Microbial extracellular polymeric substances (EPS) represent an important extracellular microbial carbon pool, yet their responses to nitrogen (N) deposition under contrasting belowground ingrowth conditions remain poorly understood. We combined three levels of N addition (0, 40, and 80 kg Nha−1 yr−1) with root–hyphal treatments in a subtropical Castanopsis carlesii natural forest to examine changes in EPS content, microbial biomass carbon (MBC)-normalized EPS (EPS/MBC ratio), and the estimated contribution of EPS-C to SOC. Low N addition significantly reduced polysaccharides and total EPS by 24% and 20%, respectively, whereas high N addition had no significant effect on EPS content. These changes coincided with declines in MBC, consistent with an association between microbial biomass and variation in the standing EPS pool under N enrichment. In contrast, EPS/MBC ratios increased across all EPS fractions, most strongly under high N, and were associated with lower microbial carbon use efficiency (CUE) and altered biomass-specific extracellular enzyme activities. This points to a shift in the standing EPS pool relative to microbial biomass rather than a change in absolute EPS content. No statistically significant effect of root–hyphal treatment on EPS content or EPS/MBC ratios was detected. Estimated EPS-C accounted for only 0.3–0.5% of SOC; its contribution decreased under low N addition and was higher under root–hyphal exclusion, the latter reflecting a smaller bulk SOC pool rather than increased EPS-C content. Overall, this study shows that N deposition alters biomass-normalized EPS responses more strongly than it alters the bulk EPS pool, identifying the EPS/MBC ratio as a sensitive indicator of shifts in the relative balance between extractable EPS and microbial biomass under N enrichment.
Soils are complex biogeochemical systems in which organic matter (OM) and mineral surfaces are heterogeneously distributed. While interactions of OM with reactive mineral surfaces are broadly recognized to shape the fate of soil organic carbon, the mechanisms governing the preferential spatial distribution of OM across mineral phases within intact microstructures remain unknown. Here, we mapped soil microstructures incubated with 15N-enriched litter using nanoscale secondary ion mass spectrometry (NanoSIMS) at a resolution of 120 nm. To identify different mineral phases within mineral-dominated regions, we developed an unsupervised machine-learning segmentation which enabled us to cluster the sample surfaces with mineral references based on elemental signatures. The spatial preferential arrangement of mineral phases with pre-existing (14N-enriched) and newly formed (15N-enriched) OM was quantified by dilation and nearest neighbor edge-to-edge distance analyses. We found that comparable mineral phases and some of their weathering products neighbored one another. Pre-existing OM was pre-dominantly spatially adjacent to calcium carbonate within the pores while newly formed OM was spatially adjacent to pre-existing OM within pores. This highlights that pores channel the distribution of OM through soil microstructure with pre-existing OM regions providing preferential binding spots while distinct mineral phases did not appear to strongly influence the arrangement due to their constrained accessibility. This indicates that the coupled effects of soil pore architecture and the heterogeneous distribution of mineral phases shape OM stabilization, rather than being primarily driven by the allocation of the most reactive mineral phases.
Soil organic matter (SOM) is a key indicator of soil quality and ecosystem functioning, directly linked to crop productivity and sustainable development. Accurately characterizing its spatiotemporal dynamics is crucial for activities related to land use planning, environmental protection, and land degradation management. Soil imaging spectroscopy has emerged as a validated technique for estimating soil properties, particularly gaining traction for quantifying organic matter. In this study, a physical retrieval model for SOM content is developed using the radiative transfer model. By incorporating the effect of moisture on reflectance, this method offers a novel strategy for addressing the challenges in retrieving soil environmental quality parameters. The physical-based, feature-based, and deep neural network-based methods were compared and cross-validated within a unified SOM retrieval framework. The coefficients of determination (R2) of the validation sets were 0.556 (physical-based model), 0.688 (feature-based statistical model), and 0.725 (deep neural network-based model), respectively. Correlation analysis results indicate good agreement among the three methods, with correlation coefficients greater than 0.6. By employing an extended triple collocation algorithm to assess the error distribution, it was found that the physical-based model achieved a squared correlation coefficient of 0.839 against the unknown true value, outperforming the other two models. The categorical triple collocation analysis revealed clear differences in the performance of the three methods at different organic matter gradients. Finally, the three SOM products were merged using TCA-derived weights to generate an integrated SOM map.
Downscaling existing digital soil maps is needed to provide the spatial detail required for fine-resolution land management and carbon accounting. This is particularly important for soil organic carbon (SOC), where existing national maps are often inadequate for local decision-making. Current downscaling approaches commonly rely on field measurements, high-resolution covariates at the target location, or simple bilinear interpolation, which can limit their applicability and accuracy in data-limited regions. Deep learning super-resolution methods have recently been explored for environmental variables and can provide a promising pathway for downscaling coarse digital soil maps. Here, we present a deep super‑resolution framework based on the Residual Channel Attention Network (RCAN) and demonstrate its application on SOC maps, showing that RCAN can downscale 90 m inputs to 30 m by learning transferable coarse‑to‑fine patterns. We found that RCAN maintained a strong correlation between estimated and reference SOC (R2 ≥ 0.992) and achieved great accuracy (median error ≤ 0.0036%) within the dominant SOC range. RCAN consistently outperformed bilinear interpolation by reducing prediction errors, improving reconstruction fidelity, and better preserving fine-scale spatial structure. These results highlight the potential of a deep learning super‑resolution framework for spatial downscaling, using SOC as an example of its applicability in data‑limited contexts.
Soil organic matter (SOM) is critical for soil health, ecosystem functioning, and climate regulation. Digital soil mapping (DSM) offers an efficient approach for estimating the spatial distribution of SOM, but its accuracy strongly depends on whether environmental covariates adequately represent soil spatial variability. In recent years, dynamic covariates that change over time, particularly remote sensing vegetation indices (VI) and soil moisture (SM), have been increasingly incorporated into DSM. However, many studies still rely on a static perspective (usually using temporally averaged representations) for establishing covariates, which leads to the fact that the intrinsic temporal dynamic features of those covariates cannot be well captured. In this study, we extracted temporal dynamic covariates from VI and SM time series data over multiple time windows (1-year, 3-year, 5-year, and 10-year) using a Fast Fourier Transform (FFT) approach and evaluated their performance in SOM mapping across paddy fields, drylands, and forests in Anhui Province, China, using a random forest model. The results indicate that the FFT-derived VI-FFT and SM-FFT covariates improved SOM mapping accuracy across all evaluated time windows, with the 1-year + 5-year combination achieving the best performance. Adding VI-FFT and SM-FFT covariates increased the metric of Lin’s concordance correlation coefficient (LCCC) by 13.9% and 12.6%, respectively. The combined use of VI-FFT and SM-FFT covariates achieved the highest predictive performance, with a model improvement of 35.0% on average. Our results underscore the importance of incorporating multi-year environmental dynamics for establishing a robust SOC-environment relationship. This suggests that future DSM efforts should move beyond static covariates by systematically extracting temporal features from historical observations of vegetation and soil moisture.
Soil microbial biomass is a sensitive indicator of soil nutrient dynamics and biogeochemical processes. Fungi, as an integral component of microbial communities, play a dominant role in critical biogeochemical functions. Although the fundamental role of fungi in global carbon and nutrient cycles is widely recognized, their biogeographic distribution remains poorly understood. Therefore, systematically mapping the spatial distribution of global fungal biomass and its dominant drivers provides an essential scientific basis for understanding terrestrial ecosystem responses to global change. This study integrates observational data of phospholipid fatty acid (PLFA) from 4,502 global sampling sites with 15 environmental factors, including space, climate, soil, and plants attributes. Using the XGBoost model with optimal predictive performance, we estimated fungal biomass carbon (FBC) stock at 0–30 cm soil depth (topsoil) and generated a high-resolution spatial distribution map. The results: (1) Global FBC stock in topsoil was estimated to be 14.07 Pg C (9.77–18.25, mean with 25% and 75% quantiles). Spatially, FBC stock was highest in high-latitude forests and peatlands in the northern hemisphere and lowest in mid-latitude desert regions. (2) Climate is the strongest predictor within the model framework regulating spatial FBC variability: negative correlation with mean annual temperature (MAT), positive with mean annual precipitation (MAP). (3) FBC exhibits nonlinear responses to environmental factors, with critical threshold values at: MAT (6.21 °C), soil pH (5.20), aridity index (0.95), and MAP (66.10 cm). (4) Sensitivity analysis indicates that under a 20% increase in future precipitation, global FBC stocks will significantly rise (p < 0.01), with the primary increase occurring in mid-latitude desert regions. The results not only provide critical parameters for global soil carbon pool estimation and enhance understanding of fungal-driven carbon cycling mechanisms, but also establish a basis for predicting soil carbon sink dynamics under climate change.
The Loess Sandy Region, a fragile ecosystem facing severe soil degradation, requires sustainable strategies to enhance soil organic carbon (SOC) and restore ecological functions. Through a long-term split-plot experiment, we evaluated straw mulching (applied (S) or not (S0)) with nitrogen fertilizer types (none (W), conventional urea (U), slow-release urea (RU)) during critical phenological periods of soybean growth. The results indicated that SU and SRU treatments significantly increased average yields by 209.03% and 356.32%, respectively, compared to S0W. Regarding carbon pools, compared to S0W, S0U and S0RU treatments reduced SOC content by 6.31% and 5.33%, respectively. The SW and SRU treatments significantly increased SOC content by 33.53% and 45.17%, respectively, and also significantly raised mineral-associated organic carbon (MAOC) content by 35.37% and 26.20%. Furthermore, SW, SU, and SRU treatments reduced the rate of organic carbon mineralization. Enzyme stoichiometric modeling revealed that SU and SRU effectively alleviated microbial nitrogen limitation. The S0U and S0RU treatments significantly reduced bacterial Chao1 and Shannon at the pod stage (R4). By the maturity stage (R8), SU and SRU treatments increased the Shannon index. Co-occurrence network analysis demonstrated that SRU treatment significantly enhanced bacterial network complexity and exhibited the lowest average variation degree. Although favorable hydrothermal conditions stimulate microbial activity and accelerate the mineralization of native organic matter in the short term, in the long run, the efficient utilization of straw-derived carbon by microorganisms, particularly the significant increase in MAOC, drives the overall increase in SOC. Overall, this study provides important theoretical support for advancing the sustainable restoration of degraded ecosystems in the Loess Region.
Redox cycling among multiple valence states of manganese in soils exerts a fundamental control over the biogeochemical cycling of virtually all elements. The redox reactive intermediate Mn(III) species complexed by organic matter has increasingly emerged as a dominant factor in these processes, and there is a need for a simple and reliable analytical method for the determination of the reactive Mn(III) species. However, the specificity of existing chemical extraction approaches remains uncertain, particularly with respect to whether the extraction process substantially dissolves other Mn(III)-bearing oxide phases and thereby biases the quantification of organically complexed Mn(III). To address this limitation, this study proposes an improved sequential extraction protocol refined for discerning manganese (Mn) fractions in soils, particularly incorporating an alkaline pyrophosphate (PP) extraction step for organic-bound reactive Mn(III) species. The method was validated using a set of Mn mineral standards and then applied to three groups of soil samples that span low, intermediate, and high Mn(III) contents based on linear combination fitting (LCF) analysis of X-ray absorption near edge structure (XANES) spectroscopy, including one group characterized by a high proportion of organically bound Mn(III). Alkaline PP extraction at pH 10.5 for 12 h caused negligible dissolution of common Mn(III)-bearing minerals in soils. Furthermore, correlation analysis between sequentially extracted Mn fractions and XANES-LCF fitting results revealed a positive relationship between PP-extractable Mn and organically bound Mn(III), suggesting that PP extraction may provide an effective approach for selectively targeting organically bound Mn(III). In addition, acidification prior to inductively coupled plasma mass spectrometry (ICP-MS) analysis may cause precipitation of organically complexed Mn in PP extracts. Validation experiments involving digestion and reduction treatments indicated that immediate analysis after acidification is essential, and that digestion is required when precipitates are present. The sequential extraction method developed in this study can supplement and refine the interpretation of spectroscopic analyses, offering a comprehensive and straightforward view of soil Mn speciation, especially for organically bound Mn(III) species.
The rapid expansion of agrivoltaic systems in karst regions has raised concerns regarding their impacts on soil hydrological processes and erosion dynamics in fragile landscapes. While these systems improve land-use efficiency, their structural configuration alters rainfall redistribution, potentially exacerbating soil erosion on karst hillslopes. However, quantitative understanding of these effects and their implications for soil and water conservation remains limited. This study quantified rainfall redistribution, soil erosion responses, and runoff harvesting potential under agrivoltaic systems in a karst mountainous area. Two field plots (30 m × 3 m) were established beneath photovoltaic (PV) arrays, including one plot equipped with a runoff collection system installed along panel drip lines. A one-year field monitoring experiment under natural rainfall conditions was conducted to assess rainfall redistribution patterns, soil loss, and runoff collection efficiency. Results showed that photovoltaic arrays significantly altered rainfall distribution patterns, with effects strongly dependent on rainfall intensity. During light to moderate rainfall events, 46.79%–66.46% of the total precipitation input during each event was concentrated along panel drip lines. During heavy rainfall to rainstorm events, 49.39%–63.21% of the event precipitation was redistributed to inter-panel areas, whereas less than 10% of the total event precipitation reached beneath panels. This spatial redistribution resulted in significantly higher soil erosion rates at panel edges and inter-panel zones compared to beneath panels (P < 0.05). Annual soil loss exceeded 5000 t km−2 yr−1, and bedrock exposure increased from 9.92% to 18.68%, indicating enhanced erosion risk. The runoff collection system effectively intercepted concentrated flow, capturing 69.0%–72.9% of rainfall and reducing localized erosion. These results demonstrate that agrivoltaic-induced rainfall redistribution is a critical driver of soil erosion in karst environments, while targeted runoff harvesting can partially offset these effects. Overall, this study provides field-based evidence of the coupled impacts of agrivoltaic systems on rainfall redistribution and soil erosion, and highlights the importance of integrating hydrological regulation measures to sustain soil resources in karst agroecosystems.
Phosphorus (P) fertilizer efficiency in highly weathered soils is constrained by phosphate sorption on iron (Fe) and aluminum (Al) oxides. Integrated soil fertility management promotes the combined use of P fertilizers and organic amendments (OAs), but the mechanisms underlying OA-P interactions remain unclear. This study evaluated how OA dose, OA quality, and time affect P mobility, assessed through phosphate sorption and diffusion from a fertilizer granule. Three Ferralsols were incubated with compost, farmyard manure (FYM), or straw at four doses (0–4 % w/w) in a full factorial design. Phosphate sorption strength (KD) was quantified after 1 and 49 days of incubation and related to the soil’s degree of carbon saturation (DCS), an operational metric that is proportional to the molar ratio of organic C to the reactive Fe and Al oxides. Across all soils and treatments, KD decreased with increasing DCS (R2 = 0.66) at 49 days after amendment, consistent with increasing competition of organic anions with phosphate for sorption sites. In contrast, the short-term (1-day) effects of amendments on KD were primarily related to soil pH. A diffusion experiment using Ferralsol samples from a field trial with three years of FYM application showed wider diffusion zones around fertilizer granules in FYM-amended than in control soils. Together, these findings suggest that OA-induced improvements in P fertilizer use efficiency observed in field trials are related to competition effects and that the effects of OA dose and quality can be predicted with a mechanistic model.
Subsoil compaction limits root growth and water uptake, often amplifying crop stress during drought. Biological “bio-drilling” with deep-rooting plants is a sustainable alternative to mechanical tillage, creating stable biopores and improving soil structure and water infiltration, but assessing its long-term effectiveness remains challenging. This study evaluates the use of a multi-coil electromagnetic induction (EMI) system to monitor subsoil responses to various loosening strategies, including biological treatments (Lucerne and Ryegrass mixture) and mechanical deep loosening with or without compost incorporation. EMI surveys prior to the experiment revealed inherent soil heterogeneity, emphasizing the importance of baseline mapping for robust experimental design. To isolate management-induced signatures, an innovative data processing workflow was implemented to effectively disentangle anthropogenic structural legacies from the dominant pedological “soil memory” (i.e., inherent natural soil heterogeneity). Time-lapse surveys tracked treatment effects over subsequent cropping seasons. Complementary EMI point measurements showed a distinct negative trend in apparent electrical conductivity (ECa), consistent with soil drying associated with crop development and root water uptake. This suggests that biological and mechanical amelioration may leave lasting structural signatures in the subsoil that modify water accessibility and hydraulic connectivity, ultimately enhancing the crop’s ability to exploit deep water reserves. By combining spatial and temporal monitoring, EMI provides a powerful tool to link soil structural changes to crop water dynamics and resource use efficiency. This work shows how non-invasive geophysical methods can reveal subtle but ecologically meaningful responses in the soil-root continuum under real cropping conditions and supports the potential of bio-drilling to reduce subsoil compaction.
Surface water browning is a widespread phenomenon in the Northern Hemisphere, although its underlying drivers remain debated. In this study, we combined 20 years of dissolved organic carbon (DOC) measurements with five years of optical characterization in soil solutions collected at multiple depths from two contrasting forest plots in the Strengbach catchment (Vosges Mountains, France): a spruce stand affected by dieback and a beech stand. We developed a conceptual framework to interpret the optical properties of DOC in soil solutions, based on commonly used indices and distinguishing between two endmembers: litter-derived and microbial-derived signatures. Our results reveal a marked browning in soil solutions associated with spruce decline, characterized by increasing DOC and aromaticity. DOC concentrations increased significantly at several soil depths between 2004 and 2024, with an intensified increase following bark beetle infestation in 2013, reaching up to 2 ppm-C/year at 5 and 30 cm depth. This increase was accompanied by a shift towards more litter-derived DOC and a concomitant decrease in microbial-derived signatures, exacerbated in deeper soil solutions. A soil sequence representing a post-decline gradient further demonstrated that spruce decline increased soil lignin content, resulting in higher DOC and aromaticity in soil solutions, driven by litter input changes. The reduced microbial activity following clear-cut may explain the lower microbial signature. Overall, the disturbance of forest cover associated with spruce decline is a major driver of soil solution browning, raising concerns regarding the quality of downstream surface waters.
Invasive alien plant species (IAPS) Japanese knotweed (Reynoutria japonica), native to East Asia, has spread to Europe, North America and other regions posing a threat to the indigenous species and ecosystems. Its invasion can be, among the aggressive growth, resilience and lack of predators, attributed to the phenolic compounds released by the plant or produced by the rhizosphere microbiome from lignin. We, therefore, isolated five lignin degrading Paraburkholderia spp. strains from the R. japonica rhizosphere and identified lignin degradation from three sources: (1) alkali lignin, (2) lignin from R. japonica plant biomass and (3) lignin from R. japonica black liquor. Furthermore, the degradation of lignin in R. japonica plant biomass was significantly more effective compared to alkali lignin. The accumulation of compounds such as hydroxycinnamic acid, nonanoic acid, and octanoic acid after degradation suggests that bacteria can partially break down lignin into intermediates, which may contribute to R. japonica invasiveness. Genome annotation analysis of the L4 strain further confirmed the ability of Paraburkholderia spp. to produce phenolic and other compounds involved IAPS promotion. Beta-ketoadipate and phenylpropanoid metabolic pathway were found to be predominant, with the lack of common lignin degrading enzymes (laccase, lignin- and manganese peroxidase).
Remote sensing images provide valuable soil spectral information for predicting soil organic matter (SOM) content. However, environmental factors, such as surface roughness, soil moisture, and crop residues, can interfere with soil spectral responses, rendering many bare soil images unsuitable for accurate SOM quantification. This study proposes a bare soil image selection approach based on spatial pattern similarity for SOM prediction. An agricultural region in central Jiangsu Province, China, was selected as the study area. Bare soil pixels were extracted using the Google Earth Engine (GEE) platform from MOD09A1 images acquired in January, November, and December from 2001 to 2024. Bare soil images exhibiting similar spatial patterns were identified from multitemporal observations and subsequently composited to generate a series of similarity-based bare soil composite images (SBSCI_p10–p90). These composite images were then evaluated for SOM prediction, with further comparison against the conventional bare soil composite image (CBSCI), both with and without soil-forming factors. Spatial cross-validation results showed that SBSCI_p70 yielded the highest R2 (0.27) and the lowest RMSE (6.66) among the evaluated SBSCIs. Compared with the CBSCI (R2 = 0.23 and RMSE = 6.86), SBSCI_p70 showed improved SOM prediction accuracy. Similar results were observed after incorporating soil-forming factors, with the SBSCI_p70-based model yielding a higher R2 (0.40) and a lower RMSE (6.04) than the CBSCI-based model (R2 = 0.36 and RMSE = 6.25). These results suggest the potential of spatial pattern similarity in selecting bare soil images for SOM prediction.
Subalpine soils store a significant amount of soil organic carbon (SOC), yet the factors driving its landscape-scale variability remain poorly constrained. Although topography, soil type, soil texture, and moisture are recognised as key drivers of SOC concentration, their interactive effects in subalpine environments remain largely unexplored. In particular, the extent to which soil type shapes landscape-scale SOC patterns remains unclear. In this study, we assessed SOC concentration, soil type, soil moisture, soil texture and elemental composition across 100 plots in two subalpine sites in the Swiss Alps. Using factor analysis, we generated three standardised continuous predictors from our dataset, namely, mineral content, sedimentary rock influence and texture index. We investigated the joint influence of these continuous predictors on SOC variability at the landscape scale in a linear mixed-effects model with topographic position (plain vs. slope areas), soil depth and soil type as additional categorical predictors. We found that soil type diversity was greater on the plain than on the slope, reflecting a greater hydrological heterogeneity. Plains supported a broad spectrum of soils, ranging from organic-rich Histosols to mineral-rich Fluvisols, while Cambisols dominated slopes. Although plains exhibited higher moisture and SOC concentration, and slopes were characterised by higher clay fractions and lithogenous elements, topographic position was not a significant predictor for the spatial distribution of SOC. Our linear mixed-effect model identified mineral content as the strongest predictor of SOC concentration, followed by the texture index and sedimentary rock influence. SOC concentration decreased as mineral content and sedimentary rock influence increased, and as the texture index shifted towards a lower proportion of clay-sized particles. Crucially, while mineral content integrates much of the variability described by the categorical variables included in the model, soil type diversity effectively captures the range of mineral content that drives SOC concentration across our sampling locations and soil depth. Our findings suggest that relying strictly on topographic position to assess SOC spatial variability overlooks important nuances in soil property distribution. We therefore suggest combining topographic position and soil type assessment into a robust framework for predicting landscape-scale SOC variability in subalpine environments.
Microbial carbon use efficiency (CUE) is a key parameter linking microbial core processes to soil organic carbon (SOC) storage. Although the 18O labeling method and the ecoenzymatic stoichiometry model have been widely used to estimate CUE, how CUE derived from these two methods varies along elevation gradients and its relationship with SOC remain poorly understood. In this study, we applied both methods to a local elevational transect in Shennongjia (800–2900 m) and a multi‑site synthesis encompassing 16 mountain ecosystems across Asia. At the multi‑site scale, 18O‑CUE increased with elevation but showed a non‑significant positive trend with SOC. Locally, 18O‑CUE exhibited a unimodal pattern peaking at mid‑elevations, and its relationship with SOC became significantly negative after accounting for climatic covariation. By contrast, CUEC: N showed no consistent elevational trend at either scale but was negatively associated with SOC in the multi‑site synthesis, with a directionally consistent trend locally. At the local scale, microbial community attributes (e.g., fungal Shannon diversity, F:B ratio) predominantly determined 18O‑CUE, whereas at the multi‑site scale, soil pH was the dominant factor. For CUEC: N, MAT was the predominant driver at the local scale, and at the multi‑site scale, soil pH indirectly regulated CUEC: N through its effect on EEAC: N. These findings demonstrate that the ecological meaning of CUE is method‑dependent, highlighting the importance of distinguishing the two metrics when predicting microbial‑mediated carbon sequestration and its feedbacks to global change.
Visible and near-infrared (vis-NIR) spectroscopy enables rapid and cost-effective soil characterization, supporting the estimation of soil chemical, physical, and biological properties. Most applications, however, focus on using spectroscopy to predict individual soil properties and augment conventional laboratory analyses, while the potential of spectral information to study soil variation in space remains largely unexplored. Latent variables derived from the dimensionality reduction of spectral data offer a compact representation of soil variability, capturing multiple soil properties simultaneously. When spatially predicted, these latent variables can provide a means to investigate soil-landscape relationships in a spatially explicit context.This study aimed to (i) model the spatial variation of latent variables derived from compressed topsoil vis-NIR spectral data across Denmark at 10 m resolution using a digital soil mapping approach; (ii) identify the drivers, in relation to SCORPAN factors, controlling their spatial variation; and (iii) examine how the predicted latent variables represent the soil-landscape relationships across Denmark. An earlier study compressed the data using principal component analysis, kernel principal component analysis, shallow autoencoders, and convolutional autoencoders. To predict the latent variables in space, we used 32 predictors comprising harmonized environmental layers representing climate, relief, and parent material. A weighted machine learning algorithm was used to model each latent variable, with bootstrap resampling providing pixel-level uncertainty estimates. Overall, climate was the main driver across all methods, followed by topography and parent material, mainly the extent of clay till deposits.By integrating compressed spectral data with diverse spatial predictors, we captured complex, non-linear soil-landscape relationships. The resulting high-resolution maps offer a spatial description of soil variation in relation to soil-forming factors and spectral signature. Our maps can be used to improve point-based predictions of soil properties or provide enhanced covariates to support digital soil mapping of soil properties and classes.
Rapid transitions in soil physical conditions during freezing and thawing (FT) processes make it particularly challenging to accurately model nitrous oxide (N2O) production and fluxes from agricultural soils in cold climates. Biogeochemical models, such as the DeNitrification and DeComposition model (DNDC), must be calibrated and validated against measured data to ensure the accurate simulation of the soil processes underlying FT N2O fluxes. This study preformed the first known direct comparison between simultaneously measured soil N2O content and daily fluxes, and DNDC predictions under FT conditions. Using high-precision data collected from large field lysimeters at Elora, Ontario, Canada, the DNDC model was modified, calibrated and validated for contrasting FT conditions (i.e., overwinter cover crops and simulated warmer winters). The model calibration process was detailed to future FT model improvement efforts. DNDC adequately predicted soil temperature (d = 0.95), daily N2O fluxes (d = 0.56) and cumulative emissions (R2 = 0.92) in a diverse crop rotation, however, faced challenges simulating FT N2O fluxes under winter warming conditions due to poor simulations of snowpack and soil N2O storage. Evaluating DNDC’s soil N2O content and flux predictions highlighted fundamental issues within the model, including the poor simulation of N2O storage during frozen soil conditions and of alternative nitrogen dynamics (i.e., microbial consumption of N2O and downward transport through the soil profile). These issues resulted in the overestimation of N2O fluxes, 192 % greater than field-observed. Changes to the submodules influencing FT hydrological processes and gas diffusion are suggested to improve N cycling predictions during FT conditions.
Environmental disturbances often drive the mixing of previously isolated microbial communities, a process termed coalescence, which reshapes community structure and interactions. While coalescence is recognized as a major driver of microbial diversity and ecosystem functioning, the specific metabolic mechanisms that stabilize coexistence remain poorly understood. In this study, we established controlled soil microcosms with single-source (derived from individual sites) and mixed-source inocula (coalesced from multiple sites) and utilized an integrated approach of metagenomics, genome-scale metabolic modeling, and vitamin supplementation to investigate the mechanisms that sustain diversity during coalescence. Coalescence increased α‑diversity but reduced β‑diversity, with most taxa originating from their single-source communities. At the genomic level, coalescence increased both metagenome-assembled genome (MAG) richness and functional trait breadth, with Bacteroidota and Pseudomonadota remaining the dominant lineages across all treatments. Coalesced communities exhibited extensive chimeric associations (newly formed associations), which were accompanied by enhanced metabolic complementarity and more frequent positive interactions among dominant MAGs affiliated with Bacteroidota and Pseudomonadota. We identified thiamine metabolism as a key mechanism stabilizing this co-dominance. Genomically streamlined Pseudomonadota prototrophs harbored the complete thiamine biosynthesis pathway, while most Bacteroidota MAGs (82%) were thiamine auxotrophs, forming the genetic basis for potential thiamine cross-feeding interactions that may alleviate widespread thiamine auxotrophy. The relative abundance of thiamine biosynthesis genes positively correlated with the abundance of cellulose and lignin derivative degradation genes (e.g., biphenyl, phenylacetate, and protocatechuic acid). In parallel, coalesced communities exhibited significantly higher soil respiration rates, indicating enhanced carbon mineralization potential. Exogenous thiamine supplementation further supported this mechanism by increasing community richness across dilution gradients. Together, our results indicate that vitamin-mediated metabolic dependencies support microbial coexistence during microbial community coalescence and are positively associated with enhanced soil carbon mineralization, with implications for understanding soil carbon processing under global change.
