Spatial variability of shallow soil moisture and its stable isotope values on a karst hillslope

Geoderma 264 (2016) 61–70

Contents lists available at ScienceDirect

Geoderma journal homepage: www.elsevier.com/locate/geoderma

Spatial variability of shallow soil moisture and its stable isotope values on a karst hillslope Jing Yang a,b,c, Hongsong Chen a,b,⁎, Yunpeng Nie a,b, Wei Zhang a,b, Kelin Wang a,b a b c

Key Laboratory of Agro-ecological Processes in Subtropical Region, Institute of Subtropical Agriculture, Chinese Academy of Sciences, Changsha, PR China Huanjiang Observation and Research Station for Karst Ecosystems, Chinese Academy of Sciences, Huanjiang, PR China University of Chinese Academy of Sciences, Beijing, PR China

a r t i c l e

i n f o

Article history: Received 17 March 2015 Received in revised form 30 September 2015 Accepted 4 October 2015 Available online xxxx Keywords: Geostatistics Karst region Soil moisture Spatial variability Stable water isotope

a b s t r a c t Soil moisture (θ) and its stable isotope values are two of the most commonly used parameters for studying hydrological processes in soils. Despite their unique ability to aid in distinguishing between soil water evaporation and plant transpiration, the spatial variability of soil water isotope values is not fully understood. In the current study, 10 m × 10 m grids were established within a 90 m × 120 m plot on a highly heterogeneous karst hillslope. Two sampling campaigns were conducted during the early growing season, on April 15, and during the midgrowing season, on August 18, 2011. Stratified soil samples were collected from the shallow soil layer (0–30 cm) to measure θ and its stable isotope values, which were represented by soil water δD values (δDθ). Related soil properties, land cover and topography were also measured and treated as influencing factors. On both sampling dates, θ decreased with depth, while δDθ stayed constant for all soil layers and were similar to the most recent rainfall values. Additionally, the variance of δDθ was smaller than that of θ, especially on August 18 when the most recent rainfalls had similar δD values. The high contrast between θ and δDθ caused the impacts of other factors on δDθ to be masked by the impact of the recent rainfall. Soil moisture presented a moderate to strong spatial dependence, which was consistent with the spatial variability of the influencing factors that were significantly correlated with θ. Soil water δD value exhibited weak spatial dependence and random spatial patterns that were different from the other influencing factors. Moreover, significant correlations between these same influencing factors and δDθ disappeared after a partial analysis with θ as a controlled variable, which means δDθ was indirectly affected by these influencing factors through θ. This suggests that the spatial variability of δDθ was being controlled at fine scales. Our results highlight the importance of analyzing the spatial variability of δDθ, its influencing factors and θ in shallow soil layers separately. © 2015 Elsevier B.V. All rights reserved.

1. Introduction The O and A horizons are the most active layers in a soil profile, as they are susceptible to rainfall, evaporation and transpiration (Grayson et al., 1997; Brocca et al., 2007). Previous studies reveal the role these horizons play in hydrological processes, such as partitioning of rainfall into surface runoff and infiltration, redistribution of infiltrated rainwater and facilitation of plant transpiration demand (Famiglietti et al., 1998; Chen et al., 2010; Legates et al., 2011). In these studies, soil moisture (θ) and its stable isotope values, were the two most commonly used parameters. Monitoring of θ at different depths can provide Abbreviations: θ, soil moisture; δDθ, soil water δD values; T, temperature; Tmax, maximum air temperature; Tmin, minimum air temperature; Max, maximum; Min, minimum; ET0, reference evapotranspiration; SD, standard deviation; CV, coefficient of variation; ρb, bulk density; CC, clay content; SOC, soil organic carbon; VC, vegetation coverage; EBR, exposed bedrock ratio; RFC, rock fragment content. ⁎ Corresponding author at: Huanjiang Observation and Research Station for Karst Ecosystems, Chinese Academy of Sciences, Huanjiang, PR China. E-mail address: [email protected] (H. Chen).

http://dx.doi.org/10.1016/j.geoderma.2015.10.003 0016-7061/© 2015 Elsevier B.V. All rights reserved.

information on flow path, infiltration mechanisms and evapotranspiration (Stothoff et al., 1999; Daly and Porporato, 2006; Lange et al., 2010). For some purposes, as when θ monitoring is problematic or for water source identification, a stable isotope technique is the best choice (Nie et al., 2010). However, despite the important role it plays in hydrological research and modeling, there is a lack of knowledge around the spatial variability of soil water isotopes. Stable water isotopes, which are ubiquitous and occur naturally in water, are ideal for tracing hydrological processes (Ayalon et al., 1998; Lee et al., 2007; Song et al., 2009). Stable water isotopes in soils are affected by the spatial and temporal variations in rainwater (Jayasena et al., 2008; Hu et al., 2013), even though stable water isotopes in rainwater vary slightly and only on small scales (Kato et al., 2013). Transpiration reduces θ but has no effect on soil water isotope values with the main reason for the positive change in isotope values being evaporation (Dawson et al., 2002). Therefore, without the interference of rainfall, a negative correlation between θ and its stable isotope values (especially in shallow soil layer) is expected. In field conditions, soil water storage capacity and plant transpiration vary across space, which make it

62

J. Yang et al. / Geoderma 264 (2016) 61–70

possible for points in space to retain properties of the antecedent rainfall (Jost et al., 2005). Due to unpredictable rainfall recharge and the stable isotope values themselves, correlation between the spatial variability of θ and its stable isotope values should be quite complicated, especially in a highly heterogeneous region. Geostatistics, first employed in soil science by Campbell (1978), was used to study the spatial variability of sand content and pH. Since then, studies have been conducted to understand the spatial patterns of soil properties and its associated variables (Bruckner et al., 1999; Brocca et al., 2007; Hu et al., 2011). Kriging (Journel and Huijbregts, 1978), a tool in geostatistics, is used to estimate values at unsampled sites, describe the spatial distribution of a variable (Snepvangers et al., 2003; Jost et al., 2005), and reveal correlations to associated factors. Recently, many studies have been conducted in typically heterogeneous regions (especially in karst regions of southwest China) to reveal the spatial variability of θ and other soil properties (Zhang et al., 2011, 2012; Peng et al., 2013). However, very few, if any, studies have looked at the spatial patterns of soil water isotope values, largely because of the prohibitively large number of samples needed, and the cost associated with analyzing them. Even with reduced costs associated with isotope testing, monitoring of θ is difficult to carry out in shallow, rocky and discontinuously distributed karst soils (Chen et al., 2011; Tokumoto et al., 2014), and it is more reasonable to gather detailed information about the spatial variability of soil water isotope values. In this study, 10 m × 10 m grids were established within a 90 m × 120 m plot on a typical karst hillslope in southwest China. Stratified soil samples were collected from the shallow soil layer (0–30 cm) to measure θ and its stable isotope values. Related soil properties and environmental factors were also collected. The objectives of this paper were to (1) study the spatial variability of shallow θ and its stable isotope values, and (2) discuss their correlations/differences and influencing environmental factors. 2. Materials and methods 2.1. Site description The study area is a small catchment (area = 1.14 km2) located in the Huanjiang Observation and Research Station for Karst Ecosystems under the Chinese Academy of Sciences (24°43′58.9″–24°44′48.8″N, 108°18′56.9″–108°19′58.4″E) in Huanjiang County of northwest Guangxi, southwest China (Fig. 1). A subtropical mountainous monsoon climate dominates with an annual (average value for Huanjiang County from 1986 to 2005) rainfall of 1389 mm and temperature of 18.5 °C (Song et al., 2010). The catchment is a typical karst area with a flat depression surrounded by mountain ranges on three sides with an outlet in the northeast. Elevation in the study area ranges from 272 m to 647 m. Hillslopes are steep with 60% having a grade greater than 25°. Exposed bedrock, underlain by weathered dolomite, covers approximately 30% of the area with thin discontinuous soil containing rock fragments (Chen et al., 2011). The vegetation of this area can be classified into three secondary communities: tussock, shrub, and secondary forest. 70% of the hillslopes are dominated by tussocks and shrub. Secondary forest is only found on the continuous dolomite outcrops or in deep soils at the foot of hillslopes (100 cm in depth). In this study, a typical southeast-facing shrubland hillslope was selected (Fig. 1). Soil depth increased while rock fragment content decreased from upslope to downslope. The vegetation is dominated by shrub with a greater density downslope than upslope. 2.2. Samples collection Two sampling campaigns were conducted on April 15 and August 18, 2011 to reflect the variations between the early and vigorous growing seasons, respectively. The study area experienced approximately one week of clear days, after the last rainfall, before the two sampling

campaigns. The rainfalls occurred on April 8 and August 12, 2011, respectively. Stratified (0–10, 10–20, and 20–30 cm) soil samples were collected by grid (10 m × 10 m) sampling using a 3 cm diameter hand auger within a 90 m × 120 m plot. A total of 130 sampling points were collected on the shrubland hillslope. Each sample was measured for gravimetric θ (expressed by the mass percentage of soil water out of dry soil, %) and water stable isotopes. In total, 390 samples (130 samples at each depth) were collected on each sampling campaign. Samples were placed in capped vials, wrapped in parafilm and stored frozen. During the second sampling period, surface (0–10 cm) undisturbed soil samples were collected by a steel cylinder with a volume of 100 cm3 and five replicates of disturbed soil samples (130 samples for each) were collected at each point. Exposed bedrock (bare rock without vegetation cover) ratio (EBR) and vegetation coverage (VC), the percentage of vegetation occupying the ground area in vertical projection, were estimated visually within an area of 2 m × 2 m around each sampling point. Elevation and slope were measured by hand-held GPS (Trimble Jono SC, accuracy 2–5 m) and compass readings at each point. Climatic data including maximum (Tmax) and minimum (Tmin) air temperature, and the amount of each rainfall event (calculated for a 24-h period between 8:00 am and 8:00 am) were obtained from a standard weather station located in the central depression of the catchment. Stable isotopes were measured from rainwater. During each rainfall event, rainwater was collected in a plastic tank with safeguards to prevent evaporation (Nie et al., 2011) to measure stable isotopes. In this study, eighteen rainfall samples were collected and measured, ten before April 15 and eight before August 18. 2.3. Lab analysis Soil moisture was measured by an oven-drying method and soil water was extracted by cryogenic vacuuming (Ehleringer et al., 2000; Goebel and Lascano, 2012; Orlowski et al., 2013). The isotopic compositions of soil water and rainwater samples were measured with a liquid water isotope laser spectroscopy instrument (Lis et al., 2008; Penna et al., 2010, 2012; Wassenaar et al., 2014) (Model DLT-100, LGR Inc.). Results are reported in δ notation relative to V-SMOW as δD


18 
 O ¼ Rsample =Rstandard −1  1000

ð1Þ

where Rsample and Rstandard are the ratio D/H or 18O/16O of a measured sample and a standard sample, respectively. The standard deviation for repeat measurements was ± 1‰ for δD and ± 0.2‰ for δ18O. The δD and δ18O values of soil water were significantly correlated (δD = 8.43δ18O + 14.56, R2 = 0.90) and showed similar trends. Previous studies state that δ18O values of soil water were greatly affected by soil carbonate but δD values were not (Meißner et al., 2014). Only the results obtained by the analyses of soil water δD values (δDθ) were introduced in this study. Undisturbed soil samples were used for bulk density (ρb, measured with a gravimetric method) and soil porosity (obtained by 1 − ρb / ρs, where ρs is the particle density of the soils, which was assumed to be 2.65 g cm−3) measurements. Rock fragment content (RFC), clay content (CC), measured by Matersizer 2000 laser particle size analyzer, and soil organic carbon (SOC), measured with a potassium dichromate heating method (Zhang et al., 2012) were measured using disturbed samples. 2.4. Data analysis A semi-variogram was used to examine the spatial structure of θ and δDθ. The semivariance γ(h) was calculated with the following formula (Bruckner et al., 1999; Hu et al., 2011): 1 X 2 ½Zðxi Þ−Zðxi þ hÞ 2NðhÞ i¼1 NðhÞ

γðhÞ ¼

ð2Þ

J. Yang et al. / Geoderma 264 (2016) 61–70

63

Fig. 1. Location of the study area and schematic map of the sampling plot in a small catchment with contour lines at 20 m intervals in Huanjiang County of northwest Guangxi, southwest China.

where h is lag distance, N(h) is the number of sampling couples of the variable z in the interval h, and Z(xi) and Z(xi + h) are the values of z at positions xi and xi + h. Normally, semi-variograms only make sense when the lag distance is within 1/2 of the maximum sampling interval (Western et al., 1998). In this study, 1/2 of the maximum sampling interval was set as the lag distance. The empirical semivariograms were fitted by theoretical semivariogram models to produce geo-statistical parameters, including nugget variance (C0), sill variance (C0 + C), and range. The nugget–sill ratio, C0 / (C0 + C), was used to characterize the spatial dependence of θ and δDθ. In general, C0 / (C0 + C) b 25% indicates a strong spatial dependence, while N75% indicates a weak or no spatial dependence, otherwise the spatial dependence is moderate. All geostatistical analyses were conducted with the software GS+ 9.0. Different theoretical semivariogram models were used to fit the empirical semivariograms and the best-fit model was chosen basing on the value of R2. The Gaussian model was the most suitable for θ and the Exponential model was the most suitable for δDθ. The formulas for the

Gaussian model (3) and the Exponential model (4) were written respectively as:   2 −h γðhÞ ¼ C0 þ C 1−e a2

ð3Þ

  h γðhÞ ¼ C0 þ C 1−e−a :

ð4Þ

pffiffiffi Range for the Gaussian model was 3a, and 3a for the Exponential model. Kriging maps were drawn in ArcMap 9.2 using the parameters obtained from GS+ 9.0. Some basic statistics, including the mean, standard deviation (SD), coefficient of variation (CV), and maximum (Max), and minimum (Min) values of θ and δDθ in each soil depth were analyzed. ANOVA was employed to detect the differences of θ and δDθ between depths. All data including θ, δDθ and environmental factors (ρb, porosity, CC, SOC, VC, EBR, RFC, elevation and slope) were checked for normality

64

J. Yang et al. / Geoderma 264 (2016) 61–70

using a one-sample Kolmogorov–Smirnov test. All except VC and EBR showed normal distribution (P N 0.05). Pearson (for normally distributed data) and non-parametric Spearman's rank (for non-normally distributed data, VC and EBR) correlation, and partial correlation (analyzed the correlation between δDθ and influencing factors by using θ as a controlled variable) analyses were employed to explore the main influencing factors of surface (0–10 cm) θ and δDθ. Reference evapotranspiration (ET0), which could be used to reflect the water depletion, was calculated by the Hargreaves and Samani (1985) model, which is expressed as:

ET0 ¼ 0:0023Ra

 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi T max þ T min þ 17:8 T max −T min 2

ð5Þ

where Ra was the extraterrestrial radiation (mm d−1). The data were analyzed by using software SPSS 18.0.

(8.5 mm) and August 12 (9.9 mm), with corresponding δD values of −77.8‰, −77.6‰ and −68.4‰, respectively. 3.2. The basic statistics of θ, δDθ and δD values of rainwater As shown in Table 1, the weighted average δD value of rainwater before the two sampling dates was − 34.9‰ and − 70.2‰, respectively. The δD value varied more between March and April than between July and August. Mean θ decreased with soil depth on both dates and was slightly higher on April 15. Average δDθ at each depth was close to the δD values of recent rainfalls (Fig. 2) and showed different vertical distributions on April 15 and August 18, which tended to decrease with soil depth on April 15 but increase on August 18. Soil moisture at each soil depth showed moderate variance (0.1 b CV b 1.0) while δDθ showed moderate variance on April 15, but weak variance (absolute CV b 0.1) on August 18. Generally, absolute CV of θ was greater than that of δDθ. 3.3. Geostatistical analysis of θ and δDθ

3. Results 3.1. Meteorological features and δD values of rainwater Fig. 2 shows the time series of rainfall, T, and ET0 from March to August in 2011 as well as the δD values of several antecedent rainfalls. The mean values of ET0 and T from July 18 to August 18 (5.9 mm d−1 and 27.0 °C) were much higher than those from March 15 to April 15 (2.8 mm d−1 and 19.2 °C). The antecedent cumulative rainfalls were 73.4 mm and 86.5 mm for April 15, 2011 and August 18, 2011, respectively. Within the month prior to April 15, relatively large rainfalls (N10 mm) occurred only on March 15 (15.0 mm) and March 17 (24.8 mm), which had δD values of −52.0‰ and −55.9‰, respectively. Rainfalls from April 2 to April 8 were small (b10 mm) with their δD values ranging from − 4.4‰ to 2.8‰. Between July 18 and August 18, two large rainfalls took place on August 5 (26.0 mm) with a δD value of −70.1‰, and on August 10 (33.6 mm) with a δD value of −76.6‰. Three small rainfalls took place on August 8 (5.4 mm), August 11

As shown in Table 2, the Gaussian model fit θ (R2 N 0.98) at all depths during both sampling times and mimicked the spatial structure of θ. The Exponential model fit δDθ (R2 N 0.70) at all depths except the 10–20 cm and 20–30 cm layers on April 15 (R2 b 0.35). Nugget values were small for θ, but large (especially for 10–20 cm and 20–30 cm depths on April 15) for δDθ. The C0 / (C0 + C) values of θ ranged from 17.2% to 38.1%, which revealed a moderate to strong spatial dependence. Soil water δD value showed an over-all weak to moderate spatial dependence. The range, or the spatial dependence distance, was 106 m to 140 m for θ. However, the range of δDθ tended to be zero due to the large C0 (no exact range was gained). 3.4. Kriging maps of θ and δDθ Soil moisture was distributed patchily and increased regularly, from top to bottom, along the hillslope (NW to SE) (Fig. 3a and b). They showed similar patterns at all the three depths on both April 15 and August 18, 2011. On the other hand, larger patches were observed in

Fig. 2. Time series of rainfall, air temperature (T), ET0 and δD value of rainwater from March to August, 2011 in the study area and surface 0–10 cm soil moisture and its δD value on April 15 and August 18, 2011. Arrows indicate the dates of samplings.

J. Yang et al. / Geoderma 264 (2016) 61–70

65

Table 1 Basic statistics of soil moisture (θ), δD value of soil water (δDθ) and rainwater on April 15 and August 18, 2011. Sampling dates

Statistic parameters

April 15

N Mean (%) Max (%) Min (%) SD (%) CV⁎

August 18

N Mean (%) Max (%) Min (%) SD (%) CV⁎

θ (%)

δDθ (‰)

δD value of water (‰)

0–10 cm

10–20 cm

20–30 cm

0–10 cm

10–20 cm

20–30 cm

130 33.5a 61.8 5.1 13.5 0.40 130 27.9a 49.7 5.8 10.6 0.38

130 28.4b 58.2 5.2 13.4 0.47 130 25.9ab 50.0 5.7 11.6 0.45

130 24.4c 52.9 3.7 14.1 0.58 130 24.1b 50.8 5.0 13.2 0.55

130 −20.5b −6.0 −50.2 7.8 0.39 130 −69.1a −49.0 −81.1 5.4 0.08

130 −25.0a −8.7 −50.3 7.6 0.29 130 −69.8a −53.7 −85.7 5.9 0.09

130 −24.0a −7.1 −50.4 9.4 0.36 130 −67.5b −48.7 −79.8 6.8 0.10

10 −34.9 2.85 −55.9 23.4 0.67 8 −70.2 −42.3 −77.6 14.2 0.20

The lowercases after the number indicate the significance between soil depths. Numbers with the same lowercase are not significant at P b 0.05. ⁎ Because the δDθ was negative, its CV was calculated to be mathematically negative. In order to compare with the CV of θ, absolute CV of δDθ was used in this study.

the kriging maps of δDθ (Fig. 3c and d). This was partially caused by the flat effects of ordinary kriging interpolation. However, certain spatial patterns were observable. The δDθ of the 0–10 cm surface tended to decrease from upslope to downslope on April 15, but increased on August 18.

θ as a controlled variable, the correlation between surface δDθ and these same influencing factors disappeared (P N 0.05). 4. Discussion 4.1. Vertical and horizontal variations of θ and δDθ

3.5. Variations of soil properties and environmental factors 3

As shown in Fig. 4, ρb decreased along the slope from 1.3 g cm to 0.8 g cm3. On the contrary, the other three soil properties (porosity, CC and SOC) increased from top to bottom, 51.4%–66.3%, 16.9%–40.5% and 31.0 to 60.7 g kg− 1, respectively. Vegetation coverage ranged from 75% to 100% and was higher downslope. The exposed bedrock ratio was distributed randomly with a wide range from 0 to 79%. Rock fragment content was high upslope (up to 57.4%) but negligible downslope. Elevation and slope decreased along the slope. 3.6. Correlations between surface (0–10 cm) θ, δDθ and influencing factors Very significant correlations (P b 0.001) were found between surface (0–10 cm) θ and δDθ on both April 15 and August 18. Interestingly, the correlations were negative on April 15 (Fig. 5a) and positive on August 18 (Fig. 5b). All factors including soil properties (ρb, porosity, CC and SOC), land cover (VC, EBR and RFC) and topography (elevation and slope) were very significantly (P b 0.01) correlated with surface θ for both sampling times (Table 3). Soil properties and land cover (except SOC and EBR) factors were significantly(P b 0.05) and topography was very significantly correlated with surface δDθ on April 15. All the factors except EBR were very significantly correlated with surface δDθ on August 18. Again, the correlations between influencing factors and surface δDθ were reversed on two days, as was found for the correlations between surface θ and δDθ. After the partial correlation analysis using

4.1.1. Higher θ in early growing season Soil moisture was higher on April 15 (early growing season) than on August 18 (vigorous growing season) (see Table 1) and was affected by antecedent rainfalls and evapotranspiration (Grayson et al., 1997; Brocca et al., 2007). The antecedent rainfalls were similar before the two sampling times, 73.4 and 86.5 mm, respectively, which might lead to similar soil water conditions. However, the evapotranspiration was completely different for the two sampling times. In August, evaporation was higher for a large T and resulted in lower θ (Gerrits et al., 2010) in August than in April. Actually, the ET0 measured in August was two times higher than that in April (Fig. 2), which also suggests greater water loss in August. On the other hand, shallow soil water is an important water source for plants with roots that are always growing horizontally in this layer (Nie et al., 2011, 2014). Plant transpiration in the vigorous growing season was higher, resulting in a lower θ (Fig. 2 and Table 1), than in the early growing season (Nie et al., 2011). 4.1.2. Vertical variation of θ and δDθ In both the early and vigorous growing seasons, mean θ decreased with an increase in soil depth (Table 1). Many previous studies show an increase in θ with soil depth because of higher evaporation in the surface soil layer (Hsieh et al., 1998). In this study, the antecedent rainfalls just before the two sampling dates were minimal (b10 mm) and were mostly depleted by canopy interception and evaporation. Modest amounts of rain reached the soil surface and even less infiltrated downward. On the other hand, the litter horizon and high soil organic matter

Table 2 Geostatistics of shallow (0–30 cm) soil moisture (θ) and soil water δD value (δDθ) on April 15 and August 18, 2011. Sampling date

April 15

August 18

Soil layers (cm)

0–10 10–20 20–30 0–10 10–20 20–30

θ

δDθ 2

Model

C0

C0 + C

Range (m)

R

C0 / C0 + C (%)

Model

C0

C0 + C

Range (m)

R2

C0 / C0 + C (%)

Gaussian Gaussian Gaussian Gaussian Gaussian Gaussian

0.005 0.008 0.008 0.006 0.005 0.008

0.029 0.021 0.026 0.018 0.022 0.026

137 106 114 141 140 137

0.99 0.99 0.99 0.99 0.99 0.99

17.2 38.1 30.8 33.3 22.7 30.8

Exponential Exponential⁎ Exponential⁎

43.8 53.5 78.5 14.8 24.1 34.2

87.6 56.4 99.6 31.6 56.7 77.7

597 − − 77.1 415 632

0.71 0.21 0.35 0.94 0.90 0.87

50.0 94.7 74.8 46.8 40.7 44.0

Exponential Exponential Exponential

Labels are nugget variance (C0), sill variance (C0 + C) and nugget–sill ratio (C0 / C0 + C). –Indicates that the range in this depth was not calculated because of the poor fitting result and large C0. ⁎ Indicates that the Exponential model did not fit δDθ very well in these two depths. We listed them here for the convenience of comparing with the other δDθ data, and no other models could fit them well either.

66

J. Yang et al. / Geoderma 264 (2016) 61–70

Fig. 3. Interpolated maps of shallow (0–30 cm) soil moisture (θ) and its δD value (δDθ). The four panels are (a) θ on April 15; (b) θ on August 18; (c) δDθ on April 15; and (d) δDθ on August 18, 2011, respectively.

content in surface soil are beneficial for water conservation (Rawls et al., 2003; Sayer, 2006). Furthermore, soil in these karst regions is shallow with a mean soil depth of about 30 cm (Chen et al., 2010). The regolith material at the deeper soil layers has low water storage capacity (Cui et al., 2002) that results in low θ. Therefore, decreasing θ with soil depth was found in this study. The vertical distribution of δDθ differed from the two sampling dates. It decreased with soil depth on April 15 but increased on August 18, despite the decrease in θ with soil depth at both sampling times. Normally, the surface layer has a more positive δDθ than the deeper layer because evaporation begins at the soil–atmosphere interface and works progressively downward (Hsieh et al., 1998). A more negative δD value for the topsoil was found on August 18. Actually, the negative stable isotope values in the upper soil layers have been found to exist due to the precipitation (Hsieh et al., 1998; Song et al., 2009; Zhao et al., 2013), and condensation has been shown to deplete the stable isotope of soil water by 5‰ at a 10–15 cm depth (Takahashi, 1998). Condensation, which occurs in the upper soil layers at night, results in more negative

stable isotopic values in topsoil (Saha et al., 2009). The average T gradient between day and night from March 1 to August 31 was 9.4 °C while the maximum T gradient was 18.9 °C. During the second sampling period, the T gradient ranged from 11.6 to 17.3 °C (14.6 °C on August 17) which indicated a strong possibility of condensation. Therefore, condensation resulting from the high T gradient could be the reason for negative δD values in the topsoil. 4.1.3. Greater variance of θ than that of δDθ Absolute CV at each depth of θ was greater than that of δDθ for both sampling times. The primary factor influencing θ and δDθ was rainfall, including rainfall amount, intensity and the δD values (Song et al., 2009; Chen et al., 2010). Although the δD values of rainwater might differ among rainfalls (Jayasena et al., 2008; Hu et al., 2013), the differences between the antecedent rainfalls before each sampling campaign were not obvious in our study (Fig. 2). Furthermore, spatial variation of the δD value of rainwater at a small scale is negligible (Kato et al., 2013). Thus, rainfall might not be the factor causing the spatial variations of θ

J. Yang et al. / Geoderma 264 (2016) 61–70

67

Fig. 4. Interpolated maps of soil properties and environmental factors. Labels are bulk density (ρb), clay content (CC), soil organic carbon (SOC), vegetation cover (VC), exposed bedrock ratio (EBR), and rock fragment content (RFC), respectively.

and δDθ. The other main influencing factor of spatial variation in θ and δDθ was evapotranspiration (Grayson et al., 1997; Dawson et al., 2002). Based on previous studies, evaporation affected both θ and δDθ, while transpiration only depleted θ without causing fractionation of δDθ (White et al., 1985; Hsieh et al., 1998; Nie et al., 2011). Plant utilization varied in space and induced spatial variation of θ but not δDθ. Furthermore, spatial variation of soil water storage capacity resulted in a variance of θ (Jost et al., 2005). Therefore, θ showed larger variances than δDθ. 4.2. Spatial patterns of θ and δDθ Moderate to strong spatial dependence of θ was found in this study. These results were in accordance with those conducted in non-karst areas (Western et al., 1998; Hu et al., 2011) and also in other karst regions (Zhang et al., 2011; Jia et al., 2013; Peng et al., 2013; Penna et al., 2013). Soil moisture showed moderate to strong spatial dependence with small C0 / C0 + C ratios (from 17.2% to 38.1%, listed in

Table 2). A moderate spatial dependence of θ was also found in a karst depression, but at a larger scale (Zhang et al., 2011). Röver and Kaiser (1999) interpreted low and moderate spatial variability as interactions that took place among physical, chemical, and biological soil components. The range, representing the zone of influence for θ, can be used to evaluate the spatial continuity. It was characterized between 100 m and 140 m on the two sampling dates (Table 2). A larger range, up to 300 m, was observed by Zhang et al. (2011) in a karst depression. The range can be used to delineate the size of spatial classes, and was used to depict θ areas, thus the spatial frequency of θ changes (Fitzjohn et al., 1998). Therefore, the ranges of θ demonstrated a variable θ pattern characterized by distinct areas, which were spatially isolated and uncorrelated with adjacent patches (see Fig. 3a and b). Spatial distribution of θ showed that it increased from high to low (Fig. 3a and b). This was correlated with the spatial distribution of soil properties and environmental factors (Fig. 4) (detailed correlations were discussed in Section 4.4). In this case, θ could be evaluated according to the characteristics of these soil properties and environmental factors.

Fig. 5. Correlations between surface (0–10 cm) soil moisture and its δD value on (a) April 15 and (b) August 18, respectively.

68

J. Yang et al. / Geoderma 264 (2016) 61–70

Table 3 Correlations between environmental factors and surface (0–10 cm) soil moisture (θ) and soil water δD value (δDθ) on April 15 and August 18, 2011. Variables

Sampling date

Parameter

ρb

Porosity

CC

SOC

VC

EBR

RFC

Elevation

Slope

θ

April 15

R P R P R P R P R’ P’ R’ P’

−0.573⁎⁎ 0.000 −0.604⁎⁎ 0.000 −0.198⁎ 0.025 0.313⁎⁎

0.616⁎⁎ 0.000 0.593⁎⁎ 0.000 0.174⁎ 0.049 −0.306⁎⁎

0.700⁎⁎ 0.000 0.670⁎⁎ 0.000 0.215⁎ 0.014 −0.375⁎⁎

0.502⁎⁎ 0.000 0.526⁎⁎ 0.000 0.033 0.709 −0.272⁎⁎

0.421⁎⁎ 0.000 0.364⁎⁎ 0.000 0.271⁎⁎ 0.002 −0.282⁎⁎

−0.692⁎⁎ 0.000 −0.627⁎⁎ 0.000 −0.211⁎ 0.017 0.331⁎⁎

−0.782⁎⁎ 0.000 −0.651⁎⁎ 0.000 −0.239⁎⁎ 0.006 0.272⁎⁎

−0.651⁎⁎ 0.000 −0.579⁎⁎ 0.000 −0.236⁎⁎ 0.007 0.302⁎⁎

0.000 0.111 0.209 −0.114 0.200

0.000 −0.082 0.356 0.085 0.341

0.000 −0.145 0.101 0.130 0.143

0.002 −0.09 0.313 −0.074 0.405

0.001 −0.143 0.106 0.170 0.055

−0.395⁎⁎ 0.006 −0.392⁎⁎ 0.000 0. 078 0.380 0.106 0.254 −0.168 0.064 −0.087 0.328

0.000 0.081 0.359 −0.128 0.148

0.002 −0.073 0.411 −0.162 0.066

0.000 0.059 0.506 −0.165 0.062

August 18 δDθ

April 15 August 18

δDθ’

April 15 August 18

Number of samples was 130. Labels are bulk density (ρb), clay content (CC), soil organic carbon (SOC), vegetation cover (VC), exposed bedrock ratio (EBR), rock fragment content (RFC), correlation coefficient (R), significant value (P), and partial correlation coefficient (R’), respectively. ⁎ The correlation was significant at the 0.05 level. ⁎⁎ The correlation was significant at the 0.01 level.

In comparison to θ, δDθ varied randomly with weak spatial dependence and large nugget effects (large C0 as shown in Table 2). The nugget effect represents the undetectable experimental error and field variation within the minimum sampling space. A large nugget effect suggests that the influencing factors of δDθ at smaller scales, such as the microhabitats, may mainly contribute to its variability (Zhang et al., 2011). Therefore, a shorter sampling interval or a larger plot could be justified. It will be helpful to employ nested samplings (Fromin et al., 2013) from a small to large scale in a future study to better understand the spatial patterns of δDθ. Although the kriging maps of δDθ were affected by the flat effects of interpolation, spatial distribution along the slope was observed (Fig. 3c and d). Soil water δD value at the surface (0–10 cm) tended to decrease from the upslope to downslope on April 15 but increased on August 18. The coarser soil, with lower water storage capacity, was more highly influenced by the most recent rainfall upslope. The δD values of the latest rainfalls before April 15 were more positive than previous rainfalls in March (Fig. 2). Therefore, δDθ was more positive in the upslope direction and decreased along the hillslope. On the contrary, δD values of the latest rainfalls before August 18 were more negative. Therefore, increasing δDθ along the hillslope was observed. 4.3. Correlations between surface (0–10 cm) θ and δDθ There were remarkable correlations between surface (0–10 cm) θ and δDθ in the two sampling campaigns (Fig. 5). Previous studies also reported significant correlations between θ and δDθ in non-karst regions (Retzlaff et al., 2001; Sun et al., 2010). It noteworthy that θ was negatively correlated with δDθ on April 15, but positively correlated on August 18. Rainfall δD values vary with time, but perhaps less so at small scales (Jayasena et al., 2008; Hu et al., 2013; Kato et al., 2013). Lateral flow may occur when the soil is saturated, but based on a previous assumption, the δDθ between points should not be high when lateral flow occurs (Goller et al., 2005). Transpiration affects θ but not δDθ (White et al., 1985; Hsieh et al., 1998; Nie et al., 2011), which leaves evaporation as the influencing factor on spatial variation in δDθ. Therefore, a negative correlation between θ and δDθ is expected. On the other hand, soil water storage capacity varies across space, especially in karst regions, which indicates that different points in space retain different components of antecedent rainfall (i.e. places with less storage capacity will be influenced more by the most recent rainfall) (Jost et al., 2005). This apparent random variation between θ and δDθ may or may not overwhelm the local evaporation signal. The δD values of latest rainfalls in April (Fig. 2) were more positive than δDθ on April 15 (Table 1). More negative δD values were found in large rainfalls in March (Fig. 2), which indicated that the present δDθ was the result of the mixing of the antecedent rainfalls containing different δD values.

The point with more storage capacity could reflect more information of earlier rainfalls. Therefore, a negative correlation between θ and δDθ was found on April 15. On the other hand, there was a long period of drought in July (Fig. 2, 13.7 mm rainfall in total with weighted average δD value of −48.5‰) and the soil water was deeply affected by evaporation, which might result in positive δD values. After being recharged by the several rainfalls with negative δD values, δDθ was driven close to the δD values of the rainfalls. The soil water containing less storage capacity was replaced more thoroughly by recent rainfalls with negative δD values. Therefore, positive correlations between θ and δDθ were observed on August 18. The different correlations between θ and δDθ could also been reflected by the kriging maps (Fig. 3). 4.4. Influencing factors of surface (0–10 cm) θ and δDθ Very significant correlations between θ and influencing factors such as soil properties, topography, and land cover were found in this study. Some researchers state that the variation of θ was mainly controlled by topography and soil properties at a hillslope scale (Famiglietti et al., 1998). In reality, multiple factors such as land use, topographic factors, soil properties and vegetation all had important effects on θ (Qiu et al., 2001; Svetlitchnyi et al., 2003; Penna et al., 2009). Soil physical properties including ρb, porosity and CC were significantly correlated with θ (Table 3). These properties had important impacts on water storage capacity, and thus affected θ (Zhao et al., 2013). Very significant positive correlations were observed between θ and SOC in this study (Table 3). Similar results were also reported in other studies (Zhang et al., 2011). SOC is an important component and had a close relationship with both θ and water retention capacity (Li et al., 2004). As SOC increased, the volume of water held at field capacity increased at a greater rate (Alan and David, 2005). Land cover factors (VC, EBR and RFC) also had significant correlations with θ (Table 3). High VC can protect θ from evaporation, but it also indicated high transpiration that reduced θ (Peng et al., 2013). The effects of VC on θ are quite uncertain and depend on the dominant process. The positive correlation between θ and VC indicates that the vegetation had greater impacts on protecting θ from evaporation. The rocky soil could strongly affect θ, soil water retention and infiltration and solute transport rates (Chen et al., 2011; Zhang et al., 2011). Previous studies indicate that areas immediately adjacent to outcrops had higher θ than areas farther away from outcrops (Conn and Snyder-Conn, 1981; Li et al., 2014). However, their results were gained based on the investigation of isolated bedrocks and their adjacent soils. At a hillslope scale, higher EBR is observed in the upslope direction and accompanied by shallow coarse soil that is not beneficial for water retention. On the contrary, in the downslope direction, EBR was relatively low and the soil was deeper with a fine texture. Therefore, negative correlations were found between θ and EBR on

J. Yang et al. / Geoderma 264 (2016) 61–70

both sampling dates. Rock fragment content also had a negative correlation with θ. Cousin et al. (2003) suggested that the high calorific characteristics of the rock fragments would lead to heating of the soil and therefore to a decrease in θ under strong evaporative conditions. Reverse spatial distributions of RFC and θ were shown in this study (Figs. 3a, b and 4). Topographic factors like elevation and slope were negatively correlated to θ. Similar result were found by Peng et al. (2013) in other karst regions. Soils at higher elevations with steep slopes were always shallower, coarser and had poor water storage capacity that resulted in lower θ (Qiu et al., 2001). Significant to very significant correlations were found between δDθ and most of the soil properties and environmental factors. δDθ was mainly affected by rainfall recharge (both present and antecedent rainfalls) and evaporation processes (Ayalon et al., 1998; Song et al., 2009). All factors that influence rainfall recharge and evaporation (e.g., vegetation, topography, and soil properties) would impact on δDθ (Hsieh et al., 198; Kato et al., 2013; Zhao et al., 2013). Interestingly, correlations between δDθ and those influencing factors were reversed on two sampling dates (Table 3), just like the reverse correlations between θ and δDθ (Fig. 5). Moreover, correlations between δDθ and any of the influencing factors disappeared after conducting the partial correlation analysis with θ as a controlled variable (Table 3). This indicates that the influencing factors might affect δDθ through affecting θ. Therefore, the reverse correlations between δDθ and the influencing factors are caused by the reverse correlations between θ and δDθ. 5. Conclusions The results showed that θ decreased with soil depth on both sampling dates. δDθ decreased with soil depth on April 15, but increased on August 18. Soil moisture showed greater variance than δDθ on both sampling dates. High variance of θ resulted from the spatial variation of soil water storage capacity and transpiration. The relatively lower variance of δDθ was the result of antecedent rainfalls with similar δD values. Soil moisture presented a moderate to strong spatial dependence, while δDθ had a weak spatial dependence and varied randomly. Significant but reverse correlations were observed between θ and δDθ for both sampling campaigns, which resulted from the joint impact of antecedent rainfall properties (amount and rainfall δD values) and local soil properties (mainly spatial variation of water storage capacity). Influencing factors such as soil properties, topography and land cover were found significantly correlated with θ and δDθ. However, correlations between δDθ and these same influencing factors disappeared after partial analysis that used θ as a controlled variable. These factors might affect δDθ by affecting θ. Despite the significant correlations between θ and δDθ, the results revealed both differences on their spatial variability and influencing factors. This suggests that we investigate θ and δDθ separately in further hydrological process studies. Acknowledgements This research was supported by the National Key Basic Research Program of China (2015CB452703) and the National Natural Science Foundation of China (41171187 and 51379205). We thank the editor and the three anonymous reviewers for their invaluable comments and suggestions on this manuscript. We also thank Magdeline Laba for checking the English. References Alan, O., David, A., 2005. Effect of organic carbon on available water in soil. Soil Sci. 170 (2), 90–101. Ayalon, A., Bar-Matthews, M., Sass, E., 1998. Rainfall-recharge relationships within a karstic terrain in the Eastern Mediterranean semi-arid region, Israel: δ18O and δD characteristics. J. Hydrol. 207, 18–31. Brocca, L., Morbidelli, R., Melone, F., Moramarco, T., 2007. Soil moisture spatial variability in experimental areas of central Italy. J. Hydrol. 333 (2–4), 356–373.

69

Bruckner, A., Kandeler, E., Kampichler, C., 1999. Plot-scale spatial pattern of soil water content, pH, substrate induced respiration and N mineralisation in a temperate coniferous forest. Geoderma 93, 207–223. Campbell, J.B., 1978. Spatial variation of sand content and pH within single contiguous delineations of two soil mapping units. Soil Sci. Soc. Am. J. 42 (3), 460–464. Chen, H.S., Liu, J.W., Wang, K.L., Zhang, W., 2011. Spatial distribution of rock fragments on steep hillslopes in karst region of northwest Guangxi, China. Catena 84, 21–28. Chen, H.S., Zhang, W., Wang, K.L., Fu, W., 2010. Soil moisture dynamics under different land uses on karst hillslope in northwest Guangxi, China. Environ. Earth Sci. 61, 1105–1111. Conn, J.S., Snyder-Conn, E.K., 1981. The relationship of the rock outcrop microhabitat to germination, water relations, and phenology of Erythrina flabelliformis (Fabaceae) in Southern Arizona. Southwest. Nat. 25 (4), 243–251. Cousin, I., Nicoullaud, B., Coutadeur, C., 2003. Influence of rock fragments on the water retention and water percolation in a calcareous soil. Catena 53, 97–114. Cui, Z.J., Li, D.W., Feng, J.L., Liu, G.N., Li, H.J., 2002. The covered karst, weathering crust and karst (double-level) planation surface. Sci. China 45 (4), 366–379. Daly, E., Porporato, A., 2006. Impact of hydroclimatic fluctuations on the soil water balance. Water Resour. Res. 42 (6). http://dx.doi.org/10.1029/2005WR004606. Dawson, T.E., Mambelli, S., Plamboeck, A.H., Templer, P.H., Tu, K.P., 2002. Stable isotopes in plant ecology. Annu. Rev. Ecol. Syst. 507-559. Ehleringer, J.R., Roden, J., Dawson, T.E., 2000. Assessing ecosystem-level water relations through stable isotope ratio analyses. In: Sala, O.E., Jackson, R.B., Mooney, H.A., Howarth, R.W. (Eds.), Methodsin Ecosystem Science. Springer, New York, pp. 181–214. Famiglietti, J.S., Rudnicki, J.W., Rodell, M., 1998. Variability in surface moisture content along a hillslope transect: Rattlesnake Hill, Texas. J. Hydrol. 210, 259–281. Fitzjohn, C., Ternan, J.L., Williams, A.G., 1998. Soil moisture variability in a semi-arid gully catchment: implications for runoff and erosion control. Catena 32 (1), 55–70. Fromin, N., Saby, N.P.A., Lensi, R., Brunet, D., Porte, B., 2013. Spatial variability of soil microbial functioning in a tropical rainforest of French Guiana using nested sampling. Geoderma 197-198, 98–107. Gerrits, A.M., Pfister, L., Saveije, H.H.G., 2010. Spatial and temporal vaiability of canopy and forest floor interception in a beech forest. Hydrol. Process. 24 (21), 3011–3025. Goebel, T.S., Lascano, R.J., 2012. System for high throughput water extraction from soil material for stable isotope analysis of water. J. Anal. Sci. Methods Instrum. 02, 203–207. Goller, R., Wilcke, W., Leng, M.J., Tobschall, H.J., Wagner, K., Valarezo, C., Zech, W., 2005. Tracing water paths through small catchments under a tropical montane rain forest in South Ecuador by an oxygen isotope approach. J. Hydrol. 308 (1–4), 67–80. Grayson, R.B., Western, A.W., Chiew, F.H.S., Blöschl, G., 1997. Preferred states in spatial soil moisture patterns: local and nonlocal controls. Water Resour. Res. 33, 2897–2908. Hargreaves, G.H., Samani, Z.A., 1985. Reference crop evapotranspiration from temperature. Appl. Eng. Agric. 96-99. Hsieh, J.C., Chadwick, O.A., Kelly, E.F., Savin, S.M., 1998. Oxygen isotopic composition of soil water: quantifying evaporation and transpiration. Geoderma 82 (1–3), 269–293. Hu, K., Chen, H.S., Nie, Y.P., Yang, J., Wang, K.L., 2013. Characteristics of seasonal variation of deuterium and oxygen-18 isotope composition of precipitation in karst peakcluster depression area, northwest Guangxi of China. Trans. Chin. Soc. Agric. Eng. 5, 53–62 (in Chinese with English abstract). Hu, W., Shao, M.A., Han, F.P., Reichardt, K., 2011. Spatial-temporal variability behavior of land surface soil water content in shrub- and grass-land. Geoderma 162, 260–272. Jayasena, H.A.H., Chandrajith, R., Dissanayake, C.B., 2008. Spatial variation of istotope composition in precipitation in a tropical environment: a case study from the Deduru Oya River Basin, Sri Lanka. Hydrol. Process. 22 (23), 4565. http://dx.doi.org/10.1002/ hyp.7060. Jia, X., Shao, M., Wei, X., Wang, Y., 2013. Hillslope scale temporal stability of soil water storage in diverse soil layers. J. Hydrol. 498, 254–264. Jost, G., Heuvelink, G.B.M., Papritz, A., 2005. Analysing the space-time distribution of soil water storage of a forest ecosystem using spatio-temporal kriging. Geoderma 128, 258–273. Journel, A.G., Huijbregts, C.J., 1978. Mining Geostatistics. Academic Press, London, p. 600. Kato, H., Onda, Y., Gomi, T., Yamanaka, T., Kawaguchi, S., 2013. Effect of canopy interception on spatial variability and isotopic composition of throughfall in Japananese cypress plantations. J. Hydrol. 504, 1–11. Lange, J., Arbel, Y., Grodek, T., Greenbaum, N., 2010. Water percolation process studies in a Mediterranean karst area. Hydrol. Process. 24, 1866–1879. Lee, K.S., Kim, J.M., Lee, D.R., Kim, Y., Lee, D., 2007. Analysis of water movement through an unsaturated soil zone in Jeju Island, Korea using stable oxygen and hydrogen isotopes. J. Hydrol. 345, 199–211. Legates, D.R., Mahmood, R., Levia, D.F., DeLiberty, T.L., Quiring, S.M., Houser, C., Nelson, F.E., 2011. Soil moisture: a central and unifying theme in physical geography. Prog. Phys. Geogr. 35 (1), 65–86. Li, S., Ren, H.D., Xue, L., Chang, J., Yao, X.H., 2014. Influence of bare rocks on surrounding soil moisture in the karst rocky desertification regions under drought conditions. Catena 116, 157–162. Li, Y.B., Wang, S.J., Li, R.L., 2004. Some soil fetures of karst ecosystem. Ecol. Environ. 13 (3), 434–438 (in Chinese with English abstract). Lis, G., Wassenaar, L.I., Hendry, M.J., 2008. High-precision laser spectroscopy D/H and 18 O/16O measurements of microliter natural water samples. Anal. Chem. 80 (1), 287–293. Meißner, M., Köhler, M., Schwendenmann, L., Hölscher, D., Dyckmans, J., 2014. Soil water uptake by trees using water stable isotopes (δ2H and δ18O) — a method test regarding soil moisture, texture and carbonate. Plant Soil 376, 327–335. Nie, Y.P., Chen, H.S., Wang, K.L., 2010. Methods for determining plant water source in thin soil region: a review. Chin. J. Appl. Ecol. 21, 2427–2433 (in Chinese with English abstract).

70

J. Yang et al. / Geoderma 264 (2016) 61–70

Nie, Y.P., Chen, H.S., Wang, K.L., Ding, Y.L., 2014. Rooting characteristics of two widely distributed woody plant species growing in different karst habitats of Southwest China. Plant Ecol. 215, 1099–1109. Nie, Y.P., Chen, H.S., Wang, K.L., Tan, W., Deng, P.Y., Yang, J., 2011. Seasonal water use patterns of woody species growing on the continuous dolostone outcrops and nearby thin soils in subtropical China. Plant Soil 341, 399–412. Orlowski, N., Frede, H.-G., Brüggemann, N., Breuer, L., 2013. Validation and application of a cryogenic vacuum extraction system for soil and plant water extraction for isotope analysis. J. Sens. Sens. Syst. 2, 179–193. Peng, W.X., Song, T.Q., Zeng, F.P., Wang, K.L., Du, H., Lu, S.Y., 2013. Spatial distribution of surface soil water content under different vegetation types in northwest Guangxi, China. Environ. Earth Sci. 69 (8), 2699–2708. Penna, D., Borga, M., Norbiato, D., Dalla Fontana, G., 2009. Hillslope scale soil moisture variability in a steep alpine terrain. J. Hydrol. 364, 311–327. Penna, D., Brocca, L., Borga, M., Dalla Fontana, G., 2013. Soil moisture temporal stability at different depths on two alpine hillslopes during wet and dry periods. J. Hydrol. 477, 55–71. Penna, D., Stenni, B., Šanda, M., Wrede, S., Bogaard, T.A., Gobbi, A., Borga, M., et al., 2010. On the reproducibility and repeatability of laser absorption spectroscopy measurements for δ 2 H and δ 18 O isotopic analysis. Hydrol. Earth Syst. Sci. 14, 1551–1566. Penna, D., Stenni, B., Šanda, M., Wrede, S., Bogaard, T.A., Michelini, M., Fischer, B.M.C., et al., 2012. Technical note: evaluation of between-sample memory effects in the analysis of δ2H and δ18O of water samples measured by laser spectroscopes. Hydrol. Earth Syst. Sci. 16, 3925–3933. Qiu, Y., Fu, B., Wang, J., Chen, L., 2001. Soil moisture variation in relation to topography and land use in a hillslope catchment of the Loess Plateau, China. J. Hydrol. 240, 243–263. Rawls, W.J., Pachepsky, Y.A., Ritchie, J.C., Sobechi, T.M., Bloodworth, H., 2003. Effect of soil organic carbon on soil water retention. Geoderma 116 (1–2), 61–76. Retzlaff, W.A., Blaisdell, G.K., Topa, M.A., 2001. Seasonal changes in water source of four families of loblolly pine (Pinus taeda L.). Trees 15, 154–162. Röver, M., Kaiser, E.A., 1999. Spatial heterogeneity within the plough layer: low and moderate variability of soil properties. Soil Biol. Biochem. 31 (2), 175–187. Saha, A.K., da Stenberg, L.S.L., Miralles-Wilhelm, F., 2009. Linking water sources with foliar nutrient status in upland plant communities in the Everglades National Park, USA. Ecohydrology 2, 42–54. Sayer, E.J., 2006. Using experimental manipulation to assess the role of leaf litter in the functioning of forest ecosystems. Biol. Rev. 81 (1), 1–31.

Snepvangers, J.J.J.C., Heuvelink, G.B.M., Huisman, J.A., 2003. Soil water content interpolation using spatio-temporal kriging with external drift. Geoderma 112, 253–271. Song, T.Q., Peng, W.X., Zeng, F.P., Wang, K.L., Cao, H.L., Li, X.K., Qin, W.G., Tan, W.N., Liu, L., 2010. Community composition and biodiversity characteristics of forests in karst cluster-peak-depression region. Biodivers. Sci. 18, 355–364 (in Chinese with English abstract). Song, X.F., Wang, S.Q., Xiao, G.Q., Wang, Z.M., Liu, X., Wang, P., 2009. A study of soil water movement combining soil water potential with stable isotopes at two sites of shallow groundwater areas in the North China Plain. Hydrol. Process. 23, 1376–1388. Stothoff, S.A., Or, D., Groeneveld, D.P., Jones, S.B., 1999. The effect of vegetation on infiltration in shallow soils underlain by fissured bedrock. J. Hydrol. 218, 169–190. Sun, S.J., Meng, P., Zhang, J.S., Huang, H., Wan, X.C., 2010. Deuterium isotope variation and water use in an agroforestry system in the rocky mountainous area of North China. Acta Ecol. Sin. 30 (14), 3717–3726 (in Chinese with English abstract). Svetlitchnyi, A.A., Plotnitskiy, S.V., Stepovaya, O.Y., 2003. Spatial distribution of soil moisture content within catchments and its modeling on the basis of topographic data. J. Hydrol. 277 (1), 50–60. Takahashi, K., 1998. Oxygen isotope ratios between soil water and stem water of trees in pot experiments. Ecol. Res. 13, 1–5. Tokumoto, I., Heilman, J.L., Schwinning, S., McInnes, K.J., Litvak, M.E., Morgan, C.L., Kamps, R.H., 2014. Small-scale variability in water storage and plant available water in shallow, rocky soils. Plant Soil 385, 193–204. Wassenaar, L.I., Coplen, T.B., Aggarwal, P.K., 2014. Approaches for achieving long-term accuracy and precision of δ18O and δ2H for waters analyzed using laser absorption spectrometers. Environ. Sci. Technol. 48, 1123–1131. Western, A.W., Blöschl, G., Grayson, R.B., 1998. Geostatistical characterizations of soil moisture patterns in Tarrawarra catchment. J. Hydrol. 205, 20–37. White, J.W.C., Cook, E.R., Lawrence, J.R., Wallace, B.S., 1985. The D/H ratios of sap in trees: implications for water sources and tree ring D/H ratios. Geochim. Cosmochim. Acta 49 (1), 237–246. Zhang, J.G., Chen, H.S., Su, Y.R., Shi, Y., Zhang, W., Kong, X.L., 2011. Spatial variability of surface soil moisture in a depression area of karst region. Clean: Soil, Air, Water 39 (7), 619–625. Zhang, W., Wang, K.L., Chen, H.S., He, X.Y., Zhang, J.G., 2012. Ancillary information improves kriging on soil organic carbon data for a typical karst peak cluster depression landscape. J. Sci. Food Agric. 92 (5), 1094–1102. Zhao, P., Tang, X.Y., Zhao, P., Wang, C., Tang, J.L., 2013. Identifying the water source for subsurface flow with deuterium and oxygen-18 isotopes of soil water collected from tension lysimeters and cores. J. Hydrol. 503, 1–10.