Introduction

Soil aggregates, composed of minerals, organic, and biotic materials, are fundamental units of soil structure and serve as crucial indicators for the formation, stabilization, and degradation of soil structure1. Physical protection within soil aggregates is an important mechanism for the storage and stabilization of particulate organic matter (POM), which is shielded from decomposition by being isolated from microorganisms and extracellular enzymes2,3. Stable aggregates can also sustain resilient and well-connected pore structures, facilitating long-term POM stabilization and sequestration4. Understanding the mechanisms by which POM is stored and stabilized by aggregates is crucial for the built-up of soil organic carbon stocks.

Soil pore structure refers to the porous medium system formed by the arrangement and composition of soil particles, which directly influences the transport of water and gas, as well as the nutrient availability5,6. Additionally, pore structure determines the fate of POM by regulating microbial activity7. However, research on POM dynamics and its protection mechanisms has traditionally relied on size and density fractionation. This method involves grinding and sieving, which destroys soil pore structure and fails to provide spatial distribution information of POM. Recently, X-ray computed tomography (CT) technology combined with machine learning has emerged as a promising approach to identify POM within intact pore structures in situ8,9. CT offers a novel protocol to investigate the distribution of POM within soil aggregates and its relationship with pore structures in soil aggregates. Studies have shown that 30-100 μm pores may play a crucial role in the processes of POM protection or loss4,10,11. These pores facilitated high microbial activity by promoting O2 and water transport, while ensuring the decomposition of fresh residues12. Furthermore, high pore connectivity and porosity within aggregates promoted POM mineralization by providing favorable conditions for microbial activities13,14. Collectively, these findings highlight the pivotal role of soil aggregate pore structure in regulating POM dynamics, influencing its accumulation and decomposition processes.

POM can be classified into fresh and decomposed fractions in CT images based on their morphological features8. Fresh POM encompasses undecomposed or semi-decomposed plant residues such as straw, roots, and leaves (fibrous structure), located inside the connected pores, which serve as hotspots for microbial decomposition. Fresh POM is incorporated into aggregates and retained within macropores through root growth, soil fauna activity, and microbial processes2,15,16. In contrast, decomposed POM tends to be more fragmented and smaller (compact structure), binding tightly to the soil matrix or small pores, which primarily originates from the further decomposition of fresh POM and some microorganism necromass6,17. Previous studies have demonstrated that climate change and agricultural management influenced POM dynamics18. In agroecosystems, fertilization not only regulates POM content by controlling exogenous organic matter inputs, but also alters soil pore structure, which governs POM loss or accumulation9,10. Notably, fresh and decomposed POM differ in origin, morphology, and distribution, and therefore respond differently to changes in pore structure. Yet, the mechanisms by which aggregate pore structure regulates their spatial distribution under various fertilization regimes across diverse climatic conditions remain unclear, particularly the role of pore size, connectivity, and spatial arrangement.

Chemical fertilizers are extensively used to enhance crop yields by supplying essential nutrients and promoting greater carbon (C) input through increased biomass production19. Nevertheless, long-term excessive fertilizer application can degrade soil structure by introducing substantial quantities of dispersive ions, such as NH4+, which reduce soil aggregation and the protection of aggregates on organic matter20. Conversely, animal manure, applied for centuries as an organic amendment, has been shown to enhance soil structure, reduce soil bulk density, and improve water retention, thereby fostering a conducive physical environment for crop growth20,21. The combined application of manure and chemical fertilizers may alleviate structural degradation caused by long-term exclusive fertilizer use22. Manure application also contributes substantial C inputs that serve as a key source for POM accumulation9. Despite these benefits, the specific influence of macropore structure (e.g., pore size distribution and connectivity) on the stabilization of fresh POM and decomposed POM within aggregates under manure application remains unclear. In this study, we hypothesized that manure application would enhance the accumulation of fresh POM and improve macropore structure, while the fate of decomposed POM could be independent of macropore structure. To test this hypothesis, we collected soils from five long-term fertilization experiments (12-34 yr) across China. X-ray CT was used to investigate the spatial distribution of POM (fresh POM, decomposed POM), macropores, and their interactions within soil aggregates. The objectives of this study were (1) to assess the effect of long-term animal manure application on characteristics of fresh POM, decomposed POM, and macropore structures within aggregates, and (2) to determine the influence of macropore structures on fresh POM and decomposed POM.

Results

Macropore structure characteristics

Across all five long-term experiments, macropore networks within aggregates exhibited greater connectivity under the NPK and NPKM treatments, as visually evidenced in Fig. S1. The aggregates under the NPKM treatment exhibited a more extensive distribution of interconnected pore networks. Compared to the control treatment, the NPK and NPKM treatments significantly increased image-based porosity at the GZL, LC, JX, and MC sites (P < 0.05, Fig. 1a). The NPKM treatment also increased surface-connected porosity at five experiment sites and isolated porosity except at the YL site (P < 0.05). In terms of pore size distributions, the NPKM treatment notably increased <30 μm porosity relative to the control treatment at five sites (P < 0.05, Fig. 1b). It also led to an increase in 30–100 μm porosity except at the LC site, and >100 μm porosity at the JX and MC sites (P < 0.05). The NPKM treatment significantly decreased the distance to pores by 30.5% and 28.0%, in the LC and JX experiments (P < 0.05, Fig. 1c). Furthermore, it increased the surface area density at five sites and connection probability except at the LC site (Table 1, P < 0.05).

Fig. 1: Aggregate macropore structures.
figure 1

Including surface-connected porosity (green) and isolated (yellow) porosity (a), pore size distributions (b), and distance to pores (c) under the control, NPK, and NPKM treatments at Gongzhuling (GZL), Luancheng (LC), Yangling (YL), Jinxian (JX), and Mengcheng (MC) experiment sites. In (b) and (c), black, yellow and green represent the control, NPK, and NPKM treatments. Different lowercase letters indicate significant differences among different fertilizations (P < 0.05). Different lowercase letters above the lines denote statistically significant differences (P < 0.05) in image-based porosity (the sum of surface-connected porosity and isolated porosity) among the treatments in (a). Vertical bars represent standard errors (n = 4).

Full size image
Table 1 Aggregate macropore characteristics
Full size table

POM distribution within aggregates

Plant debris and organic remains from manure constituted the fresh POM, while decomposed POM was more fragmented and dispersed within aggregates across all five sites (Fig. S2). At the JX site, large, spherical decomposed POM was observed. Compared to the control treatment, the NPKM treatment resulted in a significant increase in the volume density of both fresh POM and decomposed POM at all five sites (P < 0.05, Fig. 2). The NPK treatment increased decomposed POM volume density by 64.9% and 145% under LC and MC sites, respectively, and increased fresh POM volume density by 133%, 226%, and 159% at the YL, JX, and MC experiments (P < 0.05). The NPKM treatment significantly increased the surface area density of both fresh POM and decomposed POM across all five sites (P < 0.05, Table S1). Furthermore, the NPKM treatment resulted in a reduction in the sphericity of fresh POM at the LC, YL, and MC sites, as well as a reduction in the sphericity of decomposed POM at the GZL, LC, and MC sites (P < 0.05).

Fig. 2: POM volume density under different fertilizations at five experiment sites.
figure 2

Fresh POM (green) and decomposed POM (yellow). Different lowercase letters indicate significant differences among different fertilizations (P < 0.05). Different lowercase letters above the lines denote statistically significant differences (P < 0.05) in total POM (the sum of fresh POM and decomposed POM) among the treatments. Vertical bars represent standard errors (n = 4).

Full size image

The proportion of fresh POM within surface-connected pores ranged from 20% to 69%, whereas the proportion of decomposed POM was reduced to between 6% and 34% (Fig. S3). The proportion of decomposed POM distributed in isolated pores or soil matrix ranged from 66% to 94%. The NPKM treatment significantly increased the volume density of fresh POM distributed in surface-connected pores and isolated pores or matrix across all five sites relative to the control treatment (P < 0.05, Fig. 3). Additionally, the NPKM treatment also increased the volume density of decomposed POM within the surface-connected pores at five sites, and within isolated pores or matrix except at the GZL site (P < 0.05).

Fig. 3: POM distributions in soils under different fertilizations at five experiment sites.
figure 3

Fresh POM (a) and decomposed POM (b). In surface-connected pores (green) and isolated pores or matrix (yellow). Different lowercase letters indicate significant differences among different fertilizations (P < 0.05). Vertical bars represent standard errors (n = 4).

Full size image

Compared to the control treatment, the NPKM treatment significantly reduced the fresh POM-pore distance by 22.6%, 34.7%, and 35.1% at LC, YL, and JX sites, respectively (P < 0.05, Fig. 4). It also resulted in a significant reduction in decomposed POM-pore distance in GZL and JX experiments (P < 0.05). Notably, decomposed POM maintained significantly greater distances from pores than fresh POM across all treatments (P < 0.05, Fig. S4).

Fig. 4: Average distance of POM to pores under different fertilizations at five experiment sites.
figure 4

Distance of fresh POM (a) and decomposed POM (b) to pores under the control (black), NPK (yellow), and NPKM (green) treatments at five experiment sites (n = 4). Different lowercase letters indicate significant differences among different fertilizations (P < 0.05). Vertical bars represent standard errors (n = 4).

Full size image

Relationships among C input, macropore structure, and POM within aggregates

C input had positive effects on total POM and decomposed POM (Fig. 5, Table S2). The decomposed POM-pore distance had positive effects on decomposed POM. Fresh POM was positively linearly correlated with 30-100 μm porosity, distance to pores, isolated pores, surface-connected porosity, and surface area density (Fig. S5, P < 0.05). Fresh POM showed significant negative correlations with fresh POM-pore distance (P < 0.05). Decomposed POM was positively correlated with 30-100 μm porosity, surface-connected porosity, and distance to pores (P < 0.05).

Fig. 5: Redundancy analysis (RDA) of the relationships among macropore structure, C input and POM.
figure 5

Control (brown), NPK (magenta), and NPKM (blue) treatments. FPOM and DPOM indicate fresh POM and decomposed POM. Total POM is the sum of FPOM and DPOM. IP1 image-based porosity, IP2 isolated porosity, SCP surface-connected porosity, P30–100 30–100 μm porosity, CP connection probability, SAD pore surface area density, MPD mean pore diameter.

Full size image

Discussion

Long-term manure application improved macropore structures

The NPKM treatment significantly enhanced macropore structure, as evidenced by increased image-based porosity, surface-connected porosity, surface area density, and connection probability across five long-term field experiments (Fig. 1, Table 1). These findings demonstrate that manure application enhances macropore structures by improving pore connectivity and network complexity, irrespective of variations in soil properties and climatic conditions (Table S3). For instance, the GZL site exhibited a lower mean annual temperature (MAT) of 4.3 °C compared to the other four sites ( >12 °C), while the JX site experienced a higher mean annual precipitation (MAP) of 1727 mm, and the LC site had a lower clay content (15.5%). These diverse conditions further validate the consistent positive effects of manure application on pore structure enhancement. Consistent with findings from previous studies22,23, manure application facilitated the formation of pore structure within aggregates. This phenomenon can be attributed to three primary mechanisms in the study. First, manure introduces substantial amounts of organic matter, which, upon decomposition, generates numerous new pores24,25. These pores can either connect previously isolated pores or facilitate the transformation of small pores into larger ones, as demonstrated by the increased abundance of 30–100 μm pores and connected pores under the NPKM treatment (Fig. 1). The formation of new pores is accompanied by changes in pore morphology and an increase in surface area density, indicating enhanced structural roughness. The organic matter introduced by manure forms clay-granule-organic matter complexes within the soil, which strengthen cohesion within aggregates, particularly in clay soils (GZL, YL, JX, MC). This enhanced cohesion reduces the disruption of pore structures caused by wet-dry cycles or tillage practices. Second, manure is enriched with micronutrients that stimulate microbial activity, and microorganisms facilitate the development of connected pore structures by decomposing organic matter derived from manure and crop residues26. Third, manure inputs promote crop root growth and soil fauna activity, both of which play a critical role in pore network formation.

Long-term manure application promoted aggregate POM accumulation

Our study demonstrated that long-term manure application resulted in higher volume density of both fresh POM and decomposed POM across all five experiment sites (Fig. 2). This observed increase in organic matter aligns with previous studies, which attribute it to the application of organic fertilizer9,27. Although the five experimental sites differed in climate, cropping systems, and soil properties, POM exhibited a consistent response to manure application. Notably, the highest POM accumulation occurred at the GZL site, where lower temperatures suppressed POM decomposition (Table S3), which aligns with the carbon-rich characteristics of Chernozem soils28. Manure application enhanced C input, serving as a key source for fresh POM and decomposed POM accumulation (Fig. 5). Furthermore, our previous study revealed that manure application increased the biological binding agents, such as glomalin-related soil protein20, which improves small macroaggregate (0.25–2 mm) proportions. These aggregates exhibit greater structural stability when subjected to external disturbances such as wetting and drying cycles and freeze-thaw alternations, thereby enhancing the physical protection of POM1,2. Fresh POM, primarily consisting of undecomposed or semi-decomposed plant remains and organic fragments, is characterized by fast turnover, and constituted the predominant fraction of POM3,8. Cattle or pig manure, rich in coarse organic matter, can be directly incorporated into soil as fresh POM. Additionally, the application of manure, supplemented with additional nutrients such as phosphorus and potassium, enhances crop growth29. This growth results in increased above-ground and below-ground biomass, leading to higher exogenous C input (Table S4). In contrast, decomposed POM, which is more fragmented and older, results from further decomposition or translocation of fresh POM15,17. The increased fresh POM with manure addition served as a crucial source for the accumulation of decomposed POM. Additionally, manure application enhances microbial activity30 and stimulates faunal activity27, collectively accelerating the transformation of fresh POM into decomposed POM. Through bioturbation and microbial processing, these organisms further facilitate the incorporation of decomposed POM into deeper isolated pores of the soil matrix. Compared to the control, the NPKM treatment significantly increased the sphericity of decomposed POM at four sites (GZL, LC, YL, and MC; Table S1). This suggests that the enhanced morphological complexity of decomposed POM, resulting from the direct input of decomposed POM from manure and the progressive fragmentation of fresh manure-derived POM during decomposition. Consequently, manure application increases the shape complexity of decomposed POM.

Divergent roles of macropore structure on fresh POM and decomposed POM

Our study revealed distinct spatial distribution patterns of fresh and decomposed POM within soil aggregates (Figs. 3 and S3). Fresh POM was predominantly located in surface-connected pores (20%–69%), whereas decomposed POM showed a distinctly lower presence in such pores (8%–34%). Fresh POM, primarily derived from manure or crop residues, typically exhibits larger particle sizes. As a result, a substantial portion of fresh POM remains unencapsulated within soil aggregates. This spatial distribution exposes fresh POM to the external environment, enhancing its accessibility to microbial decomposition31. In contrast, decomposed POM was largely occluded within isolated pores or matrix, consistent with its having undergone multiple turnover cycles17. The encapsulation of POM by aggregates, which reduces connectivity with the aggregate exterior, is a pivotal mechanism protecting POM from microbial degradation8. RDA further demonstrated that decomposed POM-pore distance negatively influenced decomposed POM persistence (Fig. 5). Shorter POM-pore distances facilitate microbial activity, as pores serve as critical pathways for moisture and oxygen diffusion, which are key factors supporting microbial decomposition of POM11,32,33. Compared to fresh POM, decomposed POM exhibited greater distances to pores (Fig. S4), resulting in diminished pore-mediated effects on its decomposition. Processes such as freeze-thaw cycles, wet-dry alternations, and mesofauna activity (e.g., fragmentation and bioturbation) contributed to the aging and compaction of decomposed POM, strengthening its association with soil matrix34,35. Therefore, tight binding to the soil matrix or isolated pores is a crucial mechanism for the accumulation of decomposed POM within aggregates.

Positive correlations between POM volume density and macropore structures (surface-connected porosity, surface area density, 30–100 μm porosity, and connection probability) were observed in the study (Figs. S5 and 5). This indicates that fresh POM accumulation exerts a positive feedback effect on the development of macropore structure. Fresh POM contains abundant organic cementing substances, which, during their interaction with soil particles to form larger aggregates, contribute to the creation of additional pores36,37. The presence of substantial amounts of fresh organic matter supports the formation of more cohesive and resilient soil pore structures, thereby enhancing aggregate structural stability and maintaining connectivity. Furthermore, the decomposition of fresh POM generates new pore spaces that interconnect with existing ones, thereby enhancing overall pore connectivity24,25,38. Although long-term storage of the samples led to slight shrinkage of POM, this did not compromise the structural integrity of the aggregates or the POM itself, nor did it affect the validity of the results. Upon decomposition, decomposed POM contributes to pore formation within aggregates. Therefore, POM accumulation plays a critical role in the development of the macropore structure. In soils characterized by structural impediments, such as high soil strength (Vertisols), soil hardening (Fluvisols), or susceptibility to erosion (Plinthosols)39,40, POM accumulation with manure application is crucial for enhancing pore structure and nutrient availability. Well-developed pore structures (higher pore connectivity and porosity, 30–100 μm porosity) are recognized as optimal habitats for microbial activity and animal activity, where high levels of microbial and enzymatic processes occur10,11,27. Consequently, structural improvements under manure application significantly enhance POM mineralization. Our study showed that compared to the NPK treatment, the NPKM treatment significantly reduced the net fresh POM increment rate (Fig. S6), primarily because it substantially improved pore structure, creating a favorable physical environment for fresh POM decomposition10. In contrast, the net increase rate of decomposed POM showed no significant difference, resulting in its occlusion within aggregates, which reduces pore-related effects. Thus, although manure application significantly increases fresh POM content, it also leads to higher decomposition and mineralization rates. Sustained and substantial carbon inputs are required to maintain continuous POM accumulation. Future research should focus on how macropore structures influence the decomposition, transformation, and accumulation of fresh and decomposed POM during aggregate turnover (fragmentation and formation processes) under manure application, thereby elucidating the mechanisms of POM transformation.

Conclusions

This study examined the effect of long-term fertilization on the distribution and characteristics of macropore structure and POM within aggregates, using X-ray CT techniques across five long-term field experiments in China. Our findings demonstrate that long-term manure application promotes the accumulation of both fresh and decomposed POM while simultaneously improving pore structures. Fresh POM enhanced macropore structure (pore connectivity, surface area density, and 30–100 μm) and was observed in surface-connected pores, with a proportion ranging from 20% to 68%. In contrast, decomposed POM primarily was bound with the soil matrix or isolated pores and occluded in aggregates. Overall, this study provides critical insights into the distinct roles of macropore structure in fresh and decomposed POM accumulation under various fertilization regimes.

Materials and methods

Experimental design and soil sampling

Soils were collected from five long-term fertilization experiments across China (Fig. S7): Gongzhuling (GZL), Luancheng (LC), Yangling (YL), Jinxian (JX), and Mengcheng (MC). Basic information about these sites is listed in Table 2. Initiated between 1979 and 2001, these experiments had durations ranging from 12 to 34 years at the time of soil sampling. The soils were classified as Chernozems (GZL), Fluvisols (LC), Cambisols (YL), Plinthosols (JX), and Vertisols (MC)41. The soil textures were clay loam (GZL), silt loam (LC), silt clay loam (YL), silt clay loam (JX), and clay (MC), respectively. The mean annual temperature ranges from 4.5 °C at GZL to 17.2 °C at JX, while mean annual precipitation varies from 525 mm at GZL to 1727 mm at JX. The cropping systems included wheat-maize (LC and YL), maize-maize (JX), wheat-soybean (MC), and maize monocropping (GZL). Detailed information about the experiment sites can be found in Table S3. The five experiment sites were strategically selected to identify consistent patterns in POM dynamics and macropore structure under manure application across various environments and soil properties.

Table 2 Soil properties at the beginning of the five long-term fertilization experiments in China
Full size table

The experimental design included three treatments: (1) an unfertilized control, (2) inorganic fertilization (NPK), and (3) a combination of animal manure with inorganic fertilizer (NPKM). Each treatment had 3-4 replicates with different plot sizes (22–400 m2). Application rates of N, P₂O₅, and K₂O fertilizers and animal manure differed across experimental sites but were held constant during each trial period (Table 3). At all five experimental sites, animal manure (either pig or cattle-derived) was applied as a single basal dressing prior to crop seeding. Tillage was performed using rotary methods to depths of 0–15 cm.

Table 3 Treatments and application rates of animal manure at the nine long-term fertilization experiments in China (kg ha−1 yr−1)
Full size table

Soil samples were collected from the 0-20 cm plow layer after crop harvest: June 2013 (LC), June 2018 (YL), April 2013 (JX), October 2016 (MC), and April 2018 (GZL). At five experiment sites, four replicates per treatment were collected. The soil samples were passed through 3- and 5-mm sieves. Four 3–5 mm aggregates were randomly selected from each treatment for X-ray CT scanning. This resulted in a total of 60 CT samples that can be practically analyzed.

X-ray CT scanning procedures and image analysis

Aggregates were scanned using X-ray computed tomography (v|tome|x m 300) at 100 kV and 100 μA in May 2024. Images were reconstructed with Datos | x2 Rec software to generate 16-bit grayscale CT images. The 3D volume files were exported in TIFF format with VG Studio MAX 2022 at a resolution of 7 μm. Image processing was conducted using Fiji 1.5442 and VG Studio MAX 2022. Brightness and contrast were adjusted, and a median filter (radius = 2) was applied in Fiji to reduce noise. The smallest discernible features were approximately twice the size of the median filter, equivalent to 28 μm. The Adaptive Rectangle tool in VG Studio MAX 2022 was then used to delineate aggregate boundaries and prepare intact aggregate for subsequent analysis (Fig. 6a).

Fig. 6: Procedures used for visualization and quantification of POM and pores within aggregates.
figure 6

a Gray images. b Image was segmented into POM (red part), pores (blue part) and soil matrix (yellow part). c Representative examples of fresh POM (FPOM) and decomposed POM (DPOM). d POM was classified into fresh POM (green part) and decomposed POM (brown part). e Pores were classified into surface-connected pores (red part) and isolated pores (blue part). f Distance to pores. g Pore size distributions and the pore diameter increases with brightness. h 3D morphological reconstruction showing pore-contacting POM voxels and complete pore-contacting POM.

Full size image

Supervised machine learning software Ilastik was utilized to segment POM and pores in the CT images43. POM, pores, and matrix were segmented using a parallel random forest classifier. The classifier operated on a multidimensional feature space incorporating both original gray values and texture information (second derivative gray values), achieving an out-of-bag (OOB) error rate estimate of 6.5% (Fig. 6b). POM volumes greater than 1.25 × 105 μm3 were analyzed for morphological traits using the Connected Components Labeling and Analyze Region 3D tools in the MorphoLibJ plugin of Fiji. POM exhibits diverse morphological forms that can be classified into two distinct categories: fresh POM and decomposed POM. This classification is quantitatively supported by four morphological parameters: blobness, compactness, plateness, and sphericity (Table S5). Fresh POM comprises plant residues at different decomposition stages, including roots, straw, and leaves, each possessing distinct morphological features. For instance, leaves typically have a plate-like structure with a large surface area, while roots are characterized by their coherence and elongated shape. In contrast, decomposed POM, formed through microbial decomposition of residues, is more compact, and spherical, gradually integrating into the soil matrix17. For POM classification, we developed a training dataset comprising 914 manually annotated samples, with each POM cluster categorized as either fresh or decomposed (Fig. 6c). Examples of both categories are presented in Table S6. The dataset was partitioned into training (n = 640) and validation (n = 274) subsets. A random forest classifier with 200 decision trees was trained on the training subset and validated on the test subset, achieving a prediction accuracy of 93.4% in Python 3.12. The classifier underwent final training using the complete dataset, resulting in an out-of-bag (OOB) error rate estimate of 7.3%. This optimized model was subsequently applied to conduct a comprehensive analysis of POM classification across all samples. Classification results were assigned to labeled POM images using the Assign Measure to Label function in MorphoLibJ. After the classification, the Volume fraction tool was used to calculate the volume density of fresh and decomposed POM. Surface area density and sphericity of both fresh and decomposed POM were calculated using the Analyze Region 3D tool in MorphoLibJ. Sphericity was used to quantify how closely the POM resembles a spherical shape, with higher sphericity indicating greater similarity to a sphere.

We quantified multiple parameters, including image-based porosity, surface-connected porosity, isolated porosity, connection probability, surface area density, distance to pores, average POM-pore distance, and pore size distributions using Fiji/ImageJ. Image-based porosity was calculated as the ratio of pore voxels to total aggregate voxels using Volume fraction. Connection probability, a metric for pore network connectivity, ranges from 0 (many unconnected pores) to 1 (all pores forming a single interconnected cluster). It was calculated using the formula44:

$$CP={\sum }_{i=0}^{n}{V}_{i}^{2}/{\left({\sum }_{i=0}^{n}{V}_{i}\right)}^{2}$$
(1)

where CP is the connection probability, Vi the volume of each pore, and n the total number of pores. Surface area density was measured using the Connected Components Labeling and Analyze Region 3D tools in the MorphoLibJ plugin45. Distance to pores was calculated using the Euclidean distance transform of voxels excluding pore voxels, representing the shortest distance from all background voxels to the nearest pore (Fig. 6f)46. The average POM-pore distance was calculated as a weighted mean based on the frequency distribution of Euclidean distances between each POM voxel and its nearest pore. The classified POM-pore distance serves as an indicator of POM occlusion status within the soil matrix. Pore size distributions were determined with the local thickness package, based on the maximum inscribed sphere method (Fig. 6g). Pores were divided into three diameter size ranges: <30 μm, 30-100 μm, and >100 μm. Based on whether pores were connected to aggregate boundaries, we classified them into surface-connected pores and isolated intra-aggregate pores47,48. This classification was achieved using the Fill holes operation in Fiji. To determine the spatial distribution of POM within aggregates, we defined (1) POM connected to surface-connected pores as POM in surface-connected pores, (2) POM connected to isolated pores as POM in isolated pores, and (3) the remaining POM as POM in the soil matrix. We dilated both surface-connected and isolated pores by one voxel and performed a min operation with the POM image to identify POM fragments intersecting with pores, followed by 3D morphological reconstruction to obtain complete POM connected to pores (Fig. 6h). The POM connected to both pore types was classified as POM in surface-connected pores. Since only minimal POM exclusively distributed in the soil matrix was observed, we ultimately categorized POM into two classes: POM in surface-connected pores and POM in isolated pores or matrix. In the study, POM distribution and pore distribution included morphological characteristics and their spatial distribution. Visualization of the 3D pore structure and spatial distribution of POM was achieved with VG Studio MAX 2022 software (Fig. 6d, e).

C input

The carbon (C) input was estimated as the sum of carbon derived from roots (Croot+stubble), straw (Cstraw), and manure (Cmanure) using the following equations49,50:

$${C}_{input}={C}_{straw}+{C}_{root+stubble}+{C}_{manure}$$
(2)
$${{{{\rm{C}}}}}_{straw}=({Y}_{grain}\times \lambda \times O{C}_{crop})\times (1-W)\div1000$$
(3)

where Ygrain is the grain yield; λ is the ratio of straw to grain (1.2 for maize, 1.1 for wheat, and 1.6 for soybean)51; OCcrop is the C content of air-dried crop (444 g kg-1 for maize, 399 g kg-1 for wheat and 453 g kg-1 for soybean)52; W is the average water content of air-dried gain (14%).

$${C}_{root+stubble}=(({Y}_{grain}+{Y}_{straw})\times R\times {R}_{root}+{R}_{stubble}\times {Y}_{straw})\times (1-W)\times O{C}_{crop}\div1000$$
(4)

where Ystraw is the straw yield (kg ha-1); R is the ratio of root biomass to total aboveground biomass (0.351 for maize, 0.429 for wheat, and 0.235 for soybean)53,54; Rroot is the ratio of the root system within the topsoil (0-20 cm) (0.851 for maize, 0.753 for wheat, and 0.984 for soybean)52; Rstubble is the coefficient of stubble (0.03 for maize, 0.13 for wheat, and 0.15 for soybean).

$${C}_{manure}={B}_{manure}\times O{C}_{manure}$$
(5)

where Bmanure is the manure biomass and OCmanure is the C content of manure (366 g kg-1 for pig manure and 374 g kg-1 for cattle manure). The results of C input were listed in Table S4. The net POM increment rate (POMI, mm3 cm-3 Mg-1 ha-1 yr-1) under NPK and NPKM treatments was calculated by the following equation55:

$$POMI=(PO{M}_{t}-PO{M}_{control})/({C}_{input-t}-{C}_{input-control})$$
(6)

where POMt is POM volume density under NPK or NPKM treatment (mm3 cm-3), POMconntrol is POM volume density under control treatment. Cinput-t and Cinput-control is C input under NPK or NPKM treatment, and control treatment.

Data analysis

Data analysis was conducted using SPSS 23.0 and Origin 2022. One-way analysis of variance (ANOVA) was employed to assess the impact of different fertilizations on macropore structure parameters and POM characteristics. Post-hoc comparisons were performed using the least significant difference (LSD) method at a 5% significance level. When the assumptions of ANOVA were not fulfilled, the Kruskal-Walli’s test was employed to assess the statistical significance of differences. Pearson’s correlation analysis was applied to examine the relationship between POM and macropore structure parameters. Redundancy analysis (RDA) was used to analyze the relationships of macropore structure, C input, and POM-pore distance on fresh POM, decomposed POM, and total POM (sum of fresh POM and decomposed POM).

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.