Spatial and Temporal Variations in Soil Organic Matter and Their Influencing Factors in the Songnen and Sanjiang Plains of China (1984–2021)

Abstract

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.

1. Introduction

Soil organic matter (SOM) is a crucial indicator for evaluating soil fertility and quality, serving as an essential material basis for plant growth and development. It plays a clear role in regulating the global soil carbon cycle, mitigating the greenhouse effect, maintaining the stability of ecosystem functions, and improving crop yields [1,2]. However, recent studies indicate that the SOM content has been declining across many countries and regions due to population growth, climate change, and land use changes [3,4]. This decline adversely affects the physical properties of soil and disrupts the cycling of essential nutrients, such as nitrogen, phosphorus, and sulfur, ultimately leading to reduced crop yields [5]. Therefore, monitoring large-scale and long-term changes in SOM in arable land and investigating the main driving factors are critical for guiding soil fertility management, ensuring food security, and protecting arable land [6].

Most studies on the spatial and temporal changes in SOM have concentrated on critical global food production regions, areas with complex climatic conditions, fragile ecosystems, or regions with clear land-use changes. These studies have examined the specific relationships between SOM decline and various natural and economic factors [7,8,9]. As the world’s most populous country, China plays a pivotal role in global agricultural production, which is essential for meeting food demands and maintaining social stability. Consequently, research on SOM changes has garnered obvious attention across soil science, land science, environmental science, and social science [10].

The Songnen Plain (SNP) and Sanjiang Plain (SJP) in Northeast China are key areas for black soil distribution and major grain production. These regions have been focal points for studying SOM changes, demonstrating that SOM content can increase or decrease in a specific direction and form banded or patchy distributions under natural or social influences [11,12]. However, previous studies have often targeted specific hotspot areas, developed agricultural regions, or urban peripheries, resulting in limited scope [13,14,15]. This focus is due to the relative ease of obtaining small-scale SOM data through soil surveys, surface soil sample collections, and the use of the Kriging interpolation to generate SOM content maps. These methods facilitate the comparative analysis of SOM spatial and temporal evolution patterns at a lower cost [11,16]. For instance, some researchers collected soil samples in the SNP over several years to estimate SOM content in agricultural fields and analyze its spatiotemporal variations. Their results show an increasing trend in surface SOM from 2008 to 2011 in this region [17]. Other studies focused on the western part of SNP, examining the spatiotemporal variations in SOM in saline soils and their influencing factors. These findings indicate a declining trend in SOM from 1982 to 2022, with topography being the primary factor affecting surface SOM changes. Additionally, the effects of environmental humidity and land use on SOM increase with soil depth [18]. In the SJP, most studies analyze the spatial variations in SOM in specific areas, investigating change characteristics and influencing factors [19,20]. Some studies have explored the impact of farmland expansion on SOM spatial variations. These studies reveal that the SOM content in SJP farmland decreased from 1992 to 2012 as a result of expansion. [21]. Remote sensing (RS) provides spatially distributed and reproducible data on a large regional scale, offering clear advantages for studying SOM spatial and temporal changes, including the impacts of soil structure destruction and land management practices [22,23]. With advancements in machine learning algorithms, such as Random Forest, researchers can now combine RS data to map SOM distribution and investigate its spatial-temporal change patterns. This approach has become a new trend in the field [13]. Additionally, geostatistics offers unique advantages in visualizing and analyzing SOM spatial-temporal data, effectively identifying and expressing areas of SOM change based on existing data. Several studies have used geostatistical methods to reveal SOM variability at the regional scale in China [24,25,26]. In conclusion, RS mapping and geostatistical methods not only provide natural advantages in creating SOM distribution maps but also enable more intuitive representations of SOM spatial and temporal variations. Scholars have utilized 191,465 Landsat TM and OLI images, along with elevation model data, to calculate spectral indices characterizing soil formation information. Using the ML-CNN prediction model, they simulated global soil organic carbon from 1984 to 2021, creating content distribution maps (RMSE = 4.84 g/kg⁻¹, R² = 0.75, RPIQ = 2.43) [27]. Since soil organic carbon (SOC) is a crucial component of SOM, typically calculated as SOM (g/kg) = SOC (g/kg) × 1.724 [28,29], these findings provide a foundation for analyzing SOM spatial and temporal variations and exploring influencing factors on a large-scale, over a long-term basis in Heilongjiang Province, China.

Numerous studies have investigated the factors influencing SOM, including flat topography and favorable slope orientation for microbial nutrient storage in soil [30]. Large-scale changes in climate and precipitation also directly impact SOM decomposition [31,32,33]. Some studies have explored the influence of social factors, such as population and gross national product, on changes in SOM content [34]. Additionally, the conversion of dry land to paddy fields and differences between state-owned and family farms have been suggested to affect SOM [35]. However, previous research has often focused on the single-factor effects of spatial and temporal changes in SOM [36]. Commonly used methods include correlation analysis, back-propagation neural networks, and others [37]. However, SOM changes typically result from the interaction of multiple factors, both natural and anthropogenic [38]. Although some studies have examined the combined effects of structural and stochastic factors on SOM changes [39], the main influencing factors vary between study areas due to differing natural and economic conditions, even among regions at the same latitude. Geodetector models can detect the spatial heterogeneity of geographic phenomena at large scales, identifying the contribution of individual factors and their interactions and overcoming the limitations of previous methods [40]. Therefore, this study chooses elevation, slope, aspect, annual precipitation, annual average temperature, population density, gross national product, farm zoning, and dry and wetland zoning as the main influencing factors for SOM content changes in the SNP and SJP.

Heilongjiang Province, a key commercial grain production base in China, has experienced a gradual decline in arable SOM content, particularly in SNP and SJP, under continuous farming conditions. This loss reduces the soil ecosystem function, affecting soil health and, ultimately, food production [41]. SNP and SJP differ in natural and socio-economic characteristics, with SNP having a more complex topography and a longer history of economic development and reclamation [32,33]. Conversely, SJP has a shorter history of reclamation and flatter terrain. These differences result in varying SOM changes between the two regions. Thus, understanding the spatial and temporal characteristics of SOM in both regions, along with their influencing factors, is crucial for protecting and sustainably using black soil resources. However, due to limited data on SOM content over large geographical areas, no scholars have conducted such analyses and comparisons. This study is structured as follows: Part 2 describes the study area, data sources, and methods; Part 3 explores the spatial heterogeneity of SOM distribution and changes; and Part 4 discusses and summarizes the study results.

2. Materials and Methods

2.1. Study Area

The study area encompasses the SNP and the SJP within Heilongjiang Province (Figure 1). The SNP, situated in the southwestern part of Heilongjiang Province, spans from 122°16′ E to 128°19′ E and 43°06′ N to 50°17′ N, covering an area of 211,400 km² [42] and comprising 34 administrative districts. Characterized by high elevation in the northeast and low elevation in the southwest, the plain experiences a temperate continental semi-humid and semi-arid monsoon climate, with an average annual precipitation of 400–600 mm and an average annual temperature of −2 to 7 °C. Corn and soybeans dominate crop cultivation, serving as the primary food production bases of Heilongjiang Province. Additionally, the SNP boasts various soil types, predominantly black soils such as black calcium soils and meadow soils [43,44]. On the other hand, the SJP, located in the eastern part of Heilongjiang Province, covers an area of 214,000 km² [42], ranging from 129°11′ E to 135°05′ E and 43°49′ N to 48°27′ N, encompassing 22 administrative districts. The terrain features low elevation in the northeast and high elevation in the southwest, and the region experiences a temperate humid climate, with an average annual temperature of 3 °C and annual precipitation ranging from 500 to 600 mm. The SJP primarily consists of four main soil types: meadow soil, swamp soil, and black soil, collectively occupying over 95% of the total area. Economically, agricultural production activities dominate, with rice and maize serving as the main crops [45].

It is evident that the two plains exhibit clear differences in both natural and social factors. Therefore, comparing the organic matter content in arable land between the two plains and investigating influencing factors are necessary measures to further protect the black soil resources in this region.

2.2. Materials

The SOM data used in this study mainly derive from Meng’s soil organic carbon (SOC) prediction results, which have been adjusted by a specific factor. These data were gathered using cloud-free TM/OLI Landsat satellite images from Google Earth Engine (GEE) spanning seven bands from 1984 to 2021 and totaling 191,465 images. These images were processed with the LaSRC atmospheric correction and composite radiometric calibration algorithms, along with elevation data, to create a high-resolution global SOC dataset (GMR-MCNN) for black soil regions. In addition, this study proposed a local strategy (LS) to reduce the high heterogeneity of soil organic carbon content and the impact of environmental variables on prediction outcomes. In addition, this study introduces a meta-learning convolutional neural network (ML-CNN) model to tackle SOC content heterogeneity and environmental impact on predictions. The ML-CNN model integrates meta-learning (ML) with convolutional neural networks (CNNs), enhancing adaptability to new tasks, improving knowledge transfer, and reducing reliance on labeled data. ML enables CNN to automatically adjust learning rates and hyperparameters, boosting model performance and efficiency. Our study applies this model to address regional variations in SOC content. It divides the study area into four regions based on molar concentration boundaries, creating four meta-tasks. Meta-learning fine-tunes the model for each meta-task, improving predictions even with limited calibration data. The ML-CNN model achieves high accuracy with an RMSE of 4.84 g/kg, R² of 0.75, and RPIQ of 2.43, outperforming conventional CNNs and the RFE-RF model. It excels in providing high-resolution, accurate SOC predictions for the global Mollisol region and efficiently processes time-series data [27].

Topographic factors, comprising the digital elevation model (DEM), slope, and aspect, were sourced from Google Earth Engine, with slope and aspect derived from DEM data. Climatic factors (precipitation and temperature data) and socio-economic factors (GDP and population data) were retrieved from the Resources and Environment Data Platform of the Chinese Academy of Sciences (REDP). The annual spatially interpolated dataset of meteorological factors was formulated based on yearly values from over 2400 stations nationwide, utilizing the anuspl interpolation software developed by Australian scientist Hutchinson. This software employs a thin-plate smoothing spline function, which is a well-established tool for meteorological data interpolation. The incorporation of covariates, such as elevation, enhances the accuracy and smoothness of the interpolation outcomes. This dataset encompasses annual temperature and precipitation data interpolated onto a 1 km raster grid, spanning from 1960 onwards. For this study, data from 1980, 2000, and 2020 were utilized. China’s population spatial distribution dataset spans six periods: 1995, 2000, 2005, 2010, 2015, and 2019. These data are presented in 1 km raster grids, with each grid representing the population within that square kilometer. Similarly, China’s GDP spatial distribution dataset, also in 1 km raster grids, includes data from the same six periods. This dataset was generated through the spatial interpolation of national GDP statistics by county, incorporating spatial interactions among land use types, nighttime light brightness, and settlement density data, which are closely linked to human activities and GDP. Each grid represents the total GDP output of ten thousand yuan per square kilometer [46]. Farm management zoning data were obtained from unpublished records from the Northeast Institute of Geography and Agroecology, Chinese Academy of Sciences. Dry and wet field partitioning data were derived from Zhang’s research, which introduces a flexible framework for climate-assisted-supervised rice (PSPR) mapping using GEE. This framework leverages the automation of phenological assessments, the generation and refinement of training data, and the all-season classification capabilities of machine learning techniques. The methodology was validated by generating 30-meter high-resolution rice maps for Heilongjiang Province, China, covering the period from 1990 to 2020. These rice maps were validated against a comprehensive reference sample, four existing rice products, and current agricultural statistics. The results demonstrated enhanced performance and a highly linear relationship, with an average R² of 0.993 compared to previous studies [47]. This product provides a valuable tool for analyzing the effects of water and dry field zoning on changes in SOM. The types and sources of data for the factors used are shown in

Table 1. Data types and sources.

Data	Type	Unit	Resolution	Source
SOM	Raster	g/kg	30 m	https://www.sciencedirect.com accessed on 1 January 2024
DEM	Raster	m	30 m	https://earthengine.google.com/ accessed on 1 January 2024
SLP	Raster	°	30 m	/
APC	Raster	°	30 m	/
PRE	Raster	mm	1 km	http://www.resdc.cn/ accessed on 1 January 2024
TEMP	Raster	°	1 km	http://www.resdc.cn/ accessed on 1 January 2024
GDP	Raster	million	1 km	http://www.resdc.cn/ accessed on 1 January 2024
POP	Raster	million	1 km	http://www.resdc.cn/ accessed on 1 January 2024
PDZ	Vector	/		/
FAZ	Raster	/	30 m	https://github.com accessed on 1 January 2024

SOM represents soil organic matter; DEM represents the elevation; SLP represents the slope; APC represents the aspect; PRE represents the average precipitation; TEMP represents the annual temperature; GDP represents the gross domestic product; POP represents the population; PDZ represents the Paddy and dry field zones; and FAZ represents the farm zoning.

.

To study the influence of spatial and temporal changes in SOM, all data were converted to annual average changes from the 1980s and 1990s to around 2020 using a raster calculator. These data were then masked to raster data covering the range of SNP and SJP. Given that the geodetector method is suitable only for processing discrete data variables [48], nine continuous variables were converted to discrete variables in this study. The levels of discretization were determined based on the inherent properties of the data. To ensure the objectivity of the results, raster data such as topography, slope, slope direction, average annual temperature, average annual precipitation, GDP, and annual population changes were uniformly classified into eight discrete classes. After rasterizing the farm management vector data, areas belonging to farms were classified as rank 1, and areas not under farm management were classified as rank 2. The same method was applied to dry and wet field zoning, where dry field areas were classified as rank 3 and water field areas were classified as rank 4. The spatial distributions of the ranks of all factors are demonstrated in this study, as shown in Figure 2.

2.3. Methodology

2.3.1. Calculation of SOM Content

For the calculation of SOM content, we used Equation (1) to express the relationship between SOM and soil organic carbon [49,50] and obtained the results of SOM calculations for the two periods of SNP and SJP 1984–1990 and 2016–2021, as shown in Figure 3.

$$\mathrm{SOM}=\mathrm{SOC}\times 1.724$$

(1)

where SOM is the content of SOM in 1984–2021. SOC is the content of soil organic carbon in 1984–2021.

2.3.2. Calculation of the Amount of Change in SOM

Based on the spatial distribution of SNP and SJP SOM in 1984–1990 and 2016–2021, the overall magnitude of change in SNP and SJP SOM in 1984–2021 was calculated.

$$\Delta \mathrm{SOM}=\mathrm{SOM}1984-\mathrm{SOM}2021$$

(2)

where SOM is the magnitude of change in SOM from 1984 to 2021. SOM1984 and SOM2021 represent the SOM content from 1984 to 1990 and 2016 to 2021, respectively.

2.3.3. Coefficient of Variation

The CV index is a normalized measure of data dispersion, defined as the ratio of the standard deviation to the mean. This ratio allows for the direct comparison of datasets with different units and ranges. This standardized process not only helps assess the internal variability within a dataset but also enables the effective comparison of variability across different datasets. Therefore, the CV index can be used to analyze the degree of variability in long-term spatial changes in SOM [51]. The formula for calculating the CV index is as follows:

$$C_{\nu }={\displaystyle \frac{1}{SOM}}\sqrt{{\displaystyle \frac{{{\displaystyle \sum }}_{i=1}^n{\left(SO{M}_i-\mathrm{S}\overline{O}M\right)}^2}{n-1}}}$$

(3)

where Cv is the coefficient of variation; SOM_i is the SOM value in year i; S$\overline{O}$M is the mean value of SOM from 1984 to 2021; and, in this study, N was taken as 37.

2.3.4. Hotspot Analysis

Hotspot analysis is a spatial statistical method for identifying clear clusters in geospatial data, usually clustered areas of hot or cold values. In this paper, we analyze these points using the Getis-Ord Gi* statistic, which is the ratio of the local sum of a feature and its neighbors near a certain distance or scale of analysis compared to the sum of all features. A statistically obvious z-score occurs when the local sum is different from the estimated local sum and clearly varies due to random chance [52,53]. The following is the mathematical formulation of the method:

$${\mathrm{Gi}}^{*}={\displaystyle \frac{{{\displaystyle \sum }}_{\mathrm{j}=1}^{\mathrm{n}}{\mathrm{w}}_{\mathrm{i},\mathrm{j}}\left(\mathrm{D}\right){\mathrm{X}}_{\mathrm{j}}-\overline{\mathrm{X}}{{\displaystyle \sum }}_{\mathrm{j}=1}^{\mathrm{n}}{\mathrm{w}}_{\mathrm{i},\mathrm{j}}}{\mathrm{S}\sqrt{\frac{\mathrm{n}{{\displaystyle \sum }}_{\mathrm{j}=1}^{\mathrm{n}}{\mathrm{w}}_{\mathrm{i},\mathrm{j}}^2-{\left({{\displaystyle \sum }}_{\mathrm{j}=1}^{\mathrm{n}}{\mathrm{w}}_{\mathrm{i},\mathrm{j}}\right)}^2}{\mathrm{n}-1}}}}$$

(4)

where x_j is the attribute value of feature j; W_i,_j is the spatial weight between feature i and j; and N is equal to the total number of features.

2.3.5. Geodetector

The geodetector is a set of statistical methods, a statistical model to detect spatial heterogeneity and reveal the driving factors behind it, which can be used to effectively analyze the relative importance of different influencing factors on the target and detect regional spatial heterogeneity to avoid the interference of human subjectivity in the selection and calculation of evaluation factors [54]. In this study, based on the analysis of the spatial and temporal variability characteristics of SOM in SNP and SJP, a total of nine factors covering natural and socio-economic aspects were selected for factor detection and interaction detection. The goal was to quantify the effects of these factors and their interactions on the spatial and temporal changes in SOM content over the past 37 years in SNP and SJP and to clarify the driving mechanisms behind the spatial and temporal variability of soil organics in these two regions. The geodetector model includes four probes as follows: the empirical measurement of the factors is mainly used to detect the spatial division rows of the dependent variable Y and to detect the strength of the effect of the independent variable X on the spatial division rows of Y, denoted by q; the risk zone detection is used to detect whether there is a clear difference in the average values of the attributes between the two subregions; the ecological detection is used to detect whether there is a clear difference in the spatial distribution of the attribute Y between the two factors; and the interaction detection is used to detect whether there is a clear difference in the spatial distribution of the attribute Y between the two factors. The interaction test was used to detect the changes in the influence of two factors on attribute Y relative to a single factor after interactions. In this paper, factor detection and interaction detection were chosen for the study.

Divergence and factor detection are used to detect the spatial divergence of the dependent variable Y (SOM content of arable land) and the degree of explanatory power of each influential factor X on the spatial divergence of the SOM content in arable land. It is expressed in terms of the q-value with the following expression:

$$\mathrm{SSW}={\displaystyle {{\displaystyle \sum }}_{\mathrm{h}=1}^{\mathrm{L}}}{\mathrm{N}}_{\mathrm{h}}{\mathsf{\sigma }}_{\mathrm{h}}^2, \mathrm{SST}={\mathrm{N}\mathsf{\sigma }}^2$$

(5)

$$\mathrm{q}=1-{\displaystyle \frac{{{\displaystyle \sum }}_{\mathrm{h}=1}^{\mathrm{L}}{\mathrm{N}}_{\mathrm{h}}{\mathsf{\sigma }}_{\mathrm{h}}^2}{{\mathrm{N}\mathsf{\sigma }}^2}}=1-{\displaystyle \frac{\mathrm{SSW}}{\mathrm{SST}}}$$

(6)

where h = 1, 2, 3..., L denotes the stratification of the independent variable Y or the dependent variable X or the grading or classification used. N_h and N are the number of cells in stratum h and the whole district, respectively. And ${\sigma }_h^2$${\sigma }^2$ are the variance in the values of Y in stratum h and the whole region, respectively. SSW is the sum of the variances within the stratum, and SST is the total variance of the whole region. q has a value range of [0,1]. The larger the value of q, the more pronounced the spatial differentiation of the dependent variable Y is. If the grading is generated by the independent variable X, then the value of q represents the factorial explanatory power of the independent variable X on the dependent variable Y. The larger the value of q, the more the explanatory power of the independent variable X is on the dependent variable Y.

The interaction detector is used to measure the interaction between two factors to assess issues such as whether the joint action between two factors enhances or weakens the explanatory power of the independent variable Y or whether the two effects on Y are independent of each other. The interaction categories can be classified as follows.

Nonlinear attenuation:

$$q\left(X1{\displaystyle \cap }X2\right)<\mathrm{Min}\left(q\left(X1\right),q\left(X2\right)\right)$$

(7)

Single factor nonlinear attenuation:

$$\mathrm{Min}\left(q\left(X1\right),q\left(X2\right)\right)<q\left(X1{\displaystyle \cap }X2\right)<\mathrm{Max}\left(q\left(X1\right),q\left(X2\right)\right)$$

(8)

Double factor enhancement:

$$q\left(X1{\displaystyle \cap }X2\right)>\mathrm{Max}\left(q\left(X1\right),q\left(X2\right)\right)$$

(9)

Mutual independence:

$$q\left(X1{\displaystyle \cap }X2\right)=q\left(X1\right)+q\left(X2\right)$$

(10)

Nonlinear enhancement:

$$q\left(X1{\displaystyle \cap }X2\right)>q\left(X1\right)+q\left(X2\right)$$

(11)

3. Results

3.1. Dynamics of SOM in SNP and SJP

3.1.1. Differences in Spatial and Temporal Changes in SOM in SNP and SJP

We selected the two time periods, 1984–1990 and 2016–2021, for SOM extraction in the SNP and SJP regions, with the results depicted in Figure 4. The findings from the two subplots indicate clear changes in SOM, ranging from 35 g/kg to 56 g/kg, over time. Specifically, there is an increase in the proportion of the SOM content within the 42 g/kg–49 g/kg range and a decrease within the 35 g/kg–42 g/kg range. There are also some differences in the regional distribution of SOM between the two plains. In the SNP, the SOM distribution exhibits a “high in the northeast and low in the southwest” pattern across both periods. Higher SOM areas are concentrated in the northeastern part of the region, where the terrain is elevated and temperatures are lower. In contrast, the Songhua River basin, characterized by lower terrain and higher temperatures, shows relatively lower SOM reserves. Over time, areas with a lower SOM content in the SNP tend to expand towards the southeastern low-lying plains and regions with higher population concentrations. This suggests a link between SOM distribution and factors such as elevation, temperature, and population density. In the SJP, the SOM distribution during both periods shows a “high in the middle and low in the surroundings” pattern. Higher SOM areas are located in the flat plains at the center of the study area, while lower SOM areas are found in the northern part of the study area, characterized by higher latitude and lower temperatures, as well as in the southern part, where the terrain is more undulating. However, a comparison between the two periods reveals that the higher SOM areas in the center changed from “concentrated” to “scattered” over time, and the lower SOM areas in the northeast gradually expanded. This indicates that the overall SOM content in the SJP region has declined more rapidly over the last 37 years.

In recent years, influenced by global climate change and China’s reform and opening-up policies, SNP and SJP have emerged as key regions for the distribution of black soil in China. Concurrently, SOM content has exhibited clear changes. Using Equations (2) and (3), this study calculated the changes in SOM over 37 years and the CV index in two plains. The results revealed distinct trends and spatial distributions of SOM changes in SNP and SJP.

The variation in SOM content from 1984 to 2021 shows notable spatial heterogeneity in SNP and SJP (Figure 5a). In the SNP, increases in soil organic matter are primarily concentrated in the western and eastern regions with varied topography, while the northern mountainous areas with lower temperatures show insignificant changes, suggesting a correlation between SOM distribution and temperature and elevation. In contrast, overall changes in SJP over the past 37 years are less pronounced, mainly characterized by changes in the central and northern parts and minor changes in the southern parts. Areas where SOM has increased are predominantly located in surrounding flat plains, whereas declines in SOM are observed in the central and southern mountainous areas, indicating an association between SOM distribution and changes in elevation. As shown in (Figure 5b), the CV index for SNP and SJP ranges from 0.84 to 1.49. SNP exhibits greater variability in the central-southern regions, while the spatial distribution of SOM in the east has remained relatively stable over time. SJP shows higher variability in the northern region, with minimal obvious changes in SOM spatial distribution in the central and southern regions over the past 37 years. In terms of time series analysis, the study finds fluctuating declining trends in maximum, minimum, and average soil organic matter values for both plains (Figure 5c). Analysis of change in area proportions reveals that the proportion of declining areas is 63.1% for SNP and 67.49% for SJP (Figure 5d), indicating a more obvious decrease in organic matter in SJP compared to SNP. The larger share of the area that experienced a decrease in SJP aligns with the finding that the SOM reduction in SJP was greater than in SNP.

3.1.2. Analysis of Hot Spots

However, the results of SOM changes at the two time points provide only a general overview of the changes in the entire SNP and SJP. These overall increases and decreases do not elucidate the specific changes occurring within each region of the two plains, particularly at the county level. They also do not provide insights into the direction and process of these changes. To understand the specific spatial characteristics of SOM changes, more detailed information about the change process is necessary. Therefore, we employed hotspot analysis to further investigate the spatial correlation of SOM in the two plains at the county level from 1984 to 2021. Using Equation (4), we calculated the Getis-Ord Gi* indices for counties in the periods 1984–1990, 2016–2021, and 1984–2021. We then categorized the statistical values into hotspot zones, unimportant zones, and coldspot zones. The results of this analysis are presented in Figure 6.

From 1984 to 1990, the distribution of SOM in SNP displayed a “striped” pattern across the counties, with hotspots mainly in the northeastern part of the study area. In contrast, cold spots were concentrated in Longjiang County to the west, Zhaozhou County to the south, and ten other districts and counties. Concurrently, SJP’s SOM displayed a “block-like” distribution, with hotspots concentrated in nine districts and counties, such as Youyi and Jixian in the central part, and coldspot areas scattered across nine districts and counties, including Fuyuan and Yilan around the plain. In the period from 2016 to 2021, the coldspot areas and hotspot areas of SNP exhibited an expanding trend. Hotspot areas in the west extended northward, while coldspot areas in the east expanded towards the central part of the plain. The changes in coldspot areas in SJP were even more pronounced, with former coldspot areas in the western part of the plain now covered by hotspot areas. The main hotspot area in the central part contracted, with the main cold spots distributed around Jidong County in the south-central part of the plain. Overall, the hot spots and cold spots of SOM changes in SNP from 1984 to 2021 are concentrated in the southern part of the plain. Hotspots are concentrated in Daqing and Lindian in the southwest and Wuchang in the south, while Longjiang County in the west and Zhaozhou and Harbin in the center represent cold spots of these changes. The hotspots of SJP are more dispersed and concentrated in the central part of the region, such as Youyi Hulin and Jixi, while the cold spots of changes are located in the middle part of the plain, including Fuyuan in the north and Hegang in the northwest.

3.1.3. Differences in SOM Changes in Different Districts and Counties

From the previous hot and cold spot analysis results, clear disparities in SOM changes at the county level between the SNP and the SJP are evident. In SNP, the cold spots over the past 37 years are primarily located in the northwest and central counties, while the hotspots of change are concentrated in the western counties. Conversely, in SJP, the hotspots of SOM changes are situated in the central and southern regions, with cold spots predominantly found in the northeast and northwest parts. Six counties—Longjiang, Zhaozhou, Wuchang, Fuyuan, Youyi, and Hulin—were selected as obvious units for comparing SOM changes between the two plains, as shown in Figure 7.

Analyzing specific counties reveals nuanced patterns. In Longjiang County, the SOM content during the period of 1984–1990 was generally low, exhibiting a decreasing trend from northwest to southeast over time. Notably, recent protection policies may have contributed to the recent increase in SOM content in this county. Conversely, in Zhaozhou, although SOM content was high during 1984–1990, recent years have witnessed an obvious decline, particularly in the western part, possibly due to erosion caused by the nearby Songhua River. Similarly, SOM in the Wuchang area has also declined notably over the past 37 years, primarily in hilly and mountainous regions, which is likely attributable to changes in altitude. In contrast, in SJP, SOM in the northernmost Fuyuan area and southeastern Hulin area has exhibited an overall decreasing trend over time, particularly in the northern and central parts, potentially influenced by temperature differences between these regions. However, the friendship area has shown an overall increasing trend, with only the central part experiencing a decline, possibly linked to state-run farm policies favoring increases in organic matter.

3.2. Driver Analysis

To capture the spatial characteristics of independent variables sourced from different datasets, we established a 10 km × 10 km grid covering the entire cultivated area in the SNP and SJP regions. In each grid, we extracted the average SOM content as the dependent variable and collected data on elevation, slope, slope direction, annual precipitation, annual average temperature, population density, GDP level, farm zoning, and water and dry land distribution as independent variables. Considering the characteristics of the data sources, we classified each variable into eight levels using the natural breakpoint method. Outliers and null values were removed, and the data were organized into an Excel table in sequential order. To optimize continuous variables, we employed both the quartile breakpoint method and the natural breakpoint method, adhering to the principle of maximizing the q-value. Subsequently, the discretized continuous variables were inputted into the geodetector model package in R language for analysis, aiming to attain optimal detection outcomes.

3.2.1. Single Factor Detection Analysis

From the single-factor driving results on SOM content in the two plains (Figure 8), it is evident that the impact of selected driving factors on SOM content in both plains is relatively unclear. In the SNP, the average influence of driving factors on SOM content over 40 years follows this order: average annual air temperature (X5) > slope (X2) > elevation (X1) > average annual precipitation (X4) > aspect (X3) > average annual population (X7). The driving force exerted by average annual air temperature is the strongest at 0.213, followed by slope orientation with an average driving force of 0.072. Conversely, the driving forces of all other factors on SOM content are considerably less than 0.1, indicating that climatic factors, particularly air temperature, exert the most dominant influence on SOM content changes in SNP. In SJP, the average influence of driving factors on SOM content over 37 years ranks as follows: elevation (X1) > slope (X2) > temperature (X5) > GDP (X6) > aspect (X3) > farm zoning (X8) > precipitation (X4). Elevation and slope exhibit the strongest driving forces on SOM content changes in SJP, with elevation having an average driving force of 0.135 and slope following closely at 0.1. Conversely, the driving forces of other factors on SOM content changes are all below 0.1, indicating their relatively weaker explanatory power compared to elevation and slope. These findings suggest that while SNP and SJP share the same geographical dimension, their spatial and temporal changes in SOM content differ. In SNP, characterized by relatively undulating terrain, elevation and slope exert a clear influence on SOM content, with average annual temperature playing a key role in driving changes over the past 37 years. In contrast, in SJP, where the terrain is gentler, elevation emerges as the primary driver of SOM content, with temperature exhibiting a comparatively weaker driving force of 0.0728 compared to SNP. Notably, the driving force of farm management zoning and dry and wet field partitioning is not clear in either plain, suggesting their limited role in influencing spatial and temporal changes in SOM content.

3.2.2. Interaction Factor Detection

Based on the results of the interaction test between the two plains and the factors, it was observed that the interaction effect of any two elements on the spatial distribution of SOM surpassed that of a single element, as shown in Figure 9. Among these interactions, SNP exhibited a clear interaction power between (X5) average annual temperature change and other factors. Specifically, the interactions between (X5∩X4) average annual temperature and average annual precipitation, (X5∩X1) average annual temperature and elevation, (X5∩X7) average annual temperature and average annual population change, (X5∩X2) average annual temperature and slope, and (X5∩X6) average annual temperature and GDP were observed with powers of 0.351, 0.288, 0.286, 0.268, and 0.262, respectively. This underscores once again the significance of climatic factors in influencing SOM content in the SNP compared to other factors over time, which is a phenomenon closely tied to the formation and dynamics of SOM and the geographical location of the SNP within China’s high-dimensional belt. Apart from the reinforcing effect between (X5∩X6) average annual temperature and GDP, the interactions of the other factors exhibited similar effects. With the exception of the (X5∩X6) average annual air temperature and GDP interaction, all other factors showed linear enhancement, indicating that the interaction between average annual GDP and air temperature influenced the change in SOM content. It is worth mentioning that the interaction between (X5∩X8) average annual air temperature and farm zones and (X5∩X9) average annual air temperature and annual changes in dry and wet fields were 0.22 and 0.214, respectively. The dry and wet field zones showed a bifactorial enhancement of the air temperature factor, suggesting that the SOM content of dry and wet fields in the SNP was not as obvious as that of farms. In SJP, clear interactions were observed between (X1) elevation change and other factors, particularly between (X1∩X5) elevation and annual average temperature, (X1∩X6) elevation and GDP, and (X1∩X4) elevation and annual average precipitation, with forces of 0.251, 0.216, and 0.201, respectively. The influence of several factors and elevation are all examples of nonlinear growth, indicating that the change in SOM content in the SJP does not increase or decrease with the elevation changes.

4. Discussion

4.1. Differences in Dominant Factors of SOM Changes in Different Regions

In previous studies, it was demonstrated that the overall variation in SOM content exhibited clear spatial variability under both natural and anthropogenic influences [49,55]. Our findings also indicate that, under similar conditions, different regions and study scales can result in diverse patterns of SOM distribution, corroborating the observations made by Wang et al. [56]. However, the relationship between the predominant factors of the two plains and the spatial and temporal changes in SOM remains unclear. Comparing the range of average annual air temperature changes with SOM alterations, we observed a positive correlation, as shown in Figure 10. This suggests that, as average annual air temperature increases, the magnitude of SOM changes also tends to rise gradually. Notably, in the SNP, the 0 °C temperature line serves as a clear demarcation point. Below this threshold, the average annual change in SOM remains relatively low, while above 0 °C, SOM exhibits more pronounced changes. Specifically, when the average annual temperature change falls within the range of 4–5 °C, the average annual SOM change reaches its peak, with a maximum value of 3.61 g/kg. Shi et al. analyzed the spatial and temporal responses of SOM to climatic and socio-economic factors in Heilongjiang Province using Geographic Weighted Regression (GWR). Their findings indicated that over the past 25 years, SOM has responded positively to climate and socio-economic factors. Specifically, average annual temperature has had a positive impact on SOM in the Nenjiang Plain (with a regression coefficient ranging from 0 to 0.54) over this period [36], which is consistent with our results. Conversely, in the SJP, elevation emerged as the primary factor influencing SOM variability. A fluctuating trend was observed between elevation change and average annual SOM change, with the highest SOM variation (3.61 g/kg) occurring within the elevation range of 199–234 m. The smallest average annual variation in SOM (4.07 g/kg) was observed at the 234 m elevation, while the largest variation (7.32 g/kg) was recorded between the 374 and 554 m elevation. Notably, between the 200 and 550 m elevation, the average annual variation in SOM was the most obvious, reaching a maximum of 7.32 g/kg. The fluctuations observed within the 199–234 m elevation range may be attributed to the superposition of other factors, such as slope and precipitation. Zhu et al. investigated the relationship between SOM and slope elevation by analyzing surface soil samples collected from depths of 0–60 cm. Their study indicated that SOM density tended to increase with elevation [30]. However, this conclusion may be flawed, as it does not explore the specific effects of elevation changes on SOM alterations. Similarly, Chen et al. analyzed the relationship between SOM and slope elevation in the Yarlung Zangbo River Basin, showing that SOM density increased with elevation changes [30]. Their study further demonstrated the differential effects of elevation on SOM distribution in the basin. Specifically, the upper and lower reaches of the basin experienced clear organic matter loss, while the central region exhibited higher organic matter content. This analysis aligns with our findings [57].

4.2. Policy Influence

China’s policy for the protection of black soil started late. After the founding of the country, the Northeast region, with its expansive terrain and fertile land, coupled with its limited population, was conducive to the development of mechanized agriculture. In response, especially in Heilongjiang Province, several state-run farms were established to leverage advanced agricultural production technologies and optimize farmland resources, clearly enhancing land utilization rates and production efficiency. While this has improved land utilization and productivity in Heilongjiang Province, it also led to issues such as the blind expansion of farmland, excessive use of chemical fertilizers and pesticides, and pollution from industrial plastic films. These problems, including soil erosion, land salinization, and the loss of soil chemical nutrients and physical structure, were not fully recognized at the time [42].Since the early 1990s, the concept of black soil protection began to emerge in government policy documents. Starting in 2020, the Heilongjiang provincial government allocated CNY 260 million to support pilot projects in nine counties, including Hailun, Shuangcheng, and Beilin, aimed at protecting and utilizing black soil. In 2018, the second batch of pilot projects was initiated in 15 counties, including Binxian and Qinggang, aligning with regions where changes in SOM were more pronounced in our study.

Comparing the pilot areas for black soil protection with the SNP and SJP, as shown in Figure 11, we found that under government policies, the overall growth rate of SOM in Heilongjiang Province has totaled −7.79%. Except for Huachuan County (−10.3%), which had a lower growth rate than the rest of the province, the decline in SOM in other areas was mitigated to some extent. Even in Longjiang County, SOM showed a slight increase compared to the 1980s. This suggests that under policy protection, SOM content can be improved. Moving forward, the Heilongjiang provincial government should increase policy support and expand the coverage of black soil protection pilot projects to safeguard SOM across the province, ensuring the sustainable utilization of black soil resources.

4.3. Deficiencies and Prospects

In this study, we analyzed the spatiotemporal variations in SOM in SNP and SJP in the black soil region of Northeast China and discussed the main factors influencing these changes. Nonetheless, several limitations warrant attention.

Firstly, in selecting SOM drivers, this study primarily focused on factors identified in previous research. However, recent studies have highlighted additional variables influencing SOM dynamics, such as straw incorporation, shifts in fertilizer application, and variations in soil types and land management practices. For instance, investigations in Dehui County, Northeast China, revealed substantial variations in SOM content among different soil types, with black calcareous soils exhibiting the highest SOM content and wind-formed soils the lowest [58]. However, some studies have indicated that soil types across the Northeast region do not significantly influence spatiotemporal variations in SOM [59]. Similarly, analyses of the second soil census data in Shandong Province emphasized the clear impact of land use practices alongside soil type on the spatial distribution of SOM. Furthermore, studies have suggested that the increased use of chemical and organic fertilizers could contribute to SOM accumulation, particularly in systems dominated by rice and vegetable cultivation [60,61]. Additionally, research conducted in high-production areas of Huantai County, North China, underscored the positive effect of extensive straw incorporation on enhancing SOM content, advocating for its adoption as a sustainable management practice [62]. Hence, future investigations should consider a broader array of influencing factors to comprehensively elucidate the mechanisms driving SOM changes. Secondly, the scale of analysis for identifying influencing factors and the precision of utilized data products clearly impact the accuracy of the overall findings. In this study, a 10 km × 10 km analysis scale was chosen based on the study area’s extent and data resolution. However, this decision entails a degree of subjectivity, and the chosen analysis units were derived from previous research findings. Future studies should strive for a more objective and precise analysis scale, utilizing independently derived results and verifying the accuracy of the remote sensing products employed. This approach is essential to ensure the objectivity and scientific rigor of research outcomes.

5. Conclusions

By analyzing the spatial and temporal variations and drivers of SOM content in the SNP and SJP of Heilongjiang Province, the following conclusions were drawn.

(1) Over the past 37 years, the SOM content in the SNP and SJP has exhibited an overall declining trend. The period from 1990 to 2000 saw an obvious decline, followed by a slower decline from 2000 to 2021, with the decline in the SJP generally higher than that in the SNP. Spatially, both plains demonstrated absolute heterogeneity, with the central, relatively low-lying part of the SNP and the more undulating western and southern parts experiencing more obvious changes in SOM. The central and southern parts of the SNP were the main areas of SOM loss. Conversely, the overall change in organic matter in the SJP was less pronounced, with slow changes observed in the central part and surrounding areas. Organic matter growth was concentrated in the eastern part of the SJP and the southern plains with relatively flat terrain.

(2) Additionally, clear heterogeneity in SOM changes within counties in the two plains was observed. In the SNP, hot spots of SOM change were identified in the central areas of Zhaozhou and the southern areas of Wuchang, while Longjiang County in the west was the cold spot of change. In the SJP, hot spots were more dispersed, mainly concentrated in the central regions, particularly in the districts of Youyi and Hulin, while the cold spot of change was located in Fuyuan in the north of the plain.

(3) The analysis of driving factors revealed differences in the influence of factors such as topography, temperature, and precipitation on SOM changes in the two plains. In the SNP, average annual temperature is the main driving factor, with higher temperatures leading to increased SOM content. In contrast, elevation is the key factor in the SJP, where both high and low-elevation areas have a lower SOM content. The combination of different environmental factors also influences SOM changes; in the SNP, the interaction between temperature with elevation and precipitation is significant, while in the SJP, the interactions of elevation with temperature and slope with precipitation are both significant and nonlinear. Although this study has explored the main driving factors, future research could further focus on, first, the contribution efficiency of these factors to SOM changes; second, the thresholds at which these factors affect SOM changes could be explored for a deeper understanding. Furthermore, local government policies, legal protection, and attention to black soil also impact SOM change and distribution.

This study, based on high spatial resolution SOM products, explored the spatial and temporal changes in SOM in SNP and SJP between 1984 and 2021. It quantitatively analyzed the driving mechanisms of SOM changes in the two plains at a scale of 10 km × 10 km using a geodetector model. This research highlights the current status of SOM decline in the northeastern black soil region and emphasizes the need to strengthen regional differentiation in soil protection and management for the effective formulation of land resource management policies and measures for future black soil conservation efforts.