Stephen R. Heller, Karen Scott, and Douglas Bigwood
USDA, ARS, NRI, Systems Research Laboratory
Beltsville, MD 20705-2350
Received January 25, 1989
Data evaluation of physical and chemical properties of
information being entered into a database should be an integral
part of the database development process. In the initial stages
of developing a pesticide properties database, the issue of data
quality quickly arose. An example of the range of reported
solubility values for a commonly used insecticide, fenthion, is
used to show the need for thorough data collection and the need
for careful and objective data evaluation.
The explosion of information over the past two to three decades
has greatly strained our ability to store, retrieve, and digest
the vast quantities of data available. Coupled with the fact that
society relies heavily upon scientific studies to protect the
environment, it is most important that the quality of the data
being used for decision making be the best possible. For example,
in examining possible groundwater contamination, a lawyer has
recently noted that the "demand for scientifically accurate, yet
legally appropriate, contaminant flow data modeling is
mushrooming as hundreds of hazardous waste sites are added to the
EPA Superfund National Priority List.... Modelers are being asked
to answer such complex questions as how long will it take for
chemicals to enter groundwater drinking supply, at what levels,
and from which source(s)" (1). With the increased emphasis on the
use of simulation and modeling, as well as environmental
assessments, it is critical that accurate data be used. Without
good data the results of such studies are likely to be
successfully challenged in the many adversarial proceedings that
are becoming a part of daily life. Failure to defend conclusions
from scientific studies is likely to lead to adverse opinions of
science by the public.
It appears on the surface that the need for accurate reference
data should be known to everyone, and the information presented
in this paper is not necessarily original or creative. However,
from the actual state of handbook and literature data, it is
clear that there is a very practical need to raise the level of
awareness of actual data quality and to try to stimulate interest
in the collection and dissemination of well-defined data that are
as accurate as possible. The continued reporting of poor data,
while original, and perhaps creative, is not considered a
worthwhile scientific goal.
The use of pesticides to improve and sustain crops is basic to
agricultural production and has increased dramatically over the
past 40 years. At present, nearly a million tons of these
chemicals are being applied annually by farmers in the U.S. This
widespread use increases the potential for some of these
materials to move through the soil into the groundwater (2) and
to have other adverse environmental effects.
To provide the essential data on the chemical and physical
properties of pesticides, the USDA Agricultural Research Service
(ARS) has assumed the responsibility of creating, evaluating, and
maintaining a pesticide properties database (PPD) (3). While
there are several sources of information on the properties of
pesticides, (4-11) including data collected over the years in an
informal manner by ARS scientific staff (12), no one has
collected all the necessary data in one place at one time and
defined their quality. A database of known accuracy and precision
of physical and chemical properties is needed to support various
environmental modeling and assessment activities. As Kollig has
pointed out, "Data being used in the environmental and human risk
assessment for regulatory purposes must be of known reliability
for the assessment to have validity" (13). While many current
models are crude and do not now require the most accurate data,
at some point poor data will become the limiting step in
developing better models.
As part of the ARS PPD development, data have been obtained and
evaluated from the literature. In addition, the National
Agricultural Chemicals Association (NACA), as part of their
Ground Water Protection Program, has requested their members
provide ARS with the data they have submitted to the states of
Arizona and California and any other data they may wish to make
available (14).
The bibliographic literature, as searched by using Chemical
Abstracts Service Online, provided very little information.
Rarely did we find that the bibliographic abstracts contained
information that pointed to needed data. The few sources of
actual data, found in such references as the Agrochemical
Handbook (5), The Pesticide Manual (7), and The Merck Index
(11), both in hardcopy and computer-readable form, have no literature citations. Thus, when a question arises about the
validity of a value, there is no way to resolve the matter.
For example, the solubility of the widely used pesticide Alachlor is reported as 242 mg/L at 25 C in The Agrochemical Handbook (14). This compares with the values of 140 mg/L at 23 C in The Merck Index (16) and 148 mg/L at 25 C in the Yalkowsky Arizona dATAbASE Solubility Database.~7 It would be logical to assume that a typographical error has occurred, but should the value 242 be 142, or should 140 be 240? Is 148 mg/L really another typographical error and the value should be 140 mg/L? When one of the editors of The Agrochemical Handbook was contacted about this matter, the response was that their published value (242 mg/L) was "also in the Herbicide Handbook of the Weed Society of America (4) and The Pesticide Manual published by BCPC" (18). But where did these publications obtain their values?
As a result of this and many other inconsistencies encountered
(see Table I
for
additional examples), we decided that
every value must be subjected to rigorous and objective scrutiny. To do
this we require two things: (1) a data evaluation scheme that is
as consistent and as objective as possible and (2) a thorough
report of the experimental procedure and conditions. Literature
citations of experimental procedures needed for this evaluation
are difficult to obtain. An evaluation scheme is now being
developed (19) that will result in a PC-based expert system to
quickly evaluate the quality of the data on the basis of criteria
developed by experts in the field. The first property being
examined is the aqueous solubility of pesticides.
While it would be desirable to have an estimation technique
available for checking the reported solubility or for checking
the consistency of the solubility with a series of similar
compounds, such methods have not yet been developed for the
classes of compounds commonly encountered in pesticide products.
The few reported estimation techniques are too crude and cover
too limited a scope of classes of chemicals to be of use for
multifunctional group compounds as are found in the structures of
most pesticides. In a handbook on property estimation, Lyman
states, "the available estimation techniques usually yield values
that are, on the average, uncertain by less than one order of
magnitude, but errors of over two orders of magnitude occur in
about 10% of the cases with some equations-.20 Furthermore, Lyman
asserts that "few (estimation techniques) can handle complex
structures or uncommon functional groups-. In addition, there are
generally too few data for one to adequately test the validity of
an estimation techniques (21).
As a result of the cooperation of the NACA, data on fenthion were
received from Mobay Corp. Fenthion is the common name for O,O-dimethyl 0-[3-methyl-4-(methylthio)phenyl] phosphorothioate.
Fenthion was introduced in 1957 and is used as an insecticide,
primarily for ornamental flowers. The data for fenthion from
Mobay are the same as those provided to the states of Arizona and
California under the laws of those states relating to pesticides.
The Mobay data and other data found from literature searching are
given in
Table II.
Most of the citations appear to refer to the original work of
Schrader (23) reported in 1960, although there are some minor
variations in the value [e.g., a range (54-56) vs a single value
(55)1 and the temperature may differ slightly or is not reported
at all. We have also checked the US., Canadian, and German
patents (24) issued to Schegk and Schrader and found no solubility value in these publications. Rather, the data in these
patents referred primarily to toxicity and other biological data,
with values for boiling point, melting point, density, and refractive index.
As we continued our investigation it appeared that transcription
errors are probably responsible for most of the differences
reported for the values shown in
Table I. The exception is that
The Pesticide Manuals (28) reports the solubility of fenthion as
54-56 mg/L in the 5th and 6th editions, but then changes it to 2
mg/kg (=2 mg/L) in the 7th and 8th editions (33) without any
explanation. A change of a factor of nearly 30 does seem to us to
merit some explanation.
Experimental conditions were reported for the data in
Table I
from only two sources, Mobay (36) and Bowman and Sans (32,34).
On the basis of our evaluation procedures we have come to the
conclusion that the solubility data from the internal Mobay
Chemical Corp. report (4.2 mg/L) is likely to be the most
accurate value available. The quality control expert system
evaluation procedure elicits answers to a number of questions
about the data reported, in a manner similar to that used in the
SELEX expert system described previously (22). The main questions
asked are the following: (1) Is there a citation with an
analytical procedure described? (2) What type of method was used
(chromatographic, spectroscopic, or nonspecific)? (3) What is the
purity of chemicals used? (4) What is the purity of water used?
(5) How was the temperature of the experiment controlled? (6) How
many repetitions were performed, and what was the standard error?
The overall scheme of this expert system, called SOL, is shown in
Table III.
Briefly, after putting the experimental data for the solubility
of fenthion through the SOL expert system, we conclude that the
Mobay data were superior on the basis of the following
distinctions: (1) Of the two studies where experimental procedures were reported, a dynamic chromatographic technique was
used for the measurement, and only the Mobay study (36) used more
than one gas chromatographic flow rate (10 and 20 mL/h). (2) The
data from Mobay stated the purity of fenthion was 99.7%, whereas
the data from Bowman and Sans (32,34) indicated the purity of
fenthion was only 97.1%.
SUMMARY
The computer-based models for estimating potential groundwater
contamination (39-42) and environmental assessments require the
best available data for input. These data must be accurate if the
user community (scientists, government regulators, industrial
groups, and environmental groups) is to have confidence in the
predictions of these models or the results of assessments, so that
decisions based on these results will be acceptable. Data in
current literature citations, from journal publications to widely
used and cited handbooks, have not undergone sufficient evaluation
to assure their accuracy. We have shown a number of examples of the
proliferation of errors and the lack of evaluation that has spanned
a period of more than 20 years and have analyzed one example,
fenthion, in detail. We hope that the development of the ARS PPD
will be a first step in improving this situation. The ARS PPD is
expected to be used in two main ways, besides being a reference for
pesticide chemical and physical properties. The first is for the
current modeling efforts in looking into what happens to a
pesticide when it enters the soil and the possible contamination of
the groundwater that makes its way into drinking water supplies.
The second is to help in regulatory assessments of pesticides.
Predictive models are needed because time and cost make it
impossible to run actual experiments in the field to determine the
impact of chemicals on groundwater quality. Accurate and reliable
predictions will be a considerable benefit over the coming years,
but these are only possible with accurate physical and chemical
property data.
We thank R. Nash, C. Helling, D. Glotfelty, S. Rawlins, B. Acock,
and D. Wauchope of ARS for their thoughts and insight into
pesticide data and data evaluation. We also thank Patricia
Laster for her work on the design and implementation of the
computer system for the ARS PPD. We also thank T. Gilding, NACA,
for his efforts in helping to obtain data from indsutrial
agrochemical companies. Lastly, we thank Mobay Chemical Coprp.
For their report on the solubility of fenthion and T. D. Talbott
of Mobay for his assistance and helpful comments on the
manuscript.
REFERENCES AND NOTES
(1) Walsh, W. J., Ground Water Model Newsl. 1988, 7(2) 1-6.
(2) ARS Strategic Groundwater Plan. 1. Pesticides; Agriculture
Research Service, U.S. Department of Agriculture. U.S. Government
Printing Office: Washington, DC, 1988.
(3) Heller, S. R.; Bigwood, D. W.; Laster, P.; Scott, K.
The ARS
Pesticide Properties Database. Proceedings of the IIth
International CODA TO Conference; Hemisphere Publishing: New York
(in press). Heller, S. R.; Bigwood, D. W.; Laster, P.
Using dBase
for the Scientific Databases: The ARS Properties Database.
Proceedings of the National Online Meeting, May 1989; Learned
Information: Mariton, NJ; pp 179-185. For additional details,
please contact S. R. Heiler.
(4) Herbicide Handbook, 5th ea.; Weed Science Society of America
(WASA): 309 W. Clark St., Champaign, IL 61820 [(217) 356-3182],
1983.
(5) The Agrochemicals Handbook, 2nd ea., Royal Society of
Chemistry: Nottingham NG7 2RD, England (011 -44-602-507411),
1987.
(6) Farm Chemicals Handbook; Moister Publishing: 37841 Euclid
Ave., Willoughby, OH 44094 [(216) 942-20001, 1988.
(7) The Pesticide Manual, A World Compendium, 7th cd.; Worthing,
C. R., Walker S. B., Eds.; Thc British Crop Protection Council:
144-150 London Road, Croydon CR0 2TD, England, 1987.
(8) Physicochemical Properties of Pesticides; Shamshurin, A. A.,
Krimer, M. Z., Eds.; Khimiya Publishing: Moscow, 1976.
(9) Nanogen Index--A Dictionary of Pesticides and Chemical
Pollutants; Packer, K., Ed.; Nanogens International: P.O. Box
487, Freedom, CA 95019, 1975.
(10) Pesticide Index, 5th ea.; Wiswesser, W. J., Ed.; The
Entomological Society of America: 9301 Annapolis Road, Lanham, MD
20706 [(301) 731-45351, 1976.
(11) The Merck Index, 10th ed.; Delko, G., Ed.; Merck: Rahway, NJ 07065-0900, 1983.
(12) Nash, R. G.; Heller, S. R.; Lancaster, E. C. The Agricultural Research Service Pesticide
Properties Database. Abstracts of Papers, 194th National Meeting of the American Chemical
Society, New Orleans LA; American Chemical Society Washington, DC, 1987; AGRO 148.
(13) Kollig, H. P. Criteria for Evaluating the Reliability of Literature Data on Environmental
Process Constants. Toxicol. Environ. Chem. 1988, /7, 287-311.
(14) National Agricultural Chemicals Association, 1155 15th St. NW Washington, DC 20005
[(202) 296-1585].
(15) See ref 4, A004, Aug 1987.
(16) See ref 11, entry 189, p 193.
(17) The Arizona dATAbASE of Aqueous Solubility, College of Pharmacy, University of
Arizona, Tucson, AZ 85721. There were approximately 2600 records in the database as of the
end of 1988. Also see: Dannenfelser, R.-M.; Yalkowsky, S. The Arizona Database: An Aqueous
Solubility Database for Nonelectrolytes. Proceedings of the Beilstein Workshop on The
Estimation of Physical Data for Organic Compounds, Jochum, C., Hicks, M. G., Sunkel, J.,
Eds.; Springer: New York, 1988; pp 490508.
(18) Kidd, H., private communication, Jan 12, 1988.
(19) Heller;S. R.; Bigwood, D. W.; Schantz, M.; Gunther, F.; May, W.
Expert Systems for
Evaluating Physicochemical Property Values. 1. Aqueous Solubility.
Environ. SO Technol. (submitted for publication).
(20) Lyman, W., Reehl, W. F., Rosenblatt, D. H., Eds.; Handbook of Chemical Property
Estimation Methods--Environmental Behavior of Organic Compounds; McGraw-Hill, New
York, 1982; pp 2-2-2~. The book is currently out of print. The equations in the book have been
programmed into a computer system. The result system, CHEMIST, is available as an IBM PC
floppy disk system as well as an online system. For details, contact Technical Database Services,
Inc., Suite 2300, 10 Columbus Circle. New NY 10019-1202 [(212) 245-00441.
(21) Yalkowsky, S., private communication. In the example discussed there were data on the
solubility of 45 triazines. This number was found to be too small to properly test a critical
coefficient parameter.
(22) Bigwood, D. W.; Heller, S. R.; Wolf, W. R.; Schubert, A.; Holden, I. M.
SELEX: An Expert
System for Evaluating Published Data on Selenium in Food. Anal.
Chim. Acta 1987, 200, 411-419.
(23) Schrader, G. Development of the Bayer Product S 1752. HocJchenBriefc 1960, 13, 1-12.
(24) Schegk, E.; Schrader, G. U.S. Patent 3 042 703 (issued July 3, 1962); Can. Patent 631 472
(issued November 21, 1961); Ger. Patent I 116 656 ( issued November 9, 1961).
(25) Gunther, F. A. Residue Rev. 1968, 20, 67.
(26) Spencer, E. Y. Guide to the Chemicals Used in Crop Protection, Publication 1093, 6th ea.;
Information Canada: Ottawa, 1973; p 279, Catalog No. A43-1093.
(27) Verschueren, K. Handbook of Environmental Data on Organic Chemicals, 2nd cd.; Van
Nostrand Reinhold: New York, 1983; p 231.
(28) Worthing, C. R., Ed. The Pesticide Manual, A World Compendium, 6th ea.; British Crop
Protection Council: Croydon CR0 2TB, U.K., 1979; p 265.
(29) Khan, S. U. Pesticides in the Soil Environment; Fundamental Aspects of Pollution Control
and Environmental Science Series, Vol. 5; Elsevier, Amsterdam, 1980, p 220,
Table A-2.
(30) Eto, M. Organophosphorus Pesticides: Organic and Biological Chemistry; CRC Press:
Cleveland, OH 1974; p 245.
(31) See ref 11, entry 3927, p 3928.
(32) Bowman, B. T., Sans, W. W. Further Water Solubility Determinations of Insecticidal
Compounds. J. Environ. Sci. Health 1983, B18(2), 221 -227.
(33) See ref 7, p 269.
(34) Bowman, B. T.; Sans, W. W. Effect of Temperature on the Water Solubility of Insecticides.
J. Environ. Sci. Health 1985, B20(2), 625-631.
(35) See ref 4, A203, Aug 1987.
(36) Water Solubility of Fenthion Pure Active Ingredient; Mobay Corp. Agricultural Chemicals
Division, Report 94648, 1987.
(37) Suntio, L. R.; Shiu, W. Y.; Mackay, D.; Sciber, I. N.; Glotfelty, D. A Critical Review of
Hcnry's Law Constants for Pesticides. Rev. Evuiron. Contam. Toxicol. 1988, 103, 1-59.
(38) Ref 5, p C100.
(39) CREAMS, a Field Scale Model for Chemicals, Runoff, Erosion, and from Agricultural
Management Systems; Kniescl, W. G., Ed.; USDA Conservation Research Report 26. U.S.
Government Printing Office: Washington, DC, 1980; 640 pp.
(40) PRZM--User's Manual for the Pesticide Root Zone Model; Carscl, R. F., et al., Eds.; U.S.
Government Printing Office: Washington, DC, 1984; Release 1, U.S. EPA, R&D Report EPA-60~3/84-109.
(41) DeCoursey, D. G., ct al. A Model for Managing Agricultural Impacts on the Unsaturated
Zone, ACS Symposium Series; American Chemical Society: Washington, DC (in press).
(42) DeCoursey, D. G. Root Zone Water Quality Model--RZWQM; USDA, ARS, Hydro-Ecosystems Research, Ft. Collins, CO. U.S. Government Printing Office: Washington, DC,
1988.