Farmland ecosystems, as the most active carbon pool, are integral to global climate change and carbon cycling. Therefore, systematically studying the roles of carbon sources and sinks in farmland ecosystems is essential to deepening our understanding of the carbon cycle and meeting the goals of “peak carbon emissions” and “carbon neutrality” in agriculture. Using the Web of Science database, this study reviewed 1411 articles (2002–2023) via bibliometric analysis to identify key research themes, trends, future priorities and address suggestions for future directions in farmland ecosystem carbon sources and sinks. The main results include the following: (1) Over the past 22 years, global research in this field has shown a consistent growth trend, with remarkably rapid expansion in the past three years. China, the United States, and Germany are the most influential countries. As the research scope has expanded, the field has evolved into an interdisciplinary domain. (2) The diversity of this research area has become enriched, and the research content is becoming more refined and systematic. The main research topics focus on carbon sequestration, soil organic carbon (SOC), farmland management, greenhouse gas (GHG) emissions, carbon stocks, ecosystem services, land use changes, climate change, and spatiotemporal heterogeneity. (3) Current research hotspots primarily focus on studying soil microbial carbon sequestration mechanisms, the application of remote sensing technologies, and reducing GHG emissions to achieve “carbon neutrality”. While existing studies have systematically elucidated carbon sequestration mechanisms mediated by soil aggregates, microorganisms, and minerals, critical knowledge gaps persist. Regional disparities in the relative contributions of these mechanisms remain unresolved, compounded by methodological inconsistencies in carbon assessment that introduce substantial uncertainties. Although farmland management practices are identified as pivotal drivers of carbon flux variation, the interactive effects of anthropogenic interventions and natural factors on ecosystem-scale carbon balance require further mechanistic exploration. This review provides a comprehensive reference for further study on carbon sources and sinks of farmland ecosystems and devising effective emission reduction strategies.

Land-use change has a significant impact on the structure and function of ecosystems and is an important reason for the imbalance between the supply and demand of ecosystem services. Pollination services are indispensable functions of ecosystems. In recent years, land-use change has caused a decline in the abundance of pollinators, thereby affecting the supply of pollination services, which has been a major concern for governments and scholars. Currently, there is an insufficient exploration of the impact mechanism of land-use change on pollination services. The application of a pollination service evaluation model based on land-use data uses a large amount of empirical data, which greatly affects the accuracy of regional evaluation results. This study uses Huizhou as a representative example. Remote sensing images from 2015 and 2019 were used to interpret the land-use data of the region, and the spatiotemporal changes in the land use were then analyzed. Due to their high pollination dependence, litchi and longan were selected as the research objects. Basic data such as the main pollinator species of litchi and longan and floral plant species were obtained through field sampling surveys. The InVEST model was used to evaluate the abundance of pollinators in litchi and longan orchards, and the abundance of pollinators was used to represent the value of pollination services in litchi and longan orchards. Then, the Hotspot analysis method was used to analyze the change in the spatial pattern of the pollinator abundance in litchi and longan orchards. The main influencing factors of pollination service in litchi and longan orchards were analyzed by a Geographical detector. Finally, we have explored the impact mechanism of land-use change on pollination services. The following are the results of this research. The pollinator abundance in the orchards of litchi and longan and their buffer zones in Huizhou decreased by 6.64% and 13.94% from 2015 to 2019, respectively. The wild bee abundance in forest land and rainfed cropland decreased by varying degrees. The spatial aggregation characteristics of pollinator abundance in litchi and longan orchards demonstrated an increase in cold spots, whereas the hot spots decreased and were more dispersed. In the study area, the area change and land-use change of natural or semi-natural habitats, such as forest land, rainfed cropland, and grassland, affected the pollination services for litchi and longan orchards. Within the types of changing land-use, the change of forest land has the greatest impact on litchi and longan pollination services. The impact degrees of Forest land area, rainfed cropland area, area under litchi and longan orchards, and forest landscape fragmentation on the pollination services for litchi and longan orchards were 0.20, 0.16, 0.21, and 0.26, respectively.

The spatiotemporal patterns and shifts of net ecosystem productivity (NEP) play a pivotal role in ecological conservation and addressing climate change. For example, by quantifying the NEP information within ecosystems, we can achieve the protection and restoration of natural ecological balance. Monitoring the changes in NEP enables a more profound understanding and prediction of ecosystem alterations caused by global warming, thereby providing a scientific basis for formulating policies aimed at mitigating and adapting to climate change. The accurate prediction of NEP sheds light on the ecosystem’s response to climatic variations and aids in formulating targeted carbon sequestration policies. While traditional ecological process models provide a comprehensive approach to predicting NEP, they often require extensive experimental and empirical data, increasing research costs. In contrast, machine-learning models offer a cost-effective alternative for NEP prediction; however, the delicate balance in algorithm selection and hyperparameter tuning is frequently overlooked. In our quest for the optimal prediction model, we examined a combination of four mainstream machine-learning algorithms with four hyperparameter-optimization techniques. Our analysis identified that the backpropagation neural network combined with Bayesian optimization yielded the best performance, with an R² of 0.68 and an MSE of 1.43. Additionally, deep-learning models showcased promising potential in NEP prediction. Selecting appropriate algorithms and executing precise hyperparameter-optimization strategies are crucial for enhancing the accuracy of NEP predictions. This approach not only improves model performance but also provides us with new tools for a deeper understanding of and response to ecosystem changes induced by climate change.