1. Introduction
Soil organic carbon (SOC) is a significant contributor to greenhouse gas emissions and has a considerable impact on the global carbon cycle [
1] and represents the largest carbon (C) pool within terrestrial ecosystems [
2]. Globally, there is an estimated SOC pool ranging from approximately 700 Pg to 2946 Pg [
3]. Against the backdrop of global warming, the relationship between SOC, climate, and the environment has become a focal point of research [
4]. Climate and land cover drive SOC [
5]. Temperature, precipitation, and humidity, among other factors, impact SOC by affecting plant litter, soil respiration, and microbial decomposition processes [
6]. It is worth noting that the cryosphere is a critical component of the Earth’s ecosystem, covering approximately 14% of the land surface [
7]. Permafrost, as one of the components of the cryosphere, is also one of the most vulnerable carbon (C) pools [
8]. It is estimated that the global permafrost contains approximately 1320 ± 200 Pg of SOC [
9,
10]. In the Intergovernmental Panel on Climate Change (IPCC) Sixth Assessment Report [
11], it was stated with high confidence that permafrost temperatures have risen to record levels, leading to increased release of methane (CH
4) and carbon dioxide (CO
2) from permafrost regions. The rate of permafrost warming, as exemplified by the Qinghai-Tibet Plateau at 0.3 °C per decade, significantly exceeds the global average warming rate of 0.12 °C per decade [
12]. During a warming trend, permafrost undergoes significant degradation [
13,
14,
15], which affects the process of SOC decomposition [
16,
17]. Alterations in the SOC pool within permafrost can have a significant impact on atmospheric CO
2 concentrations, further accelerating the effects of climate change caused by global warming [
9].
The relationship between permafrost changes and SOC is intricately linked to the spatial distribution of permafrost. Permafrost simulation models primarily include the Stefan model [
18], the frost number model [
19], the altitude model [
20], and the top temperature of permafrost (TTOP) model [
21]. The TTOP model, specifically, has garnered significant attention for its exceptional performance and applicability. It links surface climate with subsurface thermal features and establishes a correlation between compensation effect and climatic factors [
22]. With the development of remote sensing technologies such as optical remote sensing, thermal infrared remote sensing, and microwave remote sensing, observation and mapping of perennial permafrost are carried out by developing statistical learning [
23], constructing models [
24,
25], and logical discriminations [
26]. The application of TTOP modeling has thus led to many research results [
27,
28,
29,
30]. Permafrost degradation is typically marked by a reduction in the permafrost’s extent, an elevation in soil temperature, and an increase in the active layer’s thickness [
11,
31,
32]. This alters surface drainage patterns and the structure of vegetation communities, which impacts the physical properties of soil and ecosystems [
33]. The degradation characteristics of permafrost are obvious, but the resulting environmental effects and impacts on soil organic carbon (SOC) unfold over a long period of time. The response mechanisms are multifaceted, creating complex and interconnected relationships. Permafrost degradation can hasten microbial decomposition of organic matter, rendering SOC more vulnerable to exposure and release, which could result in a decrease in SOC [
34,
35]. In contrast, increased soil temperatures and soil moisture resulting from permafrost degradation promote vegetation growth in the short run [
36]. The carbon sequestration from the increased vegetation can offset or even exceed carbon losses [
10,
37]. Such uncertainty provides a cushion against SOC release, making the study of the impact of permafrost change on SOC dynamics particularly crucial.
In Northeast China, permafrost has higher temperatures compared to the permafrost in high-altitude and Arctic regions, with a thinner permafrost thickness, making it more sensitive to climate warming [
38,
39]. According to a report [
40], the mean annual air temperature in Northeast area has risen by 0.31 °C per decade from 1961 to 2017. Climate warming has resulted in a substantial degradation of permafrost, with continuous permafrost being transformed into discontinuous permafrost and the disappearance of many isolated permafrost patches and permafrost islands [
14]. Many researchers have conducted extensive research on the extent of permafrost degradation in Northeast China, utilizing multiple sources of data such as meteorological station data, measurement data from boreholes, and remote sensing data (Landsat and MODIS inverted surface temperature data, interferometric synthetic aperture radar data, etc.) [
41,
42,
43]. However, comparing the latest results of permafrost distribution in Northeast China during the same period, it is found that the area of permafrost varies greatly, reaching 12.81 × 10
4 km
2, which indicates that further research on the distribution of permafrost in Northeast China is needed. Furthermore, since 2003, Chinese meteorological stations have switched from manual to automated measurements, resulting in significant variations in recorded ground surface temperature (GST) over different periods. This has created a lack of uniformity in GST data after this transition. Undoubtedly, permafrost is mainly concentrated in the northern part of Northeast China, where there is abundant vegetation cover, and an amount of SOC is stored in permafrost [
44]. Changes in permafrost could potentially affect SOC density, stock, and succession of land type, thus impacting ecosystem evolution [
45]. Land use and land cover (LULC) have a direct influence on the density and stock of SOC. Their capacity to release or sequester SOC differs significantly [
46]. However, the precise distribution of permafrost in Northeast China and its impact on SOC remains unexplored, as well as the impact of LULC types on SOC in permafrost degradation zones.
3. Results
3.1. Regression Model Adjusted GST
We selected seven representative national meteorological stations in the northern part of Northeast China (Mo’he, Tulihe, Jiagedaqi, Manzhouli, Hailar, Sunwu, and Boketu) to represent the permafrost regions (
Table 2), and compared the mean annual ground surface temperatures and mean annual air temperatures in China, Northeast China, and permafrost regions. As illustrated in
Figure 2a, the mean annual air temperature warming rates from 1980 to 2020 for China, Northeast China, and permafrost regions stand at 0.36 °C/10a, 0.33 °C/10a and 0.37 °C/10a, respectively. During the period from 1980 to 2002, Northeast China and permafrost regions exhibited mean annual ground surface temperature warming rates of 0.61 °C/10a and 0.77 °C/10a, respectively. Notably, the mean annual ground surface temperature warming rates for Northeast China and permafrost regions between 2003 and 2020 are recorded at 1.1 °C/10a and 1.48 °C/10a, respectively.
After adjusting the GST data, as shown in
Figure 2b, the mean annual ground surface temperature warming rates for Northeast China and permafrost regions from 1980 to 2020 are 0.53 °C/10a and 0.63 °C/10a, respectively.
3.2. TTOP and Spatial Distribution of Permafrost
Using the ANUSPLIN software for spatial interpolation of TTOP,
Figure 3a shows the TTOP results obtained for Northeast China. Upon analysis of the spatial distribution of TTOP and classification of permafrost types, the following observations were made: The highest temperature in Northeast China rose from 9.306 °C to 10.164 °C, indicating a warming trend of 0.858 °C or an increase of 9.22%. The lowest temperature increased from −6.972 °C to −3.886 °C, showing a warming of 3.086 °C or an increase of 44.26%. Seasonally frozen ground areas were defined as regions with TTOP ≥ 0 °C, while permafrost areas were defined as regions with TTOP < 0 °C,
Figure 3b. The permafrost areas in the 1980s, 1990s, 2000s, and 2010s were about 37.43 × 10
4 km
2, 24.87 × 10
4 km
2, 18.93 × 10
4 km
2, and 16.48 × 10
4 km
2, respectively. The largest stage of permafrost degradation happened between the 1980s and the 1990s, resulting in a decrease in the area of 12.57 × 10
4 km
2. Over the span of four decades, from the 1980s to the 2010s, the extent of permafrost decreased by about 20.95 × 10
4 km
2, accounting for a 55.98% decline. In contrast, the proportion of seasonally frozen ground areas increased considerably as the permafrost area percentage fell from 25.90% to 11.41%. Presently, the Da Xing’anling Mountains host a significant proportion of the remaining permafrost, raising alarm about its preservation.
3.3. Changes in Soil Organic Carbon
As shown in
Figure 4a, the regions in Northeast China that have the highest SOC density and SOC stocks are concentrated in the major forestland and grassland areas of the Da and Xiao Xing’anling Mountains, as well as the Songnen and Sanjiang Plains.
From the SOC data presented in
Figure 4b, it can be observed that the overall trend of SOC density and SOC stocks in Northeast China experienced a rapid decrease followed by an increase. SOC density and stock increased at rates of 0.036 kg C/m
2 a and 5 Tg C/a, respectively. SOC density decreased from 7.22 kg C/m
2 in the 1980s to 7.20 kg C/m
2 in the 1990s, before rising to 7.28 kg C/m
2 in the 2000s and 7.32 kg C/m
2 in the 2010s. The SOC stock started at 10.15 Pg C in the 1980s, experienced a rapid decline of 24.18 Tg C to 10.13 Pg C in the 1990s, and then substantially rose to 10.23 Pg C in the 2000s, finally reaching 10.28 Pg C in the 2010s. The trend in SOC stocks corresponds to that of SOC density. According to
Table 4, the corresponding total SOC stocks and proportions are displayed for the four periods based on various LULC types and the extent of permafrost.
The area of arable land has steadily increased between 1980 and 2020, showing a total growth of 5.8 × 10
4 km
2. However, during the same period, the area covered by forestland and grasslands has consistently decreased, with a cumulative reduction of 7.0 × 10
4 km
2. Some of these lands have been converted into construction land, while others have become unused land.
Table 5 displays the dynamics of LULC change in areas affected by permafrost degradation across various time periods. During the period from the 1980s to the 1990s, the increase in SOC stocks in arable land was the most significant in Northeast China, amounting to 267.94 Tg C. Conversely, forestland experienced the greatest loss in SOC stocks, totaling 129.71 Tg C, followed by grasslands. The region where permafrost is located witnessed a decrease in SOC stocks by 1026.97 Tg C. In the 1990s to 2000s, arable land contributed the most to the increase in SOC stocks, with an addition of 60.38 Tg C, followed by forestland, which had transitioned from losses to gains at this point, with an increase of 37.32 Tg C. Unused land experienced the greatest reduction in SOC stocks, amounting to 4.77 Tg C. In the region with permafrost, SOC stocks decreased by 494.52 Tg C. Between the 2000s and 2010s, unused land reversed its trend, registering the highest increase in soil SOC stocks, with an addition of 112.14 Tg C. Arable land followed closely behind. Both forestland and grasslands switched from gains to losses, with grasslands experiencing the most significant reduction in SOC stocks at 115.01 Tg C. The region with permafrost witnessed a decrease in SOC stocks of 202.45 Tg C.
3.4. Relationship between LULC and SOC Changes in Permafrost Degradation Areas
Permafrost degradation can cause alterations in hydrogeological conditions, which subsequently impact the succession of LULC types.
During the period from the 1980s to the 2000s, the largest increase in arable land area occurred, totaling 5201 square kilometers and resulting in a dynamic change rate of 5.23%. The highest reduction in land area during this time period occurred in forestland areas, decreasing by 4436 square kilometers and resulting in a dynamic change rate of −7.91%. Additionally, grassland also experienced a decrease of 834 square kilometers, resulting in a dynamic change rate of −1.76%. Permafrost degradation was identified as the most significant contributing factor to the decrease in forestland and grassland areas, ultimately making the land more suitable for cultivation. The changes in SOC are significantly impacted by the combined effect of permafrost degradation and human activity.
SOC changes can be divided into two parts: SOC increase (density ≥ 0) and SOC decrease (density < 0). Comparing the area occupied by various LULC types in the permafrost degradation zone with their corresponding contributions to SOC alterations, the following trends can be observed. The outcomes are displayed in
Figure 5:
The data suggest that ongoing permafrost degradation leads to a significant increase in the proportion of forestland and grassland within the affected area, accounting for over 75% during each succession period. Additionally, their respective contributions to SOC increase or decrease show the most notable changes. Forestland dominates the increase in SOC, with a contribution rate that is consistent with its proportion of the total area. The contribution rates for forestland are 50%, 75%, and 72%, for the periods of 1980s–1990s, 1990s–2000s, and 2000s–2010s, respectively. A decrease in SOC mainly occurs in grasslands, with contribution rates of 50%, 83%, and 60% for the same periods. This relationship remains unchanged, even when the area of forestland exceeds that of grassland. Unlike the situation of SOC increase, the ratio of forestland area to grassland area displays a fluctuation in strength when associated with SOC decrease. This trend may be related to the different carbon sequestration capacities and sensitivities of the two land types.
3.5. Relationship between Permafrost Change and SOC Change
When comparing the permafrost change status with the corresponding regional SOC stocks, the following observations can be made (
Table 6).
The area of permafrost degradation between the 1980s and the 1990s was 12.52 × 104 km2. Of this area, 3.77 × 104 km2 was an area of SOC increase, with a total increase of SOC stocks of 2.05 Tg C, and 8.75 × 104 km2 was an area of SOC decrease, with a total decrease of SOC stocks of 8.14 Tg C. Surprisingly, although the area of permafrost degradation in this period accounted for only 8.92% of the area of Northeast China, the total amount of the change in SOC stocks was 25.17% of the total change in SOC stocks in Northeast China, which is inconsistent with the correspondence between area and SOC and positively illustrates that the effect of permafrost degradation on SOC is significant and is especially pronounced when permafrost degradation are evident.
In the four periods of the 1980s, 1990s, 2000s, and 2010s, there was a region of permafrost between the Da Xing’anling Mountains, Genhe City, and Arctic Village with an area of 15.12 × 104 km2. The SOC stocks in this region were 1.359 Pg C, 1.346 Pg C, 1.347 Pg C, and 1.349 Pg C for the respective periods. In the 1980s and 1990s, there was a significant increase in temperature, a rapid decrease in permafrost area, and a rapid decrease in SOC stocks. During this period, the longitudinal SOC stocks in this region also decreased rapidly. From the 1990s to the 2010s, there was a gradual increase, indicating that permafrost degradation not only affects SOC through area reduction, but also influences SOC changes in the active layer of permafrost.
In the permafrost growth area, the growth area is 12,610 km2, the increase in SOC stocks is 0.24 Tg, the decrease is 0.47 Tg, and the total amount is a decrease of 0.23 Tg. Among them, the SOC growth area is 5,486 km2, which accounts for 0.56% of SOC growth, and the SOC increment accounts for 0.32%; the SOC decrease area is 7,124 km2, which accounts for 1.66%, and the SOC decrease accounts for 2.20%. The total change in SOC stocks had an impact of −0.42% on SOC stocks in Northeast China.
3.6. Model Evaluation
The regression model is a machine learning technique that exhibits efficient and stable advantages when dealing with large datasets. The trend in MAGST from 1980 to 2002 is generally consistent with the 2003–2020 MAGST for the Northeast and permafrost regions, which has been adjusted by regression modeling. The model assesses the error outcomes (
Figure 6) of a day-to-day surface air temperature simulation (N = 4748) from 1990 to 2002 with an RMSE of 1.289 °C, an MAE of 1.028 °C, and an R
2 of 0.99. Thus, the model is deemed to have an accurate fit.
In this paper, 50 borehole data (MAGT) were collected from permafrost regions of Northeast China between 2000 and 2020 [
51,
52,
53,
54,
55,
56]. The data were divided into 25 groups each from the 2000s and 2010s and compared with the TTOP of the same period in the present study. The results are presented in
Table 7:
The average temperature for the 50 boreholes’ data during both periods was −0.81 °C. In this study, the average TTOP was −0.89 °C, resulting in a difference of 0.09 °C. The root mean square error (RMSE) was 1.23 °C and the mean absolute error (MAE) was 0.94 °C. This indicates that the map of the distribution of permafrost in Northeast China developed through this study has a high degree of confidence.
As shown in
Figure 7, the correlation (R) between the model results obtained in this study and the borehole data was 0.65 (
p < 0.001), which is higher than that reported by Obu et al. [
7] (R = 0.50, N = 50). Therefore, we believe that after adjusting the GST for the years 2003–2020 through the regression model, using the TTOP model to simulate the distribution of permafrost in Northeast China region becomes more reliable.