Accurate soil organic carbon (SOC) mapping is vital for sustainable agriculture in black soil regions. Although remote sensing offers an efficient and cost-effective approach for characterizing the spatial distribution of SOC, three major bottlenecks persist: single source data cannot fully represent SOC dynamics, traditional models struggle with complex nonlinearities, and spatial heterogeneity limits global models from reflecting local pedogenic mechanisms. To address these, this study developed a hybrid framework for the Songnen Plain integrating a zonal ensemble strategy with deep learning. Using Google Earth Engine, a dataset was constructed with multi-temporal Sentinel-1 radar, Sentinel-2 optical data, and topographic covariates from 2020 to 2024, covering both bare soil (April-May) and growing seasons (June-August). A prediction model combining Fuzzy CMeans (FCM) clustering and Long Short-Term Memory (LSTM) networks was developed using 640 topsoil samples. Results indicate: (1) the FCM-LSTM zonal ensemble achieved optimal performance on the independent hold-out test set (R2 = 0.84, RMSE = 3.20 g kg-1); (2) multi-temporal data fusion effectively enhanced accuracy; compared to relying solely on bare soil data, including growing season data increased the R2 of Sentinel-2 and Sentinel-1 by 0.08 and 0.03, respectively; and (3) the synergistic use of multi-temporal radar and optical data, supplemented with environmental covariates, significantly improved SOC prediction. This study demonstrates the effectiveness of integrating multi-temporal remote sensing with a zonal ensemble strategy, providing a robust methodological template for high-precision tillage management and soil fertility preservation in critical black soil regions.
Ginseng seedling blight, caused by the soil-borne fungus Rhizoctonia solani, severely impacts ginseng yield. The increasing demand for overcoming photodegradation, off-target drift, and drug resistance of conventional synthetic fungicides has prompted the development of safe and efficient alternatives. Herein, a pathogenic microenvironment-responsive nanopesticides (OS@ZSSI@CuO2) integrating silica-modified ZIF-8 nanozymes as the porous vehicle and copper oxide (CuO2) as the capping agent was constructed to load osthole (OS) for targeted synergistic management of ginseng seedling blight. Under pathogenic acidic microenvironment, the decomposition of CuO2 resulted in the release of OS and Cu2+, and concurrently generated H2O2 to maintain peroxidase-like activity of ZSSI for the continuous production of hydroxyl radicals. Benefiting from its precise targeting ability and multimodal cooperative effects, OS@ZSSI@CuO2 demonstrated improved antifungal performance at reduced concentrations in both in vitro assays and pot trials.Transcriptomic analysis further confirmed that OS@ZSSI@CuO2 could substantially upregulate the expression of disease resistance-related genes and systematically activate the plant immune defense network. Biosafety evaluation revealed OS@ZSSI@CuO2 had negligible toxicity. This study provided an efficient and intelligent strategy for sustainable crop protection.
Soil organic matter (SOM) is a key indicator of Mollisol quality and is relevant to precision farmland management. However, regional SOM mapping remains constrained by the coarse spatial resolution of satellite imagery and the limited coverage of UAV surveys. To bridge this scale gap, this study acquired high-resolution airborne multispectral (MS) and hyperspectral (HS) data over an agricultural area of the Sanjiang Plain and integrated them with environmental covariates for SOM prediction. An improved dual-branch deep learning model, termed 1DCNN-CBAM-MLP, was developed by combining residual learning with a convolutional block attention module to capture complementary local spectral features and auxiliary background information from multi-source inputs. Three data schemes, including MS, HS, and multi-source fusion, were evaluated using traditional models (MLR, PLSR, and RF), baseline neural network models (MLP and 1DCNN), and the proposed model. Spatial discrepancy mapping was further applied to analyze cross-scale spatial differences and scale-related smoothing effects among the evaluated mapping products. Results showed that: (i) the 1DCNN-CBAM-MLP model with multi-source fusion achieved the best performance, with R2 = 0.78, RMSE = 0.32%, MAE = 0.26%, and RPIQ = 2.80, representing a modest improvement over the simpler 1DCNN-MLP variant; (ii) multi-source fusion improved R2 by 0.06 over the MS configuration and by 0.02 over the HS configuration; and (iii) most spatial discrepancies were within ±5.84%, with lower SOM values relative to the airborne-derived product concentrated near large intensified field boundaries and slightly higher values observed in degraded ridge belts. These results indicate that airborne MS-HS-environmental covariate fusion improved SOM prediction within the investigated agricultural area and represented greater local spatial variability than the coarser Sentinel-2-derived products within the evaluated product comparison.
Animal gut microbiomes—comprising bacteria, archaea, fungi, viruses, and protozoa—are fundamental to host evolution, physiology, and ecosystem resilience. This review synthesizes 21st-century advances in their diversity, spatiotemporal dynamics, and functional roles across the animal kingdom. Although high-throughput metagenomics has transformed the field, major biases remain: most studies still focus on domesticated vertebrates and fecal samples, leaving substantial “microbial dark matter” in wild hosts, invertebrates, and non-bacterial domains unexplored. We highlight how gut microbiomes mediate adaptation to environmental extremes, including hypoxia, temperature stress, and toxins, and how industrialization disrupts these communities, contributing to biodiversity loss and disease risk. We further integrate eco-evolutionary theory, multi-omics, and spatial modeling to clarify cross-kingdom interactions and functional networks. Finally, we discuss translational applications—including probiotics, fecal microbiota transplantation (FMT), phage therapy, and synthetic consortia—and emphasize the need for global collaborative initiatives, artificial intelligence (AI)-driven discovery, and standardized databases to unlock the full potential of animal gut microbiomes for biodiversity conservation, climate resilience, and planetary health in the coming decades.
Recent years have witnessed increasing damage to water ecosystems, elevating the importance of environmental protection. The activation of persulfate using modified biochar materials has attracted widespread attention due to its efficient generation of highly oxidative reactive species, enhanced degradation of refractory pollutants, facile preparation, and environmental benignity. However, comprehensive analyses of research trends and advances in this field remain scarce. To address this gap, this study retrieved relevant publications from the Web of Science Core Collection database (2011—2024). After rigorous screening, 839 research articles were selected for bibliometric visualization analysis using CiteSpace, RStudio, and VOSviewer. Through temporal analysis of publication outputs, identification of top 10 highly cited papers, keyword co-occurrence clustering, and timezone mapping, this study systematically reveals the evolutionary patterns and research hotspots in this field. By examining cutting-edge research, analyzing current frontiers, and synthesizing development trends, this work provides valuable insights and guidance for future research directions.
