Vertical changes in sulfur isotopic ratio of water flowing through a forested catchment along the coast of the sea of Japan in central Japan–a buffer against seasonal transboundary air pollution

ABSTRACT Acidic substances, specifically sulfur (S) compounds, derived from atmospheric deposition play a major role in the acidification of forest ecosystems. This study conducted field surveys to clarify a buffering system against seasonal large S inputs in a forested catchment in central Japan that has historically suffered from transboundary air pollution. Results showed that atmospheric S fluxes significantly increased in winter due to north-westerly seasonal winds from the Asian continent; fluxes were 1.1 and 0.3 kmolc ha−1 in the cold and warm seasons, respectively, due to the large effects of sea salt and transboundary air pollution. Despite the large seasonality within atmospheric deposition, SO4 2– concentrations in stream water (SW) were found to be relatively stable throughout the year. Similarly, S isotopic ratios (δ34S) in rainwater showed clear seasonal variation, increasing to 12‰ in winter and decreasing to 2‰ in summer, whereas the δ34S value of SW was stable year-round at ~9‰. Flux-weighted mean δ34S values for rainfall (RF), throughfall, stemflow, and SW were similar, i.e. 8.5, 9.5, 9.0, and 9.0‰, respectively. Both the δ34S values and the SO4 2– concentrations in RF and soil solutions appear to converge at values of SW, suggesting that atmospheric deposition is a primary S origin in SW. The sulfur adsorption-desorption in soil appears to mainly buffer the large sulfur input and prevent sudden acidification, whereas a relatively small biological sulfur cycle was suggested by litterfall. Possible disturbances within this buffering system should be carefully monitored under a changing climate.


Introduction
Acidic substances that are deposited onto Earth's surface from the atmosphere have caused the acidification of soil and inland water in Europe (Wright and Henriksen 1978;Hallbäcken and Tamm 1986), North America (Watt et al. 1983;Driscoll et al. 2001), and Asia (Matsubara et al. 2009;Nakahara et al. 2010;Qiao et al. 2016). In particular, sulfur (S) compounds, such as sulfur dioxide (SO 2 ) and sulfuric acid (H 2 SO 4 ), have played major roles in the acidification of forest ecosystems, because mobilization of sulfate ions ) in the soil-plant system is often accompanied with hydrogen ions (H + ) (Van Breemen et al. 1983). Although SO 2 emissions have started to decline or been reduced to safe levels in the aforementioned regions, the timing and mechanisms of releasing S that has accumulated in forest ecosystems are highly contested (Mitchell and Likens 2011;Vuorenmaa et al. 2017;Sase et al. 2017Sase et al. , 2019Sase et al. , 2021. SO 4 2are deposited through rainwater alongside particles such as ammonium sulfate ((NH 4 ) 2 SO 4 ); these are cycled from the atmosphere into the soil-plant system and are stored within soil in various forms, such as inorganic and organic S (Tanikawa et al. 2022). Such S compounds accumulated in soil could be mobilized due to changes in precipitation amounts and patterns (Mitchell and Likens 2011;Sase et al. 2017).
The S isotopic ratio (δ 34 S) is well known as a useful tool to investigate possible sources of S in the atmosphere, as well as S behavior in forest ecosystems (Novák et al. 2000(Novák et al. , 2001Mayer et al. 2010;Ohizumi et al. 2016;Inomata et al. 2019;Sase et al. 2019Sase et al. , 2021. The isotopic differences in fossil fuels, such as oils and coal, are reflected in atmospheric depositions such as rainwater Inomata et al. 2019) and particulate matter . Rainwater and stream water (SW) in forest ecosystems reflect the δ 34 S values of various S sources such as atmospheric deposition or storage in soil and bedrock, whereas plant tissues partially reflect these values after isotopic fractionation (Novák et al. 2000(Novák et al. , 2001Mayer et al. 2010;Ishida et al. 2015;Sase et al. 2019Sase et al. , 2021. Thus, the application of the S isotopic analysis to the waters in forest ecosystems and plant tissues is considered effective to understand the dynamics of S derived from atmospheric deposition in forest ecosystems. Atmospheric deposition along the coast of the "Sea of Japan" in central Japan is strongly affected by acidic substances from the Asian continent . Cumulative S deposition amounts over 25 years  were significantly high along the coast of the Sea of Japan . Fluxes of SO 4 2and its non-sea salt (nss) fraction by throughfall and stemflow (TF+SF) significantly increased with the fractions of sea-salt components such as Na + and Cl -, during winter in the forested catchment close to the coast of the Sea of Japan; this suggested the long-range transport of air pollutants by the seasonal northwesterly winds from the Asian continent (Kamisako et al. 2008;Sase et al. 2008). The S isotopic analysis of rainwater in the same area estimated that 60% of annual SO 4 2fluxes could be attributed to transboundary air pollution in the mid-2000s . Despite a significant large flux of SO 4 2in winter, SO 4 2concentration and pH in SW were relatively stable throughout the year at the forested catchment (Kamisako et al. 2008). The recent study showed that SO 4 2in SW was more isotopically homogenized than in rainwater, suggesting the presence of internal S cycles in forest ecosystems (Sase et al. 2021).
As described above, large seasonal fluxes in SO 4 2concentration appeared to be well homogenized in the soil-plant systems within the studied forested catchment -this seems to be the primary process buffering against atmospheric S input in the forested catchment. It is crucial to clarify the process in detail in order to understand interactions between the forest's internal S cycles and atmospheric deposition. The objectives of this study were (1) to investigate possible sources of atmospheric S deposition on forest canopy, (2) to investigate specific processes that mainly facilitate S homogenization in the aforementioned soil-plant system, and (3) to evaluate how this S homogenization process can prevent sudden forest acidification due to atmospheric S input. In addition to rainwater and SW, soil solution (SS) samples were intensively collected at different depths and slope positions within the forested catchment over four years from 2012 to 2016 in order to reach the aforementioned objectives. S isotopic analyses were applied to the waters, litterfalls, litter layers, and organic soil layers, in addition to ionic analyses. Results from this study will contribute to a deeper understanding of S dynamics in forest ecosystems under changing atmospheric deposition and climate.

Study site
This study was conducted in the Kajikawa catchment site (KJK), which is a Japanese cedar (Cryptomeria japonica D. Don) forest located in the coast of the Sea of Japan in the northern part of Shibata City (i.e. formerly called Kajikawa Village) in the Niigata Prefecture, Japan ( Figure 1; 37° 59′ N, 139° 23′ E; Kamisako et al. 2008). The annual mean precipitation was 2242 mm and the annual mean temperature was 13.4°C during 1981-2010 at the nearest meteorological station (Sase et al. 2021). The parent material at KJK is granodiorite. According to soil profile surveys conducted by our group in 2001 and the national forest soil map of Japan (Forestry and Forest Products Research Institute 2022), brown forest soils (dry type, B D , or moist type, B E ) and partly black soils (Bl) are distributed in the catchment area. These are classified according to the Classification of Forest Soil in Japan (Forest Soil Division 1976). They can be classified as Cambisols and Andosols, respectively, according to the World Reference Base for Soil Resources (WRB) (Morisada et al. 2004). Soil depths (until the B or BC horizons) varied in different slope positions, from approximately 30 cm to 190 cm, and the mean (standard error) was 81 (8) cm, according to the soil profile survey at 32 points in 2001. The previous studies described more detailed site information (Kamisako et al. 2008;Sase et al. 2008Sase et al. , 2012Sase et al. , 2021.

Rainwater collection
A collector for rainfall (RF) outside the forest canopy was installed on the northern ridge, and six collectors for throughfall (TF) and stemflow (SF) were installed in the middle portions of slopes, respectively ( Figure 1). The rainwater samples were collected monthly. The detailed sampling procedures can be found in Kamisako et al. (2008) and/or Sase et al. (2008).

Sampling of SW and SS
SW samples were collected at the bottom of the catchment ( Figure 1). SS samples were collected using a porous-cap sampler (DIK-8390-11; DIK-8390-58, DAIKI, Japan) at the upper (U), middle (M), and lower (L) slope positions at 20and/or 60-cm depths (Sase et al. 2021). For each depth at the U, M, and L positions, four samplers were installed -composite samples of all four samplers were used for δ 34 S determination. SS samples were named using a combination of the initials of position and depth, e.g. U20 for 20-cm depth on the upper position and L60 for 60cm depth on the lower position. However, the soil layer at the middle slope was too thin to collect the M60 samples. Additionally, SS samples in mid-winter (from January to March) could not be obtained because of the deep snow. We assessed monthly data of SW and SS from 2012 to 2016, during which SS samples were collected intensively. Besides the regular sampling protocol outlined above, seepage waters below litter (L) layers and fermented/humic (FH) layers were collected thrice using the porous-cap sampler at the M position at intervals of 2 weeks, i.e. 16-31 August, 14-31 October, and 31 October-15 November 2016, to determine changes in δ 34 S values on the forest floor.

Litterfall collection
Litterfall from Japanese cedar trees was collected twice a month from 18 May 2015 to 1 June 2016, except for a few collection intervals ~20-d long; winter samples (during snow cover) from 15 December 2015 were collected on 15 April 2016 (i.e. 122 d later). A total of five litter traps (each area: 0.5 m 2 ) were installed besides TF and SF samplers on three slopes; litterfall samples collected from five traps on each slope were mixed to create a composite sample. Each composite litterfall sample for each interval was divided into four plant fractions, namely, leaves, branches, female flowers, and male flowers. Sulfur contents and δ 34 S values were analyzed for each plant fraction independently.
Prior to litterfall sampling, organic matters of L and FH layers were collected in April 2015, just after the snow melted, to determine their representative δ 34 S values. Branches and other organic matters were separately applied for the S isotopic analysis. Fresh leaves, branches, and female/male flowers of Japanese cedar were collected in August 2015, when plants are considered the most active.

Chemical analyses
Rainwater samples including RF, TF, and SF, and SW and SS samples were applied for chemical analysis according to the technical manuals of the Acid Deposition Monitoring Network in East Asia (EANET) (EANET 2010a; 2010b). The pH, electrical conductivity (EC), and the concentrations of inorganic ions were determined for all water samples including rainwaters, SS, and SW, whereas the alkalinity was determined for SW samples. Detailed information on instruments used for chemical analyses can be found in Sase et al. (2021).

Sulfur isotopic analyses
The δ 34 S values of SO 4 2in rainwater, SS, and SW were determined monthly from August 2012 to November 2016. For rainwater samples, this study determined δ 34 S values for RF, TF, and SF, separately; SO 4 2in the water samples was concentrated and precipitated as BaSO 4 for δ 34 S analyses . Air-dried plant fractions in litterfall and organic matters from L and FH layers were dried at 105°C for several hours, after which they were ground into powders and decomposed by nitric acid (HNO 3 ), and the SO 4 2in the solutions was then precipitated similarly. BaSO 4 precipitates were analyzed using a EA-IRMS (EA2500-Delta Plus via Conflo II, Thermo Fisher Scientific, MA, U.S.A). The δ 34 S was determined as follows: where X and CDT indicate a sample and Canyon Diablo Troilite (as the standard substance), respectively. The analytical precision was less than 0.2‰ (Inomata et al. 2019).
For SO 4 2flux by RF and TF, contribution ratios of possible sources were estimated using a mixing model. In this assumption, three major sources, namely sea salt (SEA), transboundary air pollution derived from Chinese coal combustions (TRB), and domestic air pollution (DOM), were considered and calculated using the following equations: where f, "obs," and "seawater" indicate the fraction of each contribution to SO 4 2in the samples, observed sample concentrations, and the ratio in seawater, respectively. This study used +20.3‰ for δ 34 S SEA , 6.6‰ for δ 34 S TRB , and −2.7‰ for δ 34 S DOM Inomata et al. 2019), which have been used for the assessment of wet deposition data at all stations in Japan, including in the Niigata Prefecture (Inomata et al. 2019). We applied the same method to assess the seasonal RF and TF data at KJK for comparison. Data completeness of the δ 34 S values for the mixing model was found to be 100% and 91% for RF and TF, respectively, whereas flux data were available during the observational period. For the missing TF δ 34 S data, RF data during the same periods were used for the mixing model. Finally, seasonal contributions of SEA, TRB, and DOM to fluxes by RF and TF were estimated during the cold (i.e. October-March) and warm (i.e. April-September) seasons during the observational period.

Quality control and quality assurance (QA/QC)
The QA/QC was conducted for all the measurement processes according to the technical manuals of EANET (2010a; 2010b). Ion balances of water samples had no significant error according to the monitoring criteria.

Data analyses
All statistical analyses were conducted using R version 3.6.2 (R Core Team 2019). The Wilcoxon signed rank test was conducted to compare concentrations, fluxes, and δ 34 S values among RF, TF, and SF, using the "exactRankTests" package (version 0.8-35). Linear regression analyses between δ 34 S values and the reciprocal of SO 4 2concentrations were conducted using the "Rcmdr" package (Version 2.6-2).

Seasonal changes in atmospheric S deposition and SW chemistry
The SO 4 2concentration in RF, TF, and SF showed clear seasonality, steeply increasing in winter, and decreasing in summer (Figure 2(a)), whereas its concentration in SW was stable throughout the year (Figure 2(b)); SO 4 2concentration in RF was generally lower than that in TF and SF (Wilcoxon signed rank test: p < 0.0001 and p < 0.001, respectively). SO 4 2fluxes by RF and TF+SF showed more clear seasonality, reflecting its precipitation and concentration patterns ( Figure S1). Since the SW discharge periodically increased from late winter to spring due to snow melting ( Figure 2(b)), SW SO 4 2flux increased slightly in winter accordingly; however, the increased magnitude was lower than that seen from the fluxes of RF and TF+SF ( Figure S1).
The mean values for water fluxes and concentrations of major inorganic constituents in RF, TF, SF, SS, and SW during the warm season and the cold season are shown in Table 1. The mean concentrations of SO 4 2-, Cl -, Na + , K + , Ca 2 + , and Mg 2+ in RF were higher in the cold season than in the warm season, whereas those of NO 3 and NH 4 + were not different between the seasons (or were slightly higher in the warm season than in the cold season. Concentrations of most ions in SW remained stable between the seasons, whereas NO 3 concentration was higher in the winter, and NH 4 + concentration was higher in the summer.

Sulfur isotopic ratios in rainwater, SS, litter layers, and SW
The δ 34 S values of rainwater, including RF, TF, and SF, generally increased in winter and decreased in summer ( Figure 3(a)); seasonality was the most evident in RF. After the winter peaks, δ 34 S values in TF and SF decreased slightly slower than that in RF, and the lowest δ 34 S values in the summer were often higher in TF and SF than in RF. δ 34 S values in TF or SF were statistically different from that in RF (Wilcoxon signed rank test: p < 0.0001 and p < 0.0001, respectively). In contrast to rainwater, the δ 34 S value in SW did not show a clear seasonality, which was stable at approximately 9‰ (Figure 3(b)). The δ 34 S values in SS were also relatively stable during the observational period (Figures 3(c,d)), whereas the samples in mid-winter could not be obtained because of the deep snow. At the 20 cm-depth (Figure 3(c)), slightly higher δ 34 S values were observed in early spring than those in other seasons. Although actual peak values in mid-winter were not collected, the lowest δ 34 S values in SS in summer were much higher than those in rainwater; e.g.>6‰ and 2-4‰ in SS and rainwater, respectively. At the 60-cm depth (Figure 3(d)), the variance of δ 34 S values were significantly smaller than that at the 20-cm depth (F-test, p < 0.0001). The δ 34 S values at 60 cm were significantly higher in the lower slope position than in the higher slope position (Wilcoxon signed rank test, p < 0.0001) whereas those at 20-cm depth were not different  (p = 0.186). Seepage water δ 34 S values from L and/or A0 layers in the forest floor were largely fluctuated compared to those from SS and rather within the ranges of rainwaters ( Figure S2).

Sulfur flow by litterfall
The mean S contents in the respective plant factions of litterfall fluctuated slightly and did not show clear seasonality (Figure 4(a)). The S contents in leaves were significantly higher than those in male flowers, female flowers, and branches (Wilcoxon signed rank test: p = 0.000214, 0.00763, and 0.00195, respectively). Based on the S contents and litter amounts collected, the annual S flux by litterfall was estimated to be 2.44 kg ha −1 and 2.00, 0.07, 0.13, and 0.24 kg ha −1 by leaves, branches, female flowers, and male flowers, respectively. The mean δ 34 S values in the respective plant factions of litterfall were relatively stable (Figure 4(b)). The δ 34 S values in male flowers or female flowers were significantly higher than those in leaves (Wilcoxon signed rank test: p < 0.0001 and p = 0.000244, respectively) and branches (p = 0.00195 and p = 0.00195, respectively). The S-flux weighted mean δ 34 S value was found to be 7.3‰, which was slightly lower than the SW mean value of 9.2‰.

Seasonal contribution of transboundary air pollution to fluxes by throughfall and SW
Estimated contributions of the respective sources to atmospheric S fluxes by RF and TF were significantly different between the seasons (Figures 5(a,b)). Although TF S fluxes may include both wet (via rainwater) and dry deposition (via gases and particles), contribution tendencies of the respective sources to TF S fluxes were found to be very similar to those to RF S fluxes. Mean TF fluxes from SEA, TRB, and DOM during the observational period were 462, 432, and 213 mol c ha −1 , respectively, in the cold season, whereas those in the warm season were 55, 142, and 112 mol c ha −1 , respectively ( Figure 5(b)). Among the nss fractions, the contribution ratios of TRB and DOM to TF S fluxes during the cold season were 67 and 33%, respectively, whereas those during the warm season were 56 and 44%, respectively.

Seasonal large contribution of transboundary air pollution
An increase in atmospheric SO 4 2fluxes during winter along the coast of the Sea of Japan ( Figures. 2 and S1) has been attributed to an increase in transboundary air pollution from the Asian continent due to seasonal northwesterly winds  (Sase et al. , 2012(Sase et al. , 2021Ohizumi et al. 2016;Inomata et al. 2019). The study area is famous for heavy snow, which is derived from moisture evaporated from the Sea of Japan due to seasonal winds. Previous studies (Kamisako et al. 2008;Sase et al. 2008Sase et al. , 2012 have reported that atmospheric fluxes of Cl − and Na + increase with the SO 4 2flux at KJK in winter, suggesting the effects of seasonal winds. The large contribution of transboundary air pollution in winter along the coast of the Sea of Japan has already been suggested by S isotope Inomata et al. 2019) and modeling studies (e.g. Kuribayashi et al. 2012). The δ 34 S values in rainwater clearly indicated that the contributions of major S sources have been changing seasonally, resulting in an increase in S fluxes along with increase in the δ 34 S values Inomata et al. 2019). In addition, the δ 34 S values of nss fractions (i.e. δ 34 S nss ) in fine particles (PM2.5) also increased in winter along the Sea of Japan coast . Because it is assumed that S fluxes by TF and SF include dry-deposited components such as gaseous and particulate matter, the seasonality of the δ 34 S values in TF and SF clearly exhibits the influences of both wet (i.e. rainwater) and dry (i.e. gases/particles) deposition. As shown in Figure 3(a), the δ 34 S values of RF increased in winter and were similar to those of TF and SF. This indicates that both wet and dry deposition are influenced by transboundary air pollution during winter.
The δ 34 S values in TF and SF were not, however, identical with those in RF, showing higher values in winter and summer (Figure 3(a)). It is possible that dry-deposited components were not completely scavenged from tree canopies by TF and/or SF. Moreover, it can be hypothesized that the contributions of S sources were different among RF, TF, and SF, as observed in Europe (e.g. Novák et al. 2005). Another possibility is that biological fractionation occurred due to canopy interactions (such as uptake or leaching) (e.g. Staelens et al. 2008). Canopy interactions of SO 4 2have been considered to be relatively small. However, as suggested by S content profiles of tree rings (Ishida et al. 2015;Sase et al. 2019) and the litterfall data collected in this study, a certain portion of S appears to be cycled through a soil-plant system. Thus, the above possibilities can be suggested as avenues for future studies on S deposition processes in forest areas.
The large contributions of SEA and TRB to RF S flux in the cold season ( Figure 5(a)) have already been suggested by previous studies in the Sea of Japan's coast Inomata et al. 2019). This study confirmed that SEA and TRB also contributed largely to TF S flux in the cold season. Among the nss fractions, the contribution ratios of TRB to the TF S flux was larger in the cold season than in the warm season.

Vertical changes in the sulfur isotopic ratio
The δ 34 S-value variations became smaller vertically in waters flowing through the catchment of KJK, from rainwater including RF, TF, and SF through SS to SW (Figure 6), which was summarized based on the data shown in Figures 3 and 4. All SW δ 34 S values were ranged within the variations of rainwater, suggesting that it is the principal contribution of atmospheric S deposition. The flux-weighted mean δ 34 S values did not show a clear tendency and were found to be 8. 5, 9.5, 9.0, and 9.0‰ for RF, TF, SF, and SW, respectively. In conclusion, no clear isotopic change could be observed during hydrological processes from rainwater to SW at the KJK study site.
Homogenization (or less seasonality) of the δ 34 S values in forest ecosystems were also observed in SS and/or SW in Europe (Novák et al. 2000), North America (Campbell et al. 2006), and Asia ( (Krouse 1980;Mayer et al. 2010;Sase et al. 2019). However, we did not observe any linear relationship between SW δ 34 S values and the reciprocal of SO 4 2concentrations at KJK (i.e. p = 0.465), and no possible end member was identified for SW (Figure 7). Instead, both δ 34 S values and the reciprocal of SO 4 2concentrations in RF and SS appeared to converge at points of SW. Moreover, the S output was relatively well balanced with the input at KJK for 15 years from 2002/2003 to 2017/2018, although the S budget discrepancy was found in 2014/2015 and 2017/2018 during the recovery process from acidification (Sase et al. 2021). Thus, currently, no large effect of internal S sources, such as geological sulfur, was identified.
Relatively lower δ 34 S values in litterfall samples (i.e. fluxweighted mean: 7.3‰) suggested the possibility of biological fractionations occurring at KJK, particularly in leaves and branches, although the reason flowers had relatively high δ 34 S values has not been clarified yet. δ 34 S values of organic S in soil (Novák et al. 2001(Novák et al. , 2005 and tree-ring δ 34 S values in Japanese cedar (Ishida et al. 2015;Sase et al. 2019) have been reported to be ~ 2‰ lower than those of atmospheric deposition because of biological fractionations upon assimilation. Litterfall data collected below Japanese cedar canopy at KJK appeared to indicate the same phenomena. Moreover, the understory vegetation survey conducted in September 2016 estimated that its S flux and weighted mean δ 34 S value to be 1.7 kg S ha −1 yr −1 and 7.5‰, respectively, suggesting the occurrence of biological S fractionations as well (unpublished data). Biological S fluxes including the cedar litterfall flux (i.e. 2.4 kg S ha −1 yr −1 ) and understory vegetation flux could be estimated as 4.1 kg S ha −1 yr −1 , which was ~18% of the atmospheric S flux by TF+SF (23 kg S ha −1 yr −1 , i.e. mean for the observational period). Although the biological S fluxes could partly contribute to homogenization of δ 34 S values in the forest ecosystem, no clear effect of biological fractionations (due to assimilation by trees) on the δ 34 S values has been detected during hydrological processes from rainwaters to SW.
The significantly higher SS δ 34 S values at 60 cm in the lower slope position suggest another possibility of biological fractionation due to bacterial dissimilatory SO 4 2− reduction   (BDSR), which would enrich 34 S in the remaining SO 4 2− (Kang et al. 2014). In fact, various types of forests can emit H 2 S, particularly from deep soil layers (e.g. between 85-and 100-cm depths in the case of subtropical forest; Ke et al. 2022). The median δ 34 S value in SW was slightly higher than that in SS samples (Figure 7), whereas the flux-weighted means of SS samples could not be calculated using the porous-cap sampling method. Although no clear effect of biological fractionation has been identified by the comparison between δ 34 S values in rainwater and SW at KJK, as discussed previously, the possibility of BDSR should also be taken into consideration for further understanding S dynamics in forest ecosystems.

Buffering of seasonal acid deposition
Because there was no clear isotopic fractionation from rainwater to SW and stabilization of δ 34 S values occurred in relatively shallow soil layers, the adsorption/desorption processes of SO 4 2in soil can be suggested to be the main process of buffering seasonal acid deposition. In general, the adsorption of SO 4 2at pH-dependent charges on soil surface increase with decrease in pH and increase in SO 4 2concentration in the equilibrium solution (Rao and Sridharan 1984;Kamewada and Takahashi 1996;Shindo and Fumoto 1998;Hayashi and Okazaki 2003;Alves and Lavorenti 2004). Higher SO 4 2concentrations and lower pH in rainwaters during the cold season ( Figure 2 and Table 1) could potentially contribute to effective SO 4 2adsorption in soil. Since the climate in KJK is humid throughout the year (Sase et al. 2012), lowering rainwater SO 4 2concentrations in the summer appeared to gradually extract SO 4 2from soil into SW. As suggested by previous studies, no isotopic fractionation occurred in adsorption or desorption processes of SO 4 2in soil (Van Stempvoort et al. 1990;Novák et al. 2005). It is implied that adsorption/desorption processes of SO 4 2in soil mainly contributed to homogenization of δ 34 S values during the hydrological processes from rainwater through SS to SW at KJK. Previous studies using the Δ 17 O of SW NO 3 − suggested that NO 3 − stored in the groundwater and riparian zone was the main source of SW NO 3 − (Nakagawa et al. 2018;Ding et al. 2022). Mixing in riparian groundwater may also contribute to buffering stream SO 4 2− concentrations and δ 34 S values. However, stabilization in shallow soil layers suggests that the adsorption/desorption was more important than the groundwater mixing.
Meteorological variability, such as changes in precipitation amount and pattern under a changing climate, could alter S retention-release cycles in forest ecosystems (Nakahara et al. 2010;Mitchell and Likens 2011;Mitchell et al. 2013;Sase et al. 2017Sase et al. , 2019Ke et al. 2021;Zhigacheva et al. 2022). Mobilization of "legacy S pools" derived from atmospheric deposition in the past (Vuorenmaa et al. 2017) should carefully be monitored in the recovery process from acidification. The S budget at KJK was relatively well balanced, but the net export has increased since 2013/2014 in the recovery process (Sase et al. 2021), which is similar to the phenomena observed in forest catchments in Europe. The S retention-release cycles and adsorbed S appear to play an important role in the recovery process under a changing climate. Accumulation conditions of inorganic/organic S including adsorbed S and its isotopic ratio (Tanikawa et al. 2022) should be investigated at KJK in future studies.

Conclusions
Vertical changes in δ 34 S values were investigated for waters flowing through the KJK forested catchment in central Japan, which was largely influenced by transboundary air pollution due to northwesterly winds during the cold season. Drydeposited components collected by TF and SF appear to have (almost) the same sources as wet-deposited ones by RF, according to their similar δ 34 S values. Despite large seasonal variation in S deposition, SO 4 2concentrations in SW were well buffered and stabilized throughout the year, preventing a sudden acidification. The S input-output budget at KJK was also relatively well balanced and no large internal S source was identified. Vertical homogenization in δ 34 S values and their relationships with the reciprocal of SO 4 2concentrations suggest that adsorption and desorption processes of SO 4 2mainly contribute to the buffering potential, which caused inconsistent seasonal changes between atmospheric S deposition and SW. Plant assimilation and BDSR should also be taken into consideration for understanding of S dynamics in forest ecosystems. Further investigations on S dynamics should be continued under changing air pollution and climatic conditions.