The study was undertaken in Chepareria ward of West Pokot County located between 35° 7′ and 35° 27′ latitude and 01°15′ and 01° 55′ longitude. The area is mainly comprised of plains with altitudinal range of 1,100–1,600 m above sea level surrounded by hills and ridges with elevations of up to 3,000 m as shown in Figure 1. The surrounding hills and ridges collect large amounts of surface run-off that descends at high speeds contributing to extreme soil erosions (Mutio et al., 2021). The area experiences a temperature range between 10°C and 30°C and receives an average of 600 mm of annual precipitation (Nyberg et al., 2015). The location is dominated by four major soil types which are; Lixisols, Leptosols, Cambisols, and Luvisols. Lixisols, Leptosols, and Cambisols have weak structures with a high sand content of more than 65% (Schad et al., 2014). The study was conducted on Arenic Lixisols which is the soil type that dominates the study area selected for testing rehabilitation measures and because of its inherent weak properties that make it highly erodible and with visible signs of erosion over substantial areas in the region (Muturi, 2016), as illustrated in Figure 2.

The study scope was limited to two soil vegetation cover practices widely adopted in the area; maize-bean intercrop and pasture farms. A mechanical measure, primarily locally developed terraces (Fanya Juu and Fanya Chini) are integrated into the practices to arrest and capture surface run-off for soil and water conservation. To evaluate additive influence of vegetation cover practices and terraces, three farms with each practice (Maize-bean intercrop + terraces, and pasture + terraces) were selected as treatments. A similar number of the selected vegetation cover without terraces were also included in the sampling. Three replicate farms were selected based on similar soil fertility management practices over 5 years per each soil conservation measures combination, resulting in a total of 12 farms. In addition, three severely degraded rangeland with no management were also selected as a negative comparative treatment, and finally, to evaluate terrace age on SOC fractions in cropping system, soil samples were taken from the maize-bean intercropped farms that had hosted the terraces for five, four, three, two, and one year as shown in Figure 1. For the determination of Carbon Management (CMI, see Methods below) and related indices, a reference samples were obtained from a shrubland that had not been under cultivation of pasture for more than 15 years. Sampling of reference soil samples followed the same technique used in the selected farms.

Soil samples were taken at a depth of 0–30 cm by augering method along three diagonal line transects at 10 m interval resulting in 9 samples per each of the selected farm's horizontal interval (HI). HI is the farmable distance between two terraces as shown in Figure 3, and in each of selected farms had at least three. Collected samples were thoroughly mixed and a composite sample obtained for laboratory analysis. Three intact core samples were also collected for bulk density determination in each of the selected treatment.

Soil texture, pH water (2.5:1), total organic carbon (TOC), total nitrogen, and carbon-to-nitrogen ratio were determined across all the treatments. The soil particle sizes were determined through the hydrometer method (Bouyoucos, 1962), and TOC was determined by the wet ashing method (Walkley and Black, 1934). TOC was converted to soil organic matter by multiplying values by 1.72. Total soil nitrogen (TN) was determined through the Kjeldahl method (Okalebo et al., 2002).

Granulometric carbon fractions reflecting more stable vs. more labile properties of soil carbon based on particle size were determined using wet sieving (Cambardella and Elliott, 1992; Vieira et al., 2007). Each 20 g sample of air-dried soil was passed through a 2 mm sieve and mixed with 70 ml of sodium hexametaphosphate (HMP) solution. The mixtures were shaken for 16 h on an end-to-end shaker. Mixtures of soil were then passed through 250 μm sieve and then through a 53 μm sieve. Carbon in the fraction passing the 53 μm sieve was defined as non-labile, and that which had a size between 53 μm and 250 μm was defined as labile, in accordance with other research developing rapid assessments of such carbon fractions (Cambardella and Elliott, 1992; Vieira et al., 2007; van Wesemael et al., 2019; Ouédraogo et al., 2020). Both soil fractions were oven-dried at 50°C for 48 h. Fractions were ground using a pestle and mortar, and organic carbon in each fraction was determined by wet ashing method (Walkley and Black, 1934) as per Okalebo et al. (2002).

Several CMI related indicators (Vieira et al., 2007) were determined by using an undisturbed shrubland as a reference. This shrubland has been under enclosure for more than 15 years with relatively continuous inputs of organic litter from shrubs and forbs.

Data were curated in Excel 2016, and Shapiro-Wilk test of normality conducted on each parameter at p = 0.05. Data that were not normally distributed were subjected to log transformation upon which normality was achieved. Linear mixed models were used to compare the SOC fractions and CMI under different conservation measures. One way analysis of variance (ANOVA) was used to analyze SOC fractions, CMI, and weighted enrichment ratio mean comparisons across the selected time intervals of years after terrace construction in maize-bean intercrop. A Tukey's HSD post-hoc test was performed when significant differences between means was detected at p < 0.05. Data analyses were conducted using JMP Version 16.0. All results from transformed variables were back transformed by subjecting them to anti-log before reporting.

The soils for the selected management practices were found to have high sand content of more than 70%, hence classified as loamy Sand except for degraded rangeland which was classified as sand clay loam. TOC and TN were high in lands with soil management practices as compared to degraded rangeland without any management practice. pH of the soil was found to slightly acidic with degraded rangeland having the lowest (5.61), as illustrated in Table 1.

Vegetation cover and terracing had a significant effect on total organic carbon, labile carbon, non-labile carbon, and soil organic matter at p < 0.05 (Figure 4). Pasture significantly increased TOC. On average farms with pasture as a vegetation cover had 13.2 g C kg⁻¹. This was significantly higher than values obtained in degraded land and maize farms of 6.2 and 9.9 g C kg⁻¹, respectively (Figure 4i). Terracing significantly increased TOC by 3.1 g C kg⁻¹ when combined with a maize-bean cover crop. However, it did not significantly increase TOC when combined with pastures as illustrated by Figure 4i. Pastures and maize-bean intercrop were found to significantly improve labile organic fraction by 5.0 and 3.1 g C kg⁻¹, respectively over degraded lands. Similarly, to TOC, terracing significantly increased labile carbon only in croplands by 3.1 g C kg⁻¹ as indicated in Figure 4iii. Non-labile carbon was significantly higher in pastureland than in degraded lands by 2.2 g C kg⁻¹ as shown in Figure 4iv. Unlike in other fractions of carbon, terracing reduced significantly the non-labile carbon fractions in pasturelands by 0.9 g C kg⁻¹.

Terrace age had a significant effect on TOC, labile, and non-labile organic carbon fractions. For example, TOC and the labile organic carbon fraction significantly increased over 5 years in terraced maize-bean intercrop with increases of 4.6 and 3.0 g C kg⁻¹, respectively over those aged 1 year. For non-labile organic carbon fraction had a fluctuating trend, peaking at 3YAC and dropping after and henceforth with terraces, aged 1 and 2 years had significantly lower as illustrated by Figure 5.

The introduction of terraces in pastures and croplands significantly increased all test indices vs. non-terraced land and degraded land, particularly in the case of maize-bean intercropping Terracing in pasture and maize-bean intercrops increased LI by 0.48 and 0.31 over the practices without terraces, respectively. In contrast, the introduction of terraces led to strong significant increase in CSI in maize-bean intercropping but not in pastures, while CMI is reversed being only strongly significantly different for pastures. Combining terrace with vegetation practices were found to significantly increase CMI (p < 0.05) compared to both practices without terraces and the degraded land. For instance, terraces improved CMI four and about three times in pasture and maize-beans intercrops, respectively compared to vegetative practices without terraces. Leaving the land degraded without any intervention was found to significantly lower its CMI as illustrated by Table 2.

The lability index in croplands with terraces increased from 1YAC to 5YAC although the increase between 4 YAC and 5 YAC was significantly lower than the difference between 4 YAC and other test years. Similar to LI, the Carbon stock index (CSI) steadily increased from 1 YAC to 5 YAC. The carbon management index followed an increasing trend similar to that of LI. The highest CMI in croplands that had hosted terraces for 5 years was found to be more than triple what was recorded in the 1 year after terrace construction. Contrary to trends of all other indices, the weighted enrichment ratio depicted a dropping trend from 1 to 5 YAC. The former had the highest weighted enrichment ratio of 0.42 compared to the lowest 0.10 recorded in 5 YAC farmlands (Table 3).

High sand content of >70% was found in all the soils. This is an inherent characteristic and is the basis on which the soils were sub-classified as Arenic Lixisol (Schad et al., 2014). Contrary to both pasturelands and farmlands, clay content was high in severely degraded rangeland, indicating argic properties of sub-soil as reported in other studies (Schad et al., 2014; Blum et al., 2018). Our study established that erosion in the degraded rangeland was so severe that the A horizon had been partially or wholly washed away. Observed changes in the soil profile instigated by erosion explain the variation of pH among the vegetation cover practices and along YAC. Similar findings were reported by Hailu et al. (2012), who found that the use of soil and erosion management practices (SEMPs) in intervening erosion supervenes the rise of pH. According to the authors, Jones (2003) and Dessalegn et al. (2014), uncontrolled erosion causes the removal of basic cations like magnesium and calcium and leaves behind unbalanced acidic cations, resulting in a decline in pH. Low pH in a degraded system lowers the restoration rates and reduces carbon sequestration (Rengel, 2003; Naorem et al., 2022).

Variation of soil TOC, TN, and C:N across the vegetation cover practices, indicated by the findings were attributive to the benefits of the terraces as a conservation measure, particularly in highly disturbed systems like farmlands. Although, on average, the TOC and TN were low compared to results reported by Sainepo et al. (2018), who found an average of 22.6 g C kg⁻¹ in shrublands. TOC and TN also declined with YAC. The terraced farms with 5 YAC had a comparable high TOC than 1 YAC. The buildup of SOC is interplayed by several factors, including rainfall amount and land use management, as reported by Dessalegn et al. (2014).

Among the three vegetation cover practices, pasturelands were observed to have a stable distribution of organic carbon fractions. However, the terraced pasturelands, on average, recorded high labile SOC content compared to pasturelands without terraces. The findings are in agreement with the results of Kalambukattu et al. (2013) and Shi et al. (2020), who reported that less disturbed vegetation cover practices have less fluctuations in their SOC than any other system. Like shrublands, pasturelands are relatively less disturbed, and their organic matter source is relatively stable throughout the seasons. In addition, the soils of such systems are less eroded because of the trees, shrubs, grass, and litter, which form a natural mulch. This is consistent with, Jourgholami et al. (2019), who found that an increase in surface litter inversely correlates to surface runoff, hence, reducing erosion. In addition, permanent root systems in pastures hold soil in stable aggregates that tend to promote infiltration and reduce erosion (Gyssels et al., 2005).

Farmlands with terraces significantly surpassed those without terraces with more than twice the labile SOC fraction. Non-labile SOC fraction between farms with and without the two types of indigenous terraces had no significant difference. Labile SOC is the fraction of carbon considered critical to the agriculture system (Zhang et al., 2020), and is also the most affected by either SEMPs or soil erosion. Zhang et al. (2020) indicate further that farming at a slope greater than 10% exacerbates erosion, and increases organic carbon loss, particularly the labile fraction. The complementary effect of Fanya Juu and Fanya Chini significantly reduces soil erosion as reported by Mati (2010) and Saiz et al. (2016), which simultaneously reduce loss of SOC. However, the benefits and increment in SOC accrues over time, which is why we observed an increase in Labile SOC from 1 YAC to 5 YAC.

Terraced systems (pasturelands and farmlands) had higher carbon management index (CMI) than those without terraces. The difference in CMI between maize-bean intercrop systems with and without terrace was higher compared to difference found between pasturelands with and without terraces. CMI also increased with YAC, indicating that time is essential for significant increases or maintenance in carbon stocks as influenced by SEMPs, particularly in degraded systems. The results found by this study are highly comparable to past studies in the similar environments, as reported by Sainepo et al. (2018). Their study established a CMI range of 22–100 in different vegetation cover practices. Similarly, CMI values reported by Tang et al. (2018) from rice plantations indicated a range of 90–184. Weighted enrichment ratio decreased with an increase in YAC, indicating that as more organic matter accumulates on the land, carbon stability continues increasing, balancing between the inflow and outflows, as also reported by Chowdhury et al. (2021).