Introduction

Karst groundwater is an important water resource for humans, as about 25% of the global population completely or partially relies on drinking water from karst aquifers (Hartmann et al. 2014). Sustainable and effective management of groundwater resources requires a systematic and comprehensive understanding of groundwater recharge mechanisms to assess any changes in water budget affected by deforestation and climate change (Tayoko and Yoshio 2004). So, the study of groundwater recharge mechanism, including estimating the recharge rate, average residence time and segregating watershed evapotranspiration by evaporation and transpiration, is essential to predict a lot of the contribution to groundwater resources and their available socio-economic development.

Karst region is characterized by an extremely heterogeneous porosity, such as micropores, fissures, and conduits (Ford and Williams 2007; Hartmann et al. 2014; Gholam et al. 2016), which lead to a strong heterogeneity of the water flow at the ground and in the underground (Hartmann et al. 2014). Vegetation impact infiltration rate and flow change the recharge form of rainfall to epikarst zone (Allen et al. 2014; Hiroaki et al. 2013; Deng et al. 2019). Groundwater recharge mechanisms and rates in a forested epikarst spring catchment depend on underlying surface conditions (geology, geomorphology, rock exposure rate, soil distribution and thickness, etc.), plant physiological characteristics (plant species and growth process, etc.) and meteorological factors (rainfall, photosynthetic effective radiation, vapor pressure deficit, and wind speed, etc.) (Deng et al. 2019; Martin et al. 2020). Previous studies on groundwater recharge mechanism mainly focused on karst water process in different watershed (Fu et al. 2016; Reddy et al. 2006), ground-water recharge (Mohamed et al. 2021; Ahmet et al. 2020; Gholam et al. 2016) and karst water resources assessment (Lee et al. 2007; Hartmann et al. 2014).

Transpiration and evaporation are critical components of the water budget of forested catchments (Tayoko and Yoshio 2004; Hu et al. 2018). An important step in groundwater evaluations is to segment evapotranspiration by transpiration and evaporation. Recently, an increasing number of researchers have utilized hydrogen and oxygen isotopes to segregate watershed evapotranspiration by evaporation and transpiration, based on the Rayleigh equation (Maki and Tadashi 1998; Tayoko and Yoshio 2004; Liu et al. 2006; Hu et al. 2018). The evaporation to evapotranspiration ratio was estimated to be between 3 and 21.4%, by contrasting the differences between the initial isotopic content of groundwater and the final content at a catchment outlet (Maki and Tadashi 1998; Tayoko and Yoshio 2004; Liu et al. 2006; Hu et al. 2018). Furthermore, the empirically estimated d-excess values (d-excess = δD − 8 × δ18O) (Dansgarrd 1964), were also employed to estimate evaporation in watersheds (Huang and Pang 2012; Hu et al. 2018) based on its sensibility of non-equilibrium fractionation processes and the different diffusivities of water stable isotope during the evaporation and transpiration (Mauro et al. 2020). To improve the evaluation of groundwater process and water budgets, groundwater source, residence times, and evapotranspiration should be evaluated systematically.

The main objectives of this study were to estimate groundwater recharge source and groundwater residence times using stable water isotopes to segment evaporation and transpiration components by d-excess and water balance methods in a forested epikarst spring catchment, Southwest China. Our research could provide new insights on the quantification of components of the water cycle. Furthermore, our results may assist in developing and managing groundwater resources more effectively and systematically in karst forest catchments.

Materials and methods

Study area

The study was conducted in a typical vegetated basin (S31 epikarst spring catchment), in Guilin, Guangxi, China (Fig. 1). Humid and subtropical monsoon dominates in this area with an annual average air temperature of 18.8 °C, average annual precipitation of 1915.2 mm, ~ 80% of which occurs during the rainy season (May to October) (Deng et al. 2020), and an average annual relative humidity of 76% to 82% (Zhu et al. 2017). The main lithology is thick limestone of the upper Devonian Rongxian Formation (D3r) (Deng et al. 2020). The S31 epikarst spring catchment has an area of ~ 1 km2 and consists of three flat depressions and 15 mountains with elevations of 400 and 652 m, which are dominated by karst peak-cluster depressions. The S31 spring is located at the bottom of a hill slope, which occurs off the junction between peak-cluster depression area and peak-forest plain, and flows from the epikarst zone through developed conduits. Preliminary findings showed that No.1 and No.4 depressions (Fig. 1) have good connectivity with No. S31 spring (Chang et al. 2012). The connectivity between No.3 depression and the S31 spring is poor, which may be related to the lack of fracture development at the bottom of No.3 depression. During rainstorms, the surface runoff generated in the three depressions enters the karst conduit through a sinkhole at the bottom of the catchment (Chang et al. 2012). Based on borehole data, the thickness of epikarst zone ranges from ~ 8 to 10 m. The karst aquifers with coexisting conduits and cracks in the research areas have a high spatial heterogeneity, and the spring water is of the Ca–HCO3 type. The discharge varies from 0 to 2.5 m3/s and quickly responds to precipitation (Chang et al. 2012). The catchment has a high limestone outcrop ratio, where soils are discontinuously distributed only within gaps. The average soil thickness is 5–10 cm with a maximum value of 100 cm in depressions. The soils are well drained, gravelly, and calcareous. This catchment experienced severe deforestation in the 1950s and has been under natural restoration for almost 70 years. Currently, natural shrubwood occupies ~ 77% of the total catchment area. Caesalpinia decapetala is the main species in the area, accompanied by Toona sinensis. Caesalpinia decapetala and Toona sinensis communities occupy 70% and 30% of the catchment, respectively.

Fig. 1
figure 1

Location of the study area in China (a) sketched hydrogeology map of the Yaji experimental station (b)

Full size image

Water sampling

Rainwater, throughfall and epikarst water following individual rain events from July 2015 to October 2016 were collected for H and O isotope analysis. Throughfall water sampling was conducted in two of the main vegetation (Caesalpinia decapetala and Toona sinensis) communities. Sixteen water collectors were positioned in a fixed, albeit random, pattern on each vegetated floor to collect throughfall. The rainwater from single precipitation events was collected via a funnel attached to a plastic tank, taking care to prevent evaporation. Spring water was collected while the throughfall water was collected. All water samples were stored in 15-mL capped plastic bottles, wrapped in para-film, and frozen pending stable isotope analysis.

Isotopic analyses

Stable isotopes of oxygen and hydrogen in the precipitation, throughfall, and epikarst spring water were measured using high-precision laser spectroscopy (LWIA-24d, Los Gatos Research, USA) at the Key Laboratory of Karst Dynamics, Institute of Karst Geology, at the Chinese Academy of Geological Science. The hydrogen and the oxygen isotope values are reported with respect to Vienna Standard Mean Ocean Water (VSMOW). The precision of the measurements was within ± 0.1 ‰ for δ18O and ± 0.3 ‰ for δ D.

The d-excess = δD − 8 × δ18O was calculated for all samples (Dansgarrd, 1964). This index describes deviations from the meteoric water line (MWL) and can be used to indicate the kinetic (i.e., no equilibrium) fractionation effects of evaporation (Gat 1996).

Seasonal groundwater recharge

Mass balance methods were employed to assess the proportional groundwater recharge contributions from summer and winter rainfall (Lee and Kim 2007). The distinct d-excess values of summer and winter rainfall were related to two end members for the respective seasons as,

$$d_{{{\text{excess}} - {\text{spring water}}}} = \, Xd_{{ - {\text{Summer}}}} + \left( {{1} - X} \right) \, d_{{ - {\text{Winter}}}}$$
(1)

where d-excess was the deuterium excess value, X was the fraction of summer rainfall, and (1-X) was the fraction of winter rainfall.

Estimation of groundwater residence time

The groundwater residence time (MRT) in catchments was calculated by the seasonal isotopic fluctuation correlation between rainfall and epikarst spring water. The spring catchment was presumed to be mixed well and in a stable state, where MRT was assumed to have an exponential distribution (Reddy et al. 2006). The season isotopic fluctuation tended to a sine functional relationship as follow:

$$\delta = I + A \times \sin \, \left( {ct - \theta } \right)$$
(2)

where I was the annual average δD, δ18O, or d-excess in ‰, A was the seasonal amplitude of δD, δ18O, or d-excess, c was the fluctuation frequency of δD, δ18O, or d-excess, t was the time in days, θ was the phase lag in radians. The spring water MRT was calculated by the following formula:

$${\text{MRT}} = {\text{C}}^{{ - {1}}} \left[ {\left( {{\text{A}}_{{{\text{z2}}}} /{\text{A}}_{{{\text{z1}}}} } \right)^{{ - {2}}} - {1}} \right]^{{0.{5}}}$$
(3)

where Az1 was the seasonal amplitude of δ18O, δD, or d-excess of precipitation or throughfall, Az2 was the seasonal amplitude of δ18O, δD, or d-excess in epikarst water, and C was fluctuation frequency of δ18O, δD, or d-excess.

Estimating the evaporation rate in a vegetated area using the d-excess method

The evaporation rate in a vegetated catchment was calculated using the d-excess method including the dual water stable isotope (D-18O) (Hu et al. 2018). It was presumed that evaporation in a catchment experiences Rayleigh fractionation (Formula 4), where R was the 18O/16O or 2H/1H ratio (Clark and Fritz 1997). For convenience, we employed an equation described by Hu et al. (2018), and d replaced R, thereby using δ to indicate the Rayleigh fractionation (formula 5). So, the relation between d-excess and the spring water residual fraction was derived from formula 6. And 1-f was the evaporation ratio.

$$R = R_{0} f^{(a - 1)}$$
(4)
$$\delta = \, \left( {\delta_{0} + 1000} \right) \, f^{(a - 1)} - 1000$$
(5)
$$\begin{gathered} d = \delta D \, - 8\delta^{18} O = \left( {\delta D_{0} + 1000} \right) \, f^{(aD - 1)} - 1000 - 8 \, \left[ {\left( {\delta^{18} O_{0} + 1000} \right) \, f^{(a^{18}O - 1)} - 1000} \right] \hfill \\ = \left( {\delta D_{0} + 1000} \right) \, f^{(aD - 1)} - 8 \, \left( {\delta^{18} O_{0} + 1000} \right) \, f^{(a^{18}O - 1)} + 7000 \hfill \\ \end{gathered}$$
(6)

where ɑ (the fractionation factor between the product and the reactant) was derived as in formula (7); ɑ* (the equilibrium fractionation factor) was obtained from formula (8) and (9). And εkD or εkO (the kinetic enrichment factor) was derived from formula (10) and (11). Furthermore, δ18O0 and δD0 were the initial groundwater isotopic content (here δ18O0 or δD0 = δ18O0 or δD0 × 1000).

$$a = a_{{{\text{product}}}} - a_{{{\text{reactant}}}} = a_{v - 1} = 1/\left( {a^{*} + \varepsilon k/1000} \right)$$
(7)
$${\text{For}}^{18} O: \, \ln a^{*} = 1000 \times 1.137/T^{2} - 0.4156/T - 2.0667/1000$$
(8)
$${\text{For }}D: \ln a^{*} = 1000 \times 24.844/T^{2} - 76.248/T + 52.612/1000$$
(9)

where the ɑ* for 18O and for D (Hu et al. 2018) was given by the formula (8) and (9). Here, T was temperature (for T in K). Given an annual mean air temperature of 18.8 °C in the Yaji Basin, formula (8) yields a ɑ* value for 18O of 1.010 and formula (9) yields a ɑ* value for D of 1.086.

$$\varepsilon {\text{kD}} = {12}.{5}\left( {{1} - {\text{h}}} \right)$$
(10)
$$\varepsilon {\text{kO}} = {14}.{2}\left( {{1} - {\text{h}}} \right)$$
(11)
$${\text{ER}} = {1} - f$$
(12)

where h was the mean relative humidity given an annual mean air temperature of 80% in the Yaji Basin. ER is the evaporation ratio of the evapotranspiration rate in the catchment.

Chang et al. (2012) showed that the surface runoff only accounted for 0.63% of total discharge of S31 epikarst spring. So, evapotranspiration ET in the study area was calculated through water balance method as follows:

$${\text{ET}} = P - R$$
(13)

where P was precipitation and R was runoff. The evaporation ER (%) in basin was defined as:

$${\text{ER}} = Ef/ET = \left( {1 - f} \right) \, \times TF/\left( {P - R} \right)$$
(14)

Here, the P, R, TF, δ, and δ0 were determined by ourselves in vegetated experimental catchment, and so the ER was calculated.

Results

Seasonal variations in precipitation, throughfall, and epikarst spring water

The annual mean values of the precipitation, throughfall, and spring flow in the epikarst spring catchment during the study period (July 2015–October 2016) are listed in Table 1. The annual weighted means of isotopic content in throughfall were -4.81‰ for δ18O and -29.09‰ for δD, which are lower than those of precipitation. The isotope ratio of epikarst spring water was − 6.20‰ in δ18O and -38.65‰ in δD, which are also lower than those of precipitation and throughfall.

Table 1 Annual weighted mean water quantity, δ18O, δD, and d-excess of precipitation, throughfall and epikarst water (mean + SD)
Full size table

The relationships between δ18O and δD compositions in precipitation, throughfall, and epikarst water are plotted in Fig. 2. A local meteoric water line (LMWL) at the S31 epikarst spring catchment was: δD = 7.96δ18O + 8.67(R2 = 0.97; n = 92). Regression lines describing the isotopic data for different seasons were as follows: δD = 8.04δ18O + 9.11 (R2 = 0.96; n = 48) for the rainy season (from May to October), and δD = 8.19δ18O + 15.96 (R2 = 0.95; n = 18) for the dry season (November—April of the next year). The regression line in the epikarst spring water drawn through these data was: δD = 8.40δ18O + 13.5(R2 = 0.95, n = 40), with δD = 8.50δ18O + 14.15(R2 = 0.97, n = 29) for the rainy season (May–October), and δD =  − 4.65δ18O–73.98(R2 = 0.24, N = 11) for the dry season (November–April of the next year). The isotopes in the epikarst spring water were more stable than those in the precipitation and throughfall.

Fig. 2
figure 2

Seasonal relation between contents of 18O and D in precipitation, throughfall and epikarst spring water (PC was precipitation; TF was throughfall; EP was epikarst spring water; R was rainy season; D was dry season)

Full size image

Seasonal variations in the water quantity and isotopes in precipitation, throughfall, and epikarst water are plotted in Fig. 3. The precipitation δ18O was quite variable; however, the d-excess was stable, but δ18O value changed greatly in the spring water.

Fig. 3
figure 3

Seasonal variations in water quantity and isotopes of precipitation, throughfall and epikarst water (PC was precipitation; TF was throughfall; dashed lines were the annual amount-weighted mean of d-excess)

Full size image

Seasonal spring water recharge

Based on the mass balance formula (formula (1)), with end-member d-excess values of seasonal rainfall and epikarst spring water as input variables, the results indicated that 72.1% of the summer precipitation and 27.9% of the winter rainfall contributed to the spring water recharge. Although the study area was covered with dense vegetation, summer precipitation contributed more than that during winter, which suggested that the evapotranspiration effect was small in this area probably due to the rapid drainage of rainwater belowground through fissures and conduits.

Estimating MRT using δD/δ18O methods

The isotopic content of precipitation, throughfall, and epikarst water was fitted, as shown in Fig. 4. The seasonal variation relationships of δD, δ18O, and d-excess between the precipitation or throughfall and groundwater during the hydrologic year were well mixed, and remained in a stable state in which MRT was assumed to approximate a sine regression (Fig. 4). Each curve had a good sine fitting relationship.

Fig. 4
figure 4

Fitting curve of isotopes and d-excess (The dashed lines was the annual amount-weighted mean)

Full size image

The groundwater MRTs of the vegetated catchment were calculated using formula (3) using δD/δ18O and d-excess valves and are presented in Table 2. Deuterium isotope has a light mass, and hydrogen is eight times more affected by evaporation at the same temperature than that of oxygen. So, oxygen isotope is usually more stable than hydrogen isotope. The results of MRT calculated by oxygen isotope should be more reliable than that of hydrogen isotope and d-excess. Many studies also use oxygen isotope to calculate MRT (Maloszewski et al. 2002; Reddy et al. 2006; Rusjan et al 2019). Thus, the elapsed time from precipitation to spring discharge was 121 days using δ18O values, and the durations from throughfall to spring discharge were 131 days using δ18O values.

Table 2 The results of MRT
Full size table

Estimation of evaporation rate using d-excess methods

The d-excess method was employed to estimate the evaporation amount, according to the isotopic content of precipitation and groundwater in a hydrological year. The data in Table 1 were used for the calculation. The evaporation proportion was calculated using formula (6). The calculated evaporation rates were 3.16% using the precipitation method (Table 3). The transpiration rate in the S31 epikarst spring catchment was 51.22% (Table 4) according to the water balances method and the evaporation calculated by the d-excess. Thus, the transpiration proportion in evapotranspiration was 94.19%, which indicates that the catchment water cycle was dominated primarily by vegetation transpiration.

Table 3 δD, δ18O, and d-excess of spring water and the evaporation proportion (ER) by the d-excess method
Full size table
Table 4 Transpiration for the S31 epikarst spring catchment
Full size table

Discussion

Assessment of groundwater recharges ratio and groundwater renewal capacity

Karst acquirers are characterized by special hydrogeological conditions related to the dual media of fractures and conduits. Dispersed and centralized recharges are the two main sources of groundwater supplies. Dispersed recharge consists mainly of local flow with percolation through fissures, whereas concentrated recharge primarily consists of surface runoff, which rapidly concentrates large volumes of water that flows into sinkholes (Chang et al. 2012). Karst aquifers have well-developed drainage systems with quick responses to rainfall (Fu et al. 2016; Deng et al. 2019). Groundwater recharge occurs due to precipitation, runoff, and recharge from surface water bodies (Mohamed et al. 2021). Previous groundwater recharge studies showed that the main recharge occurs in depressions (Chang et al. 2012). Tracer experiments revealed that 80% of the tracer in No.1 depression flowed out through S31 spring with an arrival time from 3 h 32 min to 7 h 39 min, while 31% of the tracer in No.4 depression flowed out through the S31 spring with an arrival time from 8 h 12 min to 11 h 22 min (Chang et al. 2012). However, this study only studies centralized recharge, not dispersed recharge.

In our study, the contributions of epikarst water recharge were 72.1% from summer precipitation and 27.9% from winter precipitation, which were in alignment with the timing of annual precipitation, ~ 80% of which occurred during the summer (Deng et al. 2020). Our studies also revealed that the renewal speed of epikarst water was rapid ~ four months (Table 2). The surface was alternately covered with thin skeletal soils and exposed rock, which allows rainwater to recharge the karst aquifer quickly, leading to the faster speed conduit flow and the shorter MRT (Table 2). Furthermore, the MRT from throughfall to epikarst discharge (131 days) was a little longer than that from precipitation to epikarst discharge (121 days), which signified that the canopy interception prolonged the residency time of water within the system. The effects of fine-scale throughfall patterns on the generation of subsurface stormflow were minor (Hopp and McDonnell 2011). The groundwater hydrochemistry depends mainly on long-term water and rock interactions during water flow through fractures and conduits (Khan et al. 2019). Canopy interception should be given greater attention in the water management of forested karst catchments, because of uneven spatial and temporal rainfall distribution. The above results are helpful for further study of hydrological processes and water resource control in the forested catchments of karst regions.

The difference in MRT results calculated by different isotopic methods may be related to not only isotope quality and isotope diffusion but also sampling method and study area. Our uniform sampling during the year is not only used to calculate MRT but also seasonal groundwater recharge and evaporation rate. The MRT method generally calculates the average residence time of slow flow without considering the influence of fast flow in karst area. So, to improve the accuracy of calculation MRT, it is suggested to take samples after the flood peak when the spring water is relatively stable. Because of this feature, this MRT method is also more applicable in the watershed with large watershed area, relatively sufficient water–rock interaction, and small spatial and temporal distribution difference of rainfall in the watershed.

Comparison of evaporation rates in forest catchments

Accurate and precise estimations of transpiration and evaporation rate are important. Two methods (direct and indirect) have been developed to quantify evaporation. Direct methods, such as field measurements using a lysimeter, provide high accuracy data of a small area (Kubota and Suzuki 1987; Zhao et al. 2015). However, extrapolating direct measurements of evaporation ratios for a small research site to a large area is difficult (Navneet and Jaywant, 2020). The indirect method estimates evaporation ratios by integrating a number of other measured data, such as rainfall, temperature, air pressure, gust velocity, and their correlations with evaporation, such as radiation, energy (Tajchman 1972), and water balances (Lane et al. 2004). However, these methods suffer from poor spatial representation, or are time-consuming and depend on the availability of data required to calibrate and validate the areal model (Navneet and Jaywant, 2020). In our studies, evaporation rates were first quantified by d-excess, and then transpiration ratios were calculated combining water balance methods. This methodology is useful for quantifying groundwater budgets and groundwater resource assessment in karst area.

Previous research (in Table 5) has shown that the estimated ER (3.2%) using d-excess was similar to previous results for similar land cover (Tayoko and Yoshio 2004; Maki and Tadashi 1998; Liu et al. 2006), albeit lower than that of deciduous forests (8–10%) (Wilson et al. 2001), pine forests (8–16%) (Tayoko and Yoshio 2004), tropical eucalypt plantations (15–26%) (Lane et al. 2004), bare rocky land and bare soil land (7.1% and 21.4%, respectively) (Hu et al. 2018). The results from Hu et al. (2018) revealed that the evaporation rate of bare rocky land was lower than that of bare soil land, as precipitation can be quickly and directly recharged in the epikarst zone through fissures and conduits, resulting in little surface runoff. Chang et al. (2012) further confirmed that surface runoff accounted for only 0.63% of the total discharge of the catchment of the S31 spring. The karst rock surface with large pore spaces and shallow soil greatly reduced the evaporation from the catchment. Furthermore, the dense and multi-layered canopies of subtropical forests effectively reduced evaporation from the forest floor (Tayoko and Yoshio 2004; Hu et al. 2018).

Table 5 Evaporation ratios of the forest floor in previous studies
Full size table

The transpiration proportion in evapotranspiration in the S31 epikarst spring catchment was 51.22%, which suggested that the water flux from the vegetated catchment to atmosphere was primarily dominated by vegetation transpiration. In view of this advantage, vegetation transpiration should be paid more attention to in future management of groundwater in catchments. Plants with low water consumption were the first choice for the ecological restoration of rocky karst desertification. Furthermore, distinguishing evapotranspiration by transpiration and evaporation composition in catchments was significant for researching the nitrogen and carbon cycle, due to the relationship between transpiration and carbon assimilation.

Conclusions

Groundwater recharge mainly originated from precipitation during the rainy season, which averaged 72% of the total annual rainfall. The MRT from throughfall to spring discharge (131 days) was a little longer than that from precipitation to spring discharge (121 days). The evaporation rates estimated by the d-excess method in a typical epikarst spring catchment were 3.16%. The d-excess method is more suited to calculate evaporation in hot and rainy areas. Our research provides a useful reference for the accurate estimation of transpiration rate and evaporation rates in future work.