Elsevier

Chemosphere

Volume 80, Issue 5, July 2010, Pages 542-547
Chemosphere

Can site-specific heuristic toxicity models predict the toxicity of produced water?

https://doi.org/10.1016/j.chemosphere.2010.04.040Get rights and content

Abstract

An empirically derived model of major ion toxicity was combined with other toxicity assessments to account for the observed toxicity in field-collected produced water and produced water contaminated groundwater. The accuracy and precision of the ion toxicity model, calculated using model deviation ratios (MDR) and simple linear regressions, was determined for fathead minnows, Ceriodaphnia dubia, and Daphnia magna. Model accuracy for produced water fell within a factor of two for all three organisms. The precision, or variability explained by the model, was 47.9%, 56.1%, and 0.00% for fathead minnows, C. dubia, and D. magna, respectively. Incorporating other measured potential toxicants improved predictive precision for fathead minnows to 67.0% using ion toxicity and pH and to 30.9% for D. magna using ion toxicity, pH, and total ammonia. The observed toxicity to Daphnia pulex was also evaluated using D. magna model predictions and other measured parameters, but no consistent relationship was found. Dissimilar results were found for produced water contaminated groundwaters with model predictions for D. magna falling within a factor of two of and explaining 53.8% of the observed variability in D. pulex responses. These results indicate that predicted major ion toxicity, combined with other measured parameters, can accurately and precisely account for observed responses in test organisms to field-collected samples.

Introduction

Produced water is water that is co-produced during crude oil recovery. It is a complex mixture containing many potential toxicants including major ions (Na, Ca, K, Mg, Cl and SO4), ammonia, hydrogen sulfide, petroleum hydrocarbons, BTEX (benzene, toluene, ethyl-benzene, and xylenes), phenols, naphthalenes, PAHs, zinc, and other heavy metals (Fucik, 1992, Schiff et al., 1992, Kharaka et al., 2005, Benko and Drewes, 2008). These constituents vary between and within different geologic basins depending on geology and hydrology (Collins, 1985, Daly and Mesing, 1995, Benko and Drewes, 2008).

During inland oil production, produced water is commonly reinjected back into the oil seam, potentially allowing the contaminants contained in the produced water to enter nearby groundwaters and related surface water. For example, this process has been reported to result in major ion, bromide, arsenic, cadmium and copper contamination in groundwater within the area of injection (Hudak and Blanchard, 1997, Sadiq and Alam, 1997, Okandan et al., 2001). Organic contaminants, including oil, grease, and aromatic hydrocarbons, also occur in produced waters and can be found in contaminated groundwaters (Kharaka et al., 2005, Benko and Drewes, 2008). At the field site for this study, along Skiatook Lake, Oklahoma, excess major ion contamination occurs at a near-surface aquifer and the lake itself (Herkelrath and Kharaka, 2002, Kharaka et al., 2005, Kharaka et al., 2007).

As produced water leaches into surface water, there is potential for toxicity to aquatic organisms, and characterization of this toxicity may help direct management and use of the water. Prediction of the toxicity of produced water to aquatic organisms can be challenging because produced water is a complex mixture that can be composed of different potential toxicants at different sites and that composition may vary both spatially and temporally at the same site. Therefore, an approach to predict the toxicity based on concentrations of the contaminants in a specific sample of produced water requires flexible modeling approaches that can account for variability in contaminants and concentrations.

Several models exist to predict the effects of toxicant mixtures. Predictive models that only require individual toxicity testing of each contaminant are available for contaminants that generally act similarly (concentration addition, Altenburger et al., 2000) and for contaminants that act independently (independent action, Backhaus et al., 2000). Other models have been developed by empirically testing mixtures of contaminants and establishing regressions to predict toxicity. For example, derived models have been established to estimate the joint toxicity of metals (Di Toro et al., 2001), major ion salts (Mount et al., 1997), and narcotic organics such as PAHs (Swartz et al., 1995).

The model derived by Mount et al. (1997) is an empirically-derived statistical model to predict major or essential ion (Na, K, Ca, Mg, Cl, SO4, and HCO3) toxicity to three standardized bioassay organisms (fathead minnows (Pimphales promelas), Ceriodaphnia dubia, and Daphnia magna). Because these ions may account for much of the toxicity of produced water, an ion toxicity model has been suggested as a technique to predict toxicity of produced water. The results of the model can then be used to supplement toxicity identification evaluations (Tietge et al., 1997). In their study, Tietge et al. (1997) used modeled ion toxicity to screen produced waters for toxicity and suggested using modeled toxicity as a screening tool for future produced water studies. Produced waters where observed toxicity was greater than predicted (fell outside 95% confidence intervals) were then subjected to phase I toxicity identification evaluation (TIE, US EPA, 1991) procedures to determine other toxic constituents.

The goal of this study was to use the ion toxicity model and other mixture and statistical modeling approaches to assess the toxicity of produced water. The specific objectives were to (1) further evaluate the accuracy and precision of the ion toxicity model for predicting the toxicity of produced water and produced water-contaminated groundwater specific to a field site at Skiatook Lake, Oklahoma, (2) utilize TIE and regression analysis to determine the relative role of non-ion stressors within the produced water samples, and (3) evaluate the applicability of the toxicity predicted by the ion toxicity model for produced water to species other than those used to develop the model.

Section snippets

Sampling and chemical analyses

Samples were collected at a site along Skiatook Lake, a 4249 hectare reservoir in Osage County, northwest of Tulsa, Oklahoma. Salt scars, characterized by bare soil and salt crystals on exposed rocks, ran to the lake from two evaporation ponds located immediately down slope of a storage tank battery and a produced water reinjection well.

Produced water was collected from onsite storage tanks in one-liter amber bottles with Teflon coated lids, filled from the bottom with a hose to reduce volatile

Produced water

Acute 48 h LC50′s ranged from 7.44% to 11.2% for fathead minnows, 2.06% to 2.74% for C. dubia, 2.68% to 5.36% for D. magna, and 0.94% to 4.13% for D. pulex. Comparisons between observed 48 h LC50’s and those predicted from major ion concentrations differed depending on the test organism, however those differences were similar to those found using laboratory derived salt mixtures (Fig. 1). Fathead minnow lethal concentrations were most similar to predicted values with a raw MDR of 1.15, which when

Model accuracy and precision

In this study, predicted major ion toxicity alone was representative of observed toxicity for both fathead minnows and C. dubia. Model deviation ratios, indicative of effect magnitude similarity between predicted and observed or accuracy, fell within a factor of two for three species. This degree of deviation is similar to that found for other mixture models. For example, Belden et al. (2007) found that for pesticide mixture experiments where concentration addition was expected, greater than

Conclusions

The existing model for major ion toxicity to fathead minnows and C. dubia was representative of observed responses and accounted for both the magnitude and variability observed in bioassays for these two organisms. Predicted ion toxicity to D. magna was only representative of observed D. magna in groundwater samples and not produced water samples, possibly because of other toxic constituents in the produced water samples. Though the specific models developed here only apply to the study site,

Acknowledgements

We gratefully acknowledge the assistance of N. Cooper, M. Abbott, C. Boeckman, and K. Burgess for assistance with data collection. This research was funded through a cooperative agreement with the US Department of Energy (DE-FC26-04NT15544).

References (27)

  • Daly, D., Mesing, G., 1995. Atlas of Gas Related Produced Water for 1990. Gas Research Institute Topical Report...
  • D. Di Toro et al.

    Biotic ligand model of the acute toxicity of metals. 1. Technical basis

    Environ. Toxicol. Chem.

    (2001)
  • K. Fucik

    Toxicity identification evaluation and characteristics of produced water discharges from Colorado and Wyoming

  • Cited by (6)

    • A combined evaluation of the characteristics and acute toxicity of antibiotic wastewater

      2014, Ecotoxicology and Environmental Safety
      Citation Excerpt :

      However, toxicity test cannot be conducted in many Chinese pharmaceutical wastewater treatment plants because of lack of equipments, professional operators, and other items. Previous studies confirmed that wastewater toxicity could be accurately predicted by physicochemical parameters (Fischer et al., 2010). By applying stepwise linear regression analysis on the experimental data, factors that statistically influence toxicity significantly could be identified.

    • Ecotoxicological evaluation of selected pharmaceuticals to Vibrio fischeri and Daphnia magna before and after photooxidation process

      2014, Ecotoxicology and Environmental Safety
      Citation Excerpt :

      Microtox® is though a widespread acute test as it is very sensitive, reproducible, possesses high discriminant power for organic and inorganic pollutants. The ecotoxicological tests are frequently used to determine the impact of contaminants on living organisms (Fisher et al., 2010; Manusadzianas et al., 2003; Naddy et al., 2007). However, their application in the qualitative analysis of photocatalytically treated waste water containing PPCPs is quite occasional (Marugan et al., 2010; Schnell et al., 2009).

    View full text