Introduction

Human-induced atmospheric deposition of nitrogen (N) and sulphur (S) compounds have significantly altered nutrient cycling in temperate forest ecosystems and caused soil acidification and eutrophication. As a consequence, the quantification of atmospheric deposition has gained importance to establish cause-effect relationships, assess temporal and spatial trends and evaluate abatement measures and mitigation strategies (de Vries et al. 2003a; Erisman et al. 1994). Wet deposition can be measured with reasonable accuracy using wet-only or bulk precipitation collectors (Staelens et al. 2005), whereas for dry deposition (DD) a trade-off needs to be made between measurement accuracy and monitoring effort/costs (Erisman et al. 1994). The choice of a certain DD method depends on the purpose of the study and throughfall (TF) measurements are considered to be more suitable for long-term monitoring purposes and large scale monitoring networks than micrometeorological measurements (de Vries et al. 2003b; Erisman et al. 1994). Several studies have reported the use of TF measurements to evaluate trends in atmospheric deposition on forests (Graf Pannatier et al. 2011; Vanguelova et al. 2010). Furthermore, the method has often been used to compare deposition on different forest types, and deciduous versus coniferous forest stands in particular (De Schrijver et al. 2007; de Vries et al. 2007). An additional advantage of the TF method is that it also gives information on the internal nutrient cycle in forests (Emmett et al. 1998; Ferm and Hultberg 1999; Friedland et al. 1991; Parker 1983; Neirynck et al. 2008).

When incident precipitation passes through the canopy it is altered by wash-off of gases and particles deposited in dry periods prior to the precipitation event and by ion exchange, i.e. uptake or leaching, between canopy surfaces and the solutions passing over them (Draaijers et al. 1997). Therefore, to quantify total atmospheric deposition with TF measurements it is necessary to distinguish DD from canopy exchange (CE) (Parker 1983). This distinction is made by the canopy budget model of Draaijers and Erisman (1995), in which ion-exchange processes between the canopy and TF water are estimated. Various approaches of this model with respect to the time step, type of open-field precipitation data, tracer ion and ion exchange processes have been reviewed by Staelens et al. (2008). This review also assessed the sensitivity of atmospheric deposition onto two deciduous canopies to each approach compared to the reference model of Draaijers and Erisman (1995). The DD of base cations was meaningfully affected by the type of precipitation data and the tracer ion used, while canopy uptake of ammonium (NH4 +) and protons (H+) was influenced by accounting for canopy leaching of weak acids (WA). However, the authors suggested that future applications of the model could benefit from combining different approaches with each other to quantify the range of estimated DD and CE. This would also allow testing the relative influence of each approach on total N and potentially acidifying deposition.

Furthermore, Staelens et al. (2008) indicated that assuming no canopy uptake of nitrate (NO3 −) (Harrison et al. 2000) had little effect on the total N deposition when the relative uptake efficiency of NO3 − was considered to be low, but that more research was needed with regard to the uptake efficiency of NH4 + compared to NO3 − for varying tree species and environmental conditions. Adriaenssens et al. (2012b) determined NH4 +/NO3 − retention ratios from wet deposition by means of 15N labelled sources for four different tree species and phenological stadia. In this study it was also observed that inorganic N compounds were merely adsorbed to woody plant surfaces instead of being assimilated and that no relation between 15NH4 + uptake and net throughfall (NTF) of base cations was observed during the leafless season. This indicates that other processes than ion-exchange processes might play a role during the leafless season of deciduous species and that the ion-exchange assumptions made in the model might not be valid.

Therefore, in the present study we applied a combination of the different assumptions that can be made at each step in the canopy budget model (Draaijers and Erisman 1995) to the rainfall and TF data of three case studies. The steps containing the approaches that showed a high impact on CE and DD calculations according to Staelens et al. (2008) were included again, but now combined with each other. In addition, the effects of two new assumptions were tested, i.e. the NH4 +/NO3 − uptake ratio (Adriaenssens et al. 2012b) and a diameter-weighted DD factor for K+ (Adriaenssens 2012). We explored the variation in output of 600 model variants obtained by combining different model approaches and assumptions and assessed the effect of each of these approaches on (i) the range of CE, DD and total N and acidifying deposition fluxes at the level of an individual beech tree, (ii) the ratio of total N and acidifying deposition on a coniferous and deciduous stand and (iii) a time trend analysis of total N and acidifying deposition onto a mixed deciduous stand.

Materials and methods

Study sites and sample collection (see Supplementary Material for a more extensive description)

All study sites were located in Flanders, the northern part of Belgium. The region is characterized by high atmospheric N and S deposition. The European beech (Fagus sylvatica L.) tree as well as the mixed deciduous stand used for the time trend analysis both were located in the Aelmoeseneie forest (50°58.5′N, 3°48′E, 16 m a.s.l.). This is a mixed deciduous forest located near Ghent in northern Belgium, approximately 60 km from the North Sea. TF water under the canopy of the individual beech tree at 1.5 m height and bulk precipitation above the canopy at 36 m height were collected every fortnight from 22 April 2009 to 22 April 2010 by six and two funnels, respectively. Based on visual observation of the beech canopy, the following phenological periods were distinguished: leaf development (22 April–20 May 2009), fully leafed period (21 May–24 September), leaf senescence (25 September–19 November) and leafless period (20 November 2009–22 April 2010).

In the mixed oak–beech stand, precipitation, TF and stemflow (SF) were collected from 1994 till 2010 according to the guidelines of Clarke et al. (2010). The different phenological stages in this study were delineated based on K+ TF deposition, which was shown to increase during leaf development and leaf senescence (Houle et al. 1999; Neary and Gizyn 1994; Staelens et al. 2007).

The study site for the deposition ratio between a coniferous and a deciduous stand is located in the nature reserve ‘Heidebos’ in northern Belgium (Wachtebeke-Moerbeke) (51°11′N, 3°55′E, 11 m a.s.l.). Here, adjacent monospecific forest stands of pedunculate oak and Scots pine (Pinus sylvestris L.) were selected with the same soil type, stand history and tree age. Bulk precipitation and TF water were sampled biweekly from 7 December 2007 until 3 December 2008, by means of 4 and 15 collectors respectively. Based on visual observations of the oak canopy, the following phenological periods were distinguished for the oak TF data: leafless period (7 December 2007–7 May 2008), leaf development (8 May–5 June), fully leafed period (6 June–8 October) and leaf senescence (9 October–3 December 2008).

For all samples, the volume was determined and pH and electric conductivity were measured. After filtering through a 0.45 μm nylon membrane filter, NO3 −, SO4 2−, PO4 3−, Cl−, NH4 +,K+, Ca2+, Mg2+ and Na+ concentrations were determined. H+ concentrations were derived from the pH measurements.

Canopy budget models

The chemical composition of TF and SF water under a forest canopy is the result of incident precipitation, wash-off of dry deposited gases, particles or cloud droplets prior to the precipitation event, and the exchange between the canopy surfaces and the solutions passing over them (Lovett et al. 1996):

$$TF + SF = TD + CE = PD + DD + CE $$
(1)

where TD is the total deposition, PD is the precipitation deposition, DD is the dry deposition and CE is the canopy exchange.

Total potentially acidifying deposition (TDac) can then be defined as the sum of TD of NO3 − and NH4 + (TDN), SO4 2− (TDS) and Cl− (TDCl) corrected for the neutralizing effect of base cations (TDBC; Na+, K+, Ca2+ and Mg2+) (UBA 2004):

$$TD_{ac} = TD_{N} + TD_{S} + TD_{Cl} -TD_{BC} $$
(2)

The NTF of an ion is defined as the difference between TF (+SF; if available) and PD, and equals the sum of DD and CE.

$$NTF = TF + \left( {SF} \right) - PD = DD + CE $$
(3)

The aim of the canopy budget model is to distinguish DD from CE for all major ions. Positive values for CE represent canopy leaching (CL) and negative values canopy uptake (CU). However, we follow the convention of expressing CU as positive values (CU = −CE). In the model, all fluxes are expressed on an equivalent basis (molc) per unit ground surface area and time. Further on, SF is not explicitly mentioned, but is included in TF for datasets where this flux is available.

When using the model, various approaches can be chosen within each step. For a thorough overview and discussion of these approaches, we refer to Staelens et al. (2008) and the Supplementary Material. In the present study, the effect of several of these approaches reported in literature was assessed for the three case studies. Figure 1 and Table 1 give an overview of the tested approaches. The model was tested with a phenological, semi-annual and annual step and with both bulk precipitation data and wet-only precipitation data as input. Furthermore, the effect of using Na+ and SO4 2− as a tracer ion was tested. In addition, based on previous research (Adriaenssens 2012) a new approach was tested which uses a diameter-weighted DD factor for K+. WA were first excluded from the model, and then included again but calculated with three different methods, i.e. based on (i) the cation–anion balance, (ii) DOC and alkalinity and (iii) DOC and pH. In the reference canopy budget model, no CU of NO3 − is taken into account, although several studies have indicated that forest canopies can incorporate NO3 − (Adriaenssens 2012; Dail et al. 2009). To calculate CU of NO3 −, four different approaches have been reported. In the first and second approach, CU of NH4 + and H+ is set equal to CL of base cations and/or CL of WA, while in the third and fourth approach only CU of NH4 + is set equal to CL of base cations and/or CL of WA. CU of H+ is then set equal to CU of NO3 −. Additionally, in the first and third approach CU of NO3 − is calculated based on a relative uptake efficiency of NH4 + to NO3 − that equals 6, while in the second and fourth approach, this relative uptake efficiency varies by tree species and phenological period based on previous studies (Table 2; Adriaenssens et al. 2012b). Finally, previous research indicated that no CU of NH4 + and NO3 − occurs during the leafless season for deciduous species (Adriaenssens et al., 2012b), so this was included in as well excluded from the model.

Fig. 1
figure 1

Flow chart of the canopy budget model, extended from Staelens et al. (2008). The measured deposition of major ions in precipitation deposition (PD) and throughfall (+stemflow) deposition (TF (+SF)) is used to calculate NTF deposition, which consists of dry deposition (DD) and canopy exchange (CE). Ellipses are used to indicate model assumptions about the deposition factor (DDF) and on CE processes. Rounded rectangles indicate model assumptions for which various approaches were tested. Dotted horizontal lines separate equations used for different ions or ion groups. WA weak acids, Na + adj. Na+ as tracer ion with adjusted DD of K+, BC base cations, DOC dissolved organic carbon, xH exchange efficiency of H+ relative to NH4 +, xNH 4 exchange efficiency of NH4 + relative to NO3 −

Full size image
Table 1 Overview of the different tested approaches at each step of the canopy budget model (see also Fig. 1)
Full size table

Data analysis

The combination of all approaches listed above (Table 1) resulted in 720 model versions. Since the approach of no CU of NH4 + and NO3 − during the leafless period could not be calculated with an annual time step, 600 models were retained. For the case studies on the beech tree and the coniferous versus deciduous forest stands no DOC data were available, so for these datasets we evaluated 300 different models. For the comparison of forest types, deposition ratios between the coniferous and deciduous stand were evaluated. Time trends in annual deposition fluxes on the oak–beech stand were investigated through Kendall tests (Kendall 1975) to detect monotonically increasing or decreasing trends. Kendall’s tau statistic represents the non-parametric correlation between a time series of data and time. When a linear trend can be assumed, the median slope and trend line is calculated non-parametrically.

Table 2 Ammonium to nitrate uptake ratios used for the calculation of the different canopy budget model versions, based on the results of Adriaenssens et al. (2012b)
Full size table

The effect of different model versions was evaluated for two summarizing variables, i.e. the total N deposition (NH4 + + NO3 −) (TDN) and the potentially acidifying deposition (TDac; Eq. 2 in “Canopy budget models” section). For each time step considered (Table 1), the effect of different steps in the model, i.e. the type of precipitation data, the tracer ion, the CL of WA, the NO3 − uptake and the CU of N during the leafless season (Table 1) was assessed by means of a multi-way Analysis of Variance (ANOVA). To evaluate model steps for the time trend analysis, a semi-parametric Permanova test was performed on Kendall’s tau and the median slope (Anderson 2001). The effect of time step was assessed by a one-way ANOVA. All calculations and statistical analyses were performed with R version 2.13.0 (R Development Core Team 2011).

The obtained range in TDN and TDac of the different stands in this study was compared with site-specific critical loads of acidity and nutrient N (eutrophication), determined with the widely-used simple mass balance model (UBA 2004). Input data were as much as possible derived from site-specific measurements or regional data, as described by Staelens et al. (2009).

Results

Deposition onto an individual beech canopy

Application of the 300 different canopy budget models on the data for the beech canopy resulted in CE ranging from −20 to 0 kg N ha−1 yr−1 and DD ranging from 6 to 28 kg N ha−1 yr−1 (see Supplementary Material Fig. S1). The highest CU was calculated using Na+ as a tracer (without adjustment for a different DD rate of K+), excluding CL of WA and including CU of N during the leafless season. The lowest CU was obtained with SO4 2− as a tracer ion, including CL of WA, using bulk precipitation data and with a phenological time step. This lower CU of N was caused by occasionally negative CL of base cations or a high value for the estimated CL of WA, especially during leaf development. Lowest DD of N was calculated using SO4 2− as a tracer ion, including CL of WA and using bulk precipitation data, while the highest DD was obtained with Na+ as a tracer ion, excluding CL of WA, using wet-only precipitation data and including CU of N during the leafless season. When using a phenological time step, calculated DD of N during the leaf development period was negative when CL of WA was included.

The highest variance in total N deposition (TDN) could be explained by CL of WA, followed by CU of N during the leafless season (in case of a phenological or semi-annual time step), the tracer ion used and the precipitation data. For total acidifying deposition (TDac) the highest variance could be explained by the tracer ion used, CL of WA and CU of N during the leafless season (Table 3). For TDN, all factors were significant while for TDac the effect of NO3 − uptake was not significant in the semi-annual and annual time step. The time step itself had no significant effect on TDN (p = 0.838) and TDac (p = 0.998). Including CL of WA significantly lowered CU of NH4 + and consequently lowered DDN, TDN and TDac (Fig. 2). The use of SO4 2− as a tracer ion increased estimated DD of base cations compared to Na+, hereby decreasing CL of these base cations and CU and DD of NH4 +. This decreased TDN but increased TD of base cations and consequently reduced TDac. Using bulk precipitation data decreased TDN and TDac estimates. The CU of NO3 − explained the least variance in the model results. A significant difference between excluding and including CU of NO3 − was only found for TDN.

Table 3 p Values of the multi-way ANOVA analysis assessing the effect of different steps in the canopy budget model on total nitrogen (N) and total acidifying deposition to an individual beech tree (Case 1), the ratio of total N and total acidifying deposition to a coniferous and a deciduous stand (Case 2); The effect on Kendalls tau of total N and total acidifying deposition on a deciduous oak–beech stand (Case 3) was assessed by a semi-parametric Permanova test
Full size table
Fig. 2
figure 2

Box plots showing the effect of each step in the canopy budget model on total nitrogen (N) deposition (kg ha−1 yr−1) and total potentially acidifying (molc ha−1 yr−1) deposition on the individual beech tree. Stars indicate significant differences between different levels of each step

Full size image

Deposition ratio between a coniferous and a deciduous stand

Application of the 300 canopy budget models resulted in a ratio of TDN (coniferous to deciduous forest stand) between 1.02 and 1.75 and a ratio of TDac (coniferous to deciduous) between 0.68 and 1.46 (see Supplementary Material Fig. S2). The highest variance on the ratios of TDN and TDac could be explained by CU of N in the leafless season followed by the CL of WA and the tracer ion (Table 3). No significant effect of NO3 − uptake was found, except with an annual time step, and the type of precipitation data only significantly affected TDN deposition ratios for an annual time step. The time step significantly affected TDac (p < 0.001) but not TDN (p = 0.128). Including CL of WA increased the coniferous to deciduous ratio of TDN and TDac because TDN decreased relatively more for pedunculate oak than for Scots pine (Fig. 3). The use of SO4 2− also increased the ratio of TDN and TDac because the deposition ratio of SO4 2− was lower than of Na+. Excluding CU of N during the leafless season significantly decreased TDN on the pedunculate oak stand and consequently increased the ratio of TDN and TDac. Wet-only precipitation data yielded a higher TDac ratio than bulk precipitation data and a phenological time step significantly decreased the TDac ratio compared to an annual or semi-annual time step. It needs to be mentioned that in the pedunculate oak stand calculated DD values of N during leaf development were negative when CL of WA was included. For the Scots pine stand, CE of Ca2+ and/or Mg2+ was always negative during leaf development and senescence and CE of NO3 −, NH4 + and H+ was always positive during leaf development and when CL of WA was included also during other phenological periods.

Fig. 3
figure 3

Box plots showing the effect of each step in the canopy budget model on the ratio of total nitrogen (N) deposition and total potentially acidifying deposition between a Scots pine and a pedunculate oak stand. Stars indicate significant differences between different levels of each step

Full size image

Determination of trends in deposition

According to the Kendall tests, yearly TDN on the beech site calculated by the 600 different canopy budget models showed a significant (p < 0.05) time trend over 1999–2010 for 66 % of the models. Kendall’s tau value varied from −0.76 for the highly significant time trends to −0.03 for non-significant time trends. The median slope varied from −1.75 to −0.05 molc N ha−1 yr−1 (Fig. 4). Yearly TDac showed a significant (p < 0.05) time trend for 58 % of the tested models. Kendall’s tau value varied from −0.70 to −0.12 and the median slope from −197 to −26 molc N ha−1 yr−1 (see Supplementary Material Fig. S3).

Fig. 4
figure 4

Time series of total nitrogen (N) deposition and total potentially acidifying deposition calculated with canopy budget models using varying a precipitation data, b tracer ion, c canopy leaching of weak acids and d canopy uptake of N during the leafless season. The estimated median slope K (kg N ha−1 yr−1) of each approach is given

Full size image

The variation in Kendall’s tau values for TDN and TDac could mainly be explained by the CL of WA followed by the type of precipitation data, the tracer ion and the CU of N in the leafless season, although the last factor was not significant for a semi-annual time step (Table 3). Choices regarding the CU of NO3 − had no significant effect on the tau value of TDN and TDac. Using wet-only precipitation data decreased TDN by 13 % from 1999 to 2010, while TDN was reduced by 11 % with bulk precipitation data. Similarly, TDac decreased by 22 and 19 % using wet-only and bulk data, respectively (Fig. 4). Using SO4 2− as a tracer reduced TDN and TDac with 21 and 34 %, respectively, compared to 8 and 13 %, respectively, for Na+. Furthermore, no significant trend was found for TDN and TDac when CL of WA was estimated from the difference between cations and anions (−15 % TDN and −3.5 % TDac; in 12 years). In contrast, this trend was highly significant when WA were estimated from DOC and alkalinity measurements (−25 % TDN and −30 % TDac). Excluding CL of WA and including WA estimated from DOC and pH measurements resulted in intermediate but significant time trends. Estimated WA in TF were similar in magnitude for DOC combined with alkalinity measurements compared with the cation–anion balance, while in PD estimated WA from DOC and alkalinity were approximately four times higher. Excluding CU of N during the leafless season reduced TDN and TDac with 12 and 20 %, respectively, and including with 14 and 22 %, respectively.

Discussion

Effect of time step, type of precipitation data and tracer ion

Using different time steps in the canopy budget model allowed introducing different CE approaches, which are discussed further in this paragraph. However, there was also an effect of time step as such, with phenological or semi-annual time steps generally resulting in lower N and potentially acidifying deposition (TDN and TDac, respectively), lower TDac ratios between a coniferous and deciduous stand and in more steeply decreasing deposition time trends, i.e. lower Kendall tau values and median slopes. The negative DD values of NO3 −, NH4 + and H+ calculated during leaf development may be caused by 1) underestimated CL of base cations which is in turn caused by overestimated DD of base cations or 2) overestimated CL of WA by the cation–anion balance. The first cause was confirmed by negative CE values of Ca2+ and Mg2+ during leaf development. This indicates that both Na+ and SO4 2− may be subject to leaching during leaf development (Reiners and Olson 1984; Staelens et al. 2007), which makes them less suitable as tracer ion. Staelens et al. (2007; 2008) suggested solving this by calculating the DDF of Na+ for the fully leafed period and using this factor for the periods of leaf development and senescence. However, for the Scots pine stand this would not be sufficient as CE of Ca2+ and Mg2+ was also negative during the leafless period. Hence, more research is needed on the assumption of inertness of the tracer ion and the deposition efficiency of different particles.

Although the type of precipitation data was generally not the main source of variation between model versions, it significantly affected the calculated TDN and TDac in the three case studies. In particular for Na+, Cl−, K+, Ca2+ and Mg2+, which had low wet-only to bulk ratios, the NTF was significantly reduced using bulk instead of wet-only precipitation deposition. This in turn reduced CE of base cations and consequently CE and DD of N. As the wet-only to bulk ratios used in this study were generally in line with other reports (see Table 2 in Staelens et al. 2005), this conclusion may held true for other studies as well. In general, the difference in TDN or TDac between bulk and wet-only precipitation data increased when TF deposition increased, i.e. TF of Scots pine was higher than of pedunculate oak and higher in 1999 compared to 2010.

Using SO4 2− as tracer ion may significantly overestimate DD of base cations if meaningful gaseous SO2 deposition occurs, and consequently underestimate DD of N (see also Sect. 1.3.3 in the Supplementary Material). This is confirmed by the present study: TDN to the beech canopy was significantly lower with SO4 2− than with Na+ as a tracer ion; and SO2 deposition at this site was estimated to be 50 mmolc m2 yr−1, while almost no particulate SO4 2− was deposited on artificial foliage (Adriaenssens 2012). The TDN and TDac ratios between a coniferous and deciduous stand were significantly affected by the tracer ion due to the fact that the TF deposition ratio of SO4 2− was lower than for Na+. In the time trend analysis of the oak–beech stand, more steeply decreasing trends were observed with SO4 2− as tracer ion since the precipitation and TF deposition decreased more significantly from 1999 to 2010 for SO4 2− than for Na+. It is clear that for regions with high SO2 deposition, Na+ can be considered to be a more reliable tracer ion. In some cases, such as in coastal regions (Adriaenssens et al. 2012a) the use of SO4 2− may be justified if particle deposition dominates DD of S, but then still the generally finer SO4 2− particles (Lindberg et al. 1986; Ruijgrok et al. 1997) will be deposited with lower deposition efficiency (Slinn 1982) and results need to be handled with care.

In the three case studies there was no significant effect between using Na+ as a tracer ion and using Na+ adjusted for a lower DD of K+. This was mainly due to the fact that atmospheric K+ deposition was low, so that the impact of a reduced DD of K+ was not detectable in the output of the canopy budget model (here summarized by TDN and TDac). However, TDac for the individual beech tree and the ratio of TDac between the coniferous and deciduous stand were slightly increased when using Na +adj . We only adjusted the DDF of K+ as this was suggested by several authors (Ruijgrok et al. 1997; Staelens et al. 2008) and also concluded from own results (Adriaenssens 2012). Furthermore, the ratio between the MMD of K+ and Na+ was based on data of Ruijgrok et al. (1997) assuming a linear and 1:1 relationship between particle diameter and deposition efficiency for particles larger than 0.1 μm, while, according to the model of Slinn (1982) or other models (Ruijgrok et al. 1995), this relationship is more exponential. Since the steepness of this curve varies considerably among existing models that all show a large uncertainty due to variable input parameters (Ruijgrok et al. 1995) and since a more linear pattern was observed from deposition measurements (Garland 2001), our choice for a linear and 1:1 relationship in this study seems justifiable. However, future research could benefit from assessing the influence of different MMD – deposition velocity relationships for various conditions, e.g. particular meteorological conditions or different surface geometry (Garland 2001) if information on the MMD of the different aerosols is available. Further studies may suggest an adjusted DDF for Ca2+ and Mg2+ too, which will increase the impact on TDN and particularly TDac.

Effect of canopy exchange calculations: canopy leaching of weak acids, NO3 − uptake and CU of N during the leafless season

In the three case studies, a strong effect of in- or excluding CL of WA estimated by the cation–anion balance was observed. Especially for beech and oak significant CL of WA was calculated, while this was less important for the pine stand. This is in agreement with a higher base cation leaching from deciduous than coniferous canopies, as was found in our case study of a paired deciduous–coniferous stand and in other reports (De Schrijver et al. 2007; Rothe et al. 2002). In the trend analysis, CL of WA decreased significantly from 1999 to 2010, hereby decreasing the downward trend of TDN. This high impact of including WA in the canopy budget model illustrates its importance, especially in deciduous stands where CL of WA contributed on average 50 % to CL of base cations. Estimating WA by means of the cation–anion balance was considered to be a reliable approach by de Vries et al. (2001), given that all other major ions are analysed precisely and accurately (Staelens et al. 2008). However, in the trend analysis the cation–anion approach differed significantly from the two DOC-based approaches. In general, measuring WA can be preferred above estimating them. However, estimated HCO3 − from alkalinity may contain some organic acids and calculated HCO3 − from pH may be less accurate due to the uncertainty in pH measurements of weakly buffered water samples like precipitation samples. Hence, more research is needed with regard to the estimation of WA in precipitation and TF samples.

Adriaenssens et al. (2011; 2012b) observed low but significant uptake of wet deposited NO3 − by tree leaves, particularly during leaf development. Incorporating NO3 − uptake in the canopy budget model significantly increased the calculated TDN on the individual beech tree and is recommended given the clear experimental evidence. Nevertheless, including NO3 − uptake had neither effect on the comparison between a coniferous and a deciduous stand nor on the trend analysis and TDN did not differ between the different calculation NO3 − uptake approaches. This suggests that the proposed xNH4 efficiency factor of six by de Vries et al. (2001) may also be suitable and that no differentiation according to leaf phenology is necessary. However, the studied tree species were all shown to preferentially retain NH4 + compared to NO3 −. The effect of the different calculation approaches of NO3 − uptake will likely increase for species that have been shown to preferentially retain NO3 − in their foliage instead of NH4 +, such as red spruce (Picea rubens Sarg.) (Dail et al. 2009; Gomez-Guerrero et al. 2008). Therefore, it is still suggested to use a tree species specific xNH4 at a phenological time step when available. Furthermore, attributing CE of H+ and NH4 + to CL of WA often resulted in (impossible) negative DD values of H+ for the individual beech tree and the trend analysis, while this was generally not the case when CE of H+ was determined via the CE of NO3 −.

In this study we introduced the new concept of excluding the CU of NH4 + and NO3 − during the leafless season for deciduous species, based on previous observations that twigs and stems of deciduous trees retained almost no N (Adriaenssens et al. 2012b). As expected, this significantly reduced TDN and TDac for the individual beech tree (Fig. 2) and the ratios of TDN and TDac between a coniferous and deciduous stand (Fig. 3). It also decreased the estimated tau value and estimated slope in the trend analysis for TDN. However, in mature temperate deciduous stands some N can be retained by microbial assimilation processes and lichens (Reiners and Olson 1984), which mainly occur on older trees (Fritz et al. 2008; Ranius et al. 2008). Several studies have reported a significant contribution of N assimilation by epiphytic lichens to total CU in coniferous (Johansson et al. 2010; Lang et al. 1976) and tropical forests (Clark et al. 2005; Wanek and Hinko-Najera Umana 2010), but not in deciduous stands. If the latter is also significant, which needs to be investigated, attributing all NTF of N to DD during the leafless period will likely underestimate TDN. For the Scots pine stand used in this study, CU of N in winter was not excluded since needles are then still present and since significant amounts of 15N were found to be adsorbed to the bark surface, as suggested by Dail et al. (2009) and Wilson and Tiley (1998). This retained N is likely not subject to ion exchange processes with K+, Ca2+ and Mg2+ as assumed in the canopy budget model. However, this remains to be investigated.

It should also be mentioned that the CU of N calculated by the canopy budget model does not comprise immediate uptake of gases like HNO3, but only ion exchange between the water layer on the canopy surface and the canopy itself. As a consequence, DDN, TDN and TDac could be underestimated. However, the few available studies that compared N uptake from dry and wet deposition concluded that foliar retention from the liquid phase is much more important than from the gaseous phase (Harrison et al. 2000; Sievering et al. 2007), although dry deposited N may significantly contribute to canopy retention (Horvath 2004). Dry deposited N to the canopy during the dry period prior to a wet deposition event is generally mobilized by canopy wetness, after which the dissolved N is available to canopy foliage for use in photosynthesis (Sievering et al. 2007). From this we conclude that the canopy budget model takes into account most of the canopy N uptake.

Comparison with literature data and site-specific critical loads

Based on this study and other literature data (Ignatova and Dambrine 2000; Staelens et al. 2008; Zhang et al. 2006), some canopy budget model approaches can be considered more reliable than others, e.g. Na+ is a more suitable tracer ion than SO4 2−, and the use of wet-only precipitation data and including CL of WA is recommended. If we only consider these models, DD to the beech canopy was 10–20 kg N ha−1 yr−1,TDN was 19–29 kg N ha−1 yr−1 and TDac 1664–2285 molc ha−1 yr−1. This range was comparable at the Heidebos site for the oak stand (DDN: 8–16 kg N ha−1 yr−1; TDN: 20–28 kg N ha−1 yr−1; TDac: 1357–1902 molc ha−1 yr−1) but higher for the pine stand with regard to N (DDN: 23–24 kg N ha−1 yr−1; TDN: 34–35 kg N ha−1 yr−1; TDac: 1612–1895 molc ha−1 yr−1). For the oak–beech stand in the Aelmoeseneie forest in 2010 DDN was 15–31 kg N ha−1 yr−1, TDN was 25–41 kg N ha−1 yr−1 and TDac was 2228–3382 molc ha−1 yr−1. For the individual beech tree, DDN covers the results obtained by multi-layer artificial foliage and by passive sampler measurements multiplied with a fixed deposition velocity (Adriaenssens 2012). Neirynck et al. (2008) reported much higher DDN and TDN values for a Scots pine stand in the same region, however, this stand was 70 years old and consequently had a higher collecting surface for DD. Moreover, the proximity of a harbour causes higher local emissions of NOx and SO2. Canopy N uptake by the individual beech tree ranged from 4 to 14 kg N ha−1 yr−1, which is within the range of 1–57 kg N ha−1 yr−1 reported in literature (Sievering et al. 2007; Neirynck et al. 2008). The ratio of TDN between a coniferous and deciduous stand was within the range reported in meta-analyses on this topic (De Schrijver et al. 2007; Rothe et al. 2002). To our knowledge, no such comparison has been done for TDac.

The modelled critical load of acidification was 1960–2285 molc ha−1 yr−1 for the deciduous species at the two sites and 1294 eq ha−1 yr−1 for pine at the Heidebos site (Table 4) The critical loads for eutrophication amounted to 13–14 (deciduous species) and 8 (pine) kg N ha−1 yr−1. The obtained ranges for TDN reported above all exceed the critical loads for eutrophication, indicating that eutrophication is very likely in all three cases. Yet, critical loads for acidification are not all exceeded by the ranges of TDac. For instance, for the Heidebos site the obtained ranges are below the critical loads and for the beech canopy approximate around the critical load values. This highlights again, especially for the latter case, that the approaches and assumptions of the canopy budget model used to calculate atmospheric deposition on forest ecosystems are crucial when evaluating critical load exceedances.

Table 4 Modelled critical loads of acidification [CLmax(S)] and eutrophication [CLnut(N)]
Full size table

Local effects

The three case studies for which the different CE approaches were tested are all situated in the region of Flanders. In the majority of the forests in this region, critical loads for acidifying and eutrophying deposition are exceeded (Craenen et al. 2000; Staelens et al. 2009). It is therefore certainly relevant to quantify total deposition, evaluate trends or compare different forest types for mitigation strategies in this region. However, different effects of the CE calculations might be observed in regions such as China or India, where N and particularly S deposition have started to rise more recently than in Western Europe (Fowler et al. 2009), or in unpolluted boreal forests in Northern Europe (Ukonmaanaho and Starr 2002). The canopy budget model has not been properly evaluated for subtropical and tropical forests, where large amounts of N are shown to be assimilated by epiphytic and bryophytic lichens (Wanek and Hinko-Najera Umana 2010). Furthermore, even within temperate forests different results might be obtained for different tree species, e.g. when NO3 − is retained preferentially compared to NH4 + (see “Effect of canopy exchange calculations: canopy leaching of weak acids, NO -3 uptake and canopy uptake of N during the leafless season” section).

Conclusion

In this study varying approaches to calculate the canopy budget model were used. As such we explored the range in total and potentially acidifying deposition onto a deciduous forest canopy, deposition ratios between a coniferous and deciduous forest stand and the characteristics (Kendalls tau and median slope) of a deposition time trend analysis. The time step, type of precipitation data, tracer ion, CL of WA and CU of N during the leafless season all had a significant effect on the results, whereas the CU of NO3 − generally had no influence. A correction of bulk to wet-only precipitation data is recommended when available. Sodium can be considered as the most suitable tracer ion, but more research for different tree species is needed on the assumption of inertness during leaf development. Furthermore, an adjustment according to the mean mass diameter for each base cation could improve the estimated DD of these elements. Including CL of WA in the canopy budget model had usually the highest impact in all three case studies, but this could mainly be attributed to the fact that weak acid concentrations in water samples were estimated from a cation–anion balance. The strong effect of including weak acid leaching disappeared when calculating WA from DOC and bicarbonate measurements, which indicates that the accuracy of analytical measurements is of high importance, in particular for rainfall samples in which ion concentrations are generally low.

Some indications with regard to the most suitable approach were derived and since the results of this approach were in line with literature data, we may conclude that the canopy budget model can be a suitable approach to calculate total N and acidifying deposition. However, it is important to consider the various possible options of the canopy budget model when setting-up an experiment or starting data analysis. Also, this study presents evidence of the importance of CE calculations for three relevant case studies in Flanders, a region with enhanced N deposition due to anthropogenic emissions. Since local and tree species-specific effects may play a role here, similar model evaluations in other regions are recommended. Finally, it is recommended for future research to compare this approach to results obtained by other methods to estimate atmospheric deposition.