Introduction

Cultivated land is fundamental to national food security. As the foundational resource for national food security, cultivated land has garnered significant scholarly attention. Research foci encompass diverse dimensions, including Security and governance of cultivated land resources1,2, Spatial distribution patterns of existing cultivated land3, Spatiotemporal evolution mechanisms of cultivated land systems, Impacts of soil physicochemical properties on land productivity4,5, and Prediction of future cultivated land dynamics6.

Contemporary studies in Cultivated Land Prediction predominantly employ models such as FLUS7 (future land use simulation), ANN-ABM8 (artificial neural network-agent-based model), and SC-CNN-CA9 (spatio-temporal convolutional neural network-cellular automata). MaxEnt models have low computational requirements, high accuracy, and data flexibility10,11, capable of generating reliable predictions even with incomplete species distribution datasets and performing better than similar models12,13. Notably, the MaxEnt (Maximum Entropy) model remains underutilized in this domain despite its distinct advantages.

The southeast coastal areas of China have superior natural conditions, which support agricultural potential and serve as refugia for rare biological resources14,15,16. At the same time, its rapidly urbanization attracts massive migration from inland areas, driving intense competition for land resources, especially cultivated land17,18. According to China’s Third National Land Survey data, urban expansion demands have reduced regional cultivated land by 2.3% annually since 2015. Coastal cultivated land, characterized by dispersed distribution and accounting for less than 5% of the country’s total cultivated land area, exhibits high vulnerability to climate fluctuations19. Projected temperature volatility threatens 12–18% of existing coastal cultivated land through salinization and erosion. Under the pressure of rapid urbanisation and climate change, how to assess the long-term suitability of cropland along the southeast coast of China based on natural background to support the dynamic optimisation of the planning of the main functional zones and food security guarantee?

To ensure food security in the southeast coastal region, it is necessary to focus on the cultivated land degradation caused by climate change and include multiple environmental factors and future climate data into prediction and dynamic assessment, this study aims to affirm the suitable environmental factors for cultivated land, predict potentially suitable areas in the future by using the MaxEnt model in Southeast coastal areas of China. The specific objectives of this study were to (1) explore the influence and contribution of environmental factors in the spatial distribution of suitable cultivated land; (2) predict the spatial distribution of potentially suitable cultivated land under different future climate scenarios; and (3) analyse the migration route of suitable agricultural area and its driving factors.

Material and methods

Study area

The Southeast coastal area, as defined in this paper, corresponds to the region described in the 2018 first-level division of China’s natural resources dynamic regionalization (Fig. 1). This area includes Taiwan, Hainan, south-central Guangdong, eastern Guangxi, and southeastern Fujian, covering an area of 292,399 km2. It features a tropical and subtropical monsoon climate. As one of the most economically advanced regions in China, it benefits from an advantageous geographical location and robust economic conditions, supporting a diverse industrial structure that fosters scientific and technological innovation and progress in manufacturing, services, and high-tech industries. The region is also a hub of China’s international engagement, hosting numerous international ports and special economic zones. The high level of urbanization in the Southeast coastal areas attracts significant migration from inland and rural areas, leading to high population density. This influx significantly contributes to the local labor force, driving economic development and social transformation. The river network in this region is dense, and water availability is abundant. It ranks highly nationally in annual water modulus, per capita water quantity, and average land water quantity, providing significant support to the local area. Geographically, the majority of the study area lies below 25° N latitude, characterized by mountains and hills that cover 60% of the territory. Low mountains and hills are common, while mountains exceeding 1 km are primarily found in southern Fujian, northern Guangdong, southern Hainan, and central Taiwan. Influenced by regional geological structures, most mountains align northeast-southwest, with the terrain sloping from northwest to southeast. Notable nature reserves include Dai Yunshan in Fujian and Danxia Mountain in Guangdong, which protect diverse ecosystems such as Huangshan pine forests, wild orchids, mangroves, Danxia landforms, and aquatic resources20,21.

Fig. 1
figure 1

Distribution of sampling points of cultivated land in southeast coastal area (Created using ArcGIS 10.8.0, https://www.arcgis.com/index.html).

Full size image

Data collection

Land use distribution data source and processing

For this study, land use data from five different periods (2000, 2005, 2010, 2015, and 2020) were obtained from the multi-period remote sensing monitoring dataset of China (https://www.resdc.cn/). The spatial resolution is 30 m. During the execution of the MaxEnt model, it is necessary to select a set of land use data representing current land cover types. Given potential accuracy issues in land use data from specific periods, this study derived a more precise land use dataset by calculating the mode of five-phase land use data, which will serve as the final dataset for generating cultivated land distribution points. Data analysis reveals predominant distribution of forest and grassland in southeastern coastal areas, with cultivated land accounting for merely 0.23% of the total area, primarily concentrated along coastal margins. To prevent duplicate data within the same unit from affecting model prediction accuracy, and considering the limited cultivated land area, this study converted raster data to 1 km × 1 km point grids22. Environmental variables were subsequently extracted to these points, with locations containing missing environmental variables removed, ensuring only one valid distribution datum per 1 km range. This process resulted in the retention of 142 sample points for analysis. (Fig. 1, Table S1). These 142 samples effectively represent all individual cultivated land distribution locations with actual environmental variables within the study area.

Selection and processing of environment variables

Hydrothermal conditions play a fundamental role in the growth of crops on cultivated land, topographic conditions affect the redistribution of nutrients, and soils maintain and regulate cultivated land, all of which have an important impact on the suitability of cultivated land. In this study, 32 environmental variables including temperature, precipitation, topography, and geomorphology were selected for the southeast coastal region based on geographic and climatic factors (Table 1). The resolution of these variables is 1000 m. The geographic coordinate system used is GCS_China_Geodetic_Coordinate_System_2000, and the projection system is CGCS2000_3_Degree_GK_CM_114E. Current and future climate data are sourced from WorldClim2.1 (https://worldclim.org/data/index.html). Future climate projections for 2021–2040 (2040s) and 2041–2060 (2060s) are based on scenarios SSP126 and SSP585 from the CMIP6 Shared Socio-economic Path (SSP). SSP126 represents a low-emission scenario, emphasizing sustainable development and climate change mitigation. SSP585 represents an extremely high-emission scenario, emphasizing a pathway of rapid economic growth and high energy consumption.

Table 1 Variables used in the study and their sources and descriptions.
Full size table

Methods

Environmental variables, often correlated, can introduce multicollinearity, impacting result accuracy and interpretability. This study utilized the correlation coefficient from SPSS and the maximum entropy model’s contribution rate to filter variables. Variables were eliminated if two had a correlation coefficient exceeding 0.8 and the one with the lesser contribution was removed23,24. For example, 32 variables and species distribution points are input into the Maxent model to analyze the correlations among variables and their respective contribution values. If the correlation between variable A and variable B is greater than 0.8, then the contribution values of variable A and variable B are compared, and the variable with the smaller contribution value is removed, allowing the remaining variable to enter the final modeling stage.

To mitigate overfitting and enhance model transferability, adjustments to the model parameters are This paper utilizes the “Kuenm” package in R version 4.2.3 to explore various combinations of eight regularization multipliers (RM) and six feature combinations (FC). FC includes L (linear liner), T (threshold), Q (quadratic type), H (fragmented hinge), and P (product type threshold). Regularization is adjusted from 0 to 4, increasing in increments of 0.1. The optimal parameter combination, characterized by data omission rates below 5% and a small deltaAICc value, was selected to run the final model25.

The selection of remaining parameters was conducted using a random seed method. Seventy-five percent of the data was used as a training set to build the model, while the remaining 25% served as a test set to validate the model’s accuracy26. This procedure was replicated ten times with a subsampling run type. The Jackknife method evaluated the significance and contribution rate of each environmental variable in the model, presenting results in Logistic form27. The area under the curve (AUC) value indicates prediction accuracy: 0.5–0.6 suggests failure, 0.6–0.7 is poor, 0.7–0.8 is moderate, 0.8–0.9 is good, and 0.9–1 is excellent28.

The final output of the MaxEnt model was imported into ArcGIS and converted into raster data, classified into four suitability levels (high, medium, low, non-suitable) using the natural breakpoint method. The proportion and area of each level were calculated. Raster data were binarized according to suitability, with areas below the threshold of 0.1 deemed low suitability and assigned a value of 0, and others a value of 1. Overlay analysis of the binary matrices across different periods identified stable, expanding, shrinking, and non-suitable areas, calculating changes in suitability and analyzing spatial patterns in potential suitable areas over time29. Simulations of centroid shifts at different times were performed using the SDM toolbox.

Results

Model construction and accuracy checking

Among the 1160 model parameter combinations tested by FC and RM (Fig. 2), the optimal parameter combination with an Omission Rate less than 5% and a minimal deltaAICc value is LT for FC and 2.1 for RM, with an Omission Rate of 0.06 and a deltaAICc of 0. The model’s accuracy is assessed by the area under the ROC curve, the AUC value, which is 0.881 for the cultivated land training data. This high AUC value indicates that the model’s simulation precision and predictive reliability for cultivated land are excellent.

Fig. 2
figure 2

Model accuracy checking (a) Model parameter combination (b) Receiver curve (ROC) and AUC of cultivated land suitability distribution model.

Full size image

Main factors affecting the distribution of cultivated land

After initial selection, a pre-simulation was performed to further eliminate variables with low contribution rates and displacement, enhancing model accuracy. The chosen environmental variables and their contribution rates used for analyzing suitable distribution areas are detailed in Table 2.

Table 2 Environmental variables and contribution degree after removing spatial correlation.
Full size table

Thirteen variables were analyzed to determine the distribution of suitable cultivated land areas. The variables and their contribution rates are listed in Table 2, and the results of the cutting test are presented in Fig. 3. The cutting test revealed that the model performs best when using elevation alone. Without elevation, the model’s performance significantly declines, highlighting elevation as the most critical factor influencing cultivated land distribution. The main environmental factors affecting the distribution are elevation, mean daily temperature range (BIO2), clay content, and gravel, with contribution rates of 43.3%, 22.2%, 9%, and 7%, respectively, adding up to a cumulative contribution rate of 81.5%.

Fig. 3
figure 3

Knife cut test results.

Full size image

The response curves correlating cultivated land suitability with key environmental factors are illustrated in Fig. 4.There is a negative correlation between cultivated land suitability and factors such as elevation, average daily temperature range, clay content, and gravel volume percentage. Suitability decreases as these variables increase. Optimal conditions for cultivated land are defined as elevation less than 532.2 m, average daily temperature range less than 8.8 °C, clay content less than 63.1%, and gravel volume percentage less than 27.5%, corresponding to a suitability index (LOV) greater than 0.1. Since topographic soil indexes do not change significantly with time, this paper only plots the change of BIO2 range of suitable cultivated land distribution under future scenarios (Fig. 5). As can be seen from the figure, from current to future, the distribution center of BIO2 starts from Dongguan and moves southwestward. Its center remains within the territory of Guangdong throughout.

Fig. 4
figure 4

Response curve of environmental factors (a) elevation (b) daily average temperature (c) clay content (d) percentage of gravel volume range of cultivated land in southeast coastal area.

Full size image
Fig. 5
figure 5

Changes in the appropriate BIO2 range in future scenarios.

Full size image

Distribution and change of cultivated land in multiple periods

Under the current climate scenario, suitable cultivated land areas are dispersed, covering a total of 103,390 km2 along the coastal margins of the study area, which accounts for 35.4% of the southeast coastal area’s total (Table 3). These areas are primarily located along the coasts of Guangxi, Guangdong, Hainan, Fujian, and the eastern and western edges of Taiwan (Fig. 6). The distribution of suitability areas is divided into high, middle, and low suitability zones, with ratios of 1:2.6:6.3. The high suitability area encompasses 10,430 km2, or 10.1% of the total suitability area, mainly found in the coastal regions of Guangxi, Guangdong, Fujian, northeastern Hainan, and eastern and southwestern Taiwan. The middle suitability area spans 26,871 km2, or 26% of the total, predominantly in southern Guangxi and Guangdong, northern Hainan, southeastern Fujian, and western Taiwan. The low suitability area covers 66,089 km2, or 63.9% of the total, mainly distributed across central and southern Guangdong and Guangxi, Hainan, southeastern Fujian, and eastern and western Taiwan.

Table 3 Potential suitable area of cultivated land during different periods (km2).
Full size table
Fig. 6
figure 6

Potential suitable areas of cultivated land under different climate scenarios (Created using ArcGIS 10.8.0, https://www.arcgis.com/index.html).

Full size image

In future scenarios under two greenhouse gas emission models, the pattern of cultivated land in the southeast coastal area remains similar, primarily located along the coastal edge and central areas of Guangdong and Guangxi, with a general declining trend in the total cultivated land area. Under the SSP126 scenario, the total suitable area decreased from 94,253 to 85,789 km2, accounting for 32.3% and 29.4% of the study area, respectively, marking a decrease of 2.9%. Under the SSP585 scenario, it decreased from 103,642 to 93,484 km2, accounting for 35.5% and 32% of the study area, respectively, showing a 3.5% decrease. Between the current period and the 2060s, the total adaptive area decreased by 17,601 km2 under the SSP126 model and by 9906 km2 under the SSP585 model.

The analysis indicates that the overall trend for cultivated land in the future is a decrease. However, both in the 2040s and 2060s, the suitable area under the SSP585 scenario was consistently larger than under the SSP126 scenario, suggesting that lower concentrations of greenhouse gases have a more significant impact on reducing cultivated land.

According to changes in spatial patterns under different climate scenarios, the area of cultivated land suitability generally decreased, with contraction consistently exceeding expansion (Table 4). The majority of the original potential suitable areas remained stable, with minimal changes located primarily near the coastal edges of the study area (Fig. 7).

Table 4 Potential suitable area of cultivated land in different periods (km2).
Full size table
Fig. 7
figure 7

Changes of potential suitable area of cultivated land in different periods (km2) (Created using ArcGIS 10.8.0, https://www.arcgis.com/index.html).

Full size image

Under the SSP126 future climate scenario, the area suitable for cultivation is projected to reduce by 3483 km2 (3.4%) by the 2040s, mainly affecting the southwest and central Guangdong, southern Guangxi, eastern Hainan, and eastern Fujian. From the 2040s to the 2060s, the suitable area further decreased by 3096 km2, with a shrinkage rate of 3.3%, primarily in central and coastal Guangdong, northwest and southeast Hainan, and eastern Taiwan. Under the SSP585 scenario, the area shrinks by 255 km2 by the 2040s, a decrease of 0.3%, mainly in central Guangdong and southern Guangxi. From the 2040s to the 2060s, the decrease totals 3964 km2, with a shrinkage rate of 3.8%, affecting southeast Fujian, southwest and central Guangdong, and eastern and western Hainan.

Migration route of centroid in the suitable area

The current centroid of mass is in Foshan City (113.146° E, 22.983° N). Under the SSP126 and SSP585 scenarios, the centroid generally moves southeast, ultimately positioning in the South China Sea (Fig. 8).

Fig. 8
figure 8

Change of centroid of potentially suitable cultivated land in different periods.

Full size image

Specifically, in the SSP126 scenario, the centroid moves 24.04 km southwest by the 2040s (Table 5), remaining in Foshan City (112.918° E, 22.935° N). From the 2040s to the 2060s, it shifts southeast by 110.73 km, reaching the South China Sea (113.663° E, 22.213° N).

Table 5 Longitude and latitude of centroid distribution of cultivated land in different periods.
Full size table

In the SSP585 scenario, the centroid moves southeast by 98.01 km from the present to the 2040s, reaching the South China Sea (113.829° E, 22.365° N), and then moves 14.95 km southwest from the 2040s to the 2060s, remaining in the South China Sea (113.707° E, 22.292° N).

Discussions

Model construction and accuracy checking

Compared with other species distribution models, the MaxEnt model predicted the highest consistency with the actual distribution. Many scholars have shown that MaxEnt default parameters have complexity and high fit. But in previous similar studies, most modeled the distribution of species based on default parameters. Such analyses tend to lead to overfitting and high complexity, which may reduce the accuracy of the study results. Having optimized parameters improves the stability of the MaxEnt model compared to the default parameters. It makes the response curve closer to the normal distribution curve. In this study, 1160 parameter combinations were established to reduce the model error. The results show that the MaxEnt model optimized with parameters FC = LT and RM = 2.1 is optimal.

Main factors affecting the distribution of cultivated land

Previous studies have generally identified climate factors such as precipitation and temperature as the dominant drivers of cropland distribution. However, our findings demonstrate that topography plays a more critical role in shaping cropland distribution in the southeastern coastal region of China. For instance, Huang emphasized that water resource availability is the primary constraint on cropland in the North China Plain, where enhancing regional water-resource carrying capacity is essential for agricultural sustainability30. In contrast, precipitation-related variables were excluded from our final model, likely due to the monsoon-dominated climate in the southeastern coastal area, which ensures sufficient and stable annual rainfall, thereby diminishing precipitation as a limiting factor for cropland distribution. This discrepancy highlights the heterogeneity of dominant drivers across different geographical contexts. Furthermore, Paolo argued that topographic gradients critically influence cropland expansion potential in mountainous and hilly regions31, particularly through indirect effects on crop growth via light availability, surface runoff, and soil nutrient redistribution. In our study area, characterized by rugged terrain, high-altitude zones with limited heat and poor soil conditions were more suitable for forestry, while low-altitude gentle slopes with balanced hydrothermal resources emerged as optimal habitats for cropland. These results align with existing research on terrain-driven suitability differentiation and quantitatively refine the understanding of multi-scale terrain-ecosystem interactions through spatial modeling with MaxEnt32,33,34.

Distribution change of potential suitable area of cultivated land

The current and future potential for suitable cultivated land is stable along the coastal edges of southeast China, primarily influenced by topographic factors. According to the analysis, the southeast coastal region predominantly consists of low mountains and hills, comprising 60% of the land and mainly found in western Fujian, northern Guangdong, southern Hainan, and central Taiwan. The simulation results match the actual land distribution, affirming the model’s accuracy. From now to the future, there has been a downward trend in the potential suitability of cultivated land. This decrease may be attributed to mass population migrations leading to the abandonment of cultivated land in rural areas, with the land being repurposed for construction or reforested. Additionally, the long-term viability of certain crops may be decreasing due to persistent suboptimal conditions. For instance, in southeast coastal China, the main crops—rice, tea, citrus fruits, other food crops, and vegetables—are affected by climate, soil, and topography. Li Xinhua’s research suggests that rice, in particular, has a very low development potential in the Yangtze River basin and coastal areas, with a significant decline in suitable regions35, corroborating the findings of this study and highlighting the complexity and seriousness of China’s decreasing cultivated land.

Under future greenhouse gas (GHG) emission scenarios, suitable cultivated lands are contracting. Looking forward to the 2060s, the total suitable area under the SSP126 model shrinks more than under the SSP585 model, suggesting more severe land degradation under the low emission scenario, while the impact is less marked under the high emission scenario. Different climate models, especially those forecasting low GHG emissions, significantly influence the reduction in cultivated land area and the migration of its centroid. This finding aligns with Wang Haoran’s study, which found that the suitable areas for Christmas trees decreased more under low-emission scenarios36. Additionally, Chen predicted a significant reduction in cultivated land in the Changzhutan region of China under the SSP126 scenario using a similar model37, further supporting the conclusions of this study. However, Wang’s research suggests that under different warming scenarios, as the warming amplitude increases, the frequency and intensity of extreme climate events significantly increase, leading to a further reduction in cultivated land area38. This finding differs from the conclusions of this study, possibly due to the unique mountainous terrain and relatively stable climatic conditions of the southeastern coastal region studied here. These geographical and climatic features may have created local ecological environments that mitigate the negative impacts of climate change on cultivated land suitability to some extent.

In the two future scenarios, the center of mass shifts southeast, moving from high to low latitude and ultimately stabilizing in the South China Sea. In the SSP126 scenario, the migration occurs from southwest to southeast, while in the SSP585 scenario, it moves from southeast to southwest. Contrarily, Xu Tianyun’s findings indicate that under various scenarios, the potential suitable areas for mangroves along the southeast coast expand and the center of mass moves north39, contradicting the results of this paper. In addition, some studies showed that BIO6 and BIO2 were the main climate factors affecting the distribution of Xianghenuo, and there was a trend of migration to high latitude and high altitude in the future40. These results are inconsistent with the results of this paper. The reason for this difference may be because the middle and high latitudes are heat restricted, and the key factor determining crop growth is temperature. Global warming has extended the climate belt northward, and crops that were originally suitable for growing in the south are now more suitable for growing in the warmer northern regions, and eventually the agricultural center of gravity has shifted northward. However, in tropical and subtropical regions, heat is abundant and temperature is usually not a limiting factor, in contrast to diurnal temperature difference, which becomes a key environmental factor and has a large impact on crop growth, as excessive temperature differences may stress the crop and affect its growth and yield. The result of this study is that the daily temperature difference between 4.5 and 8.8 °C is suitable for crop growth in the coastal areas of central and southeast China. In the future climate scenario, the diurnal temperature difference suitable for crop growth moves to the south obviously, and the cultivated land centroid also moves to the south. In addition, the study on the relationship between cultivated land and other factors in coastal areas is largely influenced by the study scale and time41,42. In this study, the migration of the center of suitable cultivated land towards low—latitude regions contradicts the results of China’s macro—research. This discrepancy may be due to differences in research scopes. For example, the cultivated land replenishment under the guidance of the protection policy for the black soil in Northeast China dominates the national macro—migration trend43. This paradoxical phenomenon of “local southward migration—overall northward shift” indicates that there are significant turning characteristics in the scale domain of the evolution law of cultivated land. It shows that in the future of global warming, except for topographic factors, the cultivated land in the southeast coastal area is mainly affected by diurnal temperature difference, and will migrate with the distribution of the suitable diurnal temperature difference range.

Cultivated land protection proposal

As a major agricultural nation, China supports about 21% of the world’s population with only 9% of its cultivated land44. However, climate change and human disturbances have significantly degraded the quality of this land, drastically reducing its area and leading to fragmented patterns. This situation poses immense challenges for China’s agricultural development. On the one hand, through the research and development of agricultural science and technology, main grain crops such as rice and wheat that can maintain stable yield under large temperature differences can be selected, or fruit and vegetable varieties suitable for coastal climate will be promoted45,46. On the other hand, technologies such as film mulching, drip irrigation and sunshade net are used to regulate the field microclimate and reduce the impact of temperature fluctuations on crops47,48. For example, during cold wave, film insulation is applied, and the light intensity is reduced through shading net during high temperature. In addition, under these different climate scenarios, the existing cultivated land should be strictly protected, and the policy of “balance of occupied and newly added cultivated land” should be implemented to ensure that the cultivated land occupied by non-agricultural construction must be supplemented with equal quality or better cultivated land, and the red line of cultivated land should be kept49. For example, the cultivated land occupied in coastal areas should be supplemented by reclamation saline-alkali land or improving inefficient forest land. To achieve low-carbon and efficient use of space is the key to the long-term stability of China’s agriculture.

Deficiencies and prospects

In this paper, we prioritize 32 natural environmental factors related to temperature and precipitation with high confidence levels; however, it did not consider biological factors, typhoons, species interactions, and human interference. The anthropogenic factors often need to be coupled with system dynamics model or CA–Markov model to realize dynamic simulation, and we fully agree that the anthropogenic factors have an important influence on the actual distribution of cultivated land. The results of this paper are only based on natural factors, and the real distribution of cultivated land superimposed on anthropogenic activities needs to be further explored through multi-model coupling or spatial policy analysis. We also plan to integrate the land use change model with socioeconomic scenarios to improve the prediction framework in subsequent studies. Additionally, while the BCC-CSM2-MR atmospheric circulation model was used for projecting future climates, incorporating other models such as CNRM-CM6-1, CNRM-ESM2-1, and CanESM5 could provide a more reliable average simulation result for an SSP scenario, thus reducing uncertainties50. In subsequent research, it will be crucial to cultivate plant species with strong adaptability tailored to the actual conditions of the study area. Adjustments to farming methods and further predictions of potential suitable cultivated land areas will help in the protection and restoration efforts. This approach aims to strategically plan land resources in the southeast coastal area and address the conflict between urbanization and cultivated land protection.

Conclusions

R-optimized MaxEnt model was used to simulate and predict distribution of cultivated land, indicating high precision and reliability. The primary environmental factors influencing cultivated land distribution are elevation and diurnal temperature range, cumulatively accounting for 65.5%. Under the current climate situation, there are 35.4% of total area suiting for cultivation in the southeast coastal area, primarily distributed in the coastal margins and central regions of Guangdong and Guangxi.

Under different future climate scenarios of emissions, suitable cultivation land will decline, although it will remain along the coastal margin. However, the center of mass of this land is likely to shift towards lower latitudes and altitudes in the southeast. The most suitable areas for cultivation are the low-lying, plain coastal regions with loose, nutrient-rich soil and stable climate conditions; less so in the inland mountainous areas. With a well-planned farming system that fully considers environmental impacts and adapts to local conditions, the production of crops and cash crops in the southeast coastal areas could be increased, potentially addressing some global food issues. In the future, it is necessary to intensify the protection of existing cultivated land and implement a strict system of balancing the occupation and compensation of cultivated land to ensure that the amount of cultivated land does not decrease. Investigate and evaluate land resources in low-latitude areas, pay attention to the protection of cultivated land resources while promoting the development of low-carbon agriculture, and make preparations for undertaking the transfer of cultivated land.