Document JNXmJ0m3XKb6NpO5159kxK742

ATTACHMENT II CMA EXHIBIT A EVALUATION OF BIOLOGICAL DATA USED IN DERIVING ERA WATER QUALITY CRITERIA (as stated in the May IS, 1978 Federal Register Vol. 43, No. 97, pp. 21506-21518) by John Cairns, Jr. Arthur L. Buikema, Jr. James G. Geiger of Virginia Polytechnic Institute and State University Prepared for: American Petroleum Institute August 2, 1978 CMA 049439 _1 , The following comments were based on review of approximately eighty-five percent of the available literature. Some personal communications were missing or incomplete, and certain references were not available for use. .Since some of the data cited in the document were reordered, recalculated, or otherwise used, it was sometimes difficult to trace their oriqins. Consequently some of the comments herein mav not be applicable. In this report, references (numbers in parentheses) are citations specific to each table. SUMMARY OF FINDINGS The following summary is based upsn a review of the proposed criteria development methodoloqv document and approxi mately eighty-give, percent of the references cited therein. MA-iC The proponed. MATC calculations are too restrictive. The criteria preclude usinq actual MATC data even if the actual MATC data prove that a toxicant is less harmful than predicted by calculation. Inadequate Chronic Data Rase The guidelines of the criteria methods indicate that no criterion should be recommended without chronic values and acute continuous flow tests where actual toxicant concentrations are measured in the test chambers. Chronic tests require a long time, and few facilities for chronic testing exist. Since few chronic studies have been repeated, we question the reproducibility of chronic test data. Chronic studies are subject to many vari ables not usually experienced in acute te.sts--for example, seasonal and shorter cycle temperature variations, changes in test water quality over the period of exposure, and feeding and nutritional problems. Chronic test values have qreater variability than do acute tests, due to problems associated with larval viability, spawning times, egg production rates, hatchability rates, and the like. We feci that chronic data can be used, to establish water quality criteria, provided additional efforts to confirm re ported chronic results are carried out. Different Water Characteristics The criteria method development document not moot is stated objective, namely, to "provide a basis for deriving CHA. 049440 criteria specific to different water characteristics for pollutants whose toxicity varies significantly under different conditions." This is based on 'the obvious bias for soft-water data (and other factors), as indicated by the selection of data from cited references. If the criteria development methods were set up to meet this objective, the design methods and rationale were not presented. We therefore cannot comment on the methods and ap plications relative to the stated objective. Water Quality Interactions Variations in such physical parameters as temperature and in such chemical parameters as pH and hardness affect the sensitivity of organisms to many pollutants. Yet the data used to develop static vs. flow-through, nominal vs. measured, and time of ex posure correction factors were selected under a wide ranqe of temperature and water quality regimes. These data should not be grouped to develop a single correction factor under the above physical and chemical variables. The correction factor for sensitivity does not allow for water quality interactions that affect oraanism sensitivity to toxicants. "More information is needed about the effects of water quality in chronic toxicity for the sake of extrapolating from one exposure condition to another." (Andrews, e_t a_l. , un published. ) Identification of Data As noted above, much of the data in the tables was difficult to trace to original references, because the data had been re calculated and reordered (especially references in Table 4, M2 a-o) . Additionally, a number of transcription errors in our partial review of cited references were identified. Inappropriate Units Data are unitless throughout the tables. Wc found that in most of the references, data were reported as mn/1, but the units for some data included in the tables were actually ltg/l. These unitless values were treated by EPA as numerical equivalents in data manipulations, resulting in inaccurate correction factors. These 'factors should be recalculated. Data Selection Some data were selected from referenced citations. Other relevant data in the same citations were not used--a practice which may well bias the correction factors. The rationale for this Partial selection should have been presented. Most of the data used to develop the correction factors were based on pesticide toxicity. This probably skews the enr- CMA 049441 rection factors toward pesticides. Data presented in the tables were selected from predominantly soft-water conditions. Much of the hardwater data in the same references were not used. Again, this selection biases the cor rection factors, since some toxicants, especially heavy metals, are more toxic to organisms in soft water than in hard water. LCi-g Determination LC_n values in the references cited, were calculated by dif ferent methods, such as graphic interpolation (which has no con fidence limits), computerized probit analysis (several methods), and moving averages. The data should all be calculated by the same method so as to reduce error in subsequent correction factor calculations. * LCc,, and ECr, Values LC^q and EC-g values were treated as numerical equivalents in all calculations. They are not measures of the sane thine. EC50 values are measures of immobilization of an organism, whereas LC^g values are measures of death. Not only is immobilization more difficult to quantify accurately, but also it mav be temor- arilv followed by subsequent animal recovery. values being more restrictive than LCrn values, use of EC-q dai:a resulted in correction factors that arc too stringent. Species Sensitivity As expected, different species have different sensitivities to toxicants. This is due to different habitat and environmental requirements, physiological requirements, age and size differences, reproductive cycles, and other factors. We suspect that it is invalid to lump all fish data and all invertebrates data together to determine correction factors. Acclimation of Test Species Improper acclimation of test snecies mav modify the sensi tivity of organisms to toxicants. Seme references that we ex amined did not specify the source of the test organisms, and others specified field populations that mav have been stressed by pollutants m the natural habitat. Still other reference, did not state their methods for acclimating organisms to test conditions prior to toxicant exposure. We therefore do not feel that these data should be used for correction factor calculations. Test Methodology Data collected by static, static-wjth-renewal, and continuous- CMA. 049442 -4- flow techniques probably will not qive the same LC-n values. In table 4, data frofr all three exposure methods were^used to gen erate correction factors. This is inappropriate. RECOMMENDATIONS 1. EPA should conduct an experiment specifically designed to test the validity of their correction factors. The experiment should be designed with the aid of a statistician and should include simultaneous tests on the following variables: differ ent species; distinctly different classes of compounds and various states of these compounds that would be found in the environment; progressive lengths of exposure, for example, readings at 24, 48, 72, and 96 hours; different physical/chemical water quality regimes; and test conditions, such as static flow-throuqh. Analysis of variance and covariance (see Wickramarartne) should be used to establish and correct for these interrelationships prior to any calculations of correction factors, 2. Actual MATC values should be used when they exist. If actual MATC data are excluded, the reason for exclusion should be stated. 3. A more balanced data pool--that is, a pool not so heavily skewed to pesticides--should be used to calculate correction factors. EPA should not use data where the history of origin and acclimation of test organisms are not documented or are questioned. These data should be eliminated prior to re calculation. EPA should arid the data in th^ cited references which were not used in calculations and as well should seek additional data to balance soft-water and hard-water conditions in order to approximate the range of water qualities in the United States. Appropriate statistical methods should then be used to quantify these interactions. The same comment applies to pH, temperature, and type of toxicant (based on its mode of action, solubility, ionic or chelation state, and the like). 5. All data in the tables should be compared to the original literature sources, and the values should be corrected prior to recalculation of geometric means and variance (or, preferably, regression analysis, analysis of variance;, and analysis of covariance. (see Wickramarartne). The data should all be converted to consistant units before any calculations are made. 6. Only LC^q data with stated confidence limits should be used in any calculations. All values derived by graphic inter polation should be eliminated prior to recalculation. CMA 049443 7. Quotients of HC/LC^q or nc^() should be recalculated. Quotients should be based on i.C^g or HC^g divided by the J,Crg or EC5q , respectively. The LC-,g and HC^g data base should be used to compensate for random death of test orqanisms that occurs under experimental conditions. Where bC^g and EC-jn data are not available, then we suggest this value be calculated and used to correct for random death. This approach would broaden the data base and affect the geometric mean. 8. The criteria development document should (a) clearly reference the source of data points; (1)) indicate how' and why data were recalculated prior to inclusion in the tables; and (c) state which criteria were used in partial selection of data from tlie original' references. 9. The document should include a statement about how the "basis for developing criteria specific to different water characteristics" was determined. Page 21507 SPECIFIC CC'vv'N'T'S It is true, as stated, thn* I,rc g values are not a 1wavs comparable; but there are many factors in addition to tho?^ named, lencrth of exposure concentration, and type of tc^i system. The discussion in the docum'-n*: iur.orcs t!v inter action of the more obvious phvsical/chemicul parameters such as venecrature, pH, water hardness (and other water quality factors) , and the effect of previous acclimation and history on test organisms. Temperature influences organism sensitivity, and most organ isms arc more sensitive to toxicants at hichcr temperatures. Toxicities of trace metals, ammonia, certain phenciics, and some pesticides are affected by variation.; in pH and/or water hardness. These interactions are ignored throuuhout all correction fact.or calculations. Examination of cited data showed that temperatures ranged from R ' to 28C and hardness values from <40 to >3n0::om. Table 1 No comments. Table 2 Quotients presented may reflect data which would invalidate their use. Specifically: (1) Temperaturo (a) Values vanqcd from 12' to 28 V, as determined by examination of selected references in the table; and (b) Several papers did not state test temperatures. (References 2, 5, fi, d, 12, 11.1 CMA 049444 6- - (2) Hardness The majority of data were selected for "soft water" (generally a worst case), that is, hardness <60ppm. A significant amount of data for'well-defined "hard water" in these same references were ignored. This practice skews the data to a "worst case" situation and biases the correction factor. More importantly, it is inconsistent with the stated objective, of pro viding a "basis for deriving criteria specific to different water conditions." We should call attention ' to the use of one reference (19) which has several toxicity values reflecting hard-water conditions (300mg/l). (3) Water Quality Data , In one paper examined (reference 14), no water chemistry data were presented. (We did not examine all references.) Without assurance that water quality data meet strict EPA bioassay guidelines, the toxicitv data should not be used in calculations of correction factors. (4) Calculations LC50 and ECc;q values were computed in several ways-Litchfield-Wilcoxcn, computer probit analyses, graphical interpolation, and so on. It is not appropriate to combine these values in correction factor calculations. They should all be converted to a standard calculation. Graphic interpolation values, especially, either should be converted to a stronger calculation that has stated confidence intervals, or should not be used. ( 5) Incomplete Data Usage Data existed for some additional compounds in the references EPA cited, but these values were not used. For example, Reference Comncund 6 Toxaphene 4 101 compounds (only one cited) 10 Mercury For several papers, 24-, 48-, or 7?-hour ^50 values were not reported, yet ouotients were reported (Table references 3, 11, 33, 35, 38). CMA. 049445 -7- References 7 and 15 contained information for certain fish species which were not used. In reference 11, no LC^q values for mercuric chloride were identified. We presume the LC50 value for methylmercuric chloride was used. Tablo 3 Data in Table 3 (of those references we verified) wore based on toxicity studies conducted on five species of fish. Experimental conditions varied as follows: Temperature -- 17 to 25C, Hardness -- assumed 40 to 210mg/l, and Calculations -- various methods, including graphic interpolation. Please see information in Table 2 for discussion of these factors. Three notable errors were identified. Cl) Reference 4 In the nickel study, one 96-hour EC^q value, 27 mg/1, was a nominal--not measured--concentration. Th^ ratios developed in the water quality document of continuous flow to static flow data were derived from 28/32 = 0.88 and 25/27 = 0.93. It is not clear whether both the static and flow-through tests were run simultaneously. We could conceivably calculate 75/32 = 0.78 and 28/27 = 1.04 based on the quoted reference and increase the range of the ratios. (2) Reference 5 The static 96-hour LC^q value of 9 mg/I came from a different study, and it was divided into an arithmetic mean value of 10.45 mg/1. If the data in Table 7 of reference 5 are interpreted correctly, the bluomll data were not included in the water quality criteria document. In this instance, the ratios obtainable for blueqills could range from 0.77 to 1.31, in contrast to the 1.16 reported for the fathead minnow. (3) Reference 6 The 96-hour LC^q value for endrin (0.3w) was qivon in the text, but it is not clear whether the value came from research presented in the paper (6) , other un reported research, or another paper. Reference 6 suggests that fathead minnow sensitivity to endrin varied by a threefold factor when com,pa r i no various studies. Furthermore, it is not clear whether theoreti cal concentrations of endrin were usi\l to estimate LCr n values. CMA, 049446 8- - In conclusion, most of the data are for pesticides and may introduce a bias in the correction factor for other compounds. able 4 Most of the data, especially the last sections, were difficult to comprehend due to the method of citation. This was especially true for reference 42, and where there were many citations and the numbers were rearranged. The data might be comparable for any given toxicant or cita tion. It would be hard to acknowledge, however, that all data in Table 4 taken as a whole are comparable. In light of the data we examined, we found: (1) Physical/Chemical Parameters (see Table 2 for a dis cussion of significance) Temperature -- 10s to 268C (4, 13) Hardness -- soft to 375 mq/1 (30) pH -- varies with hardness (2) Biological Parameters The data base consisted of at least 20 species of fish (12) and different sizes or ages of fish (4). (3) Testing Procedures (a) Calculations -- different methods (b) Conditions -- static, flow-through, and static- renewal (c) Concentrations -- some values were nominal not measured (39) (d) Solvent controls -- not reported (39) (4) Units of Reporting Not all values were corrected to mg/1, and some data were reported as ug/1 (21, 29). These values were not corrected to mg/1 before the geometric mean was calculated. .(5) Incipient Values Incipient LC^g values were used for comparative purposes when 96-hour values were available (23). LCriQ inci pient values are not the same as standard acuue tests because the time interval on an incipient value is not fixed. The 7-day incipient lethal value is 0.009 mg/1 endosulfan, and it is reported as 0.09. .'able 5 We had difficulty determining how some MATC values were calculated from the cited literature. Not all of the MATC values were in the same units (mg/1 and ug/1). For examples, see refer ence 15 and 18. CMA 049447 -9- The authors of reference; 15 concluded that the safe con centration of HCN for bluegills was 5.2 mg/1 . They arrived at this value because 5.2 mg/i inhibited spawning behavior, and if a "complete life history is to be accomplished in cyanide-polluted waters," the level must be lower. The water quality criteria document uses 17.4 as the MATC from reference 15. Table 6 The general comments in Tabic 2 apply to Table 6. Errors in data transcription are noted below. Reference Criteria Citation Correct Value 1 0.,48 2.,06 1 0 ., r. 7 0 ., 9 4 0. 31 * 0 ,, 00 3 0., 97 0 ,, 097 2 0..40 0 ., 04 0. 7 6 0 .,07 6 `t 0,,43 0 .. 5 7 5 n,, ni 0 ,,007 5 o,, 01 0 ,.004 5 0,. 6 5 0,. 0 G 5 5 n.. 2 2 0 ,.25 5 0., 4 3 0 ,, 1 3 f. 0.,55 n,, 08 5 0 o,, 03 o,,05 7 0 ., 0 6 n,, 6 The correction factors should be recalculated. Wo were unable to check the values in reference .`5. Table 7 Ke have no problems with the data presented. However, since all data used in this table are based on pesticides, the correction factor will be biased against all other compounds. Table 8 Those data wc could trace were accurate. The upper part of the table was inconsistent in units used (me/] and ;.g/l). We assume that these were not corrected prior to correct.ion factor calculation. .Errors in data transcription include reportina a 24-hour TLm (1). The values for zinc, nickel, and cadmium were for snail eggs, adult snails, or both (1). For thiodan (31), the value should be 75 and not 175. Data for l.vmhnia manna were re ported in reference 13, but they were not Trfcl uded-in-the document. The temperature range used in reference 13 was from 5 to 19C. This table includes fish data even though it is labelled for in vertebrates (reference 2). For cooper, F.PA cites data from references 8 and 9; we could not identify the copper value for reference 8. CMA. 049448 0- ! - ble 9 The MATC value for toxaphene could not be traced. If we assume that it was from reference 42F, then no value was cal culated or implied in the reference. Aaain MATC values were presented in mg/1 and ug/l, and we assume that calculations were made without correction to comparable units. ble 10 Only references 3 and 9 were available for study. Again, many of the comments discussed above apply, for example, lumping of data sets for species, habitat, different LC^g methodologies, and the like. For reference 3, 4.8 was the 96-hour LC5q for LAS obtained from a static test where the concentrations were cal culated, not measured. For the continuous-flow test, the 96-hour LC5A value was 4.35, and this value was measured. The document implied that the data from reference 9 are on an effluent, but they are for pure industrial chemicals. There were errors in reporting the data. Line 6 (p. 21518? should be 0.25/0.55 = 0.45; line 15 (p. 21518) should be 5000/1000 = 0.5; and line 16 (p. 21518) should be 3200/6000 = 0.53. The table includes quotients of the highest concentration effecting or killing 0 - 10% of the organisms divided by the 96-hour LC^g or FC^g value. Where 0 and 10% mortality data were available, the data were selected to include only the hiqhest concentration killing 0% of the population; data where there was a 10% kill (or random death) were ignored. For example, see page 21517, column 3, lines 73 (25/40=0.63), 76 (7/11* = 0.64 ) 77 (450/550 = 0.82), and 83 (180/290 = 0.62). Pertinent data where there was only a 10%, and not 0%, mortality reported (see tables of reference 9) were ignored. The 0.4^ mean value was calculated for data where there was no observed effect or an observed 10% effect. Whey could not a calculated LC^g value be used to obtain the same information? \ 1 >, cm 049449