured at a particular point in time. Therefore, data from 3 time points (i.e. Apr 2008, Dec 2009 and Apr 2010) has been used as data for the pooled time point ¨before injection¨. For example, if the concentration of TCE at Apr, 2010 was measured the data was used in ¨before injection¨ if it was missing it as replaced by a measured value from Dec 2009 or Apr 2008 if data from Dec 2009 was also missing. This combination of data from different sampling dates resulted in a data set with 7 sampling times. The detailed determination time and sampling sites of the dataset can be seen in the Table 1.

(13)

In addition, some of the 27 variables besides time, such as PCE and nitrate were always below detection limit and these parameters have consequently been excluded from the analysis. In addition, the measured field indicators (pH, conductivity, redox, temperature) were in a preliminary analysis shown to vary extensively over time and this variation is most likely due to external factors rather than the ISB and consequently they have not been used in this study.

Hence, the objective variables of this paper include 6 organic parameters — trichloroethene (TCE), trans-1,2-dichloroethene (tDCE), cis-1,2-dichloroethene (cDCE), vinyl chloride (VC), methane and ethene (unit: µg/L) and 6 chemical parameters of groundwater — iron (Fe²⁺), manganese (Mn²⁺), sulfate, chloride, dissolved organic carbon (DOC) (unit: mg/L) and alkalinity (unit: mg HCO^3-/L).

As the data set, as stated above, was very fragmented and thus not a priori collected with the purpose of using it for data reduction or classification using multivariate analysis several attempts were made to analyze the data. In the work presented several smaller PCAs was finally conducted trying to make maximal use of the data in the complicated data set.

(14)

Table 1 Overview of the large but fragmented dataset Variables Times Number of sites Sampling sites

Organic variables (TCE, cDCE, tDCE, VC, methane, ethene);

Chemical parameters (Fe²⁺, Mn²⁺, sulfate, chloride, DOC and alkalinity)

Before injection 20 GV1, GV102, GV103, GV105, GV106, GV107, GV2, GV22, GV25, GV26, GV52, GVK14, GVK15, GVK17, GVK19, GVK24, GVK25, GVK3, GVK6, GVK8

Jan 2011 15 GV1, GV103, GV106, GV2, GV22, GV25, GV52, GVK15, GVK17, GVK19, GVK24, GVK25, GVK3, GVK6, GVK8

Oct 2011 18 GV1, GV102, GV103, GV105, GV106, GV107, GV22, GV25, GV52, GVK15, GVK17, GVK19, GVK24, GVK25, GVK28, GVK3, GVK6, GVK8

Oct 2012 11 GV1, GV103, GV105, GV106, GV22, GV25, GV52, GVK17, GVK-3, GVK-6, GVK-8 Nov 2013 11 GV102, GV103, GV106, GV2, GV22, GV25, GV52, GVK24, GVK3, GVK6, GVK8

Nov 2014 8 GV1, GV102, GV106, GV2, GV52, GVK24, GVK6, GVK8

Sep 2015 8 GV1, GV102, GV106, GV2, GV52, GVK24, GVK6, GVK8

(15)

2.3 Data analysis methods

All the data will be processed by the software – STATISTICA. To uncover possible latent structures in the dataset, multivariate analysis with principle component analysis (PCA) were considered suitable. Multivariate analysis is a tool for explaining why observations vary and PCA groups variables that vary together. PCA converts a set of observations that might be correlated into a set of values of linearly uncorrelated variables. These values are called principle components. To find out which variables that varied together in this case PCA were done by Eigen-value decomposition to project component cores and loadings. In addition, repeated ANOVA can be used to compare the differences among different time points of variables and can see the trends over time of variables. So, the main statistical methods include PCA and repeated ANOVA. The analysis steps in this thesis are as following,

I. Multivariate analysis using the PC and classification module in the software Statistica 13.1 will be used to reduce the variables and to identify groups of sampling points at the different stages of bioremediation. The analysis will enable the identification of outliers, explain their presence and if scientifically justified remove them from further analysis to ease the interpretation of the bioremediation process.

II. PCA grouping analysis: by detecting and comparing the screened components to compare the performances of bioremediation treatment over time for all sampling points, so that we can judge whether or not or how much this treatment can remediate the contamination within a certain time.

III. After removing outliers conduct repeated measures ANOVA to analyze the trends of different variables over time, and which time points is significantly different with others.

2.4 Statistics

- When the correlation coefficient was higher than 0.5 or less than -0.5, the correlation between variable and a principle component with Eigen value larger or very close to 1.0 is higher was considered a strong correlation in this paper.

- Using F-value and p-value to judge the difference between variables at different

(16)

times whether or not is significant. When p<0.05, the difference is significant.

2.5 Databases

Web of science, ESCO, Springer, google scholar.

(17)

3 Results and Analysis

3.1 Correlation between variables and principle components (PCs)

According to the results of the principle component analysis of variables at different determination times in STATISTICA, the correlation results between variables and principle components can be attained, we can make use of the selected principle components whose Eigen value are higher than 1.0 to attribute and reduce the variables in different datasets. This is benefit for us to find the biodegradation patterns of TCE in the contaminated sites.

Before injecting electron donor, 4 principle components whose Eigen value were higher than 1 were identified, they explain 31.50%, 24.77%, 16.66% and 9.01% (totally 81.94%) of the total variance, respectively (Table 2, The detailed correlation coefficients between variables and PCs can be seen in Appendix 1). PC1 is positively correlated with TCE, tDCE, cDCE, ethene, Fe²⁺ and Mn²⁺. PC2 is negatively correlated with VC, methane, sulfate and alkalinity. PC3 is positively correlated with tDCE and negatively correlated with ethene, Fe²⁺ and Mn²⁺; PC4 has a positive correlation with sulfate and has a negative correlation with chloride.

After injection, at Jan, 2011, 4 principle components were attained for their Eigen value higher than 1, they explain 45.52%, 25.03%, 15.88%, 8.74% and 9%

(totally 95.17%) of the total variance, respectively. PC1 is positively correlated with tDCE, cDCE, VC, methane, ethene, Fe²⁺, Mn²⁺ and DOC, and is negatively correlated with chloride, sulfate and alkalinity. PC2 has a positive correlation with TCE and Mn²⁺

and has a negative correlation with tDCE and alkalinity. PC3 is found to negatively correlated with cDCE and DOC.

At Oct, 2011, 4 principle components whose Eigen value higher than 1 were selected, they can respectively explain 25.35%, 22.75%, 21.02% and 11.35% (totally 80.47%) of the total variance. PC1 is negatively correlated with cDCE, Mn²⁺, chloride, DOC and alkalinity. PC2 has a positive correlation with methane, ethene, Fe²⁺ and DOC, and has a negative correlation with TCE. PC3 is positively correlated with tDCE, cDCE,

(18)

VC and ethene, is negatively correlated with chloride. PC4 is positively correlated with sulfate and is negatively correlated with methane.

At oct, 2012, 2 PCs whose Eigen value higher than 1 are selected, they can explain 75.86% and 20.41% (totally 96.27%) of the total variance, respectively. PC1 is positively correlated with ethene, is negatively correlated with TCE, tDCE, cDCE, VC, Fe²⁺, Mn²⁺, chloride, sulfate, DOC and alkalinity. PC2 has a negative correlation with methane, ethene and Fe²⁺.

At Nov, 2013, 3 PCs were selected whose Eigen value are higher than 1. They can explain 87.64% of the total variance in all, and respectively explain41.42%, 33.24%

and 12.99% total variance. PC1 is negatively correlated with TCE, cDCE, VC, methane, Fe²⁺, Mn²⁺, chloride, sulfate and alkalinity. PC2 is positively correlated with TCE, tDCE and cDCE, is negatively correlated with methane, chloride and DOC. PC3 is positively correlated with ethene.

At Nov, 2014, 3 PCs whose Eigen value higher than 1 were selected and they can respectively explain 60.09%, 16.93% and 13.12% (totally 90.14%) of the total variance, PC1 is negatively correlated with TCE, tDCE, cDCE, VC, Fe²⁺, Mn²⁺, chloride, sulfate and alkalinity. PC2 has a positive correlation with VC, methane and ethene, has a negative correlation with DOC. PC3 has a positive correlation with Fe²⁺ and has a negative correlation with DOC.

At Sep, 2015, 4 PCs whose Eigen value higher than 1 were selected, they can respectively explain 53.77%, 23.42%, 9.20% and 8.57% (totally 94.96%) of the total variance. PC1 is negatively correlated with TCE, tDCE, cDCE, VC, Fe²⁺, Mn²⁺, chloride, sulfate and alkalinity. PC2 has a positive correlation with TCE, tDCE and cDCE and has a negative correlation with Methane, Chloride and DOC. PC3 is negatively correlated with methane. PC4 is positively correlated with DOC.

(19)

Table 2 Correlation between variables and PC whose Eigen value is higher than 1.0 at different determination times Sampling

time

Principle

components (PCs)

Total variance

Eigen Value

Correlated variables

(Green symbolizes positive correlation. Red symbolizes negative correlation)

Before injection

PC1 31.50% 3.77983 TCE, tDCE, cDCE, Ethene, Fe²⁺, Mn²⁺

PC2 24.77% 2.97221 VC, Methane, Sulfate, Alkalinity PC3 16.66% 1.99937 tDCE, Ethene, Fe²⁺, Mn²⁺

PC4 9.01% 1.08156 Chloride, Sulfate

TOTAL 81.94%

2011-Jan

PC1 45.52% 5.462957 tDCE, cDCE, VC, Methane, Ethene, Fe²⁺, Mn²⁺, DOC, Chloride, Sulfate, Alkalinity PC2 25.03% 3.003152 TCE, Mn²⁺, tDCE, Alkalinity

PC3 15.88% 1.905352 cDCE, DOC

PC4 8.74% 1.048999

TOTAL 95.17%

2011-Oct

PC1 25.35% 3.041838 cDCE, Mn²⁺, Chloride, DOC, Alkalinity PC2 22.75% 2.729637 TCE, Methane, Ethene, Fe²⁺, DOC PC3 21.02% 2.522440 Chloride, tDCE, cDCE, VC, Ethene PC4 11.35% 1.362030 Methane, Sulfate

TOTAL 80.47%

2012-Oct

PC1 75.86% 9.102700 TCE, tDCE, cDCE, VC, Fe²⁺, Mn²⁺, Chloride, Sulfate, DOC, Alkalinity, Ethene PC2 20.41% 2.449397 Methane, Ethene, Fe²⁺

TOTAL 96.27%

(20)

Sampling time

Principle

components (PCs)

Total variance

Eigen Value

Correlated variables

(Green symbolizes positive correlation. Red symbolizes negative correlation)

2013-Nov

PC1 41.42% 4.970043 TCE, cDCE, VC, Methane, Fe²⁺, Mn²⁺, Chloride, Sulfate, Alkalinity PC2 33.24% 3.988387 Methane, Chloride, DOC, TCE, tDCE, cDCE

PC3 12.99% 1.558456 Ethene

TOTAL 87.64%

2014-Nov

PC1 60.09% 7.210580 TCE, tDCE, cDCE, VC, Fe²⁺, Mn²⁺, Chloride, Sulfate, Alkalinity PC2 16.93% 2.031528 VC, Methane, Ethene, DOC

PC3 13.12% 1.574292 DOC, Fe²⁺

TOTAL 90.14%

2015-Sep

PC1 53.77% 6.452609 TCE, tDCE, cDCE, VC, Fe²⁺, Mn²⁺, Chloride, Sulfate, Alkalinity PC2 23.42% 2.809840 Sulfate, VC, Methane, Ethene

PC3 9.20% 1.104295 Methane

PC4 8.57% 1.028441 DOC

TOTAL 94.96%

(21)

3.2 Trends of variables over time

This study tried to use multivariate analysis with PCA to what extent of PCs can be affected by the cases, and which ones are the most different cases with others that can be thought of as outliers which are needed to be removed in the later repeated ANOVA so as to reduce the standard deviation of variables and aid in the interpretation of the time course of remediation.

3.2.1 Outliers of cases for datasets from PCA

According to the figures of projection of sampling sites on the factor-plane at different times attained by PCA at different time points (see these figures in Appendix 2), the outliers of cases for different datasets at different determination times is presented in Table 3.

Table 3 Outliers of cases for variables at different determination times Determination time Cases to be removed as outliers

Before injection GV1, GVK8, GVK6, GVK17, GVK25

Jan 2011 Each case varies differently, NO cases to be removed

Oct 2011 GV1, GVK3, GVK6, GVK8

Oct 2012 Each case varies differently, NO cases to be removed

Nov 2013 GVK3, GVK6

Nov 2014 GVK6, GVK8

Sep 2015 GVK6, GVK8

3.2.2 Trends of variables over time from repeated ANOVA

The repeated ANOVA was used to evaluate the trends of variables over time during remediation. In the repeated measure ANOVA analysis, the sampling point(s) with the most deviant value(s) were classified as outliers (Table 3). The outlier data was excluded from the data set and the analysis was repeated.

For TCE and its metabolites

The concentration of TCE is that it generally decreases with time although there is a slight increase in Oct, 2012 (Figure 1). TCE stays at a low concentration from Nov, 2013. In addition, there are two time points - before injection and Oct2012 which shows

(22)

very big variation which cannot be reduced by removing the outliers. Looking for the reason in the original dataset, it is easily to identify two groups of values, the existence of either very low or high value group is not casual, this can be attributed to the very different diffusion process in different direction and the uneven distribution of TCE in the different sampling sites.

The trend of tDCE over time is firstly that it increases and then it decreases to a stable low level close to 5 µg/L (Figure 2). At Jan, 2011 and Oct, 2012, the standard deviation is large and it may be an indirect result from the very uneven distribution of TCE for its reductive dechlorination process, and directly affected by the distribution of tDCE.

The trend of cDCE over time is firstly that it increases and then decreases to a stable low level close to 200 µg/L (Figure 3), the fluctuation of cDCE over time is very similar to tDCE. At Jan, 2011, its standard deviation is very big, may indirectly resulted from the very uneven distribution of TCE for its reductive dechlorination process, and directly affected by the distribution of cDCE at this time points.

The trend of VC over time is firstly that is increased and then decrease to a stable low level (Figure 4), the turning peak point is Oct, 2012. At Oct, 2012, the standard deviation of VC is very big, may resulted from the very uneven distribution and dramatic biodegradation of tDCE and cDCE at this time points, and resulted from the different transformation paces of DCE at different sampling sites which provide various environmental conditions, for example, flow direction of groundwater and terrain.

For methane and ethene, their trends over time – increase-decrease-slight increase - are very similar (Figure 5, 6), the difference among time points of methane is very significant (F(6, 30)=4.4847, p=.00236<0.005), the difference among times of ethene is very close to the significant level (0.05<p=0.0992<0.1), they indicated the increase

(23)

variables have big standard deviation in Oct, 2012. It may can be explained by the dramatically interaction between TCE and its metabolites, because all of them have big fluctuation in Oct, 2012. In addition, ethene is a clear indicator of the successful remediation performance, so it general increases over time (Figure 6). Final concentration is higher than in before injection indicating that the transformation processes of TCE are in progress after implementing the bioremediation technique showing the successful remediation of ISB in a certain extent.

TIME; LS Means

Current effect: F(6, 24)=.48420, p=.81350 Effective hypothesis decomposition Vertical bars denote 0.95 confidence intervals

Before injection Jan-2011

Oct-2011 Oct-2012

Nov-2013 Nov-2014

Sep-2015 Time

-80 -60 -40 -20 20 40 60 80 100 120 140 160

0

TCE (礸/L)

Figure 1 Trend of TCE at different time points

TIME; LS Means

Current effect: F(6, 30)=1.6186, p=.17643 Effective hypothesis decomposition Vertical bars denote 0.95 confidence intervals

Oct-2011 Oct-2012

Nov-2013 Nov-2014

Sep-2015 Time

-5 5 10 15 20 25 30 35 40 45 50

0

tDCE (礸/L)

(24)

Figure 2 Trend of tDCE at different time points

TIME; LS Means

Oct-2011

Oct-2012

Nov-2013

Nov-2014

Sep-2015

Time -400

-200 200 400 600 800 1000 1200 1400 1600 1800 2000

0

cDCE (

礸 /L )

Figure 3 Trend of cDCE at different time points

TIME; LS Means

Oct-11

Oct-12

Nov-13

Nov-14

Sep-15

Time -400

-200 200 400 600 800 1000

0

V C ( 礸 /L)

Figure 4 Trend of VC at different time points

(25)

TIME; LS Means

Oct-11

Oct-12

Nov-13

Nov-14

Sep-15 Time

-10000 -5000 5000 10000 15000 20000 25000 30000 35000 40000 45000 50000

0

Methane (礸/L)

Figure 5 Trend of methane at different time points

TIME; LS Means

Oct-11

Oct-12

Nov-13

Nov-14

Sep-15 Time

-150 -100 -50 50 100 150 200 250 300 350 400

0

Ethene (礸/L)

Figure 6 Trend of Ethene at different time points

(26)

For the chemical parameters of groundwater

The general trend of Fe²⁺ over time (Figure 5) is “increase- decrease-steady level”

(Figure 7), one obvious peak concentration which appeared at Jan, 2011. From Nov, 2013, the concentration of Fe²⁺ can steadily keep relative low level. The differences among the concentration of iron at different time points are very significant (F(6, 24)=3.3750, p=0.0148<0.05), this means injecting electron donors can significantly increase the concentration of Fe²⁺ in the groundwater. In, addition, the peak value appearing in Jan, 2011 has a relative high standard deviation. There is a very large variation in concentration in different sampling points ranging from 0.0015 to 51 mg/L.

hence, what results in the desperate distribution of Fe²⁺ between the high concentration sites and the low concentration sits are interested to identify.

The general trend of Mn²⁺ over time (Figure 8) is “slight increase-general decrease”, the turning point is Jan, 2011, this denotes to the maximum concentration of Mn²⁺ presented in Jan, 2011. In addition, the concentration of Mn²⁺ has a very big standard deviation and a sudden rise in Nov, 2014 making the concentration of Mn²⁺ at this point very strange, which maybe resulted from the change of external environment.

The differences among the concentration of Mn²⁺ at different time points are very significant (F(6, 24)=3.9580, p=0.00683<0.05).

The general trend of chloride, sulfate and alkalinity over time (Figure 9, 10, 12) are very similar - “increase-steady”, chloride, sulfate and alkalinity are generally increased until Oct, 2012 results in the maximum value appeared at this time point.

However, the standard deviation of chloride and sulfate from Oct, 2012 to Sep, 2015 are all very large, looking from the original dataset, two extreme and not casual separate distribution of these three variables’ concentration can be found. Specially, there is a significant difference of alkalinity between different time points (F(6,24)=2.5723, p=0.0456<0.05), indicating the ISB significantly increase alkalinity.

The general trend of DOC over time (Figure 11) is “increase-decrease-stable level”.

(27)

The figure showed the concentration of DOC appeared an acute increase after injecting electron donors, and an acute decrease making the DOC fall to original level basically after the acute increase, the turning point is Jan, 2011. The peak value in Jan, 2011 which indicate the ISB have influenced the concentration of DOC significantly (F=6.8097, P=00012<0.001). However, the variance of DOC at Jan, 2011 is very big resulted from the uneven distribution at different sampling sites, they can be attributed to the different diffusion and remediation paces of electron donors.

TIME; LS Means

Oct-2011

Oct-2012

Nov-2013

Nov-2014

Sep-2015 Time

-20 -10 10 20 30 40 50 60

0

Fe2+ (mg/L)

Figure 7 Trend of Fe²⁺ at different time points

(28)

TIME; LS Means

Oct-2011

Oct-2012

Nov-2013

Nov-2014

Sep-2015 Time

-0.2 -0.1 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9

0

Mn2+ (mg/L)

Figure 8 Trend of Mn²⁺ at different time points

TIME; LS Means

Oct-2011

Oct-2012

Nov-2013

Nov-2014

Sep-2015

Time -100

-50 50 100 150 200 250 300 350 400

0

Chloride (mg/L)

Figure 9 Trend of chloride at different time points

(29)

TIME; LS Means

Oct-2011

Oct-2012

Nov-2013

Nov-2014

Sep-2015

Time -150

-100 -50 50 100 150 200 250 300 350 400 450

0

Sulfate (mg/L)

Figure 10 Trend of sulfate at different time points

TIME; LS Means

Oct-2011

Oct-2012

Nov-2013

Nov-2014

Sep-2015

Time -50

50 100 150 200 250 300 350

0

DOC (mg/L)

Figure 1 Trend of DOC at different time points

(30)

TIME; LS Means

Before injection

Jan-2011 Oct-2011

Oct-2012 Nov-2013

Nov-2014 Sep-2015 Time

-400 -200 200 400 600 800 1000 1200 1400

3- Alkalinity (

m g H C O /L

) 0

Figure 2 Trend of alkalinity at different time points

(31)

4 Discussion

4.1 Transformation patterns and paces

- Is the remediation occurring at the same pace and with the same efficiency in all sampling points?

From the trends of TCE, tDCE, cDCE, VC, methane and ethene over time, they showed that a general decrease with time of TCE (Figure 1), an increase of tDCE, cDCE presented in Jan, 2011 (Figure 2-3), a dramatic increase of VC, methane and ethene presented in Oct, 2012 (Figure 4-6). These fluctuation of these 6 organic variables corresponds to the biodegradation sequence properties of TCE, denoting to these variables varied together in patterns supports the stepwise transformation theory.

Therefore, different sampling sites can have a general transformation consistence of TCE with time after injecting electron donors because of the function of different effective bacterial. Most chlorinated ethene-respiring communities were enriched from contaminated anaerobic environments, and various chlorinated ethene dechlorinating communities have been enriched from pristine anaerobic environments that are not contaminated by chlorinated ethenes (He et al., 2005; Wen et al., 2017).

From the correlation results between selected principle components (PCs) and variables at sampling time points, we can find the correlation between screened PC1 and cDCE, Mn²⁺, chloride and alkalinity are always kept the unity of negative correlation and have relatively high correlation intensity at all detection times from Oct 2011 to Sep 2015.

Relative stable concentration of cDCE, Mn²⁺, chloride and alkalinity can be observed in a year after injecting 3DMe. Hence, cDCE, Mn²⁺, chloride and alkalinity can keep consistent change pattern one year after injection

4.2 The most indicative variables

- Which variables can be used for effectively describing the bioremediation performance of the contaminated sites?

First of all, time since injecting electron donors is an important variable to remediate the contamination sites, it is an indicator which can judge the performance of the

(32)

selected bioremediation techniques (Ostojić et al., 2014; Paul and Smolders, 2015, 2014;

Schaefer, 2016). In addition, all the organic substances detected in this project tend to low concentration and kept at low level steadily or smaller fluctuation from Nov, 2014 to Sep, 2015. This can prove the successful remediation of the TCE-contaminated sites, the bioremediation time is almost uniform to the study result of Florey et al. (2017), a significant decreasing trend of TCE were exhibited from Sep 2010 (injection time) to Mar 2015 and also indicate the successful bioremediation time.

The repeated measures ANOVA analysis showed that the difference of methane, Fe²⁺, Mn²⁺, DOC and alkalinity, among different time points is significant (p<0.05), denoting to these indicators’ detection values have appeared sharp fluctuation within the detection period. Among the four variables, Fe²⁺ and Mn²⁺ can affect the competition of electrons, methane can indicate the procedures of transformation, alkalinity stands for the buffer capacity of the dechlorinating bacteria, DOC represents how much organic carbon sources can be availed by the dechlorinating bacteria. The increased of Fe²⁺ and Mn²⁺ (Paul and Smolders, 2015), the production of methane and the increased dissolved organic chemical and alkalinity can be the evidence to show the enhanced bioremediation of TCE (Lien et al., 2016) proved the indicating function of methane, Fe²⁺, DOC and alkalinity.

4.3 Interaction between variables

- Is there any interaction between measured variables indicating the remediation performance at different sampling sites?

For all variables, they can interact with each other and the interaction changes over time.

the bioremediation processes often accompanied with the reduction of TCE, depletion of the electron acceptor (nitrate, and sulfate, both of them can inhibit dechlorination as these substances both can act as alternative electron acceptors concentrations), production of Fe²⁺, Mn²⁺ and methane, production of TCE degradation byproducts, increased chloride, alkalinity and microbial populations (Lien et al., 2016; Paul and Smolders, 2015, 2014), all of them are indicators of the successful remediation of TCE-

(33)

There appears to be three general types of pattern of change with time for some of the variables. The first group includes methane and ethane, the second group includes chloride, sulfate and alkalinity and the third group consists of cDCE and tDCE.

4.4 Uncertainty and limitations

- About the six chemical parameters of groundwater and the four field indictors, they are easily affected by the external environment, and the extent and level of this influence can be more than the influence brought by bioremediation technique itself.

Hence, it is very hard to analyze how the bioremediation techniques to affect the groundwater chemical and field physical properties.

- When removing outliers or extremely different cases in repeated ANOVA analysis, some useful information may be missed leading to the unreliable results.

- Only some superficial indicators, the concentration of pollutants and some metabolites, physico-chemical properties of the contaminated sites were determined in this project, the different microbial communities, their biomass and microcosms were not investigated in the statistical analyses leading to we cannot know how the internal dechlorinating bacterial to transform the pollutants into nontoxic substances and bioremediate the sites contaminated by TCE.

- Due to the time, fund and techniques restraint of the project only one site was examined. Although a lot of data from a lot of sampling points within the site were gathered and used, more real cases analyses should be performed to add credibility and reliability to this mathematically statistical way and to find more remediation patterns of the contaminated sites by TCE. More measurements for more cases and variables (such as microbial bacterial, biomass and so on) are needed to clearly show the change patterns and paces of variables over time and the differences between cases. This is also benefit to use PCA to show the overall trend more precisely.

Environmental factors may influence the remediation potential and pace of remediation at each sampling point. Field indicators were measured but they were excluded from this study due to time constraints. It is possible that a better understanding of the remediation process would be obtained if these were included or accounted for in the

(34)

analyses and should consequently be included in further studies.

5 Conclusion

- The general trend of tDCE, cDCE, VC and ethene supporting the reductive dechlorination of TCE with time after injecting electron donors.

- cDCE, Mn²⁺, chloride and alkalinity keep a consistent correlation with the screened first principle component already one year after injection.

- The variables that can effectively describe this bioremediation performance are Fe²⁺, DOC, Mn²⁺, methane and alkalinity. Their dramatic changes with time can indicate the active function of dechlorinating bacteria to remediate the contamination.

- Three group indicators can be identified for their trends over time have a certain consistence these three groups are 1) methane and ethane 2) chloride, sulfate and 3) alkalinity; cDCE and tDCE.

(35)

References

AULENTA, F., MAJONE, M. & TANDOI, V. 2006. Enhanced anaerobic bioremediation of chlorinated solvents: environmental factors influencing microbial activity and their relevance under field conditions. Journal of Chemical Technology &

Biotechnology, 81, 1463-1474.

AULENTA, F., PERA, A., ROSSETTI, S., PETRANGELI PAPINI, M. &

MAJONE, M. 2007. Relevance of side reactions in anaerobic reductive dechlorination microcosms amended with different electron donors. Water Res, 41, 27-38.

DAVID, M. M., CECILLON, S., WARNE, B. M., PRESTAT, E., JANSSON, J. K.

& VOGEL, T. M. 2015. Microbial ecology of chlorinated solvent biodegradation.

Environ Microbiol, 17, 4835-50.

DOLINOVA, I., STROJSOVA, M., CERNIK, M., NEMECEK, J., MACHACKOVA, J. & SEVCU, A. 2017. Microbial degradation of chloroethenes: a review. Environ Sci Pollut Res Int, 24, 13262-13283.

FERGUSON, J. F. & PIETARI, J. M. H. 2000. Anaerobic transformations and bioremediation of chlorinated solvents. Environmental Pollution, 107, 209-215.

GOVENDER, A., SHAIK, R., ABBAI, N. S. & PILLAY, B. 2011. Dehalogenase gene detection and microbial diversity of a chlorinated hydrocarbon contaminated site.

World Journal of Microbiology and Biotechnology, 27, 2407-2414.

HE, J., RITALAHTI, K. M., AIELLO, M. R. & LOFFLER, F. E. 2003. Complete Detoxification of Vinyl Chloride by an Anaerobic Enrichment Culture and Identification of the Reductively Dechlorinating Population as a Dehalococcoides Species. Applied and Environmental Microbiology, 69, 996-1003.

KAO, C. M., LIAO, H. Y., CHIEN, C. C., TSENG, Y. K., TANG, P., LIN, C. E. &

CHEN, S. C. 2016. The change of microbial community from chlorinated solvent- contaminated groundwater after biostimulation using the metagenome analysis. J Hazard Mater, 302, 144-150.

KOUZNETSOVA, I., MAO, X., ROBINSON, C., BARRY, D. A., GERHARD, J.

I. & MCCARTY, P. L. 2010. Biological reduction of chlorinated solvents: Batch-scale geochemical modeling. Advances in Water Resources, 33, 969-986.

MAJONE, M., VERDINI, R., AULENTA, F., ROSSETTI, S., TANDOI, V., KALOGERAKIS, N., AGATHOS, S., PUIG, S., ZANAROLI, G. & FAVA, F. 2015. In situ groundwater and sediment bioremediation: barriers and perspectives at European contaminated sites. New Biotechnology, 32, 133-146.

MATTURRO, B., PRESTA, E. & ROSSETTI, S. 2016. Reductive dechlorination of tetrachloroethene in marine sediments: Biodiversity and dehalorespiring capabilities of the indigenous microbes. Sci Total Environ, 545-546, 445-52.

NEMECEK, J., POKORNY, P., LHOTSKY, O., KNYTL, V., NAJMANOVA, P.,

(36)

STEINOVA, J., CERNIK, M., FILIPOVA, A., FILIP, J. & CAJTHAML, T. 2016.

Combined nano-biotechnology for in-situ remediation of mixed contamination of groundwater by hexavalent chromium and chlorinated solvents. Sci Total Environ, 563- 564, 822-34.

NEMECEK, J., STEINOVA, J., SPANEK, R., PLUHAR, T., POKORNY, P., NAJMANOVA, P., KNYTL, V. & CERNIK, M. 2018. Thermally enhanced in situ bioremediation of groundwater contaminated with chlorinated solvents - A field test.

Sci Total Environ, 622-623, 743-755.

NI, Z., VAN GAANS, P., RIJNAARTS, H. & GROTENHUIS, T. 2018.

Combination of aquifer thermal energy storage and enhanced bioremediation:

Biological and chemical clogging. Sci Total Environ, 613-614, 707-713.

PANT, P. & PANT, S. 2010. A review: Advances in microbial remediation of trichloroethylene (TCE). Journal of Environmental Sciences, 22, 116-126.

RUSYN, I., CHIU, W. A., LASH, L. H., KROMHOUT, H., HANSEN, J. &

GUYTON, K. Z. 2014. Trichloroethylene: Mechanistic, epidemiologic and other supporting evidence of carcinogenic hazard. Pharmacol Ther, 141, 55-68.

SCHERR, K. E., NAHOLD, M. M., LANTSCHBAUER, W. & LOIBNER, A. P.

2011. Sequential application of electron donors and humic acids for the anaerobic bioremediation of chlorinated aliphatic hydrocarbons. N Biotechnol, 29, 116-25.

SHEU, Y.-T., TSANG, D. C. W., DONG, C.-D., CHEN, C.-W., LUO, S.-G. & KAO, C.-M. 2018. Enhanced bioremediation of TCE-contaminated groundwater using gamma poly-glutamic acid as the primary substrate. Journal of Cleaner Production, 178, 108-118.

Florey, J., Adams, A., Michel, L., 2017. Reductive dechlorination of a chlorinated solvent plume in Houston, Texas. Remediation 28, 45–54.

https://doi.org/10.1002/rem.21541

He, J., Sung, Y., Krajmalnik-Brown, R., Ritalahti, K.M., Löffler, F.E., 2005. Isolation and characterization of Dehalococcoides sp. strain FL2, a trichloroethene (TCE)- and 1,2-dichloroethene-respiring anaerobe. Environ. Microbiol. 7, 1442–1450.

https://doi.org/10.1111/j.1462-2920.2005.00830.x

Lien, P.J., Yang, Z.H., Chang, Y.M., Tu, Y.T., Kao, C.M., 2016. Enhanced

bioremediation of TCE-contaminated groundwater with coexistence of fuel oil:

Effectiveness and mechanism study. Chem. Eng. J. 289, 525–536.

https://doi.org/10.1016/j.cej.2016.01.011

Ostojić, B.D., Stanković, B., Dordević, D.S., 2014. Theoretical study of the molecular properties of dimethylanthracenes as properties for the prediction of their

biodegradation and mutagenicity. Chemosphere 111, 144–150.

https://doi.org/10.1016/j.chemosphere.2014.03.067

(37)

trichloroethylene as affected by iron reducing conditions: Model systems and environmental samples. Chemosphere 119, 1113–1119.

Paul, L., Smolders, E., 2014. Inhibition of iron (III) minerals and acidification on the reductive dechlorination of trichloroethylene. Chemosphere 111, 471–477.

Schaefer, C.E., 2016. Naturally occurring dechlorination reactions in rock matrices:

Impacts on TCE fate and flux. Environ. Technol. Innov. 6, 115–122.

https://doi.org/10.1016/j.eti.2016.08.003

Wen, L.L., Zhang, Y., Chen, J.X., Zhang, Z.X., Yi, Y.Y., Tang, Y., Rittmann, B.E., Zhao, H.P., 2017. The dechlorination of TCE by a perchlorate reducing

consortium. Chem. Eng. J. 313, 1215–1221.

https://doi.org/10.1016/j.cej.2016.11.021

Chlorinated solvents in groundwater, Government of South Australia. 2017-11-30.

http://www.sahealth.sa.gov.au/wps/wcm/connect/public+content/sa+health+internet/h ealth+topics/health+conditions+prevention+and+treatment/chemicals+and+contamina nts/chlorinated+solvents+in+groundwater.

3-D Microemulsion (3DMe)-Regenesis. 2006. https://regenesis.com/wp- content/uploads/2014/08/HRC_Adv_Tech_Bulletin_1_0_Intro.pdf.

Using PCA & Repeated ANOVA to evaluate the In Situ Bioremediation performance of sites contaminated by trichloroethylene

Master Thesis

HALMSTAD

Master's Programme in Applied Environmental Science, 15 credits

Using PCA & Repeated ANOVA to evaluate the In Situ Bioremediation performance of sites contaminated by trichloroethylene

Applied Environmental Science, 15 credits

Halmstad 2018-5-31

Xinyao Chen

Using PCA & Repeated ANOVA to evaluate the In Situ Bioremediation performance of sites contaminated by

trichloroethylene

Abstract

Contents

1 Introduction

2 Materials and Methods

3 Results and Analysis

礸 /L )

V C ( 礸 /L)

m g H C O /L

4 Discussion

5 Conclusion

References