Aquatic Chemistry of Lakes and Reservoirs in Taiwan

The composition of lake and reservoir waters in Taiwan is mainly con­ trolled by inputs from chemical weathering and by concentration through water evaporation. Inputs from seasalt aerosols are secondary. Sulfate in the water samples mostly originates from oxidation of sulfide minerals in source rocks, such as marbles, gneiss and s�hists, in the Central Range of Taiwan. Pco2 values estimated from pH and alkalinity data are mostly highe r than the atmospheric Pco2, indicating a net input of C02 from biological respiration over photosynthesis in lakes and reservoirs. Those water samples are also undersaturated with respect to calcite. Some coastal water samples with high alkalinity have unusually high pH (or low Pco2) and are satu­ rated or slightly oversaturated with respect to calcite, probably suggesting a net removal of C02 by active photosynthesis along with evaporative con­ centration of alkalinity in these water bodies. Some high mountain lakes are low in alkalinity and may be susceptible to acidification by acid rain. However, exceptionally high chemical weathering rates in the Central Range may mitigate the acid rain problem. Kaolinite is a stable secondary mineral in high mountain lakes (>2000 m), whereas Ca-smectite and/or kaolinite are stable secondary minerals in other water bodies. The chemical weath­ ering rates of the Taiwan Island are extremely high, but the chemical com­ positions of lakes and reservoirs in Taiwan are, in general, similar to the composition of the world mean river water. (


INTRODUCTION
The island of Taiwan has a length of 385 km, a maximum width of 143 km and a total area of 36,000 km 2 • The Central Range is the backbone of the island ( Figure 1) and has more than 25 mountain peaks with altitudes exceeding 3000 m. The highest mountain, Mt.Yueshan, is 3997 m above sea level. Taiwan is situated on the collision boundary between the Eurasian and the Philippine Sea plates and thus has beenvery active geotectonically since its formation in the late Paleozoic to Mesozoic. Age-dating of raised coral reefs and fossil shells (Konishi et al, 1968: Peng et al, 1977Liew et al, 1990;Wang and Burnett, 1990) and the measurement of the elevation changes of benchmarks by geodetic leveling (Liu and Yu, 1990) all indicate rapid uplift rates (mostly in the range of 2 to 10 mm/yr but sometime as high as 35 mm/yr in some places) of the island from the early Holocene to the present. The high uplift rate and high relief of the island result in the observed rapid physical and chemical weathering rates of the Central Range today (1300 mg/cm2 yr and 65 mg/cm 2 yr, respectively; Li, 1976).
The composition of river and lake waters often reflect the chemical weathering processes in the upstream drainage areas. Factors controlling the chemistry of river and lake waters are summarized by Drever (1985Drever ( ,1988, Stumm (1985), Meybeck ( 1987Meybeck ( a, 1987b and Lerman et al. (1995). Many case studies are also summarized by Degens (1982) and Degens etal. (1983Degens etal. ( , 1985. The case study of the Amazon River by Stallard and Edmond (1981and 1987) is one outstanding recent example. They demonstrated the importance of precipitation inputs, the influence of geology and the weathering environment as well as the thermodynamic con straints on the chemistry of the Amazon River waters.   Wang, 1980).
In this paper, some chemical composition data of lakes and reservoir waters in Taiwan are presented. The objectives are (1) to illustrate spatial variation in the water chemistry of lakes and reservoirs in Taiwan; (2) to show the relationships among the concentrations of major cations and anions; (3) to elucidate factors controlling the chemistry of water samples, such as seasalt and rainwater inputs, chemical weathering of surficial rocks, chemical equilibria, bio logical activities and evaporation; (4) to evaluate the susceptibility of lakes and reservoirs to acidification by acid rain; and (5) to compare the compositions with those of world-wide sur face waters so as to detect any significant differences.
Because rainfall and its subsequent evaporation provide important background informa tion, the contour maps of the average annual precipitation and evaporation rates of water ( in cm/yr) over Taiwan are given in Figures 2a and 2b for reference (modified after the Central Weather Bureau, 199 1; and Wang and Yi, 1992). In general, th� Central Range and the north eastern coast report high precipitation, mainly as a result of typhoons and monsoons. The highest precipitation rate is 480 to 500 cm/yr in northeastern Taiwan. In contrast, the western coastal area is low in precipitation but high in evaporation. In some areas, especially in the most western parts of Taiwan and the Penghu Islands, the evaporation exceeds the precipita tion (compare Figs. 2a and 2b Wang and Yi, 1992).

SAMPLING AND ANALYTICAL METHODS
Water samples from lakes and reservoirs were collected with plastic buckets and a Hydro Bios water sampler, usually from the top two meters at two different locations within each water body. The temperature and pH of water samples were measured immediately in the field. Samples were filtered with 0.45 mm Nalgene filters, stored in amber plastic bottles at 4 °C and then shipped back to the laboratory for chemical analysis.
The 4.004 and 7.4 15 pH buffers from the National Institute of Standard and Technology (NIST) USA were adopted to calibrate the Radiometer PHM 85 system. The reproducibility is within ± 0.003 pH units. Alkalinity was measured with the Gran Titration. The concentrations of K+, Na+, NH4+, Ca+2, Mg+ 2 , Cl-, N03-and SQ 4 -2 wer . e analyzed by a Dionex 2000i Ion Chro matograph. A Perkin Elmer 2380 AA provided additional analysis for ca+ 2 , Mg+ 2 , K+ and Na + . Certified rainwater standard samples (2694I and 26941I) supplied by the Environmental Re source Associate (ERA) were used as references. The overall analytical error is equal to or less than 6 %. The details of those analytical methods are given elsewhere (Chen et al., 1988.

Data
More than 600 water samples from about 120 lakes and reservoirs throughout Taiwan were analyzed (Chen et al., 1988;Chen, 1994). From these we have selected one representa tive sample for each lake and reservoir. Each selected sampl e is internally consistent with regard to the charge balance between major cations and anions (i.e., the total cationic and anionic charge ratio in the range of 1.0 ± 0.3). The chemical compositions of those selected water samples are summarized in Table 1, and sampling locations are shown in Figure 3. All are expressed in meq/l unless noted otherwise. Readers may contact the authors for the indi vidual names of lakes and reservoirs. The so-called mountain samples in Figure3 (solid circles � m) were obtained from water bodies at altitude greater than 500 m. For ease of identification, water samples from coastal areas less than 500 m in altitudes are further grouped into north (open circles= n), south (open triangles= s), and east (open squares= e) along with four offshore island samples (crosses = i).
In addition to the charge balance, the degree of agreement between measured conductiv ity (kb) and calculated conductivity (k 1) provides an additional check on theintemal corisiso s ca tency of the data set given in Table 1. The kcal (in µ S/cm) can be obtained by the equation: where Ui =mobility of ion i (cm2/ohm/equivalent); mi= concentration of ion i ( µmol/cm3); Jzi I = ab solute value of ion charge; and µ s = 10-5 ohm-1 • The ui values for cations and anions m infinitely diluted solutions at 25°C were obtained from the Handbook of Chemistry and Physics (CRC Press, 1994). As shown in Figure 4, the general agreement between kobs and k cal values is excellent, except for a few samples with low conductivity where the accurate mea- surements of both conductivity and ionic concentrations were less precise. As shown in the ternary diagram of alkalinity-S04-Cl data (Figure 5a), alkalinity accounts for more than 50 % of the total anions for most samples, as is also the case for the world mean river water (the large open circle in Figure 5a; Meybeck, 1987a, b). Chloride is predominant only in a few coastal samples (il, i4, ml, el and s27). For samples m2, m5, e4 and e5, more than 50 % of the total anions is SO 4 (Figure 5a). Figure 5b illustrsates the ternary diagram of non-seasalt major cations (ns-Ca, ns-Mg and ns-Na) after the seasalt contributions were subtracted by assuming that all Cl is of seasalt origin, i.e., ns-X = X -Cl* (X/Cl) , where X and Cl are the concentrations of a given seawater cation and chloride in a sample respectively, while (X/Cl) seawater is the ratio in seawater. The ns-Ca accounts for more than 50 % of the total non-seasalt cations for two-thirds of the samples, just as in the case of the world mean river water (the large open circle in Figure 5b). The predominance of ns-Na in samples nl, n3, n4, ,s3, s5, s27, m30 and m33 ( Figure 5b) may indicate Na inputs from the chemical weathering of surficial rocks high in feldspars and Naplagioclase. In comparison with the mean composition of Taiwan's upper crust ( the large open star in Figure 5b; Lan et al., 1991), Ca is preferentially leached out over Mg, Mg over Na, Li et al.    The large open circles represent the data for the world mean river water (Meybeck, 1987a and b). The large open star represents the data point for the av erage crust rocks of Taiwan (Lan et al., 1991), which is very similar to the world's average upper crust. Numerals are specific sample numbers given in Table 1.
and as is discussed later Na over K by the water samples (except for the ns-N a dominant samples), again this is also the case for the world mean river water. It was noted that the mean composition of Taiwan's upper crust is very similar to that of the world average (Lan et al., 1991).

Spatial Distribution Patterns of Major Ion Concentrations
The isopleth maps for chlorine (Cl), alkalinity (Alk) and sulfate (S04), as shown in Fig  ures 6a, 6b and 6c respectively, point out a trend of generally low concentrations of these species in high mountain areas, but high concentrations toward coastal areas, especially those toward the west. Figures 7a to 7d also show the concentration changes in Cl, Alk, S04 and Na as a function of the altitude of lakes and reservoirs.
The Cl concentrations in the mountain samples are uniformly low ( < 0.05 meq/l; Figure 6a and 7a) when compared to the average rainwater ofTaiwan (0. 15 meq Cl/I, at the bottom of Table 1; rainwater data are from Chen et al., 1994), whereas alkalinity is high between the altitude 500 and 2300 m (Figure 7c). The implications here are that the seasalt aerosols are the main source of Cl in the samples, and a large fraction of these seasalt aerosols carried by air masses is washed out by rainfall before reaching the high mountain or inland areas (Stallard and Edmond, 1981;Li, 1992). Also, seasalt is diluted by high rainfall in the high mountain area. In contrast, the high Cl concentration in the western coastal area (from about 0.5 meq/l to as high as 7 .1 meq/l in the Penghu Islands; Figure 6a) indicates the excess amount of evapora tion over precipitation (Figure 2a and 2b), and the high inputs of seasalt aerosols and sea sprays. Alkalinity results from the chemical weathering of aluminosilicate and carbonate rocks. The low alkalinity ( < 0.2 meq/l ) in some parts of the high mountain area (Figure 6b) is mainly caused by dilution through high rainfall there . The low-alkalinity lakes may be susceptible to acidification by acid rains. However, in view of the extremely high chemical weathering rate in the Central Range (about 65 mg/cm 2 yr; Li, 1976), acidification of lakes is less likely. For example, the chemical weathering rate of 65 mg/cm 2 yr roughly corresponds to a consumption rate of 0.5 meq H + /cm 2 yr (Li, 1972). The hydrogen ions are mainly provided by oxidation of the organic matter in soils. For comparison, the most polluted rainwater from Taipei (Table 1) can deliver only about 0.01 to 0.02 meq H+/cm 2 yr. However, one cannot totally rule out the possibility that the chemical weathering rate in the drainage areas of some high mountain lakes may be low, such that the acidification of lakes by acid rain may become feasible. The drastic increase in alkalinity yet with the constantly low Cl concentration in the mountain samples (Figures 7a, 7c and 8e) indicates that the increase in alkalinity is mainly caused by intense chemical weathering and not by evaporative concentration. In contrast, high alkalinity samples from the southwestern coast (Figure 6b) suggest additional evaporative concentra tion.
The sources of SO 4 in the water samples may include inputs from seasalt aerosols, fossil fuel burning, and chemical weathering of sulfur containing minerals, such as chalcopyrite, pyrite, gypsum and others. However, gypsum may be excluded because there are no known evaporite or gypsum outcrops ill the Central Range area (only a few secondary gypsum depos its in the southwestern part of Taiwan are known to exist). As shown in Figure 8a, the SO jCI ratio of most samples is much higher than those in seawater and polluted rainwater. Therefore, most sulfates are chemical weathering products. The close correlation between alkalinity and SO 4 , as shown in Figure 8d ., also provides further proof of this.

Relationships Among Major Ions
The correlation coefficient matrixes obtained fr om the raw chemical data (Table 1) and from the seasalt corrected data (i.e. non-seasalt components) are summarized in Tables 2a and  2b. The high to moderate correlations among alkalinity, S04, Ca, Mg and Na (correlation coefficients y > 0.5 in Tables 2a and 2b) are graphically presented in Figure 8d, 8f, 9a, 9b, 9d and 9e. These correlations can be explained by the following schematic weathering reactions of carbonate (marbles, limestones) and aluminosilicate rocks (gneiss and schists) which con tain many pyrite veins or grains in the Central Range of Taiwan, as stated earlier (Ho, 1975): where M is Ca and Mg. According to the above reactions, the equivalents of Ca+ 2 + Mg+ 2 should be balanced by those of HCQ 3 -+ SO 4-2 • This is the case for most samples shown in Figure 9c. It is clear that seasalt inputs for Ca, Mg and alkalinity are negligible.
The high to moderate correlations among Cl, Na and K ( y> 0.5) in Table 2a and weak correlations among Cl, ns-Na and ns-K ( y < 0.4) in Table 2b are again presented graphically in Figsure 8b, 8c and 9f. The Na/Cl and K/Cl ratios (Figure 8b and 8c) as well as the SOjCl ratios ( Figure Sa) of most samples are much greater than those in seawater or in polluted rainwater (dotted lines), thus again demonstrating the importance of the chemical weathering inputs for Na ,K and SO 4• The high SO jCl ratios in the polluted rainwaters as compared to that in seawater (Figure 8a) also indicate additional S04 sources other than seasalt in rainwater, such as the oxidation of natural dimethyl sulfide (DMS) and fossil fuel burning. The ns-Ca/ns Mg ratios of most samples are greater than that of the average upper crust of Taiwan ( Figure  9d; Lan et. al., 1991), implying preferential weathering of calcium-rich carbonate rocks over gneiss and schists. The ns-Na/ns-Mg ratios scatter around the average upper crust value of  Taiwan (Figure 9e). The ns-Na/ns-K values are mostly higher than the upper crust value (Fig  ure 9t), indicating preferential mobilization of Na over K during weathering processes as also shown by Colman (1982). The nitrate concentrations of the samples do not correlate appreciably with any other ion concentrations (Table 2a) and can be lower or higher than the values predicted from rainwater inputs (Figure 10, dotted lines). An NO/Cl ratio lower than that of rainwater may indicate an effective removal of nitrate by biological activity, such as photosynthetic uptake and/or micro bial denitrification, whereas high nitrate (>0.1 meq/l) samples may indicate serious pollution inputs from fertilizer usage and municipal sewage. Nitrate concentrations greater than 0.02 meq/l may have the potential to give rise to lake eutrophication (Sawyer, 1947). However, Vollenweider (1975) and Schindler (1974Schindler ( , 1977 demonstrated that phosphorus rather than nitrate loading and the mean depth of lake or the mean residence time of lake bottom water concertedly affect lake eutrophication. Lake eutrophication is an important issue which is to be discussed in a separate paper. Only four lake samples exceed the nitrate drinking water standard (0. 16 meq/l) recommended by the U. S. Public Health Service (McCabe et al., 1970).

Carbonate Chemistry and Activity-Activity Diagrams
The partial pressure of C0 2 (Pco 2 in atmospheric units), pH and the activity or concentra tion of bicarbonate ) are related by the following equation (Li, 1988): (2) where Kco 2 = [H 2 C03] I Pco 2 = 10-1.47 and K 1 = [H+][HC0 3 -] I [H 2 C0 3 ] = 10-635 Equation 2 is plotted in Figure 1 la at different given Pco 2 values. The samples with measured pH and calculated bicarbonate concentrations are also plotted in the same figure. Bicarbonate concentrations were calculated from pH and alkalinity data, where alkalinity was assumed to be equal to carbonate alkalinity. Most samples have Pco 2 values greater than the atmospheric Pco 2 (about 10-3.5 atm), which is indicative of active microbial oxidation of organic matter in lakes and reservoirs. Some samples with high bicarbonate concentrations have unusually high pH or low predicted Pco 2 values (e.g. s2, s3, s7, n20, n23, n26, el in Figure lla), probably caused by evaporative concentration and active photosynthesis in the water bodies.
The activity-activity ratio diagrams adopted from Drever (1988) are shown in Figure 12a and 12b, and depict the stability fields of Ca-, Mg-and Na-smectites and kaolinite. If lake and reservoir waters are in thermodynamic equilibrium with solid phases, kaolinite is a stable secondary clay mineral in high mountain (> 2000 m) water bodies, whereas Ca-smectite and/ or kaolinite are stable secondary clay minerals in other water bodies. However, as emphasized by Drever (1988), thermodynamic equilibrium between solution and clay minerals in natural aquatic environments is rarely attainable.
The tight clustering of water samples along a 1: 1 slope in Figure 12a and along a 2: 1 slope in Figure 12b, as also shown by Bluth and Kump (1994)  conclusion can be derived for Figure 12b. This means that there is no need to invoke Norton's (1974) notion that Ca, Mg and Na are released from solid phases into river waters in equal molar concentrations at equilibrium conditions.

Classification Scheme of Natural Waters by Gibbs
In the plots of the total dissolved salts versus the weight ratios of Na/(Na +Ca ) and Cl/ (Cl+HC0 3 ) (Figure 13), the chemical data of numerous rivers and lakes around the world (Livingstone, 1963) mostly fall within the fields enclosed by the dashed lines (Gibbs, 1970). The waters with low dissolved salts and high weight ratios of Na/(Na+Ca) and Cl/(Cl+HCO) reflect dominant sources of atmospheric precipitation (refyrred to as being rain-dominated). Increasing Ca and HC03 inputs from the chemical weathering of surficial rocks produce wa ters with low weight ratios of Na/(Na +Ca) and Cl/(Cl+HCO) and moderate contents of total dissolved salts (also referred to be rock-dominated). Further evaporation of water and concur rent precipitation of CaC0 3 produce waters with high dissolved salts and high weight ratios of Na/(Na+Ca) and Cl/(Cl+HC0 3 ) (also referred to be evaporation-crystallization dominated; Gibbs, 1970). As shown in Figure 13, most of the data from Taiwan 's lakes and reservoirs similarly do fall within those fields, especially within the rain-dominated and rock-dominated segments. The obvious exception is the following: about one half of the mountain samples (solid circles in Figure 13a and b) tend to have low dissolved salts as compared to the coastal samples and to the world-wide surface waters at the given Na/(Na+Ca) and Cl/(Cl+HCO) values. As stated earlier, this fact can be explained both by the wash-out of a large fraction of seasalt aerosols before reaching the high mountain are a and by the dilution effect of seasalt by high rainfall in this high mountain area. Even though most of the coastal samples from Taiwan fall within the fields of world-wide surface waters, their rain-dominated segment tends to have high dissolved salts (n l, n4, and s27 in Figure 13), reflecting the high dissolved salts in the local coastal rainwaters (Table 1, and the open stars in Figure 13). Samples s3, s5, s14, s29, il and i13 (enclosed by the dotted line in Figure 13) fall outside the field of world-wide surface waters, especially with respect to Na/(Na+Ca) ratios ( Figure 13a) . Again thisindicates appreciable Na inputs from chemical weathering of feldspars and Na-rich plagioclase. Cl/CCl+HC03l Fig. 13. Plots of the total dissolved salts versus the weight ratios of N a/(N a + Ca) and Cl/(Cl + HC03) for the world-wide surface waters (fields enclosed by dashed lines, given by Gibbs, 1970); and for water samples from Taiwan's lakes and reservoirs. The dotted lines enclose normal coastal samples and some unusual samples discussed in the text. Numerals are specific sample numbers given in Table 1.
Finally, the ratio of a given pair of elements for the world mean river water is shown as a solid diagonal line in each diagram of Figure 8 and 9. The data from Taiwan 's lakes and reservoirs more or less cluster along those solid diagonal lines. Therefore, despite the excep tionally high chemical weathering rates of the island, the chemical weathering products from Taiwan are not much different from those of other areas around the globe (Li, 1976).

SUMMARY AND CONCLUSIONS
(1) Systematic variations in the .chemical composition of for lake and reservoir waters from Taiwan as functions of distance from the coast ( Figure 6) and altitude (Figure 7) can be explained by three factors: (i) the diminishing seasalt inputs from coastal to high mountain areas; (ii) the inputs from chemical weathering of carbonate and silicate rocks in upstream drainage areas of lakes and reservoirs; and (iii) the dilution by high rainfall in the high moun tain areas, and evaporative concentration in coastal areas, especially along the southwestern coast of Taiwan.
(2) The close correlations among the concentrations of Ca, Mg, Alk and S0 4 are best explained by the chemical weathering of marble, gneiss and schists in the Central Range of Taiwan, which also contains chalcopyrite and pyrite veins and ores. Large fractions of Na and K in lake and reservoir waters are also the end products of chemical w· eathering.
(3) High mountain lakes with low alkalinity may be susceptible to acidification by acid rains. The extremely high chemical weathering rates in the Central Range, on the other hand, may mitigate acidification.
( 4) The Pco 2 values of water samples as calculated from pH and alkalinity values ( Figure   1 la) are mostly greater than the atmospheric Pco 2 , indicating active microbial oxidation of organic matter in lakes and reservoirs. A fe w coastal samples have calculated Pco 2 values much lower than the atmospheric Pco 2 , probably caused by active photosynthesis in the water bodies, and become oversaturated with respect to calcite. Otherwise, most of the water samples are under-saturated to saturated with respect to calcite (Figure 11 b ).
(5) The activity-activity diagrams ( Figure 12) indicate that kaolinite is a stable secondary clay mineral in high mountain (> . 2000 m) lakes and reservoirs , whereas Ca-smectite and/or kaolinite are stable secondary clay minerals in the other lakes and reservoirs. The observed 1: 1 and 2: 1 slopes in Figures 12a and 12b are shown to be artifacts.
(6) In spite of exceptionally high chemical weathering rates in Taiwan, the compositions . of water samples from Taiwan's lakes and reservoirs are mostly comparable to those of the world-wide surface· waters (Livingstone ,1963 ;and Gibbs, 1970). Exceptions are some moun tain samples which tend to have low dissolved salts relative to the world-wide surface waters due to the wash-out of seasalt aerosols before reaching to the high mountain areas ( Figure 13). Additionally, some coast<';tl samples have unusually high Na/(Na+Ca) values when compared with the world-wide surface waters at moderate total dissolved salt concentrations; they also show low Cl/(Cl+HC0 3 ) values (Figure 13), suggesting Na inputs from the chemical weather ing of Na-rich minerals, such as feldspars and plagioclases. and the Environmental Protection Administration (EP A82-G 104-09-12) to CTC and JJH . Ac knowledgment is also due to two anonymous reviewers. This is SOEST publication number 4567.