Effects of grazing management practices, topographic position, and land cover type on soil organic carbon fractions in semi-arid rangelands of Kenya

Abstract

Background

This study evaluated the effects of grazing management practices, topographic position, and land cover types on mineral-associated organic carbon (MAOC) and particulate organic carbon (POC) in a semi-arid rangeland of Kenya. Research was conducted at Mpala Research Centre (controlled grazing) and Ilmotiok Community Group Ranch (continuous grazing) in Laikipia County. A factorial experimental design with a split-plot arrangement was used in this study. Grazing management practices (controlled and continuous grazing) and topographic positions (midslope, foot slope, and bottomland) were assigned to the main plots, while land cover types (bare ground, grass patches, and tree mosaics) were designated as subplots. Soil samples were collected at 10 cm intervals, 0–10 cm, 10–20 cm, and 20–30 cm depth for MAOC and POC analysis. Data analysis was done using R software, where nonparametric tests were done when the assumptions of normality and homogeneity of variance were violated.

Results

Controlled grazing resulted in higher MAOC (0.361%) and POC (0.683%) compared to continuous grazing (0.352% and 0.548%, respectively), indicating an increase of 2.56% in MAOC and 24.64% in POC under controlled grazing. This can largely be attributed to improved vegetation recovery, especially in midslope areas. The highest MAOC (0.367%) was found in the bottomland, likely due to reduced erosion and improved water retention. The midslope and foot slope positions had lower MAOC means of 0.358% and 0.344%, respectively. Depth analysis showed peak MAOC at 20 cm (0.390%), with controlled grazing resulting in better carbon retention at 30 cm. Similarly, controlled grazing yielded a mean POC of 0.683% versus 0.548% for continuous grazing, with bottomland having the highest POC of 0.754%. A Kruskal‒Wallis tests showed significant differences in MAOC and POC across land cover types (χ² = 42.701, p < 0.001 for MAOC, and χ² = 83.53, p < 0.001 for POC), with tree mosaics and bare land contributing most to POC and MAOC, respectively.

Conclusions

These findings highlight the beneficial role of controlled grazing and diverse land cover in enhancing soil carbon storage. To promote sustainable rangeland management, it is recommended that rangeland managers adopt controlled grazing practices and allow diverse land cover types, such as tree mosaics, to increase carbon sequestration and ecosystem resilience.

Supplementary Information

The online version contains supplementary material available at 10.1186/s13021-025-00319-y.

Introduction

Semi-arid rangelands in Kenya have been traditionally managed through grazing practices [1–3], which have significantly influenced landscape dynamics, including an increase in tree density [4–6]), soil compaction, vegetation loss, altered plant composition, and invasive species spread. Two primary grazing management practices observed in Kenyan rangelands and beyond include continuous and controlled grazing [7, 8]. Continuous grazing involves year-round grazing without a rest period for vegetation recovery [9, 10]. In contrast, for controlled grazing, communities traditionally have specific places reserved for dry season grazing within community land, prioritizing both livestock and wildlife and including provisions for drought scenarios and rest periods [6, 11]. The intensity of grazing varies, with continuous grazing characterized by heavy grazing and controlled grazing by light to moderate intensity [10, 12].

Grazing intensity has a profound effect on soil carbon fractions, particularly on the surface, which is closely correlated with plant species composition and functioning [12–14]. The soil labile and stable carbon fractions constitute the soil organic carbon (SOC) reservoir [15]. The soil labile carbon fractions are highly sensitive to management practices and potentially serve as significant sources of atmospheric carbon dioxide (CO₂) when mismanaged through activities such as overgrazing [16]. The active labile carbon pool is indirectly influenced by grazing through changes in species richness and ecosystem functioning [17, 18].

However, beyond grazing, topographic position significantly influences soil carbon dynamics by affecting erosion rates, moisture retention, and vegetation distribution. Upper slopes tend to experience higher erosion and runoff, reducing SOC retention, while lower slopes typically accumulate more organic matter due to sedimentation and improved moisture [19]. Similarly, land cover types (e.g., bare ground, grasslands, and tree mosaics) play a crucial role in soil carbon input and protection. Vegetation cover enhances litter input and reduces erosion, thus improving both labile and stable carbon fractions [20]. Despite their importance, few studies in semi-arid Kenyan rangelands have concurrently examined how land cover and topographic variation influence soil carbon fractions. Studies have shown that continuous grazing has negative effects on soil properties such as total nitrogen, phosphorus, and cation exchange capacity (CEC) [21, 22], in contrast with controlled grazing practices. However, the effects of increasing depth on soil carbon pools in semi-arid rangelands, where grasses and trees coexist, are less understood [23, 24].

Labile carbon fractions are derived primarily from vegetation and animal droppings [25, 26], with urine and faecal matter from herbivores containing high levels of labile carbon that are highly decomposable and act as hot spots for greenhouse gas emissions. Grazing directly impacts vegetation composition and litter, thereby influencing evapotranspiration and shading. This further impacts the decomposition and mineralization rates of labile carbon by affecting soil moisture, oxygen availability, and temperature [27, 28]. In a broader context, in temperate regions experiencing freezing and thawing, grazing has been observed to increase labile carbon quantities [29], which serve as energy sources for denitrifying microbes and can contribute to nitrous oxide emissions [30].

Most studies on soil carbon in Kenya focus on soil organic carbon in the top 0–30 cm depth, with limited documentation on various carbon fractions and their trends in deeper soil profiles [31–33]. Soil depth provides insights into carbon accumulation over time, reflecting the historical influences of weather, land management practices, and environmental factors on SOC dynamics [34, 35]. SOC typically decreases with depth, and the subsoil tends to have more stable carbon than the topsoil does. The current literature concerning semi-arid rangelands in Kenya predominantly addresses total soil organic carbon, with insufficient data on the carbon fractions necessary for policy formulation regarding grazing management [4, 36].

While there is extensive literature on grazing effects, there remains a critical gap regarding the interactions among topographic position, land cover types, and soil carbon fractions, particularly in semi-arid environments. Addressing this gap is vital to inform localized soil conservation and grazing management strategies that consider terrain and vegetation structure. Particulate organic carbon (POC) and mineral-associated organic carbon (MAOC) are two key fractions of soil organic carbon that differ in composition, stability, and function. POC consists primarily of undecomposed or partially decomposed plant fragments, is nutrient-poor, and has a particle size greater than 53 μm [37], making it more labile and easily decomposed by soil microorganisms. In contrast, MAOC is composed of microbial and plant-derived organic matter with low molecular weights, is relatively nutrient rich [38], and has a particle size smaller than 53 μm. While POC contributes to short-term nutrient cycling and soil structure, MAOC is tightly bound to fine soil particles, offering greater stability and playing a vital role in long-term carbon storage and soil fertility [39, 40]. Both fractions are essential for maintaining soil health and regulating carbon dynamics in ecosystems. The aim of this study was to assess the effects of different grazing management practices, topographic positions and land cover types on MAOC and POC in semi-arid rangelands of Kenya. This study sought to provide comprehensive insights into how these interacting factors influence soil carbon dynamics, which is crucial for sustainable land management in arid and semi-arid ecosystems.

Materials and methods

Site description

Laikipia County (Fig. 1) is situated between latitudes 0° 25’ 47.1” N and 0° 17’ 53.7” S and longitudes 36° 57’ 32.9” E and 36° 47’ 39.9” E and is characterized as semi-arid to arid rangelands within agroecological Zone V. The County experiences two distinct rainy seasons: short rains, which occur between October and November, and the primary rainy season, which lasts from April to May [41]. The average annual rainfall is approximately 600 ± 50 mm, with temperatures ranging from 25 to 30 °C in the highland areas and 12–17 °C in the lowlands. Laikipia County supports various economic activities, including livestock rearing, tourism, both private and public conservancies, ranching, large- and small-scale farming, horticulture, sand harvesting, and quarrying. Pastoralism and charcoal production are vital for the local economy.

Fig. 1.

Map showing the study site Mpala and neighbouring Ilmotiok ranches

Topographically, Laikipia County lies on the Laikipia Plateau, a gently undulating landscape with prominent hills and ridges interspersed with low-lying plains [42]. Elevation ranges from 1,500 m near the Ewaso Nyiro Basin to 2,611 m above sea level, influencing both temperature and moisture gradients [43]. The area is drained by several seasonal rivers and tributaries that feed into the Ewaso Nyiro River, a critical watercourse for both human and wildlife use [44, 45]. Drainage patterns are predominantly dendritic, and water retention is often constrained by shallow soils and high surface runoff, especially in sloped areas.

In 2016, two neighbouring ranches, Mpala Research Centre and Ilmotiok Community Group Ranch, were chosen for study. The Mpala Research Centre, which is privately owned, follows a controlled grazing system with low to moderate grazing intensity and has a stocking density of 24 TLU/km² (Tropical Livestock Units per square kilometer), excluding elephants. On the other hand, Ilmotiok Community Group Ranch, with a stocking density of 37 TLU/km², experiences heavy grazing intensity [6], as detailed in Table 1. TLU is a standardized unit that allows comparison of different livestock species by converting them into an equivalent of one mature bovine animal weighing 250 kg. In the study area, common livestock species include cattle, goats, sheep, and camels, with cattle being the dominant group. These animals are herded in mixed groups and serve as the primary livelihood source for the local pastoralist communities. The county’s elevation ranges from 1,500 m at the Ewaso Nyiro Basin to 2,611 m above sea level. The two dominant soil types are Ferric and Chromic Luvisols and Pellic Vertisols [46], with further soil details provided in Table 1.

Table 1.

Soil properties and grazing management in the Ilmotiok and Mpala ranches

Grazing management		Ilmotiok	Mpala
Total livestock biomass (TLU)		897	2074
Total livestock and wildlife (TLU)		1334.4	4572.7
Land area (Ha)		3651	19,000
Stocking density (TLU/Km2)		37	24
Stocking intensity		Heavy	Light to moderate
Soil property
pH (H2O 1:2)		7.06	6.63
EC (dS/m)		0.13	0.07
Sand %		62.25	55.49
Silt %		26.99	32.25
Clay %		10.76	12.33
Texture class		Sandy clay loam	Sandy loam
Bulk density (g/cm3)		1.48	1.51
Total N (g/kg)		1.11	1.51
CN ratio		5.19	5.05
Total OC (%)		0.85	1.01

Source: [6]

Experimental design and treatment

The study employed a split-plot factorial experimental design. The main plots were arranged in a factorial combination of grazing management practices (controlled vs. continuous) and topographic positions (midslope, foot slope, and bottomland). Within each main plot, land cover types (bare ground, grass patches, and tree mosaics) were assigned as subplots (Fig. 2).

Fig. 2.

Hierarchical overview of experimental design

The design was replicated using blocks to account for field variability. The land cover types, including bare ground, grass patches, and tree mosaics, were randomly selected on the basis of their prevalence in the study area, with more tree density being more pronounced in the bottomland. The topographical positions were categorized into midslope, foot slope, and bottomland, following the classification by [47] (Fig. 3). A permanent reference point was established within each gradient (A-F), with the coordinates and altitude recorded via a GPS device.

Fig. 3.

Topographical classification. (Source: Almquist et al. (2002)

The geographical position of each grazing management practice was measured via a clinometer, and two rods were marked at equal heights. At the reference point (Point A), one rod was held vertically, whereas the second was placed 30 m away, parallel to the first, as shown in Fig. 3. Topographical variation was calculated via Eq. (1):

1

Where: ∆ T = change in topographical point, D = distance between sampling points, and C = clinometer reading.

A cloth tape was used to mark each transect, with 10 points along each transect classified as bottomland, midslope, or foot slope, each comprising 33% of the transect’s total length of 200 m. The ratio of each topographic class within the transect was calculated. Each position was classified as midslope (A-B), foot slope (B-C), or bottomland (C-D) on the basis of the elevation relative to adjacent points. For each topographical location, land cover types (bare ground, grass patches, and tree mosaics) were chosen to establish three transects, each measuring 200 m in length and 50 m in width (totalling 1 hectare per transect). Each 200 m transect was subdivided into four continuous 50 m by 50 m plots (2500 m² each) and assigned the dominant land cover type based on visual classification. Soil samples were taken from each land cover type to a depth of 30 cm at 10 cm intervals and analysed for MAOC and POC. Soil depth (0–10 cm, 10–20 cm, and 20–30 cm) was treated as a repeated measure within each land cover subplot, resulting in a split-split-plot design structure with grazing management and topographic position as main plots, land cover types as subplots, and depth as a within-plot factor.

Soil sampling and analysis

Disturbed soil samples (approximately 250 g) from each land cover type were collected from a cleared, flat surface, free of roots and grass, using a calibrated soil auger (1 m height, 5 cm diameter) at depth intervals of 10 cm from the surface to 30 cm, and analysed for MAOC and POC. In total, 162 soil samples were collected, representing a combination of 2 grazing management practices, 3 topographical positions, 3 land cover types, 3 soil depths, and 3 replicates. Soil samples were placed in sterile, airtight polyethylene bags immediately after collection and stored in a cooler with ice packs in the field. The soil sampling was done within one day by 8 trained personnel from the University of Nairobi and Kenya Forestry Research Institute (KEFRI). Upon arrival at the laboratory, samples were stored shortly at 4 °C and subsequently air-dried within 48 h. All visible organic residues, including fine roots and mycelia, were manually removed prior to analysis. Prior to carbon analysis, soil samples were subjected to standard pre-treatment procedures; after air-drying, they were gently crushed and passed through a 2 mm sieve to remove coarse fragments, roots, and visible organic matter. Subsamples for MAOC and POC determination were further ground to pass through a 0.5 mm sieve to ensure homogeneity.

These samples were analysed for POC and MAOC and classified into different carbon fractions as described by [48]. For each sample, 50 g of fresh soil was dispersed in a 10% calgon solution for 6 h, followed by wet sieving via a sequence of sieves with mesh sizes of 2 mm, 250 μm, and 50 μm. The soil fractions retained on each sieve were oven-dried at 65 °C for 24 h, after which they were weighed to determine the percentages of POC (in the 2–250 μm fraction) and MAOC (in the > 53–38 μm fraction). The carbon concentration for each fraction was then measured via the Walkley‒Black method [49], with the percentage of carbon content calculated via Eq. 2. This study primarily captured spatial variation in land use through different grazing management systems, land cover types, and topographic positions, while temporal changes were not within the scope of the current sampling design.

2

Where: m.e. = milliequivalents. The value 0.3 is a constant derived from the stoichiometric relationship between the amount of dichromate consumed and the corresponding amount of carbon oxidized, assuming that 1 mL of 1 N dichromate is equivalent to 0.003 g of carbon; thus, when converted to percentage and factored for unit consistency, the coefficient becomes 0.3. The ferrous sulfate (FeSO₄) is used to titrate the excess potassium dichromate (K₂Cr₂O₇) that was not reduced by the organic carbon in the soil sample.

Data analysis

The data analysis was conducted to explore the effects of grazing management, land cover type, and soil depth on MAOC and POC via various statistical methods. Given the split-plot factorial experimental design, with grazing management as the main plot factor and land cover type as the subplot factor, the data structure was first examined for suitability with linear mixed-effects models. However, due to violations of model assumptions and persistent non-normality and heteroscedasticity even after transformation, a non-parametric approach was deemed more robust and appropriate. Descriptive statistics, from the pastecs package [50], were utilized to summarize central tendencies and data distributions. The homogeneity of variance among the grazing management and land cover type groups was evaluated via Levene’s test from the car package [51], whereas the normality of the data distribution was assessed via the Shapiro‒Wilk test from the stats package. The graphics utilized the ggplot2 package [52] via R Core Team software version 4.3.3 [53].

Results

Effects of grazing management, land cover, and soil depth on MAOC

The mean MAOC values varied across the grazing management systems, with controlled grazing management resulting in a mean MAOC of 0.361% compared with 0.352% under continuous grazing management, representing a 2.56% increase in MAOC under controlled grazing compared to continuous grazing (Fig. 4). Controlled grazing also resulted in lower variability in MAOC values. Topographical slope position influenced the distribution of MAOC, with bottomland exhibiting the highest mean MAOC of 0.367, followed by midslope at 0.358 and foot slope at 0.344%; this corresponds to a 6.27% decrease from bottomland to foot slopes and a 2.45% decrease from bottomland to midslope. The midslope position presented the greatest variability in MAOC values across all slope positions. Interactive effects of grazing management and slope position were observed. The bottomland had slightly higher MAOC values under continuous grazing (Fig. 4), whereas controlled grazing was associated with increased MAOC in the midslope areas. Across all the grazing systems, bottomland consistently presented the highest MAOC values.

Fig. 4.

MAOC across grazing management practices, topographic position, and land cover types. Mgt stands for management

The mean MAOC varied across land cover types, with bare land exhibiting the highest mean value of 0.417%, followed by tree cover at 0.331% and grass cover at 0.326%, indicating a 20.62% decrease from bare land to tree cover and a 21.82% decrease from bare land to grass cover. Under bare land, continuous grazing resulted in higher MAOC values than did control grazing. Conversely, in both grassland and tree land, controlled grazing resulted in higher MAOC values than did continuous grazing, with tree cover resulting in slightly higher MAOC values than did grassland, recording a 1.53% increase from grass to tree cover under controlled grazing.

The mean MAOC was highest at the 20 cm soil depth, at 0.390, followed by 0.342 at 10 cm and 0.338 at 30 cm depths, showing a 12.31% decrease from 20 cm to 10 cm and a 13.33% decrease from 20 cm to 30 cm. At the 10 and 20 cm depths, continuous grazing resulted in higher MAOC values than did the control, with greater variability observed under continuous grazing. At the 30 cm depth, the MAOC levels under controlled grazing were similar to those under continuous grazing.

Effects of grazing management, land cover, and soil depth on POC

The mean POC was greater under controlled grazing, at 0.683%, compared to 0.548% under continuous grazing, a 24.64% increase under controlled grazing compared to continuous grazing. Controlled grazing also exhibited greater variability. The mean POC varied by slope position, with bottomland having the highest mean of 0.754%, followed by foot slope at 0.640% and middle slope having the lowest mean of 0.454%, reflecting a 15.12% decrease from bottomland to foot slope and a 39.79% decrease from bottomland to middle slope (Fig. 5). Compared with continuous grazing, controlled grazing resulted in higher POC levels in the bottom slope and foot slope areas. In the midslope areas, continuous grazing resulted in higher POC values than did controlled grazing.

Fig. 5.

POC across grazing management practices, slope positions, and land cover types

The mean POC was highest in tree-covered areas, at 0.891%, followed by grass-covered areas at 0.666%, and lowest in bare land at 0.291%, corresponding to a 25.25% decrease from tree-covered areas to grass-covered areas and a 67.34% decrease from tree-covered areas to bare land. The tree-covered areas presented the widest range of POC values, from 0.440 to 1.520%, reflecting variations in tree density, organic inputs, and erosion control. Grass-covered areas presented moderate POC levels, ranging from 0.110 to 1.430%, which were influenced by grazing intensity, soil type, and seasonal dynamics. Bare land, with the lowest mean POC, presented minimal organic inputs and high susceptibility to erosion. Compared with continuous grazing, controlled grazing resulted in higher POC levels in both bare and tree-covered areas. However, in grass-covered areas, continuous grazing resulted in higher POC values than did controlled grazing.

The mean POC was lowest at the 30 cm depth, at 0.414%, with the highest variability ranging from 0.090 to 0.940%. Intermediate depths (20 cm) presented a mean POC of 0.631%, whereas the shallowest depth (10 cm) presented the highest mean POC, at 0.802%, indicating a 21.32% decrease from 10 cm to 20 cm and a 48.38% decrease from 10 cm to 30 cm.

Trends in combinations of grazing management and land cover type

The highest mean MAOC values were observed for controlled grazing, particularly for midslope bare land, at 0.4867%, foot slope bare land, at 0.4433%, and bottomland bare land, at 0.3967%. In contrast, the lowest mean MAOC values were recorded under continuous grazing, specifically for midslope tree cover, at 0.20%, foot slope grass, at 0.2533%, and bottomland grass, at 0.2967%. For POC, the highest mean values were found under controlled grazing, with bottomland cover at 1.4933%, bottomland cover at 1.3900%, and foot slope tree cover at 1.3533%. The lowest mean POC values were observed under continuous grazing with bare land.

Interactions

The results (Fig. 6) suggest that the effect of grazing management on the percentage MAOC differs depending on the land cover type. For controlled grazing, the MAOC for grass increases significantly, potentially surpassing that of tree cover under continuous grazing. Notably, MAOC increases from controlled to continuous grazing on bare land and from continuous to controlled on grass land. The mean POC for bare land was the lowest across both grazing management types. Compared with continuous grazing, tree cover under controlled grazing significantly increased the mean POC. Grassland maintains a relatively stable POC level, with a slight increase under controlled grazing.

Fig. 6.

Grazing management-land cover type interaction for MAOC and POC

For MAOC, bare land consistently shows the highest values across all topographic positions, peaking at midslope, around 0.45%, while grass and tree cover have lower, more stable MAOC levels, around 0.30–0.35%, with grass slightly increasing from bottomland to midslope (Fig. 7). Grass has a slightly higher MAOC in bottomland, around 0.32% compared to tree, around 0.30%, but this trend reverses as tree surpasses grass at footslope and midslope, stabilizing around 0.33% while grass dips to around 0.30%. In contrast, POC exhibits a stark decline in bare land from bottomland, 0.6% to midslope, 0.2%, while grass and tree cover maintain higher POC levels, with tree cover peaking at bottomland, 0.9% and grass showing a slight increase toward midslope, around 0.7%. The interactions highlight that bare land significantly influences MAOC and POC differently across topographic positions, with MAOC favouring midslope and POC favouring lower positions, while grass and tree cover show more consistent trends.

Fig. 7.

Interaction between topographic position and land cover type

Analysis of variance

For MAOC, the homogeneity of variance was not violated for grazing management, land cover type, or their interaction (p values > 0.05). However, normality was violated for MAOC in most groups, particularly under the bare and tree land cover types and continuous grazing management. For POC, there was significant differences in variance across grazing management, land cover type, and their interaction (p values < 0.001). There were nonnormalities for POC in several groups, especially under continuous grazing management under bare and grass land cover types.

For MAOC, Levene’s tests indicated no significant differences in variances (e.g., grazing: F(1,160) = 2.2455, p = 0.136; interaction: F(5,156) = 1.2237, p = 0.3005), but Shapiro-Wilk tests showed non-normality in several groups (e.g., Bare: W = 0.94272, p = 0.01212; Tree: W = 0.93601, p = 0.006377). For POC, variances were unequal (e.g., grazing: F(1,160) = 19.531, p < 0.001; interaction: F(5,156) = 16.995, p < 0.001), and non-normality was also observed (e.g., Continuous: W = 0.96221, p = 0.01732). A log-transformation was applied to the melted dataset; however, re-testing still showed violations of assumptions: Levene’s test showed unequal variances for grazing (F(1,26) = 9.7966, p = 0.004283), and Shapiro-Wilk tests continued to show non-normality in some groups (e.g., Grass: W = 0.83972, p = 0.04378; Continuous.B: W = 0.92575, p = 0.009338).

These results support the use of non-parametric tests to ensure robust analysis and address data validity concerns. Although the factorial split-plot design initially warranted the use of mixed-effects ANOVA to assess both main and interaction effects of grazing management and land cover type, violations of ANOVA assumptions (normality and homoscedasticity), even after transformation, led to the adoption of non-parametric Kruskal–Wallis tests for group comparisons, followed by Dunn’s post hoc tests with a Bonferroni-adjusted significance level of α = 0.05.

Mineral-associated organic carbon

The Kruskal‒Wallis test revealed no significant differences in MAOC levels among the different grazing management types (χ² = 0.6998, p = 0.4029). However, significant differences were found among land cover types (χ² = 42.701, df = 2, p < 0.001). The Dunn post hoc test revealed significant differences between bare and grass (p < 0.001) and between bare and tree (p < 0.001) but no significant difference between grass and tree (p = 0.7277) (Table 2).

Table 2.

Mean MAOC by land cover type

Land cover	Mean ± SD
Bare	0.412 ± 0.07 a
Tree	0.331 ± 0.06 b
Grass	0.326 ± 0.06 b

Grass and tree-covered plots exhibited comparable MAOC levels, which were statistically lower than those in bare areas. Grazing management and slope position did not exert significant main effects on MAOC (χ² = 0.7, p = 0.4029; χ² = 2.691, p = 0.2605, respectively) (Table 3).

Table 3.

Kruskal-Wallis test results for MAOC (All Effects)

Effect	Chi_Squared (χ²)	Df	P_Value
Grazing.mgt	0.7	1	0.4029
Slope.position	2.691	2	0.2605
Land.cover.type	42.701	2	5.30E-10
Soil.depth	18.432	2	9.90E-05
Grazing.mgt: Slope.position	15.357	5	0.0089
Grazing.mgt: Land.cover.type	48.498	5	2.80E-09
Grazing.mgt: Soil.depth	19.391	5	0.0016
Slope.position: Land.cover.type	58.522	8	9.10E-10
Slope.position: Soil.depth	25.16	8	0.0015
Land.cover.type: Soil.depth	94.967	8	4.50E-17
Grazing.mgt: Slope.position: Land.cover.type	80.575	17	3.00E-10
Grazing.mgt: Slope.position: Soil.depth	39.834	17	0.0014
Grazing.mgt: Land.cover.type: Soil.depth	103.894	17	1.70E-14
Slope.position: Land.cover.type: Soil.depth	119.372	26	6.40E-14
Grazing.mgt: Slope.position: Land.cover.type: Soil.depth	148.95	53	4.70E-11

A significant interaction between land cover and grazing management was detected (χ² = 48.498, p < 0.001), where controlled grazing enhanced MAOC accumulation in grass and tree covers. Under grass cover, controlled grazing yielded a MAOC concentration of 0.348%, compared to 0.304% under continuous grazing (Table 4).

Table 4.

Mean MAOC by grazing management and land cover type

Grazing management	Land cover type	Mean ± SD
Continuous	Bare	0.421 ± 0.07 a
Controlled	Bare	0.404 ± 0.06 a
Controlled	Grass	0.348 ± 0.06 b
Continuous	Tree	0.331 ± 0.07 b
Controlled	Tree	0.331 ± 0.06 b
Continuous	Grass	0.304 ± 0.04 b

Particle organic carbon

POC content varied significantly across land cover types (χ² = 83.530, p < 0.001), slope positions (χ² = 19.715, p < 0.001), and soil depth (χ² = 27.747, p < 0.001) (Table 5).

Table 5.

Kruskal-Wallis test results for POC (All Effects)

Effect	Chi_Squared (χ²)	Df	P_Value
Grazing.mgt	3.058	1	0.0803
Slope.position	19.715	2	5.20E-05
Land.cover.type	83.527	2	7.30E-19
Soil.depth	27.747	2	9.40E-07
Grazing.mgt: Slope.position	27.356	5	4.90E-05
Grazing.mgt: Land.cover.type	90.156	5	6.20E-18
Grazing.mgt: Soil.depth	30.924	5	9.70E-06
Slope.position: Land.cover.type	111.863	8	1.60E-20
Slope.position: Soil.depth	47.792	8	1.10E-07
Land.cover.type: Soil.depth	112.528	8	1.10E-20
Grazing.mgt: Slope.position: Land.cover.type	129.488	17	2.40E-19
Grazing.mgt: Slope.position: Soil.depth	55.601	17	5.50E-06
Grazing.mgt: Land.cover.type: Soil.depth	119.479	17	1.90E-17
Slope.position: Land.cover.type: Soil.depth	141.526	26	7.30E-18
Grazing.mgt: Slope.position: Land.cover.type: Soil.depth	159.996	53	1.10E-12

Tree-covered sites had the highest POC concentrations, followed sequentially by grass and bare plots (Table 6).

Table 6.

Mean POC by land cover type

Land cover	Mean ± SD
Tree	0.891 ± 0.29 a
Grass	0.666 ± 0.34 b
Bare	0.291 ± 0.13 c

Controlled grazing consistently increased POC content across all land cover types. The highest value (1.04%) was recorded under tree cover with controlled grazing, whereas the lowest value (0.253%) occurred in bare land under continuous grazing (Table 7; Fig. 6).

Table 7.

Mean POC by grazing management and land cover type

Grazing management	Land cover type	Mean ± SD
Controlled	Tree	1.04 ± 0.29 a
Continuous	Tree	0.738 ± 0.2 b
Controlled	Grass	0.678 ± 0.44 b
Continuous	Grass	0.654 ± 0.19 b
Controlled	Bare	0.328 ± 0.13 c
Continuous	Bare	0.253 ± 0.11 c

Interactive effects of grazing management and land cover on MAOC and POC

Multiple significant two-way and higher-order interaction effects were observed. For MAOC, land cover × grazing (χ² = 48.498), land cover × soil depth (χ² =94.967), and slope × land cover (χ² = 58.522) interactions were all significant at p < 0.001 (Table 3). For POC, significant interactions were found for land cover × grazing (χ² = 90.156), grazing × soil depth (χ² = 30.924), and slope × soil depth (χ² = 47.792) (Table 5). Four-way interactions (land cover × grazing × slope × depth) were also significant for both MAOC (χ² = 148.950) and POC (χ² = 159.996), indicating that soil organic carbon fractions are co-regulated by complex interdependencies among biophysical and management variables.

Discussion

Bare land exhibited the highest mean MAOC but the lowest mean POC across both grazing management practices reflecting minimal organic input and increased susceptibility to erosion. This suggests that the absence of vegetation exposes soil to aerobic conditions, facilitating the conversion of leaf litter and other residues into mineral-associated organic carbon (MAOC), while reducing particulate organic carbon (POC). Under both grazing treatments, continuous grazing resulted in greater MAOC compared with controlled grazing on bare land, a trend likely caused by intensive grazing and high livestock stocking density. The higher MAOC observed on bare land under continuous grazing reflects the tendency for intensive grazing to reduce vegetative cover, leaving soil exposed and promoting the mineralization of surface residues. In this way, intensive grazing appears to shift organic matter dynamics towards the more stabilized mineral-associated fraction. Similar observations have been reported by [54], who attributed increased MAOC accumulation to the loss of protective vegetation and the resulting exposure of organic matter to aerobic decomposition.

While bare land under continuous grazing promoted MAOC accumulation, controlled grazing supported higher POC across both bare and treecovered areas. Reduced soil disturbance and improved vegetation cover under controlled grazing fostered particulate organic matter retention by limiting erosion and supporting soil structure. This is consistent with [55], who noted that managed grazing promotes soil aggregation and plant growth. Similarly, [56], observed that such practices reduce compaction and maintain moisture. In contrast, in grasslands, continuous grazing resulted in higher POC, likely due to increased trampling and the accelerated breakdown of plant residues, as also observed by [57]. Nonetheless [58], cautioned that this surface accumulation of organic matter may be shortlived, with longterm effects potentially degrading soil health and carbon storage. The absence of vegetation further compounded these dynamics, leaving soil exposed and susceptible to erosion, a trend similarly reported by [36], where bare soils suffered greater carbon losses due to nutrient depletion. Together, these findings highlight the pivotal role of grazing management and vegetation cover in shaping soil organic matter dynamics and sustaining longterm carbon storage. Significantly higher mean MAOC values on bare land under controlled grazing resulted from reduced soil disturbance, which promoted organic carbon accumulation. This aligns with findings from [59], who noted that managed grazing practices lead to improved soil health by stabilizing soil aggregates and enhancing organic matter retention. The lower values under continuous grazing than under tree cover are likely due to soil compaction and increased erosion. Continuous grazing often leads to overgrazing, which compresses soil and diminishes its ability to store carbon, as highlighted by [60], whereas the higher POC under controlled grazing reflects better soil structure and reduced erosion. Enhanced vegetation cover not only contributes to organic matter inputs through litterfall and root biomass but also protects the soil from erosion, facilitating greater carbon sequestration. These findings are consistent with those reported by [59] and [60], both of which emphasize the importance of grazing management in maintaining soil health and increasing carbon storage.

Bottomland positions presented the highest MAOC across grazing management practices, reflecting their favorable conditions for organic matter accumulation due to reduced erosion and enhanced water retention. In continuous grazing, slightly higher MAOC was observed in the lower slope positions relative to the upper slopes, likely as a result of organic matter accumulation and reduced runoff. In contrast, controlled grazing significantly increased MAOC in midslope areas, suggesting that improved ground cover and erosion control supported greater stabilization of soil organic matter. These patterns align with observations by [61] and [62], which highlight the role of topographic position and grazing management in promoting soil carbon retention. Similarly, controlled grazing supported higher POC levels than continuous grazing in bottom slopes and foot slope positions, likely due to reduced soil disturbance and better vegetation cover, which enhances organic matter accumulation. In contrast, continuous grazing resulted in higher POC on midslope positions, likely due to increased grazing pressure and trampling, which accelerated the breakdown and redistribution of surface organic matter. This pattern closely aligns with the observations of [57], who reported trampling accelerated decomposition and elevated organic carbon formation. Similarly [58], observed that trampling enhances organic incorporation in mid-gradient zones, contributing to higher POC concentration.

The lower MAOC and POC levels at the foot slope in both grazing management practices can be attributed to more frequent soil disturbances and less organic matter deposition than at the bottom, although it still benefits from some runoff accumulation. These findings underscore how slope position significantly influences organic carbon dynamics, with flatter areas favouring the retention of carbon-rich materials [63]. Since MAOC is usually bound to clay minerals, a relatively high concentration is expected to be found on bottom slopes as a function of geological sorting, which is consistent with the findings of [63].

Controlled grazing supported higher POC across all soil depths compared with continuous grazing, indicating its role in enhancing organic matter accumulation throughout the profile. In contrast, continuous grazing resulted in higher MAOC at 10 and 20 cm depths, with greater variability observed under this regime, while both treatments exhibited comparable MAOC levels at 30 cm. This suggests that controlled grazing may be more effective in maintaining MAOC at deeper soil layers. Similar depthdependent variations in MAOC have been reported in earlier studies [59, 64, 65]. The higher POC under controlled grazing can be attributed to reduced grazing pressure, allowing for better vegetation recovery and increased organic matter inputs across the soil profile. These findings highlight the benefits of managed grazing practices in promoting soil health and longterm carbon storage, as also emphasized by [66] and [56].

Interaction effects between land cover, slope position, and soil depth were the dominant drivers of soil organic carbon dynamics, with MAOC and POC responding strongly to these combined influences. Land cover and soil depth primarily shaped MAOC indicating that the nature of surface vegetation and the depth of the soil profile govern mineralassociated carbon stabilization. In contrast, POC responded significantly to slope position, land cover, and soil depth, underscoring its sensitivity to topographic and surface organic matter dynamics. These findings closely agree with [67], who associated higher POC on the topsoil while deeper soil layers had less due to reduced fine root biomass. In shrub and herbaceous vegetation in high topography root biomass is higher in the topsoil, while woody vegetation allocates root biomass to deeper soil layers [68]. The lack of significant direct effects for grazing management suggests that its role in soil carbon stabilization operates within this interactive context, where its impacts emerge only when combined with sitespecific land cover, topography, and depth conditions. The interaction between land cover type and soil depth for MAOC (p ≤ 0.001) and for POC (p ≤ 0.001) likely reflects varying organic matter inputs and decomposition rates across vegetation types and soil depths. Tree-covered areas enhanced MAOC through greater root inputs at deeper depths. The highly significant interaction effect on POC in tree-covered areas was enhanced through increased litter inputs and reduced erosion. These findings agree with [69] and [70], who attributed that different vegetation types have varying decomposition rates and biomass accumulation, which influences SOC input and distribution.

POC levels were lowest in bare land across both grazing management types, aligning with observations by [71] and [54], and indicating minimal organic carbon retention under exposed conditions. Conversely, compared with continuous grazing, tree cover under controlled grazing significantly increased the mean POC, suggesting that effective management practices enhance carbon storage. Grassland areas maintained a relatively stable POC level, with a slight increase under controlled grazing, indicating its role in carbon sequestration. The higher MAOC observed on bare ground compared with grass and treecovered patches in continuously grazed areas may be attributed to high temperatures at the site, which accelerate decomposition and erosion of particulate organic matter, thereby facilitating its conversion to mineralassociated forms. This trend is consistent with findings from earlier studies [72, 73], which have linked reduced vegetation cover and increased soil exposure to altered soil carbon dynamics.

The significant interactions among grazing management, slope position, land cover type, and soil depth (MAOC: p ≤ 0.001, POC: p ≤ 0.001), highlight the highly synergistic nature of soil organic carbon dynamics. The significant MAOC interactions is due to differences in soil microbial activity, carbon stabilization mechanisms, and erosion patterns across the landscape. POC accumulation is a product of complex synergies, possibly driven by differences in root exudates, microbial activity, and soil stability across varying landscape and management conditions. These findings suggest that management strategies aimed at increasing organic carbon should consider the combined effects of these environmental factors rather than focusing on any single variable, emphasizing the need for integrated approaches in soil carbon management. Similar conclusions have been reported by [74–76], reinforcing the need for integrated approaches to optimize soil carbon sequestration across varied agroecological settings.

Conclusions

Compared with continuous grazing, controlled grazing resulted in higher levels of MAOC and POC, particularly in areas with improved vegetation recovery and reduced soil disturbance. The bottomland positions presented the highest carbon accumulation due to better water retention and organic matter deposition. The results underscore the critical role of land cover types and elevation, with tree-covered areas showing enhanced POC levels due to consistent organic inputs, whereas bare land demonstrated high MAOC levels, likely influenced by reduced competition for nutrients. Overall, this study highlights the necessity of adopting sustainable grazing management practices to increase soil carbon storage and promote ecosystem resilience.

To promote sustainable land management in semi-arid rangelands, we recommend the adoption of controlled grazing practices to increase carbon sequestration and improve soil health. The incorporation of diverse land cover types, such as tree mosaics, is essential for maximizing organic carbon inputs and reducing soil erosion. Additionally, land management strategies should focus on restoring vegetation cover in heavily grazed areas to bolster organic matter accumulation and enhance overall ecosystem functioning. Continuous monitoring and adaptive management approaches are crucial to ensure that grazing practices align with ecological goals, ultimately contributing to improved carbon dynamics and resilience in these vulnerable ecosystems.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

AG, SM1, and JM conceptualized the idea, and RO and SK provided funding through the SLEEK and USAID NAS-PEER projects. AG conducted the field data collection and laboratory analysis. SM1 and SM reviewed the literature, analysed the data and co-drafted the manuscript. SM, RO, JM, JC, and SK reviewed the manuscript.

Funding

The study was funded by USAID NAS-PEER and SLEEK (System for Land-based Emission Estimation in Kenya, http://www.sleek.environment.go.ke/).

Data availability

No datasets were generated or analysed during the current study.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.