Acid Deposition: Long-Term Trends (1986)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

7 Streams and Lakes fames R. Kramer, Anders W. Andren, Richard A. Smith Arthur H. Johnson, Richard B. Alexander, and Gary OehZert INTRODUCTION Various attempts have been made to assess changes in the acidity of surface waters that may have occurred over the past 50 years. An excellent review of the methods used in the analyses as well as the basic conclusions from a large number of investigations may be found in the EPA's Critical Assessment Review Papers (Environmental Protection Agency 1984). For example, Henriksen (1979, 1980, 1982), Almer et al. (1978), Dickson (1980), Christophersen and Wright (1981), Thompson (1982), Galloway et al. (1983a), and Wright (1983) used basic geochemical concepts to derive general models that are designed to predict both the sensitivity and the degree of response of a water body to acid inputs. More complex mechanistic and time-dependent models that are intimately synchronized to hydrological, biological, and geomorpho- logical factors have also been used (for example, Chen et al. 1982, Schnoor et al. 1982, Booty and Kramer 1984). These models have provided much useful information, but currently none can provide detailed, quantitative assess- ments of past or future trends in lake and stream acidification from acid deposition. (For an evaluation of the strengths and limitations of these models see Church (1984).) Another approach has been to compare the acidic status of lakes, rivers, and ponds in North America using his- torical and current data on pH, alkalinity, and sulfate. Studies in the United States include, for example, those of Schofield (1976) for New York lakes, Davis et al. (1978) for lakes in Maine, Johnson (1979) for streams in New Jersey, Pfeiffer and Festa (1980) for lakes in New York, Hendrey et al. (1980) and Burns et al. (1981) for 231

232 headwater streams in North Carolina and New Hampshire, Arnold et al. (1980) for Pennsylvania streams, Crisman et al. (1980) for lakes in northern Florida, Norton et al. (1981) for lakes in New Hampshire, Maine, and Vermont, Lewis (1982) for lakes in Colorado, Haines et al. (1983) for lakes in New England, and Eilers et al. (1985) for lakes in northern Wisconsin. Similar studies have also been published on lakes in Canada (Beams and Harvey 1972, Beamish et al. 1975, Dillon et al. 1978, Watt et al. 1979, Thompson et al. 1980) and Scandinavia (Wright and Gjessing 1976, Rebadrosf 1980). These studies analyzed water quality data using a variety of methods. Some sets include continuous data records and others only data at discrete points in time. Many of the analyses account for such complicating factors as changes in analytical methods and land use practices. The conclusions from a variety of studies are that at least some lakes exhibit a decrease in pH and alkalinity when historical and recent data are compared. Similarly, a review by Turk (1983) compares historical and geographical trends in the acidity of precipitation with trends in North American surface waters. Based on the synthesis of a variety of data collected during discrete time periods (mainly the 1930s and late 1970s) and examined by Schofield (1976), Pfeiffer and Festa (1980), Davis et al. (1978), Norton et al. (1981), Burns et al. (1981), and Haines et al. (1983), he concludes that decreases in pH and alkalinity have occurred in lakes in New York and New England. In addition, water quality data are available from the U.S. Geological Survey's Hydrologic Bench-Mark Network. These data, collected since the middle 1960s, provide a continuous record of high-quality data for pa, alkalinity, and sulfate as well as the major_cations and anions for small watersheds in the United States. Smith and Alexander (1983) have studied these records and conclude that "In the northeastern quarter of the country, SO2 emissions have decreased over the past 15 years and the trends in the cited chemical characteristics (alkalinity and sulfate) of Bench-Mark streams are consistent with a hypothesis of decreased acid deposition in that region. Throughout much of the remainder of the country, SO2 emissions have increased and trends in stream sulfate, alkalinity and alkalinity/total cation ratios are con- sistent with a hypothesis of increased acid deposition. There are, however, several acknowledged problems in the interpretation of "historical n lake water chemistry

233 data. Perhaps the chief drawback is the lack of docu- mentation of sampling and analytical procedures. A major problem encountered in our study has been the lack of sufficient documentation, before 1950, of the procedures for calculating pH and alkalinity. For example, whereas the current method for measuring pH and alkalinity employs standard electronic instruments, before about 1950 the method of choice was calorimetry using indicator dyes, which, when added in small quantities to solutions of unknown pH, cause color changes at specific pHs. Methyl orange, a commonly used indicator for measuring the alkalinities of lake waters in the 1930s, undergoes subtle color changes over a range of pH values, and the exact endpoint used by analysts is often not clearly specified in the historical records. In this chapter, we review some of these records and assess trends based on the most likely endpoints employed in the historical studies. Furthermore, for reasons that are described in detail in Appendix D, titration of lake water solutions to a methyl orange endpoint requires a correction for "over- titration." Assumptions about the magnitude of the correction are especially important when analyzing low- alkalinity waters. In most previous studies, researchers have assumed that the correction factor is a constant independent of the original alkalinity. Values ranging from zero to greater than 100 microequivalents per liter (peq/L) have been employed in such studies. A uniform method for analyzing these types of samples is highly desirable, and one goal of this chapter is to present a method for determining the historical corrections by calculations rather than by assumption. The method accounts for the fact that the relation between the correction factor and alkalinity is sensitive to the alkalinity of the test solution. In addition, the method provides a means for checking the internal consistency of the data set. Additional factors that must be considered in inter- preting current and historical data include (1) clima- tology, since the hydrology of lakes and rivers may vary considerably from year to year in response to fluctuations in precipitation and climatic factors affecting evapora- tion, (2) regional patterns of emissions of acid pre- cursors, since natural waters downwind of sources may be expected to exhibit greater changes in water quality than those situated upwind, (3) changes in land use (although Drably et al. (1980) found that land use changes in Norway did not seem to be related to regional acidifica-

234 tion patterns, local changes in alkalinity and pH regimes were evident), (4) the statistical validity of using data separated by 20 to 40 years to examine trends, and (5) representativeness of data sets for various regions. Inferences about trends obtained from data on a limited number of lakes within a region should reflect the distribution of values of alkalinity and pH (and other physical, chemical, and biological parameters) for the whole region. In the current study, we have tried to overcome some of the difficulties mentioned above by examining temporal and spatial variations in several water quality parameters--pH, alkalinity, sulfate, and selected major anions and cations--that simultaneously give some information about the acidic status of lakes. The analysis is restricted to a few sets of data of high quality selected to provide both a determination of regional trends in lake and stream acidification and an illustration of a consistent methodologic approach. In this chapter, we first examine the hypothesis that sulfate levels in lakes and wet precipitation in the northeastern United States and eastern Canada are correlated. Many concepts of lake acidification focus on the titration of a surface water body by sulfuric acid deposited from the atmosphere (e.g., Henriksen 1980). The result of this simple titration is to substitute sulfate ion for bicarbonate ion, thus lowering the alkalinity and the pH of the affected surface water. Wright (1983, 1984) has shown a linear correlation between averages of sulfate concentrations for selected lakes and average sulfate concentration in precipitation in eastern North America, and Thompson and Hutton (1985) show a similar relationship in eastern Canada. To link sulfate ion concentration in atmospheric deposition to the sulfate ion concentration of surface waters, a direct relationship must exist between atmo- spheric and lake sulfate fluxes for a steady state assumption. A fundamental condition for this relation- ship to be valid is that there be no overriding internal sources or sinks of sulfate ions in the watersheds. If such sources and sinks do occur, then the method of averaging sulfate concentrations of large sets of lakes may mask the large variability in the data, particularly if averaging cancels factors equal in magnitude but opposite in sign. We examine data from a set of 626 lakes and assign to each a value of precipitation sulfate flux. We draw general conclusions about spatial

235 associations between precipitation and lake sulfate fluxes from a statistical analysis of both the aggregated data and the data subdivided by regions. Next we address sulfate-driven changes in stream-water chemistry, using data from the U.S. Geological Survey Bench-Mark Network. The network provides a continuous record of monthly data on major ions in stream water beginning in the middle 1960s. Data are from pre- dominately undeveloped watersheds in 37 states. Con- tinuous sampling and unchanged analytical methods for the period make the data particularly well suited for examining atmospheric influences on water quality as a function of spatial and temporal SO2 emission patterns. The study of sulfate-driven changes in surface water chemistry is accomplished with the aid of a general paradigm that links acidic deposition to changes in the water quality in watersheds with acidic soil. Finally, we compare pH and alkalinity data of lakes taken at discrete time intervals, i.e., from the 1920s and 1930s to the 1970s and l980s. Three lake surveys were chosen for analysis; two in the northeastern United States (New York and New Hampshire) and one in the upper Midwest (Wisconsin). These lakes were selected because they provide, in our judgment, some of the most complete historical and recent data available, allowing us to test them for internal consistency according to the protocol of Kramer and Tessier (1982). We present this analysis to illustrate a chemically rigorous interpretation of historical pH and alkalinity measurements and to estimate the magnitude of changes in the acidic status of these lakes that may have occurred over the past 50 to 60 years. In our approach we limited our data to those lakes for which historical data on pH, alkalinity, and free CO2 acidity were available. The availability of simultaneous data for the three parameters allows us to perform a check for internal consistency of the data as an indicator of their reliability. Without this constraint a larger historical data base is available over wider geographic areas. Although our method allows us to screen the data for internal consistency, a larger (unscreened) data base may provide more expansive regional information and be amenable to more powerful statistical tests, although the data may be less reliable and hence the results of the tests less meaningful. Differences that may exist between the results obtained in this chapter and elsewhere must be viewed in light of the assumptions made and how the assumptions affect the selection and analysis of the data.

236 SULFATE FLUXES IN PRECIPITATION AND LAKES In this section, we test the hypothesis that the fluxes of sulfate into lakes from atmospheric deposition generally determine the fluxes of sulfate outflow from lakes over a broad geographical area. Such a relationship must be demonstrated to exist before we can assume with any confidence that water chemistry and related data on fisheries records and lake sediment stratigraphy (de- scribed in the following chapters) are accurate indicators of atmospheric acid deposition. Studies of the mass balance of sulfate in a few water- sheds have been reported for which parameters such as atmospheric deposition, streamflow, and geology are sufficiently well determined to permit more detailed analyses (e.g., the Hubbard Brook watershed, Likens et al. 1985; and three lakes in the Adirondacks, Galloway et al. 1983b). The studies conclude that sulfate concentra- tions in these waters are controlled by atmospheric deposition of sulfur. In a different type of analysis, Wright (1983) found an approximately linear relationship between the mean concentration of sulfate in 15 groups of lakes in North American and mean excess (i.e., nonmarine) sulfate in wet precipitation. In our analysis, we compare sulfate input fluxes in wet deposition with the estimated sulfate output fluxes from 626 lakes in New York, New England, Quebec, Labrador, and Newfoundland. Fluxes of sulfate from wet deposition were obtained by multiplying the concentration of sulfate in rain by the intensity of rainfall (meters/year). Lake output fluxes of sulfate were estimated as the product of sulfate concentration in lake water and the intensity of net precipitation (i.e., meters/year measured as rainfall minus evaporation). (See Appendix B for specific methods and references.) It was possible to perform a detailed statistical analysis to test the relation between sulfate inputs and outputs for linearity (or some other monotonic relationship) and regional variance on subdividing the area into smaller regions. Such an analysis suited the particular needs of this study, which addresses the question of broad regional patterns in eastern North America. However, we note that this analysis is subject to uncertainty because for each of these lakes we generally lack important information on dry deposition, transpiration, observed streamflow from the watershed, soil chemistry, and processes within the watershed that are sources of sulfur.

237 Results Figure 7.1(a) is a plot of sulfate inputs to 626 faxes from wet-only deposition as a function of annual sulfate outputs from the lakes based on data for 1980-1982. It is apparent that there is a greater variability in lake sulfate outflow fluxes than in precipitation sulfate fluxes. The scatter is greatest for lakes in Massachusetts (Figure 7.1(b)) owing to some extreme outliers and is of similar, smaller magnitude for other regions. In some cases, the output flux of sulfate from [ekes is greater than the flux from wet precipitation, suggesting that significant sources of sulfur other than from wet deposition exist in these watersheds. In other cases, the reverse is apparently true, perhaps suggesting that atmospheric sulfur may be retained in the watershed, or that sulfate is chemically reduced in the lake and deposited in the sediment. The apparent lake sulfate output flux normalized to the wet-only precipitation sulfate input flux is analyzed by geographic region in Table 7.1. Deficiencies in lake sulfate outputs compared with inputs from wet-only deposition Generally occur in remote regions such as Newfoundland and Labrador, whereas excesses occur in more populated areas such as New York, Connecticut, and Massachusetts. The origins of the deficiencies and excesses are not known on a lake-specific basis, but dry deposition, watershed processes involving sulfur cycling, and other sulfur sources of unknown origin are expected to be important contributors. Analysis In an effort to stabilize the variability in the sulfate output fluxes of lakes, the data were analyzed on the logarithmic (base e) scale (Figure 7.2). As a whole the data do not follow the y = x line (i.e., the case in which sulfate outputs equal sulfate inputs). Linear regression yields an estimate of a line log y = a log x + mi' in which the intercept, ml', is -0.306 (standard error, 0.041) and the slope, a, is 1.593 (standard error, 0.046). The slope of this line is significantly different from one, so the overall relationship on the natural scale, y = mixa in which mi is +0.736 (antilogarithm (base e) of ml') and a is 1.593, is significantly different from linear.

238 1 5 14 13 12 11 ., 10 Q 9 ° 8 J 7 IL u~ 6 ~ 5— u7 4- y 3 2 1 o o o o ° 0 °~o o~ ~; O 0 ~ ~o 0.5 (a) o o o o o8° - ° :~10 0 fi~°C~ O 0 0 O 0 0 ° o o e} c ~ 0 3~ ~ _ '~ ~p ~ 0 0 ~uO - : o 1.5 2.5 WET PRECIPITATION SULFATE FLUX (input) 9 m~2 yr~1 36 - 34 - 32 - 30- E 28- ~ 26— Q 241 O 22— x 20- 18— 16- 1 4— ~ 12- y 10- 8— 6- 4 4.5 o 0 0 0 oB o o o ° ° ° C~ O o 2- 1 1 1 1 1 r 2.1 2.3 2.5 2.7 (b) o c~ o o 0 0 2.9 3.1 WET PRECIPITATION SULFATE FLUX (input) 9 m~2 yr 1 FIGURE 7.1 (a) Comparison of sulfate input fluxes to lakes from wet-only deposition with sulfate output fluxes from lakes in the northeastern United States and eastern Canada. (b) A similar plot for lakes in Massachusetts only.

239 o U: ._ o a~ ~ Ct ~Q c~.. .C) Ct o q:: ~ O au U) ~ _ ~ ^ V) .= 3 ~ ' ·_ Ce ~ G =,, .~ O Ct ~ p~,, o z o _ _ .= · C~ .c · ~ Ce C~ (D 8< ~ o E~ Z ~ (,, ~ .= m ~ ~ zo O C) ~L) _ ~ . _ ·_ e~ := C) X ~ C) C~ ~ mX 3 o A Ch X o Vl oo X V o o Vl ~ X V U~ - U~ - Vl ~o X V o - o - Vl ~ X V o V~ o Vl X V o o Vl ~ X V O. O. Vl X Vl ol ol v - x ~0 oD 0.4 'e '~ c~ ~ ko ~n ~ ~ ~ cr 0 oo ~ - - ~ ~ - - - ~ o o ~ - - - ~ o o oX ~ ~ ~ - x - - t - ) ~ ~ ~ ~ — ~D ~ ~ ~ ~t {~) — c~ - ) - - - ~ u~ ~ ~ ~) - ~ x ~ ~ ~ ~ ~ oo - - ~ ~ - - ~ - oC o o ~ - o ~ ~ ~ x ~ o ~ ~ x - ~ 0 ~ t~ } ) ~ ~ ~ ~D ~ O ~ ~ - - os ~ ~ ox ~ - ~ ~ ~ ~ ~ ~ t - ~ ') ') - O' m~ = - - O ~ r 0 ~ _ _ _ cM r~~ ~ 0 Cx ~o 0 ~ U~ ~ _ _ ~ ~ —) ~ _ ~ _ ~ ~ — O O ~ ~ ~ — m_ _ _ ~ _ ~ ~ cr~ r ~c v~ 0 0 — ~ ~ o~ _ X ~ ~ ~ _ - ] _ t ~ _ ~ oo _ 0 o0 r~ cx~ _ ~ _ ~ —1 oo 0 x O == — '] _ .= '_ ~0 O ~ C) mm ~ ~ ~ z ~ z ~ z ~ ~ 3 ~ 9 ~ ~ ~ =g, ~ e 2 E . s ~ ' 2 s s ~ u E e ~ c 0 9 u ° u =°gEc o~!c ^c 0 0 ~ s ~ sc 0= ~ 9 9-sE 9 ~ ,, 3 3 3 D U ~ — O O E ,., v s O ~ ~ u 9 s u ~ 2 0 s c E E -3 c ,~ _= _ 5 3 c ,~ 2 -e ~ ; ~ ~ ~ ~-2g-~!~-g~Ei~ce

240 41 1 @ 3 r x 2 CL 3 1 a: J ~ O UJ A a: J o 1 o . . -2 · — . |.'r~ . · _ . _3 ~ 1 1 1 -0.5 0.0 0.5 1.0 1.5 LOG OF WET PRECIPITATION SULFATE INPUT FLUX FIGURE 7.2 Comparison of the logarithm (base e) of sulfate fluxes to lakes from wet-only deposition with the logarithm (base e) of sulfate fluxes from lakes in the northeastern United States and eastern Canada. Solid line is the y = x line (i.e., condition in which sulfate flux from the lakes equals the sulfate flux to the lakes from wet-only deposition); dashed line is the line estimated from linear regression with slope of 1.593. As noted earlier there is a tendency for lakes in some regions to lie above the line of equal fluxes (excess lake sulfate flux in Table 7.1) and other regions to lie below it (excess precipitation sulfate flux in Table 7.1). This tendency may result from regional differences in geology, dry deposition, or other factors causing the regional ratios of input to output fluxes to differ from the overall relationship. To account for regional differences we performed an analysis of covariance in which we allowed each region to have its own intercept, but fitted the same slope for all regions. To analyze the data on a regional basis, we have chosen the following six regions: Connecticut, Massachusetts, and Rhode Island (CT-MA-RI); Maine, Vermont, and New Hampshire (ME-VT-NH); New York (NY); Newfoundland (NF); Labrador (LB); Quebec (QU). The linear regressions shown for each region in Figure 7.3 yield the following estimates:

241 Region Intercept (mi') Standard Error CT~ RI 0.691 ME—VT—NH 0 . 2 57 NY 0.280 NF -0.004 LB -0 .515 QU 0.098 Slope (a) 1.089 0.0791 0.0729 0.1082 0.0626 0.0407 0.0781 0.0712 ~ The improvement in fit in going from a single regres- sion line to the six regional lines is highly significant, having a p value less than 10-6. Furthermore, the common slope (a) is not significantly different from one. Thus, there is no evidence that the relationship on the natural scale (y = mica) is nonlinear once regional differences have been taken into account. The multipliers mi' are regionally specific. Thus lake sulfate output flux y is estimated to be a multiple of precipitation sulfate input flux x. (Note, that on the natural scale, our model will not in general estimate the arithmetic mean of the lake sulfate flux.) The multiple depends on such factors as the ratio of total deposition to wet-only deposition, the rate at which sulfur is demobilized in the watershed or sediments, the magnitude of sulfur sources in the watershed, and the magnitude of any bias in our estimate of wet-only sulfate flux. The values of mi may be calculated from the antilogarithms of the intercepts (mi') shown in Figure 7.3. The values are as follows: 2.00 (CT-MA-RI), 1.32 (NY), 1.29 (ME-VT-NH), 1.10 (QU), 1.00 (NF), and 0.60 (LB). These multiples are greatest in regions close to strong sulfur source areas and smallest in remote areas, suggesting that dry deposition plays a major role. Since the 0.60 value for Labrador indicates that output fluxes are smaller than input fluxes, watershed processes that retain sulfate may be important. To obtain more detailed information of sulfur mass balance from these data, each location and perhaps each watershed needs to be considered separately. In some cases internal sources of sulfate may be important contributing factors to the sulfur mass balance (Wagner et al. 1982). For example, in Table 7.2, the geological setting is described for the locations of lakes in Massachusetts and Maine having large excesses of lake sulfate fluxes. Since strike zones, amphibolites,

242 ', 4 _ x 6 2 _ ; 1~—- ~ O _ O O log x axis UJ -1 _ -1 ~ -2— 1 ~ _3 1 1 1 `, {).5 0.0 0.5 1.0 1.5 o 4 _ Ox _ ~ ~ _ O m'i = 0.280 _ tic .5 0.0 0.5 ~5 o -1 -2 CT-MA-A I NY ma x axis 2 1 4 o _ I -3 _ 1.0 1.5 -0.5 ~ F -3 _3 ~~' -0.5 0.0 0.5 1.0 1.5 -0.5 o . a.. .~. ~ ·.e .'` m''=~)515 LOG SULFATE INPUT (9 m~2 yr~1 ) 4 3 _ x ME-VT-Ni ~ m'= 0.257 _ ~ I . . 0 ~ -2 _ 3 1 _0 5 0.0 log x axis 1 1 1 0.5 1.0 1.5 QU m' = 0.098 log x axis 0.0 1 1 1 0.5 1.0 1.5 a) . _ x o ~ - ~~ _ · ~ m'i = -0.004 1 1 1 1 0.0 0.5 1.0 1.5 NF log x axis LOG SULFATE INPUT (9 m~2 yr~1 ) FIGURE 7.3 Regional relationships between inputs and outputs of sulfate for lakes in northeastern North America assuming that the only source of sulfate input is from wet deposition. volcanics, and basalts are associated with sulfide minerals, it is quite possible that the excess sulfur in lakes with these associations are due to contributions from sulfide minerals. On the other hand, there are four lakes listed in Massachusetts (Big Bear Pond, Milham Street Pond, Round Pond, and Stearns Pond) for which

243 TABLE 7.2 Geological Characteristics of Watersheds in Massachusetts and Maine with Large Excess of Outflows of Lake Sulfate Name Location Geological Setting Massachuetts Wheeler Pond 42° 25' 08" On a major strike fault, amphibolites 71° 3 1 ' 17" and Andover granite. Holts Pond 42° 09' 46" On a major strike fault, amphibolites. 71° 19'01" Bear Pond 41° 56' 27" Molasse. Probable other anthropo- 70° 59' 38" genie source. Little Farm Pond 42° 15' 00" Mattapan volcanics. 71° 22' 41" Milham Street Pond 42° 19' 17" Near major strike fault with diabase 71° 36' 13" dikes, but probably other anthropo- gen~c sources. Round Pond 42° 35' 53" Alkaline complex and fault. Sulfur 70° 49' 01 " excess probably anthropogenic. Stearns Pond 42° 37' 05" Fishbrook gneiss. Sulfur excess prob- 71° 04' 04" ably anthropogenic. Maine Iron Bound Pond 45° 46' 04" Devonian slate unit with fault. 70° 05' 23" Ferguson Pond 46° 51' 58" Adjacent to recently found sulfide ore 68° 41' 55" deposit. Bluffer Pond 46° 22' 30" Bluffer Pond volcanics. 69° 05' 58" SOURCE: G. M. Boone, Syracuse University, personal communication, 1985. there are no obvious geological sources of sulfur, and these may be affected by anthropogenic contributions. TRENDS IN THE CHEMISTRY OF HEADWATER STREAMS OVER THE PAST 15 TO 2 0 YEARS In this section we consider trends in alkalinity, strong-acid anions, and nonprotolytic (base) cations* in *Protolyte--any electrolyte that reacts with protons (Ht ions) in the system of interest and especially in the pH range of interest. The proton condition is the ion balance condition for protolytes (e.g., Stumm and Morgan 1981, pp. 138-139).

244 small streams of the U.S. Geological Survey's Bench-Mark Network. The data are analyzed to assess trends and to determine whether the trends are consistent with the mechanisms thought to govern changes in surface water alkalinity driven by changes in sulfate deposition. In Chapter 1 these trends were related to other types of regional trends thought to be related to acid deposition. A general model linking acidic deposition to changes in surface water chemistry in watersheds with acid soils has been prominent for a number of years, (i.e., Seip 1980, Overrein et al. 1980, Cronan and Schofield 1979), and we use the version below (National Research Council 1984), which suggests that sulfate-driven changes in surface water chemistry occur as follows: 1. Sulfur deposition increases. 2. Sulfate concentrations in surface waters increase 3. Concentrations of protolytic (H+, A13+) and nonprotolytic (Ca2+, Mg2+, Na+, K+) cations in surface waters increase. 4. The increase in protolytic cations leads to decreases in surface water alkalinity. This model is a consequence of the fact that charge balance must be maintained when the concentration of a mobile anion increases in a solution moving through an acid-dominated ion-exchange system. While the model does not take into account the complexities involved in the flow of Son ~ through the terrestrial portion of the watershed, it has been widely accepted on the strength of empirical support from studies of short-duration (e.g., a few years) time series (i.e., Galloway et al. 1983a) and in spatial comparisons of surface water chemistry in high versus low sulfate deposition areas (i.e., Mohn et al. 1980; Environmental Protection Agency 1984). The areas most prominently studied have been glaciated regions with thin, acidic soils, where clear-water lakes and streams having low alkalinity. Base cation/acid anion--generally refers to a strong base cation/acid anion that for the system of interest can be considered to be totally dissociated, and hence its concentration is not a function of pH (e.g., base cations: Ca2+, Mg2+, Na+, K+; acid anions, C1-, NOT, etc.).

245 In assessing trends in surface water alkalinity that span periods of decades, it is important to take into account the fact that internal changes can cause acidification of soils (particularly surface horizons) and perhaps acidification of surface waters. Major internal sources of HE in forest soils are dissociation of carbonic acid, cation accumulation in biomass, mineralization of organic sulfur and nitrogen compounds, vitrification, oxidation of reduced mineral compounds (e.g., pyrite), and humus formation (i.e., Driscoll and Likens 1982, Ulrich 1983). The major sinks for H+ are reduction reactions, anion accumulation in biomass, mineral weathering, and destruction of organic acids. Any shifts or trends in the balance of processes governing the H budgets of soils can change the net production of H+ in the soil. In this regard Rosenquist et al. (1980) and Krug and Frink (1983) suggest correctly that the balance of H+ production and consumption (par- ticularly in the upper soil horizons) can be strongly acidifying during the development of maturing forests or after recovery of forests from major fires. In acid forest soils not subjected to the input of strong-acid anions, most highly acidic solutions are the result of organic acids. This acidity is not easily transported through soils to surface water because the organic anions are retained or immobilized in the subsoil as a part of the regime of podsol development. LeachateS are normally most acidic in the O and E horizons (the organic forest floor and uppermost mineral horizon, respectively). Acidification of the upper soil horizons does not necessarily translate into acidified surface waters because a mobile anion is needed to transport the protolytes. In cases where water from acidic surface horizons or from acid organic soils drains directly into streams, the stream water can be highly acidic because of organic acids. In theory, if water drains directly into streams through acidifying soil horizons, a trend of surface water acidification could be observed. However, to date, long-term changes in soil acidity owing to natural biogenic processes have not been shown in field studies to be paralleled by changes in surface water chemistry. Decreases in surface water alkalinity that might result from the effect of runoff routed through naturally acidifying soils can be distinguished from sulfate-driven decreases in alkalinity in that long-term decreases in alkalinity resulting from acidifying soils should in most

246 cases be accompanied by a decrease in the concentration of nonprotolytic cations. This follows because as the soil acidifies, cation exchange sites will be increasingly occupied by protolytic cations at the expense of non- protolytic cations, and the ratio of protolytic to non- protolytic cations in the soil solution is controlled by their ratio on exchange sites and by the anion concentra- tion in the soil solution (Reuse 1985). Acidification of soils tends to increase the sulfate adsorption capacity, and reduce the dissociation of weak acids such as carbonic acid and organic acids, subsequently decreasing the soil- solution anion concentration and the effectiveness of cation leaching. The overall result of soil acidification from natural biogenic processes is that the soil solution becomes more acidic with lower concentrations of non- protolytic cations. As a first approximation, alkalinity trends in surface waters can be due to changes in strong-acid anion flux plus changes in internal processes in accordance with Eq. (7.1): ~AlkAA + AAlki = ^£Bi + h7BAA—(BAA + 6£Ai) where hAlkAA is the change in alkalinity due to changes in atmospheric strong-acid anion flux; hAlki is the change in alkalinity due to internal processes; b~Bi is the change in base cation concentrations due to internal processes; A£BAA is the change in base cations due to changes in atmospheric acid anion flux; h£AAA is the change in strong-acid anion concentrations due to changes in atmospheric flux; and h7Ai is the change in strong-acid anion concentrations due to changes in internal processes. It should be noted that Eq. (7.1) ignores the possible secondary effects of acidic deposition on soil character- istics that affect alkalinity, and the addition of inter- active terms may be appropriate. Equation (7.1) is, how- ever, sufficient to demonstrate that several combinations of long-term changes in alkalinity, base cations, and strong-acid anions are possible. The balance of those processes will be governed by local soil and vegetation characteristics, disturbance history, and mobile anion deposition. Having recognized that internal sources of anions and H+ as well as atmospheric sources of mobile anions could affect trends in surface water alkalinity, and having suggested that there could be interactions between

247 acid deposition and natural hydrogen-ion budgets, we point out that the deposition of strong acids or acidify- ing substances such as Nat to soils can promote soil acidification (Van Breeman et al. 1982, Ulrich 1983) by increasing proton input to the soil. The rate at which a soil will acidify under such influences is a function of the magnitude of the ionic input compared to the pool of exchangeable ions in the soil, and the ratio between acidic and basic cations in the solution added to the soil. In strongly acid forest soils, the annual ionic input is generally a very small fraction of the exchange- able pool, so the effect of atmospheric deposition in changing bulk soil chemistry would likely be slowly realized in most cases in the United States. In summary, the pathways by which changes in surface water alkalinity may occur are through changes in the deposition of mobile anions and, under appropriate hydro- logic conditions, changes in soil acidity, which can be natural or promoted by acidic deposition. Additional complexities are introduced by changes in processes that alter alkalinity within surface water bodies and by changes in weathering rates that may accompany changes in acid flux. Thus any trends in surface water alkalinity will be the result of numerous processes, some of which are interrelated. The balance of these processes is expected to be highly dependent on local conditions. Methods Since 1964 the U.S. Geological Survey has operated the hydrologic Bench Mark Network of 47 streamflow and water quality monitoring stations in small, predominantly undeveloped stream basins (Cobb and Biesecker 1971). The network includes stations in 37 states and was originally established to define baseline hydrologic conditions in a variety of climatic and geologic settings. Because of the application of consistent sampling and analytical methods (Skougstad et al. 1979) for a 15- to 20-year period at each site, records from the network are useful for investigating recent trends in water quality. Char- acteristics of the chemistry of the streams, and the geochemistry and land use histories of the basins are given in Appendix Tables C.2 and C.3, respectively. Trends in anion and cation concentrations were estimated using the Seasonal Kendall test. This test is nonparametric and is intended for analysis of time trends

248 in seasonally varying water quality data from fixed, regularly sampled monitoring sites such as those in the Bench-Mark Network (Hirsch et al. 1982, Smith et al. 1982). In addition to a test for trend, the statistical procedure includes an estimate of the median rate of change in concentration over the sampling period (called a trend slope in this report) and a method for adjusting the data to correct for effects of changing stream flow on trend in the water quality record. Trend is defined here simply as monotonic change with time occurring as either an abrupt or a gradual change in concentration. Quality-assured data sets for each stream typically contain anion (alkalinity, NO3, SOi~, C1-) and cation (pH, Ca2+, Mg2+, Na+, K+, Nat) concentrations for 100 to 150 samples collected over 15 years, along with data on specific conductance and stream discharge, suspended sediment, etc. Data used in our analysis were screened using electroneutrality and conductivity balances (Z anion equivalents = ~ cation equivalents + 15 percent; and calculated specific conductance = measured specific conductance + 15 percent) and by assessment of the quality-assurance measures used in sample collection, processing, and analysis. Although our inability to account for aluminum and organic anions in some highly acidic streams (e.g., McDonalds Branch, New Jersey) may have introduced some bias into the analysis, we do not believe that this limitation is serious. Sulfate output from the Bench Mark wafer sheds was estimated as the product of discharge-weighted mean sulfate concentrations and mean basin runoff for the 2-year period from 1980 and 1981. Sulfate deposition at each basin was estimated by linear interpolation of wet-only sulfate deposition data from the National Acid Deposition Program (NADP) for 1980 and 1981. Results Sulfate Input and Output Critical to interpreting trends in stream sulfate is an understanding of watershed sulfur budgets and ecosystem sulfur cycling. Figure 7.4 indicates the major pathways of sulfur retention and release that need to be considered in determining how closely SOi~ inputs will be reflected in S024- yield in streams. Watersheds that

249 Input from Atmosphere (wet + dry deposition) Reduction to H2S < Output to Surface Waters 4 Above-ground Biomass =, 14 ,, 1 Dry deposition washoff, sulfate leaching from foliage = Inorganic S Soil Solution = Organic S - Geochem ice I Sulfur (Subsoil/Geological substrate) Litterfall, dissolved organic sulfur leaching ,\;%~ .~--,\~.. ' Soil Pools FIGURE 7.4 Flow diagram of the movement of sulfur in an undisturbed forested watershed. are in a steady state with respect to sulfur would be the least ambiguous cases in which to judge the impact of changing rates of atmospheric sulfate deposition since changes in inputs are expected to be reflected in changes in output of similar magnitude and direction. Since

250 2.0 1.0 o o ~ 2.0 - ~— 1.0 By <1: 0 111 2.0 1.0 o ME NY NJ PA MN1 MN2 Ml Wl NO RTH EAST _ ~ ~ n ~ ~ n n ~ ~ n rrrrl~ 9~ hit ~ ~tt4 ffl~ VA NC TN SC1 SC2 GA F L A L MS LA AR K SOUTH EAST mall n n I I ~ n n in n I I SD NE OK1 OK2 NM CO WY AZ 101 ID2 WA OR CA WEST FIGURE 7.5 Ratio of stream basin sulfate yield to sulfate input from wet-only deposition at Bench-Mark stations that have ratios less than 2.0. Hatched bars correspond to streams with mean alkalinities less than 500 peq/L. White bars correspond to streams with mean alkalinities greater than 500 peq/L. watershed-scale studies lack quantitative measurements of dry deposition, a precise understanding of sulfur cycling and budgets has not been achieved even in the most inten- sively studied sites (i.e., Likens et al. 1977, Richter et al. 1983). As a result, any explanation of the cause of trends in stream sulfate is subject to uncertainty since trends in internal sources and sinks as well as trends in atmospheric deposition may contribute to trends in surface water SOi~. For Bench-Mark watersheds in the northeastern United States, many variations of sulfur budgets are possible, with an overall tendency for output to exceed input owing to the effects of dry deposition and/or internal S sources (Figure 7.5). Richter et al. (1983) showed that major amounts of S ~ ~ were retained in the B horizon of some of the

251 Paleudults at Walker Branch, Tennessee, attributable in part to So2 adsorption (Johnson and Henderson 1979). Thus, some soils (particularly Ultisols and Oxisols) will serve as effective sinks for atmospherically deposited SO4 until the sulfate-sorption capacity becomes limited. Another means of sulfate retention is through incorporation into soil organic-S pools (Fitzgerald et al. 1982), and this mechanism could be quantitatively important in some instances. Sulfate flux (as wet-only deposition) to and from all the U.S. Geological Survey's Bench-Mark watersheds for 1981 is shown in Appendix Table C.2. In trying to assess trends that might be related to changes in atmospheric sulfate flux, we have used watersheds whose ratios of yield to wet input were less than 2.0, judging that greater values indicated basins with important internal sources of S~~ (Figure 7.5). We have done this without definitive information on what the input of sulfate from dry deposition might be, but, based on precipitation and throughfall studies in forests, it is reasonable to expect that inputs from dry deposition could be equal to the input from wet-only deposition in some cases (Morrison 1984). Similarities of sulfate budgets within some regions are notable. So2~ output equals or is greater than wet input in watersheds of the Northeast, reflecting the influence of dry deposition, internal sources, or both. Sulfate appears to be strongly retained in watersheds of the Southeast. Soils in those watersheds are predominantly Ultisols (Appendix Table C.3), which typically have high So2 sorption capacities (Johnson 1980, Olson et al. 1982). Thus, sulfate-mass balance at the southeastern stations may be influenced by sorption sites in subsoil mineral horizons and/or by incorporation of sulfur into soil organic pools. Ratios of sulfate input to output at the western sites are variable, and they are difficult to account for in many cases given the information available on geological and soil characteristics. Trends in Stream-Water Sulfate in Bench-Mark Streams Figure 7.6 shows trends in stream sulfate concen- trations over the periods of record for the watersheds that were judged not to have dominating internal S sources.

252 W>-- ~ · _. · No trend (p>0.1) · Trend up(p~0.1) ~ Trend down(pSO.1) Of FIGURE 7.6 Trends in sulfate conc~nt-rAt-;On~ Ah- R"n~h- . · . . mar K stations naming a ratio of basin sulfate yield to basin sulfate deposition (wet-only) of less than 2.0. Regions B. C, D, and E are highlighted by darkened lines. Sulfate concentrations tended to increase during the 15- to 20-year period at stations over a broad area of the continental United States extending from the Southeast to the mountain states and the Northwest. By contrast, stations in the northeastern quarter of the nation show either no trend or decreases in sulfate concentrations during the same period. This geographical pattern of trends occurs more or less independently of the sig- nificance criteria used in trend testing (Smith and Alexander 1983). Although the statistical significance of trends at many of the stations is high (p < 0.01), the magnitude of change has been small in most cases (Appendix Table C.2). Among stations showing a statis- tically significant trend in sulfate concentration (either a significant increase or a significant decrease), there is a median relative cnange in stream sulfate of 1.2 percent per year, or an overall change of about 20 percent over the period of record. The trends in sulfate observed in Bench-Mark streams of the Northeast (Region B) and Southeast (Region C) regions are consistent with regional emission trends (Chapter 2, Figure 2.24) and with the contention that

253 changing sulfate levels in small streams reflect changing emission and deposition patterns in watersheds that do not show evidence of dominating internal sulfate sources. Obviously, it would be preferable to have detailed sulfur cycling data for each watershed for further confirmation that sulfur output actually-reflects atmospheric sulfur input. Insight into whether there is a cause-and-effect relationship between regional emissions and stream sulfate levels would be improved by site specific, process-level information regarding (1) the applicability of regional emission trends to the atmospheric sources contributing to deposition at each site and (2) the sources and sinks aovernina I'll fat" filly Huh arch watershed. Although considerable sulfate is retained in the Bench- Mark watersheds of the southeastern United States, there is a consistent pattern of increasing stream-water sulfate concentrations over the period of record. The trends in stream sulfate strongly reflect the regionwide trend in emissions, but we believe that more needs to be known about sulfur dynamics in the soils of those watersheds before we conclude that increased emissions have driven the increases in stream sulfate. , ~ . ~ ~ =—, ~ ~ _ . . Trends in Alkalinity and Cations The effect that increased sulfate flux through strongly acid soils may have in reducing surface water alkalinity is of particular practical concern. In acid soils charge balance requires that an increased cation flux must accompany an increased sulfate flux in the soil solution, and the balance between protolytic and nonprotolytic cations is determined by the ability of the soils to supply nonprotolytic cations through mineral weathering or cation exchange. In watersheds with thin, extremely acid soils, exchange sites are dominated by A13+, and H+, and mineral weathering is usually limited. Thus, changes in Sol ~ flux are expected to be accommodated mostly by changes in protolytic cation concentrations, thereby altering alkalinity in receiving water. In watersheds dominated by high base status soils and/or readily weatherable minerals that contain base cations, short-term changes in SO4 flux are expected to be balanced mostly by changes in nonprotolytic cation flux. Thus, surface waters are expected to respond to increasing sulfate flux by both reduction of alkalinity and increases in non-

254 ~1~1, 1 ~ 1N · No trend (p > 0.1 ) · Trend up(pSO.1) Trend down(p60.1) Or I; 7~ I, \_g' FIGURE 7.7 Trends in alkalinity at Bench-Mark stations having a mean alkalinity of less than 800 peq/L. Regions B. C, D, and E are highlighted by darkened lines. protolytic cation concentrations. In general, the magnitude of the alkalinity changes will therefore be less than the magnitude of the sulfate changes. Figure 7.7 shows trends in alkalinity at stations with low to moderate mean alkalinity (<800 peq/L). The regional pattern of trends is the approximate inverse of that of trends in sulfate (i.e., decreases in the South and West and increases in the Northeast) but station-by- station comparisons of statistically defined trends (p < 0.1) fail to show a consistent inverse relationship (see Appendix Table C.2). Following is a more detailed analysis of short-term changes in sulfate and alkalinity at these stations. Nitrate is also a strong acid anion added from the atmosphere, and, although NO3 is apt to be strongly retained in the terrestrial ecosystem by plants and soil organisms in most cases, changes in NO3 concentrations in surface waters must be considered along with sulfate changes in interpreting alkalinity changes. For cases in which changes in strong-acid anion flux drive changes in stream chemistry, the relationship between changes in strcng-acid anions, alkalinity, and cation concentrations given in Eq. (7.1) can be summarized in general terms with the equation

255 AZA = 6£B - AAlk, (7 .2) where ZA is the total strong-acid anion concentration (nitrate plus sulfate, peq/L), £B is the total base (nonprotolytic) cation concentration (Ca2+ + Mg2+ + Na+ + K ) (peq/L), and Alk is the alkalinity (peq/L). Equation (7.2) is essentially a statement of the charge balance that must prevail when a change in strong-acid anion flux is the variable affecting a change in stream chemistry. Equation (7.2) can be differentiated with respect to the strong-acid anion concentration to give dIB - dAlk = 1. d£A dIA (7.3) The derivatives d£B/d£A and dAlk/d£A are the rates of change of base cation concentration and alkalinity, respectively, per unit change in acid anion concentration. Their values will differ from one geochemical setting to another and provide a measure of the ability of a basin to accommodate short-term changes in sulfate flux by supplying base cations. Estimates of the derivatives for Bench-Mark stations were obtained from linear regressions of major cation concentrations and alkalinity on sulfate plus nitrate concentration. In contrast to the 15- to 20-year trends presented in Figures 7.6 and 7.7, these regressions tend to reflect short-term variations in stream chemistry. Values of the estimates are given in Table 7.3. The observed effect of strong-acid anion change on cations is due almost entirely to changes in S ~ ~, since both concentrations and changes in nitrate are very small. A plot of d£B/dEA versus d Alk/d£A is given in Figure 7.8 for streams with mean alkalinity of <500 peq/L. For a majority of eastern stations and a number of lower- alkalinity western sites, dAlk/dEA is negative and d£B/~IA ranges from O to 1. Increases in acid anion concentra- tions are balanced, in part, by decreases in alkalinity and, in part, by increases in base cations. Alkalinity changes in the low-alkalinity streams (~500 peq/L) balance about 30 percent of the change in SOi~, and changes in base cations balance about 50 percent of the change in SOi~. We have not conclusively determined why the residuals (1 - (d£B/d:A - dAlk/dIA)) are generally positive, but this may be due to the effect of ions that we did not use in the model (i.e., C1-, NHt, A13+).

256 0.5 ~ 0.0 C: Z _0.5 CA . ·SC DISC WASHe NY · · OR PA- · WYO _ . _ -1 .0 / 0.0 0.2 · NC - LA:/ - ·MS - 0.4 0.6 0.8 1.0 d ~ BASE CATIONS/d ~ ACID ANIONS FIGURE 7.8 Plot of rate of change in base cation concentrations versus rate of change in alkalinity per unit change in acid anion concentration for stations with mean alkalinities less than 500 peq/L. Some stations in the group have small positive values of dAlk/dIA that are not statistically distinguishable from zero (Table 7.3). At these stations, acid anion changes are essentially completely balanced by changes in base cations. Figure 7.9 shows a larger set of stations, including for comparison sites with mean alkalinity values between 500 and 800 peq/L. At higher-alkalinity sites, values of the derivatives are usually greater than 1.0, indi- cating that alkalinity and base cation changes cannot be entirely driven by changes in So2 . Mechanisms other than mobile anion flux must therefore be important at those sites. At low-alkalinity sites, the data are consistent with the hypothesis that changes in S ~4 flux drive changes in cation flux, resulting in changes in base cation concentrations and/or in alkalinity.

257 o .= ._ ._ u, ct ._ —a. ._ ._ - Hi: o ._ ct ~ ._ - ._ - o ._ ., ~3 ~ so Ct m ~ Ct lo: ~ 0 Ct ~ 0 o o . _ ~ Ct ._ -= o As U: ~ Am ~ _ S: o A') . _ A_ A: 3 m ~ 1 so cat .o ,<O as, I' _- 0 ._ ct ._ ct ~ ct ~' _ _ ~ +1 _- 0 ._ ct ct <¢ en cat m_ al = o - ~n ~ o ~ - - ~ ~ o ~ oo ~ oo - ~ ~ o ~ ~ ~ ~ o ~ o ~ o ~ ~ ~ - ~ - ~ ~ ~ o ~ ~ oo ~ ~ . . . . . . . . . . . . . . . . . . . . oooooooooooooooooooo 1 o oo ~ ~ o o u~ ~ ~ oo ~ o oo ~ ~ ~ ~ oo ~ ~ ~ - ~ - ) . . . . . . . . . . . . . . . . . . . . ooooooomooooooo - oooo 1 1 1 1 1 1 1 1 1 1 1 1 1 ~ ~ oo c~ ~ ~ o ~ ~ ~ c~ ~ ~ o o ~ ~ ~ o o o o - o o o o 1 1 1 1 1 1 1 1 O ~ - - ] o o o o - o o o o 1 1 1 1 1 1 1 ~ - oo ~ ~ ON c')—] ~ ~) . . . . . . . . . . . o o o o o o o - o o o 1 1 1 1 1 1 1 1 ~ ~ ~ ~ — — ' ~ ~ ~ ~ ~ ~ oo ~ ~ o ~ - - ~ ~ ~ ~ ~ - o ~ - ~ ~ . . . . . . . . . . . o o o o o o o - o o o 1 1 1 1 1 1 1 ~ - ~ ~ ~ ~ ~ ~ ~ - ~ - ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ _ ~t ~ ~ ~ cr~ ~ ~ 0 0 ~ — O oo '} ~ - - ~ ~ ~ ~ o ~ ~ ~ ~ ~ ~ o ~ - ~ ~ ooooooom - oooooo~oooo ~ o ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ o ~ ~ oo O — ~ ~ oo oo ~ ~ ~ ~ ~ ~ ~ ~ 0 _ ~ . . . . . . . . . . . . . . . . . O 0 0 0 — ~ 0 0 0 0 0 0 — ~ 0 0 0 1 O ~ ~ 0 0 0 0 0 0 ~ ~ ~ ~ _ ~ ~ ~ ~ ~ ~ ~ ~ ~ _ ~ O~ ~ _ ~ ~t 0 - ) ~ ~ ~ ~ ~ ~ 0 ~D ~ ~ O0 ~ ~ ~ ~ u~ . . . . . . . . . . . . . . . . . 0000000~0000000~0~ z `~o~ ~ Z 3 Z C ~ ~ V A ~ ~ ~ ~ ' E ' ' C 2- 2 g E u ~ - _ 2 c, . ' _ ~ ; O a ~ G ~ C ~ ~ _ ~ ~ E 3 ~, ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ x ~ ~ Z ~ x ~ ~ ct Q0 00 O ._ c' =: u, _ c~ ct ct C: C: c~ o ct c~ C~ O _ ct _ ct ~ ~ .= c.> O O Z ~

258 2.0 z en 1.0 C) 0.5 - , 0.0 Hi: J _0 5 ~ VA -1.0 ~ I I I l l 0.0 AL ~ / _ CO CA ~ eMS WASH Nit- PA · OK' ME /1 ~ ~~\~ /`,\0~ ,L,\0~ J 0.5 1.0 1.5 d ~ BASE CATIONS/d ~ ACID ANIONS 2.0 2.5 3.0 FIGURE 7.9 Plot of rate of change in base cation concentration versus rate of change in alkalinity per unit change in acid anion concentration for stations with mean alkalinities less than 800 peq/L. Interpretation of 15-Year Trends in Alkalinity In addition to long-term trends in surface water chemistry, which might be caused by changes in atmospheric deposition of strong acids, processes in soils or in streams controlling hydrogen-ion budgets may have changed through time to produce changes in base cations, strong- acid anions, and alkalinity. Changes in mineralization rates of nitrogen and sulfur compounds, changes in weathering rates, and natural processes of soil acidifi- cation could result in such changes. The trends in cations and alkalinity apparent in the Bench-Mark stream data afford an opportunity to evaluate the possible role of internal Processes i n round (15-year) trends in alkalinity and strong-acid anions. Figure 7.10(a) chows trend slopes for alkalinity minus base cations plotted against trend slopes in sulfate plus nitrate for the 12 stations that display charge balance (Eq. (7.2)) errors of less than 1.5 peg L 1 yr 1.

259 In On ~ 6.0 of o Z 2.0 At LU I: 0.0 ZJ -2.0 A: J At at us 6.0 - .0 -6.0 (a) Am, ~ ME NY- . MN ~CO NC: OK- ID \ ·SC -4.0 -6.0 -6.0 - .0 1 1 1 -2.0 0.0 2.0 4.0 6.0 TR EN DS I N SU LFATE PLUS N ITRATE (,ueq L-1 yr~1 ) (b) O Predicted Alkalinity Trend (d Alk/d~A PEA) · Observed Alkalinity Trend NJ \ NY M E>! t &4 ~ ~ MN '_ CO SC 9! ISC OK \ it, 1 1 1 1 1 1 -6.0 - .0 -2.0 0.0 2.0 4.0 6.0 TREND IN SULFATE PLUS NITRATE (,lleq L-1 yr~1 ) FIGURE 7.10 (a) Trend in alkalinity minus trend in base cations versus trends in sulfate plus nitrate. (b) Trend in alkalinity versus trend in sulfate plus nitrate.

260 The condition that charge balance is maintained within reasonable limits suggests that the trends are probably accurately modeled. If trends in acid anion fluxes were the only cause of observed trends in stream chemistry, it would be expected that most alkalinity trend slopes would be considerably less than strong-acid anion trend slopes since some of the change in the latter is accompanied by changes in base cations. Figure 7.10(b) shows the alkalinity trend slope that would be predicted from the values of (dAlk/d£A)~IA (Table 7.3) for the 12 stations plotted against aZA. Observed values of AAlk are also plotted in Figure 7.10(b) for comparison. In a few cases alkalinity reductions have occurred (in Minnesota, Colorado, and Oregon) that are considerably in excess of the alkalinity change predicted on the basis of strong- acid anion changes. It is noteworthy that the large decreases in alkalinity are mostly balanced by increases in S~~ and decreases in base cations. The decrease in base cations suggests that souse processes other than or in addition to mobile anion leaching are important in causing alkalinity reductions. On the other hand, increasing alkalinity in streams of the Northeast can be accounted for almost entirely by the reduction in sulfate deposition within the error constraints of the analysiS. PROTOLYTIC CHEMISTRY OF SURFACE WATERS OF NEW HAMPSHIRE, NEW YORK, AND WISCONSIN Extensive sets of historical and recent data exist for lakes in New Hampshire, New York, and Wisconsin from which conclusions may be drawn regarding changes in the acidic status of poorly buffered lakes. These data are perhaps unique in that detailed records of both sampling and analysis exist and measurements of alkalinity, acidity (free CO2 acidity), and pH were routinely made. Having measurements of the three parameters permits an assessment of the internal consistency--and hence the quality--of the data because the chemistry of the system is mathe- matically overdetermined if carbonate protolytes are assumed. Analysis of the data is also illustrative of many of the factors that should be considered when attempting to assess time trends in data on aquatic protolytes. The assessment procedure is as follows: (1) Define the underlying chemical principles. (2) Use the

261 principles to adjust the data for bias owing to changes in analytical techniques. (3) Define a consistent data set by comparing independent estimates of alkalinity and pa. (4) Analyze the data for temporal changes. Chemical Principles Techniques for measuring alkalinity, pH, and acidity underwent significant changes between the 1930s and the 1980s. Early methods all relied on calorimetric tech- niques, which employ special chemicals, or indicators, that change color at specific values of pH. Recent methods entail electrometric techniques. Considerable uncertainties exist regarding the calorimetric data. First, they depend upon observations of color change that can vary from analyst to analyst. A1SO, we often do not know whether formulations of indicators have changed over time, thus altering the endpoints of titrations. Further- more, the indicators themselves may be acids or bases, and even when they are added in small quantities they can alter the pH of the test solution and introduce systemati bias if unaccounted for. Such methodological biases may be large in historical data on poorly buffered waters. At the same time, recent data may also contain methodo- logical biases or inaccuracies. For example, electro- metric measurement of pH in dilute solutions is often unreliable (Blakar and Digernes 1984). As will be shown, the pH at the true equivalence point in an alkalinity titration varies as a function of the original alkalinity of the test samples. Modern methods, such as the Gran titration technique (Gran 1952), take this into account, whereas the older titration technique, using the indicator methyl orange (MO), does not. That is, the endpoint pH in the MO technique is fixed, usually well below the true equivalence point. The MO indicator changes color according to the sequence yellow-orange- pink-red, spanning a pH range from about 4.6 to 3.1 (Clark 1928). The exact values of pH at which the color changes appear, as well as which color changes were used in particular circumstances, are not always evident (vice infra). Where corrections for this overtitration have been made in previous work, investigators have assumed values of the pH at the endpoint ranging from 4.04 to 4.5 (Kramer and Tessier 1982, Church 1984). Zimmerman and Harvey (1979) concluded that no corrections could be made because the true endpoint pH varies in an indeterminant c

262 manner. However, several other authors have assumed that the pH of the true equivalence point is invariant at pH = 5 for low-alkalinity waters, which is not precisely true (for a review, see Church 1984). The correction, simply made as the difference between the hydrogen-ion concen- tration at the indicator endpoint (about 10 4 N) and the assumed true endpoint (10-5 N) is a better approximation but not so good as can be achieved. we offer an alternative--and chemically more rigorous-- approach (Kramer and Testier 198Z). The calculations for this approach consider the correction of the MO and phenolphthalien endpoints for alkalinity and acidity and the correction of the calorimetrically determined pH to the actual pH that is due to the acid/base effect of the indicator. Furthermore, we employ the fact that a car- bonate system is overdetermined if data on pH, alkalinity, and CO2 acidity are available. Thus two independent values of alkalinity and three pH values are obtained. The following discussion outlines the approach, whereas Appendix D discusses the derivations and calculations in detail. Alkalinity is the "excess base" relative to the ion-balance condition for the solubility of CO2 in water and is zero when all the strong-acid concentrations equal all the strong base concentrations in the presence of CO2:* [Ark] = [HCO3] + 2[CO3] + [OH] - [H], whereas CO2 acidity ([Acy]) is defined as the "excess" acid relative to the solution condition for MHCO3: [Acy] = [H2CO3] + [H] - [OH] - [CO3], (7~4) (7.5) where square brackets [ ] indicate molar concentrations. Equations (7.4) and (7.5) are valid for a system of carbonate protolytes only. The possible influence of other protolytes is considered in Appendix D. Addition of Eqs. (7.4) and (7.5) gives the total carbonate Ct: *For simplicity, charges are not designated unless needed for meaning. [H2CO3] is defined as [CO2]aq + [ H2CO3] .

263 Ct= [Ark] + [Acy] = [H2CO3] + [HOCK] + [Cab] (7.6) The pH for titration of [Ark] and [Acy] varies with Ct. The true endpoint values of pH for [Alk ] and [Acy] can be determined by setting Eqs. (7.4) and (7.5) equal to zero and solving for [H]; the results for the endpoint concentrations of hydrogen, for [Hla and [Hlb for [Ark] and [Acy], respectively, are approximately [H]a = (CtKl) 1/2 _ K2/2 and [Hlb = [K1 (K2 + KW)/Ct] 1/2 (beta) (7.7b) where K1, K2, and Kw are the first and second dis- sociation constants for carbonic acid and the dissociation constant for water, respectively. Most fixed-endpoint titrations would not have been titrated to the correct value of [H] because of the dependence of the equivalence point [H] on total carbonate (Eq. (7.7a,b)), which would vary from sample to sample. Fixed endpoints can be adjusted to the correct endpoint value, however, if the actual titration endpoints pHX[Alk] and pHy [Acy] are known. The following equations relate the actual alkalinity and CO2 acidity to the titrated fixed endpoint values [Alk]X and [Acy]y. These relationships are derived knowing that [Ark] and [Acy] are each zero at their respective endpoints: [Ark] = [Alk]x + [HCO3]X + 2[C~]X + [OH]x [ ]x ( and [Acy] = [Acy]y + [H2CO3]y + [H]y - [OHly - [C~]y, (7~9) where the subscripts x and y refer to the concentrations of each parameter at pH = x or pH = y. In practice, equations using Ct. K1, K2, Kw, and [H] are derived from Eqs. (7.8) and (7.9), and [Ark] and [Acy] are derived in an iterative fashion using these two equations (Appendix D, Eqs. (7e) and (7f)) and the definition of Ct. MO was the indicator used for alkalinity titrations until about 1950 and sometimes even later. The pH of the

264 MO color change (and the phenolphthalein color change for [Acy]) must be known if one is to correct the historical alkalinity values for a comparison to values determined today. MO undergoes several subtle color changes from yellow to pink in the pH range of 4.5 to 3.1; this change is unaffected by ionic strength (Clark 1928, Kolthoff 1931). Different pH values of the MO color change appear in the literature. Most titration procedures used before the mid-1940s followed the protocols stated in Standard Methods for Examination of Water and Wastewater (1926, 1930, 1933, 1936). The 1933 and 1936 volumes call for titration "until the faintest pink coloration appears; that is until the color of the solution is no longer pure yellow." These reference texts do not quote an exact value of the pH at the color change but do state that the pH is "about" 4.0. The 1960 edition of Standard Methods states more precisely: "The indicator changes to orange at pH 4.6 and pink at 4.0. n Kolthoff and Stenger (1947) state "Thus with methyl orange . . . the color begins to change at a pH of about 4n; and later they note that the blank adjustment is 100 peq/L which is equal to a pH of 4.0 for distilled water. They also note that the color changes from orange-yellow to orange at this point. Kramer and Tessier (1982) repeated the MO titration on a low-alkalinity sample; independent technicians following the Standard Methods (1933) procedure with the "faintest pink n directive found the pH for the color change to be 4.04. Recently, several investigators have reported higher pH values for the MO color change. Eilers et al. (1985), using known solutions of CaCO3, obtained a correction factor of 86 peq/L using the "pink endpoint" reference; this correction is equivalent to a oH for this endpoint of 4.07 or less. In addition the authors reported the appearance of a "very faintest pink" color at a pH of about 4.3. G. E. Likens (personal communication, the New York Botanical Garden, 1985) reported the same color change at 4.25, and T. A. Haines (personal communication, Fish and Wildlife Service, Orono, Maine, 1985) reported a value for the "faintest pink n endpoint between 4.3 and 4.5 depending on the alkalinity of the original solution. , _ _ ~ Haines also noted a lower "definite orange" endpoint between about 4.0 to 4.27 again depending on the original alkalinity. In an effort to determine which pH value of the MO endpoint was used in the historical surveys, we examined the relevant literature for documentation. We could find

265 no records in the New York survey citing the value of pH used in the MO titration. In New Hampshire we could only find a reference to titration to "salmon pink" (R. E. Towne, New Hampshire Water Supply and Pollution Control Commission, personal communication, 1983). The best evidence is provided in the Wisconsin historical lakes survey. In a letter to J. Juday, a principal investigator of the survey from the University of Wisconsin, F. L. Taylor (1933), an analytical chemist on the survey team, describes his dissatisfaction with the "fainter endpoint'' and discusses his rationale for changing the titration procedure from the "fainter endpoint" to an endpoint lower than a pH of 4.0: The titrations made in August for fixed CO2 were considerably higher than those of previous years, or even of July, 1932, because they were the recorded titrations made to a deeper end-point of the indicator than that previously used. . . . I found that I could detect the color change at that point easier than at a fainter color, and hence could duplicate titrations easier. . . Further, Taylor describes the pH range of the "fainter" endpoint: Titrations made previous to these were all made with this fainter end-point for which blanks probably ranged from 0.3 cc to 0.4 cc, i.e. from 1.5 to 2.0 ppm. In calculating earlier titrations by me I feel the safest blank to use is the lower (1.5 ppm). Since Taylor did his blank study with boiled distilled water, the only adjustment should be for H+. Thus the MO endpoints equivalent to 1.5 and 2.0 ppm (CO2) would be 66 and 88 peq/L, equal to values of the MO endpoint pH of 4.18 and 4.06 respectively. Eilers et al. (1985) attempted to resolve the question of the MO endpoint by examining the relationship between specific conductance in an historical and a recent set of data. From regressions of alkalinity and specific con- ductance for Wisconsin lakes, they obtained a correction of 64 peq/L for the MO endpoint, equal to a pH of 4.19 or less depending on total carbonate content (e.g., Eqs. (7g) and (7h), Appendix D). At this time we have no explanation for the disparity in the values of pH reported for the MO color change. As

266 noted above, some investigators have reported higher pH values (than 4.2) for the MO endpoint. However, we believe that the weight of the evidence suggests that a range of values between 4.0 and 4.2 is more probable. In the analyses of alkalinity changes presented in the following sections, we consider two possible endpoints for the MO adjustment: pH values of 4.04 and 4.19. The lower value is supported by the findings of Kramer and Tessier (1982), Kolthoff and Stenger (1947), Standard Methods (1960), and the recommended lower limit of the pH range noted by Taylor (1933); the higher value reflects the analysis by Eilers et al. (1985) and the upper limit of the pH range noted (and preferred) by Taylor (1933). The actual alkalinity correction, as noted before, also depends upon the total carbonate (Ct). Thus Eq. (7e) of Appendix D is used for the MO adjustment with values of [H]x of 10 4 04 and 10-4-19 peq/L. It is possible that the pH of color change may lie outside this range for specific cases owing to differ- ences in human perception of the color change as well as other factors, such as analytical set-up, sample color interference, and the overall lack of sensitivity of the method. With respect to the last-named factor, it is important to realize that for the historical methods each 0.1 cm3 of titrant is equivalent to 22 peq/L alkalinity, which is about equal to the reported precision of the method. The lack of sensitivity of the early methods probably adds additional variance to the endpoint. Phenolphtalein was the indicator used for base titration in order to determine "free" CO2. The endpoint pH (PHv = [AC]) for this titration is about u.~, wick results in a small adjustment (Eq. (1.9) ) of less than 10 peq/L. The adjustment for MO-determined alkalinity can be quite large (up to about 91 peq/L), whereas that for the free CO2 acidity correction can be very small (about 2 peq/L). A small variation in the MO endpoint of a few tenths of a pH unit can result in a variation in alkalinity of 10 to 20 peq/L. Thus, one must have con- fidence that the MO titration was carried to the appro- priate pH. A consistency check can be carried out whereby an independent estimate of [Ark] is made from the combination of [Acy] and pH measurements. The two values can be compared, and after establishing an acceptable margin of difference one can determine whether the values of [Alk], pH, and [Acy] obtained from the colorimetric pH, [MO], and [AC] are consistent or inconsistent. ~ ~ _ . . .

267 Equation (13) of Appendix D gives the relationship for obtaining Ct from [Acy] and pH. Ct and [Acy] can be applied to Eq. (7.6) to obtain [Ark] by difference. Colorimetric pH must also be corrected because the pH indicator dyes are weak acids and bases and alter the actual pH of a poorly buffered solution. The change for such a solution has been shown to be as large as 3 pH units (Blakar and Digernes 1984). For example, using one half of the indicator concentration as specified for old comparators, Blakar and Digernes show that an actual pH of 4.8 in a poorly buffered solution would be read as a pH of 6.0 if bromthymol blue, a commonly used indicator in historical surveys, were used. Kramer and (1982) showed that addition of bromthymol blue to the comparator recipe to a sample with near alkalinity changes the pH from an actual value Tessier according zero of about 5.2 to a value of nearly 7. At the same time, the same measurement carried out in a well-buffered solution will have a negligible effect on the pH. Therefore, to adjust for the indicator effect, one must know the calorimetric pH, the kind of indicator, details of concentrations and volumes, and the concentration of one other protolyte (usually balk]). The actual pH can be determined in three ways by combinations of [Alk]/pHC, [Acy]/pHc and [Alk]/[Acy], where PHC is the calorimetrically determined pa. The results of the three determinations are compared for equivalence as a consistency check, and the data may be accepted or rejected for an acceptable margin of differ- ence. Equations (10) to (12) of Appendix D give the relationships for obtaining the pH from the three combinations of [Alk], [Acy] and pHc. To illustrate differences between the above method for correcting historical pH and using calorimetric pH values alone, we examined historical pH values of Wisconsin lakes. Some of the lakes were of high alkalinity and required little adjustment, whereas other lakes required a large correction to the calorimetric pH value. Figure 7.11 shows the magnitude of corrections of historical data, which ranged from +0.5 to -1.25 pH units. It is also noteworthy that Birge et al. (1935) used MO alkalinity and phenolphthalein free CO2 in a way similar to the use of [Alkl/[Acy] here to calculate a pH. In their use of the Kolthoff expression, they made the assumptions that [HCO3] = adjusted MO alkalinity and [H2CO3] = free CO2; they calculate pH, then, from [H] = [H2CO3]K1/[HCO3].

268 92 69 In LL, By at: o 46 111 m z o m 1 -1.0 -0.5 0 0.5 DIFFERENCE IN pH 71 53 y 11 o 36 ~ an G LIJ 18 Q O FIGURE 7.11 Correction of historical pH data for Wisconsin lakes, showing differences that can occur when pH is calculated by two methods of correction. pHS obtained from the method of applying a constant correction factor of 63 peq/L are subtracted from pHs obtained from the method of Kramer and Tessier (1982). In summary, (1) fundamental principles of ion balance are applied to a carbonate ion system in order to derive definitions of alkalinity and acidity; these concepts are also used to reevaluate data titrated to the uncorrected endpoint as well as to adjust calorimetric pH data. (2) If alkalinity, acidity, and pa data exist, two independent estimates of alkalinity and three estimates of pH can be made. If the estimates agree within an acceptable margin, the data may be considered internally consistent. Data and Methods For this report we analyzed historical data from New York (1929-1936), New Hampshire (1930s and 1940-1950), and Wisconsin* (1925-1932) and compared them with recent *Records exist for 1925-1941. However, we included only data for 1925-1932 because of a change in titration to a lower MO endpoint that occurred in 1932. See earlier · . discussion.

N.Y. 1970s and 1980s N.H. 1930, 1940-50 N.H. 1970s Wisconsin 1925-41 Wisconsin 1980 Surface Profile Profile Surface Profile Surface 269 TABLE 7.4 Analytical Aspects of Lakes Data from New York, New Hampshire and Wisconsin Data Set Type of Sample N.Y. 1930s Surface Bottom Alkalinity Acidity pH M.O.a Phen.b Comments Gran M.O. Phen.b to pH 4.5 M.O. Phen.e Gran One survey only at various locations in same lake one data source. 248 lakes. Elec.0 Four data sources. Col. Elec. Col. Elec. Replicate depths common. One data source. 1 13 lakes. One data source. Multiple measure- ments on some lakes common. 161 lakes. One data source. Replicate measure- ments for many lakes. a Methyl orange alkalinity. bPhenolphthalein (titrant NaOH). CColor~metr~c. ~Electrometric. ephenOlphthalein (titrant Na2CO3~. data. The historical data are described in detail for New York lakes in reports of the New York Department of Conservation (1930, 1931, 1932, 1933, 1934), for New Hampshire lakes in Hoover (1937, 1938), Warfel (1939), and Newell (1970, 1972, 1977), and for Wisconsin lakes by Juday et al. (1935). There are two recent surveys for New York lakes (Pfeiffer and Festa 1980, Colquhoun et al. 1984), one recent survey for New Hampshire lakes (Towne 1983), and one recent survey for Wisconsin lakes (Eilers et al. 1985). Table 7.4 summarizes some of the important aspects of the lake surveys. As indicated above techniques for analysis of pH have changed over time. From the 1920s to the 1950s colori- metric comparators were used, and different indicators were used for different pH intervals. New York and New Hampshire lake surveys used the Hellige comparator, whereas pH was determined in Wisconsin by using the LaMotte kits. Specific indicator dyes available for use

270 were generally named in New Hampshire and Wisconsin records; we assumed that the same indicators and ranges were employed for New York surveys. For example, methyl red (pH < 6) and bromthymol blue (6 < pH < 7.6) were available for use in the 1930s New Hampshire surveys, whereas chlorophenol red was substituted for methyl red in the 1940s New Hampshire surveys. Kramer and Tessier (1982) have published the available infor- mation on indicator pH range, indicator concentrations, and their acid dissociation constants. Additional details were compiled from laboratory notebooks and discussion with personnel who conducted the surveys in New Hampshire, New York, and Wisconsin. Alkalinity and free CO2 acidity were commonly measured following Standard Methods (1926, 1933) using MO and phenolphthalein, respectively, as indicators. Because there is ambiguity about the correct endpoint of MO, we incorporated two different values of pH, 4.04 and 4.19, in our analysis. The endpoint of 8.25 pH units was assumed for phenolphthalein. For recent surveys in New York and Wisconsin the Gran Technique (Gran 1952) was used for analysis of alkalinity. The Gran function is a special case in the titration expression for alkalinity. When it is used correctly, there should be no requirement for endpoint adjustment. When [H+] predominates over all other terms in Eq. (7.4), the conventional acid-base balance becomes Fat = [H](Vs~v) = ca(v - ve) = be + blv; (7.10) where Vs is sample volume, v and Ca are acid volume and concentration, ve is the equivalence point volume, and be and b1 are linear coefficients obtained from regression of F1 on v; ve = -bo/b1 because ve = v when F1 = 0. Only values of pH of about one unit less than PHe should be used so that the assumptions of excess [H] are met. The recent alkalinity data for New York and Wisconsin were not adjusted since the Gran method was used. Alkalinity was determined for recent New Hampshire data by titrating to a fixed endpoint pH of 4.5. Equation (7e) of Appendix D, calculated for the endpoint value of 4.5, was used to correct these data. Electro- metrically determined pa in all of the recent surveys were accepted as given. A specific consistency check can be applied when the alkalinity is less than about -10 peq/L. In this case,

271 Eq. (7.4) simplifies to: [Ark] = -[H]. This check can be applied to some of the recent New York and New Hampshire data. Various calculations and consistency checks have been carried out on all of the historical data. These procedures were specified in the previous section and in Appendix D. They consist of correction of [MO], [AC], and PHC to obtain [Alk], [Acy], and pa. Generally two independent estimates of alkalinity and three independent estimates of pH were obtained for the historical data. In choosing the three data bases described here the primary consideration was the existence of data on pH, alkalinity, and free Cod acidity and detailed records of sampling and anaylsis that allowed us to check the internal consistency of the data and to evaluate their quality. Another important consideration, to which we have paid less attention, is representativeness, i.e., the degree to which the chosen lakes in a given region reflect the characteristics (morphology, water quality, history of land use in the watershed, etc.) of the complete set of lakes in the region. Two problems that we encountered were the lack of a consistent set of criteria to define representativeness and the lack of data for the complete set of lakes in the region required to characterize it accurately. Nevertheless, we note here a few general facts. The Wisconsin Department of Natural Resources has found that of the approximately 14,000 lakes in Wisconsin, 1,100 (or about 8 percent) are "acid sensitive. n If we apply the definition to the 161 Wisconsin lakes chosen in this study, 43 (or 27 percent) of the lakes are acid sensitive. We thus believe that it is likely that the lakes chosen for study in this report do not underestimate the regional trend in acidity but may in fact overestimate it. In New York State, lakes have been classified by elevation and area groupings. Larger lakes tend to occur at lower elevations, whereas smaller and more acidic lakes tend to occur at higher elevations. The data set of New York lakes analyzed in this study includes only a limited number of high- elevation lakes; just one lake is above 800 m, and only 39 lakes are above 600 m in elevation. The data set overall includes about 10 percent of all the lakes within the Adirondack region, but it includes only about 4 percent of the lakes at the higher elevations. A dif- ferent alkalinity trend may have been observed had we been able to assess more high-elevation lakes. Further- more, the number of lakes increasing or decreasing in

272 alkalinity could have been subdivided by lake-surface area because many of the larger lakes have high alkalinities. In this case, there might have been different trends for different area groupings. Replication and Consistency The laboratory precision of the MO titration method was reported as 20 peq/L, and a recent study (Kramer and Tessier 1982) gives a similar value of 24 peq/L. The calorimetric precision was reported to be 0.1 to 0.2 pH units. In the historical New Hampshire survey, investigators measured a depth profile for each lake and commonly duplicated the analysis at one depth, referred to here as a "depth replicate." In addition, in some cases entire depth profiles were repeated on the same date, or, in some instances, at a later date (up to 2 years). Averages of the duplicate profile data are referred to as "duplicate profile averages." The procedure applied here is (a) to compare the variation in alkalinity for depth replicates with the laboratory precision, (b) to determine the variation in alkalinity of duplicate profile averages obtained on the same date to establish an acceptable difference for the two alkalinity estimates,* and (c) to determine the variation in duplicate profile averages obtained on different dates to establish a value reflecting natural short-time (2 years or less) variation in addition to variability introduced during analysis and sampling. In all the estimates the root-mean-square deviation was calculated. Table 7.5 gives the results for the replication studies using New Hampshire data. The column headed MO refers to the average variations in alkalinity determined by MO titration. The variation in depth replicate mea- surements of 18 peg/L is nearly the same as the laboratory precision of 20 to 24 peq/L. Therefore it *This difference estimate probably reflects a minimum value when applied to one method, whereas the consistency check considers the difference between two different methods of estimating alkalinity; thus there may be addition of errors for the multiple method technique that would result in a larger margin of error.

273 TABLE 7.5 Vanation in Replicate Samples from Historic New Hampshire Lake Surveys Sample Sample Alkalinity (,u~eq/L) by Size MO pHc/CO2 Acidity Depth replicate Duplicate profile averages, same date 10 10 21 14 53 Duplicate profile averages, 21 67 different date 10 91 92 70 NOTE: Results are shown for different groupings. They are the root mean square values for differences in duplicate sample results. appears that the precision of field analyses was of the same magnitude as the laboratory analysis. The variation in duplicate profile averages of samples obtained on the same date was 53 peq/L, more than two times the analytical precision and represents both analytical variation due to different analysts and spatial variations of alkalinity. Variation in duplicate profile averages of MO-determined alkalinity on different dates was 67 peq/L. Compared with 53 peq/L for samples obtained on the same date, this result suggests that historically the short-term (2 years or less) variation in alkalinity on average was about 41 peq/L (root-mean-square deviation of 53 and 67 peq/L), which is slightly less than the precision or the profile variability. The column of Table 7.5 labeled "pHC/CO2 Acidity" shows the root-mean-square variations of alkalinity determined from calorimetric pH and free Cot measure- ments. The average differences for these estimates (91, 92, and 70 peq/L) are larger than the corresponding differences for MO alkalinity, probably owing to the combination of free CO2 and colorimetric pH errors in addition to the sampling and handling errors. It is also worth noting that the larger variations are due in part to three poor replicate measurements. If these data are ignored, the average replication for the second alkalinity estimate is quite similar to the values obtained by the MO technique. In our analysis of the three sets of lakes in Wisconsin, New York, and New Hampshire, we consider the

274 data to be consistent if the two estimates of alkalinity agree within 50 peq/L. The tolerance of 50 peq/L is the rounded equivalent of 53 peq/L, the variation in alkalinity for duplicate profile averages of samples obtained on the same date. This value reflects, on average, the cumulative differences that are due to sampling and analysis errors and variability within the lake being surveyed. A similar analysis of replicated data for Wisconsin indicates that the relative errors of the types presented in Table 7.S are similar or less. ThuS the same criterion of 50 peq/L was used for Wisconsin data. There were no replication studies available for historical New York lakes data. The same criterion of 50 peq/L was used in screening New York data, assuming that the same analytical techniques were employed as for the New Hampshire survey. When the criteria for consistent data are applied to the historical data sets assuming a MO endpoint of 4.04, 68 percent of the New Hampshire data and 39 percent of the New York data are consistent. If a MO endpoint of 4.19 is assumed, 48 percent of the New Hampshire data, 52 percent of the New York data, and 90 percent of the Wisconsin data are consistent. The method of consistency checking for alkalinity and pH is not without potential problems. First, the basic assumptions are that only carbonate protolytes are important and that the total carbonate remains unchanged during measurement of the calorimetric pH and free Cot . Although the alkalinity will not change as a function of total CO2, pH and acidity will change. The magnitude of possible effects on alkalinity that are due to other protolytes is discussed in Appendix D. Inspection of Egs. (7) and (10) to (12) of Appendix D shows that the relationship between alkalinity and pHc/[Acy] is not linear. To assess the probable error in the second alkalinity estimation, various estimates of alkalinity were made using values of precision about given values Of PHC and [Acyl. Precisions for pH and free CO2 were taken as +0.2 and +5 peq/L, respectively. Figure 7.12 summarizes the results of these calculations. The average probable deviation for alkalinity was determined from the combinations of ranges Of PHC and free CO2 as defined by the previsions. This potential error is compared with [Acy] for various calorimetric pH values; the actual alkalinity for the stated PHC and [Acy] is shown in parentheses. The worst-case error of alkalinity (when both PHC and [Acy] deviations give a maximum error) would be about five times that shown. . -

275 40 J ~ 30 ~ _ 20 10 ~ I - ' I 1/ / 26; 4/ 84) OL ~ ,'~ o t (1207) ·~; ,~ / (/ ,~o,J 7435) pH = 6.5 PHC = 6.0, 5.5 (169) - (-73, 33) CO2-ACIDITY (,ueq/L) FIGURE 7.12 Average probable deviation to be expected for alkalinity calculated from free CO2 acidity and calorimetric pH (pHc). The deviation is obtained by varying PHC by +0.2 unit and CO2 acidity by +5 peq/L about assumed values. Numbers in parentheses next to the data points are actual calculated alkalinities. 300

276 The plot shows that the error increases with increasing alkalinity; for the kinds of lakes discussed here, the probable error in the calculation would be about 20 peq/L, equal to the analytical precision for the MO alkalinity titration. However, if the errors are cumulative, the probable error would be about 100 peq/L; this may explain the few large variations for replicate New Hampshire samples described earlier. Another potentially significant uncertainty in the analysis arises from the use of calorimetric indicators near the limits of their ranges. The change from one indicator (bromthymol blue) to another indicator (chloro- phenol red) occurs at a PHC Of 6. If bromthymol blue is used at a pH of 6 or slightly less, the potential error for the alkalinity is enlarged compared with that possible when chlorophenol red is used. It is probable that most of the historic calorimetric pH measurements near a pH of 6 were made with bromthymol blue (Juday et al. 1935). In addition, if bromthymol blue were used below its lower pH limit on occasion, the effect would be to increase the historical values of pH and alkalinity calculated from pHc/[Acy] . After the corrected pH was calculated by using Eqs. (10) to (12) in Appendix D, historical pH data were considered consistent if values agreed within +0.2 pH unit for pH < 5 and +0.5 pH unit for pH > 5. These constraints were set following the arguments by Blakar and Digernes (1984). These authors showed that a precision of +0.2 pH unit or less is achievable by using calorimetric techniques. In this regard, it is worth noting that the probable error for adjustment of pH, using the same approach as discussed earlier for alkalinity, is less than 0.2 unit and is 0.5 unit in the worst case. Thus the data that pass the consistency check of 50 peq/L for alkalinity should pass the pH consistency check. Recent data for New Hampshire, New York, and Wisconsin could not be screened for consistency because of the lack of replicate data or redundant protolyte measurements. The New Hampshire alkalinity data were adjusted for a fixed pH endpoint of 4.5, as discussed earlier. Since the recent Wisconsin data also included major ions, the data were assessed for ion balance (Eilers et al. 1985). Baker and Harvey (1985) have compared the findings of two surveys of New York lakes (Pfeiffer and Festa 1980, Colquhoun et al. 1984). Noting that the pH measurements in the 1980 report were consistently low, they suggest

277 that these measurements are in error, a finding also proposed by Schofield (1982). There are no means of confirming this hypothesis since there were no quality- assurance programs in any of the recent surveys. Rather than exclude the 1980 survey, we compared both sets of recent data with historical data. However, from this example it appears that in some cases recent electro- metrically derived data may also exhibit variance in sample replication. Trends in Alkalinity and pH of Lakes in New York, New Hampshire, and Wisconsin To determine changes in the pH and alkalinity of lakes over the past 50 to 60 years, we corrected and screened the historical data as described earlier and compared them with comparable data for the 1970s and 1980s. The results are shown in Tables 7.6, 7.7, and 7.8 for lakes in New York, New Hampshire, and Wisconsin, respectively. In all cases negative values signify trends toward increasing acidification, and positive values signify trends toward increasing alkalization. To avoid excessive weighing of large outliers, the median and the 10th and Both precentiles were calculated rather than the mean and the standard deviation. In compiling the historical data, whenever possible we obtained the alkalinity averaged from the MO titration and from the calorimetric pH and free CO2 acidity measurements. We determined the pH from the average of the three calculated values of pH. Because of uncertainty in knowing the precise pH employed for the MO endpoint in the historical studies of New York and New Hampshire lakes, we have analyzed the data assuming two different pH values, 4.04 and 4.19. The lower value corresponds to the findings of Kolthoff and Stenger (1947) and Kramer and Tessier (1982). The value of 4.19 is supported by Taylor (1933) and Eilers et al. (1985). Since the latter two studies refer specifically to the Wisconsin data set, we assumed that the pa of 4.19 was the best estimate to use in the compilation of historical values of pH and alkalinity in Wisconsin lakes. As is shown in Table 7.6 and Figures 7.13(a) and 7.13(b) for New York lakes, the choice of the MO endpoint may influence substantially the calculated changes in pH and alkalinity. Also, the results appear to be influenced

278 TABLE 7.6 Summary of Alkalinity and pH Changes for the Adirondack Region of New York ALKALINITY median 10% 90% pH (MO) n (~eq/L) (~eq/L) (~eq/L) 4.04 1984 report 97 (248) 1 (8) - 121 ( - 233) 166 (165) 1980 report 97 (248) - 28 I- 14) - 141 ~ -244) 165 (140) 4.19 1984 report 130 (248) - 44 ~ - 69) - 143 ~ - 377) 60 (61) 1980 report 130 (248) - 69 ~ - 89) - 187 ~ - 389) 41 (41) pH median 10% 90% pH (MO) n (units) (units) (units) 4.04 1984 report 97 (248) -0.12 (0.10) -1.04 (- 1.05) 1.18 (1.32) 1980 report 97 (248) -0.14 (0.02) -1.60 (-1.21) 1.18 (1.11) 4.19 1984 report 130 (248) - 0.63 ~ - 0.55) - 1.66 ~ - 1.66) 0.47 (0.53) 1980 report 130 (248) - 0.74 ~ - 0.66) - 1.86 ~ - 1.90) 0.04 (0.39) NOTE: Values of the median, 10th, and 90th percentiles are given. Changes are defined as the recent minus the historical values. Changes are given for the consistent data, all the data (in parentheses), and relative to the 1980 and 1984 New York Department of Environmental Conservation reports (i.e., Pfeiffer and Festa 1980 and Colquhoun et al. 19841. The data are further compared for calculations assuming MO alkalinity endpoints of 4.04 and 4.19. by the choice of the recent data set. When an endpoint pH of 4.04 is assumed and the historical data are com- pared with the 1984 data set, the median value for the change in alkalinity is +1 peq/L, and the median value for the change in pH is -0.12 unit. Given the uncer- tainties in these analyses, the results are indistinguish- able from zero. When the historical data are compared with the 1980 data set however, the calculated median change in alkalinity, -28 peq/L, may be significant, although the median change in pH, -0.14 unit, is not. When the same data are calculated using 4.19 as the MO endpoint, there appear to be distinct trends toward decreasing alkalinity (-44 and -69 peq/L for the 1984 and 1980 data, respectively) and decreasing pH (-0.63 and

279 TABLE 7.7 Summary of Alkalinity and pH Changes for Lakes in New Hampshire. ALKALINITY median 10% pH (MO) na (~eq/L) (~eq/L) 4.04 115 (170) 8 (8) 4.19 81 (170) - 7 ~ - 21) 90% (~eq/L) 75 (80) 87 (90) -47 (-59) - 73 ~ - 157) pH median pH (MO) n (units) 4.04 114 (169) 0.43 (0.44) 4.19 81 (170) 0.25 (0.20) 10% (units) 90% (units) - 0.29 ( - 0.23) 1.05 (1.05) - 0.84 ( - 0.46) 0.87 (0.90) NOTE: Values of the median, 10th, and 90th percentiles are given for consistent data and all the data (in parentheses). Data are calculated for MO pH endpoints of 4.04 and 4.19. Differences are expressed as the recent value minus the historic value. aIn the case of New Hampshire, n refers to the number of samples, including duplicate samples for the same lake. The total number of lakes in the data set is 113. -0.74 for the 1984 and 1980 data, respectively), regardless of the choice of recent data set. To test whether liming activities may have biased our results, we examined the liming histories of New York lakes (data were provided through the courtesy of F. Vertucci, Cornell University, personal communication 1985). Our complete data set of 248 lakes included 16 limed lakes. Our consistent data set contained three limed lakes. These, however, had a negligible effect (about 1 peq/L) on our analysis. Changes in New Hampshire lakes do not appear to be so sensitive to the assumption of the MO endpoint (Table 7. 7) . If the 4.04 endpoint is assumed, the median value for the change in alkalinity over the 50 year time period in these lakes is +8 peq/L, and the median for change in pH is +0.43 unit. If the higher endpoint is applied, median values of -7 peq/L and +0.25 unit are predicted. Thus, although there are numerous cases of both increases and decreases in alkalinity and pH, no significant change in alkalinity is apparent. However, the pH on average may have increased. The pH and alkalinity in the Wisconsin lakes apparently have increased significantly since the 1920s and 1930s. Table 7.8 and Figure 7.14 (derived from an MO endpoint pH

280 TABLE 7.8 Summary of Alkalinity and pH Changes for Lakes in Wisconsin. ALKALINITY median 10% 90% n (~eq/L) (~eq/L) (~eq/L) _ 145 38 -50 110 pH median 10% 90% n (units) (units) (units) . 145 0.51 0.01 1.12 NOTE: Values of the median, 10th, and 90th percentiles are given for consistent data. Data are calculated for a MO pH endpoint of 4.19 only. Differences are expressed as the recent value minus the historic value. Because data for 90% of the lalces passed the consistency check, calculations for all the data are not included. of 4.19) indicate that the median of the change in alkalinity is +38 peq/L and the median of change in pa is +0.51 unit. If an MO endpoint of 4.04 were applied, the calculated increases in alkalinity and pH would be even greater. However, an endpoint of 4.19 is the most likely value in this analysis, based on documentation provided by Taylor (1933) and the findings of Eilers et al. (1985). Since a large percentage of the data was eliminated by our consistency checks in New York and New Hampshire, we tested whether use of the consistent data alone introduced any bias. Employing a nonparametric chi-square statis- tical test, we concluded that there is no evidence of such bias. The major effect of using consistent data only is demonstrated for alkalinity changes in New York lakes, assuming an MO endpoint value of 4.19 (see Figure 7.15). Calculations based on consistent data resulted in the elimination of most of the lakes with alkalinity changes greater in magnitude than 200 peq/L. The median change in alkalinity increased from -69 peq/L based on compilation of all the lakes to -44 peq/L based on consistent data only. This difference reflects the fact that most of the lakes eliminated were those with high negative values for changes in alkalinity. Nevertheless, a few larger outliers remain, even for the consistent set of data.

281 40 cn ye 30 At: 11 O 20 m z 10 40 (n 30 by J 0 20 m z pH (MO) = 4.04 n = 97 10% 0 1 ~ ~ ~ ;;-; -; 4- ~ ;~ ;; ;;; ;;;5;~~ ~ ;~t ~ ~ ~ ~ ~ 1 -600 -400 -200 0 +200 +400 (a) ~ ALK (recent alkalinity minus historical alkalinity) pH (MO) = 4.19 n= 130 10 O L ~ 1 ~ . 109 ~ ,~ 90% I I (,ueq/L) -600 -400 -200 0 +200 +400 /\ ALK (recent alkalinity minus historical alkalinity) FIGURE 7.13 (a) Two scenarios for change in alkalinities of Adirondack lakes over five decades calculated from methyl orange endpoint pHs of 4.04 and 4.19. (b) Two scenarios for change in pHs of Adirondack lakes over five decades calculated from methyl orange endpoint pHs of 4.04 and 4.19. NOTE: Analyses are based on comparison of historical data with data from Colquhoun et al. (1984). Short-Term Variability The maximum changes in alkalinity of lakes that can result from atmospheric inputs of acid are estimated to be approximately 100 peq/L (Galloway 1984). Our analysis demonstrates that a number of lakes in the three regions of study have changes of this magnitude or higher. These lakes are listed in Table 7.9. We have not examined the factors responsible for these large changes, but such a study would be useful.

282 20 _ In UJ y a: J O 10 UJ m z y a: J 11 0 1n CC LU z FIGURE 7.13 (continued). pH (MO) = 4.04 n = 97 (b) 50% 90% an= I ~ .. ~ . ........ , , , ~ , , . . .] .... ~ pH (recent pH minus historical pH) pH (MO) = 4.19 n= 130 10% 20 ,..~..t,.,.~..—,...... O tN~@T I -2 -1 0 +1 90% +2 /\ pH (recent pH minus historical pH) Comparison of recent and historical data considers the possible changes in alkalinity of only two or three data points over about five decades. In assessing these results and the magnitude of changes, one must consider the magnitude of changes from short-term (seasonal) fluctuations and how such changes compare with the changes considered over five decades. This subject was addressed to some degree in the earlier discussion of the replication of historical data for New Hampshire. Average variation in the analyses of duplicate lake profiles collected at different times over short time periods (up to 2 years) was compared with variation in duplicate lake profiles collected on the same date. The conclusion was that the short-term variation was about 41 peq/L on average compared to the variation in duplicate profile averages of about 50 peq/L. Intensively measured, low-alkalinity lakes may also provide an analysis of the magnitude of short-term

283 60 LL y 40 6 o m 20 o 1 60~ en Y 40 of: J o m ~ 20 He o pH ( MO) = 4.19 n= 145 ................. ................ , ., ., ., .,: .: .: ,............... ,,, ., ., . ...: .: ,, .,,,::,... .. .. ................. . . .. , . ...... ·. .. ..... 10% 50% : :> : i in : , . . . ... .. ... ............. 281: ~~ ~ S ~ ~ if.: :.:,: - - - - . -,:::.: :,:,:,:.:,:,:,: ,: :,:. . -: . - . . .:- -:. ::-: :,: :-::: :-:: :.: :-: :-:.: :,:.:-:-::.: ::,:.: :.: :. . - . a. - -:.:.:.: :-:-,.: ,-: --:-:: :-: :-:: :-::. :-:.:: :::-.:: ::.: ::-:::::.::::::.:::::.::.:.:-:::::-: -.:-:: :-:::$-: :.:.:.:.:-:.: I ~ ~~ ~~ I (,ueq/L) So% -200 -100 0 +100 +200 ~ ALK (recent alkalinity minus historical alkalinity) pH (MO) = 4.19 n= 145 1 1 10% -1.0 -0.5 0 +0.5 +1.0 +1.5 in pH (recent pH minus historical pH) FIGURE 7.14 Changes in alkalinity and pH in Wisconsin lakes from the 1920s and 1930s to the l980s, assuming a MO endpoint pH of 4.19. changes. The Vermont Department of Water Resources and Environmental Engineering (1983) has been carrying out an intensive program of monitoring lakes since 1982. Lakes ranging in alkalinity from less than O to greater than 100 peq/L were chosen for assessment. All data used were tested for consistency by calculating the departure from electroneutrality and comparing calculated and measured specific conductance; deviations of less than 15 percent were accepted. Monthly trends for alkalinity for each lake are shown in Figure 7.16 along with the means and standard deviations. The deviations in alkalinity are quite small for all lakes. The largest standard

284 40 30 O 20 m z 10 o 50 40 o m 20 z pH (MO) = 4.19 n= 130 consistent data set A ..... ~ l I . I (peq/L) 200 0 = -44 ~eq/L -600 -400 to ALK (recent alkalinity minus historical alkalinity) 50% = -69 peq/L +200 +400 n pH (MO) = 4.19 n= 248 all the data v v -600 -400 -200 0 +200 +400 /& ALK (recent alkalinity minus historical alkalinity) (,ueq/L) FIGURE 7.15 Plots of change in alkalinities of Adirondack lakes calculated with consistent data only and with all of the data, assuming an endpoint of 4.19. Recent data are from Colquhoun et al. (1984). deviation of 52 peq/L is about the same as the variation obtained in the historical duplicate profile analyses of New Hampshire lakes. Thus we may conclude that short-term variations in alkalinity may be quite small and are often of less than the tolerance permitted for consistent historical data. This conclusion would not deny the occurrence in individual lakes of large, episodic variations in alkalinity of perhaps 200 peq/L or greater (Driscoll and Newton 1985). Because changes in alkalinity caused by acid deposition may be of the same magnitude as the analytical error in our analysis of these lakes, the question arises whether any effects of acid deposition can be detected or rejected using the historical data from lakes in the 1920s and 1930s. However, the sample size and standard deviation

285 TABLE 7.9 Lakes That Have Experienced Large Changes in Alkalinities Since the 1930s Lake Increase ~ + ~ / Decrease ~—~ New Hampshire: Opechee Bay, Laconia Potanipo Pond, Brookline Swain's Lake, Barrington Beaver Lake, Derry Canobie Lake, Windham - French Pond, Henniker Great East Lake, Wakefield Lovell Lake, Wakefield Rocky Bound Pond, Croydon Wash Pond, Hampstead Wheelwright Pond, Lee New York: Ballston Lake (7-1090) Irving Lake (7-0732) Pine Lake (7-0724) Spruce Lake (7-0706A) Elm Lake (5-0304) Lilypad Pond (6-0025) Silver Lake (3-03551) Sterling Pond (6-0029) Valentine Pond (5-0370) Wisconsin: Dog Lake Goodyear Lake Virgin Lake Big Crooked Lake Booth Lake Cranberry Lake Diamond Lake Frank Lake Garth Lake Katherine Lake Little Arbor Vitae Lake Little Muskie Lake Little Portage Lake Mill Lake Silver Lake Spirit Lake Squash Lake Sumach Lake Yawley Lake + + + + + + + + + + + + + + NOTE: The magnitudes of changes are greater than 100,ueq/L for New Hampshire and Wisconsin and 200 ,ueq/L for New York.

286 200 150 _ _ 100 ~ ~ 40 my 20 _ .~ -20 _ 1 1 1 · Beebe P n=6, O Bourne P n=7, · Branch P n=8, O Hardwood P n=6, · Kettle P n=7, ~ Osmore P n=5, * Stratton P n=7, ~ X Wheller P n=5, ~ \ \ 1/ 1 1 1 1 1 1 1 1 1 0~ in In/ \: ~ \ 01 03 1 1 1 05 07 09 11 07 09 11 1 982 1983 DATE (months) . FIGURE 7.16 Monthly changes in alkalinity calculated from quality-assured data for lakes in Vermont. Mean S.D. _5.9 + 8.8 -1.0 + 9.9 -11.5 + 8.8 28.6 + 5.4 Mean, S.D. 92.5 + 19.9 1 38.0 + 52.0 22.2 + 8.4 100.0 + 22.1 Means+ 1a

287 of the alkalinity changes are such that with high probability (greater than 0.9), we would be able to detect a regional decrease of 30 to 50 peq/L. SUMMARY Relationship Between Sulfate Input Flux to Lakes from Wet Deposition and Sulfate Output Flux from Lakes In a large area of northeastern North America the mathematical relationship between fluxes of sulfate to lakes from wet deposition and fluxes of sulfate output from lakes is significantly linear when the data are examined on the basis of smaller contiguous regions (southern New England, New York, northern New England, Quebec, Newfoundland, and Labrador). Regional differ- ences arise because of different multiplying factors in the regression analysis. One physical interpretation of this result is that the multiplying factor is proportional to the ratio of total atmospheric sulfur deposition (i.e., wet plus dry deposition) to wet-only sulfur deposition. Thus the aggregated data appear to exhibit nonlinearity because of the varying importance of dry deposition from region to region. In some cases internal sources of sulfur in the watershed may also contribute to the flux of sulfur into the lake. Some likely examples are cited. Bench-Mark Streams Trends in sulfate are discernible in Bench-Mark streams over a period extending from the mid-1960s to 1983. On a regional basis, those trends in stream sulfate are con- sistent with trends in SO2 emissions. Analysis of sulfate-mass balance in Bench-Mark water- sheds of the southeastern United States, which are dominated by Ultisols, show a net retention of sulfate. Similar analyses for sites in the Northeast suggest that there is an excess of sulfate yield over wet-deposition sulfur input, but the contribution of sulfate to the excess from internal sources and from dry deposition cannot be determined with the current data. In general in the northeastern United States (Region B) since the mid-1960s the Bench-Mark streams have expert enced no change or decreases in sulfate concentrations and no change or increases in alkalinity. In the -

288 Southeast (Region C), streams have shown increases in sulfate concentrations and have either shown no changes or have decreased in alkalinity. In Bench-Mark streams with alkalinities less than 500 peq/L, short-term changes in the flux of strong-acid anions (SO4 , Now) are balanced by changes in alkalinity and nonprotolytic (base) cations. For these softwater Bench-Mark streams, short-term alkalinity changes balance from 0 to 78 percent of the sulfate change, averaging about 30 percent. This result supports the contention that changes in strong-acid anion fluxes affect surface water alkalinity in watersheds that have acid soils. Alkalinity and nonprotolytic cation changes in low- alkalinity Bench-Mark streams over the 15- to 20-year period of record are consistent with changes caused by a changing flux of strong-acid anions from the atmosphere, but there is also evidence suggesting that internal watershed processes or in-stream processes govern trends in nonprotolytic cations and alkalinity. At many eastern low-alkalinity stations changes in strong-acid anions are sufficient to account for alkalinity and nonprotolytic cation changes, but at some western stations that receive low rates of SO4 deposition substantial reductions in alkalinity have occurred that have a large component not accounted for by changes in strong-acid anions. In these cases the alkalinity reductions are accompanied by reductions in nonprotolytic cation concentrations and may result primarily from changes in internal-terrestrial or in-stream processes. Lakes in Wisconsin, New York, and New Hampshire Data from historical surveys of lake water parameters in Wisconsin, New York, and New Hampshire are of varying quality, but sufficient data and documentation are avail- able to test for internal consistency. Applying the criteria of Kramer and Tessier (1982), we obtained consistent sets of pH and alkalinity data for approxi- mately 300 lakes for which recent data also exist. Comparing historical with recent data, it is possible to estimate pH and alkalinity changes over the past half-century. This method is not without problems. Use of the MO indicator in the historical studies called for titration to the "faintest pink" endpoint. Estimates in the

289 literature of the pH of this endpoint range between about 4.0 to higher than 4.3. The color of the endpoint in this range has been variously described as "faint pink," "pink," and "salmon pink." The Wisconsin survey provides the best documentation of the pH of the MO endpoint, citing a range of pH values from 4.06 to 4.18 for the "fainter" pink color and recommending that a value of 4.18 be used in calculating titrations. Our analysis shows that changes in alkalinity and pH in New York lakes are sensitive to the assumption made about the pH of the MO endpoint. Assuming an endpoint pH of 4.2 (or greater), New York lakes, on average, appear to have increased substantially in acidic status over the past 50 years, as reflected in reductions in both alkalinity and pH. Alternatively, if a value of the pH endpoint close to 4.0 is assumed and if the historical data are compared with the 1984 data set, then there appears to have been little change, on average, in the acidic status of New York lakes over the past 50 years. However, if the historical data are compared with the 1980 New York data set, there may have been an overall decline in alkalinity. New Hampshire lakes do not appear to be so sensitive to the choice of tne endpoint pH of MO. If an endpoint pH close to 4.2 is assumed, these lakes, on average, show little or no decrease in alkalinity and show a slight increase in pH. With an endpoint near 4.0 there is again little or no change, on average, in alkalinity, but an increase in pH of about 0.4 pH unit. Wisconsin lakes, on average, appear to show a significant increase in alkalinity (38 peq/L) and pH (+0.5 unit) if an MO endpoint of 4.2 is assumed. The increases are even larger if an endpoint close to 4.0 is assigned. Currently, the question regarding the correct endpoint pH in historical alkalinity titrations is unresolved. In the judgment of the authors of this chapter the endpoint lies in the range from 4.19 to 4.04, but evidence is not sufficient to specify a most likely value. Tests done with indicator dyes which have been preserved for the past 50 years may be valuable in resolving this question. We have not attempted to indicate the mechanisms for change in acidic status of these lakes. Any discussion of mechanisms must consider each lake separately and include all the hydrological and biogeochemical factor For most lakes, data are not available currently to permit this kind of analysis.

290 One possible avenue of future research is to conduct a detailed study of the outliers in the three data sets. If the maximum change in lake alkalinity from deposition of atmospheric sulfur is about 100 peq/L, then study of the larger changes subjected to the most careful scrutiny for quality may offer a rational explanation for other acidification mechanisms. REFERENCES Almer, B., W. Dickson, C. Eckstrom, and E. Hornstrom. 1978. Sulfur pollution and the aquatic ecosystems. Pp. 273-311 of Sulfur in the Environment, Part Il. Ecological Impacts. J. O. Nriagu, ed. New York: J. Wiley and Sons. Arnold, D. E., R. W. Light, and V. J. Dymond. 1980. Probable effects of acid precipitation on Pennsylvania waters. U.S. Environmental Protection Agency. EPA-600/3-80-012. 21 pp. Baker, J., and T. Harvey. 1985. Critique of acid lakes and fish population status in the Adirondack region of New York State. Draft final report for NAPAP Project E3-25, U.S. Environmental Protection Agency. Beamish, R. J., and H. H. Harvey. 1972. Acidification of the La Cloche Mountain Lakes, Ontario and resulting fish mortalities. J. Fish. Res. Board Can. 29:1131-1143. Beamish, R. J., W. L. Lockhart, J. C. van Loon, and H. H. Harvey. 1975. Long-term acidification of a lake and resulting effects of fishes. Ambio 4:98-102. Blakar, I. A., and I. Digernes. 1984. Evaluation of acidification based on former calorimetric determination of pH: The effect of indicators on pH in poorly buffered water. Verh. Int. Ver. Limnol. 22:679-685. Booty, W. G., and J. R. Kramer. 1984. Sensitivity and analysis of a watershed acidification model, Phil. Trans. R. Soc. London Ser. B 305:441-449. Burns, D. A., J. N. Galloway, and G. R. Hendry. 1981. Acidification of surface waters in two areas of the eastern United States. Water Air Soil Pollut. 16: 277-285. Chen, C. W., J. D. Dean, S. A. Gherini, and R. A. Goldstein. 1982. Acid rain model: hydrologic module. J. Env. Eng. Div. Am. Soc. Civ. Eng. 108:455-472.

291 Christophersen, N., and R. F. Wright. 1981. Sulfate budget and a model for sulfate concentrations in stream water at Birkenes, a small forested catchment in southernmost Norway. Wat. Resour. Res. 17:377-389. Church, M. F. 1984. Predictive modeling of acidic deposition on surface waters. Pp. 4-113--4-128 of The Acidic Deposition Phenomenon and Its Effects. Critical Assessment Review Papers. Vol. II. Effects Sciences. A. P. Altshuller and R. A. Linthurst, eds. EPA-600/8-83-016 BE. U.S. Environmental Protection Agency. Clark, W. M. 1928. The Determination of Hydrogen Ions. Baltimore, Maryland: The Williams and Wilkins Company. Cobb, E. D., and J. E. Biesecker. 1971. The national hydrologic Bench-Mark network. Circular 460-D, U.S. Geological Survey. 38 pp. Colquhoun, J., W. Kreter, and M. Pfeiffer. 1984. Acidity states of lakes and streams in New York State. New York Department of Environmental Conservation, Raybrook, New York. 49 pp + viii app. Crisman, T. L., R. L. Schulze, P. L. Brezonik, and S. A. Bloom. 1980. Acid precipitation: the biotic response in Florida lakes. Pp. 296-297 of Ecological Impact of Acid Precipitation. D. Drably, and A. Tollan, eds. Proceedings of an international conference, Sandefjord, Norway. Sur Nedb~rs Virkning Pa Skog Og Fisk (SNSF) Project, Oslo. Critical Assessment Review Papers (CARP). 1984. The acidic deposition phenomenon and its effects. A. P. Altschuller and R. A. Linthurst, eds. Report EPA-600/8-83-016AF. U.S. Environmental Protection Agency. Cronan, C. S., and C. L. Schofield. 1979. Aluminum leaching response to acid precipitation: effects on high-elevation watersheds in the Northeast. Science 204(20):304-306. Davis, R. B., M. O. Smith, J. H. Bailey, and S. A. Norton. 1978. Acidification of Maine (U.S.A.) lakes by acidic precipitation. Verh. Int. Ver. Limnol. 10:532-537. Dickson, W. 1980. Properties of acidified waters. Pp. 75-83 of Proceedings of an International Conference. Ecological Impact of Acid Precipitation, D. Drabl<S and A. Tollan, eds. Sur Nedb~rs Virkning Pa Skog Og Fisk (SNSF) Project, Oslo, Norway.

292 Dillon, P. J. 1983. Chemical alteration of surface water by acidic deposition in Canada. Pp. 275-286 of Ecological Effects of Acid Deposition. Report P. 1636 National Swedish Environmental Protection Board. Dillon, P. J., D. S. Jeffries, W. Snyder, R. Reid, N. D. Yan, D. Evans, J. Moss, and W. A. Scheider. 1978. Acid precipitation in south-central Ontario: recent observations. J. Fish. Res. Board Can. 35:809-815. Drabl~s, D., D. I. Sevaldrud, and J. A. Timberlid. 1980. Historical land-use changes related to fish status development in different areas in southern Norway. Pp. 367-369 of Proceedings of an International Conference. Ecological Impact of Acid Precipitation, D. Drably and A. Tollan, eds. Sur Nedb~rs Virkning Pa Skog Og Fisk (SNSF) Project, Oslo, Norway. Driscoll, C. T., and G. E. Likens. 1982. Hydrogen ion budget of an aggrading forested ecosystem. Tellus 34:283-292. Driscoll, C. T., and R. M. Newton. 1985. Chemical characteristics of Adirondack lakes. Environ. Sci. Technol. 19:1018-1024. Eilers, J. M., G. E. Glass, and A. K. Pollack. 1985. Water quality changes in northern Wisconsin lakes, 1930-1980. RL-D Draft 539. U.S. Environmental Protection Agency. Environmental Protection Agency. 1984. The Acidic Deposition Phenomenon and Its Effects. In Critical Assessment Review Papers, Vol. II. Effects Sciences. A. P. Altshuller and R. A. Linthurst, eds. EPA-600/8-83-016 BE. Fitzgerald, J. W., T. C. Strickland, and W. T. Swank. 1982. Metabolic fate of inorganic sulfate in soil samples from undisturbed and managed forest ecosystems. Soil Biol. Biochem. 14:529-536. Galloway, J. N. 1984. Long-term acidification. In The Acidic Deposition Phenomenon and Its Effects. Critical Assessment Review Papers. Vol. II. Effects Sciences. A. P. Altshuller and R. A. Linthurst, eds. EPA-600/8-83-016 BF. Galloway, J. M., S. A. Norton and M. R. Church. 1983a. Freshwater acidification from atmospheric deposition of sulfuric acid: A conceptual model. Environ. Sci. Technol. 17:541A-545A. Galloway, J. N., C. L. Schofield, N. E. Peters, G. R. Hendrey, and E. R. Altwicker. 1983b. The effect of atmospheric sulfur on the composition of three Adirondack lakes. Can. J. Fish. Aquat. Sci. 40:799-806.

293 Gran, G. 1952. Determination of equivalence point in potentiometric titrations. Analyst 77:661-671. Haines, T. A., J. J. Akielagzek, and P. J. Rago. 1983. A regional survey of chemistry of headwater lakes and streams in New England: vulnerability to acidification. Air Pollution and Acid Rain Series. U.S. Fish and Wildlife Service Report. Hendrey, G. R., J. N. Galloway, S. A. Norton, C. L. Schofield, P. W. Schaffer, and D. A. Burns. 1980. Geological and hydrochemical sensitivity of the eastern United States to acid precipitation. ~ EPA-60013-80-024. U.S. Environmental Protection Agency. 100 pp. Henriksen, A. 1979. A simple approach for identifying and measuring acidification of freshwater. Nature 278:542-545. Henriksen, A. 1980. Acidification of freshwaters--a large scale titration. Pp. 68-74 of Ecological Impact of Acid Precipitation. D. Drabl~s and A. Tollan, eds. Sur Nedb~rs Virkning Pa Skog Og Fisk (SNSF) Project, Oslo. Henriksen, A. 1982. Changes in Base Cation Concentrations Due to Freshwater Acidification. Acid Rain Research Report 1/1982. Norwegian Institute for Water Research (NIVA), Oslo. Hirsch, R. M., J. R. Slack, and R. A. Smith. 1982. Techniques of trend analysis for monthly water quality data. Water Resour. Res. 18:107-121. Hoover, E. E. 1937. Biological Survey of Androscoggin Saco and Coastal Watersheds. New Hampshire Fish and Game Commission, Concord. Hoover, E. E. 1938. Biological Survey of the Merrimack Watershed. New Hampshire Fish and Game Commission, Concord. Johnson, A. H. 1979. Evidence of acidification of headwater streams in the New Jersey pinelands. Science 206:834-836. Johnson, D. W. 1980. Site susceptibility to leaching by H2SO4 in acid rainfall. Pp. 525-536 of Effects of Acid Precipitation on Terrestrial Ecosystems. T. C. Hutchinson and M. Havas, eds. New York: Plenum Press. Johnson, D. W., and G. S. Henderson. 1979. Sulfate absorption and sulfur fractions in a highly weathered soil under mixed deciduous forest. Soil Sci. 128:34-40.

294 Juday, J., E. A. Birge, and V. W. Meloche. 1935. The carbon dioxide and hydrogen content of the lake waters of northeastern Wisconsin. Trans. Wis. Acad. Sci. Arts Letters 29:1-82. Kolthoff, I. M. 1931. Pp. 30-31, 52-53 of The Colorimetric and Potentiometric Determination of pH. New York: John Wiley and Sons. Kolthoff, I. M., and V. A. Stenger. 1947. Titration methods. Pp. 61-62 of Volumetric Analysis, Vol. II. New York: John Wiley and Sons. Kramer, J. R., and A. Tessier. 1982. Acidification of aquatic systems: a critique of chemical approaches. Environ. Sci. Technol. 16:606A-615A. Krug, E. C., and C. R. Frink. 1983. Acid rain on acid soil: a new perspective. Science 221:520-525. Lewis, W. M. 1982. Changes in pH and buffering capacity of lakes in the Colorado Rockies. Limnol. Oceanogr. 27:161-171. Likens, G. E., F. H. Bormann, R. S. Pierce, and N. M. Johnson. 1977. Biogeochemistry of a Forested Ecosystem. New York: Springer-Verlag. 146 pp. Likens, G. E., F. H. Bormann, R. S. Pierce, and J. S. Eaton. 1985. The Hubbard Brook Valley. Chapter II of An Ecosystem Approach to Aquatic Ecology: Mirror Lake and Its Environment, G. E. Likens, ed. New York: Springer-Verlag. Mohn, E., E. Joranger, S. Kalvenes, B. Sollie, and R. Wright. 1980. Regional surveys of the chemistry of small Norwegian lakes: a statistical analysis of the data from 1974-1978. Pp. 240-241 of Ecological Impact of Acid Precipitation, D. Disables and A. Tollens, eds. Sur Nedb~rs Virkning Pa Skog Og Fisk (SNSF) project. Morrison, I. K. 1984. Acid rain: a review of literature on acid deposition effects in forest ecosystems. For Abstr. 4S:483-506. National Research Council. 1984. Acid Deposition: Processes of Lake Acidification. Washington, D.C.: National Academy Press. Newell, A. E. 1970. Biological survey of the lakes and ponds in Cheshire, Hillsborough, and Rockingham Counties. New Hampshire Fish and Game Commission, Concord. Newell, A. E. 1972. Biological survey of the lakes and ponds in Coos Grafton and Car rol Counties. New Hampshire Fish and Game Commission, Concord. .

295 Newell, A. E. 1977. Biological survey of the lakes and ponds in Sullivan, Merrimack, Belknap and Strafford Counties. New Hampshire Fish and Game Commission, Concord. New York Department of Conservation. 1930. Biological Survey of the St. Lawrence Watershed, Vol. V. Albany, New York. New York Department of Conservation. 1931. Biological Survey of the Oswegatchie and Black River Systems, Vol. VI. Albany, New York. New York Department of Conservation. 1932. Biological Survey of the Upper Hudson Watershed, Vol VII. Albany, New York. New York Department of Conservation. 1933. Biological Survey of the Raquette Watershed, Vol VIII. Albany, New York. New York Department of Conservation. 1934. Biological Survey of the Mohawk-Hudson Watershed, Vol. IX. Albany, New York. Norton, S. A., R. B. Davis, and D. F. Braekke. 1981. Responses of Northern New England Lakes to Atmospheric Inputs of Acids and Heavy Metals. Completion Report: Project A-048-ME. U.S. Department of the Interior, Office of Water Research and Technology. 90 pp. Oliver, B. G., E. M. Thurman, and R. L. Malcolm. 1983. The contribution of humic substances to the acidity of colored natural waters. Geochim. Cosmochim. Acta 47:2031-2035. Olson, R. J., D. W. Johnson, and D. S. Shriner. 1982. Regional assessment of potential sensitivity of soils in the eastern U.S. to acid precipitation. OaK Ridge National Laboratory, Environmental Sciences Division, Oak Ridge, Tenn. Publication Number 1899. 50 pp. Overrein, L. M., H. M. Seip, and A. Tollan. 1980. Acid Precipitation-Effects on Forests and Fish. Final report. Sur Nedb~rs Virkning Pa Skog Og Fisk (S~SF) project, 1972-1980. Oslo. 175 pp. Pfeiffer, M. H. and P. J. Festal 1980. Acidity status of lakes in the Adirondack region of New York in relationship to fish sources. New York Department of Environmental Conservation, Albany, N.Y. FW-P168(10/80). 36 pp.

296 Rebsdorf, A. 1980. Acidification of Danish softwater lakes. In Proceedings of the International Conference on Ecological Impact of Acid Precipitation, D. Drably and A. Tollan, eds. Sur Nedb~rs Virkning Pa Skog Og Fisk (SNSF) project, 1432 As, Norway. 383 pp. Reuss, J. O. 1985. Implications of the Ca-A1 exchange system for the effect of acid precipitation on soils J. Environ. Qual. 14:26-31. Richter, D. D., D. W. Johnson, and D. E. Todd. 1983. Atmospheric sulfur deposition, neutralization and ion leaching in two deciduous forest ecosystems. J. Environ. Qual. 12:263-269. Rosenquist, I. Th., P. Jorgensen, and H. Rueslatten. 1980. The importance of natural H production for acidity in soil and water. Proceedings of an International Conference. Pp. 240-241 of Ecological Impact of Acid Precipitation, D. Drabl~s and A. Tollan, eds. Sur Nedb~rs Virkning Pa Skog Og Fisk tSNSF) project, Oslo. Schnorr, J. L., G. R. Carmichael, and F. A. Van Schepen. 1982. An integrated approach to acid rainfall assessments. Chapter 13 of Vol. II. Energy and Environmental Chemistry: Acid Rain, L. H. Keith, ed. Ann Arbor, Michigan: Ann Arbor Science. Schofield, C. L. 1976. Lake acidification in the Adirondack mountains of New York: causes and consequences. In Proceedings of the First International Symposium on Acid Precipitation and the Forest Ecosystem, L. S. Dochinger and T. A. Seliga, eds. U.S. Department of Agriculture Forest Service Technical Report NE-2. Schofield, C. L. 1982. Historical fisheries changes as related to surface water pH changes in the United States. Pp. 57-68 of Acid Rain/Fisheries, R. E. Johnson (ed.). Bethesda, Md.: American Fisheries Society. Seip, H. M. 1980. Acidification of Freshwater--Sources and Mechanisms. Pp. 358-366 of Ecological Impacts of Acid Precipitation, D. Drably and A. Tollan, eds. Proceedings of an international congress in Sandefjord, Norway. Sur Nedb~rs Virkning Pa Skog Og Fisk (SNSF) Project, Oslo. Skougstad, M. W., M. J. Fishman, L. C. Friedman, D. E. Erdmann, and S. Duncan, Eds. 1979. Methods for Determination of Inorganic Substances in Water and Fluvial Sediments Techniques of Water-Resources Investigations of the U.S. Geological Survey. Book 5, Chapter A1. Washington, D.C.: U.S. Government Printing Office. .

297 Smith, R. A., and R. B. Alexander. 1983. Evidence for acid precipitation induced trends in stream chemistry at Hydrologic Bench-Mark stations. Circular 910. U.S. Geological Survey. 12 pp. Smith, R. A., R. M. Hirsch, and J. R. Slack. 1982. A study of trends in total phosphorus measurements at NASQAN stations. USGS Water-Supply Paper 2190. Standard Methods for the Examination of Water and Sewage. 1926. 4th ed. American Public Health Association, New York. Standard Methods for the Examination of Water and Sewage. 1930. 6th ed. American Public Health Association, New York. Standard 1933. York. Standard 1936. York. Standard 1960. York. Methods for the Examination of Water and Sewage. 7th ed. American Public Health Association, New Methods for the Examination of Water and Sewage. 8th ed. American Public Health Association, New Methods for the Examination of Water and Sewage. 11th ed. American Public Health Assocation, New Stung, W., and J. J. Morgan. 1981. Aquatic Chemistry. New York: John Wiley and Sons. Taylor, F. L. 1933. Letter of February 23, 1933, from F. Lowell Taylor to J. Juday, Department of Chemistry, University of Wisconsin. Archives of the University of Wisconsin, Madison. Thompson, M. E. 1982. The cation denudation rate as a quantitative index of sensitivity of eastern Canadian rivers to acidic atmospheric precipitation. Water Air Soil Pollut. 18:215-226. Thompson, M. E., and M. B. Hutton. 1982. Sulfate in lakes of eastern Canada: calculated atmospheric loads compared with measured wet deposition. National Water Research Institute, Environment Canada, Burlington, Ontario. Thompson, M. E., and M. B. Hutton. 1985. Sulfate in lakes of Eastern Canada: calculated yields compared with measured wet and dry deposition. Water Air Soil Pollut. 24:77-83. Thompson, M. E., F. C. Elder, A. R. Davis, and S. Whitlow. 1980. Evidence of acidification of rivers of Eastern Canada. Pp. 244-245 of Ecological Impact of Acid Precipitation, D. Drabl0`s and A. Tollan, eds. Proceedings of an international conference. Sur Nedbbrs Virkning Pa Skog Og Fisk (SNSF) project, Oslo.

298 Towne, R. E. 1983. Past and present pH and alkalinity of New Hampshire lakes and ponds, New Hampshire Water Supply Pollution Control Commission, Concord. Unpublished manuscript. 20 pp. Turk, J. T. 1983. An evaluation of trends in the acidity of precipitation and the related acidification of surface water in North America. Water Supply Paper 2249. U.S. Geological Survey. 18 pp. Ulrich, B. 1981. Okosystemare Hypothese uber die Ursachen des Tanneusterbens. Forstwiss. Centralbl. 100:228-238. Ulrich, B. 1983. Stabilitat van waldokosystemen unter dem einfluss des "Sauren regens. n Allg. Forstz. 26/27:670-677. Van Breeman, N., P. A. Burrough, F. J. Velthorst, H. F. van Dobben, T. deWitt, T. B. Ridder, and H. F. R. Reijuders. 1982. Soil acidification ammonium sulfate in forest canopy throughfall. Nature 299:548-550. Vermont Department of Water Resources and Environmental Engineering. 1983. Vermont acid precipitation program--long-term lake monitoring. Vermont Department of Water Resources and Environmental Engineering, Montpelier. Wagner, D. P., D. S. Fanning, J. E. Foss, M. S. Patterson, and P. A. Snow. 1982. Morphological and mineralogical features related to sulfide oxidation under natural and disturbed land surfaces in Maryland. Chapter 7 of Acid Sulfate Weathering, J. A. Kittrick, D. S. Fanning, and L. R. Hossner, eds. SSSA Special publication 10. Soil Society of America, Madison, Wisc Warfel, H. E. 1939. Biological survey of the Connecticut watershed. New Hampshire Fish and Game Commission, Concord. Watt, W. D., D. Scott, and S. Ray. 1979. Acidification and other chemical changes in Halifax County lakeS after 21 years. Limnol. Oceanogr. 24:1154-1161. Wright, R. F. 1983. Predicting Acidification of North American Lakes. Norwegian Institute for Water Research (NIVA) report 4/1983. Oslo, Norway. 165 pp. Wright, R. F. 1984. Regional pattern of sulfate concentrations in lakes in Eastern North America. Paper submitted to Committee on Monitoring and Assessment of Trends in Eastern North America, National Research Council. May 1984. .

299 Wright, R. F., and E. T. Gjessing. 1976. Acid precipitation: changes in the chemical composition of lakes. Ambio 5:219-223. Zimmerman, A. P., and H. H. Harvey. 1979. Sensitivity to acidification of waters of Ontario and neighboring states. Final report to Ontario Hydro. University of Toronto, Ontario. 118 pp.