Published in J. Environ. Qual. 32:2290-2300 (2003).
© ASA, CSSA, SSSA
677 S. Segoe Rd., Madison, WI 53711 USA
TECHNICAL REPORTS
Surface Water Quality
Modeling Surface Water Critical Loads with PROFILE
Possibilities and Challenges
L. Rapp* and
K. Bishop
Department of Environmental Assessment, Swedish University of Agricultural Sciences, SE-750 07 Uppsala, Sweden
* Corresponding author (lars.rapp{at}md.slu.se).
Received for publication August 7, 2002.
 |
ABSTRACT
|
|---|
The critical load concept has become a valuable tool for policymakers in the European negotiations on emission reductions. Despite the international acceptance, ongoing validation of critical load methodology is of the utmost importance to avoid a situation where the calculation results are difficult to defend. In this paper we explore the potential of using the steady state soil chemistry model PROFILE as an alternative to the Steady State Water Chemistry (SSWC) method for calculating critical loads of acidity. The hypothesis is that the uncertainty in prediction of preindustrial leaching of base cations is reduced when soil properties instead of lake chemistry are used as input data. Paleolimnological reconstructions of preindustrial lake chemistry are used to test PROFILE. As PROFILE requires soil data that are not generally available on a catchment level, we used distributions of crucial parameters from soil survey data within the vicinity of five lakes for which paleoecological pH reconstructions were available. An important concern is the characterization of catchment hydrology. A calibration of the "effective" soil depth, needed to give PROFILE predictions that coincided with paleolimnology, suggested that approximately 0.6 m of the total soil depth was hydrologically active in supplying acid neutralizing capacity (ANC) to runoff through weathering. At present, there is insufficient evidence to either recommend or reject the PROFILE model for surface water critical loads. Before such a judgement can be made, the approach presented here has to be tested for other regions, and the definition of catchment hydrology needs to be investigated further.
Abbreviations: ANC, acid neutralizing capacity • BC, base cations • DOC, dissolved organic carbon • SSWC, Steady State Water Chemistry
|
INTRODUCTION
|
|---|
ACIDIFICATION of soils and surface waters remains a major problem in Europe and North America. New areas like southern Asia, southern Africa, and South America will also increasingly be subject to acidification due to industrialization and elevated levels of acidic depositions (Rodhe et al., 1995). Since acidification is a transboundary problem, international cooperation is necessary to address the problem. An important basis for such cooperation is agreement on the problem. One way of defining the acidification problem is to determine the critical load of acid deposition for a region. A critical load is defined (United Nations Economic Commission for Europe, 1994) as a "quantitative estimate of an exposure to one or more pollutants below which significant harmful effects on specified sensitive elements of the environment do not occur, according to present knowledge."
The critical load concept has become a valuable tool for policymakers in the European negotiations on coordinated emission reductions. This concept was first used as a basis for an international agreement on pollution control in the second Sulfur Protocol established in Oslo (United Nations Economic Commission for Europe, 1994) within the Convention on Long Range Transboundary Air Pollution (CLRTAP) of the United Nations Economic Commission for Europe (UNECE). As a result, Europe has embraced the critical load concept for determining national emission control polices (Bull, 1995) and new applications of the concept are being explored. The Gothenburg Protocol (United Nations Economic Commission for Europe, 1999) deals with four types of pollutants (S, nitrogen oxide + nitrogen dioxide [NOx], volatile organic compounds [VOCs], and NH3) thus tackling other environmental problems in addition to acidification.
Despite the international acceptance and usefulness of the critical load concept, ongoing validation and improvements of its methodology are of the utmost importance. The concept has been criticized (Skeffington, 1995) for not adequately evaluating ecological effects of energy options on the same terms as the costs. Moreover, the costs of emission control rise steeply as the degree of emission reduction approaches 100% (Skeffington, 1995; United Nations Economic Commission for Europe, 1993). It is also argued that the results of the international negotiations are as much a product of policy as of science. One reason for the latter statement is the large uncertainty about how the critical load ought to be calculated (Skeffington, 1995). As for the future, validation of the critical load concept must remain high on our priorities list (Cresser, 2000).
Data uncertainty alone may substantially decrease the number of distinct critical load classes that can be defined, and incorporation of uncertainty management within the critical load concept may improve its usefulness (Barkman et al., 1995, 1999). Furthermore, as emissions decline (due to a combination of economic restructuring and pollution control agreements), acid deposition is coming closer to the critical load. This increases the likelihood that deposition levels fall within the uncertainty range of critical load models. Therefore, problems will arise if critical load numbers are treated as hard numbers known with absolute accuracy (Skeffington, 1999). Therefore, a more realistic view of the critical load concept would be to treat critical load numbers in a risk perspective (Barkman, 1997) instead of threshold values above which environmental damage occurs.
These developments all tend to increase the importance of accurately defining critical loads for surface waters. One way of defining the critical load of acidity is (Hettelingh et al., 1991):
 | [1] |
where CL(acid) is the critical load of actual acidity (mmolc m-2 yr-1), BCw is the weathering rate of base cations (mmolc m-2 yr-1), and ANCle.crit is the acceptable leaching of acid neutralizing capacity (mmolc m-2 yr-1). For surface waters Eq. [1] is often modified to Eq. [2] because lake concentrations are used as input data:
 | [2] |
where Q is runoff (m yr-1), [BC*]o is the preindustrial (o) concentration of base cations (µmolc L-1), ANClimit is the critical chemical value (µmolc L-1), and BC*d is the nonmarine component (*) of base cation deposition (mmolc m-2 yr-1). The term ANC is the total buffering capacity of the surface water, derived from base cations and the anions of strong acids (Munson and Gherini, 1993):
 | [3] |
where all concentrations are given in mmol L-1.
This paper investigates the potential for improving on the method currently used for calculating critical loads of acidity for surface waters in the Nordic countries by using soil data instead of surface water data to estimate the catchment weathering rate.
|
THE STEADY STATE WATER CHEMISTRY (SSWC) METHOD
|
|---|
In Fenno–Scandia, the surface water critical loads have generally been calculated using the SSWC method (Henriksen et al., 1990a, b, 1992, 1993) and the First-Order Acidity Balance (FAB) model (Kämäri et al., 1992; Henriksen et al., 1993; Downing et al., 1993; Posch et al., 1997). The FAB model is an extension of SSWC in which nitrogen processes, especially retention in the catchment and the lake, are also taken into account. The calculation of the critical load for acidity is, however, that of the SSWC method discussed in this paper. The SSWC model has the advantage of being easy to use with limited data requirements that are readily determined from contemporary lake chemistry. A key feature of the model is the empirical F factor that is used to estimate the fraction of acidity from deposition of sulfur and nitrogen compounds reaching the lake that has been neutralized by ion exchange reactions in the catchment soil. This makes it possible to estimate preindustrial surface water ANC from current ANC and an estimate of preindustrial sulfate concentrations.
The basic difficulty with SSWC is that it estimates the fundamental basis of critical load calculations, the weathering rate of catchment soils under pristine conditions, from contemporary surface water chemistry. This is a rather indirect method fraught with difficulties and results in different critical loads depending on when the lakes are sampled. That contradicts the critical load concept because the critical load should be a stable quantity for an ecosystem. However, the connection between changing present water chemistry and changing critical loads is not surprising (Cresser, 2000). The implications of changing lake chemistry on surface water critical load estimates have been investigated in Rapp (2001).
|
THE PROFILE MODEL AS AN ALTERNATIVE TO SSWC?
|
|---|
A potential alternative to the SSWC method for determining surface water critical loads is the PROFILE model (Warfvinge and Sverdrup, 1992; Sverdrup and Warfvinge, 1993), a steady state biogeochemical process-oriented model that calculates soil chemistry, including the weathering rate, in a soil profile (Fig. 1)
. This model has been used for soil critical loads in Sweden and Denmark (Posch et al., 1997) and was once used to predict surface water critical loads (Downing et al., 1993). It has also been applied in other regions of the world, such as the state of Maryland in the USA (Sverdrup et al., 1992, 1995) and northern Asia (Semenov et al., 2000).
View larger version (38K):
[in this window]
[in a new window]
|
Fig. 1. Conceptual model of a forest soil as used by the PROFILE model. The soil layers represent the organic (O horizon), eluvial (E horizon), illuvial (B horizon), and the underlying mineral soil that are little affected by pedogenic processes (C horizon). The figure is extracted and modified from Alveteg (1998).
|
|
The PROFILE model is based on an approach entirely different from that of the SSWC method. Instead of using contemporary lake chemistry to determine the preindustrial weathering rate (included in [BC*]o, Eq. [2]), PROFILE bases the weathering estimate on the properties of the soil, together with related climatic and hydrological factors. The working hypothesis of this paper is that direct use of soil properties for calculating the weathering rate could reduce the uncertainty in the calculation of surface water critical loads. A similar approach where catchment topography, land use, and riparian zone geology was used for calculating freshwater critical loads proved promising (Kernan et al., 1998; Cresser et al., 2000).
Application of PROFILE to surface waters, however, involves challenges not present in the application of PROFILE to soil critical loads. For soil critical loads, the chemistry in the rooting zone (0–0.5 m) on an individual plot or stand is of interest because the soil solution that is taken up by the roots is a chemical criterion connected to forest health. When applying PROFILE to surface water critical loads, though, soil thickness and water flow paths must be considered. Furthermore, the model must integrate the variation of soil properties and hydrology across the entire lake catchment. The SSWC method avoids this integration problem by relying on lake chemistry that has already integrated the output of the entire catchment. So while the direct consideration of soil properties promises a more reliable and stable estimate of [BC*]o, there are problems of data acquisition and integration that will have to be overcome if PROFILE is to be a viable alternative for calculating surface water critical loads.
The implications of uncertainties about input data on the many soil properties needed for the implementation of PROFILE have been investigated with regard to weathering rates for a specific soil profile. In Jönsson et al. (1995), an uncertainty analysis was performed involving Monte Carlo simulations where the input parameters were varied systematically either simultaneously or in groups. The simulations were performed for three Nordic sites covered by pine and spruce forests, with podzol as the dominant soil type. The study revealed that soil moisture saturation, which involves soil bulk density and volumetric water content, had a great influence on the weathering rate, while parameters like soil stratification, precipitation, and percolation were of the least importance. However, percolation and acid deposition were of great importance for the ANC leaching calculations. The authors concluded that the weathering rate could be calculated accurately provided that the input data errors were within a specified range. However, a sensitivity analysis of the PROFILE weathering rate (Hodson et al., 1996), which allowed the parameters to vary across a wider range that corresponded to observed variations, concluded that the PROFILE model should be used with caution, and applications of the model should report on the uncertainty present in the input data. PROFILE was tested further by Hodson et al. (1997) in a study that demonstrated that the basis for the calculation of the exposed mineral surface was impaired by considerable uncertainties and drawbacks.
Other studies (Barkman et al., 1995, 1999) have looked into the effect of data uncertainties on critical loads of acidity and exceedances for forest soils. These studies involved Monte Carlo simulations as in Jönsson et al. (1995). Uncertainties in input data resulted in considerable overlaps between different percentiles (i.e., the number of distinct critical load classes decreased). Thus, it is not clear to what extent an area is protected from acid deposition.
Despite the uncertainties and criticism involved in using the PROFILE model, it still offers an interesting alternative to the SSWC method for surface water critical loads that should be explored. Its application to lakes has been limited (Downing et al., 1993) and no validation of its applicability to lakes has been conducted. It is also likely that such an analysis would bring out more information and knowledge about the PROFILE model with reference to its possibilities and limitations in other applications.
The objective of this study is to investigate the possibilities and challenges of using the PROFILE model for determining critical loads of acidity for surface waters. To evaluate the success of these determinations, paleolimnological reconstructions of preindustrial lake chemistry from five lakes in northern Sweden were used. Since an estimate of catchment soil parameters has to be made, but such data are not available (nor would it be clear how to average samples from different parts of catchments), all available soil data from the vicinity of the study area were treated as potentially representative lake catchments. The PROFILE model was run in a Monte Carlo mode. The resulting distribution of preindustrial lake chemistry (ANCo), calculated by the PROFILE model, was then compared with the paleolimnological reconstructions.
|
MATERIALS AND METHODS
|
|---|
Preindustrial Lake Chemistry from Paleolimnology
An important feature of the PROFILE model is that it is based on the assumption of steady state between atmospheric deposition and soil solution. Consequently, short-term capacity processes like ion exchange and sulfur retention are not taken into account. Therefore, a comparison of the PROFILE predictions, using contemporary deposition, to lake chemistry would not be an appropriate test of the model due to variation in deposition of pollutants during the last several decades.
Paleolimnological research in Sweden (Renberg et al., 1993; Korsman et al., 1994; Korsman and Birks, 1996; Korsman, 1999) has, however, made it possible to compare critical load models to preindustrial lake chemistry (i.e., before human impact when conditions were closer to steady state due to little anthropogenic pollution). However, a true steady state situation does not exist in nature, as natural climatic variations cannot be circumvented. Nevertheless, the paleolimnological reconstructions are the most reliable predictions for preindustrial lake chemistry. In Korsman's regional study (Korsman and Birks, 1996; Korsman, 1999), pH (±0.4 pH unit) and alkalinity (±30 µmolc L-1) were predicted using a calibration set based on water chemistry data from the period 1985–1989. In this study, a subset of five lakes was taken from Korsman's (Korsman and Birks, 1996; Korsman, 1999) study in northern Sweden (Fig. 2)
. These reconstructions of preindustrial alkalinity and pH were based on sediment cores taken at a depth of 25 to 30 cm. Depending on the sedimentation rate, the reconstructed lake chemistry represents a time period as far back as the 16th century and earlier. By assuming that these values represent a long-term average they can be compared with the steady state PROFILE model. These lakes were selected due to the location of available input data (soil depth), as well as a relatively homogenous bedrock and soil composition in the study area. There tends to be an increase in alkalinity (and ANC), as a result of changed land use (e.g., forest burning) since preindustrial times, except for Lill-Siksjön (Table 1). No dramatic changes have taken place, except for Vantsjön, where alkalinity has increased dramatically since preindustrial times.
Preindustrial lake buffering capacities, derived from paleolimnological reconstructions, hereafter referred to as "diatom inferred," were given as alkalinity values (Table 1). The critical load concept for surface waters and PROFILE, however, use ANC, the total buffering capacity of the water solution. Therefore, the reconstructed alkalinity numbers were converted to ANCo using the assumption that all buffering below pH 5.4 results from organic acids (Hemond, 1990; Köhler et al., 2000):
 | [4] |
where alkalinityo(5.4) is the diatom-inferred alkalinity titrated to pH 5.4 (µmolc L-1), ß is the charge density (µmolc mg-1), and DOCo is the preindustrial dissolved organic carbon (mg L-1). Thus, a fixed amount of ANC per mg DOC was added to the alkalinity numbers. The charge density (ß = 6.3) was arrived at by a calibration (Rapp, 2001) where the diatom-inferred preindustrial pHo values were predicted from ANCo (Eq. [4]). The calibration criterion was to yield a consistent pHo–ANCo relationship, meaning that the aim was to distribute the model error evenly from pH 4 to pH 8. The DOCo values are based on preindustrial diatom-inferred color (Korsman and Birks, 1996) and a relationship between modern DOC and color (Bishop, 1996):
 | [5] |
where color is given in mg Pt L-1.
The PROFILE Model
The PROFILE model has been thoroughly described in Warfvinge and Sverdrup (1992)( 1995), so the following description is an overview with an emphasis on how the model predicts surface water chemistry. The PROFILE model is the steady state version and predecessor of the dynamic model SAFE (Warfvinge et al., 1992). Natural soil horizons are simulated by a series of continuously stirred tank reactors in which the soil processes occur (Fig. 1). In principle, the model procedure starts with the inflow of precipitation to the first layer thus yielding a starting ANC. On the way down through the soil layers, ANC may either increase (weathering, uptake of nitrate) or decrease (uptake of base cations and ammonium, nitrification). By assuming chemical equilibrium, a corresponding pH can be calculated by quantifying the carbonate, aluminum, and organic acid buffering systems. The ANC of the water leaving the soil is the key parameter for surface water critical loads, which are based on an ANC limit needed to protect sensitive aquatic species.
The version of PROFILE used here is an extension of the model described in Warfvinge and Sverdrup (1995). The difference lies mainly in the parameterization of input data to allow for soil depth greater than the rooting zone. The catchment soil was modeled using four compartments to represent the organic (O horizon), eluvial (E horizon), illuvial (B horizon), and the underlying mineral soil that are little affected by pedogenic processes (C horizon) according to USDA (1998). An implicit assumption is that lake chemistry is represented by the runoff water chemistry leaving the C horizon (Fig. 1) (i.e., no lake processes producing or consuming ANC are taken into account). Studies (Kelly et al., 1982; Schindler et al., 1986) have shown, however, that reduction reactions of sulfate and nitrate as well as exchange and diffusion reactions in the sediments may buffer lakes undergoing acidification. To what extent these processes control lake chemistry is beyond the scope of this paper.
An important issue when applying PROFILE to surface waters instead of soils is how the water is routed through the soil profile before emerging as runoff. The model version for soil critical loads (Warfvinge and Sverdrup, 1995) does not include any horizontal flow out of upper soil horizons (i.e., all runoff water percolates down through the whole soil profile before entering a lake). This routing is, however, a major simplification that does not reflect what is known about runoff generation in the till soils that cover much of Sweden. The lower soil horizons often contain clay fractions or more compact tills with low hydraulic conductivity that transmit much less water compared with upper layers where the conductivity is higher (Rodhe, 1988; Bishop, 1991). Thus, a large portion of the annual runoff water flows in the upper horizons especially during seasonal events like spring floods. A natural catchment includes many different combinations of flow depths and pathway variations, making it difficult to define an average flow pattern that represents the catchment. Therefore, this study examines the effect of two hydrological extremes, on the assumption that the actual situation for different catchments is bracketed by these extremes. In one case, the whole soil profile is hydrologically active and thus contributes to weathering. The other case is where only 0.5 m is hydrologically active (i.e., conducts lateral flow). While there are times and locations when even less than the upper half meter is hydrologically active, we deem that on an annual basis restricting flow to the uppermost 0.5 m is a reasonable estimate of that hydrological end of the spectrum for Swedish conditions.
Distributions and Input Data
The PROFILE model requires a great deal of soil data for its implementation. In Sweden, these data demands can be satisfied to some extent by the database for forest critical loads, based on the Swedish National Survey of Forest Soils and Vegetation. The available database (Warfvinge and Sverdrup, 1995) includes 1883 sites with full chemical and physical descriptions of the soil, distributed all over Sweden. However, the soil data density is not sufficient to characterize individual lake catchments. Therefore, we have chosen to create distributions of soil data for an area that surrounds the study lakes. Using this approach, the PROFILE prediction of ANCo is a distribution of potential catchments that is compared with the paleolimnological reconstructions.
Distributions of parameters crucial for lake chemistry have been constructed (i.e., soil depth, mineralogy, texture, and soil water content). These were selected partly on the basis of findings in earlier studies (Jönsson et al., 1995; Hodson et al., 1996) and partly on what has not been explored earlier (soil depth). These are the factors that, to a large extent, determine the weathering rate. In creating the parameter distributions, all measurements from within 25 km of the study lakes were used (Fig. 2). The borderline was selected as it covers an area of uniform bedrock consisting of granite and greywacke. The study area consists of forests (52%), open fields (15%), peatland (26%), lakes (6%), and small areas of nonproductive forestland (1983–1987 Swedish National Survey of Forest Soils and Vegetation). The mineral soils are predominantly (90%) derived from glacial tills.
Soil depths were available from an archive compiled when a powerline (Fig. 2 and 3)
was surveyed in the late 1960s (Stegmann, 1998). The powerline transect extends through the landscape where the five lakes are situated and includes soil depth measurements at 300-m intervals.
Texture and soil water content distributions (Fig. 4 and 5)
are based on the 1983–1987 Swedish National Survey of Forest Soils and Vegetation (n = 112) The parameterization of data is outlined in Warfvinge and Sverdrup (1995). Texture classes range from boulders and rock (Class 1, particle size >20 mm), gravel (Class 2), sand (Class 3), loamy sand (Class 4), sandy loam (Class 5), sandy loam (Class 6), silt loam (Class 7), and clay (Class 8) according to USDA (1998). The corresponding surface areas are also illustrated. Class 5 dominates the distribution in the study area. The soil water contents are related to distance to the average ground water level. Class 1 represents xeric conditions while Class 6 represents aquic conditions according to USDA (1998). The distribution of soil water content is dominated by Classes 3 and 4, corresponding to 0.20 to 0.25 m3 m-3.
View larger version (14K):
[in this window]
[in a new window]
|
Fig. 4. Distribution of soil texture illustrated as classes and corresponding mineral surface area. Loss on ignition (2.5%) has been used to account for nonmineral surfaces.
|
|
The distribution of mineralogy was taken from the critical load database compiled by Warfvinge and Sverdrup (1995). The distribution was assumed to be rectangular, that is, all mineralogy sets (n = 10, Table 2) had the same probability of being included in the distribution. Another feature of this approach is that individual minerals are not varied within a site, as in Jönsson et al. (1995), but instead complete sets of mineralogy are varied within the study area. In this way, covariance effects are circumvented.
Other input data were kept constant (Table 3). To represent preindustrial conditions, the deposition, uptake, and litterfall data were estimated using the 1990 data as the reference for modern conditions. The large reduction in uptake relies on a much smaller biomass removal in preindustrial times (Östlund et al., 1997). Litterfall is not of importance for the runoff chemistry in the model as the nutrients are recirculated in the upper layers. Soil stratification was estimated by keeping the thickness of the O and E horizons constant while B and C horizons were varied depending on soil depth. A maximum depth was assigned for the B horizon (0.29 m) and the C horizon was not less than 0.01 m (Table 3).
Monte Carlo Sampling
The PROFILE model was run using Monte Carlo sampling to generate a set of parameter values from the distributions of soil depth, soil texture, soil water content, and soil mineralogy. Then the PROFILE model was run using these parameter sets, together with other input data to yield a distribution of preindustrial ANCo. To get reproducible results, 1500 samples from each distribution of soil depth, soil texture, soil water content, and soil mineralogy were sufficient.
At present, the PROFILE model has not been developed for land-use types other than forest soils. As a consequence, it is not clear how to treat peatland and other nonforested land areas in the model. As a first approximation, only the area covered by mineral soils is assumed to weather and contribute to weathering. Organogenic soils are assumed not to contribute weathering and ANC to runoff because the fraction of mineral soil is negligible. The water leaving such areas, including lakes, is characterized by precipitation chemistry (Table 3). In this way, the ANC of the runoff water leaving the distribution area is calculated by area-weighting the different ANC contributors:
 | [6] |
where ANCo,distribution is the preindustrial ANC of the runoff water leaving the distribution area (µmolc L-1); ANCprofile is the ANC calculated by the PROFILE model (µmolc L-1); areaweathered is the area covered by mineral soils such as forest soils, grazing, and open land (m2); ANCprecipitation is the ANC of the precipitation (µmolc L-1); area1-weathered constitutes peatland, bedrock outcrops, and lakes (m2); and areatotal is the total distribution area (m2). The distribution area is comprised of about 33% peatland and lakes. In the following sections, if not otherwise stated, this has been used when comparing PROFILE to paleolimnological reconstructions.
|
RESULTS
|
|---|
The ANC distribution calculated by the PROFILE model was compared with the paleolimnological reconstructions (five lakes), hereafter referred to as the Diatom model. Running PROFILE using distributions of soil depth, texture, mineralogy, and soil water content, with hydrologic routing of all water vertically downwards to the lowermost soil profile, results in a wide distribution (Fig. 6) . The other hydrological extreme, in which water movement is confined to the uppermost 0.5 m of the soil profile, results in a more narrow distribution (Table 4) and lowers ANCo because less soil contributes to weathering. The actual situation lies between these flow patterns, but probably shifted toward the shallower routing. The median value of the Diatom model predictions is bracketed by the medians of the PROFILE distributions.
View larger version (36K):
[in this window]
[in a new window]
|
Fig. 6. Frequency plot of preindustrial acid neutralizing capacity (ANCo) calculated by the PROFILE model (n = 1500). Dotted lines correspond to the medians of the distributions. The paleolimnological reconstructions (n = 5) are indicated including the uncertainty range (standard error of prediction, ±30 µmolc L-1).
|
|
View this table:
[in this window]
[in a new window]
|
Table 4. Comparison of the PROFILE model with the Diatom model when modeling preindustrial acid neutralizing capacity (ANCo).
|
|
To evaluate the relative importance of depth, texture, moisture, and mineralogy, they were varied one by one using the distributions. The parameters not varied were set at their median values. As an example, when the influence of soil texture was studied, soil depth (1.0 m), soil moisture (Class 3), and mineralogy (Set 2) were used. The mineralogy distribution is different from the other distributions as it was assumed rectangular and it is not clear how to rank the mineralogy sets. Therefore, Set 2 (Table 2) was selected as it gave the ANC median if only mineralogy was varied.
By plotting each run's cumulative distribution together with a run where all parameters are varied simultaneously, the contribution of each parameter to the whole variation in ANCo is visible (Fig. 7)
. The difference in ANCo between 10th and 90th percentiles (Table 5) is largest when varying soil depth. Soil texture can be crucial as low and high texture classes cause the extremes for any single parameter. Mineralogy was not crucial, as expected, because the study area is situated on fairly uniform bedrock consisting of granite and greywacke. The highest ANC, when varying mineralogy only, was caused by Mineralogy Set 10, the only set consisting of biotite and no vermiculite. Soil moisture content was of the least importance among the parameters investigated.
View larger version (23K):
[in this window]
[in a new window]
|
Fig. 7. Cumulative distributions showing the individual parameters' contribution to the whole variation in preindustrial acid neutralizing capacity (ANCo).
|
|
|
DISCUSSION
|
|---|
The results suggest that the effective soil depth is less than the total soil depth that can be expected in the study area. A calibration was performed to see what effective depths are needed to get the PROFILE model and the Diatom model to coincide for each lake. The same distributions were used with the exception of soil depth. The correction for peatland and lake area was made, but now estimates from each specific catchment were used (Table 1). The median of the PROFILE distribution was compared with the Diatom model's prediction of ANCo. The optimized effective depths varied from 0.59 to 0.79 m (Table 6), which is a relatively narrow range compared with the observed distribution of soil depth.
The PROFILE predictions should be interpreted as a population of 1500 potential catchments, each with a corresponding ANCo. Therefore, a more statistically sound way to compare the two models would be to treat the diatom predictions of the five lakes as a population as well. A calibration was done (Fig. 8) , as described, with a 33% correction for peatland and lake area. The active soil depth needed was about 0.62 m (i.e., within the range that was found for the individual lakes). Before generalizing this depth, the PROFILE model would need to be tested in other regions, where paleolimnological data are available, to see the variation in effective depths using the approach presented here.
View larger version (26K):
[in this window]
[in a new window]
|
Fig. 8. The PROFILE model (frequency plot), using a calibrated annual average hydrological soil depth, compared with the Diatom model. Error bars indicate the associated uncertainty (standard error of prediction, ±30 µmolc L-1) of the Diatom model.
|
|
While the effective depth of about 0.6 m is reasonable, the definition of hydrology is not completely resolved. An implicit assumption in the current model formulation is that the whole soil depth is hydrologically active (i.e., all runoff water percolates down through the entire soil profile). This is only a first approximation of soil hydrology, and one that makes the model output very dependent on soil depth (Fig. 7, Table 5). An important future task is to introduce lateral flow to the model, but this raises the question of how to route the water. It is clear that seasonal events like spring floods and frozen soils have to be considered as well as the spatial variation of the soil hydraulic conductivity.
Another important concern when distributing lateral flow is how to limit ANC accumulation in soil horizons with longer residence times. If a small amount of water, for instance 5% of runoff, percolates through a thick bottom layer, a long residence time for water can lead to unrealistically high ANC levels. To get realistic ANCs in these long-residence-time waters, the weathering rate has to slow down as the concentration of weathering products increases. In PROFILE, that is taken care of by the rate reduction functions (Warfvinge and Sverdrup, 1992, 1995), here presented as the "brake" for mineral dissolution caused by the accumulation of the weathering end-products of BC and aluminum (Al):
 | [7] |
where fH is the rate reduction factor (dimensionless), [BC2+] is the base cation concentration expressed as a divalent cation (mol L-1), CBC is the rate reduction concentration with reference to BC (µmolc L-1), 
BC,H is the rate reduction factor coefficient with reference to BC and H+ (dimensionless), [Al3+] is the aluminum concentration (mol L-1), CAl is the rate reduction concentration with reference to Al (µmolc L-1), and Al,H is the rate reduction factor coefficient with reference to Al and H+ (dimensionless).
These expressions, however, have to be calibrated to new hydrological conditions to achieve realistic ANC levels. To illustrate the response of the brakes, the coefficients BC,H, BC,H20, and the concentration, CBC, were manipulated to get a higher brake response at longer residence times. Different brake sets were investigated (Table 7, Fig. 9)
. With proper calibration, these brakes should make it possible to model soil horizons with long residence times, but conducting such a calibration is a challenge that remains to be addressed. If lateral flow was introduced, with an appropriate braking function, the sensitivity of model output to variation in soil depth would be diminished. However, this "gain" would be offset by uncertainties in lateral routing and brake functions.
View larger version (19K):
[in this window]
[in a new window]
|
Fig. 9. Acid neutralizing capacity as a function of residence time in a soil profile using different sets of weathering brakes. Changing the amount of water flow through the profile created the time series.
|
|
The large part of the study area covered by peat (26%) was considered inert, and this is another source of uncertainty. We simply assumed that most runoff water runs through the upper peat layers where the fraction of mineral soil is negligible and the release of organic anions is accompanied by protons, which does not change ANC. However, if the counter ion is a cation such as sodium, ANC will certainly change. It is important to keep in mind the time perspective for such processes as short-term processes should not be taken into account in critical load calculations due to the underlying steady state assumption. How to treat peatland is an important future concern, especially when applying the model in northern Sweden where a large part of the landscape is covered by peat.
Another future field of interest is how to incorporate lake processes that may have a large influence on ANC in a lake system. This issue involves the question of how to handle the difference between runoff chemistry and lake chemistry in critical load calculations.
|
CONCLUSIONS
|
|---|
The traditional methodology of calculating critical loads for an individual lake is not feasible when using the PROFILE model due to the general lack of catchment-specific soil data. Therefore, we have explored the possibility of circumventing this problem by using distributions of crucial parameters that correspond to the region in which a group of lakes is situated. It should be emphasized that our approach explores an alternative way to estimate the weathering rate. Nitrogen retention processes in lakes have to be taken into account using the First-Order Acidity Balance (FAB) model to arrive at a critical load of acid deposition. This "regional distribution" approach is possible since critical loads used in the international negotiations are not for individual sites but for distributions of ecosystems in large areas (50 x 50 km or larger). Furthermore, the statistical design of the Swedish National Survey of Forest Soils and Vegetation, in combination with other data sources, offers an appropriate way to compile a regional database to be used for calculating "soil-based" surface water critical loads, as has been done for forest critical loads.
One of the most important concerns when modeling lake chemistry with PROFILE is to determine the hydrology of the catchment, which is not needed for the application to forest soils (nor is it readily determined from field data). When testing the PROFILE model by comparison with the Diatom model, we used two hydrological extremes, as there are so many possibilities for how to route the water in a soil profile that lie between these two extremes. The diatom predictions (median) were bracketed by the two hydrological extremes using the PROFILE model, but closer to the shallower routing. A calibration revealed that approximately 0.6 m of hydrologically active soil was needed to get the PROFILE predictions to coincide with the Diatom model, with reference to medians of ANCo. This seems reasonable given what is known about the hydrology of the region.
An important issue related to the effective depth is incorporation of lateral flow in the current model formulation. This will diminish the sensitivity in model output caused by variation in soil depth, but raises the question of how to route the water and how to adapt the weathering submodel to waters with residence times of months to decades. Besides the importance of soil depth and hydrology, texture was also found to be of greater importance than mineralogy or soil water content. Mineralogy was not expected to be crucial, though, as the study area is situated on fairly uniform bedrock consisting of granite and greywacke.
At present, there is insufficient evidence to either recommend or reject the PROFILE model for surface water critical loads. The approach presented here must be tested for other regions and the hydrology issue needs to be investigated further before such a judgement can be made. It is clear that PROFILE includes the key long-term process, the weathering rate, quantified from soil properties that are fairly stable over time. This is a great advantage compared with the SSWC method, which relies on more variable surface water input data. A further analysis of PROFILE's ability for lake critical loads would also bring out useful information for other applications of PROFILE and similar models.
|
ACKNOWLEDGMENTS
|
|---|
This work was financed by the Swedish Environmental Protection Agency, Lilli and Oscar Lamms Foundation Stipend and the Environmental Data Centre at the Swedish University of Agricultural Sciences. Part of the dataset used in this publication was made available by the Swedish National Survey of Forest Soils and Vegetation performed by the Department of Forest Soils, Swedish University of Agricultural Sciences. The authors are solely responsible for the interpretation of data.
|
REFERENCES
|
|---|
- Alveteg, M. 1998. Dynamics of forest soil chemistry. Ph.D. diss. Rep. in Ecol. and Environ. Eng., no. 3. Dep. of Chem. Eng. II, Lund Univ., Lund, Sweden.
- Barkman, A. 1997. Applying the critical loads concept: Constraints induced by data uncertainty. Rep. in Ecol. and Environ. Eng., no. 1. Dep. of Chem. Eng. II, Lund Univ., Lund, Sweden.
- Barkman, A., P. Schlyter, M. Lejonklev, M. Alveteg, P. Warfvinge, T. Arnström, and H. Sverdrup. 1999. Uncertainties in high resolution critical load assessment for forest soils: Possibilities and constraints of combining distributed soil modelling and geographic information systems. Geogr. Environ. Model. 3:125–143.
- Barkman, A., P. Warfvinge, and H. Sverdrup. 1995. Regionalization of critical load under uncertainty. Water Air Soil Pollut. 85:2515–2520.
- Bishop, K. 1991. Episodic increases in stream acidity, catchment flow pathways and hydrograph separation. Ph.D. diss. Univ. of Cambridge, England.
- Bishop, K. 1996. Dissolved organic carbon and the modeling of surface water quality. Rep. Ser. no. 35. Swedish Univ. of Agric. Sci., Dep. of For. Ecol., Umeå.
- Bull, K.R. 1995. Critical loads—Possibilities and constraints. Water Air Soil Pollut. 85:2101–2112.
- Cresser, M.C. 2000. The critical loads concept: Milestone or millstone for the new millennium? Sci. Total Environ. 249:51–62.[Medline]
- Cresser, M.S., R. Smart, M.F. Billet, C. Soulsby, C. Neal, A. Wade, S. Langan, and A.C. Edwards. 2000. Modelling water chemistry for a major Scottish river from catchment attributes (Suppl. 1). J. Appl. Ecol. 37:171–184.
- Downing, R.J., J.-P. Hettelingh, and P.A.M. de Smet (ed.) 1993. Calculation and mapping of critical loads in Europe: Status report. Rep. no. 259101003. Coordination Centre for Effects, Natl. Inst. of Public Health and Environ. Protection (RIVM), Bilthoven, the Netherlands.
- Hemond, H.F. 1990. Acid neutralising capacity, alkalinity and acid–base status of natural waters containing organic acids. Environ. Sci. Technol. 24:1486–1489.
- Henriksen, A., M. Forsius, J. Kämäri, M. Posch, and A. Wilander. 1993. Exceedance of critical loads for lakes in Finland, Norway and Sweden: Reduction requirements for nitrogen and sulphur deposition. Acid Rain Res. Rep. 32/1993. Norwegian Inst. for Water Res. (NIVA), Oslo.
- Henriksen, A., J. Kämäri, M. Posch, G. Lövblad, M. Forsius, and A. Wilander. 1990a. Critical loads to surface waters in Fennoscandia, intra- and inter-grid variability of critical loads and their exceedance. Miljörapport 1990:17, Nord 1990:124. Nordic Council of Ministers, Copenhagen.
- Henriksen, A., J. Kämäri, M. Posch, and A. Wilander. 1992. Critical loads of acidity: Nordic surface waters. Ambio 21:356–363.
- Henriksen, A., L. Lien, and T.S. Traaen. 1990b. Critical loads for surface waters—Chemical criteria for inputs of strong acids. Acid Rain Res. Rep. 22/1990. Norwegian Inst. for Water Res., Oslo.
- Hettelingh, J.-P., R.J. Downing, and P.A.M. de Smet (ed.) 1991. Mapping critical loads for Europe. CCE Tech. Rep. no. 1. RIVM Rep. no. 259101001. Coordination Centre for Effects, Natl. Inst. of Public Health and Environ. Protection, Bilthoven, The Netherlands.
- Hodson, M.E., S.J. Langan, and M.J. Wilson. 1996. A sensitivity analysis of the PROFILE model in relation to the calculation of soil weathering rates. Appl. Geochem. 11:835–844.
- Hodson, M.E., S.J. Langan, and M.J. Wilson. 1997. A critical evaluation of the PROFILE model in calculating mineral weathering rates. Water Air Soil Pollut. 98:79–104.
- Jönsson, C., P. Warfvinge, and H. Sverdrup. 1995. Uncertainty in predicting weathering rate and environmental stress factors with the PROFILE model. Water Air Soil Pollut. 81:1–23.
- Kelly, C.A., J.W.M. Rudd, R.B. Cook, and D.W. Schindler. 1982. The potential importance of bacterial processes in regulating rate of lake acidification. Limnol. Oceanogr. 27:868–882.
- Kernan, M.R., T.E.H. Allot, and R.W. Batterbee. 1998. Predicting freshwater critical loads at the catchment scale: An empirical model. Water Air Soil Pollut. 105:31–41.
- Korsman, T. 1999. Temporal and spatial trends of lake acidity in northern Sweden. J. Paleolimnol. 22:1–15.
- Korsman, T., and H.J.B. Birks. 1996. Diatom-based water chemistry reconstructions: A comparison of reconstruction techniques. J. Paleolimnol. 15:65–77.
- Korsman, T., I. Renberg, and N.J. Anderson. 1994. A paleolimnological test of the influence of Norway spruce (Picea abies) immigration on lake-water acidity. The Holocene 4:132–140.
- Kämäri, J., D.S. Jeffries, D.O. Hessen, A. Henriksen, M. Posch, and M. Forsius. 1992. Nitrogen critical loads and their exceedance for surface waters. p. 161–200. In P. Grennfelt and E. Thörnelöf (ed.) Critical loads for nitrogen. Nord 1992:41. Nordic Council of Ministers, Copenhagen.
- Köhler, S., H. Laudon, A. Wilander, and K. Bishop. 2000. Estimating organic acids dissociation in natural surface water using total alkalinity and TOC. Water Res. 34:1425–1434.
- Linder, S., and O. Tenow. 1980. Needle-eating insects and grazing dynamics in a mature Scots pine forest in central Sweden. Ecol. Bull. 32:269–306.
- Munson, R.K., and S.A. Gherini. 1993. Influence of organic acids on the pH and acid neutralizing capacity of Adirondack lakes. Water Res. 29:891–899.
- Mylona, S. 1993. Trends of sulfur dioxide emissions, air concentrations and depositions of sulfur in Europe since 1880. Technical report. EMEP/MSC-W Rep. 2/93. Norwegian Meteorol. Inst., Oslo.
- Posch, M., J.-P. Hettelingh, P.A.M. de Smet, and R.T. Downing. 1997. Calculation and mapping of critical thresholds in Europe: Status Report 1997. RIVM Rep. no. 259101007. Coordination Centre for Effects, Natl. Inst. of Public Health and Environ. Protection, Bilthoven, The Netherlands.
- Rapp, L. 2001. Critical loads of acid deposition for surface water, exploring existing models and a potential alternative for Sweden. Ph.D. diss. Acta Universitatis Agriculturae Sueciae, Silvestria 207. Swedish Univ. of Agric. Sci., Uppsala, Sweden.
- Rapp, L., and K. Bishop. 1995. Sulphur deposition and changes in Swedish lake chemistry 1988–1993. Water Air Soil Pollut. 85:2449–2454.
- Renberg, I., T. Korsman, and H.J.B. Birks. 1993. Prehistoric increases in the pH of acid-sensitive Swedish lakes caused by land-use changes. Nature (London) 362:824–826.
- Rodhe, A. 1988. On the generation of stream runoff in till soils. Nord. Hydrol. 20:1–8.
- Rodhe, H., P. Grennfelt, J. Wisniewski, C. Ågren, G. Bengtsson, K. Johansson, P. Kauppi, V. Kucera, L. Rasmussen, B. Rosseland, L. Schotte, and G. Selldén. 1995. Acid reign 95? Conference summary statement. Water Air Soil Pollut. 85:1–14.
- Schindler, D.W., M.A. Turner, M.P. Stainton, and G.A. Linsey. 1986. Natural sources of acid neutralizing capacity in low alkalinity lakes of the precambrian shield. Science (Washington, DC) 232:844–847.[Abstract/Free Full Text]
- Semenov, M., V. Bashkin, and H. Sverdrup. 2000. Application of biogeochemical model PROFILE for assessment of north Asian ecosystem sensitivity to acid deposition. Asian J. Energy Environ. 1:143–161.
- Skeffington, R.A. 1995. Critical loads and energy policy. p. 91–95. In R.W. Batterbee (ed.) Acid rain and its impact: The critical loads debate. Ensis, London.
- Skeffington, R.A. 1999. The use of critical loads in environmental policy making: A critical appraisal. Environ. Sci. Technol. 33:245–252.
- Stegmann, B. 1998. Positionsbestämning av kraftledningsstolpar med GPS i Västerbotten. (In Swedish.) Swedish Univ. of Agric. Sci., Umeå.
- Sverdrup, H., and P. Warfvinge. 1993. Calculating field weathering rates using a mechanistic geochemical model—PROFILE. J. Appl. Geochem. 8:273–283.
- Sverdrup, H., P. Warfvinge, and D. Britt. 1995. Assessing the potential for forest effects due to soil acidification in Maryland. Water Air Soil Pollut. 83:1–22.
- Sverdrup, H., P. Warfvinge, A. Janicki, R. Morgan, M. Rabenhorst, and M. Bowman. 1992. Mapping critical loads and steady state stream chemistry in the state of Maryland. Environ. Pollut. 77:195–203.
- Swedish Meteorological and Hydrological Institute. 1994. Årmedelvärden 1961–1990 av nederbörd, avdunstning och avrinning. SMHI Hydrologi nr 49. SMHI, Norrköping.
- Taesler, R. (ed.) 1972. Klimatdata för Sverige. (In Swedish.) Svensk Byggtjänst, Stockholm.
- United Nations Economic Commission for Europe. 1993. The state of transboundary air pollution. Update. Air Pollution Studies 9. UNECE, New York.
- United Nations Economic Commission for Europe. 1994. Protocol to the 1979 convention on long-range transboundary air pollution on further reduction of sulphur emissions. UNECE, New York.
- United Nations Economic Commission for Europe. 1999. The 1999 Gothenburg protocol to abate acidification, eutrophication and ground-level ozone [Online]. Available at http://www.unece.org/env/lrtap/multi_h1.htm (verified 2 May 2003). UNECE, New York.
- USDA. 1998. Keys to soil taxonomy. 8th ed. USDA Natural Resour. Conserv. Serv., Washington, DC.
- Warfvinge, P., M. Holmberg, M. Posch, and R.F. Wright. 1992. The use of dynamic models to set target loads. Ambio 21:369–376.[Web of Science]
- Warfvinge, P., and H. Sverdrup. 1992. Calculating critical loads of acid deposition with PROFILE—A steady state soil chemistry model. Water Air Soil Pollut. 63:119–143.
- Warfvinge, P., and H. Sverdrup. 1995. Critical loads of acidity to Swedish forest soils, methods data and results. Reports in Ecol. and Environ. Eng., Rep. 5. Dep. of Chem. Eng. II, Lund Univ., Lund, Sweden.
- Östlund, L., O. Zackrisson, and A.-L. Axelsson. 1997. The history and transformation of a Scandinavian boreal forest landscape since the 19th century. Can. J. For. Res. 27:1198–1206.
Related articles in JEQ:
- This Issue in Journal of Environmental Quality
JEQ 2003 32: 1931-1938.
[Full Text]
TOP
ABSTRACT
INTRODUCTION
