Introduction

Widespread atmospheric deposition of strong acid anions of sulfur (S) and nitrogen (N) aerosolized by anthropogenic processes is an important driver of plant nutritional status and uptake of forest soils and is inherently linked to plant growth and transpiration1,2,3. For example, acid deposition can have a fertilization effect on plant growth by increasing the soil N availability or base cation mobilization4,5. In addition, acid deposition can cause soil acidification and the leaching of base cations, such as calcium (Ca), magnesium (Mg) and potassium (K), thus reducing the availability of these nutrients for plants6. Moreover, acid deposition may aggravate the release of metal ions into the soil and soil solution in the form of hydrated oxides, particularly aluminum (Al)7. The increased mobilization of Al in the soil may inhibit root cell division and elongation8,9 and damage the root membrane structure and function, leading to blocked Ca uptake10,11,12. Consequently, long-term acid deposition may impair plant growth because it leads to nutrient imbalances and hinders the nutrient uptake process. The resultant positive or negative effects associated with acid deposition are largely determined by the biogeochemical functions of the affected nutrients. However, the integrated physiological responses of plants to soil nutrient availabilities with regard to hydrological changes remain elusive, particularly in regions suffering from long-term acid deposition in subtropical or tropical forest ecosystems.

The hydrological cycle of forested watersheds involves three main fluxes: precipitation (P), evapotranspiration (ET), and streamflow (Q)13. Plant transpiration supports the continuous upward movement of water and nutrients from the soil to the vegetation14,15, a crucial component linking soil biogeochemical processes and forest hydrology. Notably, soil calcium plays a crucial role in modulating plant physiological responses, such as stomatal closure regulation, carbohydrate metabolism and cell division16. Nutrient imbalances may drive plants to acquire nutrients for physiological demands through soil uptake and transpiration17. Thus, plant transpiration may be particularly sensitive to soil nutritional conditions18,19. A study from the Hubbard Brook Experimental Forest (HBEF; New Hampshire) observed a short-term (3-year) increase in plant transpiration following soil Ca restoration because the increased Ca availability enhanced the primary production of trees20. Another study from the Fernow Experimental Forest (FEF; West Virginia) reported that acid deposition could intensify plant water transport when plants underwent soil Ca deficiencies as observed in a long-term (30-year) acidified watershed21. Additionally, the elevated plant transpiration and reduced water drainage loss resulted from additional N being added into a field plot in an N-rich subtropical forest at the Dinghushan Biosphere Reserve (DBR, Guangdong Province)22. These results highlighted the soil biogeochemical effects on plant physiological processes, ultimately altering watershed hydrology. However, the abovementioned studies did not examine the stomatal or xylem hydraulic response to acid deposition, and Smith and Shortle23 also pointed out the need for a microscale understanding of xylem hydraulic traits within a living tree. Tree leaf stomatal and xylem hydraulic traits may reflect the morphological and physiological adjustments of trees to changing environmental conditions, thus providing important information about water transport when coping with nutrient availability alterations resulting from acid deposition. Thus, the lack of direct evidence from leaf stomatal and stem xylem hydraulic structures at the individual tree level significantly constrains our understanding of the physiological mechanism behind the hydrological responses of forested watersheds to acid deposition.

Acid deposition affects soil biogeochemical processes and causes subsequent cascading changes in plant physiological processes and hydrological dynamics. However, acid deposition does not affect plant transpiration in isolation; concurrent changes in P, air temperature, vapor pressure deficit (VPD), and elevated atmospheric carbon dioxide (CO2) can also affect plant transpiration24,25,26,27,28. For example, mounting studies have suggested that CO2 has far-reaching impacts on ET through modified plant physiological processes29. On the one hand, CO2 may reduce stomatal conductance and diminish ET, thus increasing Q30. On the other hand, this effect may be neutralized through the promotion of plant leaf growth, thus increasing the surface area on which transpiration occurs31. In addition, plant transpiration largely depends on stomatal apertures, which are profoundly regulated by VPD and air temperature32,33. VPD reflects the atmospheric evaporative demand and connects ecosystem water cycles by regulating plant transpiration34,35. These interactions that affect plant transpiration and streamflow highlight the need for additional observational datasets to separate the effects of acid deposition on plant physiological performance and hydrological processes. Acid deposition exerts previously unrecognized impacts on the hydrological cycle, the contribution of acid deposition to hydrological processes from the forested watershed perspective remains uncertain.

The primary objective of this study is to examine how the responses of plant physiological traits to acid deposition at the microscale connect with hydrological processes at the watershed scale. To fill this knowledge gap, we uniquely combine data from a controlled plot experiment, forest and soil inventory data, and watershed hydrology monitoring data. First, we initiated a long-term plot-scale experiment in 2009 to investigate leaf stomata, stem radial growth and xylem hydraulic conductivity through dendrochronological techniques in the DBR watershed. We addressed how the physiological traits of trees adjust to acid addition. We also used foliar stoichiometry to examine the potential base cation deficiency corresponding to acid input, particularly for the Ca-regulated stomatal closure process. Second, we established a permanent plot (1 ha) to monitor the soil geochemical properties and forest species composition under ambient acid deposition. From these data, we assessed changes in the base cation content corresponding to the plant nutrient availability in the forest soil. Combining the observed results of the physiological responses of trees to acid deposition, we used forest inventory data to extend the data from the individual tree level to the watershed scale. Third, we measured the hydrological cycle in the DBR watershed and applied the Budyko framework, a widely used tool coupled water supply (annual P) and energy availability (potential evapotranspiration) balance, to examine the relative importance of hydrological drivers. From this analysis, we quantitatively attributed the dynamics of ET and Q in the watershed with regard to acid deposition, climate factors and atmospheric CO2 through an attribution analysis. Therefore, our findings of the physiological responses of trees to acid deposition experiments at the microscale and the hydrological processes at the watershed scale enabled us to improve the mechanistic understanding of how long-term acid deposition influences the watershed hydrological cycle.

Results

Tree stem xylem and leaf traits in response to acid deposition

Tree growth and xylem hydraulic trait responses to experimental acid addition were observed directly from tree stem anatomy in this work. Tree rings were collected from Castanopsis chinenes, which is the most abundant tree species in the study area and makes up 44% to 59% of the total stand basal area derived from forest inventory data during 1999–2015. The tree ring widths (RW) during 2000–2008 showed very similar values prior to the treatment period (Fig. 1a). The RWs of trees growing in the treated plots gradually became larger than those in the control plots since the start of the acid addition in 2009 (Fig. 1a). The maximum RW deviation between the treated and control plots was observed after a 6–7 year period. Then, the RWs declined rapidly, and a similar stem radial growth was maintained during 2017 and 2019. Altered stem radial growth under acid addition could subsequently impact the xylem hydraulic traits; thus, acid addition might affect the xylem hydraulic conductivity (Ks) and mean vessel diameter (Dh) of annual rings. The Ks trend was consistent with the RW values after acid addition. Generally, the Ks and Dh of trees growing in the control plots remained relatively stable during the study period (Fig. 1b). For the treated plots, the Ks showed an obvious increasing trend during the first 6–7 years after acid addition (2009–2016) followed by steep decreases during 2017 and 2019. The Dh growing in the pH3.0 plots showed a slightly increasing trend, whereas a slightly decreased in Dh was observed in the pH4.0 plots during the whole experimental period (Supplementary Fig. 1). Overall, acid deposition has a short-term (6–7 years) fertilization effect on tree physiological processes, and the stem radial growth-enhancing effect ceases as indicated by the tree RW and Ks values with continuous acid application.

Fig. 1: The responses of tree growth and xylem hydraulic traits to experimental acid addition.
figure 1

a, b The tree-ring widths (RW) and theoretical xylem-hydraulic conductivity (Ks) within annual ring of Castanopsis chinensis in the treatment (pH3.0, pH4.0) and control plot (CK). The gray shaded areas indicate the period of experimental acid addition. The lines indicate the mean values measured from 6, 8 and 8 tree cores sampled from pH3.0, pH4.0 and CK, respectively.

Full size image

For tree leaf traits, the leaf stomatal densities of trees growing in the acid addition plots were consistently higher than those of trees growing in the CK plots, while the stomatal densities of Aidia canthioides Masam growing in the pH4.0 was slightly lower than in the CK (Table 1). The stomatal guard cell lengths in the acid addition plots were slightly larger (though not statistically significant) than those in the control plots. In addition, the foliar Ca contents in the acidified plots were significantly lower than those in the control plot (Table 1). These reduced foliar Ca contents were an expected result of the loss of soil base cations induced by acid deposition. This foliar cation content is in line with previous observations of other tree species from plot-scale control experiments performed at the same site22. In addition, Castanopsis chinenes, Schima superba and Pinus massoniana (the tree species tested for their foliar Ca contents) were the dominant tree species in the study area, and the basal area of these three species increased from 94 to 96% between 1999 and 2015 (Fig. 2a). Similar foliar Ca content decreased with acid treatment for all tested tree species suggesting that the decrease is a generalized tree response to acid deposition. In addition, the iWUE of Castanopsis chinenes, a dominant tree species accounted for 44% to 59% of the total stand basal area, was 16% lower in the pH 3.0 plots compared to the control plots (Supplementary Fig. 2). Overall, acid addition showed no significant differences in iWUE of trees growing at treatment and control plots.

Table 1 The responses of leaf traits to experimental acid addition.
Full size table
Fig. 2: A conceptual diagram showing how soil biogeochemical processes altered by acid deposition affect plant transpiration at tree level and plot-level as well as affect hydrological processes at watershed scale.
figure 2

a The forest inventory data at watershed scale. b, c Acid deposition driven soil acidification with decreased base cations and elevated aluminum content. d, e The roles of precipitation (P), acid deposition, net radiation (Rn), vapor pressure deficit (VPD) and atmospheric CO2 in the dynamics of evapotranspiration (ET) and streamflow (Q) from the forest watershed. The error bars represent the standard deviation obtained from independent measurements. SI supplementary information.

Full size image

Soil base cation availability changed with acid deposition

The N and S deposition rates ranged from 25.1 to 53.4 and 37.7 to 72.5 kg ha-1 yr-1, respectively, and the rainwater pH generally varied from 4.5 to 5.6 from 2000 to 2018 (Supplementary Fig. 3). Although the acid deposition decreased by approximately 17% for N and 19% for S, the forested watershed still received large amounts of N and S, which continued to influence the soil properties. In general, the soil pHwater decreased from 4.1 to 3.8 (S = −37) under ambient acid deposition over the study period (Fig. 2b). Consistent with soil acidification, the leaching of soil Exch. Ca into the subsurface layer or Q losses (Supplementary Fig. 4) from the watershed would reduce the soil Ca content and stocks (Supplementary Fig. 5). The soil Exch. Al substantially increased and accounted for approximately 30.1 to 84.4% of the extractable cation pools (Fig. 2c). Therefore, long-term ambient acid deposition could promote soil acidification, the leaching loss of soil Ca, and the release of Al into the soil, which in turn may induce nutrient imbalances and cause nutrient deficiencies in vegetation.

Dynamics of Q and ET in the DBR watershed

The dynamics of annual P and Q in the DBR watershed were monitored over the period from 2000 to 2018. Overall, P exhibited a significant increasing trend ranging from 1278.6 to 2869.2 mm (S = 56) (Supplementary Fig. 6), while Q ranged from 368.5 to 1322.2 mm with a significant decreasing trend (S = −52) over the study period (Fig. 3a). The soil water storage of 0–50 cm was relatively stable and the annual difference was less variable with ranged from −26.1 to 20.2 mm (Supplementary Fig. 6). The annual ET derived from the water-balance method exhibited a significant increasing trend (S = 114) (Fig. 3b). This significant positive trend arose from the increasing trend in P, coupled with a decreasing trend in Q from 2000 to 2018. The ratio of ET and P (ET/P) indicates the fraction of P returned to the atmosphere. We found that ET/P varied from 22 to 72% and exhibited an increasing trend, whereas the annual runoff coefficient declined during the past two decades (Fig. 3a, b). These results indicated that despite the increase in P, the ET increase outpaced the P increase during the study period, resulting in a decreased Q.

Fig. 3: Dynamics of observed streamflow (Q) and evapotranspiration (ET).
figure 3

a Dynamics of Q and the runoff coefficient, calculated as the ratio of Q to P, during 2000-2018. b Water balance-derived ET and ET/P, calculated as the ratio of ET to P, in the DBR watershed during 2000–2018. The dash line in a and b indicate average Q and ET trends during study period, respectively. c Scatterplot of potential evapotranspiration (PET)/P against ET/P and the solid line is the relationship represented by Fu’s equation. The 1:1 dash line in c indicate that ET is limited by available energy, and the horizontal dash line in c indicate that ET is limited by available water for a watershed.

Full size image

Potential evapotranspiration (PET) is a key component for understanding the watershed water balance in the Budyko framework, as shown in Fig. 3c. In this study, PET was derived from climate variables, including the air temperature (T), net radiation (Rn), relative humidity (Rh) and wind speed (U2), and showed an increasing trend (S = 27) (Supplementary Fig. 7). In addition, Rh significantly increased (S = 47), while Rn (S = 23), U2 (S = 3) and T (S = 1) showed slightly increasing trends during the study period. We applied a partial correlation analysis to evaluate the PET sensitivity to each climate variable. These results indicated that PET was attributable to Rn, which contributed approximately 60% of the PET increase (Supplementary Fig. 8). Thus, Rn was regarded as the main driver of the PET increase during the study period. In the Budyko framework, the hydrological components of ET and Q are shaped by the P (water supply), PET (energy supply) and watershed characteristics (ω). Combining the abovementioned changes in PET and P, the elasticities of ET to P, PET and ω were 0.68, 0.62 and 0.75, while the elasticities of Q to P, PET and ω were 0.87, −0.61 and −0.74, respectively. This implies that ET and Q were most sensitive to ω and P, respectively. In addition, ω contributed 62.9% of the increased ET and 58.4% of the decreased Q during the study period. Our findings suggested that in addition to P and PET, watershed characteristics (ω), including the watershed topography, soil properties and vegetation dynamics, greatly contributed to the dynamics of ET and Q. Therefore, the potential impacts of ω on ET and Q become an important factor for investigating watershed hydrology.

In addition, the acid deposition amount could explain 72% of the ω variability (Supplementary Fig. 8). Given that acid deposition plays a fundamental role in soil nutrient and tree physiology processes, it is reasonable to examine the contribution of acid deposition to watershed hydrological processes. Therefore, we evaluated the roles of P, acid deposition, Rn, VPD, and atmospheric CO2 in the dynamics of ET and Q from the DBR watershed by conducting a partial correlation (R-value) analysis. We found that ET and Q were significantly and positively correlated with P (Fig. 4). An increased Rn is expected to increase ET (RET-Rn = 0.079), although not statistically significant, and thus reduce Q (RQ-Rn = −0.091) in our watershed (Fig. 4a). The elevated CO2 and VPD exhibited a negative correlation with ET and a positive correlation with Q, although these correlations were not statistically significant (Fig. 4b). Specifically, we found that ET was significantly and positively correlated with acid deposition (RET-Acid deposition = 0.695), while Q was significantly and negatively correlated with acid deposition (RQ-Acid deposition = −0.685) (Fig. 4). Our results corroborate recent findings of an intensification of vegetation water use drawn from an acidification watershed experiment21. Additionally, we calculated the relative contribution of each factor to ET and Q to further reinforce the partial correlation analysis results. The results of the relative importance analysis showed that 54.9% and 51.6% of the ET and Q were attributable to acid deposition. The P contributed 23.3% and 27.3% of the observed variability in ET and Q, while VPD explained 19.3% and 18.1% of the observed variability in ET and Q, respectively (Fig. 2d, e). Therefore, these results suggested that acid deposition acts as an important contributor to the increased ET and decreased Q observed over the study period.

Fig. 4: The comparison of partial correlation coefficients between evapotranspiration (ET), streamflow (Q) and potential drivers.
figure 4

a, b Partial correlation coefficients between ET, Q and potential drivers, respectively. The black column indicates the partial correlation coefficient. The selected drivers are precipitation (P), net radiation (Rn), vapor pressure deficit (VPD), atmospheric CO2 concentration (CO2), and acid deposition. * and ** indicate that the significant partial correlation coefficients at p < 0.05 and p < 0.01, respectively.

Full size image

Discussion

Within our field experimental design, data representing the leaf and xylem hydraulic traits of trees growing under simulated acid deposition allowed us to separate the effects of acid addition from other factors. Our evidence of the morphological and physiological traits of leaves and stems at the living tree level revealed that the trees experienced a short-term increase in stem radial growth with a greater xylem hydraulic conductivity to meet the large water use during the initial 6–7 years of treatment. For the initial short-term period, tree growth responded positively to the improved soil fertility due to the unparalleled level of nutrient availability released into the soil solution after acid addition in the N-rich subtropical forest. Our results were consistent with a temporary increase in ET due to the fertilizer effect of calcium addition at the HBEF20. The direct restoration of a limiting nutrient in the HBEF case study20 and the input N and increased base cation mobilization after acid addition in our field experiment essentially improved forest fertility in the early years of treatment, thus stimulating plant growth and boosting tree water use. Our results further provide physiological evidence of stem anatomical traits characterizing the increased plant transpiration response to the initial phase of acid addition.

However, this increased stem radial growth cannot be sustained, as indicated by the reduced RW under continuous acid addition compared with the RW at CK (Fig. 1). Similar patterns of initial increasing radial growth followed by decreasing growth under prolonged acid deposition have been found in other studies36,37,38. In contrast to acid addition not significantly changing soil N concentration after 12-year, continuous acid input will eventually reach an excessive level and accelerate the base cation leaching loss, thus reducing the stocks and availability of base cations (Supplementary Fig. 9), which in line with the previous finding that the base cations depletion as a consequence of ongoing acid deposition39. The observed decreased foliar Ca content suggested that plant Ca availability was consistently reduced under acid addition (Table 1). Given the signaling roles of Ca in regulating stomatal closure16 coincided with lowered soil Ca availability induced by continuous acid addition, trees with a higher leaf stomatal density may uptake more water to drive soil Ca flows to roots and satisfy Ca demand and thus induce a lower water use efficiency21,22. Noteworthy is that litterfall (a proxy of plant primary productivity) showed no significant differences among acid addition and controlling plots (Supplementary Fig. 10). These results indicated that lowered base cations availability may drive plants to allocate more photosynthates to belowground for growing roots and increase root exudations to mycorrhiza to improve uptake of soil water and nutrients, thus reducing photosynthates for stem radial growth. This interpretation is strongly supported by a recent work in tropical forests, suggesting who found that exchangeable base cation availability primarily drive plant to allocate more photosynthates to belowground that in turn reduce aboveground stem growth40. Moreover, an acid addition study at the DBR site found that the plant allocates more photosynthates to mycorrhizal fungi to alleviate the acidification-induced soil phosphorous limitation in the same acid additionplots41. This result probably explained that no significant differences in litterfall among treated plots. Indeed, recent studies highlighted that soil nutrient availability plays a key but overlooked role in plant carbon allocation and growth. In line with the recent conceptual framework42,43, our data on lowered soil base cation availability induced by long-term acid addition may drive plants to deliver more photosynthates to belowground to enhance nutrient availability, subsequently reducing photosynthates to stem radial growth (Fig. 1). Combining the decreasing soil Exch. Ca, shifts in the foliar Ca contents of dominant trees and forest inventory data, the increased plant transpiration adjusted to the nutrient imbalances induced by long-term acid deposition at the individual tree level can be scaled up to a higher vegetation community. These findings potentially have great implications in soil geochemistry effects on watershed hydrology through mediating plant physiological functions.

The watershed-scale ET, calculated by the water-balance method, exhibited an increasing trend; consequently, Q decreased from 2000 to 2018. The abovementioned results raise important questions about the contribution of acid deposition to regulating hydrological processes in a forested watershed. Our results showed that ET and Q are more sensitive to ω than P or PET in our watershed, as derived from the Budyko framework. In addition, the relative contributions of ω to the increased ET and decreased Q were approximately 62.9% and 58.4%, respectively; these values were higher than the relative contributions from P and PET. These results indicated that ω has become a dominant driver regulating ET and Q changes and represents the watershed topography, vegetation cover and soil properties. These results are in line with previous studies that found ω is the major cause of streamflow dynamics44,45. At our study site, we assume that the topography and vegetation cover remain stable because this watershed is located within a biosphere reserve and has not been disturbed. Thus, we argue that the observed changes in ET and Q at the watershed scale result from tree transpiration adjusted to altered soil geochemical properties induced by acid deposition. Indeed, our data suggested that the forest soils experienced a pronounced acidification process with reduced base cation contents and substantially increased soil Exch. Al over the study period (Fig. 2b, c, Supplementary Fig. 5). The reduced availability of soil Ca, its physiological importance in stomatal regulation, and its coincidence with the increased soil Al stimulated trees to take up more water and thus reduced the streamflow.

In addition, we applied a partial correlation analysis to investigate the responses of ET and Q to climate factors, CO2 and acid deposition. The results indicated that the increased ET and decreased Q were dominantly attributable to P and acid deposition. However, the relative contribution of acid deposition was higher than the relative contribution of P to ET. Although the increased P had positive effects on Q, the negative effects of acid deposition on Q largely outweighed the positive effects of P, thus leading to an overall decrease in Q. In addition, our results indicated that the raised CO2 had little impact on ET during the past two decades, in line with the previous findings that the low nutrient availability may explain these limited CO2 effects46. Indeed, our data on the decreasing soil Ca content support the notion that nutrient availability may play an important role in mediating the vegetation physiological response to the rise of CO2. In particular, subtropical forest ecosystems are enriched with N, and increased plant transpiration is more likely in these areas due to acclimation to low base cations conditions driven by excessive acid input into deeply weathered soils. Our current results illustrated that soil calcium, a rock-derived nutrient, has profound effects on watershed hydrology through tree physiological processes and thus highlighted the importance of the calcium biogeochemical function in ecosystem processes in sub- or tropical forests. Indeed, the recent findings call attention to the impacts of calcium availability, apart from N and P, on tropical forest physiological and ecological processes47. Given the assumption that the nutrient imbalance induced by acid deposition continues, the morphological and physiological adjustment responses of plants to nutrient availability may modulate plant transpiration and nutrient acquisition processes, and thereby regulate the hydrological cycle. In this context, a recent meta-analysis suggested that N deposition could increase leaf area and transpiration rate, thus lead to enhanced vegetation water requirement48. Collectively, we built a conceptual framework, together with the recent findings21,22, to illustrate how acid deposition alters soil biogeochemical properties linked to tree physiological acclimation at the tree level and further extends to hydrological processes at the watershed scale (Fig. 2). These results highlighted the critical role of vegetation physiological response to nutrient availability in watershed hydrology.

We found that the response of plant growth to acid deposition has a detectable impact on watershed discharges based on field manipulation experiments and watershed-scale observations. Our results thus improve the understanding of how acid deposition affects hydrological cycles through vegetation physiological responses. We do not mean to diminish the vital roles of climate conditions and land-use changes in determining hydrological processes32,49,50. We also understand that our results have uncertainties. Since the DBR watershed is located in a biosphere reserve and vegetation cover is nearly complete, we used ET as a proxy for vegetation transpiration, not separating soil evaporation. We quantified and interpreted long-term changes in ET based on the water-balance method, which relied on field observations of P, Q and soil moisture storage through meteorological and hydrological measuring stations. The observed P data agreed well with P extracted from remote-sensing datasets, indicating that the field observation-based P is reliable in our study (Supplementary Fig. 11). Some uncertainties come from variations in soil water and groundwater storage, which have been assumed to be negligible at an annual scale when calculating watershed ET. Particularly, recent studies emphasized that plants can access deep moisture stored in the soil or regolith layer regardless of the soil moisture content in the shallow layer51,52. Thus, the storage of deep soil or rock moisture can be used to support plant transpiration and cannot be neglected, particularly in subtropical soil associated with highly weathered regolith and deep-rooted plants53,54. The watershed hydrological dynamics coevolved with the climate, vegetation, soil, and regolith properties55. To reduce the uncertainties in assessing the magnitude of these factors impact on water balance components, it is necessary to continue field manipulation experiments and observations on meteorological, hydrological and edaphic properties and other ancillary data, such as watershed architecture and groundwater depth. We also acknowledge that we only monitored the stem xylem anatomy and hydraulic traits of one dominant tree species (Castanopsis chinenes), a ring-porous species that exhibited an increase from 44% to 59% of the total basal area in our study watershed. During this inventory measurement period, Schima superba, a diffusive-porous species, retained 21 to 24% of the total basal area, whereas Pinus massoniana, a nonporous species, decreased from 28% to 17% in the total basal area. The response of tree growth to chronic acid deposition is regarded as species-specific and that diffusive-porous species exhibit reduced growth56. If this species-specific response to acid deposition applies to our study site, in addition to our results, we need to examine more interspecific physiological adjustments to nutrient availabilities, particularly for decreasing base cations and N-rich subtropical forests. In addition, it should be noted that the enhanced Ks of trees result from an increased Dh during the period of plant adjustment to low levels of base cations caused by acid deposition (Supplementary Fig. 1). These results can be interpreted as an increased Ks increasing the risk of hydraulic failure, which could result in tree mortality. Ultimately, an increased Ks drives a shift in species composition and thus changes the watershed hydrology, indicating that a shift from ring-porous to diffusive-porous dominant species can lead to a long-term reduction in streamflow57. These adaptive changes can increase the plant water use and transpiration processes, thus driving nutrients into plant tissues and consequently inducing changes in plant hydraulic structures to fulfill the water transportation demand. Regarding this aspect, our results further motivate the need to clarify how soil nutrient availability mediates tree growth, species succession and its interactions with climate change on forest hydrology in acid-impacted and N-rich forest ecosystems.

Overall, our results provided evidence to advance our mechanistic understanding of increased plant transpiration due to increased growth in the initial years and the reduced availability of base cations under prolonged acid deposition. Despite remaining uncertainties, we expect our results are robust to highlight how plants physiological traits adapt to deposition-driven changes in forest nutrients and provide insights into water cycling at forested watersheds in response to acid deposition. Such results explicitly reinforced that plant response to acid deposition plays a non-negligible factor in mediating hydrological cycle, and provided a biogeochemical understanding of forest watershed hydrology in addition to climate factors.

Methods

Study site

This study was conducted at the Dinghushan Biosphere Reserve (DBR) (23°10′-23°12′N, 112°31′-112°34′E), located in Zhaoqing city, Guangdong Province, China (Supplementary Fig. 12a). The DBR covers an area of 1155 ha and ranges in elevation from 14 to 1000 m above sea level. This region has a typical subtropical monsoon climate. The annual mean minimum and maximum temperatures are 12.0 °C in January and 28.0 °C in July, respectively. The average annual P is approximately 1900 mm, of which approximately 80% occurs from April to September. The highly-weathered soils in this region are classified in the Ultisols order of the United States Department of Agriculture (USDA) soil taxonomy system and are derived from sandstone and shale parent material58. The major tree species are Castanopsis chinensis, Schima superba, Pinus massoniana, Cryptocarya chinensis, Aporusa yunnanensis, Blastus cochinchinensis, Ardisia quinquegona and Psychotria rubra. The DBR joined the Chinese Ecosystem Research Network (CERN) initiated by the Chinese Academy of Sciences (CAS) in 1991 (http://dhf.cern.ac.cn/content?id=40684). The DBR was selected as the pilot station of the National Field Research Station network organized by the Ministry of Science and Technology of China in 1999. A meteorological station (M520, Vaisala, Finland), installed at a hilltop, is used to monitor the precipitation amount, air temperature, relative humidity etc. according to the mission of the Chinese Ecosystem Research Network (CERN) (Supplementary Fig. 12b). For validating the precision of monitoring precipitation dataset, monthly precipitation in DBR for 2000-2018 was also obtained from the China Meteorological Forcing Dataset (CMFD), Multi-Source Weighted-Ensemble Precipitation (MSWEP) and Climate Hazards group Infrared Precipitation with Stations (CHIRPS). The CMFD is based on remote sensing products, reanalysis datasets and in-situ station data and is specifically developed for studies of land surface processes in China. This dataset has a 3-hourly temporal resolution and 0.1° spatial resolution59. The MSWEP is a recently released global precipitation dataset with a 3-hourly temporal resolution, covering the period 1979 to the near present60,61. The dataset is based on the complementary strengths of gauge-, satellite-, and reanalysis-based data to produce global precipitation estimation. The CHIRPS is a new quasi-global (50°S-50°N) precipitation dataset with daily, pentadal and monthly precipitation at a spatial resolution of 0.05°, covering the period 1981 to the near present62.

Sampling and measurement in experimental acid addition plots

The acid deposition simulation in the DBR watershed was initiated in June 2009. All treated plots had similar forest compositions and topography, and the differences among plots arose from different acid addition levels. These plots were surrounded by concrete panels inserted 15 cm into the topsoil layer to prevent surface runoff from flowing out of each plot. Two acid intensities were simulated with solution pH values of 4.0 (pH4.0) and 3.0 (pH3.0). Plots applied to local lake waters with a pH was about 4.5 and a low base cation concentration (Ca: 0.2-1.8 mg L−1, Mg: 0.3-0.7 mg L−1, K: 0.3-1.1 mg L−1 during 2009–2020) were selected as the control group (CK). Further, the large quantity of lake water used each time is convenient to obtain in the actual field condition. Each treatment had three replicated plots and thus, nine plots (10 m × 10 m) were established. The simulated acidic solution was prepared by mixing H2SO4 and HNO3 at a 1:1 mole ratio into the lake water. At each application time, 40 liters were applied to each plot at an interval of two weeks throughout the experimental period. Thus, the added H+ amount was 9.6 and 96.0 mol H+ ha−1 year−1 for the pH 4.0 and pH 3.0 treatments41,63. Meanwhile, the added base cation amount was about 0.23, 0.19 and 0.13 mol H+ ha−1 year−1 for Ca, Mg, K, which was substantially lesser than acid input. The experimental acid solution was sprayed below the tree canopy using a gasoline engine sprayer. Litterfall was monthly collected during 2012, 2020 and 2021. A nylon litter trap (1.0 m × 1.0 m in size and 1.0 m above the ground) with a mesh size of 1 mm was installed in each plot.

Tree foliage was collected in each treated plot in December 2020. Foliar samples were divided into two parts. One part was used for stomatal density and guard cell length measurements, and the other was oven-dried at 60 °C for subsequent nutrient content and carbon isotopes analysis. The measurement methods are described in the supplementary material (Supplementary Table 1). Due to the small plot area (10 m × 10 m), the main tree species in the experimental plots were Castanopsis chinensis, and this species was found to be the dominant species at our study site based on forest inventory data. Thus, two cores per tree with stem diameter >20 cm were collected at breast height using a 5-mm increment borer, and the total numbers of tree-ring samples taken were 6, 8 and 8 for pH 3.0, pH 4.0 and control plots, respectively. The procedure of the tree-ring cores processing follows standard dendrochronological techniques64. Each core sample was glued on a wooden strip and air-dried. The cores were polished with progressively fine sandpaper at different grits (300/600/800/1200) to a clear and flat surface, and then chalk was added to the vessel lumina. Digital images of the polished cores were captured by a scanner (Epson TSD4800 flatbed scanner) at a 2400-dpi resolution. The tree-ring width (RW) was conducted using the ROXAS software from scanned images. The RW from two cores of the same tree was visually cross-dated and also statistically checked by t-test and Gleichläufigkeit values65. Then, the quality of cross-dating was evaluated by the COFECHA program and then the cross-dated samples were detrended with the negative exponential model method using the Dendrochronology program library in R (dplR) conducted in R statistical software66,67. Finally, all vessel parameters (e.g. vessel number, tangential and radial diameter, vessel area) were measured for each dated tree-ring and then calculated hydraulic conductivity (Ks) and mean vessel diameter (Dh) of annual rings in ROXAS software68.

Sampling and measurement in the permanent plot

Southern China has suffered severe acid deposition within the past decades, and the DBR is an appropriate site for studying the effects of acid deposition on the processes and functions of subtropical forest ecosystems1,69,70,71. Forest inventories were conducted in the 1.0-ha permanent plot in 1999, 2004, 2010 and 2015. Trees with diameters at breast height larger than 2.5 cm were included, and the trees were identified and counted by species name. Soil samples were also collected with four replicates from the 0–20 cm layer in the permanent plot. The soil samples were air-dried, one part was passed through 2-mm mesh, and the other part was passed through 0.25-mm mesh for further chemical measurements. The soil chemical properties, including the soil pHwater, cation exchange capacity (CEC), soil exchangeable potassium (Exch. K), calcium (Exch. Ca), magnesium (Exch. Mg) and aluminum (Exch. Al), were measured in the laboratory. Details of these measurements are provided in the supplementary material (Supplementary Table 1). Soil samples were collected to test soil moisture content (SMC) monthly using a 30-mm diameter hand-held auger at depths of 0–15, 15–30, and 30–45 cm in each permanent plot. The soil samples were oven-dried at 105 °C for 24 h. For each soil profile, soil samples were collected at intervals of 15 cm for soil bulk density measurements using a 100 cm3 soil corer. Thus, the soil water storage at certain depths was calculated based on the average monthly SMC and bulk density.

Watershed hydrology

The daily streamflow (Q) was automatically measured at the outlet of the DBR watershed (6.1 km2) since 2000 and used digital equipment (WGZ-1, Huazheng, Chongqing, China) installed as a rectangular hydrological weir (Supplementary Fig. 12c)72. The staff of DBR resided in this field observatory station and regularly checked streamflow data and collected streamflow samples. Streamflow samples were manually collected at the hydrological weir in January, April, July and October during 2000–2018. Three streamflow samples were collected at each time and in total 228 samples were collected during the study period. The collected streamflow samples were filtered through 0.45 μm filters and stored in a cooler (4 °C). The concentrations of Ca (Stream. Ca), Mg (Stream. Mg) and K (Stream. K) in the streamflow were measured by a flame atomic adsorption spectrophotometer (GBC932AA, Australia), and the average values in stream water were used as the annual mean concentrations. In this study, we selected Fu’s equation73, one of the commonly used functional forms for estimating ET, and examined the dominant drivers of hydrological changes45,74. The details of calculations were provided in the supplementary information.

Data analysis

The annual mean atmospheric CO2 concentration (eCO2) was derived from published datasets obtained from Mt. Waliguan station (35°17′N, 100°54′E), China, under the framework of the World Meteorological Organization/Global Atmosphere Watch75,76,77,78. The water balance-derived annual ET was estimated by subtracting the observed annual Q and interannual soil water storage variations from the observed annual P at the watershed scale20. The annual runoff coefficient was calculated as the ratio of annual Q to the annual P amount. Since acid deposition continuously impacts soil biogeochemical processes, we used cumulative acid deposition as a variable to characterize the effects of acid deposition on hydrological processes.

The nonparametric Mann-Kendall test was used to analyze trends and their significance over the study period. A positive “S” value indicates an increasing trend, while a negative “S” value indicates a trend that is decreasing over time21,79. A partial correlation coefficient was also calculated to detect the statistical contribution of each factor after statistically excluding the effects of other factors80,81. All analyses were conducted using Statistics Package for Social Science (SPSS) Version 2 with significance levels of p < 0.05 and p < 0.01. All the figures were created using SigmaPlot 14.0.