Analysis of Soil δ¹³C and δ¹⁵N Along Precipitation Gradient: Critical Insights into Tree–Grass Interactions and Soil C Sequestration in Savannas

Abstract

In situ observations of belowground tree–grass interactions are sparse in savanna ecosystems. In this study, we analyzed stable carbon and nitrogen isotopes (δ¹³C and δ¹⁵N) in soils and plants from four study sites in an African savanna ecosystem along the Kalahari moisture gradient. The homogeneous soil texture, primarily sandy soils, is well-drained and nutrient-poor, influencing vegetation and water retention uniformly across the region. At each site, soil samples were collected from a 120 cm deep soil profile. We used a 2-endmember mass balance approach to calculate the relative contributions of C₃ and C₄ plants to SOC in the 120 m soil profile. The wettest site was dominated by trees, whereas the driest site was dominated by shrubs. The intermediates had the highest amount of grass biomass. Our results revealed that tree- and shrub-derived SOC was highest in the wettest and driest sites, respectively. The contribution of C₃ plants was 63.8% and 55.8%, in the wettest and driest sites, respectively, when integrating the 120 cm depth. Grass-derived SOC was highest (69.4%) in the middle of the precipitation gradient when integrating the 120 cm depth. The δ¹⁵N values were highest in the middle of the precipitation gradient (10.7‰) and lowest in the wettest site (5.2‰). Our findings indicate that belowground tree–grass interactions and nitrogen cycling in savanna ecosystems are more complex than previously thought and do not conform to the traditional concept of the two-layer roots hypothesis. This lack of conformity could be attributed to several factors, including overlap in rooting depth and ecological drivers, such as wildfires and herbivory, which could stimulate production of belowground biomass. We used space-for-time substitution to leverage the region’s steep north–south precipitation gradient and homogeneous soil texture. Our results indicated that trees and shrubs would become an important SOC source in the two extreme sites of the transect, while grass would become an important SOC source in the middle of the precipitation gradient.

1. Introduction

Savannas are characterized by coexistence of trees and grasses and cover about 23–33 million km², or about one-fifth, of the Earth’s land surface [1,2,3]. They are found in the latitude band ~15–30° in the south and north of the equator, where the mean annual precipitation (MAP) ranges between 100 and 1500 mm yr⁻¹ [4,5]. Equilibrium theory proposes that competition for water between woody and grass plants is a key determinant of the composition of the two life forms (trees and grasses). The theory suggests that competition is excluded based on the assumption that trees and grasses have ecological niches [6,7,8,9,10]. Because soil moisture varies with depth, grasses are efficient at exploiting moisture in the topsoil, whereas trees have exclusive access to water in the subsoil.

Savannas are an important component of drylands and global carbon stock, containing 15–20% of global soil organic carbon (SOC) [11,12,13,14,15]. SOC stocks in savannas are facing potential challenges from anthropogenic climate change and associated extreme weather events such as drought and windstorms [16,17,18,19]. Global climate models predict that savannas, particularly in southern Africa, would experience a 15% decrease in mean annual precipitation (MAP) and a 1.5 to 4 °C increase in mean annual temperature (MAT) by the end of the 21st century [18,19,20]. Regionally, the models predict an increase in the frequency and magnitude of associated extreme weather events such as wildfires, lightning, and droughts in Southern Africa. At the local scale, Batisani and Yarnal [21] observed a trend where rainfall decreases over time for the past decade in Botswana Kalahari (i.e., the area inside Botswana underlain by Kalahari Sands).

Dintwe and Okin [22] estimated that global savannas will lose ~28.4 Tg C y⁻¹ and 64.1 Tg C yr⁻¹ under Representative Concentration Pathway 2.6 (RCP2.6) and RCP8.5, respectively. Furthermore, they cautioned that the rapid loss of C from savanna soils through soil respiration could accelerate global warming and strengthen positive feedback mechanisms between climate change and processes controlling SOC. Vegetation models predict that plant productivity and SOC stocks in savannas, particularly in Africa, would decrease due to a decrease in precipitation [23]. Few studies in the Kalahari and across the world indicate that a decrease in precipitation causes a decline in net primary productivity (NPP), which ultimately leads to a decline in soil C sequestration [22,24]. Zeng and Neelin [25] used the QTCM model and concluded that a decrease in precipitation could increase vegetation productivity in African savannas, and ultimately in SOC sequestration. In Australia savannas, [26] estimated total carbon stock to be 204 ton C ha⁻¹, with about 84% belowground, 16% aboveground, and SOC accounting for about 74% of the total carbon content. Grace, Post [27] used the SOCRATES terrestrial carbon model to estimate changes in SOC for the Australian continent between the years 1990 and 2100 in response to climate change. They found that under a high greenhouse gas emissions scenario, the Australian continent becomes a source of CO₂ with a net reduction of 6.4% (518 Mt) in topsoil carbon, when compared to no climate change. Similarly, ref. [28] projected a decline in SOC stocks over south-eastern Australia until ~2070 due to changes in temperatures and rainfall in the region. In South American savannas, the Cerrado stores a significant amount of SOC, estimated at over 13 gigatons in the top 1 m of the soil, with a vertical distribution pattern of 26% in the upper 0–30 cm layer, 38% in the 30–100 cm layer, and 36% in the 100–200 cm layer [29,30]. However, SOC in South America, particularly in Brazil, is expected to decrease due to the combined effects of climate change and agricultural expansion [31,32]. Wei, Shao [33] showed that at the global scale, climate and soil factors affected the decrease in SOC stocks. Precipitation, depending on its intensity and frequency, could have varying outcomes on vegetation productivity, and ultimately soil carbon sequestration and nutrient cycling. For instance, small increases in precipitation intensity may increase plant productivity by decreasing interception and evaporation [34]. Alternatively, a large increase in precipitation intensity and associated decrease in frequency may decrease plant productivity due to deep soil infiltration and lower soil moisture in surface layers [35,36]. Moreover, vegetation responds to short-term variation in precipitation through adjustments in plant physiology, whereas long-term variation is mediated by plant establishment, composition, and disturbance regimes [37,38]. However, it is still not clear how climate change, particularly precipitation, will affect soil C sequestration and nutrient cycling in savanna ecosystems. With these contradictory analyses, it is clear that more work needs to be done in order to assess and understand the potential impacts of climate change on SOC balance in the savanna [39,40,41].

The objective of this study is to determine the impact of climate change on soil C sequestration in savanna ecosystems. We used stable C and N isotopes (δ¹³C and δ¹⁵N) to assess the long-term impacts of climate on C and N pools in the soil profile in a savanna ecosystem along a steep precipitation gradient (180–540 mm yr⁻¹). Stable isotopes of nitrogen (δ¹⁵N) are used to evaluate whether the abundant Mimosaceae in the Kalahari and other plants contribute nitrogen (N) to the system. Nitrogen is an essential plant nutrient and one of the limiting nutrients in savanna ecosystems [42,43,44]. Because plants are the major terrestrial producers and sources of soil organic C (SOC), soil carbon stocks record and reflect the long-term interactions between trees and grasses [45,46,47]. δ¹³C values of SOC relate to the ¹³C of plants residues mainly through root turnover [48]. Thus, the soil profile δ¹³C may be used as a proxy to identify sources of SOC and thus infer the belowground tree–grass relationships. This analysis of δ¹³C along a precipitation gradient elucidates both belowground interactions between plant functional types as well as the effect of this interaction under different precipitation regimes. Based on the previous work in the Kalahari [48,49,50], we hypothesized that the drier sites of the Kalahari will have relatively low soil C input from trees, shrubs and grasses, and inversely wetter sites will have relatively more soil C input from trees, shrubs and grasses. However, it is important to point out that this study differs from earlier studies in the sense that more intensive work (e.g., more sites and deeper soil profiles) was done in soil sampling and analysis.

2. Materials and Methods

2.1. Study Site Description

The study was conducted in the Kalahari Transect (KT) in southern Africa (Figure 1). The Kalahari basin covers an area of about 2.5 million km² between 12–29° S and 14–28° E, along a steep north–south precipitation gradient [16,51,52]. While the mean annual precipitation (MAP) in the north could be as high as 1200 mm, the south receives as little as 150 mm. The Botswana Kalahari stretches 1000 km in the north–south direction across the western part of the country along a rainfall gradient ranging from 650 mm yr⁻¹ in the north to 180 mm yr⁻¹ in the south [49]. Potential evapotranspiration rates vary from 1000 mm yr⁻¹ in the north to >2000 mm yr⁻¹ in the south [53].

Soils in the Kalahari are deep homogeneous aeolian sand deposits reaching 100–200 m depth in the center of the basin [52,54,55]. The soils in the Kalahari are slightly acidic, about ~95% sand, and poor in organic carbon and total nitrogen content [14,22,56]. Throughout much of the Kalahari, groundwater is deep (>20 m) [57], meaning that plants rely almost exclusively on precipitation (

Table 1. Geographic coordinates, climate parameters, soil, and vegetation characteristics of the study sites. The δ¹³C values for trees and grasses are obtained from [48,58], whereas tree δ¹⁵N values are obtained from [48,59].

Site Feature	Bokspits	Tshane	Kuke	Shakawe
Geographic	26°53′39″ S	24°01′01″ S	20°58′36″ S	18°21′51″ S
Coordinates	20°41′54″ E	21°52′08″ E	22°28′48″ E	21°50′31″ E
Mean Annual Precipitation (mm)	180	360	440	540
Mean Min Temp (°C)	2.1	3.9	4.4	5.8
Mean Max Temp (°C)	34.3	33.5	33.4	33.7
Soil pH	6.7	5.6	5.9	5.4
Organic Carbon (g m−2)	1397.0	2506.0	2136.0	1982.0
Woody biomass (g m−2)	0.15	0.31	0.40	0.52
Nitrogen (g m−2)	78.6	151.0	135.0	107.0
C3 Foliar δ13CVPDB (‰)	−23.2	−25.9	−25.9	−26.7
C4 Foliar δ13CVPDB (‰)	−14.0	−14.2	−13.8	−13.4
C3 foliar δ15Natm (‰)	7.40	6.82	7.27	3.50

).

The vegetation structure (e.g., diversity and canopy cover) in the Kalahari is strongly tied to the north–south precipitation gradient [55,60]. Shrubs and trees in the dry portions of the Kalahari contribute significantly to the patterning of soil resources, whereas in wet portions, only trees contribute to the patterning of soil resources [61]. The dominant plant species in the wettest part of Kalahari (north) are nutrient-poor broad-leaf Combretaceae and Caesalpinaceae, whilst in the driest (south) part are C₄ grasses, with sparsely distributed nutrient-rich fine-leaf Mimosaceae [49,54,55,60,62,63,64]. Among the Poaceae, the Aristidae (Aristida and Stipagrostis spp.) dominate the arid south, the Panicoidae in the semi-arid center, while the Andropogonae dominate in the mesic north [55].

The KT was identified by the IGBP (International Geosphere-Biosphere Programme) as one of the ‘mega-transects’ for global change studies [55]. With a remarkably steep precipitation gradient, lack of relief, and homogeneous soil texture, the KT provides an ideal setting to study the impact of climate on carbon dynamics with minimal confounding variables.

2.2. Sample Design

Four study sites were selected along the Kalahari transect (KT) along the north–south precipitation gradient (Figure 1). Three 20 m × 20 m plots were established at each site, with each plot resulting in 400, 1 m × 1 m quadrats. Vegetation was characterized on each plot before removal, see [62]. The cleared plots were gridded into 1 × 1 m quadrats. Eighty (80) of the 1 m × 1 m quadrats were randomly selected and sampled in the center with a 4-inch (~10 cm) diameter sand auger (AMS Inc, Pittsburgh, PA, USA) at four depths. The auger was inserted in the center of each quadrat until its tip was at 10, 30, 70, and 120 cm depth [65]. Upon reaching each of these depths, the auger was removed, and a mixed sample from the auger was extracted. Thus, the average depth for the samples was approximately 5, 20, 50, and 95 cm. About 50 g of soil was taken at each depth and stored in a labeled zip-lock bag. The bags were left open for a week to air dry before being shipped to the USA for laboratory analysis.

2.3. Laboratory Analysis

Each labeled bag of soil was sieved, thoroughly mixed, and sub-sampled using a riffle sampler. Soil grain size distribution was measured in a subset of soil samples following a method outlined by [66,67]. For this, approximately 0.5 cm³ of sample was boiled in DI water and pretreated with about 50 mL of 30% H₂O₂ to remove organic matter, 10 mL of 1 N HCl was used to remove carbonates, and 10 mL of 1 N NaOH was used to remove biogenic silica. Grain size was measured using the Malvern Mastersizer 2000 G (Malvern Panalytical, Malvern, Worcestershire, United Kingdom) large-volume sample dispersion unit. To verify the equipment’s accuracy, repeatability and stability, a tuff standard (TS2) with a known distribution between 1.0 μm and 16 μm (mean = 4.53 μm ± 0.07 μm, n = 2998) was run regularly. The TS2 was run every few samples and once at the end for a final assessment. The results were reported as volume percent and divided into sand (74.00–2000 μm), silt (3.90–73.99 μm) and clay (0.02–3.89 μm) [66].

Subsamples of about 4–5 cm³ were ground into a powder using a ball mill for three minutes at 3450 rpm (Cianflone Scientific Instruments Corporation, Pittsburgh, PA, USA). Thirty-three samples for each depth at 10 and 30 cm, and fifteen samples each depth at 70 and 120 cm were randomly selected for each site for δ¹³C and δ¹⁵N analysis. A small portion, approximately 190 mg, of the ground, homogenized soil was placed into a 9 × 10 mm tin capsule (Costech Analytical, Valencia, CA, USA). Before being sealed, the soils in the open silver capsules were fumigated in an atmosphere created by a beaker of concentrated HCl placed in a bell jar for 24 h in order to remove carbonates [68]. After being sealed, the sample was combusted on an Elemental Combustion System (Costech Analytical, Valencia, CA, USA) interfaced by continuous gas flow ConFlo IV (Thermo Scientific, Bremen, Germany) to a stable isotope ratio mass spectrometer (Delta V Plus, Thermo Scientific, Bremen, Germany). SOC and total nitrogen (TN) results are from [14]. Tree δ¹⁵N values are obtained from [48,59], whereas δ¹³C values for trees and grasses are obtained from [48,58].

The δ¹³C in the soil samples was normalized to the international standards USGS-40 (L-glutamic acid) and Buffalo River Sediments (SRM 2704), while ammonium sulfate (IAEA-N-1) was used as the standard for δ¹⁵N [69,70,71,72,73,74,75,76]. The isotope ratios were calculated using Equation (1), where R was the element measured, x was the heavier isotope and y was the lighter isotope of the element measured [65,70,72,77,78,79,80].

$${\delta }^{x}R\mathrm{‰}=\left({\displaystyle \frac{{(}^{x}R{/}^{y}R{)}_{sample}}{{(}^{x}R{/}^{y}R{)}_{standard}}}-1\right)\ast {10}^3={\displaystyle \frac{\left(R_{sample}-R_{standard}\right)}{R_{standard}}}\ast {10}^3$$

(1)

By convention, the δ¹³C and δ¹⁵N were corrected and reported as parts per thousand (per mil) relative to that for Vienna Pee Dee Belemnite (VPDB) and atmospheric N2, respectively [71,76,81]. Reproducibility of these measurements was approximately ±0.2‰.

2.4. Statistical Analysis

Descriptive statistics such as mean, median, and standard deviation were used to analyze the measured variables and to indicate the degree of overall variation. Analysis of variance (ANOVA) was used to test for differences between samples, depth, and sites. Fisher’s least significant difference (LSD) method was used to distinguish differences following the ANOVA test. We further used linear and multiple regression techniques to evaluate the relationship between isotope values and climate variables.

We calculated the difference between C₃ foliar δ¹⁵N and the potential sources of plant N, which are soil and the atmosphere. The difference (Δ¹⁵N) was used to identify the dominant source of plant N as well as to estimate the isotopic composition of plant available N [82,83,84,85]. By definition, the δ¹⁵N of the atmospheric N₂ is 0‰, whereas that of the soil is usually greater than 0‰, especially in drier environments [86]. Therefore, δ¹⁵N of N-fixing plants should be very close to 0‰ (±2‰), with slight modification by diffusion fractionation and the isotope effect associated with N fixation [87,88]. At present, there is no easy and straightforward technique for isolating and analyzing δ¹⁵N in the soil pools that are available for plant uptake [89].

To estimate the proportion of SOC from trees and grasses, we assumed that SOC was made up of the sum of the fractions (F_i) contributed from the sources (δ_i), where the sources were C₃ and C₄ plant species (Equation (2)).

$${\delta }^{13}{C}_{soil}={\sum }_{i=1}^n{F}_i{\delta }_i$$

(2)

Assuming the soil is a mixture of organic matter from C₃ and C₄ plants, we used a simple 2-endmember mass balance equation to calculate the relative contributions of C₃ and C₄ plants to SOC in the 120 m soil profile (Equation (3)):

$$C_x\left(\%\right)={\displaystyle \frac{{\delta }^{13}{C}_{soc}-{\delta }^{13}{C}_y}{{\delta }^{13}{C}_x-{\delta }^{13}{C}_y}} \times 100$$

(3)

where the δ¹³C of SOC (δ¹³Csoc) comprises end-members with unique δ13C values (δ¹³Cx and δ¹³Cy) [90,91]. Values of δ¹³C-C₃ and δ¹³C-C₄ derived from the literature for measurements made in areas of the Kalahari with MAP comparable to our locations were used for these calculations [48,58] (Table 1). Studies in the Kalahari show that dominant plant species belong to C₃ and C₄, and that CAM plants are almost absent [62,92], suggesting that CAM plants are not a significant SOC source.

Because the amount of SOC declined with depth (e.g., Dintwe, Okin [14], an exponential decay function, $C=C_o{e}^{-kz}$, was used to estimate total C contribution from trees and grasses in the soil profile, where C (concentration at depth) and Co (surface concentration) were in units of g m⁻³. The parameters of the equation, Co and k (m⁻¹) were determined by least squares fitting of measured concentrations with depth, z. The total C contribution was calculated by taking the integral of the equation from 0 to 1 m:

$$\int C_o{e}^{-kz}dz=-C_o{e}^{-kz}|k\left|{\displaystyle \genfrac{}{}{0pt}{}{1\mathrm{m}}{0\mathrm{m}}}\right.$$

(4)

We calculated Z₅₀ and Z₉₅ of the soil profile, which are the depths above which 50% and 95% of the total C contribution occurs, respectively. Furthermore, we calculate the e-folding depth $\left({\displaystyle \raisebox{1ex}{$1$}\!\left/ \!\raisebox{-1ex}{$k$}\right.}\right)$. The e-folding depth is the depth at which C contribution decreases to 1/e of the surface value.

3. Results

3.1. Soil Texture

The soils in the Kalahari contained more than 90% sand at all sites and depths except in Bokspits at 120 cm, where sand content was 88% (

Table 2. Soil texture in the Kalahari Transect. Kalahari soils contain more than 90% sand.

Site	Depth	Clay	Silt	Sand
(MAP)	(cm)	(% Composition)
Bokspits (180 mm)	10	0.55	3.10	96.34
	30	0.67	3.56	95.76
	70	0.81	4.36	94.82
	120	1.30	10.36	88.32
	0–120	0.97	6.62	92.39
Tshane (350 mm)	10	1.54	5.01	93.43
	30	1.13	6.30	92.55
	70	1.52	7.76	90.70
	120	1.58	8.17	90.22
	0–120	1.48	7.46	91.04
Kuke (440 mm)	10	1.15	3.44	95.40
	30	0.87	3.88	95.23
	70	1.05	4.56	94.38
	120	1.00	4.59	94.39
	0–120	1.01	4.37	94.61
Shakawe (540 mm)	10	0.00	1.02	98.98
	30	0.00	0.83	99.16
	70	0.00	0.00	99.99
	120	0.00	1.21	98.79
	0–120	0.00	0.73	99.27

). Shakawe had the highest sand content, containing more than 98% at all depths. Clay content was highest in Tshane, followed by Kuke, Bokspits, and Shakawe, respectively. Silt content increased with soil depth in all the sites except in Shakawe.

The mean soil δ¹³C values ranged between −22.7‰ and −17.3‰ (Figure 2). The highest δ¹³C value at 10 cm was recorded at Tshane (−18.9‰), followed by Bokspits with −19.0%. The lowest δ¹³C value at 10 cm depth (−22.2‰). was recorded at Shakawe. Similarly, Tshane had the highest depth-weighted δ¹³C average for the soil profile, followed by Bokspits, Kuke and Shakawe.

The highest δ¹³C values occurred at 30 cm in all the sites, and the second-highest values were found at 10 cm. Soil organic carbon was highest at 10 cm in all the sites, followed by the 10–30 cm layer. Shakawe (wettest) had the lowest mean δ¹³C values, and Tshane had the highest values. In general, soil δ¹³C values decreased with MAP, consistent with the pattern observed in other studies around the world [43,56,93,94,95,96,97,98]. The mean SOC was highest in Tshane and lowest in Bokspits (driest). There was a relatively strong correlation between mean δ¹³C and clay content in the soil profile along the KT (r² = 0.63, p-value < 0.001).

3.2. SOC Input from Trees and Grasses

The percentage contribution of trees (C₃) to SOC, as estimated from δ¹³Csoc, was highest in Shakawe (wettest), and lowest in Tshane (Table S1). The highest contribution of C₃-derived organic matter to the SOC pool was recorded at 10 cm in all the sites. The C₃ contribution at 70 cm was higher than at 30 cm in all the sites, except in Tshane, where the C₃ contribution at 30 cm and 70 cm were statistically indistinguishable. The depth-weighted mean tree SOC contribution in Shakawe was about 64% (0.66 mg C g⁻¹), while SOC in Tshane was approximately 31% (0.48 mg C g⁻¹) (Figure 2). Overall, the contribution of C₃ to SOC increased with MAP, with Bokspits recording 769.4 g C m⁻², whereas Shakawe recorded 1288.1 g C m⁻² (Figure 3).

The intermediate sites had higher grass (C₄) SOC contribution than the wettest and driest sites (Figure 2). The highest C₄ contribution was at 30 cm in all the sites. Bokspits has the lowest absolute C₄ contribution, while Tshane had the highest absolute contribution, 648 g C m⁻² and 1721.0 g C m⁻², respectively (Figure 3). Grass carbon contribution had a non-linear relationship with MAP. The contribution increased with MAP, reached a maximum in Tshane, and then decreased with an increasing MAP. In contrast, tree carbon contribution increased with increasing MAP. The SOC contribution from C₃ plants was higher than the C₄ input at 10 cm, except in Tshane, where C₄ had a higher contribution. The C₄ contribution was higher than the C₃ contribution at 30 cm in the intermediate sites (Tshane and Kuke), lower than the C₃ contribution in the two extreme sites (Bokspits and Shakawe). The e-folding depth for C₃-derived carbon was lowest in Shakawe and highest in Bokspits, 0.6 m and 2.1 m, respectively (Figure 3).

3.3. Soil δ¹⁵N

The mean soil δ¹⁵N values ranged between 1.6‰ and 10.7‰ (Figure 4). The two intermediate sites, Tshane and Kuke, had the highest δ¹⁵N values of 10.7‰ and 8.6‰, respectively, at 10 cm depth. The two extreme sites (driest and wettest), Bokspits and Shakawe, had 7.4‰ and 5.2‰, respectively. Similarly, the intermediate sites had the highest depth-weighted mean δ¹⁵N values, and the two extreme sites had the lowest depth-weighted mean δ¹⁵N values. Although the mean TN in Bokspits was the same as in Shakawe, the mean δ¹⁵N in Bokspits was twice as high as in Shakawe (Figure 4). Shakawe had the overall lowest δ¹⁵N values in the soil profile compared to other sites.

The difference between plant and soil δ¹⁵N was highest in Tshane (the second driest site) and lowest in Kuke. The difference between plant and atmospheric δ¹⁵N was highest in the driest site and lowest in the wettest site. The difference between foliar and atmospheric δ¹⁵N was higher than the difference between foliar and soil δ¹⁵N in all the sites.

4. Discussion

4.1. Soil Texture

The soil texture results indicate that Kalahari soils contain more than 90% sand (90% quartz), which is consistent with other research findings [52,99,100]. The lack of relationship between grain particle size and precipitation could be attributed to the fact that Kalahari sand deposits are a result of aeolian processes [52], and are independent of precipitation and biological weathering. Although precipitation decreased in the north–south direction, there was no relationship between soil texture and the north–south precipitation gradient. No trends were observed between soil texture and depth, which suggests that biological weathering does not play any significant role in the soil texture of the Kalahari. Furthermore, there was no correlation between soil texture and vegetation structure (woody biomass), which again suggests that vegetation structure, particularly root activity, does not influence soil texture.

4.2. Soil Profile δ¹³C and Tree–Grass Belowground Interaction

In general, soil δ¹³C values decreased in the south–north direction, following an increase in the south–north MAP, a pattern observed in other studies around the world [43,66,94,95,96,97,98,99]. The relationship between trees and precipitation suggests that as MAP increases, trees become more dominant and consequently, an important source of SOC, consistent with the global-scale distribution of woodlands [1,2,101].

Soil δ¹³C values and estimated tree vs. grass C inputs show a distinct and consistent pattern in the soil profile in all the sites, with δ¹³C increasing to 30 cm (or 70 cm, in the case of Tshane) and then decreasing below the 30 cm depth (Figure 2). This pattern is inconsistent with what is expected purely from Rayleigh distillation, e.g., [102], in which enrichment occurs as microbial processes break down SOM during downward translocation in the soil, resulting in a monotonic increase in δ¹³C with depth. This is not surprising, as Wynn, Bird [103] showed that though the Rayleigh distillation model may work for heavy-textured soils, this model is inappropriate for sandy soils like those in the Kalahari. Indeed, ref. [46] have suggested that changes in δ13C are not, in fact, due to Rayleigh fractionation, leaving the validity of this model in question.

Wang, Okin [96] measured surface spatial variations of δ¹³C in the Kalahari. The reported δ¹³C values from their sites were highly consistent with the spatial distribution of trees and grasses, thus suggesting that the isotopic compositions in the soil are meaningful representations of the source of carbon. This pattern gives us confidence that the differences observed with soil depth in the present study are meaningful. Thus, we conclude that the relatively depleted δ¹³C values in the top 10 cm and below 30 cm suggest the importance of tree input at these depths; whereas the relatively higher δ¹³C values in the 30 cm layer observed at all sites suggest the importance of grass input (Figure 2). Our findings are consistent with those of [104], who used stable C isotope analysis in African humid savannas and found that fine roots from trees peaked at 10 cm and below 30 cm, while fine roots from grasses peaked at 20 cm. We therefore believe that the best explanation for the observed peaked δ¹³C pattern (i.e., enrichment in ¹³C at 30 cm, or 30 cm and 70 cm in the case of Tshane, compared to the other soil layers) is that the measured values represent the carbon emplaced by plant roots at those depths. In mesic and semiarid savanna ecosystems in Australia and southern Africa, stable C isotope analysis revealed that tree and grass fine roots were concentrated in the top 20 cm of the soil [105,106].

In the Sonoran desert, Nilsen, Sharifi [107] found that the tree Prosopis developed two distinct root systems, whereby root mats occurred at 10 cm and below 60 cm. Kambatuku, Cramer [108] investigated belowground interactions between trees and grasses, and found that grasses exclusively exploited the 15 cm layer, whereas trees exclusively exploited the deeper subsoil layer at 35 cm. Although the [108] experiment was under a controlled environment, its findings highlight the importance of niche separation between trees and grasses in a savanna ecosystem.

4.3. Soil Profile δ¹³C and Precipitation Gradient

The contribution of C₃ plants to SOC was consistent with vegetation structure in the Kalahari, where tree cover and height increased with MAP [4,55,60] (Table 1). Studies from other parts of the world have shown a similar pattern, where woody contribution increased with MAP [43,48,93]. The e-folding depth and Z₉₅ values for SOC from trees were lowest in Shakawe (

Table 3. Organic C contribution from trees and grasses. Organic C contribution from trees was highest in the wettest site and lowest in the driest site. The e-folding (1/k) is the depth at which SOC contribution decreases to 1/e of the surface value, whereas Z₅₀ and Z₉₅ refer to 50% and 95% of tree-derived SOC in the soil profile.

Site	Trees						Grasses
	Co (g m−3)	Tree Input (g m−2)	k (m)	e-Folding Depth (m)	Z50 (m)	Z95 (m)	Co (g m−3)	Grass Input (g m−2)
Bokspits (180 mm)	965.5	769.4	0.5	2.1	1.5	6.3	881.0	648.0
Tshane (350 mm)	1216.7	808.4	0.9	1.1	0.8	3.4	2084.4	1721.0
Kuke (440 mm)	1423.2	1121.9	0.5	2.0	1.4	6.0	1470.3	1073.5
Shakawe (540 mm)	2729.5	1288.1	1.8	0.6	0.4	1.7	1583.9	830.6

), suggesting that trees in the area could be shallow-rooted. These results are consistent with the ones from both [100,109] studies for the same sites, where they found that 90% of tree root biomass was in the top 75 cm of the soil profile. It is important to reiterate that Kalahari sands have a homogeneous texture and, as a result, would have a uniform impact on δ¹³C values. Therefore, variations in δ¹³C values would be attributed to vegetation structure, which in turn is controlled by precipitation.

The contribution of C₄ to SOC increased from Bokspits, reached a maximum in Tshane, and decreased with an increasing MAP. Our results are consistent with Scanlon, Albertson [110], who showed a non-linear relationship between aboveground grass biomass and MAP, and explained that grass cover peaks at the location with approximately 450 mm MAP. Soil δ¹³C values indicated that the contribution of grass to SOM increases with MAP, reaching a maximum in Tshane, and decreasing at the wetter sites. At these intermediate sites, we speculate that lower overall precipitation (than in the north) reduces the growth of trees, thus allowing water to penetrate deeper into the soil to reach the intermediate-depth grass roots. The result of this would be production of more grass root biomass, which translates into high C sequestration. Furthermore, previous studies in the Kalahari indicated a peak in surface SOC and herbaceous cover in the middle of the Kalahari transect [110,111,112]. In light of this pattern, we follow Scanlon, Albertson [110] assessment in suggesting that low grass cover and productivity in the drier portion of Kalahari could be attributed to low soil moisture, whereas in the wet portion, the low grass cover and productivity could be attributed to belowground competition from trees for water or aboveground competition for light. The intermediate sites appear to provide optimal conditions for grass productivity, and this is reflected by the grass contributions to SOC. Trees in the intermediate sites may facilitate grass growth through improved soil water availability related to hydraulic lift, or through a reduction in incoming solar radiation [113,114]. Furthermore, facilitation might also be a result of improvement in nutrient availability related to litter inputs from trees [115].

4.4. Soil Profile δ¹⁵N Along Precipitation Gradient

Nitrogen is one of the factors limiting productivity in the Kalahari, and plants rely on soil N fixed by cyanobacteria [42,59]. Our results show that the difference between foliar and soil δ¹⁵N was lower than the difference between foliar and atmospheric δ¹⁵N (Table S2). Conversely, foliar δ¹⁵N values were closer to soil δ¹⁵N values than atmospheric values, suggesting that plants in the Kalahari obtain most of their N from the soil rather than through fixation, even in the drier sites where the main woody species are Fabaceae. Aranibar, Otter [59] suggested that plants in the wetter portion of the Kalahari, where Fabaceae are largely absent, could be fixing nitrogen based on their low foliar δ15N values. We have found that soil δ¹⁵N values are also the lowest in Shakawe, our wettest site. It is possible that the low foliar δ¹⁵N values found by Aranibar, Otter [59] in the wetter portion of the Kalahari do not necessarily suggest plant N-fixation but rather reflect the source of N. In fact, the relatively low soil δ¹⁵N in Shakawe could be due to higher soil moisture, which facilitates fixation of atmospheric N by soil microbes. The rate of N fixation by microbes could be higher in the wetter portions of the Kalahari, and lower in the drier portions. The relatively high N-fixation by soil microbes in the wettest portion of the Kalahari could cause soil δ¹⁵N to be closer to that of the atmosphere, compared to the drier portion of the Kalahari. A recent study [14] found that the C-to-N ratios of fine roots of Fabaceae in the Kalahari were not significantly lower (higher N) than those of other species/families, which suggested that Fabaceae were not fixing N₂.

Thus, it is likely that C₃ and C₄ plants in the drier parts of the Kalahari have evolved under low soil N, and have developed adaptation mechanisms to access the limited soil N. Priyadarshini, Prins [116] suggested that grasses, which usually exhibit N limitation, benefit from N redirected by trees. Furthermore, the structure of the Kalahari soils, which are characterized by low quantities of clay and humus colloids (Table 2), low cation exchange capacities, and high porosity and permeability, could potentially enhance N leaching and redirection, leading to a relatively homogenous distribution of δ¹⁵N in the soil, thus making N available to non-N-fixing plants [116].

4.5. Potential Impacts of Climate Change on C Cycling

Future climate change is expected to significantly alter the precipitation totals, increase warming by the end of this century, and lead to extreme weather events [18,19,20,117]. Anthropogenic climate change may alter vegetation composition in dryland ecosystems, particularly savannas [118,119,120]. Because vegetation is the primary source of SOC, changes in vegetation structure will have a direct impact on SOC stocks. In this study, our analysis indicated that SOC input from trees increases with MAP, which is consistent with results from our previous studies where we found low SOC quantities in drier sites of the Kalahari and higher SOC content in the wetter parts Dintwe, Okin [14]. Similarly, studies from other parts of the world indicated that soil C derived from woody plants increases as precipitation increases [43,93].

Because the Kalahari basin is characterized by homogeneous sandy soil and the north–south precipitation gradient, we used a space-for-time substitution approach to assess future climate conditions in soil C stocks. Our results showed that a decrease in precipitation would cause a decrease in the overall soil C sequestration, whereas an increase in precipitation would increase soil C. Dintwe and Okin [22] showed that a decrease in precipitation causes a decrease in net primary productivity (NPP), whereas an increase in temperature results in the overall decrease in NPP and soil respiration. Furthermore, their study revealed that the effect of increased temperature in SOC was more pronounced in the driest and wettest parts of the Kalahari, whereas the effect of precipitation (decrease) was more pronounced in the intermediate sites.

Some parts of the Southern African savannas may become woodier due to increasing CO₂ concentration and thicketization [120,121]. Under these perceived circumstances, C₃ contribution to SOC input will become important, as in the case of Bokspits (driest part), where shrubs play an important role in SOC input. Using our C isotopic results, we concluded that under drier conditions (e.g., Bokspits), shrubs would become an important source of SOC, whereas under wetter conditions (Shakawe), trees would become a significant source of SOC. Grasses, on the other hand, become an important source of SOC under optimal climate conditions (intermediate sites). Furthermore, changes in species composition and structure would affect N cycling in savanna ecosystems. For instance, as shrubs and thickets dominated the drier landscapes, most of the plants required N will be derived from the atmosphere through the process of N fixation. It is likely that under drier conditions, microbial activities would decrease, leading to a decrease in the nitrification process and ultimately resulting in very low quantities of plant-available N.

5. Conclusions

This field-based study aimed to shed light on how climate change would affect soil C sequestration in the savannas. Here, we used the Kalahari as a representative of global savannas. The steep precipitation gradient with homogeneous soil texture provided an environment conducive to decouple climatic effect from biogeophysical factors such as relief and soil texture. Our findings strongly indicate that belowground tree–grass interactions, and soil C sequestration in savannas are more complex than we thought and warrant a detailed soil C and N isotopic analysis. The soil profile of δ¹³C and δ¹⁵N is quite a lot more complex than simply representing root activity of C₃ and C₄ plants at depth. The co-existence of trees and grasses in savannas may be maintained by a vertical differentiation of root functional groups [109], contrary to the commonly assumed separated two-layer root system Walter [122]. However, this is outside the scope of this research.

We found that C₃ plants play an important role in SOC sequestration in the driest and wettest parts of the Kalahari. However, a detailed analysis revealed that in the driest part, shrubs contributed significantly to SOC, whereas in the wettest site, trees played a significant role. C₄-dominated SOC sequestration in the intermediate sites. At these intermediate sites, we speculate that lower overall precipitation (than in the north) reduces the growth of trees, thus allowing water to penetrate deeper into the soil to reach the intermediate-depth grass roots. The result of this would be the production of more grass roots biomass, which translates into high C sequestration.

Using the Kalahari transect as a space-for-time substitution to assess the impact of climate change on soil C sequestration, we conclude that under extreme dry conditions, grasses and shrubs would become an important component of C sequestration. Conversely, under wetter conditions, trees would contribute significantly to soil C. Under moderate climate conditions, grasses have a competitive advantage over trees, and as a result, grasses would become an important source of soil C. It is, however, critical to acknowledge that our study covered only four sites along a 1000 km transect. Because savannas are found on all the continents, save for Antarctica, more detailed and intensive studies are needed to fully understand the potential impact of climate change on carbon cycling in savannas.