Environ. Sci. Technol. W9Q, 24, 926-927

ordinary kriging known as indicator kriging (7) could be used to better handle problems with loss of precipitation from the samples.

The deposition estimates made by Voldner and Alvo based on direct interpolation of deposition values provide useful information. Their deposition estimates based on direct interpolation of concentrations are useful only as an example of the bias possible when unequal support sizes are ignored.

Registry No. Sulfur, 7704-34-9; nitrogen, 7727-37-9.

Literature Cited (1) Voldner, E. C.; Alvo, M. On the estimation of sulfur and

nitrogen wet deposition to the Great Lakes. Enuiron. Sci. Technol. 1989,23, 1223-1232.

(2) Bilonick, R. A. Risk-qualified maps of hydrogen ion con- centration for the New York State area for 1966-1978. Atmos. Environ. 1983, 17, 2513-2524.

(3) Bilonick, R. A. The space-time distribution of sulfate de- position in the Northeastern United States. Atmos. En- viron. 1985, 19, 1829-1845.

(4) Journel, A. G.; Huijbregts, Ch. J. Mining Geostatistics; Academic Press: New York, 1978.

(5) Guertin, K.; Villeneuve, J.; Deschenes, S. The choice of working variables in the geostatistical estimation of the spatial distribution of ion concentration from acid precip- itation. Atmos. Enuiron. 1988, 22, 2787-2801.

(6) Granat, L. Sulfate in precipitation as observed by the European Atmospheric Chemistry Network. Atmos. En- viron. 1978, 12, 413-424.

(7) Bilonick, R. A. Monthly hydrogen ion deposition maps for the northeastern U.S. from July 1982 to September 1984. Atmos. Environ. 1988, 22, 1909-1924.

R. A. Bllonick

Consolidation Coal Company Consol Plaza Pittsburgh, Pennsylvania 1524 1

SIR: Two possible approaches for estimating sulfur and nitrogen deposition to the Great Lakes and basins are discussed. The first proposes to estimate deposition by interpolating depositions at the sites whereas the second proposes to interpolate concentration and precipitation at the sites separately and then multiply the interpolated values on a grid by grid basis. In situations where there are a large and well-spaced number of sites with adequate deposition data, the first approach is entirely suitable and prevents the difficulties cited above. Bilonick fails to recognize that the measured deposition at a site may not represent the actual deposition. Frequently, the site precipitation is underestimated due to undercatch of the collector, missing events, or evaporative losses. Measured deposition in the first two cases provides an underestimate of the annual deposition, even though the precipitation weighted mean concentration may approach the actual annual value. A simple scaling with the annual precipi- tation amount as recorded by a colocated rain gauge, and done for the major networks, reduces the error [Olsen et al. (I)]. Since the GLAD and GLP networks were not equipped with rain gauges, such a correction was not feasible. A simple interpolation of site deposition would propagate the errors spatially. If the low precipitation value, however, is caused by evaporative losses and no chemical degradation occurs in the sample, the measured concentration is too high, while the measured deposition is essentially correct. In this case, interpolation of site deposition is valid. Adjustment of site precipitation on the other hand, would lead to overestimation.

The precipitation values for the GLAD and GLP net-

works, even for sites with 100% data capture, appear low compared to recordings by neighboring rain gauges. The reason for this discrepancy is not clear. Hence, the two approaches for estimating deposition were used. The method of estimating concentration and precipitation separately was further motivated by the observation that spatial variability in concentration for sulfur and nitrogen is generally less than spatial variability in precipitation. To illustrate the problem, suppose that site precipitation in the example provided is underestimated by 30% due to undercatch or missing events, but that concentration is correct.

location I 1 X 2 x3

precipitation, L 3.5 ? 0.7 deposition, mg 3.5 ? 2.1

The deposition at x 2 is estimated to be 2.8 mg by Bilonick, a 30% underestimate compared to the real value of 4.0 mg. On the other hand, separate interpolation of the concen- tration and precipitation yields the estimates 2.0 mg/L and 2.1 L, respectively. If deposition is obtained by multiplying the interpolated values of concentration and precipitation, the estimate is 4.2 mg, which is close to the real value of 4.0 mg.

The question of how best to take advantage of the data available from the denser precipitation networks also re- mains open. The suggestion made by Bilonick to develop a regression equation relating deposition to precipitation is fraught with danger since it does not take into account any spatial considerations. It would indeed be welcome if the problem were that simple since it would render unnecessary the development of any physically based comprehensive deposition models. The suggestion to use cokriging “would require the estimation and modeling of semivariograms for deposition and precipitation as well as the covariogram between deposition and Precipitation”, as indicated by Bilonick. It would be difficult to estimate the latter accurately with the existing site data.

concentration, mg/L 1.0 ? 3.0

Registry No. Sulfur, 7704-34-9; nitrogen, 7727-37-9.

Literature Cited (1) Olsen, A. R.; Bigelow, D. S.; Chan, W. H.; Clark, T. L.; Lusis,

M. A.; Misra, P. K.; Vet, R. J.; Voldner, E. C. Unified Wet Deposition Data Summaries for North America: Data Summary Procedures and Results for 1984. Atmos. En- viron., in press.

E. Voldner Atmospheric Environment Service Environment Canada Downsview, Ontario, Canada K1N 6N5

M. Alvo” Department of Mathematics University of Ottawa Ottawa, Ontario, Canada K I N 6N5

Comment on “Characterization of Environmental Tobacco Smoke”

SIR: In reporting their analyses of components of en- vironmental tobacco smoke (ETS), Lofroth et al. ( I ) characterized some of the substances as “genotoxic” and “hazardous”. The implications of human toxicity to ETS components were not supported by the authors and are of questionable appropriateness. Neither ETS nor com-

926 Environ. Sci. Technol., Vol. 24, No. 6, 1990 0013-936X/90/0924-0926$02.50/0 @ 1990 American Chemical Society