Materials and methods

Site description

Langtjern is a forested, boreal lake catchment in southeast Norway (Figure 1) (4.8 km2; 510–750 m.a.s.l; 60.371 N, 9.727 E) where monitoring of water chemistry started in 1972 under the national monitoring programs for air pollution effects on surface waters (Lund et al. 2018; Vogt & Skancke 2022). The catchment includes a lake (0.23 km2) and its land cover is dominated by forest (75% forested with low productivity Scots pine forest and 5% of productive Norway spruce forest), located on shallow mineral soils, and peatland (20%). Since the 1950s, when some small-scale forest harvest was conducted, the catchment has been free from direct human disturbances. The geology consists of till generated from felsic gneisses and granites (de Wit et al. 2018). Mean annual temperature, precipitation, and discharge (1974-2022) are 2.6°C, 834 mm, and 634 mm, respectively.

Data sources

Streamwater discharge and chemistry
Annual values for discharge and flow-weighted streamwater chemistry (1974–2022) were derived from the routine monitoring conducted by NIVA as part of the Norwegian national environmental monitoring programmes (Vogt & Skancke 2022). Water level is monitored continuously by a weir at the outlet and converted to discharge using standardized stage-discharge relationships. Streamwater samples are collected weekly at lake outlet and analysed for major chemical components (pH, labile Al, alkalinity), major anions (SO4, Cl, NO3), major cations (Ca, Mg, Na, K) and TOC at NIVA. Analytical methods have evolved since monitoring began in 1973 and currently entail automated ion-chromatography. In 1986, monitoring of TOC started. Acid Neutralizing Capacity (ANC) is calculated as the sum of base cations minus the sum of strong acidic anions. ANC corrected for organic acidity, ANCoaa (Lydersen et al. 2004), is calculated as ANC + 1/3*site density of TOC, where site density is -10.2 µeq/mg C. The charge balance of the streamwater chemistry was estimated as the difference between the equivalent sum of all major cations and major anions, and the anion deficit was attributed to organic acidity (OA). Charge density (CD) of TOC was calculated as the anion deficit divided by TOC. For NH4+ and F-, no full time series were available, but earlier data show that their concentrations are low and that they largely outweigh each other in terms of equivalent concentrations. Labile Al was assumed to have charge density of 2+ while the contribution of bicarbonate was set at zero, an acceptable assumption at pH below 5.5.
Element fluxes at the catchment outlet were calculated for each day using measured water discharges at the sampling point multiplied by the solute concentration interpolated from the weekly samples. The daily calculated data were then aggregated to annual fluxes of each element.
Soil element pools
Soils were sampled with the purpose to estimate catchment soil pools of exchangeable cations, and following the same sampling design, in 1983 (Reuss 1990), 1991 (Stuanes et al. 1995) and 2001 (SFT 2001) and reported in Larssen (2005). Standard soil depth for Langtjern was estimated to be 40 cm, including the O horizon. Catchment exchangeable base cation soil pools were calculated by multiplying bulk density with layer depth and soil base cation concentration, where the B horizon was assumed to extend to 40 cm soil depth. Only mineral forest soils were sampled, and the soil base cation stores reported in in Larssen (2005) only represent forest soils.
Atmospheric deposition
Total annual deposition inputs (the sum of wet and dry deposition of SO4, NH4, NO3, Cl and major cations) to MAGIC for the period 1850-2100 are derived from air quality monitoring at the station Brekkebygda (390 masl., 60.29o N, 9.76o E, 7 km south of Langtjern) for 1974-2022, a hindcast for 1860-1973 and projections for 2023-2100. The air quality and deposition monitoring for 1974-2022, resulting in annual wet deposition, follows analytical methods, quality control and flux calculations are described in Aas et al. (2022). The mass balance approach incorporated in the MAGIC model requires adjustment of the wet deposition data, since the deposition station data are not entirely representative for the catchment: dry deposition is not measured, and precipitation amounts may differ from the catchment given the lower elevation of the deposition station. The adjustment is done in two steps. First, dry deposition is estimated using Cl as a conservative tracer, assuming that all Cl originates from sea salt aerosols. For the 1974–2022, the ratio of inputs of Cl (wet deposition) to outputs (streamwater flux of Cl) was 0.898, where the difference (1-0.102 ) is assumed to be from dry deposition. The dry deposition of Cl is assumed to be associated with dry deposition of cations and marine sulphate (mSO4) following their relative proportions in seawater (as given by the following molar elemental ratios: Ca/Cl : 0.037, Mg/Cl : 0.196, Na/Cl : 0.856, K/Cl : 0.018, SO4/Cl: 0.103). Second, we assume that long term catchment outputs of S (in streamwater) equal anthropogenic (SO4*) and marine inputs (mSO4) to the catchment, resulting in the scaling factor 0.98 for anthropogenic SO4 deposition. Historical (1860-1973) and future (2023-2100) deposition of sea salts and base cations was assumed to equal averaged total deposition over 1974-2022. For SO4*, NH4 and NO3, historical deposition was estimated from the modelled deposition for the Langtjern grid square by EMEP based on historical emissions of air pollutants in Europe (Schopp et al. 2003), and scaled to Langtjern using the ratio of the long term averaged EMEP deposition of the grid square and Langtjern deposition, estimated as described above. Future deposition of SO4*, NH4 and NO3 was assumed to follow the European emissions of the CLRTAP current legislation scenario (CLE) for the Langtjern grid square forward to the year 2050 and then held constant to the year 2100.The scenarios were supplied by the Coordination Centre for Effects of the CLRTAP. Future deposition was scaled to Langtjern deposition 1974-2022 data to follow the same relative trends as in the CLE grid square.
Climate
Weather data, including daily average temperature (tm, ̵ͦC), daily precipitation (rr, mm) and evaporation (gwb_eva, mm), was downloaded using NVE’s Grid Time Series (GTS) API. The catchment area of Lake Langtjern was first delineated based on a 10×10 m digital elevation model for Norway (https://www.geonorge.no/, DTM 10 Terrengmodell (UTM33)) as described by de Wit et al. (2023). Weather data were then downloaded for each 1x1km grid (partially or entirely) overlapping with the catchment area and then area-weighted averaged to one value for the catchment. Projected 90% confidence interval of changes in temperature, precipitation, and evaporation for scenario RCP8.5 for 2071-2100 relative to reference period 1971-2000 were taken from Hanssen-Bauer et al. (2015) to generate climate data over 2023-2100. Temperature (delta T), precipitation (delta P) and discharge (delta Q) increases from 1971-2000 to 2071-2100 were +3.0°C to +5.6°C (median: +4.2°C), +8% to +29% (median: +15%) and -2% to +16% (median: +8%), respectively. A locally adapted scenario RCP8.5* was derived by adding delta T, P and Q to 2023 based on the difference between averaged T, P and Q at Langtjern for 2000-2022 and the reference period 1971-2000.

Model description and calibration

MAGIC (Model of Acidification of Groundwater In Catchments) is a process-oriented semi-distributed mass balance model for biochemical processes involving ions and nutrients at catchment scale (Cosby et al. 2001; Cosby et al. 1985), originally used to predict air pollution effects on surface waters and further developed to also include effects of land use and climate change. This setup of MAGIC includes two connected compartments, e.g., soil and surface water (stream or lake), where the soil runoff is sent to the surface water compartment. The solutes (SO4, NO3, NH4, Cl, Ca, Mg, Na, and K) concentrations are computed as the mass balance between atmospheric deposition, bedrock weathering, retention (for N species) and export through runoff. A new version of MAGIC, MAGIC-Forest (Norling et al. this issue) with modules for hydrology, forest growth, soil carbon accumulation and SO4-dependent organic matter solubility, has been implemented in the open-source Mobius modelling framework (Norling et al. 2021). The Mobius framework has a modern graphical user interface and scripting for interaction with models, allowing for user-friendly advanced auto-calibration and sensitivity analysis.
In the current MAGIC application, we start from the calibration conducted by Larssen (2005) which described streamwater chemistry responses to reductions in S deposition during the period 1974–2003. We included monitoring data for the subsequent period 2004–2022. New features in MAGIC used in the current application are i) the SO4-solubility control of TOC (\(c_{\text{OA}}=c_{OA,0}-c_{\text{SO}_{4}}f_{\text{SO}_{4}})\)background concentrations of organic acids reduced with factor fSO4 multiplied with SO4) (Norling et al. this issue) and ii) a factor to change weathering rates (only used in a sensitivity analysis;\(w_{y}=w_{0}(1+max(0,\left(y-y_{0}\right){)\alpha}_{w})\), i.e. the rate stays constant at \(w_{0}\) until year \(y_{0}\), and increases by a linearly with slope \(\alpha_{w}\) after that). Table SI 1 describes which parameters are fixed and which are calibrated. Soil characteristics are based on soil element pools as described earlier. The model was calibrated manually, and optimization of the model parameters (Table SI 1) was done by comparing observed and modelled annual FWM concentrations of ANC, H+, labile Al, SO4, NO3, Cl, Ca and Mg and organic acids. During calibration, we also ensured that other modelled variables were showing an acceptable fit to observations such as total and cation-specific soil base saturation and FWM concentrations of Aln+, Na and K. The best set of parameters was selected using expert judgement anchored in an evaluation of combined performance metrics: the Nash-Sutcliffe Efficiency (NSE), the root-mean square error (RMSE), the coefficient of determination (r2) and the bias (summation of difference between model and empirical estimate, divided by nr of estimates), in addition to comparison with soil base saturation. The automated optimization implemented in Mobius resulted in similar model performance as the manual calibration for base cation weathering rates, but manual optimization was chosen over automated since the automated optimization did not result in converging parameter values.

Scenarios

Scenarios for future climate and deposition always used CLE for future deposition. The scenarios are referenced as ConstantClimate (CLE deposition and no climate change), RCP8.5 (CLE scenario and RCP8.5) and RCP8.5* (CLE scenario and RCP8.5*).