Terrestrial soils possess a massive potential for carbon sequestration (~3.2 Pg Cyr⁻¹), highlighting the urgent need for cost-effective strategies to enhance soil carbon storage. Croplands, however, generally exhibit lower soil organic carbon (SOC) levels compared to natural ecosystems due to agricultural practices and crop harvest. Practices like fallowing, cultivation, and biomass removal decrease SOC by reducing carbon inputs or increasing decomposition rates. Strategically managing croplands, particularly with significant harvest residue inputs, offers an opportunity for enhanced SOC storage. Studies indicate that long-term return of crop residues can significantly improve cropland SOC stocks. Globally, cropland soils could potentially sequester 0.90–1.85 Pg Cyr⁻¹, representing ~10% of current annual fossil fuel emissions. Therefore, cropland carbon sequestration via effective management is a crucial large-scale approach for climate change mitigation. Soil microorganisms play a vital role in regulating SOC decomposition and formation. Microbial-derived necromass carbon (including dead cells, cell parts, and extracellular polymeric substances) can contribute significantly (50–80%) to SOC. The persistence of this necromass depends on its chemical composition and interactions with soil minerals. Microbial community properties (diversity, composition) also influence necromass formation and persistence, impacting SOC sequestration. Microbial community composition and species interactions regulate microbial death pathways, influencing necromass quantity and molecular composition (e.g., fungal necromass is richer in carbon than bacterial necromass). Microbial necromass available for mineral stabilization is also linked to the efficiency of microbial biomass production. Microbial communities with higher carbon use efficiency and fungal abundance accumulate more microbial-derived organic carbon. The quantitative relationships between microbial community properties, biomass production, and necromass, however, remain largely unknown, making it difficult to effectively integrate these factors into SOC prediction models. Northeast China's croplands face significant SOC depletion, with SOC decreasing from 50 to 24 g kg⁻¹ after 150 years of cultivation. A comprehensive understanding of microbial processes involved in SOC formation is crucial for restoration. Crop types also influence SOC dynamics and sequestration potential; SOC declines are less pronounced in rice paddies than in maize fields. The microbial mechanisms underlying these crop-driven differences, however, are unclear, limiting our ability to make crop-specific predictions and recommendations for SOC sequestration. This study, therefore, investigated the relationships between microbial community properties, microbial carbon pools, and SOC in cropland soils of Liaoning province, northeast China, aiming to develop crop-specific predictive models incorporating microbial parameters.

Existing research highlights the significant potential of soil carbon sequestration, particularly in terrestrial ecosystems. Studies have shown the impact of agricultural management practices on SOC levels, with practices like fallowing, cultivation, and biomass removal leading to decreased SOC. Conversely, the strategic management of croplands, particularly through the return of crop residues, has shown promise in enhancing SOC stocks. The role of soil microorganisms in regulating SOC decomposition and formation has been extensively studied. Microbial necromass, a significant component of SOC, has been shown to contribute substantially to soil organic matter. However, research on the quantitative relationships between microbial community properties, biomass production, and necromass in croplands is limited. The impact of crop type on SOC dynamics and sequestration potential has also been observed, but the underlying microbial mechanisms remain unclear. This study addresses these knowledge gaps by investigating the relationships between microbial community properties, microbial carbon pools, and SOC in cropland soils, with a focus on developing crop-specific predictive models.

This study surveyed 468 cropland soils (349 maize, 119 rice) from Liaoning province, northeast China, sampled after crop harvest in September and October 2019. Fields had been under cultivation for at least 20 years. At each site, five sampling plots (100 m²) were established, and five soil cores (0–15 cm depth) were collected per plot, creating one composite sample per site. Samples were sieved (2.0-mm mesh), visible plant material removed, and stored appropriately for various analyses. **Soil Physiochemical Analysis:** Soil pH (1:2.5 soil:water ratio), water content (oven-drying), SOC and total nitrogen (elemental analyzer), total phosphorus (digestion method), and available phosphorus (Olsen-P method) were measured. **Microbial Biomass and Necromass Carbon Analysis:** Microbial biomass carbon (MBC) was determined using the fumigation extraction method. Amino sugars (glucosamine, galactosamine, muramic acid) were measured to quantify microbial necromass carbon, using established equations to calculate bacterial and fungal necromass carbon. Total necromass carbon was the sum of bacterial and fungal necromass carbon. **Amplicon Sequencing and Data Processing:** Soil DNA was extracted, and bacterial and fungal communities were characterized by amplifying and sequencing the V4–V5 regions of the 16S rRNA gene and the ITS1 region of the ITS genes, respectively. Raw sequencing data were processed using VSEARCH and QIIME2, clustering sequences into amplicon sequence variants (ASVs) at 100% similarity. Alpha and beta diversity metrics were calculated. **Network Construction:** Bacteria-fungi interaction networks were constructed for each crop using Spearman correlation analysis. Network topological parameters (node number, average connectivity, centralization of betweenness, clustering coefficient, centralization of degree, density, average path length) were extracted, and principal component analysis (PCA) was used to reduce dimensionality, creating network complexity indices (PC1 and PC2). **Statistical Analyses:** One-way ANOVA tested differences in SOC, MBC, necromass carbon, and other variables between maize and rice soils. PERMANOVA compared microbial community composition. Wilcoxon rank-sum test compared network parameters. Multiple regression models predicted SOC, using climate, soil properties, microbial carbon pools, and community properties as predictors. Model selection used AIC. Structural equation modeling (SEM) explored relationships between microbial parameters and SOC.

This large-scale study revealed significant relationships between microbial parameters and SOC in cropland soils of northeast China. **Soil Carbon Pools and Properties:** SOC and MBC differed significantly between maize and rice soils, with rice showing higher values. However, maize soils had greater microbial necromass carbon (MNC), bacterial necromass carbon (BNC), and fungal necromass carbon (FNC). MNC accounted for a larger percentage of SOC in maize (41.2%) than in rice (27.1%). **Microbial Diversity and Network Complexity:** Distinct bacterial and fungal community compositions were observed between maize and rice soils, with maize exhibiting greater alpha diversity. Network analysis revealed that maize and rice soils had different microbial co-occurrence patterns, with maize networks showing different topological properties compared to rice networks. Taxonomic composition of the modules within the networks also differed. **Linking Microbial Community, Biomass, and Necromass C to SOC:** Microbial necromass carbon (MNC) and its components (BNC and FNC) were positively correlated with SOC in both maize and rice soils. MBC positively correlated with SOC in maize but not rice. The best model for predicting SOC included all predictors (climate, soil properties, microbial carbon pools, and community properties), showing the lowest AIC and highest R². Removing microbial pools or community properties significantly reduced model predictive power. In the best model for maize, MBC, MNC, bacterial diversity, and network complexity (network PC1) were significant predictors. Microbial carbon pools accounted for 31.7% of SOC variance, and community properties for 23.9%. For rice, soil properties explained the largest proportion of SOC variance (57.7%), with the soil N/P ratio as the most important predictor. Microbial necromass carbon, network PC1, and bacterial diversity together explained 25.7% of SOC variance in rice. **Structural Equation Modeling (SEM):** SEM analysis further elucidated the pathways between microbial parameters and SOC. For maize, necromass C and biomass C had direct positive effects on SOC, while microbial network and diversity influenced SOC directly or indirectly via microbial biomass C and necromass C. For rice, necromass C positively influenced SOC, but microbial biomass C did not. Microbial network properties were linked directly to SOC and indirectly via diversity and necromass C. In maize, microbial diversity and biomass C were the most important microbial predictors of SOC, while in rice, network properties and necromass C were the most important.

This study demonstrates the importance of considering both microbial community properties and carbon pools (biomass and necromass) for accurately predicting cropland SOC. The findings support the hypothesis that microbial necromass is a substantial contributor to soil organic matter. However, its relative importance varies between crops, possibly due to differences in soil conditions (e.g., oxygen limitation in waterlogged rice soils). Fungal necromass appears more important than bacterial necromass for SOC accumulation in these croplands, possibly due to its slower decomposition rate and role in aggregate formation. The study also provides evidence of the influence of microbial community properties (diversity and network complexity) on SOC accumulation, likely through their effect on microbial biomass and necromass production and carbon use efficiency. The integration of microbial parameters significantly improves SOC prediction models, although the relative importance of these parameters differs between maize and rice soils. Soil properties also play a crucial role in determining SOC concentration, with soil nutrient ratios being particularly influential.

This large-scale study highlights the critical role of microbial communities in regulating SOC formation and accumulation in croplands. Both microbial carbon pools (biomass and necromass) and community properties (diversity and network complexity) significantly influence SOC, improving the accuracy of SOC prediction models. Fungal necromass appears to be a particularly important contributor to SOC. Management strategies aimed at enhancing microbial diversity, complexity, biomass, and necromass could be effective in increasing cropland SOC levels. Future research could focus on investigating the specific mechanisms underlying these relationships and the impacts of various agricultural management practices on these microbial parameters.

While this study provides valuable insights, certain limitations should be acknowledged. The study focused on a specific region of Northeast China, limiting the generalizability of the findings to other regions with different climates and soil types. The study's temporal scale (a single sampling point) may not fully capture the dynamic nature of SOC and microbial interactions. Further research incorporating longer-term monitoring and broader geographical representation would strengthen the conclusions. Additionally, the study's reliance on correlation analysis does not establish direct causation between microbial parameters and SOC.