Soil organic matter (SOM) in cultivated land is vital for land quality and food security. This study examines SOM distribution and influencing factors in northeastern China, providing insights for sustainable agriculture. Utilizing 10 m resolution SOM data, the analysis covers regions including the Greater and Lesser Khingan Mountains, Liaohe Plain, Sanjiang Plain, Songnen Plain, the northwest semi-arid region, and the low hilly areas of Paektu Mountain. The Geodetector method is employed to assess various influencing factors. The key findings are as follows: (1) The average SOM content in Northeast China (37.70 g/kg) surpasses the national average, is highest in the Greater and Lesser Khingan Mountains (49.32 g/kg), and lowest in the northwest semi-arid region (26.15 g/kg). (2) SOM content is maximized in regions with high altitudes, steep slopes, low temperatures, and moderate precipitation. (3) The annual average temperature is the primary factor influencing SOM distribution, with a combination of temperature and administrative divisions providing better explanatory power. (4) SOM trends vary across protected areas, with slope being critical in semi-humid plains, elevation in arid regions, and no dominant factors identified in the Sanjiang Plain. These findings underscore the need for tailored black soil protection policies to effectively leverage local resources and preserve ecosystem integrity.
Soil organic matter (SOM) is essential for assessing land quality and enhancing soil fertility. Understanding SOM spatial and temporal changes is crucial for sustainable soil management. This study investigates the spatial and temporal variations and influencing factors of SOM content in the Songnen Plain (SNP) and Sanjiang Plain (SJP) of Heilongjiang Province, China, based on high-precision SOC content data (RMSE = 4.84 g/kg−1, R2 = 0.75, RPIQ = 2.43) from 1984 to 2021, with geostatistical analyses and geodetector models. This study aims to quantitatively reveal and compare the long-term spatial and temporal characteristics of SOM changes and their influencing factors across these two plains. The results show that SOM content in both plains has decreased over the past 37 years. In the SNP, the average SOM decreased from 48.61 g/kg to 45.6 g/kg, representing a reduction of 3.01 g/kg, or a 6.10% decrease; SOM decreased spatially from northeast to southwest, covering 63.1% of the area. In the SJP, the average SOM declined from 48.41 g/kg to 44.31 g/kg, a decrease of 4.1 g/kg, or an 8.50% decrease; no pronounced spatial pattern was observed, but the declining area comprises 67.49%. Changing SOM hotspots are concentrated in southern SNP and central and northwestern SJP, showing clear heterogeneity across counties. Geodetector model analysis indicates annual mean temperature as the primary driver of SOM variations in SNP; while elevation is the main driver in SJP, the combined explanatory power of multiple factors surpasses individual ones. There is a positive correlation between SOM and temperature in SNP, and policy protection positively influences SOM in both plains. These findings provide insights into the differential protection of SOM in SNP and SJP.
Soil organic matter (SOM) plays a critical role in maintaining soil fertility, sustaining ecosystem stability, and mitigating climate change impacts, making its conservation essential for agricultural sustainability. However, systematic county-level assessments of SOM spatial heterogeneity and its drivers across Northeast China remain limited, constraining region-specific soil management strategies. Understanding the spatial distribution and drivers of SOM is therefore vital for effective black soil protection in Northeast China. This study investigated the spatial heterogeneity and driving mechanisms of SOM in Northeast China, covering 289 counties across Heilongjiang, Jilin, and Liaoning Provinces. High-resolution (10 m) SOM data combined with 15 natural, climatic, soil, vegetation, and socioeconomic variables were analyzed using spatial autocorrelation (global and local Moran’s I) and the Geodetector model. Results showed that SOM exhibited a clear spatial pattern of “higher in the north and east, lower in the south and west,” with significant spatial clustering (Moran’s I = 0.730, p < 0.001). At the regional scale, climate factors were the dominant drivers, with potential evapotranspiration (q = 0.810) and mean annual temperature (q = 0.794) exerting the strongest explanatory power. At the provincial scale, dominant factors varied: topographic controls in Liaoning, climate–topography interactions in Jilin, and climate dominance in Heilongjiang. Anthropogenic footprint had limited overall influence but showed amplifying effects in certain local areas. These findings highlight the multi-scale, multi-factor nature of SOM heterogeneity and underscore the need for region-specific management strategies.
Soil organic matter (SOM) is essential for ecosystem health and agricultural productivity. Accurate prediction of SOM content is critical for modern agricultural management and sustainable soil use. Existing digital soil mapping (DSM) models, when processing temporal data, primarily focus on modeling the changes in input data across successive time steps. However, they do not adequately model the relationships among different input variables, which hinders the capture of complex data patterns and limits the accuracy of predictions. To address this problem, this paper proposes a novel deep learning model, 2-Channel Network (2C-Net), leveraging sequential multi-temporal remote sensing images to improve SOM prediction. The network separates input data into temporal and spatial data, processing them through independent temporal and spatial channels. Temporal data includes multi-temporal Sentinel-2 spectral reflectance, while spatial data consists of environmental covariates including climate and topography. The Multi-sequence Feature Fusion Module (MFFM) is proposed to globally model spectral data across multiple bands and time steps, and the Diverse Convolutional Architecture (DCA) extracts spatial features from environmental data. Experimental results show that 2C-Net outperforms the baseline model (CNN-LSTM) and mainstream machine learning model for DSM, with R2 = 0.524, RMSE = 0.884 (%), MAE = 0.581 (%), and MSE = 0.781 (%)2. Furthermore, this study demonstrates the significant importance of sequential spectral data for the inversion of SOM content and concludes the following: for the SOM inversion task, the bare soil period after tilling is a more important time window than other bare soil periods. 2C-Net model effectively captures spatiotemporal features, offering high-accuracy SOM predictions and supporting future DSM and soil management.