Fruit tree-based agroforestry as a nature-based solution for carbon sequestration in the vulnerable Eastern Himalayan ecosystem: plant–soil microbe feedbacks in the context of global change

1 Introduction

Terrestrial ecosystems store approximately 2,000–2,500 Pg of organic carbon in their soils, nearly three times the carbon currently held in the atmosphere (Batjes, 1996; Lal, 2004). Land use change, agricultural intensification, and climate warming are collectively destabilising this vast reservoir, converting terrestrial ecosystems from net carbon sinks to net sources at an accelerating rate (Le Quéré et al., 2018; IPCC, 2019). Montane tropical and subtropical ecosystems including the Himalayan arc are characterised by steep erosional gradients, high rainfall, fragile soil structures, and a long history of subsistence agriculture, these ecosystems exhibit particularly rapid carbon turnover rates and heightened sensitivity to disturbance (Ahirwal et al., 2021; Choudhury et al., 2022). Under the Representative Concentration Pathway (RCP) 8.5 scenario, the Eastern Himalayan region is projected to experience temperature increases of 2.5–4.5 °C and intensified monsoon variability by 2,100, further accelerating soil carbon mineralisation and erosional losses (Krishnan et al., 2020). The urgent need to identify land use strategies capable of simultaneously sequestering carbon, protecting soils, and supporting human wellbeing in these landscapes has driven growing scientific and policy interest in nature-based solutions (NbS) (IUCN, 2020; Cohen-Shacham et al., 2016).

NbS defined as actions that work with and enhance nature to simultaneously address societal challenges and improve human wellbeing while providing ecosystem services and benefits to biodiversity (IUCN, 2020) include ecosystem restoration, improved forest management, climate-smart agriculture, and agroforestry. Among these, agroforestry has been identified as one of the most promising NbS for carbon sequestration in tropical and subtropical regions, with global estimates of 12–228 Mg C ha⁻¹ in soils and biomass depending on system age, composition, and climate (Nair et al., 2010). In India, the National Agroforestry Policy (2014) explicitly recognises agroforestry as a carbon sink strategy under the country’s Nationally Determined Contributions (NDCs) to the Paris Agreement, targeting an additional carbon sink of 2.5–3.0 billion tonnes of CO² equivalent by 2030 [Ministry of Agriculture and Farmers Welfare (MoAFW), 2014]. However, a comprehensive, species-level integrated assessment of biomass carbon, soil carbon fractions, and microbial biomass carbon across agroforestry systems remains scarce in the Eastern Himalayan region.

The North Eastern Hill Region (NEHR) of India covers approximately 26.2 million hectares (8% of India’s total geographical area), of which more than 77% is classified as hilly terrain, and more than 60% falls under forest cover (FSI, 2019). Despite this apparent ecological richness, the region supports a population exceeding 50 million people, 86% of whom depend on traditional shifting cultivation (Jhum or slash-and-burn agriculture) as the primary livelihood strategy (Mishra and Francaviglia, 2021). The Jhum cycle, combined with the expansion of soil-exhausting cash crops particularly ginger (>54,000 ha) and turmeric (>24,000 ha) has resulted in severe soil degradation and carbon loss (Spices Board India, 2023). Annual soil loss under Jhum cultivation has been documented at 80.0–229.5 t ha⁻¹ yr.⁻¹ (Choudhury et al., 2022), driven by water erosion responsible for approximately 30% of total degraded land, alongside soil acidification, leaching of base cations, and decline in soil organic carbon (Singh et al., 2019). The region’s total degraded land area stands at approximately 4.60 million hectares equivalent to 18.2% of the total geographical area (Department of Land Resources, Ministry of Rural Development, Government of India, 2019).

Against this backdrop, fruit tree-based agroforestry systems (FBS) have been proposed as a sustainable land use alternative capable of reversing soil degradation trends while generating additional income for farmers (Do et al., 2020; Gupta et al., 2013). India is the world’s second-largest producer of fruits and vegetables (National Horticulture Board (NHB), 2020), and the NEHR contributes approximately 6.5% of national fruit production area. Perennial fruit trees including guava, peach, pear, plum, Assam lemon, and Khasi mandarin are well-adapted to the region’s subtropical climate and acidic soils, and their integration with seasonal crops offers multiple co-benefits: improved soil organic carbon through litter addition and root turnover; enhanced nutrient cycling; reduced erosion; improved soil moisture retention; and crop diversification for food and nutritional security (Sharma et al., 2017; Ramesh et al., 2019; Hussain et al., 2019). A 20-year study period, as examined here, provides a robust window to evaluate the long-term soil carbon storage trajectory of FBS relative to adjacent cropland under identical edaphic and climatic conditions. Fruit tree integration with turmeric and ginger intercropping has been reported to increase soil organic carbon by up to 30% within a few years (Ramesh et al., 2019; Hussain et al., 2019), highlighting the potential synergies between perennial tree carbon inputs and intercrop biomass in enhancing soil carbon sequestration in the region.

Plant–soil microbe feedbacks the bidirectional interactions by which plants shape soil microbial community composition and activity, which in turn influence plant growth and nutrient availability are now recognised as fundamental regulators of ecosystem carbon cycling and the long-term persistence of soil organic matter (Wardle et al., 2004; Bardgett and Wardle, 2010; van der Putten et al., 2013). These feedbacks operate through multiple pathways, including litter quality effects on decomposer community structure (Cornelissen et al., 2004), root exudate-mediated priming of soil organic matter decomposition (Fontaine et al., 2007; Kuzyakov, 2010), mycorrhizal mediation of soil carbon allocation (Averill et al., 2014), and microbial necromass contribution to stable mineral-associated organic matter (Cotrufo et al., 2013; Liang et al., 2019).

The rhizosphere, the volume of soil directly influenced by root activity, typically extending 0–5 mm from the root surface is the most biologically active compartment of terrestrial soils and the primary locus of plant–soil microbe feedbacks (Hinsinger et al., 2009; Blagodatskaya and Kuzyakov, 2013). Root exudates include low-molecular-weight compounds (sugars, amino acids, organic acids), high-molecular-weight mucilages, and border cells collectively constituting up to 40% of net photosynthate in some perennial woody species (Jones et al., 2009). These exudates create carbon hotspots that stimulate microbial activity 2–25 times above bulk soil levels (the ‘rhizosphere effect’), and can either accelerate decomposition of existing soil organic matter through priming (Kuzyakov, 2010) or promote the formation of stable organo-mineral complexes through microbial biomass turnover and aggregate formation (Cotrufo et al., 2013; Kallenbach et al., 2016). In perennial tree systems like FBS, the rhizosphere effect extends through deep root networks across the full soil profile a fundamentally different spatial and temporal pattern from annual crop root systems potentially enabling tree-based systems to sequester carbon at greater depths and over longer timescales (Cardinael et al., 2015; Jobbágy and Jackson, 2000).

Soil microbial biomass carbon (SMBC) is widely used as a sensitive, early-warning indicator of land use effects on soil biological fertility and carbon cycling dynamics (Jenkinson and Ladd, 1981; Anderson and Domsch, 1990). The ratio of SMBC to TOC (the microbial quotient, qMB) provides additional information on the efficiency of microbial carbon use relative to total organic carbon availability, and has been used as an indicator of ecosystem carbon sequestration potential and resilience (Insam and Domsch, 1988; Anderson and Domsch, 1990). Higher qMB values indicate efficient microbial use of available substrates, often associated with high-quality, low-C: N litter inputs and stabilised microbial communities characteristic of mature agroforestry systems (Matos et al., 2022; Debnath et al., 2015). The quality of organic inputs (particularly C: N ratio and lignin content) governs microbial enzyme induction patterns, decomposer community composition, and ultimately the proportion of input carbon that is retained in the soil versus emitted as CO₂ (Cornelissen et al., 2004; Wardle et al., 2004).

Despite this theoretical foundation, empirical studies simultaneously characterising the full carbon spectrum from biomass inputs through labile and stable soil carbon fractions and interpreting results explicitly through the lens of plant–soil microbe feedbacks and rhizosphere processes in subtropical montane fruit tree agroforestry are scarce. Most existing studies in the Eastern Himalayan region have focused on either biomass carbon or total organic carbon, without integrating microbial biomass dynamics, carbon fractionation, and global change implications into a coherent framework (Ansari et al., 2022; Ahirwal et al., 2022; Choudhury et al., 2021). This study fills that gap by providing a comprehensive integrated assessment of biomass carbon, soil carbon fractions, and microbial biomass carbon across six fruit tree species at ICAR-RCER-NEH. It is important to note that while plant–soil–microbe feedback mechanisms are discussed as interpretive frameworks, the underlying biological processes (e.g., enzyme activities, root exudate chemistry, microbial community composition) were not directly measured and are inferred from measured carbon and SMBC data in the light of existing literature.

The specific objectives of this study are: (i) to quantify and compare above- and belowground biomass carbon, seasonal leaf litter fall, and annual leaf carbon and nitrogen inputs across six fruit tree species; (ii) to characterise the complete soil carbon inventory including TC, TOC, TIC, SMBC, TN, and carbon fractions (C_VL, C_L, C_LL, C_NL) and their stocks across six soil depths to 1.0 m; (iii) to interpret the vertical distribution of SMBC, carbon fractions, and stratification ratios in the context of microbial biomass and activity, drawing on established plant–soil–microbe feedback literature as an interpretive framework; and (iv) to evaluate the potential of fruit tree-based agroforestry as a nature-based solution for carbon sequestration in the vulnerable Eastern Himalayan ecosystem, with implications for global change adaptation.

We hypothesise that (i) FBS will exhibit significantly higher SMBC, C_VL, and active carbon pools than cropland, driven by greater organic substrate inputs via litter and root exudation; (ii) the vertical distribution of SMBC and labile carbon will show more gradual depth-attenuation in FBS than in cropland, reflecting deeper rhizosphere influence of perennial fruit tree root networks; (iii) species-specific differences in litter quality (C: N ratio, leaf N content) and root architecture will generate distinct plant–soil microbe feedback signatures among FBS, resulting in species-level differences in soil carbon fractions and microbial activity; and (iv) the passive carbon pool (C_LL + C_NL) will be significantly elevated in FBS relative to cropland, conferring greater resistance to warming-induced decomposition under projected global change scenarios.

2 Materials and methods

2.1 Study area and climate

The study was conducted at the agroforestry section of the Division of Systems, Research, and Engineering, ICAR Research Complex for the North Eastern Hill Region (ICAR-RCER-NEH), Umiam, Meghalaya, India (25°41′39″N, 91°55′14″E; 1,000–1,030 m above mean sea level) (Figure 1). The ICAR-RCER-NEH campus lies within the Indo-Burma mega-biodiversity hotspot, recognised for exceptional levels of plant endemism and biological richness (Myers et al., 2000). The region experiences a humid subtropical climate (Köppen: Cwa) with a pronounced monsoon season. Mean annual precipitation exceeds 2000 mm, with approximately 70% concentrated in the July–September monsoon period (Choudhury et al., 2012). Mean daily maximum temperature peaks at 29.4 °C in April, while minimum temperature falls to 7.6 °C in December. Mean daily sunshine duration is 5.42 h. The area is bounded by Bangladesh to the south and west, and by the Indian state of Assam to the north and east (Choudhury et al., 2012).

The soil is classified as kaolinitic typic kandiudults (USDA Soil Taxonomy) and grouped under Acrisols in the FAO/WRB classification system (Yadav et al., 2021). These soils are inherently acidic (pH 4.44–6.04 across land uses and depths), reflecting sandstone-derived parent material and intense chemical weathering characteristic of high-rainfall tropical environments. Acidic reaction results primarily from leaching of base cations (Ca, Mg, K, Na), accumulation of kaolinite clay, and production of organic acids from decomposing organic matter (Patiram, 2007; Choudhury et al., 2021). Soil texture is sandy clay loam, with sand contributing 52.7–65.0%, clay 15.5–33.5%, and silt 10.6–25.2% across study depths and land uses (Table 1). The relatively low clay content combined with high rainfall creates conditions of moderate to high leachability, making soil organic carbon maintenance particularly dependent on continuous biological inputs from perennial vegetation cover.

Table 1. Vertical distribution of sand, silt, clay, soil bulk density (ρb), pH and moisture content across different land use systems in the surface and subsurface soils in the foothills of Eastern Himalaya (Meghalaya).

LUS^$: Land use system; DMRT: Duncan’s Multiple Range Test values at p < 0.05. Means in the column followed by different small letters in superscript (a–e) based on DMRT values represent statistical significant differences among the land uses at p < 0.05. Interaction effect of soil depth was significant at p < 0.0001.

2.2 Experiment details (land use evaluation)

In 2002, primary Pinus kesiya (chir pine) forest on sloppy upland terrain (>40% slope) was converted to terraced agroforestry plots at ICAR-RCER-NEH. Terraces of 1.2–1.5 m width were constructed, and six distinct fruit tree species were established: guava (Psidium guajava L.), peach (Prunus persica L.), Assam lemon (Citrus limon L.), pear (Pyrus communis L.), plum (Prunus salicina L.), and Khasi mandarin (Citrus reticulata Blanco). Trees were planted at a uniform spacing of 5 m × 5 m (400 trees ha⁻¹). Seasonal intercrops alternating turmeric (Curcuma longa L.) and ginger (Zingiber officinale Roscoe) annually were maintained within all FBS plots. Adjacent terrace strips without fruit trees were maintained as cropland control, receiving the same seasonal intercrops. Farm yard manure (FYM) was uniformly applied at 10 t ha⁻¹ across all land uses. By the time of this study (2021), all FBS were 20 years old, fully established, structurally homogeneous, and actively fruit-bearing.

The experimental unit was the land use system (plot), with three spatially independent quadrats per land use serving as replicate observations. Each quadrat was located at least 10 m from adjacent quadrats to ensure spatial independence and minimise pseudo-replication. Soil depths were treated as repeated measures nested within quadrats, not as independent replicates. Ten representative trees (based on their health status, normal growth and similarity to average stand condition) per FBS were selected by stratified random sampling across the full plot area to ensure representative coverage of within-plot structural variability. Soil properties at establishment were not significantly different among land uses, as confirmed by the uniform experimental soil management history. The soil bulk density ranged 1.07–1.47 gcm⁻³among FBSs while the cropland was marginal higher in soil bulk density (1.19–1.53 gcm⁻³) (Table 1). Soil texture was classified as sandy clay loam where sand and clay contributed 52.7–65% and 15.5–33.5%, respectively, while the remaining portion was contributed by the silt (Table 1). The detailed information of all the land use systems is given in the Supplementary Table 1.

2.3 Soil sampling protocol

The soil profile of the region varies due to multiple attributes including valley, parent materials, vegetation, altitude, slope, and aspects. A total of 126 composite soil samples were collected up to 1.0-meter soil profile depth to evaluate the lateral variations in soil carbon and other soil properties. The samples were collected from 6 fruit trees-based agroforestry systems and 1 control site (cropland adjoining the fruit trees-based agroforestry systems), with a total of 3 quadrats randomly laid down in each system and soil samples collected from all the corners and middle of the quadrats. For each soil profile depth, soil samples were properly mixed from the center and each corner of the quadrat (L1 + L1 + L1 + L1 + L1 = L1) resulting in one composite sample to represent each depth. The first four soil samples were collected with a 15 cm interval up to a depth of 60 cm, while the other 2 were collected from 60–80 cm and 80–100 cm soil profile depth. Each soil profile layer was labeled according to the layer distribution, where L1 denotes the surface layer (0-15 cm), L2 (15–30 cm), L3 (30–45 cm), L4 (45–60 cm), L5 (60–80 cm), and L6 (80–100 cm), respectively. The soil samples were collected during the pre-monsoon period in the month of May, 2021 and standard protocols were followed during collection to ensure accuracy and consistency. The collection of composite soil samples up to a depth of 1.0 meter allowed for a comprehensive assessment of lateral variations in soil properties, including soil carbon and other important soil attributes. This provides valuable insights into the spatial distribution of soil properties of different land uses and can help in formulating appropriate land management strategies for sustainable soil use and conservation.

2.4 Above and below ground tree biomass estimation

A total of 10 representative trees from each agroforestry system were selected to take the measurements of height and diameter at breast height (DBH) with the help of Haglöf Vertex Laser clinometer/hypsometer (for height) and tree calliper (for DBH and collar diameter) to calculate the biomass of each tree. The biomass was calculated through previously developed allometric equations for respective tree species. To represent biomass on hectare basis, the average of the representative trees’ biomass was multiplied by the corresponding tree density. The below-ground biomass and carbon were calculated by the conversion factors 0.26 (for below ground biomass) (Cairns et al., 1997) and 0.40 (for carbon content) (IPCC, 2006) respectively.

The following biomass equations were used to estimate the above-ground biomass.

Table 2.

2.5 Leaf litter bag experiment and leaf carbon

The leaf fall of different tree species was estimated by randomly placing 6 quadrates of 1 m × 1 m quadrate in each agroforestry system. The tree leaves that fell into all the quadrates for each system were then separately collected quarterly round the year, dried, weighed, and expressed as on hectare basis quarterly and annually. The leaf total carbon percentage (%) was estimated following the standard protocol of TOC analyser (Model: Elementar TOC Analyser, Vario TOC Select).

2.6 Laboratory analysis of soil carbon, carbon pools, and other soil properties

The soil bulk density (pb) of oven dried (104 ± 1 °C) soil sample were estimated by using the core sampler (5.15 cm height and 4.7 cm diameter). The following formula was used (Equation 1):

Bulk density ( gcm − 3 ) = Mass of dry soil ( g ) / Volume of dry soil ( cm − 3 ) (1)

After removing debris, stones, roots, and other foreign materials, half of the pre-processed samples were immediately stored in sealed plastic bags, while the other half were subjected to air drying, crushing, and sieving (using a 2.0 mm size) to analyze various soil properties. Soil pH was measured in a 1:2.5 soil:water suspension using a calibrated digital pH meter. Particle size distribution was determined by the hydrometer method. Total carbon (TC) and total inorganic carbon (TIC) were measured using a Vario TOC Select Elementar analyser with a TIC module; total organic carbon (TOC) was calculated as TC − TIC. Carbon fractionation was performed using the modified Walkley-Black oxidation method (Chan et al., 2001) with three concentrations of H₂SO₄ (12 N, 18 N, and 24 N). Carbon oxidised by 12 N H₂SO₄ was classified as very labile carbon (C_VL); the difference between 18 N and 12 N was labile carbon (C_L); between 24 N and 18 N was less labile carbon (C_LL); and the difference between TOC and 24 N oxidation was non-labile carbon (C_NL). Active pool = C_VL + C_L; passive pool = C_LL + C_NL. Total nitrogen (TN) was determined by the Kjeldahl digestion–distillation–titration method (AOAC, 1995).

Soil microbial biomass carbon (SMBC) was determined using the chloroform fumigation–extraction (CFE) method (Vance et al., 1987). Briefly, 25 g field-moist soil was fumigated with ethanol-free chloroform under vacuum for 24 h in the dark to lyse microbial cells. Fumigated and non-fumigated soils were extracted with 0.5 M K₂SO₄ (1:4 soil:extractant ratio) and filtered through Whatman 42 paper. Organic carbon in extracts was measured by the TOC analyser. SMBC was calculated as (TOC of fumigated − TOC of non-fumigated extract) / kEC, where kEC = 0.45 (Wu et al., 1990). The microbial quotient (qMB = SMBC/TOC × 100%) was computed for each land use and depth as a proxy for microbial community efficiency and substrate utilisation (Anderson and Domsch, 1990; Insam and Domsch, 1988). The C_VL fraction was used as a proxy for microbially accessible substrate availability operationalising the concept of rhizosphere-derived labile inputs. Stratification ratios (SR) of SMBC were computed as SR = SMBC at L1 / SMBC at L(n), providing a depth-dependent index of how steeply microbial activity declines from surface to subsurface soils an indirect measure of the vertical extent of rhizosphere influence.

2.7 Estimation of soil profile C stock

The soil carbon and Kjeldahl total nitrogen stocks (TC, TIC, TOC, TN, SMBC and soil carbon pools) for each soil profile layer, soil profile average (PWA) and soil profile sum (up to 1.0 m) were estimated through the measured corresponding concentrations (0.15 mm sieved), bulk density (ρb) values, coarse fragments (CF: coarse roots/ gravels/ rocks) larger than the 2.0 mm diameter (fraction by mass, %/100) and from the respective thickness of layers following the (Ahirwal et al., 2022) of Ellert et al. (2001) (Equations 2–4).

C Stock ( m g ha − 1 ) = [ ρb ( gcm − 3 ) x C ( % ) x ( 1 − CF ) x soil depth ( m ) x 104 m 2 ha − 1 / 100 ] (2)

Soil profile ( 0.0 − 1.0 m ) weighted average C stock = Soil profile ( 0.0 − 1.0 m ) C stock 100 cm ( total profile depth ) . (3)

Soil profile ( 0.0 – 1.0 m depth ) C stock = ∑ L = 1 L = 6 C stock for each soil depth (4)

2.8 Stratification ratio (SR)

The Stratification Ratio (SR) measures the value of a soil parameter in the surface soils divided by the value in the subsurface soils, as described by Qu et al. (2020). For soil quality assessment, the SR of Total Organic Carbon (TOC) is often used as an indicator, as an increase in SR is associated with greater rates and amounts of TOC sequestration (Franzluebbers, 2002). In this study, the SRs of various soil carbon inventories were calculated between the surface layer (L1: 0–0.15 m) and five subsurface layers: SR1 (L1: 0–0.15/L2: 0.15–0.30 m), SR2 (L1: 0–0.15/L3: 0.30–0.45 m), SR3 (L1: 0–0.15/L4: 0.45–0.60 m), SR4 (L1: 0–0.15/L5: 0.60–0.80 m), and SR5 (L1: 0–0.15/L6: 0.80–1.0 m, the bottom-most layer).

2.9 Contribution of different carbon inventories

We have calculated the contribution percentage (ratio) for all the soil parameters, i.e., the ratio (contribution) of total inorganic carbon to total carbon (TIC/TC), total organic carbon to total carbon (TOC/TC), total nitrogen to total carbon (TN/TC or C: N ratio), total nitrogen to total organic carbon (TN/TOC), soil microbial biomass carbon to total organic carbon (SMBC/TOC), very labile carbon to total organic carbon (C_VL/TOC), active pool to total organic carbon (AP/TOC), and passive pool to total organic carbon (PP/TOC) under different land use systems.

2.10 Statistical analysis

All the collected data of different land uses were checked for normality of distribution, and the design of sampling across land use systems led to a simple balanced two-way analysis of variance (ANOVA) with seven land uses (6 fruit trees-based agroforestry systems and 1 cropland) and six depths (L1-L6) as factors/ treatments. All the soil carbon inventories and soil microbial properties including total nitrogen (TN, %) across the land use systems and soil depths were compared by two-way ANOVAs followed by Duncan’s Multiple Range Test (DMRT) at Post Hoc stage with 5% probability level. In addition to two-way ANOVA, Pearson correlation analysis was performed among all soil carbon variables, SMBC, and TN to identify bivariate relationships (n = 126). Principal component analysis (PCA) was applied to the full matrix of soil carbon fractions (TC, TOC, TIC, SMBC, TN, C_VL, C_L, C_LL, C_NL, active pool, passive pool) to explore multivariate structure among land uses.

3 Results

3.1 Basic soil physical and chemical properties across land uses

All soils were acidic across the full 1.0 m profile, with pH values ranging from 4.44 to 6.04 (Table 1). Cropland soils were consistently and significantly (p < 0.05) more acidic than FBS soils at every depth (cropland pH: 4.44–4.94; FBS pH: 5.06–6.04). The least acidic soils were under Khasi mandarin (pH 0.90 units higher than cropland in L1), while all FBS were comparable in surface (L1) acidity. Soil pH increased irregularly in subsurface layers across all land uses, with cropland maintaining the lowest pH at every depth: a pattern reflecting differential base cation inputs through litter and root activity. Soil bulk density (ρb) ranged 1.07–1.53 gcm⁻³ across all land uses and depths (Table 1), with cropland consistently recording higher ρb than FBS (cropland surface L1: 1.19 Mg m⁻³; FBS: 1.07–1.18 gcm⁻³). The higher ρb under cropland reflects compaction from repeated tillage and reduced organic matter inputs. Soil moisture was significantly (p < 0.05) higher in FBS (+2.9 to +5.8%) relative to cropland (27.7%) across the profile, with peach and mandarin FBS recording the highest surface moisture (33.5%) (Table 1). Soil texture was sandy clay loam across all land uses, with FBS soils consistently showing higher clay content than cropland soils in both surface and subsurface layers, consistent with reduced erosional loss and improved aggregate stability under perennial canopy cover.

3.2 Soil carbon fractions and microbial biomass across land uses

3.2.1 Total carbon, organic carbon, and inorganic carbon

The profile-weighted average (PWA) total carbon (TC) ranged from 1.53% (cropland) to 2.15% (peach FBS), with FBS soils 5–29% higher than cropland (Table 2). Peach FBS recorded the highest TC (2.15%; +28.8% above cropland), followed by Khasi mandarin (2.12%; +27.8%). Pear-based FBS was lowest among FBS (1.61%; +5.0%). Mean TC across all land uses was 1.83%. TOC contributed 88.9–91.0% of TC across all land uses, with the remaining 9.0–11.1% accounted for by TIC indicating that the dominant carbon form is organic, reflecting strong biological carbon inputs from perennial vegetation (Table 2). Peach FBS soils recorded TOC 29.5% above cropland (TOC: 1.36%), while pear was least among FBS (+6.8%). TIC concentration ranged 0.15–0.22% (PWA), with peach and mandarin recording the highest values (+22.7% above cropland). Mean TOC across all land uses was 1.65% (Table 2).

Table 2. Effect of diverse land uses (fruit-based systems and cropland) on soil profile (PWA: Weighted average of 1.0 m depth) carbon concentration (%, g per 100 g).

PWA, Profile weighted average; LUS^$, Land use system; TC, Total Carbon; TIC, Total inorganic carbon; TOC, Total organic carbon; SMBC, soil microbial biomass carbon; TN, Total nitrogen; C_VL, C_L, C_LL, and C_NL stand for very labile, labile, less labile, and non-labile organic carbon pools; DMRT, Duncan’s Multiple Range Test values at p < 0.05; Means in the column followed by different small letters in superscript (a–e) based on DMRT values represent statistical significant differences among the land uses at p < 0.0001.

3.2.2 Soil microbial biomass carbon and microbial quotient

SMBC (PWA) ranged from 352.3 μg g⁻¹ (cropland) to 551.1 μg g⁻¹ (peach FBS), a 36.1% difference (Table 2). All FBS soils were significantly (p < 0.05) higher in SMBC than cropland, with enrichment ranging from +12.4% (pear) to +36.1% (peach). Mean SMBC across all land uses was 459.7 μg g⁻¹, which was 23.4% above cropland. Pearson correlation analysis confirmed significant positive relationships between SMBC and TOC (r = 0.87, p < 0.001, 95% CI: 0.82–0.91), SMBC and C_VL (r = 0.79, p < 0.001, 95% CI: 0.73–0.84), and SMBC and TN (r = 0.81, p < 0.001, 95% CI: 0.75–0.86) across all land uses and depths (n = 126). These correlations are illustrated in the PCA biplot (Figure 2a) and Pearson correlation matrix (Figure 2b). The SMBC: TOC ratio (microbial quotient, qMB) ranged from 1.8% (Khasi mandarin) to 3.0% (guava) across land uses (Table 2), indicating that 2–3% of the total organic carbon pool is held in living microbial biomass. This interspecific variation in qMB suggests that different fruit tree species stimulate distinct microbial biomass-to-substrate ratios, reflecting differences in litter quality, root exudate composition, and substrate availability. Total nitrogen (TN, PWA) was 16.7–37.5% higher in FBS than cropland (0.10%) (Table 2) with peach and Khasi mandarin recording the highest TN (0.16% each). Mean TN across all land uses was 0.14% (Table 2).

3.2.3 Carbon fractions: labile and stable pools

C_VL (PWA) was higher in all FBS compared to cropland (0.46%), with a reduction of 13.2–36.1% observed in cropland relative to FBS. Peach FBS recorded the highest C_VL (0.72%), while pear was lowest among FBS (0.53%) (Table 2). Labile carbon (C_L) showed wide variation (CV < 22%) among land uses, ranging −25 to +35.5% relative to cropland (0.20%), reflecting the dynamic and transient nature of this fraction. The active pool (C_VL + C_L) was 13.2–25.8% higher in all FBS than cropland (0.66%). Less labile carbon (C_LL) was 22.2–50% higher in FBS relative to cropland (0.21%), reflecting accumulation of physically and chemically protected carbon (Table 2). The passive pool (C_LL + C_NL) was significantly (p < 0.05) higher in all FBS compared to cropland, with the exception of pear, which was statistically equivalent to cropland in passive pool (Table 2). The contribution of C_VL, C_L, C_LL, and C_NL to TOC ranged 30.5–39.8%, 8.8–16.4%, 15.4–23.0%, and 27.3–37.4%, respectively (Table 3).

Table 3. Effect of land use on different fractions of carbon to TC and TOC concentration.

LUS^$, Land use system; TC, Total Carbon; TIC, Total inorganic carbon; TOC, Total organic carbon; SMBC, soil microbial biomass carbon; TN, Total nitrogen; C_VL, C_L, C_LL, and C_NL stand for very labile, labile, less labile, and non-labile organic carbon pools.

3.3 Soil carbon stocks across land uses

The PWA TC stock across FBS ranged from 32.83 (pear) to 45.54 Mg ha⁻¹ (Khasi mandarin), compared to 32.25 Mg ha⁻¹ in cropland (Table 4). Khasi mandarin recorded the highest PWA TC stock (+29.2% above cropland), followed by peach (+36.7%). TOC stock contributed 88.8–90.5% of TC stock (Table 4). FBS were higher in SMBC stock by 14.9–35.1% and in TN stock by 16.9–43.1% compared to cropland. The active pool stock (PWA) was +11.3 to +25.9% higher in FBS than cropland (13.90 Mg ha⁻¹), and the passive pool was +11.1 to +37.1% higher in all FBS except pear (which was statistically equivalent to cropland at 13.84 Mg ha⁻¹) (Table 4). For profile sum stocks (to 1.0 m total depth, Table 5), the pattern was consistent: TC ranged 200.5 (cropland) to 279.0 (Khasi mandarin) Mg ha⁻¹. FBS soils were higher in TC by 5–28%, in TOC by 2–29%, in SMBC by 13.5–35%, in TN by 16–38%, and in C_VL by 8.5–33.7% compared to cropland (Table 5). The passive pool (C_LL + C_NL) was significantly higher in all FBS except pear, which recorded 6.4% lower passive pool stock than cropland, consistent with the pattern observed in the PWA analysis (Table 5).

Table 4. Effect of diverse land uses (Fruit based systems and cropland) on soil profile (PWA: weighted average of 1.0 m depth) carbon stock (Mg C ha⁻¹).

PWA, Profile weighted average; LUS^$, Land use system; TC, Total Carbon; TIC, Total inorganic carbon; TOC, Total organic carbon; SMBC, soil microbial biomass carbon; TN, Total nitrogen; C_VL, C_L, C_LL, and C_NL stand for very labile, labile, less labile, and non-labile organic carbon pools; DMRT, Duncan’s Multiple Range Test values at p < 0.05; Means in the column followed by different small letters in superscript (a–d) based on DMRT values represent statistical significant differences among the land uses at p < 0.0001.

Table 5. Effect of diverse land uses (fruit based systems and cropland) on soil profile (sum of 1.0 m depth) carbon stock.

LUS^$, Land use system; TC, Total Carbon; TIC, Total inorganic carbon; TOC, Total organic carbon; SMBC, soil microbial biomass carbon; TN, Total nitrogen; C_VL, C_L, C_LL, and C_NL stand for very labile, labile, less labile, and non-labile organic carbon pools; DMRT, Duncan’s Multiple Range Test values at p < 0.05; Means in the column followed by different small letters in superscript (a–d) based on DMRT values represent statistical significant differences among the land uses at p < 0.0001.

Pairwise mean comparison plots of soil profile carbon stocks (0–1.0 m) revealed distinct differences among the FBS for TC, TIC, TOC and SOC (Figure 3). In general, Khasi mandarin and peach-based systems occupied the upper end of the carbon stock gradient, whereas cropland, pear and guava were associated with lower carbon stocks. Significant pairwise differences were most frequent for TC, TOC and SOC, particularly between the high-carbon fruit-based systems, while TIC exhibited comparatively fewer significant contrasts. The SOC and TOC plots showed the clearest separation among systems, indicating that fruit tree-based land uses primarily enhanced the organic carbon fraction of the soil profile. The frequent non-significant comparisons among some fruit-based systems, especially plum, lemon and guava, suggest that although these systems differed numerically in carbon storage, their overall carbon accumulation potential was broadly comparable.

To contextualise these stocks in terms of carbon dynamics, approximate mean annual carbon storage rates were estimated by dividing profile sum TC stocks by the 20-year system age. Estimated mean annual storage rates ranged from 10.0 Mg ha⁻¹ yr.⁻¹ (cropland) to 13.95 Mg ha⁻¹ yr.⁻¹ (Khasi mandarin) and 13.72 Mg ha⁻¹ yr.⁻¹ (peach), representing a net additional annual storage of 0.50–3.95 Mg C ha⁻¹ yr.⁻¹ in FBS relative to cropland over the 20-year period. These represent time-averaged storage rates rather than instantaneous sequestration rates, and do not capture non-linear accumulation dynamics.

3.4 Vertical distribution and stratification ratios

TC and TOC declined with depth in all land uses, from maxima of 2.64–3.58% (L1) to minima at L6, a reduction of 8.9–79.2% across the profile (Figures 4a,b). Total inorganic carbon showed a comparatively irregular depth-wise distribution across the different land-use systems, unlike the more consistent decline observed for TC and TOC. TIC was generally higher in the surface layer, with values roughly in the range of 0.24–0.32%, but declined at about 20–40 cm depth in most land uses (Figure 4c). FBS maintained 9.6–26.3% higher TC than cropland in L1, with peach dominating in L1–L4 and Khasi mandarin in L5–L6. SMBC declined consistently with depth (Figure 4d), from surface maxima of 741.48–938.44 μg g⁻¹ (L1) to substantially lower values in subsoils. Peach FBS surface SMBC was 21% higher than cropland at L1, and maintained the highest SMBC throughout the profile to L6. Cropland consistently recorded the lowest SMBC at every depth, with the steepest decline from surface to subsurface layers. Total nitrogen exhibited a clear depth-dependent pattern, although the decline was not uniformly linear. TN concentrations were comparatively high in the surface soil, approximately 0.17–0.21%, reflecting greater accumulation of organic matter and biologically derived nitrogen near the soil surface. Concentrations declined somewhat at 20–40 cm depth, followed by a modest increase in several land uses around 50–55 cm, where TN reached approximately 0.17–0.22%. Below this layer, however, TN decreased sharply, with values generally falling below 0.10% at 70 cm and remaining low at 90–95 cm (Figure 4e). SOC exhibited the clearest and most consistent vertical stratification. SOC was greatest in the upper soil layer and progressively decreased with increasing soil depth across all land-use systems (Figure 4f).

Carbon pool vertical distributions showed C_VL declining constantly from L1 to L6, while C_L, C_LL, and C_NL oscillated irregularly across depths (Figure 5). The active pool tracked C_VL (declining with depth), while the passive pool showed a complex, non-monotonic depth distribution likely reflecting the interplay between physical protection at shallow depths and chemical stabilisation through organo-mineral associations in deeper layers. The stratification ratio of total carbon showed a clear increasing trend with increasing soil depth across all land-use systems. The SR was relatively low when the surface layer was compared with L2 (0–0.15 vs. 0.15–0.30 m), generally around 1.1–1.5, but increased progressively to approximately 3–5.5 when the surface layer was compared with the deepest layer (L6: 0.80–1.0 m) (Figure 6a). The SR of TOC was similar to TC, with SR increasing consistently as the comparison extended from L2 towards L6. The TOC stratification ratio was close to unity in the upper soil comparison but increased markedly at greater depths, reaching approximately 3–6 in the L1/L6 comparison (Figure 6b). SMBC stratification ratios (SR1–SR5) revealed a steeper gradient in cropland than in FBS (Figure 6c) confirming that the reduction in microbial biomass with depth was significantly more abrupt in cropland soils. This constitutes indirect evidence that the rhizosphere influence of fruit tree root networks sustains microbial activity at depth. TN stratification ratios (Figure 6d) showed an atypical plateau or reduction in SR3 under several FBS, coinciding with elevated TN in mid-profile layers (L3–L4), likely reflecting active N mineralisation from root turnover at those depths. The SR of active-pooldisplayed a pronounced depth-dependent increase in stratification. SR values were relatively low at L1/L2 and L1/L3, but increased considerably at L1/L5 and particularly at L1/L6. At the deepest comparison, values reached approximately 3.5–5.5, with the guava system showing the greatest stratification, followed by lemon and cropland, whereas mandarin and some other systems showed comparatively lower values (Figure 6e). The passive carbon pool exhibited the most pronounced stratification among the four carbon indicators considered. Although SR values were relatively close to unity in the upper soil comparison, they increased sharply with increasing depth, reaching approximately 3–6.5 at L1/L6 (Figure 6f).

Principal component analysis (PCA) of the full soil carbon matrix (TC, TOC, TIC, SMBC, TN, C_VL, C_L, C_LL, C_NL, active pool, passive pool) yielded two principal components with eigenvalues >1.0, together explaining 71.3% of total variance. PC1 (48.6% variance) was positively loaded by SMBC, TOC, C_VL, and TN, and negatively by C_NL, effectively separating land uses along a biological activity gradient. PC2 (22.7% variance) contrasted the active pool (C_VL + C_L) against the passive pool (C_LL + C_NL), separating surface from subsurface layers. Peach and Khasi mandarin FBS were separated from cropland in positive PC1 space, confirming their superior biological carbon cycling performance (Figure 2a).

3.5 Biomass, leaf litter, and leaf carbon/nitrogen

Aboveground biomass (AGB) ranged from 22.68 Mg ha⁻¹ (plum) to 31.98 Mg ha⁻¹ (guava), with guava (+29.1%) and peach (+28.0%) recording the highest AGB (Figure 7a). Belowground biomass (BGB) mirrored this species ranking, ranging 5.9–8.31 Mg ha⁻¹ (Figure 7b). Guava contributed the highest biomass carbon: 12.79 Mg ha⁻¹ (aboveground) and 3.32 Mg ha⁻¹ (belowground). Annual leaf fall ranged 0.567 (plum) to 0.810 Mg ha⁻¹ (pear), with peak deposition in November–January (35.8–43.1% of annual total) (Figure 7e). Annual leaf carbon input ranged 0.256 (plum) to 0.357 Mg ha⁻¹ yr.⁻¹ (guava) (Figure 7f). Total leaf carbon content ranged 40.4% (pear) to 46.5% (peach), while leaf nitrogen ranged 2.0% (guava) to 3.8% (mandarin). The C: N ratio of leaves ranged from 11.64 (lemon) to 22.6 (guava), with higher leaf N and lower C: N ratios in mandarin and lemon positively correlated with SMBC and soil TN, supporting the hypothesis of litter quality-mediated plant–soil microbe feedbacks (Figures 7c,d).

4 Discussion

4.1 Fruit tree agroforestry as an effective NbS: ecosystem-level evidence

Our results provide compelling evidence that 20-year-old fruit tree-based agroforestry systems in the Eastern Himalayan region function as highly effective nature-based solutions for carbon storage. Total carbon stocks under Khasi mandarin and peach FBS (279.0 and 274.3 Mg ha⁻¹, respectively, to 1.0 m depth) are among the highest reported for tropical and subtropical agroforestry systems in India (Naik et al., 2017; Ansari et al., 2022), substantially exceeding those of adjacent cropland (200.5 Mg ha⁻¹; representing a net difference in carbon storage of approximately 74–79 Mg C ha⁻¹ over 20 years). These values compare favourably with carbon stocks reported for tropical agroforestry systems globally (30–358 Mg C ha⁻¹; Nair et al., 2010; Singh et al., 2024), and align with India’s NDC commitments to expand carbon sinks through agroforestry.

The NbS value of FBS in this context extends well beyond soil carbon sequestration. FBS significantly enhanced biomass carbon and supported higher intercrop biomass (turmeric +10.5 to +37%; ginger +20.6 to +34.1% above cropland), reflecting improved microclimate conditions and soil fertility under the fruit tree canopy. Simultaneously, FBS showed improved soil pH, lower bulk density, and higher moisture content co-benefits that promote soil biological activity, nutrient cycling, and aggregate stability (Lenka et al., 2012; Sahoo et al., 2019; Yadav et al., 2021). This suite of outcomes soil carbon, biomass, food production, soil quality improvement is the hallmark of an effective, multifunctional NbS consistent with IUCN (2020) criteria, and positions FBS as a priority land use for climate-smart agriculture policy in the Eastern Himalayan region.

Our TOC values (PWA 1.53–2.15%) are consistent with Ansari et al. (2022), who reported 1.78–2.50% TOC under agroforestry systems in the foothills of Eastern Himalaya, and with Thangavel et al. (2018), who reported higher TOC in FBS soils at 0.75 m depth under similar agroclimatic conditions. Naik et al. (2017), working at lower altitude (650 m) and rainfall (1,400 mm), reported substantially lower TOC stocks (53.3–62.47 Mg ha⁻¹ to 0.60 m) than our values (178.3–250.3 Mg ha⁻¹ to 1.0 m), consistent with the strong positive effect of higher rainfall on organic matter inputs in our study region.

4.2 Microbial biomass and activity: the biological engine of carbon sequestration

The significantly higher SMBC under all FBS relative to cropland (+12.4 to +36.1%) underscores the critical role of the soil microbial community as the biological engine of carbon sequestration in these systems. SMBC is not only a sensitive indicator of land use effects on soil biological quality but also an active agent in organic matter transformation, nutrient mineralisation, and carbon stabilisation (Jenkinson and Ladd, 1981; Anderson and Domsch, 1990). Our SMBC values (352.3–551.1 μg g⁻¹) fall within the range reported by Meetei et al. (2020) for dominant agroforestry systems in NEHR (609–796 μg g⁻¹ surface layer) and Matos et al. (2022) for agroforestry in southeastern Brazil (429–727 μg g⁻¹), and are broadly consistent with western Himalayan fruit orchards (825–1,000 μg g⁻¹ surface; Debnath et al., 2015). The somewhat lower profile-weighted average values in our study reflect the integration across the full 1.0 m depth, where SMBC declines substantially in subsoils.

While microbial community composition and enzyme activities were not directly measured in this study, three mechanisms inferred from the SMBC and carbon fraction data in combination with published literature are proposed to explain the higher SMBC in FBS. First, the quantity and diversity of organic substrate inputs are markedly greater: leaf litter fall, fine root turnover, fruit and twig deposition, and intercrop residues collectively provide a continuous, multi-source carbon supply sustaining diverse microbial guilds across trophic levels (Lenka et al., 2012; Ahirwal et al., 2022). Second, the physical environment for microbial activity is improved: higher soil moisture, lower soil temperature fluctuations under canopy shade, lower bulk density (favouring aeration and optimal water-filled pore space), and higher clay content (protecting microbial biomass from desiccation) collectively create a more stable and favourable microhabitat (Six et al., 2006; Matos et al., 2022). Third, reduced soil disturbance under perennial FBS compared to annual tillage in cropland preserves microbial community structure, reduces disruption of aggregate networks, and maintains the spatial organisation of microbial niches (Bossio and Scow, 1998; Guo and Gifford, 2002). These three mechanisms act synergistically under the FBS canopy, explaining the consistent and significant SMBC advantage across all six fruit tree species.

The species-specific variation in SMBC with peach highest and pear lowest among FBS is particularly informative. Peach leaves recorded the highest leaf carbon content (46.5%) and moderate leaf nitrogen (2.6%), yielding a C: N ratio of approximately 17.8. Thangavel et al. (2018) reported that peach produces the highest fine root biomass among FBS species in the region, providing the highest belowground substrate input to the rhizosphere. Pear, despite high leaf fall, had the lowest leaf carbon and moderate C: N ratio, suggesting that litter quality rather than quantity governs SMBC responses. This is consistent with the Microbial Efficiency-Matrix Stabililization (MEMS) framework (Cotrufo et al., 2013), which proposes that high-quality, low-C: N litter promotes microbial biomass accumulation and subsequent formation of stable microbial-derived carbon, rather than simply accelerating decomposition and carbon loss. The mirobial quotient (qMB) variation among species (1.8–3.0%) further confirms that carbon use efficiency differs among FBS, with implications for the proportion of incoming carbon substrates retained in stable pools versus mineralised.

The depth distribution of SMBC and its stratification ratios are particularly informative about the spatial extent of microbial activity under FBS. In cropland soils, the steeper SR gradient reflects dependence on surface litter inputs alone, with essentially no biological subsidy from root networks below 0.30 m (Jobbágy and Jackson, 2000; Lorenz and Lal, 2005). Under FBS, the more gradual SMBC depth-attenuation reflects continuous biological activity sustained by fine root exudation, mycorrhizal networks, and root turnover throughout the 1.0 m profile. This is functionally significant: carbon sequestered at depth (>0.30 m) is more stable and persistent than surface carbon, being less susceptible to warming-induced mineralisation, erosional loss, and agricultural disturbance (Jobbágy and Jackson, 2000; Lorenz and Lal, 2005). The deeper maintenance of SMBC under FBS thus has direct implications for the long-term stability of sequestered carbon and the resilience of FBS carbon stocks under future global change.

4.3 Rhizosphere processes: root exudation, priming, and carbon stabilisation

The rhizosphere is the most biogeochemically dynamic compartment of terrestrial soils, where the convergence of root activity, microbial metabolism, and mineral surface chemistry determines the fate of incoming carbon whether mineralised to CO₂, transformed into microbial biomass, or stabilised into mineral-associated organic matter (Hinsinger et al., 2009; Blagodatskaya and Kuzyakov, 2013). Our data provide multiple lines of evidence for active rhizosphere processes in FBS soils operating throughout the soil profile.

The systematic enrichment of C_VL in FBS surfaces soils (+8.5 to 33.7% above cropland) provides the strongest indirect evidence of enhanced rhizosphere substrate inputs. CVL (oxidisable carbon by 12 N H₂SO₄) represents the most recently deposited and biologically accessible organic matter, primarily comprising simple carbohydrates, proteins and other low molecular weight compounds derived from fresh litter decomposition and root turnover (Chan et al., 2001). Its systematic enrichment under FBS is consistent with continuous inputs of fresh carbon substrates that support microbial metabolism and reflects patterns commonly associated with rhizosphere priming in perennial tree systems, although root exudates and priming processes were not directly measured in this study (Kuzyakov, 2010; Fontaine et al., 2007). The concurrent enrichment of C_VL with significantly higher SMBC across FBS further indicates increased microbial access to readily decomposable substrates suggesting enhanced biological carbon cycling under perennial fruit trees. These observations are therefore consistent with rather than direct proof of greater rhizosphere derived organic matter inputs and active plant–soil-microbe interactions in FBS.

Critically, the enrichment of C_VL under FBS was not accompanied by net depletion of less labile or non-labile fractions. Instead, C_LL was 22.2–50% higher in FBS than cropland, and the passive pool was also significantly elevated. This pattern is compatible with the MEMS framework (Cotrufo et al., 2013) and the stoichiometric decomposition model (Chen et al., 2014), which predict that high-quality substrate inputs promote microbial biomass accumulation and subsequent organo-mineral stabilisation. Microbial necromass was not directly quantified in this study, but the co-occurrence of high SMBC and elevated C_LL under FBS is consistent with published evidence that microbial necromass contributes substantially to stable soil carbon pools (Liang et al., 2019; Kallenbach et al., 2016). Under FBS, the combination of elevated C_VL, SMBC, and reduced bulk density is consistent with simultaneous active and passive pool enrichment a dual benefit that distinguishes tree-based systems from annual crops.

The C_LL enrichment under FBS also reflects the role of soil aggregate formation in protecting organic matter from decomposition. Higher microbial biomass under FBS produces more extracellular polysaccharides and fungal hyphae, acting as binding agents in the formation of micro- and macro-aggregates (Six et al., 2004). Aggregate formation occludes recently deposited organic matter from further decomposition, converting labile C_VL into physically protected C_LL, a carbon stabilisation pathway that is fundamentally driven by rhizosphere microbial processes. The lower bulk density in FBS soils corroborates this interpretation: better aggregation creates more pore space and structural architecture, further enhancing root penetration, water retention, and microbial habitat quality in a self-reinforcing feedback loop. Root architecture drives the spatial patterning of rhizosphere effects. Peach, with the deepest and most proliferative fine root system among the FBS species studied (Thangavel et al., 2018), created the most spatially extensive rhizosphere zone, enriching SMBC, C_VL, and TN from L1 to L4. Khasi mandarin, with fine roots extending into L5–L6, dominated carbon stocks at the deepest profile layers. These species-specific patterns suggest that multi-species FBS combining shallow- and deep-rooted fruit trees could maximise rhizosphere coverage across the full soil profile; a key design principle for optimising carbon sequestration in agroforestry systems (Garrity, 2004; Oelbermann et al., 2004).

4.4 Plant–soil microbe feedbacks and long-term carbon persistence

Plant–soil microbe feedbacks operate across multiple timescales and spatial scales in perennial agroforestry systems (Wardle et al., 2004; Bardgett and Wardle, 2010; van der Putten et al., 2013). Based on the measured data and literature context, three feedback pathways are proposed as plausible explanations for the superior carbon storage performance of FBS; these are inferred patterns rather than directly demonstrated causal mechanisms.

The first is a litter quality–decomposer community feedback. Leaf nitrogen content varied strongly among species (2.0–3.8%), and higher leaf N was positively correlated with SMBC and soil TN under respective FBS. Higher leaf N inputs promote nitrogen-mineralising microbial communities that release plant-available N, further stimulating root growth, exudate production, and litter quality in subsequent seasons; a positive reinforcing loop (Cornelissen et al., 2004; Ahirwal et al., 2021). Guava, with the lowest leaf N (2.0%) and highest C: N ratio (22.6), exhibited comparatively lower SMBC enrichment despite producing the highest leaf fall volume consistent with dominance of slow-decomposing fungal communities under high-C: N litter regimes (Wardle et al., 2004). In contrast, mandarin and lemon (leaf N: 3.6–3.7%) supported broader decomposer diversity and faster nutrient cycling, consistent with stronger positive feedbacks to microbial biomass. The second feedback pathway is a root exudate–microbial activity–soil pH loop. Fruit tree root exudates sustain microbial activity in the rhizosphere; this microbial activity in turn releases organic acids and base cations from soil minerals, progressively raising soil pH under FBS (up to +0.90 pH units in mandarin relative to cropland). Higher pH reduces Al³⁺ and Mn²⁺ toxicity, improves phosphorus availability, and creates conditions more favourable for diverse microbial communities (Rousk et al., 2010). This pH-mediated feedback is particularly important in the kaolinitic Acrisol soils of the study region, where natural acidity is a primary constraint on soil biological activity. The third feedback pathway is a microbial biomass–aggregate formation–carbon stabilisation loop. Higher SMBC under FBS is associated with more extracellular polymers and fungal hyphae, consistent with aggregate formation (Six et al., 2004). Better aggregation reduces bulk density, enhances root penetration, and maintains a stable microhabitat for microbial communities in turn sustaining high SMBC and carbon stabilisation in a self-reinforcing cycle. The concurrent enrichment of both C_VL and C_LL in FBS driven by this feedback demonstrates that tree-based systems do not simply accelerate carbon cycling but simultaneously enhance both substrate availability (for biological activity) and carbon protection (in aggregates and organo-mineral associations). A fourth, emerging feedback pathway concerns the role of arbuscular mycorrhizal fungi (AMF) and ectomycorrhizal associations. While not directly measured in this study, the perennial root systems of fruit trees provide a stable and extensive host network for mycorrhizal fungi, which both extend the rhizosphere zone of nutrient and water acquisition and contribute substantial hyphal biomass to soil carbon pools (Averill et al., 2014; Leake et al., 2004). Mycorrhizal hyphal turnover rates (30–60 days) are rapid relative to bulk root turnover (months to years), contributing a high-frequency pulse of labile carbon to soil microbial communities potentially explaining some of the C_VL enrichment observed in FBS soils, particularly in subsurface layers beyond the influence of surface litter decomposition.

4.5 Implications for global change adaptation and carbon permanence

In the context of projected global change including temperature increases of 2.5–4.5 °C and altered precipitation in the Eastern Himalayan region (Krishnan et al., 2020) the carbon sequestration potential and resilience of FBS relative to cropland take on critical importance. Temperature sensitivity of soil organic matter decomposition is well established: a 1 °C increase can raise microbial respiration by 5–10% in tropical surface soils (Davidson and Janssens, 2006; Bond-Lamberty and Thomson, 2010). However, the temperature sensitivity of different carbon fractions differs substantially: labile fractions (C_VL, C_L) are more temperature-sensitive than stable fractions (C_LL, C_NL), and mineral-associated carbon is particularly resistant to warming-induced decomposition (Conant et al., 2011; Poeplau and Don, 2015).

Our data show that FBS particularly peach and Khasi mandarin have systematically higher passive pool (C_LL + C_NL) stocks than cropland, potentially conferring greater resilience to warming-induced carbon loss. Furthermore, elevated SMBC under FBS indicates higher microbial community diversity, associated with greater functional redundancy and resilience to climate perturbations (Allison and Martiny, 2008). The greater soil moisture retention under FBS canopy (+2.9 to +5.8% above cropland) is also directly relevant to adaptation, as higher moisture reduces soil carbon oxidation during drought episodes and maintains conditions for biological nitrogen fixation and nutrient mineralisation (Hu et al., 2018). Elevated atmospheric CO²; a central driver of global change is expected to stimulate fine root production in perennial trees via CO² fertilisation, increasing belowground carbon allocation and rhizosphere exudate flux (Norby and Zak, 2011). This could amplify the rhizosphere priming and microbial biomass stimulation effects documented here, potentially creating a positive global change feedback loop that further enhances soil carbon sequestration under FBS relative to cropland. Under warming scenarios, the transition from labile to stable carbon fractions may also be accelerated under FBS given the higher C_LL formation rates evidenced in our data, providing a buffer against temperature-driven carbon losses.

From a policy perspective, our findings support the integration of fruit tree-based agroforestry into India’s NDC implementation strategy and international carbon accounting frameworks including the Forest Carbon Partnership Facility (FCPF), the ART/TREES Carbon Standard, and the Verified Carbon Standard for agriculture, forestry, and land use (AFOLU). Specifically, the permanence of carbon sequestered under FBS evidenced by elevated passive pool stocks and microbial community resilience is a critical parameter for carbon credit verification. Our study provides the empirical basis for demonstrating both the additionality (FBS store significantly more carbon than adjacent cropland) and permanence (passive pool enrichment, aggregate protection, deeper profile carbon) of carbon sequestration in fruit tree agroforestry systems.

4.6 Species recommendations and agroforestry design implications

From carbon storage perspective and noting that economic performance, fruit yield, and farmer adoption were not evaluated in this study, peach showed the highest near-term SMBC and labile carbon performance, while Khasi mandarin demonstrated the greatest long-term total carbon stock. Guava, despite high leaf fall and biomass, showed moderate soil carbon outcomes, likely associated with its high litter C: N ratio. Pear consistently recorded the lowest soil carbon metrics among FBS. Any species selection for NbS programmes should additionally consider socioeconomic factors, fruit yield performance, and farmer adoption rates, which are beyond the scope of this study. From a system design perspective, the complementarity of peach (L1–L4 rhizosphere enrichment) and Khasi mandarin (L5–L6 deep carbon enrichment) suggests that mixed FBS combining these species could maximise whole-profile carbon sequestration through root niche partitioning consistent with the broader agroforestry design principle of exploiting functional diversity across ecological niches in space, time, and soil depth (Garrity, 2004; Oelbermann et al., 2004). The maintenance of turmeric and ginger intercrops within FBS is also recommended, as their elevated biomass production within FBS demonstrates the food security co-benefits essential for farmer adoption in the NEHR context.

4.7 Study limitations and future direction of research

This study has some limitations that should be considered when interpreting the findings. First, the spatial coverage and range of fruit-tree agroforestry systems examined may not fully represent the substantial environmental and management heterogeneity of the Eastern Himalayan region. Second, the observed plant–soil–microbial associations should not be interpreted as definitive evidence of mechanistic feedbacks. Third, carbon-stock estimates are subject to uncertainties associated with below-ground biomass estimation, soil bulk density and sampling depth. Furthermore, important drivers such as root exudation, microbial functional genes, extracellular enzyme activities, mycorrhizal interactions and seasonal soil moisture dynamics were not fully resolved. Finally, extrapolation of plot-level findings to the broader Eastern Himalayan landscape should be undertaken cautiously because carbon sequestration varies with species composition, site conditions and management practices. Future research should prioritise: (i) direct characterisation of microbial community composition and enzyme activity under FBS; (ii) quantification of root exudate chemistry across fruit tree species; (iii) isotopic (¹³C) fate-tracking of litter-derived carbon; and (iv) long-term monitoring under experimental climate warming to validate the enhanced carbon permanence predicted by passive pool enrichment data.

5 Conclusion

This study demonstrates that fruit tree-based agroforestry is an effective nature-based solution (NbS) for carbon storage and soil restoration in the Eastern Himalayas. After 20 years, all six fruit tree-based systems consistently improved soil carbon, microbial biomass, and nutrient status throughout the 1.0 m soil profile compared with adjacent cropland. These benefits are associated with continuous organic matter inputs, perennial root systems, and elevated microbial activity; patterns consistent with positive plant–soil–microbe interactions, though the underlying biological mechanisms were not directly measured and should be investigated in future studies.

Key findings include:

• Fruit trees produced substantial aboveground (22.68–31.98 Mg ha⁻¹) and belowground biomass (5.90–8.31 Mg ha⁻¹), with annual leaf carbon inputs of 0.256–0.357 Mg ha⁻¹ yr.⁻¹.
• Agroforestry significantly increased total carbon, organic carbon, microbial biomass carbon, total nitrogen, and both active and passive carbon pools across the soil profile.
• Peach showed the highest microbial biomass and labile carbon, while Khasi mandarin stored the greatest total and subsoil carbon.
• Perennial root systems sustained deeper microbial activity and carbon cycling than cropland.
• Positive plant–soil–microbe feedbacks enhanced nutrient cycling, soil aggregation, and long-term carbon stabilisation.

Overall, fruit tree-based agroforestry offers a practical and scientifically robust NbS for climate change mitigation, soil restoration, and sustainable livelihoods in the Eastern Himalayas. Future research should directly measure microbial community composition, enzyme activities, and root exudate chemistry to confirm the proposed plant–soil–microbe feedback mechanisms.

Generative AI statement

The author(s) declared that Generative AI was not used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Supplementary material

The Supplementary material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/ffgc.2026.1921800/full#supplementary-material

Supplementary material Supplementary material at the source

Footnotes

Edited by: Kiril Manevski, Aarhus University, Denmark

Reviewed by: Naresh Singh, Uttarakhand University of Horticulture and Forestry, India

Azimul Hasan, VIT Bhopal University, India