Model Prognoses for Future Acidification Recovery of Surface Waters

File failed to load: .... Long-term (30 years) monitoring data were used here to constrain the calibration of the acidification model MAGIC to three ...
0 downloads 0 Views 249KB Size
Download PDF
Environ. Sci. Technol. 2005, 39, 7970-7979

Model Prognoses for Future Acidification Recovery of Surface Waters in Norway Using Long-Term Monitoring Data THORJØRN LARSSEN* Norwegian Institute for Water Research (NIVA), Post Office Box 173, Kjelsås, N-0411 Oslo, Norway

During the past 20 years, acid deposition in Europe has decreased by more than 60%, yet still a large number of lakes and streams in southern Norway have not recovered to a water quality sufficient to support sustainable populations of trout or salmon. Long-term (30 years) monitoring data were used here to constrain the calibration of the acidification model MAGIC to three Norwegian calibrated catchments. The model accounted for 60-80% of the variance in the year-to-year variations in concentrations of most of the major ions in streamwater. The results support the use of the lumped parameter acid neutralizing capacity (ANC) to link chemical parameters to biological response, as the calibration efficiency for ANC is considerably higher than for other biologically important parameters such as inorganic aluminum (Aln+) and pH. Three different scenarios for future deposition of sulfur were run: current legislation, maximum feasible reductions, and an illustrative scenario removing all anthropogenic deposition. These analyses show that much of the potential improvement in water quality has already occurred and that only limited further improvement can be expected from the current legislation. The current legislation is unlikely to produce ANC values sufficiently high to allow self-reproducing populations of trout at two of the three sites. Most of the response in water chemistry to reduced acid deposition has been rapid; the water chemical responses largely occur the same year or a few years after reduction in the input. The soil pool of exchangeable base cations depleted during 150 years of acid deposition, however, requires several centuries for replenishment. The uncertainties in future predictions come from several factors, such as future nitrogen dynamics and impacts from changes in seasalt and precipitation events. The differences in future water chemistry predicted from changed seasalt deposition or nitrogen dynamics are larger that the differences between different deposition scenarios. Hence, these factors must be included in future assessments of recovery from acidification.

Introduction During the last two decades the policy goals of acidification modeling have changed. Initially, the objective was to illustrate the effects of anthropogenic acid deposition and the benefits of emission reductions. As substantial reductions * Corresponding author phone: +47 22 18 51 94; fax: +47 22 18 52 00; e-mail: [email protected]. 7970

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 39, NO. 20, 2005

in emissions have been implemented, the focus has increasingly changed to include issues such as (a) whether the latest internationally agreed measures are sufficient and (b) when waters and soils will recover from acidification (1). To address these questions, policy makers have taken increasing interest in dynamic models, especially as the time component is lacking in the steady-state models used to calculate critical loads (2). As acid deposition in Europe decreases as a result of reductions in sulfur (S) and nitrogen (N) emissions, there is an increased need to know whether these reductions are sufficient to achieve ecosystem recovery and when this recovery will occur. To address these issues, dynamic acidification models are required. The predictive power of dynamic acidification modeling is a crucial issue for determining the usefulness of such models to support policy making. In Norway, S deposition has decreased by about 60% over the period 1974-2003 (3). Sulfate (SO4) concentrations in surface waters have decreased correspondingly and at approximately the same rate (4). The decreased SO4 concentrations in streams have been accompanied by decreased concentrations of the acid cations H+ and inorganic aluminum (Aln+) and to a lesser extent the base cations calcium (Ca) and magnesium (Mg) (4). Acid-neutralizing capacity (ANC), defined as the equivalent difference between the sum of base cations and the sum of strong acid anions, has increased significantly (4). Dynamic models for surface water acidification have been developed during the last two decades. Models have been refined and enlarged through inclusion of additional processes as new data and new knowledge have become available (e.g., ref 5). Dynamic acidification models should be calibrated against observed data. The predictive performance of a model is therefore related to the availability and quality of monitoring data. Deposition and runoff composition and amount have been monitored since the early 1970s at three calibrated catchments (Birkenes, Storgama, and Langtjern) in southern Norway. These sites are well-suited for model calibration and evaluation, and the long time series provide considerable constraints in model calibration. Here the MAGIC model version 777 (5-7) was applied to these three Norwegian catchments using the long-term (1973/4-2004) monitoring data. After calibrating the model to the available time series data, predictions were run for three different future deposition scenarios: current legislation, maximum feasible reduction, and no anthropogenic emissions. Major confounding factors contributing to the uncertainty in the predictions were evaluated.

Material and Methods Monitoring Sites. Monitoring data from three calibrated catchments in southern Norway (Birkenes, Storgama, and Langtjern) were selected for this study (Figure 1). These data are among the best long-term data available. Samples of runoff have been collected at least weekly and of precipitation weekly or daily. There has been no disturbance in the catchments by human activities. The sites are considered representative for three dominating nature types in the most important acid-sensitive areas in southern Norway, a region in which fish populations have been damaged in thousands of lakes and streams as a result of acidification (8). All sites are heavily influenced by acid deposition, were severely acidified when measurements started in the mid-1970s, and have exhibited substantial recovery during the late 1980s 10.1021/es0484247 CCC: $30.25

 2005 American Chemical Society Published on Web 09/21/2005

FIGURE 1. Map of southern Norway showing the location of the three calibrated catchments.

TABLE 1. Characteristic Data for the Birkenes, Storgama, and Langtjern Calibrated Catchments (data from ref 4) parameter catchment area (km2) latitude (degrees north) longitude (degrees east) altitude (meters above sea level) surface water measured in seasalt influence (see Cl- concentrations in Table 2) water DOC concentration (mg L-1) annual precipitation (mean 1961-1990) (mm)a

Birkenes Storgama

Langtjern

0.41 58.38 8.25 190 stream high

0.60 59.06 8.66 587 stream medium

4.69 60.37 9.73 516 lake outlet low

5.2

4.8

9.6

1490

994

747

a Data are from the nearest long-term meteorological station and may deviate somewhat from the precipitation at the site; see ref 4.

and 1990s. They represent the range of acidified waters in southern Norway. All three sites are part of the Norwegian monitoring program for long-range transported air pollutants (4) and are also part of the international ICP-Waters monitoring program run under the auspices of the UNECE Convention on Long-Range Transboundary Air Pollution (CLRTAP) (1, 9). The three sites differ in certain important characteristics, such as dominant vegetation, catchment size, hydrology, seasalt influence, as well as acid deposition inputs (Table 1). The Birkenes catchment is located in West-Agder County about 20 km north of Kristiansand on the south coast (Figure 1). The catchment area is 0.41 km2, and the elevation is 200-300 masl. The vegetation is mainly 100-year old Norway spruce (Picea abies) with some Scots pine (Pinus sylvestris) and birch (Betula pubescens) and an undergrowth of mosses, blueberry, and fern. Mineral soils have developed in a shallow layer of glacial till on granitic bedrock. Mineral soil types are acid brown earth and podzols. Peaty deposits have developed on poorly drained sites in the catchment. On the slopes, well-drained thin organic layers on gravel or bedrock are common. The catchment is drained by three small secondorder streams, which converge about 150 m above the V-notch weir. The site for precipitation and air sampling is located about 500 m north of the catchment. The site is substantially influenced by seasalt deposition, which has considerable influence on the streamwater chemistry. Langtjern is a lake catchment located 120 km northwest of Oslo in the county of Buskerud in southeastern Norway

(Figure 1). The area is underlain by felsic gneisses and granites; thin soils are developed on till of generally the same lithology as the bedrock. Within the Langtjern catchment, 63% of the area is covered by mixed spruce (Picea abies), pine (Pinus sylvestris), and birch (Betula pubescens) forest, 16% by peat deposits or bogs, and 16% is exposed bedrock. The catchment area of Langtjern is 4.69 km2. Langtjern itself has a surface area of 0.227 km2. Detailed descriptions of the site can be found in ref 10. The station for precipitation and air chemistry is located about 6 km east of the catchment; it was moved in 1995 from Gulsvik to nearby Brekkebygda. The Storgama catchment is located in Telemark County in southernmost Norway about 50 km from the coast (Figure 1). The catchment area is 0.6 km2, and the elevation is 580-690 masl. The vegetation is predominantly sparse unproductive forest of Scots pine (Pinus sylvestris) and some birch (Betula pubescens) with undergrowth of heather. Mineral soils have developed in a shallow layer of glacial till on granitic bedrock. The main mineral soil type is podzol. Shallow peaty deposits have developed on poorly drained sites in the catchment. The weir at the bottom of the catchment is at the outlet of a small pond. The station for precipitation and air chemistry is located about 6 km from the catchment at Treungen. Storgama is the most barren of the three sites. The MAGIC Model. MAGIC is a lumped-parameter model of intermediate complexity, developed to predict the longterm effects of acidic deposition on surface water chemistry (5-7). The model simulates soil solution and surface water chemistry to predict average concentrations of the major ions. MAGIC calculates for each time step (one year in the present applications) the concentrations of major ions under the assumption of simultaneous reactions involving sulfate adsorption, cation exchange, dissolution-precipitationspeciation of aluminum and dissolution-speciation of inorganic carbon. MAGIC accounts for the mass balance of major ions in the soil by book-keeping the fluxes from atmospheric inputs, chemical weathering, net uptake in biomass, and loss to runoff. More detailed model description and illustrative applications can be found in refs 11-16. The model was set up with one soil box. All water and ions deposited on the catchment (i.e., not the water deposited directly on the lake) are assumed to pass through the soil. Model Input Data. Data inputs required for MAGIC comprise soil chemical and physical characteristics, input and output fluxes for water and major ions, and net uptake fluxes for vegetation. Annual deposition input fluxes (water and major ions) were obtained from the monitoring station closest to the site (data from ref 3). The deposition monitoring stations are located relatively close to the calibrated catchments, but the precipitation collector is not necessarily fully representative for the whole catchment because of different exposure, slopes, altitudes, etc. In addition, dry deposition is not accounted for in precipitation samples. As dry deposition patterns may be different form seasalt aerosols and gaseous SO2 (assumed to be the major components of dry deposition), these were treated separately. In the model calibration, chloride was assumed to be a conservative tracer, not taking part in chemical processes, but simply following the water through the soil. Standard seasalt ratios with chloride were used to adjust the seasalt fractions of the other ions in precipitation. For total S deposition, the flux was adjusted to surface water data assuming constant SO4 weathering over time, no uptake in the soil, and negligible adsorption. Hence the dry deposition flux of S was assumed to be the difference between the long-term net input and output fluxes adjusted for a small weathering flux. The assumption of limited SO4 adsorption was tested by comparing the shapes of the long-term trends in the SO4 input and output fluxes. VOL. 39, NO. 20, 2005 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

7971

TABLE 2. Model Inputs and Calibrated Parameters parameter depositionb Ca depositiona Mg depositiona Na depositiona K depositionb NH4 depositionb SO4 depositiona Cl depositionb NO3 discharge, annuala soil depth bulk density CEC SO4 adsorption half saturation SO4 maximum adsorption capacity pCO2 soil pCO2 stream organic acids, soil organic acids, stream vegetation uptake Ca vegetation uptake Mg vegetation uptake Na vegetation uptake K net uptake N soil net uptake N aquatic organic acidity constant pKa1 organic acidity constant pKa2 organic acidity constant pKa3 weathering S Al(OH)3 solubility constant, soil Al(OH)3 solubility constant, stream weathering Ca weathering Mg weathering Na weathering K selectivity coeff. Al-Ca selectivity coeff. Al-Mg selectivity coeff. Al-Na selectivity coeff. Al-K Ca saturation (pre-industrial) Mg saturation (pre-industrial) Na saturation (pre-industrial) K saturation (pre-industrial) total base saturation (pre-industrial)

unit

Birkenes

Storgama

Langtjern

13.6 30.3 131.1 5.7 65.8 116.3 153.6 71.2

4.9 6.4 26.1 1.8 28.0 49.3 30.7 33.4

4.2 2.6 9.2 2.5 23.4 30.8 9.9 22.4

1.15 0.4 655 113 100 0.1 0.33 0.07 65 10 16.6 3.8 0.0 7.3 93 0 3.04 4.51 6.46 20

0.91 0.32 432 146 100 0.1 0.33 0.07 30 16 0.0 0.0 0.0 0.0 85 0 3.04 4.51 6.46 6.4

0.60 0.4 681 109 100 1 0.33 0.07 35 30 10.3 3.0 0.0 15.0 97 50 3.04 4.51 6.46 6.6

Calibrated or Optimized Parameters log 10 7.8 log 10 8.6 meq m-2 50.0 meq m-2 4.0 meq m-2 12.0 meq m-2 4.0 log -1.73 log -1.58 log -1.88 log -7.44 % 14.8 % 9.3 % 2.4 % 2.9 % 29.4

9.2 8.2 16 0.2 6.0 0.0 -2.51 -3.03 -2.97 -7.31 17.0 9.0 1.7 2.6 30.3

9.0 8.5 30 7.0 5.5 14.2 -1.03 -0.96 -2.37 -6.65 17.5 5.0 1.0 3.0 26.5

Deposition Fluxes meq m-2 meq m-2 meq m-2 meq m-2 meq m-2 meq m-2 meq m-2 meq m-2 Fixed Parameters m m kg m-3 meq kg-1 meq m-3 meq kg-1 atm atm mmol m-3 mmol m-3 meq m-2 meq m-2 meq m-2 meq m-2 % %

a Long-term annual values, calculated from al observations, used when year-specific observations are not available. the five-year period 1988-1992.

The entire time series (1974-2004) of deposition data was used as model input. Prior to the period of monitoring, historic deposition for SO4, oxidized N (NOx), and reduced N (NHy) were taken from calculations with the EMEP model based on historic European emission estimates (17). The EMEP model results are available as annual deposition fluxes for each year since 1880 for grid cells of 50 km by 50 km. A sequence for deposition of each compound was found relative to the five-year average for 1990. To get site-specific historic deposition, the site observed deposition in 1990 was then multiplied with the scale factor for all years prior to monitoring. For base cations such historic data do not exist, and the historical Ca deposition was given a time trend by assuming 50% of the nonmarine deposition measured in 1988-1992 being scaled to the historical sulfur deposition pattern. The other base cations were kept constant through time, as no time trend is evident in the observations. The assumptions of historical base cation deposition have been shown to be of minor importance for the forecast results when the seasalt fraction of base cation deposition dominates (18). Sensitivity 7972

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 39, NO. 20, 2005

b

Average deposition for

analyses of changing the deposition of non seasalt Ca were carried out and showed very low sensitivity (results not shown). The historical deposition sequences were combined with the long-term observed average discharge to calculate fluxes. Since averages were used, in combination with smooth deposition sequences, the model outputs were also smooth for the time period up to 1973, after which monitoring data with year-to-year variations were used. For the forecast scenarios the model outputs became smooth again, as forecasts were based on emission estimates and EMEP model calculations (17). To evaluate the importance of the climatic variations (water flux and seasalt inputs) forecasts were also run with an estimated year-to-year variation in the future. Soil data were available from four different plots in each of the three catchments. Soils have been sampled three times, at eight-year intervals at all three sites, and analyzed for exchangeable cations, bulk density, and other parameters (19-21). Five soil layers were sampled at each plot, all in four replicates. To calculate one overall catchment value, the samples were weighted by density and depth, according

to the procedure described by ref 13. The catchment average soil depth has been estimated at 0.40 m for Birkenes and Langtjern and 0.32 m for Storgama. The soil data show considerable variation between sampling sites and horizons for each of the sampling years. Because of this large variation, there are usually no significant differences between years for these catchments (19-22). The variation is most likely related to the sampling and analytical procedures and probably not to actual changes in the soil chemistry (22). As model inputs for bulk density and cation exchange capacity (CEC), the average of the three years samples was used. For the base saturation, only data from the two last years were used because a change in the analytical methodology took place between the first and second sampling (21). Soils data used are summarized in Table 2. For the soil CO2 pressure, a default value of 10 times atmospheric CO2 content was used. Data for vegetation uptake of base cations were taken from the ICP Forest sites (the International Cooperative Program on Assessment and Monitoring of Air Pollution Effects on Forests (1); data from refs 23, 24). As the model input is the net uptake, that is, the net removal over a long time span, only the nutrients measured in the tree stems were used in estimating the uptake. The input data are summarized in Table 2. Model Calibration. The observed surface water chemistry and base saturation were not used as model inputs, but rather as targets in the model calibration. This means that the model was run with a set of inputs and the model outputs compared to the observations. The model was calibrated by adjusting parameters until modeled surface water chemistry and soil base saturation matched the observed data. The SO4 adsorption parameters were calibrated by comparing the long-term slopes of observed and modeled SO4 concentrations in surface water. The organic acid concentration in soil solution was calibrated by comparing the slopes of observed base cation concentration in surface water with the model outputs. The organic acid concentration was used to adjust the initial acidity of the soil solution and hence indirectly was used to calibrate the cation exchange constants. The concentration of NO3 was calibrated by first assuming 100% nitrification of the incoming reduced N, and then setting NO3 uptake in the catchment at a constant fraction of the total N inputs such that modeled water concentrations matched observations. Concentrations of NH4 in these waters are negligible. This uptake fraction was kept constant over time in the present calibration, as there is little information supporting the choice of other mechanisms. However, the fate of N in the catchment in the future has a strong influence on the forecast results and different scenarios for future N retention were evaluated. The use of long-term trend data for surface water chemistry in calibration limits the choices of parameter values in model calibration and increases confidence in the model predictions (14). For the exchangeable base cations in the soil there was no significant time trend, and calibration was done to fit an average between the two years with observations. The base cation concentrations in surface water and the exchangeable base cation pools were calibrated by adjusting weathering rates, initial values of base saturation (i.e., preindustrial, 1850), the dissolution constant for aluminum trihydroxide, and organic acid concentration in soil solution simultaneously until modeled values matched the observations. To quantify the model calibration performance, an efficiency value was calculated for each of the surface water parameters. In the calibration process, most weight was given to the dominant ionssthose that contribute most to the ion balance and ANC. The efficiencies (Eff) for parameter M

FIGURE 2. Historic and future deposition scenarios used (based on ref 17). CLE: current legislation; MFR: maximum feasible reduction; NOD: no anthropogenic deposition. See text for description. (model outputs that are compared to observations in the calibration process) were calculated on the basis of the deviations between modeled and observed values relative to the variation in the observations and were calculated as

x

n

∑(M

EffM ) 1 -

obs,i

- Mmdl,i)2

i)1

Varobs,M

‚ 100%

where Mobs,i represents observed concentrations and Mmdl,i represents modeled concentrations of parameter M in year i. Varobs,M is the variance in the observations over all years. The goal here was to predict acidification impacts on surface waters in relation to future recovery. The outputs were limited to chemical parameters that relate to possible impacts on indicator organisms or ecosystem functions. For surface waters, ANC has been shown to correlate well with damage to fish populations in Norway (25). For sustaining a trout population, an ANC value of 20 µeq L-1 is commonly used as a critical limit for Norway (25), with a possible correction for lakes with high concentrations of dissolved organic matter (26). In model calibrations, the aim was to maximize the fit to observations for the model outputs contributing most to the ANC. Forecast Scenarios. Forecast scenarios for S and N deposition have been estimated at the EMEP grid scale on the basis of emission scenarios (17). The deposition sequences used were averages for the four EMEP grid cells covering most of southern Norway, scaled to current deposition (Figure 2). Three future deposition scenarios were applied: current legislation (CLE), maximum feasible reduction (MFR), and no deposition from anthropogenic sources (NOD). The CLE scenario is a combination of the present legislation in Europe. This mainly includes the 1999 Gothenburg protocol of the Convention on Long-Range AirPollutants (CLRTAP) of the United Nations Economic Commission for Europe (27) and the National Emission Ceilings directive of the European Union (28), in addition to other national legislation implemented or agreed (17). For Norway VOL. 39, NO. 20, 2005 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

7973

FIGURE 3. Modeled (lines) and observed (dots) data and calibration efficiencies for the three calibrated catchments: (a) Birkenes; (b) Storgama; (c) Langtjern. The lines show modeled concentrations of SO4, Ca, and ANC in streamwater and %BS in soil; the dots show the corresponding annual average observed values. The bars show calibration efficiencies. The left panels show the time period 19702010; the right panels show the extended time period 1850-2050. The bars show calibration efficiencies. High efficiency means good agreement between model outputs and observations, while low values mean poor agreement. Negative efficiencies mean that the model outputs are further away from the observations than the variation in the observations. The color patterns on the bars show how much the variables contribute to the ANC. SAA means Sum of Acid Anions; SBC means Sum of Base Cations. 7974

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 39, NO. 20, 2005

the scenario gives slightly lower deposition then expected with only the Gothenburg protocol (29). The MFR scenario assumes that best available technology will be fully implemented in Europe between 2010 and 2015. The deposition trends were then estimated using the same tools as for the CLE scenario (i.e., the EMEP model and the RAINS database (30)). This scenario entails considerable reductions of both S and N emissions compared to the CLE scenario. The NOD scenario assumes a linear decrease to zero from 2010 to 2015 in all deposition of anthropogenic origin. Although clearly unreasonable, this scenario provides a measure of maximum recovery in the future.

Results and Discussion Calibration Results. The model outputs captured most of the important observed trends in SO4 and Ca concentrations, ANC, and base saturation, although there are differences among the sites (Figure 3). SO4 is an important anion and a major driver in the model calculations. The observed trends in SO4 were, in general, reproduced by the model, and the efficiencies were relatively high (87-91%). The model outputs indicated that the assumption of low SO4 adsorption is reasonable. For Langtjern, however, the match between observations and modeled SO4 concentrations was poor for the last years of the observations, where modeled concentrations were too high. The efficiency was high for chloride (Cl) at Birkenes, the site as which Cl comprises more than 50% of the anionic charge. The high efficiency was expected, as the Cl concentration in the stream was used to scale the level of seasalt ions in deposition. At the two other sites, the seasalt influence and hence the Cl concentration is smaller, leading to a larger relative variation from year to year not accounted for in the model runs. The levels as well as the trends in the predominant base cations were reproduced reasonably well by the model. The dominant base cations partly come from seasalt inputs and partly from catchment sources. At Birkenes, where the seasalt contribution is high, the efficiencies for the base cations were the highest. Sodium (Na), the major base cation at Birkenes, had a high efficiency. At the two other sites, Ca is the dominant base cation and also had the highest efficiency. Mg had lower efficiencies than Ca, but the overall trend and level were reproduced, although inter-annual variation was not fully captured (figure not shown). Potassium (K) and NO3 had low efficiencies, due to large inter-annual fluctuations controlled by biological processes not included in the model. These ions are present in generally low concentrations and thus contribute only little to the ANC; the low efficiencies are therefore of low significance here. H+ and Aln+ also had relatively low efficiencies. These ions are more difficult to model, as values for the parameters in the processes controlling these ions have proven to be difficult to generalize. Al and H+ are not part of the ANC calculation and as long as ANC is used as the model output related to the biological response, modest calibration results for H+ and Al can be accepted. However, if dose-response functions between chemical and biological status including, for instance, Al toxicity directly were to be applied, more effort would be required on the modeling of the Al chemistry. For ANC, the efficiency is a result of the calibration of the efficiencies for the contributing ions. Efficiencies for ANC ranged from 70 to 74% for the three sites, which is rather good considering the many simplifications inherent in the modeling process (e.g., annual time step, lumped catchment scale approach, one soil box, little climate dependency). Comparison with long-term trends was used as the means by which the model calibration was evaluated. Several other approaches have been used to evaluate the ability of the

FIGURE 4. Long-term sequences (1850-2500) for % BS, Ca concentration, and ANC for the MFR scenario at Storgama. MAGIC model to capture long-term trends. Such approaches include evaluation against large-scale experiments (31-33), paleolimnological data (15, 34), and long-term historical biological data (35). Comparisons with long-term observations have also been conducted previously (36, 37), although not with data spanning three decades as done in this study. The current assessment confirms the conclusions from these studies, that the model adequately simulates the major long-term trends of surface water acidification. However, as the decline in the SO4 deposition is leveling off, as seen in the most recent years (2000-2004), there are larger deviations between the model outputs and the observations. Forecast Results. Expected reductions in SO4 concentrations under different scenarios follow the deposition changes. The response in surface water to reductions in SO4 deposition is fast because there is little SO4 adsorption in the soils. SO4 concentrations well below the present levels are expected after implementation of the CLE scenario. Further reductions, beyond the current legislation, will lead to additional reductions in the surface water SO4 concentrations, but the future changes will be small relative to the observed changes over the past 20 years. For the base cation concentrations, illustrated by the Ca (Figure 3), the decrease observed so far is likely to level off and possibly increase in the future. The sources of base cations in the catchment are deposition, weathering, and the exchangeable pool in the soil. As the ionic strength in the soil solution decreases with decreasing concentration of strong acid anions in the deposition, the release from the exchangeable pool will decrease. The predicted slow increase further into the future after the deposition has stabilized can be explained by the buildup of the exchangeable base cation pool at a constant ionic strength. The buildup of the pool of exchangeable base cations under a constant deposition scenario was predicted to be very slow. To illustrate this, the model forecasts were run 500 years into the future using the MFR scenario and the deposition kept constant from 2010 onward. The model predicted a continued increase in the base saturation even after 500 years, with a subsequent increase in the base cation concentration and ANC in surface water (Figure 4). However, most of the increase in ANC was predicted to occur much faster: of the predicted increase from 5 to 11 µeq L-1 from VOL. 39, NO. 20, 2005 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

7975

FIGURE 5. Illustration of the importance of the uncertainties in future dynamics of N and the potential change in seasalt deposition compared with the CLE and MFR scenarios. The MFR deposition scenario for Birkenes was applied as base case and scenarios with increasing N leaching and increasing seasalt deposition were added to this scenario. The ”constant N leaching” scenario assumes a constant relative N retention at the currently observed rate. The “increased N leaching” scenario assumes constant N retention until 2010 and then a linear decrease in retention to zero retention in 2090. The seasalt scenario assumed a linear increase in the seasalt deposition from 2010 to a double level in 2090. 2015 to 2500, half of the increase comes in the first 30 years. This long-term change is very small compared to the direct response to the deposition changes already experienced. The direct response to deposition changes is larger and much faster. ANC is expected to increase as the deposition decreases in all the forecasts, even though the changes appear to be rather small for the CLE scenario as much of the reduction in S deposition included in the scenario already has taken place. At Birkenes, which has the lowest ANC among the sites, the model calculations indicated that still several decades are still needed before positive ANC values can be expected. Under the CLE scenario, model calculations suggest that ANC will remain negative for the next 50 years. At Storgama, positive ANC was observed for the first time in 2001. The model predictions did not simulate the high ANC values in the last three years. The main reason for this is a shift in the ratio between the observed SO4 concentrations in input and output. Output concentrations have continued to decline, while the decline in the inputs has leveled off. From the modeled ANC level, a small increase should be expected in the future for the different scenarios. A slight increase of about 6 µeq L-1 is expected over the next decade under the CLE scenario and about 12 µeq L-1 under the MFR scenario. The NOD scenario suggests a long-term increase in ANC at Storgama to about 20µeq L-1. At Langtjern, the ANC has already recovered to a relatively high value (35-40 µeq L-1). The model calibration to Langtjern was not as good as for the other surface water sites, and the observed slopes in the SO4 concentration and the ANC in the recent years were poorly simulated. The main reason for the lack of fit in recent years (2001-2004) is the 7976

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 39, NO. 20, 2005

same as that for Storgama: the observed SO4 concentrations in the input have leveled off and stabilized while the concentrations in the outputs have continued to decline. The mechanism for this shift is not yet understood and requires further investigations. From the modeled present levels, further improvements in ANC are expected. The difference for modeled ANC for the CLE and MFR scenarios in 2050 is only 3 µeq L-1, while the NOD scenario gives an additional increase of 5 µeq L-1. At Birkenes and Storgama, the model predictions suggest that the CLE scenario is not sufficient to produce future annual average ANC values sufficient to host sustainable populations of trout. By applying the MFR scenario, the ANC is predicted to reach approximately 20 µeq L-1 at Storgama by 2020. At Birkenes, an emission reduction beyond the MFR scenario is needed in order to reach a sufficient water quality to support a trout population. At Langtjern the ANC is already relatively high. There are currently populations of stocked trout in Langtjern, but reproduction is still limited. Langtjern has a relatively high concentration of dissolved organic carbon (DOC; around 10 mg L-1) (4), and organic acids play an important role. A slight further increase in the ANC (of around 5 µeq L-1) can be expected after implementing current legislation. The MFR scenario is predicted to give a slight further improvement of a few µeq L-1. A slight increase from the current situation may be sufficient to give water quality adequate to support a self-reproducing trout population at Langtjern. Future Nitrogen Dynamics as a Confounding Factor. In Europe, deposition of S is decreasing faster than deposition of N. The relative importance of N as a driver of acidification

is thus increasing. If all deposited N leached to surface waters as NO3, in many areas NO3 would contribute as strongly to acidification as SO4. However, most terrestrial systems are N-deficient and consequently have a pronounced retention of inorganic N. When N deposition exceeds uptake capacities and/or when root damage caused by acidification reduces N uptake capacities, NO3 concentrations in runoff water increase (i.e., “nitrogen saturation”) (38, 39). The 1999 Gothenburg protocol (27) is based on the precautionary principle, in which all N deposition over a certain catchment-specific threshold value is assumed to leach as NO3 in runoff (FAB model (40)). This leakage is the potential contribution of N to acidification. Today, most catchments leak only a small fraction of the potential. In southern Norway the retention is typically 70-95%, and increases from the southwest to the northeast (41). The extent of N retention in the future, and consequently the influence of N on surface water acidification, therefore represents a key uncertainty in future recovery from acidification. To illustrate this uncertainty the MFR scenario was run for Storgama assuming increased N leaching in the future. N retention was held at the same relative rate to the input as observed at present until 2010, and thereafter a linear decrease in the retention to zero in 2090 was assumed. This can be regarded as a long-term “worst case” scenario in terms of increased N leaching. The result was a linear increase in NO3 concentration and thus a decrease in ANC from 2010 (Figure 5). The decrease in ANC is substantial, further recovery is prevented, and in this illustrative scenario the increased N saturation results in re-acidification even after implementing emission reductions. The increased acidification following N saturation more than offsets the reduced deposition from the CLE to the MFR scenario. Seasalt Episodes as a Confounding Factor. Seasalt episodes are well-known to have shortOLINIT-term negative impacts on surface water chemistry (42). The driving process is cation exchange in the soil in which sodium and magnesium in seasalt-enriched rainwater exchange for Al and H+. This mobilization of Al is most severe a few days or weeks after the seasalt event, but it also influences the annual average chemistry. Here the MAGIC model was run with a yearly time step, based on calendar years, and hence did not focus on shortOLINIT-term seasalt events per se, but the impact on the annual average values can be assessed. The predicted ANC values in the most extreme seasalt years were simulated lower than the observations (seen most clearly in the model calibration for the years 1991 and 1993). This deviation is related to the simplified hydrology and water retention time used in the model. An inter-annual variation in seasalt deposition and rainfall was included to illustrate the impact of the salt input. The observed variation in seasalt deposition 1974-2003 was repeated into the future. Two scenarios were prepared, one assuming the observed seasalt sequence in the forecast, the other assuming a 50% increase in the seasalt input. The latter provides a simple illustration of a possible climate change scenario in which increased storm frequency results in increased seasalt deposition. The model suggested that the dip in ANC observed with seasalt events will be smaller as the acidification pressure decreases (Figure 6). The model also showed that the large variation from year to year is related to the variation in seasalt and water inputs. The biological relevance of the seasalt scenarios is important, as fish-kills in acidified waters typically occur during or immediately after seasalt events. Even though seasalt events also in the future are predicted to push ANC to values below what is considered acceptable for fish, the severity of the episodes will decline with declining acid inputs.

FIGURE 6. Illustration of the future impact of two different seasalt deposition scenarios at Birkenes. Both scenarios are based on repetition of the observed (1974-2003) sequence of seasalt deposition and water fluxes. Scenario “seasalts” repeats the observed sequence; scenario “seasalt 50% incr” assumes 50% increased annual deposition in seasalts from 2003 and onward. One response to the increased seasalt input is increased Ca concentration in streamwater. This is partly due to increased seasalt input and partly to increased ion exchange with Na and Mg. The ANC is reduced in response to the increased seasalt input, but the difference in ANC produced by the two scenarios levels off as the base saturation increases. The increased base cation input leads to a faster increase in the soil base saturation. As the base saturation builds up faster in the soil, the difference in ANC between the two scenarios decreases. Similar modeling results have been reported for other sites in Scandinavia (33). The uncertainty introduced by varying the long-term inputs of seasalts is of magnitude comparable to the uncertainty related to nitrogen leaching and to the uncertainty arising from the different future S deposition scenarios (Figure 5). Addition of increased seasalts to the MFR scenario results in lowering ANC to a level similar to the CLE scenario. Policy Implications. Considerable improvement in surface water quality has been observed during the period 19802000 in Norway in response to reduced acid deposition (43). For many lakes and rivers, however, further improvements will be required to obtain surface water quality that can support sustainable populations of trout and salmon. The predictions here suggest that implementation of the Gothenburg protocol and other related measures (the CLE scenario) will give a modest amount of improvement in the surface water quality. The predictions, however, also show that for the most-sensitive surface waters these measures will be insufficient to reach acceptable water quality. A largescale European assessment of recovery from surface water VOL. 39, NO. 20, 2005 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

7977

acidification under the CLE scenario showed that southern Norway will still likely have widespread acidification problems long into the future, while many other parts of Europe are well on the way to recovering to acceptable water chemistry (11). European emission reductions beyond the CLE will be beneficial for bringing surface waters in southern Norway closer to a state with sustainable reproducing stocks of trout and salmon. Confounding factors related to potential changes in climate (as illustrated here by changed seasalt deposition as well as the uncertainties related to the unknown fate of future N deposition, may influence the recovery to a greater extent than further reductions in S deposition. Hence, uncertainties related to such factors should be taken into account in policy development. There is an increasing need to integrate impacts from different drivers of environmental change and to assess the policy options in an integrated fashion.

(13) (14) (15)

(16) (17)

(18)

Acknowledgments This work has been funded, in part, through grants from the Norwegian Pollution Control Authority, The Norwegian Directorate for Nature Management, and the Norwegian Institute for Water Research. Many thanks to Dick Wright for helpful discussions throughout this work and in preparation of the manuscript. The constructive comments from three anonymous referees are also appreciated.

(19) (20) (21)

Literature Cited (1) Bull, K. R.; Achermann, B.; Bashkin, V.; Chrast, R.; Fenech, G.; Forsius, M.; Gregor, H.-D.; Guardans, R.; Haussmann, T.; Hayes, F.; Hettelingh, J.-P.; Johannessen, T.; Krzyzanowski, M.; Kucera, V.; Kvæven, B.; Lorenz, M.; Lundin, L.; Mills, G.; Posch, M.; Skjelkvåle, B. L.; Ulstein, M. J. Coordinated effects monitoring and modeling for developing and supporting international air pollution control agreements. Water, Air, Soil Pollut. 2001, 130, 119-130. (2) UNECE, Convention on long-range transboundary air pollution. Executive Body for the convention. Informal document no. 1 for the Eighteenth session 28 Nov.-1 Dec. 2000. 2000. http:// www.unece.org/env/lrtap/. (3) Aas, W.; Solberg, S.; Berg, T.; Manø, S.; Yttri, K. E. Monitoring of long-range transported air pollution and precipitations atmospheric deposition 2003. The Norwegian Pollution Control Authority (SFT), 2004. (4) SFT. The Norwegian monitoring programme for long-range transported air pollutants. Annual reportseffects 2003. The Norwegian Pollution Control Authority (SFT), 2004. (5) Cosby, B. J.; Ferrier, R. C.; Jenkins, A.; Wright, R. F. Modeling the effects of acid deposition: Refinements, adjustments and inclusion of nitrogen dynamics in the MAGIC model. Hydrol. Earth Syst. Sci. 2001, 5, 499-518. (6) Cosby, B. J.; Hornberger, G. M.; Galloway, J. N.; Wright, R. F. Modeling the effects of acid deposition: Assessment of a lumped parameter model of soil water and streamwater chemistry. Water Resour. Res. 1985, 21, 51-63. (7) Cosby, B. J.; Hornberger, G. M.; Galloway, J. N.; Wright, R. F. Time scales of catchment acidification: A quantitative model for estimating freshwater acidification. Environ. Sci. Technol. 1985, 19, 1144-1149. (8) Hesthagen, T.; Sevaldrud, I. H.; Berger, H. M. Assessment of damage to fish populations in Norwegian lakes due to acidification. Ambio 1999, 28, 112-117. (9) Skjelkvåle, B. L.; Stoddard, J. L.; Jeffers, J. N. R.; Tørseth, K.; Høgåsen, T.; Bowman, J.; Mannio, J.; Monteith, D.; Mosello, R.; Rogora, M.; Rzychon, D.; Vesel’y, J.; Wieting, J.; Wilander, A.; Worsztynowicz, A. Regional scale evidence for improvements in surface water chemistry 1990-2001. Environ. Pollut., in press. (10) Henriksen, A.; Grande, M. Lake Langtjernsfish studies in the Langtjern area 1966-2000. NIVA-report 4537/2002. Norwegian Institute for Water Research, Oslo, 2002. (11) Wright, R. F.; Larssen, T.; Camarero, L.; Cosby, B. J.; Ferrier, R. C.; Helliwell, R. C.; Forsius, M.; Jenkins, A.; Kopa´cek, J.; Majer, V.; Moldan, F.; Posch, M.; Rogora, M.; Scho¨pp, W. Recovery of acidified European surface waters. Environ. Sci. Technol. 2005, 39, 64A-72A. (12) Wright, R. F.; Henriksen, A. Gap closure; Use of MAGIC model to predict time required to achieve steady-state following 7978

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 39, NO. 20, 2005

(22) (23)

(24) (25) (26)

(27)

(28) (29) (30)

(31)

(32) (33)

(34)

(35)

implementation of the Oslo protocol. NIVA-Report 4012/1999. Norwegian Institute for Water Research, Oslo, 1999. Wright, R. F.; Cosby, B. J. Future recovery of acidified lakes in southern Norway predicted by the MAGIC model. Hydrol. Earth Syst. Sci. 2003, 7, 467-483. Larssen, T.; Cosby, B. J.; Høgåsen, T. Uncertainties in predictions of surface water acidity using the MAGIC model. Water, Air, Soil Pollut. Focus 2004, 4, 125-137. Jenkins, A.; Whitehead, P. G.; Cosby, B. J.; Birks, H. J. B. Modeling long-term acidification: A comparison with diatom reconstructions and the implications for reversibility. Philos. Trans. R. Soc. B 1990, 327, 435-440. Ferrier, R. C.; Wright, R. F.; Jenkins, A.; Barth, H. Predicting recovery of acidified freshwaters in Europe and Canada: An introduction. Hydrol. Earth Syst. Sci. 2003, 7, 431-435. Scho¨pp, W.; Posch, M.; Mylona, S.; Johansson, M. Long-term development of acid deposition (1880-2030) in sensitive freshwater regions in Europe. Hydrol. Earth Syst. Sci. 2003, 7, 436-446. Lo¨vblad, G.; Persson, C.; Klein, T.; Ruoho-Airola, T.; Hovmand, M.; Tarrason, L.; Tørseth, K.; Rapp, L.; Larssen, T.; Moldan, F. The deposition of base cations in the Nordic countries. IVLReport B 1583, Swedish Environmental Research Institute, Gothenburg, Sweden, 2004. SFT. The Norwegian monitoring programme for long-range transported air pollutants. Annual reportseffects 1999. The Norwegian Pollution Control Authority (SFT), 2000. SFT. The Norwegian monitoring programme for long-range transported air pollutants. Annual reportseffects 2000. The Norwegian Pollution Control Authority (SFT), 2001. SFT. The Norwegian monitoring programme for long-range transported air pollutants. Annual reportseffects 2001. The Norwegian Pollution Control Authority (SFT), 2002. Stuanes, A. O.; Abrahamsen, G.; Røsberg, I. Acidification of soils in five catchments in Norway. Water, Air, Soil Pollut. 1995, 85, 635-640. Kvindesland, S.; Jørgensen, P.; Frogner, T.; Aamlid, D. Hydrogeochemical processes in a forested watershed in Southern Norway. Aktuelt fra Skogforsk. Norwegian Institute for Forest Research, 1994. Røsberg, I.; Stuanes, A. O. Table NS-1. Atmospheric Deposition and Forest Nutrient Cycling. A Synthesis of the Integrated Forest Study. Springer-Verlag: New York, 1992. Lien, L.; Raddum, G. G.; Fjellheim, A. Critical loads of acidity to freshwater -fish and invertebrates. NIVA-Report 2732/1992. Norwegian Institute for Water Research, Oslo, 1992. Lydersen, E.; Larssen, T.; Fjeld, E. The influence of total organic carbon (TOC) on the relationship between acid neutralizing capacity (ANC) and fish status in Norwegian lakes. Sci. Total Environ. 2004, 362, 63-69. UNECE. The 1999 Protocol to Abate Acidification, Eutrophication and Ground-level Ozone. United Nations Economic Commission for Europe, 1999. http://www.unece.org/env/lrtap/ multi_h1.htm EU. Directive on national emission ceilings for certain atmospheric pollutants. The European Parliament and the Council, 2001. EU. Information Note DGXI. D.3/MLu D(99). 2000. Amann, M.; Cofala, J.; Heyes, A.; Klimont, Z.; Mechler, R.; Posch, M.; Scho¨pp, W. The RAINS model. Documentation of the model approach prepared for the RAINS peer review 2004. IIASA, International Institute for Applied Systems Analysis, Laxenburg, Austria, 2004. http:// www.iiasa.ac.at/rains/ Wright, R. F.; Emmett, B. A.; Jenkins, A. Acid deposition, landuse change and global change: MAGIC7 model applied to Risdalsheia, Norway (RAIN and CLIMEX projects) and Aber, UK (NITREX project). Hydrol. Earth Syst. Sci. 1998, 2, 385-397. Cosby, B. J.; Wright, R. F.; Gjessing, E. An acidification model (MAGIC) with organic acids evaluated using whole-catchment manipulations in Norway. J. Hydrol. 1995, 170, 101-122. Beier, C.; Moldan, F.; Wright, R. F. Terrestrial ecosystem recoverysModeling the effects of reduced acidic inputs and increased inputs of sea salts induced by global change. Ambio 2003, 32, 275-282. Wright, R. F.; Cosby, B. J.; Hornberger, G. M.; Galloway, J. N. Comparison of paleolimnological with MAGIC model reconstructions of water acidification. Water, Air, Soil Pollut. 1986, 30, 367-380. Kroglund, F.; Wright, R. F.; Burchart, C. Acidification and Atlantic salmon: Critical limits for Norwegian rivers. NIVA-Report 4501/ 2002. Norwegian Institute for Water Research, Oslo, 2002.

(36) Hruska, J.; Moldan, F.; Kram, P. Recovery from acidification in Central EuropesObserved and predicted changes of soil and streamwater chemistry in the Lysina catchment, Czech Republic. Environ. Pollut. 2002, 120, 261-274. (37) Jenkins, A.; Camarero, L.; Cosby, B. J.; Ferrier, R. C.; Forsius, M.; Helliwell, R. C.; Kopa´cek, J.; Majer, V.; Moldan, F.; Posch, M.; Rogora, M.; Scho¨pp, W.; Wright, R. F. A modeling assessment of acidification and recovery of European surface waters. Hydrol. Earth Syst. Sci. 2003, 7, 447-455. (38) Aber, J. D.; Nadelhoffer, K. J.; Steudler, P.; Melillo, J. Nitrogen saturation in northern forest ecosystems. Bioscience 1989, 39, 378-386. (39) Stoddard, J. L. Long-term changes in watershed retention of nitrogen: Its causes and aquatic consequences. In Environmental Chemistry of Lakes and Reservoirs; Baker, L. A., Ed.; American Chemical Society: Washington, DC, 1994; pp 223284.

(40) Henriksen, A.; Posch, M. Steady-state models for calculating critical loads of acidity for surface waters. Water, Air, Soil Pollut. Focus 2001, 1, 375-398. (41) Kaste, Ø.; Skjelkvåle, B. L. Nitrogen dynamics in runoff from two small heathland catchments representing opposite extremes with respect to climate and N deposition in Norway. Hydrol. Earth Syst. Sci. 2002, 6, 351-362. (42) Hindar, A.; Henriksen, A.; Tørseth, K.; Semb, A. Acid water and fish death. Nature 1994, 372, 327-328. (43) Skjelkvåle, B. L.; Mannio, J.; Wilander, A.; Andersen, T. Recovery from acidification of lakes in Finland, Norway and Sweden 19901999. Hydrol. Earth Syst. Sci. 2001, 5, 327-338.

Received for review October 7, 2004. Revised manuscript received May 16, 2005. Accepted June 9, 2005. ES0484247

VOL. 39, NO. 20, 2005 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

7979