The United Nations Framework Convention on climate change defines a “carbon source” as the process, activity or mechanism that releases greenhouse gases, aerosols or their precursors into the atmosphere, while “carbon sink” refers to the activity, process or mechanism of removing greenhouse gases, aerosols or their precursors from the atmosphere ^[1][33]. The knowledge acquired through the definition of carbon source and sink is that they are relative concepts. “Carbon source” refers to the matrix that releases carbon to the atmosphere in nature, and “carbon sink” refers to the deposit of carbon in nature.

Many scholars use two indicators, net primary productivity (NPP) and net ecosystem exchange (NEE) of vegetation, to describe the carbon sources and sinks of ecosystems ^[2][3][4][5][22,27,34,35]. NPP refers to the amount of organic matter in photosynthetic products fixed by plants per unit time and unit area after deducting the part consumed by the plants respiration ^[6][36]. It is an important index to evaluate the production capacity of the plant community under natural environment conditions and to measure the carbon sequestration capacity of vegetation ^[4][5][34,35]. Wang et al. ^[7][37] evaluated the temporal and spatial variation characteristics of the NPP of arable land ecosystems in China from 2001 to 2010 by combining MOD17A3 NPP data and GIS techniques, and observed that only 22% of NPP was significantly correlated with precipitation, and only 7% with temperature, indicating that arable land ecosystems were greatly affected by human activities. NEE is an important indicator to measure the carbon balance of ecosystems. It is the result of the balance between the total photosynthesis and total respiration of an ecosystem ^[8][38]. Zhang et al. ^[9][39] used a novel geospatial agricultural modeling system to calculate the NPP of crops, so as to estimate the NEE of arable land ecosystems. A positive value indicated that an area was a carbon source, while a negative value denoted a carbon sink. Li et al. ^[10][40] found that net radiation directly affected the seasonal variation of evapotranspiration and NEE in a winter wheat- summer maize system. Xu et al. ^[11][41] observed that in a rice–wheat system, seasonal variation in daily NEE and daytime NEE was directly affected by crop vegetation growth, and nighttime NEE and soil temperature at 10 cm during the wheat season exhibited a significant exponential relationship when accounting for grain removal and the return of straw to the field, indicating that the system was a weak carbon sink.

In addition, when calculating the carbon sources and sinks of arable land ecosystems, carbon absorption and carbon emission are usually estimated separately ^[12][13][2,26]. Carbon absorption is estimated based on crop-yield data, economic coefficient and carbon absorption rate, while carbon emission is estimated based on different carbon-emission pathways and combined with the carbon conversion coefficient ^[14][24]. The following parameters are often used ^[15][42]: biological yield is the total amount of dry matter (mostly above ground) harvested from crops per unit area of land; economic yield refers to the dry matter weight of grains or other organs of crops harvested for food or other uses per unit area of land; and economic coefficient is the ratio between economic yield and biological yield, varying with plant species, varieties, natural environment and cultivation measures. In general, the data of economic yield can be obtained from national or regional statistical data ^[13][14][24,26], but it is difficult to obtain the small-scale data ^[16][28]. Most of the carbon-emission coefficients of production activities are directly quoted from the research results of West (Oak Ridge National Laboratory) ^[17][43], Lal ^[18][44] and IPCC.

Global terrestrial soil organic carbon (SOC) stocks are about 1400–1500 PgC within 1 m depth soils, and are the largest carbon pool on the earth’s surface, 2–3 times greater than that of terrestrial vegetation and more than twice that of the atmospheric carbon pool ^[19][45]. Small changes in terrestrial stocks cause changes in CO₂ concentrations in the atmosphere, thus affecting global climate ^[20][46]. SOC in arable land accounts for 10% of total organic carbon in soils ^[21][47]. Therefore, SOC has a certain function in the regulation of atmospheric CO₂ concentrations. Whether an arable land ecosystem is a carbon source or sink largely depends on the balance between the fixation of arable land SOC and the release of greenhouse gases ^[22][31]. Therefore,  research on the arable land SOC pool has gradually become the focus of the international community.

Many countries had completed the estimation of their SOC stocks on national or regional scales ^[23][48]. SOC stocks were mainly estimated by soil types, vegetation types or model methods, and the determination of relevant estimation factors was mainly obtained by collecting historical data and satellite images and through hyperspectral remote sensing technology. Song et al. ^[24][49] estimated that topsoil SOC stocks were about 5.1 PgC, based on the second soil survey data from 1979 to 1982 in China. Considering the entire arable land category, Tommaso et al. ^[25][50] estimated that the average SOC stock in the topsoil (30 cm) in Italy was 52.1 ± 17.4 Mg C ha⁻¹, which was similar to that reported by other European countries. Sleutel et al. ^[26][51], combining SOC data with arable land area data, estimated that the SOC stocks of arable land was about 49,000 tons in Belgium. In France, Arrouays et al. ^[27][52] estimated SOC stocks at 0–30 cm soil depth according to land use and soil type using data from geo-referenced databases. The results showed that SOC stocks were 15–40 Mg C ha⁻¹ in central France, and SOC stocks were 40–50 Mg C ha⁻¹ in northern and southwestern regions.

2. Main Factors Affecting the Dynamics of Carbon Sources and Sinks in Arable Land Ecosystems

2.1. Effects of Natural Environmental Factors on Arable Land Carbon

The rate of soil carbon loss is related to the soil environment, which is strongly controlled by climatic conditions ^[28][19]. Climate controls NPP above and below the ground, and thus the input of organic matter, while climate also contributes to carbon loss by driving the output of organic matter through microbial activity in the soil ^[29][30][82,83]. A large number of studies have confirmed a negative correlation between SOC and temperature in arable land ecosystems. Low temperature may reduce the mineralization of SOC through thermodynamic mechanism ^{[29][31][32][33]}[3,20,82,84]. This conclusion was verified in arable-land ecosystems in different regions ^[34][35][36][85,86,87], and was also confirmed in forest ecosystems ^[29][82] and shrub ecosystems ^[37][30]. However, the situation was different in some low-temperature areas. Increasing temperature could stimulate the input of plant productivity ^[38][88], which was more conducive to the accumulation of SOC. On the Qinghai Tibet Plateau, Nie et al. ^[39][89] concluded that the increase of annual average temperature had a positive impact on SOC density, which might ascribe to the increase of soil carbon input exceeded the carbon loss caused by elevated temperature. Therefore, the increase of temperature might make the soil accumulate carbon under warming conditions. In addition, the types suitable for crop growth will vary under different temperatures, leading to differences in photosynthetic rate and carbon absorption rate. For example, temperature led to differences in crop structure between South and North China, further resulting in differences in carbon sources and sinks in arable land ecosystems ^[14][24]. In the research of Wang et al. ^[40][72], although the season length of maize (113 days) was 52% shorter than that of wheat (235 days), more than 55% of CO₂ emissions come from maize season, and the interaction of soil temperature and moisture better explained the variations of the ecosystem respiration and soil respiration from the relatively colder and drier wheat growing season to the warmer and wetter maize growing season.

2.2. Impact of Human Management Measures on Arable Land Carbon

2.2.1. Tillage Measures

In arable land ecosystems, it is generally believed that agricultural farming strategies have more far-reaching impacts on SOC than natural factors ^[34][41][85,92]. The decrease of SOC content in arable land systems was mainly caused by cultivation. Tillage changes the quality and quantity of carbon input in soil and the physical properties of soil that affect carbon decomposition ^[28][19]. In the process of cultivation, the topsoil environment is often changing. It was generally believed that the loss of SOC mainly occurs in the 0–30 cm soil layer ^[42][43][93,94]. Tillage methods could also significantly affect SOC content ^[44][95]. Traditional farming methods, such as fallow in bare land, burning or removal of crop residues and inverted farming, have promoted the loss of SOM ^[32][45][20,96]. By reducing soil tillage and adopting conservation tillage measures, such as retaining crop residues in arable land, SOC can be fixed ^[12][28][46][2,13,19]. No tillage and less tillage have significant effects on the increase of SOC stocks and the change of microbial biomass carbon ^[22][28][19,31]. West and Post ^[17][43] estimated that with the conversion from traditional tillage to no tillage, the global SOC retention rate was 57 ± 14 g cm⁻² yr⁻¹. Dachraoui and Sombrero ^[12][2] compared the carbon footprint of corn under traditional tillage and no tillage management, showed that no tillage system reduced greenhouse gas emissions and contributed to carbon sequestration in the soil at the depth of 0-30 cm. Zhang et al. ^[47][97] also noted that SOC of the topsoil (0–30 cm) increased significantly under no tillage system compared with conventional tillage. In addition, increasing the complexity of crop rotation and straw return could also increase SOC and reduce greenhouse gas emissions ^[48][98]. In monoculture system, crops uptake less than half the amount of nitrogen fertilizers normally, through crop rotation, other types of crops could absorb nitrogen during the fallow period of bare land, also straw might lead to the richness and diversity of plant litter and increase the acquisition of carbon ^[49][58]. Because different natural factors shape the background of different arable land ecosystems, the spatial variability affecting site characteristics must be considered in the land use planning and implementation of strengthening carbon sequestration, so as to scientifically and effectively increase the carbon sink capacity of arable land ecosystems.

2.2.2. Fertilization Measures

The impacts of fertilization on carbon source and sink of arable land ecosystems are mainly reflected in two aspects ^[28][50][19,80]. First, it improves the vegetative environment for plant growth and increases biomass, so as to increase the input of soil organic residues and promote the accumulation of organic carbon. Second, by affecting the population, quantity and activity of soil microorganisms, it has an impact on soil respiration. Many researchers have reported positive effects on SOC sequestration due to increased fertilizer and organic inputs. Morell et al. ^[51][103] found that after 15 years of application of mineral nitrogen, the amount of carbon retained in the soil increased due to the increase of crop residue production. Trost et al. ^[52][104] concluded that the combination of irrigation and fertilization may lead to a significant increase in SOC content, especially in light soil with low initial organic carbon content. Yue et al. ^[53][15] showed in a meta-analysis that nitrogen application significantly increased the total carbon storage of soil by 5.82%, and increased the carbon content of aboveground and underground parts of plants by 25.65% and 15.93% respectively. Globally, nitrogen addition significantly increased aboveground net primary productivity by 52.38%, indicating that with the increase of nitrogen deposition, terrestrial ecosystems may be enhanced as carbon sinks. Moharana et al. ^[54][105] observed that the SOC accumulation effects of farmyard manure (FYM) and FYM + NPK (N: nitrogen; P: phosphorus; K: potassium) treatments were better than that of NPK alone. In intensive agriculture, production depended on the extensive use of synthetic fertilizers, especially nitrogen fertilizer ^[55][106]. Some studies found that nitrogen addition was considered to be the largest contributor to the impact of different management practices on carbon emissions, the emissions caused by nitrogen fertilizer exceed 50% of the total emissions ^[56][57][107,108]. Due to the decline of nutrient use efficiency, the use of chemical fertilizer to maintain crop yield has been increasing, which leads to higher direct emissions of greenhouse gases from soil ^[58][109]. The mechanism of nitrogen fertilizer affecting CO₂ emission is that nitrogen application promotes microbial growth and soil respiration that depends on the SOM content. When the soil carbon source is sufficient, applying N fertilizer will promote soil respiration and increase CO₂ emission, while when the carbon source is insufficient, soil respiration is inhibited ^[59][110]. Jiang et al. ^[56][107] found that increasing the amount of nitrogen application could improve rice yield, but when the amount of nitrogen application exceeded 225 kg N ha⁻¹, it had little impacts on rice yield and even caused some adverse effects. This highlights the importance of improving nitrogen management practices, preventing economic losses to crop producers, thus achieving a balance between reducing carbon emissions and expanding net carbon sinks in arable land ecosystems.

2.2.3. Irrigation Measures

Irrigation and its scheduling affect soil and crop water status, thereby affecting microbial function and greenhouse gas emissions ^[60][111]. Proper soil moisture would enhance root respiration and microbial activities, accelerate the decomposition of SOM and increase CO₂ emission ^[52][104]. However, high humidity reduces soil aeration and inhibits soil respiration and CO₂ emissions ^[61][112]. The effect of irrigation on carbon sequestration appear to be highly dependent on location/conditions ^[40][72]. In arid areas or soil with low initial carbon content, irrigation can increase the content of SOC ^[62][63][16,113]. Wetting the soil with irrigation after drought could release the accumulated SOM during drought periods and produce a large amount of nutrients and organic carbon ^[64][114]. In desert areas, due to large soil pores, the infiltration of moisture can transport SOM and fine particles to deep layers ^[32][20]. However, in areas with humid climate and high initial SOC content, irrigation might lead to a decrease in SOC ^[65][115]. The commonly used irrigation methods in arable land include border irrigation, sprinkler irrigation and drip irrigation, which may cause different effects on arable land carbon. In southwestern Nebraska, Gillabel et al. ^[62][16] compared the carbon stocks between irrigation and dryland management treatment, it was found that the carbon stocks of irrigation were 25% higher than that of dryland cultivation, and the carbon input of crop residue under drip irrigation was estimated to be 2.5 times higher than that under drought. However, it was found that increasing carbon input under irrigation could not improve the level of large aggregates and the greater carbon stock was mainly due to the higher carbon sequestration in micro aggregates. The impact of tillage damage on the overall level was greater than the increased residue input under irrigation ^[62][16]. Li et al. ^[60][111] concluded the CO₂ flux and cumulative emission of drip irrigation plot were significantly higher than that of border irrigation plot. The increase of CO₂ emission of drip irrigation might due to the better water and soil environment created by irrigation, resulting in higher plant root respiration and stronger microbial activity. However, there were no unified conclusions on the impacts of irrigation method on CO₂ emission. Li et al. ^[66][116] reported that under the condition of film covering, the CO₂ emission of drip irrigation in clay loam was lower than that of flood irrigation. Therefore, information on different management practices and other irrigation systems in different regions is needed to more accurately understand the overall impact of irrigation on soil carbon storage.

2.2.4. Land Use Change

Land use change is considered to be the second largest cause of carbon emissions after fuel consumption ^[67][117]. Houghton et al. ^[68][118] concluded that 156 Pg C was released into the atmosphere globally due to land use change and management during 1850–2000. The growing population’s demand for food, fiber and fuel had accelerated the transformation of natural land into managed land, such as from forest or natural grassland to pasture or arable land ^[69][119]. The transformation from natural ecosystems to agricultural ecosystems would consume organic carbon pool, mainly due to: (I) low return of biomass carbon, (II) large loss of organic carbon caused by erosion, mineralization and leaching, and (III) large changes in soil temperature and water status ^[70][14]. Some studies had also focused on changes in carbon stocks between specific ecosystems. Don et al. ^[71][120] found that the conversion from virgin forest to arable land resulted in 25% to 30% SOC loss by a meta-analysis using 385 existing studies in tropical land. Clair et al. ^[72][121] found that the replacement of existing forest land by rape field would lead to net emissions, while the replacement of existing arable land by perennial miscanthus and short rotation shrub would produce significant net greenhouse gas benefits. In a study on the black soil area in Northeast China, Song et al. ^[33][84] found that the conversion of grassland to arable land would lead to the loss of C and N in 0–30 cm soil layer to a certain extent. DeFries et al. ^[73][122] concluded that 25–30% of the carbon in the topsoil would be released into the atmosphere when the forest was transformed into permanent arable land. It could be seen that the mutual transformation of forest and arable land will strengthen or reduce the carbon sequestration capacity of soil to a certain extent. Houghton and Nassikas ^[74][123] used land use change rate and carbon density data to compare the interaction between different ecosystems, considering five land use types: Arable land, pasture, plantation, industrial wood and fuelwood. The results showed that the net carbon flux of land use change from 1850 to 2015 was 145 ± 16 Pg C. Most of the emissions came from tropical regions (10² ± 5.8 Pg C). The average global net emissions in the last decade (2006-2015) were 1.11 (±0.35) Pg C yr⁻¹, including the net carbon source in tropical regions (1.41 ± 0.17 Pg C yr⁻¹), the net carbon sink in northern mid-latitudes regions (−0.28 ± 0.21 Pg C yr⁻¹), and the neutrality in southern mid-latitudes regions. Recently, some studies combined RS and GIS to obtain land use change information and then estimate carbon emissions. Zhu et al. ^[75][124] combined remote sensing, GIS and IPCC method to quantify changes in vegetation carbon storage and SOC storage resulting from land use change during 1970–2010 in Zhejiang province of China. The result showed that land use change has resulted in huge amounts of carbon emissions, mainly caused by decrease of farmland with high SOC content, attributing to urban expansion. Li et al. ^[76][125] used the land use change data from 2000–2020 of Anhui province in China, evaluated the net carbon emissions and clarified the carbon emission effect from three aspects of carbon footprint, ecological carrying capacity and ecological deficit. They found that forestland is the main carbon sink, while construction land is the main carbon source, also the carbon footprint of increases rapidly, the ecological carrying capacity changes slowly, and the ecological deficit becomes larger and larger, indicated that with economic development, carbon emissions from construction land will become more and more significant, and the low-carbon development will face great pressure. The improvement of the availability and quality of multi-spatial and multi-temporal remote sensing data and the emergence of new analysis technologies have deepened the understanding of impact of land use change on carbon emissions ^[75][124].

3. Problems and Prospects of Carbon Source and Sink Research in Arable Land Ecosystems

3.1. Problems of Carbon Source and Sink in Arable Land Ecosystems

3.1. Research Problems of Carbon Source and Sink in Arable Land Ecosystems

Although many scholars have made significant progress and achieved important results in the research of carbon source and sink in arable land ecosystems, due to the differences of natural and social environment in different regions, there still remains insufficient understanding on mechanisms of carbon cycle in arable land system and the influencing factors of carbon source and sink changes. In particular, due to the difficulty of obtaining statistical data, the change of carbon source and sink at small scales is not well understood, so it cannot provide guidance for carbon sequestration and emission reduction of arable land system. There is still a lot of room for improvement in relevant research. Also, balancing the need for agricultural products and other land use with reducing carbon emissions is a big challenge in developing sustainable management in arable land system. In addition, there are some problems in the studies of carbon source and sink of arable land system, such as fuzzy system boundary, incomplete accounting index and parameter inconsistent with reality, making the differences between results, and even making the results are not comparable. The measurement accuracy of carbon absorption and emission is relatively at a low level, mainly because most scholars used the production activities input coefficient or carbon respiration coefficient of different crops published by the IPCC for calculation, and did not take into account the differences caused by climate or soil conditions, inducing larger errors in carbon emission and carbon absorption estimates. Besides, while most studies considered the impacts of agricultural machinery and chemical fertilizer factors on carbon emissions, but did not consider the carbon emissions caused by the power consumption of people engaged in agriculture production; however some studies included the latter factors, providing different estimates.

3.2. Prospects of Carbon Source and Sink in Arable Land Ecosystems

3.2. Research Prospects of Carbon Source and Sink in Arable Land Ecosystems

In view of the problems existing in quantification of the carbon source and sink of arable ecosystems, the research on the following aspects should be strengthened. First, the accuracy of data calculation shall be improved. When determining the carbon conversion rate of different agricultural inputs, the conversion coefficient should be adjusted according to different soil texture, climate and farming conditions. At the same time, itwe should comprehensively analyze the role of various influencing factors. Models and methods should be compared to select the best method according to different scresearch scales. Second, the use of high-tech can not only improve the accuracy of research, but also provide a more scientific basis for the rational development of agriculture and the protection of the global ecological environment. Since there are considerable uncertainties in the global long-term and large-scale study of arable land system, combining models with remote sensing and GIS technology provides a good opportunity to evaluate the spatial and temporal distribution pattern of carbon sink and carbon source activities in arable land system. Finally, the impact of management measures on the carbon source and sink of arable land ecosystems shall be considered. Due to the complexity and diversity of influencing factors of arable land carbon source and sink, the research results cannot be blindly generalized. To determine the amount of irrigation, fertilizer application and mechanized use according to local conditions, it is necessary to carry out studies at different scales.