Acute toxicity of storm water associated with de-icing/anti-icing activities at Canadian airports

Sample locations and collection

The Canadian airports in Toronto (YYZ) and Ottawa (YOW), Ontario; Saint John (YSJ), New Brunswick; and Halifax (YHZ), Nova Scotia, were the selected sampling locations. All samples were collected by Inland Technologies (Mississauga, ON, Canada). Attempts were made to collect samples for each of three de-icing events from two sites at each airport. The two sites were an environmental sampling site, which consisted of runoff water, and samples of the concentrated de-icing/anti-icing chemical collected either from the clean-up vehicles (sweeper trucks) or from the airport ramp. To maximize the potential toxic response (e.g., worst-case scenario, highest recorded glycol levels), all specific sampling sites were preapproved by on-site airport personnel. The original study design required collection during three de-icing events from two sites at each airport; however, weather conditions at the Saint John airport made it impossible to collect samples from both sites during two of the three de-icing events. In total, two events from Halifax (four samples) and three events each from Toronto (six samples), Ottawa (six samples), and Saint John (four samples) were collected.

All toxicity testing was conducted by ESG International (Toronto, ON, Canada). All samples requiring temporary storage were held at a temperature range of 4 to 6°C. Three 20-L buckets of each sample (e.g., the environmental sample or the sweeper truck sample) were collected for testing and mixed immediately before testing in a plastic container lined with a food-grade polyethylene bag. Care was taken to prevent entrainment of air in the sample during mixing. As the sample was mixed, subsamples were taken for the rainbow trout, D. magna, IQ, Microtox, Rotoxkit, and Thamnotoxkit tests.

External quality-assurance testing

Two government laboratories provided external quality-assurance (QA) testing on a selected number of samples. Personnel from Environment Canada (Dartmouth, NS, Canada), conducted QA testing for the Thamnotoxkit and Rotoxkit bioassays. Personnel from British Columbia Environment (North Vancouver, BC, Canada) conducted QA testing for the rainbow trout, D. magna, IQ, and Microtox bioassays. Sweeper truck samples from the first de-icing event from each airport were provided to the external QA/quality-control (QA/QC) laboratories. Samples were sent directly from each airport to the laboratories, which also tested a sample of pure Type I de-icing fluid (i.e., reference material). The results of the QA laboratory tests were compared with those of ESG International to provide an indication regarding the reproducibility of the tests. Toxicity test comparisons were conducted using the least significant difference test [11].

A sample of Type 1 de-icing fluid (1995/1996 Union Carbide XL54 Type 1 De-icing Fluid; Union Carbide, Rye, NY, USA), which is referred to as the reference material, was obtained from the Air Canada de-icing storage facilities in Toronto (ON, Canada). The reference material, which consisted of approximately 54% glycol, 1% additive, and 45% water, was used to determine the response of the test organisms to the full-strength Type 1 de-icing fluid used during the 1995/1996 de-icing period. Test concentrations for the rainbow trout, D. magna, and IQ bioassays were individually prepared by adding a measured volume (by wt) of the test substance directly to the laboratory dilution water. For the Rotoxkit and Thamnotoxkit tests, serial dilutions were prepared from a stock solution consisting of laboratory dilution water and a measured weight of the test material. The Microtox bioassay concentrations were prepared on a volume/volume basis. For comparison, all reference material test results are presented here on a weight/volume basis.

Laboratory dilution water

Natural groundwater was used as a source of laboratory water in the rainbow trout, D. magna, and IQ tests. Reconstituted water or diluent supplied by the manufacturer was used with the Microtox, Rotoxkit, and Thamnotoxkit tests. Laboratory water quality was monitored semiannually and, before use, was filtered through a 20-μ, cellulose-acetate filter and sterilized using ultraviolet radiation. A continuous supply of oil-free compressed air was provided to bring the pH and the concentration of dissolved oxygen as well as other gases into equilibrium with air. The concentration of dissolved oxygen in the water was maintained at greater than 80% saturation. Water used for culture or holding of the test animals was identical to that used for testing purposes.

Test organisms

Rainbow trout used for testing were obtained from a licensed fish hatchery (Rainbow Springs Trout Farm, Thames-ford, ON, Canada). Tests with rainbow trout were conducted using similarly aged fish of uniform size (weight, 0.3–1.0 g). Tests with D. magna (age, <24 h) were conducted using organisms obtained from in-house laboratory cultures. To provide an on-going assessment regarding the fitness of the stock cultures, a 21-d test to assess the health of the daphnid culture was initiated at the beginning of each month. The Microtox, Thamnotoxkit, and Rotoxkit tests did not require culturing of organisms. Lyophilized bacteria for the Microtox test and freeze-dried cysts for the other tests were provided by the manufacturer. For all bioassays, a reference toxicant was used to establish the validity of the effluent toxicity data generated during testing.

Toxicity bioassays

Tests using rainbow trout, D. magna, and Microtox were conducted in accordance with Environment Canada test protocols [12-14]. The IQ, Rotoxkit, and Thamnotoxkit tests were conducted in accordance with the manufacturer's methodology [15-17]. Each bioassay (one effluent sample) included at least one group of control organisms in 100% laboratory dilution water or diluent but otherwise exposed to the same conditions as the test specimens. All bioassays were conducted under static conditions, with no renewal of the test solution. Dissolved oxygen (87–08-00; Orion Research, Beverly, MA, USA), pH, and conductivity were measured in each sample both at the start and the end of testing. Hardness (ethylene-diaminetetraacetic acid titrimeteric method) in the full-strength effluent and dilution water control was measured at the start of each test.

The 96-h rainbow trout bioassays were conducted in temperature-controlled water baths held at 15 ± 1°C. Exposure vessels consisted of 20-L plastic pails fitted with a polyethylene plastic liner. Each test consisted of one replicate, with 10 fish per replicate being exposed to a minimum of five effluent concentrations. Solutions were aerated (˜7.5 ml/L/min) throughout the exposure period. Observations for trout immobility or mortality were recorded after 24, 48, 72, and 96 h. A fish was considered to be dead if it showed no evidence of opercular or other activity and no response to gentle prodding. The 48-h D. magna bioassays were conducted in 55-ml glass test tubes using four replicates per concentration and three daphnids per replicate. All daphnid tests were conducted in temperature-controlled rooms at 20 ± 1°C. Observations for daphnid immobility or mortality were recorded after 24 and 48 h. A daphnid was considered to be dead if it had no visible heart beat at microscopic examination. The test end-point for the rainbow trout and the D. magna bioassays was mortality.

The IQ tests were conducted in a self-contained test system consisting of a six-cell chamber. Daphnids (age, 2–5 d) were starved for 6 to 24 h before testing. Six daphnids were added to each individual test cell, which contained 15 ml of test solution (three replicates per concentration for a total of 18 organisms per concentration). At the end of a 1-h exposure period, the IQ Additive™ was added to each chamber and allowed to incubate for 15 min, after which time each chamber was illuminated with an ultraviolet light and scored visually. The test measured fluorescence of the organisms after the addition of IQ Additive. Decreases in fluorescence related to decreases in galactose metabolism, indicating a toxic effect. The test endpoint was inhibition of fluorescence.

The 15-min acute Microtox test measures the change in bioluminescence of a naturally luminescent marine bacteria (Vibrio fischeri) after exposure to a toxicant or effluent. Changes in bacterial luminescence were assessed using a temperature-controlled photometer (Microbics Model 500 Analyzer; Azur, Carlsbad, CA, USA). All tests were conducted at 15 ± 1°C. The full-strength effluent sample was adjusted to 2% salinity using analytical-grade NaCl. Effluent dilutions and controls were prepared in duplicate. Because of addition of the bacterial reagent and diluent to each concentration, the highest concentration tested was 50%. Measurements of luminescence were made after a 15-min exposure to the toxicant. The test endpoint was inhibition of light production.

The 24-h Rotoxkit tests were conducted using newly hatched (age, 0–2 h) rotifers (Brachionus calyciflorus). Bioassays were conducted in a self-contained test system, which consisted of a 36-cell chamber. Five newly hatched rotifers were added to each individual test cell containing 0.3 ml of test solution (six replicates per concentration, for a total of 30 organisms per concentration). The 24-h Thamnotoxkit tests were conducted in a similar manner, using second- and third-instar fairy shrimp larvae (Thamnocephalus platyurus). Bioassays were conducted in a self-contained test system, which consisted of a 24-cell chamber. Ten larvae were added to each individual test cell containing 1.0 ml of test solution (three replicates per concentration, for a total of 30 organisms per concentration). Test chambers were incubated in the dark for 24 h. At the completion of each test, each chamber was microscopically examined, and the number of living and dead organisms were recorded. The test endpoint for both tests was mortality. An organism was considered to be dead if it showed no movement in 5 to 10 s of observation.

Chemical analyses

All chemical analyses were conducted by Zenon (Burlington, ON, Canada). The environmental and sweeper truck samples were analyzed for 15 chemical parameters. Individual analyses were conducted in accordance with the methods outlined in Standard Methods for the Examination of Water and Wastewater [18] and included pH (4500-H B), total suspended solids (2540 D), biological oxygen demand (5210 B), dissolved oxygen (5210 B), nitrate and nitrite (4500 F), ammonia (4500 H), total kjeldahl nitrogen (4500-N Org B), total phosphorus (4500-P F), phenolics (5530 D), and oil and grease (5520 B). Glycols (i.e., propylene, ethylene, diethylene) were analyzed using gas chromatography with a flame ionization detector. Total petroleum hydrocarbons were extracted using methylene chloride/hexane and analyzed using capillary column gas chromatography with a flame ionization detector.

Statistical analysis of toxicity data

The LC50/EC50 endpoints and 95% confidence intervals for the rainbow trout, D. magna, IQ, Rotoxkit, and Thamnotoxkit tests were calculated using either probit or binomial methods [19]. Results were adjusted for control mortality using Abbott's correction. If immobile organisms were observed in any test for which the EC50 was calculated, the EC50 rather than the LC50 value was used in all comparisons of toxicity and chemical parameters. The EC50 endpoints for the Microtox test were calculated using least-square regressions. For data analysis, samples that were either nonlethal or had an LC50/EC50 of greater than 100% were assigned a value of 100. Microtox tests that had an EC50 of greater than 50% were assigned a value of 50.

For bioassay comparisons, toxicity results from the environmental samples for each of the four locations were pooled. Between-location comparisons were not conducted because of the limited number of samples collected from each location. Sweeper truck toxicity results were treated in a similar manner. To meet the standard assumptions required for analysis, non-parametric methods that assumed no specific distribution of the data were used for all comparisons of toxicity and chemical parameters. Two sets of analyses were conducted on the toxicity data. First, to compare the rapid screening bioassays with the standard toxicity bioassays, toxicity results (i.e., LC50, EC50) for the IQ, acute Microtox, Rotoxkit, and Thamnotoxkit tests were statistically compared with results from the rainbow trout and D. magna tests using the Wilcoxon signed rank test [20]. The signed rank test, which is a nonparametric test that assumes no specific distribution of the data, was used to determine if the LC50 or EC50 values of the standard bioassays were significantly different from those of the rapid screening bioassays. If these values are not significantly different from those of a standard bioassay, the rapid screening test may have the potential for use in place of the standard test. Alpha levels (a) were set to 0.05. The second set of analyses involved use of a ranking procedure to determine the relative sensitivity of each of the six bioassays. The most sensitive test (e.g., that with the lowest LC50 or EC50) was assigned the lowest number, and the least sensitive test was assigned the highest number. The ranks were then averaged for each test to yield a mean rank for the environmental and the sweeper truck samples [21]. The ranking process is limited, however, because a statistical difference among the response parameters could not be established.

Correlation of chemistry with toxicity data

The LC50 and EC50 test results were compared with the chemical parameters using nonparametric correlation analysis. Kendalls' tau (τ) was used to test for a significant correlation between the toxicity bioassays and each of the measured chemical parameters. Bonferroni's p value adjustment was used to put a more stringent limit on rejection of the null hypothesis. Consequently, any correlation with a p of greater than 0.0036 was considered to be nonsignificant.

Toxicity results

Summaries of the toxicity data, expressed as percentage effluent by volume (% v/v), are presented in Table 1 for the sweeper truck samples and in Table 2 for the environmental samples. Four of the 10 environmental samples were lethal (LC50 < 100%) to trout, and three were lethal to D. magna. The many nonlethal or no-effect responses to the environmental samples did not allow for statistical analyses of the toxicity or the chemistry data. Only the results of the sweeper truck samples could be analyzed statistically. For each sweeper truck sample, the toxicity results of the rainbow trout and the D. magna bioassays were statistically compared with the results obtained using the rapid screening acute lethality tests. The final sweeper truck sample from Toronto was not tested using the Thamnotoxkit test because of problems associated with hatching of the cysts; thus, this sample was not included in the statistical analyses. The IQ and acute Microtox test results were not significantly different from those of the rainbow trout bioassay (p = 0.58 and 0.09 for the IQ and Microtox comparisons, respectively), and the Thamnotoxkit results were not significantly different from the those of D. magna bioassay (p = 0.58). All other test results were significantly different from those of the trout and the D. magna bioassays (p < 0.05). Results of the ranking procedure used to determine the relative sensitivities of the six bioassays indicated that for the environmental samples, the IQ bioassay was the most sensitive test, followed in decreasing order by the rainbow trout, D. magna, and Thamnotoxkit tests. The Rotoxkit and Microtox tests were the least sensitive for the environmental samples. Similarly, the IQ test was the most sensitive test for the sweeper truck samples, followed in decreasing order by the rainbow trout, Microtox, D. magna, Thamnotoxkit, and Rotoxkit tests (Table 3).

The toxicity results from the environmental samples could not be analyzed statistically, but inspection of these data indicated that the toxicity responses of rainbow trout and D. magna varied among the samples collected at the four locations. Three of the 10 samples tested were observed to be lethal to trout but nonlethal or only moderately toxic (LC50 > 100%) to D. magna. Similarly, two of the 10 samples tested were lethal to D. magna, but nonlethal or only moderately toxic to trout. Effluent collected from the Ottawa airport during the third de-icing event was the only environmental sample observed to be nonlethal to both trout and D. magna. Analysis of the sweeper truck toxicity results indicated that the IQ and rainbow trout bioassays were comparable. However, inspection of the environmental sample toxicity results suggested that compared with the trout bioassay, the IQ test overestimated the acute lethal toxicity of the environmental samples. For five of the 10 environmental samples, the IQ test indicated that the samples were toxic (EC50s, 3–75%), whereas the rainbow trout bioassays indicated that the samples were nonlethal (Table 1). These results also suggest that the IQ test may be responding to toxicants in the environmental samples that are not present in the sweeper truck samples.

Chemistry

A summary of the chemical analyses and toxicity tests are presented in Tables 4 and 5. Ethylene glycol was the only glycol detected in the sweeper truck or environmental samples. All propylene and diethylene glycol measurements were less than the detection limits. This was not unexpected, however, because the airports studied used ethylene-based ADAFs. Only six of the 15 detected parameters had significant negative correlations with the toxicity tests (p < 0.001). Trout and D. magna toxicity correlated with pH and nitrate. Daphnia magna toxicity also correlated with nitrate + nitrite concentrations. The IQ test results correlated with pH, TSS, and nitrate, and the Microtox test results correlated with nitrate and phenolics. The Rotoxkit test results correlated with pH and nitrate, and the Thamnotoxkit test results correlated with pH and dissolved oxygen. Except for the Thamnotoxkit, all bioassays correlated significantly with extractable petroleum hydrocarbons. Although no significant correlation was established between toxicity and dissolved oxygen concentration, inspection of the rainbow trout and D. magna raw data indicated that in many (but not all) cases, several exposure concentrations had measured dissolved oxygen concentrations at or below levels that would contribute to trout mortality (<5 mg/L). Inspection of control dissolved oxygen concentrations demonstrated that the fish themselves were not responsible for the decrease in oxygen; therefore, another cause must have depleted the dissolved oxygen concentrations. Glycols have the potential to contribute to oxygen depletion in aquatic ecosystems [1, 2], but whether ethylene glycol may have caused a depletion of dissolved oxygen in the trout bioassays, which in turn contributed to mortality, is not known.

The number of significant correlations to a specific parameter was noted for the rapid screening tests to determine how each agreed (in terms of chemistry) with the standard bioassays. To give a percentage agreement value with the standard bioassays, the number of chemical parameters for a rapid screening test that were similar to the rainbow trout or D. magna toxicity tests were divided by the number of parameters that correlated with the rainbow trout or D. magna toxicity tests [22]. Comparisons of the chemical parameter results between the standard and the rapid screening bioassays indicated that the IQ and Rotoxkit test agreed (in terms of chemistry) 100% with the rainbow trout bioassay and 75% with the D. magna bioassay (Table 6). However, compared with the trout bioassay, the IQ bioassay also significantly correlated with an additional parameter (agreement >100%).

Reference material test results

The reference material tests demonstrated good agreement with the QA laboratories, except for the IQ test (Table 7). No significant differences between test facilities were observed for the D. magna, Rotoxkit, or Thamnotoxkit tests (p > 0.05), whereas only slight differences were observed for the rainbow trout and the Microtox tests. Differences in mixing procedures, laboratory dilution water quality (in the case of the rainbow trout test), or pipetting technique (in the case of the Microtox assay) may explain the differences observed. Differences in IQ test results may have resulted, in part, from differences in the mixing procedures or laboratory dilution water quality, but are more likely a result of subjectivity in endpoint measurements.

Type II anti-icing fluid was not tested, but recent results indicate that these solutions are more toxic than either pure glycol products or de-icing solutions [2, 7]. Recent toxicity data on the ethylene glycol–based ADAFs currently used in Canada indicate LC50s for rainbow trout of 22,810, 17,368, and 251 mg/L for pure ethylene glycol, de-icer, and anti-icer products, respectively [1, 2]. Enhanced toxicity to aquatic invertebrates has also been reported, with 7-d EC25s for Ceriodaphnia sp. of 9,226, 1,181, and 54 mg/L for pure ethylene glycol, de-icer, and anti-icer products, respectively [1, 2]. The difference in toxicity between the Type I and II products was thought to relate to the particular suite of additives in the formulated anti-icing product.

Expression of toxicity test results as ethylene glycol

A relationship between ethylene glycol concentrations and toxicity could not be established using correlation analyses. Therefore, a second approach was used to further clarify if measured glycol concentrations could explain the toxicity of the storm water samples. Toxicity values (%, v/v) for the sweeper truck and environmental samples (Tables 1 and 2) were calculated in terms of mg/L ethylene glycol, and the measured concentration of ethylene glycol in the full-strength effluent was used to convert the toxicity values from percentage effluent by volume (%, v/v) to mg/L ethylene glycol (Tables 8 and 9). Toxicity results for samples that were non-lethal or had an LC50 or EC50 of greater than 100% (or <50% in the case of the Microtox test) are presented as values greater than the highest concentration of ethylene glycol as measured in full-strength effluent.

The toxicity of ethylene glycol as a pure compound was not tested during this study, but published LC50s for ethylene glycol range from 17,800 to 50,800 mg/L (96 h) for rainbow trout and from 46,300 to 56,400 mg/L (48 h) for D. magna [1, 23]. Comparisons of the LC50s for pure ethylene glycol with the sweeper truck and environmental toxicity tests, expressed as mg/L ethylene glycol (Tables 8 and 9), indicated that ethylene glycol was not the only substance contributing to toxicity. In various instances, the samples were either more or less toxic than would be predicted based on the LC50s for ethylene glycol alone. Furthermore, comparisons of results with the reference material (i.e., pure ADAF) tests (Table 7) to the sweeper truck and environmental toxicity tests (mg/L ethylene glycol) also suggests that other substances were contributing to toxicity.

Expression of the sweeper truck and environmental sample toxicity data in terms of mg/L ethylene glycol can only be used indirectly to explain toxicity, because the concentrations of ethylene glycol were measured only in the full-strength effluent and not directly in any of the exposure concentrations. However, the data indicated that for the sweeper truck and the environmental samples, ethylene glycol was not predictive of acute lethal toxicity to rainbow trout or D. magna. These findings are supported by those of other studies, which have shown that whereas glycols may contribute to acute toxicity, their presence alone cannot explain all observed ADAF toxicity. Pillard [10] tested the toxicity of various pure glycol compounds as well as glycol-based ADAFs. Compared with the pure glycol product, the formulated products were more acutely and chronically toxic to both fathead minnows and Ceriodaphnia dubia. The 96-h LC50s for fathead minnows were 72,860 mg/L using pure ethylene glycol and 8,050 mg/ L using the formulated product. Similarly, the 48-h LC50s for C. dubia were 34,000 mg/L using pure ethylene glycol and 13,140 mg/L using the formulated product. A similar response was also observed for the chronic toxicity test data. Fisher et al. [8] observed that storm water LC50s based on total glycol were lower than what would be predicted from the pure glycol data. They suggested that glycols were a main source of toxicity, but that additives in the ADAFs also contributed to acute toxicity.

Additives such as rust inhibitors, surfactants, wetting agents, antioxidants and dyes are all possible sources for the significantly enhanced toxicity of formulated ADAFs (particularly anti-icers) compared with glycols [1, 7]. Unfortunately, information regarding the specific additives is unavailable, because they are considered to be propriety by the manufacturer. The source of ADAF toxicity is also confounded by frequent changes in the formulation chemistry as products become more efficacious. Recently, Cancilla et al. [6] attempted to isolate and to identify the most toxic ADAF additives. The benzotriazole and tolytriazole antioxidant components of the additive package were identified as having a significant effect in the Microtox assay; both products are commonly used as corrosion inhibitors. Other contaminants such as urea, petroleum hydrocarbons, and cleaning solvents may also contribute to airport storm water toxicity during de-icing and non de-icing events. Pillard et al. [10] observed that the toxicity of storm water to daphnids during non de-icing events appeared to relate to fuel spills, and the concentration of total petroleum hydrocarbons was measured at 28 mg/L. In field studies, Turnbull and Bevan [7] found that increases in receiving-water ammonia concentrations coincided with applications of urea as a de-icer to airport runways. Results from in situ bioassays suggest that ammonia generated from the urea application was responsible for Gammarus pulex mortality. Decreases in benthic diversity in the vicinity of the runoff were also attributed to these increased ammonia concentrations.

In comparison, laboratory studies conducted by Fisher et al. [8] concluded that while elevated levels of total nitrogen were observed in composite storm water samples collected during a de-icing event, the concentration of un-ionized ammonia was less than that expected to contribute to daphnid or fathead minnow toxicity. For most samples tested during this study, concentrations of un-ionized ammonia were also not expected to contribute to daphnid or trout toxicity. However, concentrations of unionized ammonia and nitrite in two of the 10 storm water samples were at or greater than the levels that would be expected to cause trout mortality (assuming LC50s for ammonia and nitrite in the range of 0.16–1.1 and 0.19–0.39 mg/L, respectively) [24]. Concentrations of unionized ammonia and nitrite in the second sample of storm water collected from Saint John were measured at 1.1 and 4 mg/L, respectively, and concentrations of unionized ammonia and nitrite in the first sample of storm water collected from Ottawa were measured 0.25 and 0.23 mg/L, respectively. The corresponding toxicity data indicates that storm water collected during the second event at Saint John was the most acutely lethal sample tested during this study (LC50 for trout, 12%) (Table 2). The first event from Ottawa was also the only sample from this location that was toxic to rainbow trout. This does not imply that ammonia or nitrite are solely responsible for the toxicity of this sample, but it does indicate that concentrations of nitrogen-based substances, and of ammonia in particular, can reach lethal levels in airport storm water.

Performance evaluation of rapid screening bioassays

Results of the bioassay comparisons indicated that the rapid screening tests were easy to conduct, required little time to obtain results, and in several cases, produced results similar to the “traditional” rainbow trout and D. magna bioassays. The rapid screening tests are not intended to replace the traditional bioassays, but their speed makes them practical for use in screening a large number of samples. The main objective of this study was to evaluate the effectiveness of these alternative rapid screening bioassays to predict the responses of the standard regulatory tests; therefore, if an alternative test did not produce toxicity results that were similar to those of the traditional bioassays, it was not considered to be an appropriate test. The Thamnotoxkit was the only rapid screening test that produced toxicity results similar to the D. magna bioassay. Likewise, the IQ and Microtox tests were the only rapid screening tests that produced results similar to the rainbow trout bioassay.

The IQ and Microtox tests were comparable to the rainbow trout bioassay in terms of toxicity, but only the IQ test agreed in terms of both toxicity and chemistry. However, several disadvantages are associated with the IQ bioassay, including the lack of a standardized QA/QC program, subjectivity in end-point measurements, and replacement of tests that measure lethality with one that measures a sublethal response. The lack of reference toxicant and QA/QC data for the IQ test may influence the quality and reliability of the test data. Without proper QA/QC, the quality or health of the test organisms, abilities of the technician conducting the test, and conditions in the laboratory at the time of testing cannot be measured. If the IQ test is to be chosen as the best alternative acute toxicity test for routine monitoring of de-icing operation effluents at airports, then better criteria for endpoint estimation, mandatory reporting elements, and QA/QC requirements (e.g., reference toxicity testing) must be defined in the test method. (After completion of this study, a newer version of the IQ test kit was released that addresses issues related to QA/QC.)

Subjectivity in endpoint measurements was a disadvantage associated with the IQ test. Completion of the IQ test involved visually comparing the fluorescence of each daphnid in the exposure concentrations to the controls. A decrease in fluorescence related to a decrease in metabolism, indicating a toxic effect. Because the degree of fluorescence is based on visual observations only, subjectivity in endpoint measurements may result in variable test results, both within and between laboratories, as seen in the reference material IQ test results. Most important, relative to the rainbow trout toxicity test results, the IQ test may overestimate the acute lethal toxicity of an effluent. In several cases, the IQ test was significantly more sensitive compared with the rainbow trout test. These results are not surprising considering that the IQ test measures a sublethal endpoint (i.e., metabolism), whereas the trout bioassay is strictly designed to measure lethality.

The Thamnotoxkit test produced toxicity results comparable to the D. magna bioassay, but the two main disadvantages associated with the Thamnotoxkit were the lack of agreement with the D. magna bioassay in terms of chemistry and the problems related to hatching of the cysts. In terms of chemistry, none of the rapid screening tests agreed completely with the D. magna bioassay. However, the Thamnotoxkit test was the least similar to the D. magna bioassay, whereas the IQ and Rotoxkit were the most similar. As mentioned, the second disadvantage of the Thamnotoxkit test related to problems associated with hatching of the Thamnotoxkit cysts. The process of freeze drying the Thamnocephalus sp. cysts will often vary, resulting in variations of the time to hatch (M. Sukkel, personal communication). Most cysts received for testing in this study required 24 h to hatch. However, several batches of cysts obtained during the latter part of the study required 72 h to hatch. The 72-h cysts appeared to be of poor quality, because numerous vials did not produced any neonates, even after 96 h. Because of the unpredictable timing of de-icing events, a principal requirement of any bioassay for monitoring airport storm water is that it can be initiated as soon as the weather conditions dictate collection of a sample. Thus, use of the Thamnotoxkit may be less practical than use of the standard bioassays given the often variable and unreliable cyst hatching times.