Comparison of simulated forest soil response to acid deposition reduction with two models of differing complexity

Abstract. Great effort has been dedicated to developing soil acidification models for use on different scales. This paper focuses on the changes in model performance of a site scale soil acidification model (NUCSAM) and a national to European scale soil acidification model (SMART 2). This was done to gain insight into the effects of model simplification. Because these models aim to predict the response to reduction in acid deposition, these models must be tested under such circumstances. A straightforward calibration and validation of the regional model, however, is hampered by lack of observations over a sufficient time period. Consequently, NUCSAM was calibrated and validated to a manipulation experiment involving reduced acid deposition in the Speuld forest, the Netherlands. SMART 2 was then used with calibrated input data from NUCSAM. The acid deposition was excluded by a roof beneath the canopy. The roofed area consists of a plot receiving pristine deposition levels of nitrogen (N) and sulphur (S) and a control plot receiving ambient deposition. NUCSAM was calibrated on the ambient plot, followed by a validation of both models on the pristine plot.
Both models predicted soil solution concentrations within the 95% confidence interval of the observed responses for both the ambient plot and the pristine plot at 90 cm depth. Despite the large seasonal and vertical (spatial) variation in soil solution chemistry, the trends in annual flux- weighted soil solution chemistry, as predicted by SMART 2 and NUCSAM, corresponded well.The annual leaching fluxes below the root zone were also similar although differences exist for the topsoil. For the topsoil, NUCSAM simulated the nutrients and acid related constituents better than SMART 2. Both models overestimated the ammonium (NH₄) concentration at 10 cm depth. SMART 2 underestimated calcium and magnesium (BC²⁺) concentration at 10 depth, whereas NUCSAM overestimated BC²⁺ concentration at 90 cm depth. NUCSAM predicted the effect of deposition reduction on N concentrations at both depths, whereas SMART 2 underestimated the effect of deposition reduction at 10 cm depth. Both models predicted faster effects of deposition reduction on aluminum (Al), sulphate (SO₄) and base cations than was observed. Generally, it appeared that the differences were large during the period of profound deposition changes whereas small differences occurred during slight variations in deposition level. It is concluded that a simpler model description does not affect the model's performance significantly as regards flux-weighted annual average concentrations at greater depth. Model improvements must focus on processes related to N-dynamics.