Impact of Farmland Use Transition on Grain Carbon Sink Transfer in Karst Mountainous Areas

Abstract

Farmland use transition (FUT) not only reshapes agricultural production systems but also significantly impacts cross-regional carbon sink transfers in the grain trade. However, comprehensive studies exploring connections between FUT and grain carbon sink transfer (GCST) are limited. We constructed an indicator system and transformation framework for FUT by considering dominant and recessive dimensions. Moreover, we estimate GCST based on grain supply–demand balance and fixed carbon coefficients. Fixed effects and threshold models are employed to identify both linear and nonlinear relationships between FUT and GCST. Results show that FUT significantly reshapes carbon sink flows. In terms of dominant FUT indicators, cultivation land rate (CLR) and grain planting area proportion (GPAP) positively drive GCST by expanding the carbon sink supply and exporting ecological services. Regarding recessive FUT indicators, both grain yield per unit area (GYield) and pesticide-fertilizer intensity (PFI) promote GCST, highlighting the role of efficiency and inputs, while rural per capita disposable income (RPCDI) suppresses GCST due to agricultural marginalization. A grain yield threshold of 2.092 t/ha is identified. Below this value, FUT exerts substantial positive effects on GCST. Above it, the effects weaken. This study explains the relationship between FUT and ecosystem carbon sinks, providing a scientific basis for advancing green agriculture in karst mountainous areas.

1. Introduction

The current global warming conditions pose increasingly severe threats to human survival [1,2]. Carbon sequestration on agricultural land offers substantial mitigation potential [3]. Promoting FUT has thus become a critical pathway for building climate-resilient agriculture and an effective means to enhance regional carbon sink capacity. The synergistic realization of SDG2 (Zero Hunger) and SDG13 (Climate Action) demands the spatial reorganization of cropland systems and the optimized allocation of land and technological resources to address the dual challenge of food security and carbon sequestration. However, regional disparities in resource endowments, barriers to technology diffusion, and inconsistent policy responses have led grain trade to evolve into a hidden channel of carbon sink redistribution, where the carbon sink services embedded in grain production are unilaterally transferred from surplus (production) regions to deficit (consumption) regions. This spatial imbalance in grain carbon sink transfer (GCST) exacerbates the asymmetry of carbon responsibility allocation. In karst mountainous areas, declining marginal returns to grain production and limited ecological resilience compound the constraints on enhancing carbon sink supply. Hence, investigating the influence of FUT on GCST has emerged as a pivotal scientific question in facilitating the transition towards low-carbon and climate-resilient agriculture in ecologically fragile mountainous regions.

The theoretical system of land use transition originated from forest transition, with its core paradigm emphasizing the dynamic coupling relationship between land use patterns and socio-economic stages; that is, specific development stages correspond to characteristic land use configurations [4]. Existing international research primarily focuses on transitions at the national level, such as forests [5,6] and grasslands [7], with a particular emphasis on ecological and environmental effects [8] and driving mechanisms [9]. In China, it extends from the theory of broad land use transformation [10,11], proposing two forms: explicit and implicit. The research framework covers transformation diagnosis, mechanism, and effect research [12]. Within farmland use transition research, a robust system has emerged addressing transition pathways [13,14], spatiotemporal pattern evolution [15,16], ecological effects [17], driving mechanisms [18], and regional heterogeneity [19]. This scope has expanded to interdisciplinary domains such as rural homestead dynamics [20] and functional transitions [21]. Spatial econometric models have further elucidated the regulatory role of farmland use transition in carbon emissions [22]. Under carbon neutrality constraints, cross-regional carbon competition has intensified the cross-border transfer of embedded environmental burdens in food trade [23,24]. With the increase in human activities and inter-regional interactions, the role and impact of cross-boundary carbon sequestration service flows are becoming increasingly evident [25,26]. FUT serves as a pivotal driving force influencing ecosystem carbon sequestration [27]. While converting farmland to forests and returning farmland to wetlands can enhance local carbon sequestration capacity, such practices may also alter the regional carbon budget dynamics by shifting food production areas [28]. Therefore, when assessing carbon sequestration services in a particular region, it is crucial to incorporate the spatial flow factors of carbon services to minimize biases in the evaluation results [29,30]. Despite these advancements, several research gaps remain. First, existing studies have yet to explore the impact of FUT on GCST. Second, the spatial heterogeneity of GCST, particularly in terms of regional imbalances in grain supply and demand, has received relatively little attention. Third, the mechanism of FUTs influence on GCST is rarely considered for its threshold effect.

This study addresses three objectives to fill existing research gaps: First, it constructs a FUT indicator system and identifies its influence pathways on GCST. Second, based on regional grain production, grain demand for urban and rural residents’ food consumption and animal feed, combined with unit farmland carbon absorption factors, it calculates regional grain supply-demand relationships and quantifies the spatial transfer characteristics of GCST. Third, it identifies the evolution patterns and mechanisms of FUTs impact on GCST and proposes optimization pathways for FUT, as well as regionally differentiated carbon sink enhancement strategies.

Research contributions include: (1) It constructs a FUT indicator system integrating both dominant and recessive dimensions, and combines regional grain supply-demand gaps with carbon sink coefficients to quantify GCST, thereby revealing the spatial redistribution of carbon sink services driven by interregional grain flows; (2) It employs fixed-effects and threshold regression models to uncover the linear and nonlinear effects of FUT on GCST, identifies a critical yield threshold (2.092 t/hm²), and confirms the stage-dependent regulation of carbon sink transfer; (3) It provides empirical evidence and regional policy recommendations to support the coordinated promotion of food security and carbon sink capacity in mountainous areas.

2. Mechanistic Analysis

2.1. Conceptual Definitions

To examine the impact of FUT on GCST, it is necessary to clarify both concepts. FUT refers to the dynamic adjustment process in farmland use patterns, structure, and functions, encompassing dominant (DFUT) and recessive (RFUT) transitions. These changes are driven by socio-economic development, policy interventions, and ecological conditions, and they evolve across different development stages [31,32]. DFUT represents the quantitative dimension of change—such as alterations in farmland area; cropping structure; and spatial configuration—which directly influence grain production capacity and land use efficiency. In contrast, RFUT reflects the qualitative dimension, focusing on improvements in farmland management models, technological adoption, and production efficiency.

Grain Carbon Sink Transfer (GCST) refers to the spatial and temporal reallocation of carbon sink services embedded in grain production, as grain flows from surplus (production) regions to deficit (consumption) regions. Through interregional grain trade, farmland-based carbon sequestration in producing areas is effectively transferred to consuming areas, forming a virtual flow of carbon sinks along the agri-food supply chain. The GCST concept is rooted in the broader virtual ecosystem service trade framework [33]. It reflects a form of carbon sink outsourcing, where urban consumption centers benefit from carbon sequestration performed in rural or ecologically fragile areas without directly bearing the environmental responsibility [34]. This process highlights regional disparities in carbon sink contributions and reveals the need for more equitable carbon accounting. In less-developed areas—such as mountainous regions with weak regulatory capacity—GCST may also represent carbon sink leakage; where ecosystem services are transferred without recognition or compensation. Recognizing GCST as a key feature of agri-environmental interactions underscores the importance of territorialized carbon sink accounting, interregional ecological equity, and differentiated governance strategies for sustainable agricultural transitions.

2.2. Mechanism Analysis of GCST

2.2.1. Direct Effects

(1)

FUT

DFUT directly drives GCST through cropland quantity and cropping structure adjustments. Changes in cropland quantity are often accompanied by cropland conversion to non-agricultural and non-grain uses [35]. Changes in cropland area imply intensified land use through reclamation or multiple cropping, which expands total photosynthetic capacity and increases the embedded carbon sink in regional grain output. As surplus grain is traded across regions, the associated carbon sink services are also redistributed, thus increasing GCST. Additionally, the configuration of the planting system influences the quantity of carbon sequestered within the soil. Non-grain conversion has altered the traditional farmland use pattern, weakened the carbon sink effect of farmland, and enhanced its carbon source attribute [36].

In contrast, RFUT affects GCST through technological progress and input optimization. GYield increases indicate improved land productivity and stronger carbon fixation per hectare, driving up GCST through more efficient output [27]. When managed efficiently, PFI supports yield gains without proportional environmental costs, enhancing the carbon sink capacity of cultivated systems [37]. However, RPCDI may reduce GCST by accelerating the marginalization of grain production in favor of higher-value agricultural or non-agricultural land uses, weakening regional grain output, and reducing the carbon sink services available for interregional transfer [38].

2.2.2. Threshold Effect

The impact of FUT on GCST exhibits clear nonlinear characteristics, with GYield serving as a structural threshold dividing two stages of development.

In the low-yield stage, agricultural growth relies on land expansion and the intensive use of inputs. These practices increase grain output and total carbon sink capacity, generating larger surpluses available for interregional transfer and thereby intensifying GCST [39]. In the high-yield stage, the logic of agricultural development shifts toward efficiency. FUT becomes increasingly driven by technological innovation, soil conservation, and precision in inputs. However, ecological constraints and the plateauing of productivity gains reduce the marginal growth of carbon sink output [40]. As a result, the positive impact of FUT on GCST persists, but with diminishing returns, reflecting the ecological limits of high-yield agricultural zones. As a result, the positive impact of FUT on GCST persists, but with diminishing returns, reflecting the ecological limits of high-yield agricultural zones (Figure 1).

3. Materials and Methods

3.1. Study Area

Guizhou is in the Yunnan-Guizhou Plateau, with a higher elevation in the west and lower in the east. It belongs to the subtropical humid monsoon climate zone and has various biological resources. The province’s total area is 176,100 square kilometers, characterized by a relatively steep slope of arable land and relatively outdated agricultural production technology. However, the characteristics of three-dimensional agriculture are significant, making it suitable for developing regional agricultural green production and modern high-efficiency mountain characteristic agricultural industrial clusters. As of 2020, the total population was 38.56 million, of which the primary industry employed 6.43 million people, accounting for 16.67% of the total population. The arable land area is 3415.03 hectares, the grain sowing area is 2754.13 hectares, the effective irrigation area is 1165.49 hectares, the agricultural mechanization is 30.01 million kilowatts, the use of agricultural fertilizers is 787,800 tons, and the total grain output is 10.5763 million tons.

3.2. Data Description

The basic input-output data on cultivated land production used in this study, including cultivated land area, permanent resident population, total sown area of crops, sown area of grain crops, agricultural fertilizer application amount, pesticide usage, disposable income of rural residents, and grain output, are all sourced from the “Guizhou Statistical Yearbook” over the years, based on the actual data of each year. In addition, according to the national social grain supply and demand balance survey provided by the National Food and Reserve Administration, the demand for food and feed grains for urban and rural residents accounts for 80% of the total grain demand, with food grains accounting for 38% and feed grains accounting for 42%. Therefore, the grain demand in this study mainly considers the demand for food and feed grains of urban and rural residents in the region. Specifically, the data on food grains for urban and rural residents is sourced from the food consumption survey data of residents in Guizhou Province in the “Guizhou Statistical Yearbook”. In contrast, the demand for feed grains is estimated based on the “grain consumption per main product” of some livestock products in the “Compilation of National Agricultural Product Cost and Income Data”, calculated as “feed grain demand = adjusted livestock product output × feed conversion rate”. This calculation covers the feed grains consumed for food such as pork, beef, mutton, poultry meat, poultry eggs, aquatic products, and dairy products.

It is worth noting that these areas are considered unsampled regions in this study due to significant data missing in Qiannan, Qiandongnan, and Qianxinan prefectures. To address the missing data, this study imputed the missing data for specific years and regions using linear interpolation and adjusted the data based on relevant indicators.

3.3. Methodology

(1)

Entropy evaluation method

This study employs an entropy evaluation method to measure the FUT, with the specific calculation formula as follows [41]:

$$\begin{array}{ll}\hfill F_i& ={\displaystyle \sum _{i=1}^n{x}_{qij}{w}_j},w_j=f_j/{\displaystyle \sum f_j}\\ \hfill f_j& =1+\ln (rn){\displaystyle \sum _q{\displaystyle \sum _i\frac{x_{qij}}{{\displaystyle \sum _q{\displaystyle \sum _i{x}_{qij}}}}}}\ln {\displaystyle \sum _i\frac{x_{qij}}{{\displaystyle \sum _{q}i{\displaystyle \sum x_{qij}}}}}\end{array}$$

(1)

In Equation (1): F_i represents the FUT in the city area; xq_ij is the standardized value of the city area q on the th indicator i in the th year, and the standardization of indicators is carried out using the maximization method and minimization method; w_j is the weight of the j th indicator, and to avoid subjectivity in determining the weight, the entropy method is used. Since the traditional entropy method can only handle cross-sectional data, while this study is based on panel data, the entropy method incorporating time variables is adopted for determining the weight. f_j is the information redundancy of the j th indicator, with r and n representing the research time limit and the number of evaluation units in the city, respectively.

In the selection of indicators, we referred to relevant articles on FUT and ecosystem carbon sinks. For dominant transformation, many scholars believe that changes in cropland area reflect trends in cropland transformation and are the core data for studying cropland transformation [42]. Therefore, we used the cultivated land rate to reflect quantitative changes in the dominant transformation of farmland utilization. In prior research, the structure of cultivated land use has commonly mirrored the internal structural adjustments within farmland. Various crops grow through different biophysical processes and exert diverse impacts on regional ecosystems [43]. By drawing upon previous studies and considering the specific circumstances of the study area, GPAP was utilized to quantify the shifts in cultivated land use structure.

Regarding recessive transformations, we focused on the function dimensions. GYield reflects the output level of cropland utilization and is an important indicator used to assess the degree of FUT. Related studies usually take them as implicit transformation indicators that represent the production function of arable land [44,45]. Therefore, we selected GYield as a measure of the production function dimension. Research indicates that pesticides and fertilization significantly affect crop photosynthesis, thereby influencing grain yield and ecosystem carbon sequestration [46,47]. Therefore, using pesticide-fertilizer intensity as an indicator to represent the transition in ecological function is of great significance for understanding the role of agricultural activities in the evolution of FUT and GCST. Changes in land-use patterns and regional ecological effects resulting from rural residents’ income have attracted widespread attention [48]. We choose rural per capita disposable income to reflect the functional transformation of livelihoods.

(2)

Indicator calculation method

To quantify GCST, this study constructs a carbon sink flow matrix by multiplying the inter-city grain trade volume by the fixed carbon sink coefficient of grain production in the corresponding source city [46]. The trade volume is estimated based on population, per capita grain consumption, feed grain demand, and local grain output [47].

Given Guizhou’s relatively consistent agroecological conditions, a fixed carbon sink coefficient is adopted to ensure comparability across regions and reduce uncertainty in estimation. This approach captures the interregional balance of grain supply and demand, allowing the embedded carbon sink services—carried by grain flows—to be spatially allocated. The calculation formula is as follows:

$$GCST={\displaystyle \sum _{i=1,j=1}^{i=6,j=6}(C_{ij}\cdot Q_i})={\displaystyle \sum _{i=1,j=1}^{i=6,j=6}(C_{ij}\cdot \frac{NP{P}_i}{C{Y}_i}})$$

(2)

In Equation (2), GCST represents the total grain carbon sink transfer (t), C_ij denotes the volume of grain traded from source city i to recipient city j (t)., Qi is the fixed carbon emission per unit of grain production in city i (t), NPP_i is the net primary productivity of farmland in city i (4.243 t/hm⁻²) [48], and CY_i is the grain yield per unit of cropped area (t/hm²).

(3)

Fixed effects model

This study employs a fixed effects model based on the earlier theoretical analysis. The baseline model is established as follows:

GCST_i,t = a_i + β₁CLR_it + β₂GPAP_it + β₃GYield_it + β₄RPCDI_it + β₅PFI_it+ ε_i,t

(3)

In Equation (3), CLR, GPAP, GYield, RPCDI, and PFI represent cultivation land rate, grain planting area proportion, grain yield per unit area, rural per capita disposable income, and pesticide-fertilizer intensity, respectively. a_i denotes individual fixed effects, β₁–β₅ are the regression coefficients, and ε is the error term.

(4)

Threshold effect model

To test the nonlinear relationship between FUT and GCST, the panel threshold regression model [49] is used, as follows:

$$GCS{T}_{it}={\mu }_i+{\beta }_{1}FU{T}_{it}\cdot I(q<\eta )+{\beta }_{2}FU{T}_{it}•I(q\ge \eta )+{\displaystyle \sum _{k=1}^4{\theta }_k}\cdot C_{k,it}+{\epsilon }_{it}$$

(4)

In Equation (4), X is the core explanatory variable FUT, η is the threshold value, q is the threshold variable, I (·) is the indicator function, and θ are the regression coefficients of control variables. Other variables are defined as in Equation (3).

4. Results

4.1. Spatiotemporal Evolution Characteristics of FUT and GCST

From 2001 to 2020, the FUT index in Guizhou Province generally showed a fluctuating upward trend (Figure 2). From a regional perspective, the farmland use transition index in Zunyi was relatively high, with an average of 0.431, significantly higher than in other areas (

Table 1. Guizhou FUT from 2001 to 2020.

Year	Anshun	Bijie	Guiyang	Liupanshui	Tongren	Zunyi
2001	0.307	0.313	0.352	0.29	0.291	0.324
2002	0.285	0.32	0.334	0.295	0.29	0.32
2003	0.307	0.317	0.343	0.302	0.295	0.332
2004	0.308	0.332	0.348	0.305	0.304	0.343
2005	0.308	0.346	0.347	0.305	0.305	0.353
2006	0.311	0.342	0.349	0.307	0.316	0.343
2007	0.273	0.333	0.303	0.301	0.295	0.359
2008	0.293	0.353	0.324	0.31	0.315	0.374
2009	0.29	0.311	0.328	0.31	0.3	0.381
2010	0.299	0.361	0.348	0.307	0.318	0.393
2011	0.28	0.359	0.352	0.308	0.285	0.368
2012	0.313	0.393	0.376	0.347	0.326	0.417
2013	0.325	0.412	0.407	0.379	0.34	0.43
2014	0.369	0.438	0.438	0.396	0.37	0.494
2015	0.377	0.464	0.461	0.413	0.393	0.522
2016	0.396	0.487	0.489	0.425	0.407	0.55
2017	0.423	0.482	0.511	0.44	0.401	0.535
2018	0.447	0.517	0.52	0.435	0.409	0.553
2019	0.492	0.536	0.569	0.465	0.432	0.597
2020	0.51	0.565	0.605	0.482	0.452	0.631

). The reason is that Zunyi has a higher level of economic development and a faster process of agricultural modernization, leading to a notable improvement in the intensification and efficiency of farmland use. The policy support and technological applications related to FUT in these areas have a pioneering effect, resulting in relatively higher FUT indices. Regarding the distribution characteristics of the FUT index, Zunyi City, Bijie City, and Guiyang City ranked at the top, while Tongren City, Anshun City, and Liupanshui City ranked lower. The possible reasons include the relatively lagging economic development in the eastern and southern regions, traditional agricultural production methods, complex terrain, and weak agricultural infrastructure in some areas, which lead to insufficient FUT momentum and lower transition indices.

The spatial and temporal evolution of GCST in Guizhou Province reveals a distinct core–periphery structure with dynamic regional shifts (

Table 2. Spatiotemporal patterns of GCST in Guizhou from 2001 to 2020 (10⁶t).

Year	Anshun	Bijie	Guiyang	Liupanshui	Tongren	Zunyi
2001	0.202	1.074	−0.263	0.176	0.576	0.860
2002	−0.075	0.992	−0.607	0.118	0.515	0.703
2003	0.110	0.754	−0.373	0.114	0.511	0.774
2004	0.100	0.941	−0.356	0.098	0.537	0.703
2005	0.104	1.513	−0.554	0.121	0.639	0.791
2006	0.031	0.885	−0.453	0.054	0.409	−0.154
2007	−0.021	0.683	−0.398	0.028	0.179	1.131
2008	0.068	0.771	−0.209	0.118	0.416	1.135
2009	0.048	0.271	−0.230	0.107	0.330	1.058
2010	−0.006	0.537	−0.223	0.028	0.303	0.934
2011	−0.278	−0.273	−0.701	−0.259	−0.439	−0.129
2012	−0.139	0.157	−0.574	−0.014	−0.092	0.389
2013	−0.188	0.059	−0.646	0.021	−0.214	0.114
2014	−0.156	0.197	−0.595	0.015	−0.015	0.409
2015	−0.179	0.241	−0.610	0.010	−0.015	0.458
2016	−0.138	0.373	−0.592	−0.003	0.045	0.558
2017	−0.138	0.438	−0.592	0.011	0.038	−0.073
2018	−0.015	0.346	−0.572	−0.091	−0.162	−0.157
2019	−0.001	0.254	−0.542	−0.142	−0.186	−0.072
2020	−0.010	0.270	−0.605	−0.189	−0.239	−0.199

“+” = exports carbon sink services; “−” = receives carbon sink services.

). From 2001 to 2010, Bijie and Zunyi were the province’s principal carbon sink output regions, driven by large-scale grain production and favorable dryland conditions. Bijie recorded the highest GCST levels (e.g., 1.513 in 2005), while Zunyi maintained consistently strong output. After 2011, both cities experienced moderate declines but continued to be net contributors to carbon sink services. Guiyang consistently exhibited negative GCST values throughout the study period (e.g., −0.263 in 2001 to −0.605 in 2020), indicating a persistent carbon sink deficit. Its high urbanization rate, reduced arable land, and rising food demand have made it increasingly reliant on external grain-derived carbon sink inputs. Liupanshui shifted from a weak output area in the early 2000s to a net sink recipient after 2011. This transition was driven by industrial restructuring, yield instability under mountainous climatic conditions, and adjustments to land use policies. Anshun maintained a modestly positive GCST from 2001 to 2009, benefiting from its ecological agricultural capacity. However, after 2010, the city transitioned into a net input zone as ecosystem service capacity plateaued. Though not a major contributor, Tongren sustained positive GCST values before 2011 but experienced a gradual decline thereafter, reflecting growing ecological and structural constraints.

The province’s GCST structure evolved from a “Bijie–Zunyi output belt” and “Guiyang input pole” in the early 2000s to a more fragmented pattern after 2011. Traditional output regions weakened, urban demand intensified, and a transitional fluctuation zone emerged between Liupanshui and Zunyi. These shifts highlight the interplay between urbanization, farmland pressure, and the ecological vulnerability of mountainous agricultural systems.

4.2. Benchmark Test Analysis

This paper selects five variables to characterize the various forms of FUT and employs a spatial panel model to analyze the impact of GCST.

Before conducting a regression analysis, it is necessary to check for multicollinearity issues among independent variables through correlation analysis. The correlation analysis examined the tolerance and variance inflation factor (VIF) between the FUT and GCST variables from 2001 to 2020. As shown in

Table A1. Multicollinearity test results.

Constant	Tolerances	VIF
CLR	0.185	5.403
GPAP	0.548	1.825
GYield	0.269	3.722
RPCDI	0.247	4.053
PFI	0.322	3.101

Dependent variable: GCST.

of the results, there was no significant multicollinearity, allowing regression analysis to proceed. Conduct the unit root test (ADF) for variables related to FUT and GCST. The test results indicate that the sequence data from 2001 to 2020 exhibit no unit roots and are stationary, suggesting that the data are suitable for panel regression analysis. To select the appropriate panel data regression model, we conducted the Hausman test. The results showed a chi-square statistic of 72.78, with a p-value less than 0.05 (p = 0.0000). Therefore, we rejected the null hypothesis, indicating that the assumption of the random effects model does not hold, and we chose the fixed effects model for further analysis.

The benchmark test results are presented in

Table 3. Regression results of the FUT fixed effects model on GCST.

Variable	Coefficient	Standard Error	t	P	R2	F
Const	−3.679	0.548	−6.710	0.000	Within = 0.5628 between = 0.4192 overall = 0.0006	F = 28.07 p = 0.000
CLR	5.136	0.800	6.420	0.000
GPAP	0.730	0.383	1.910	0.059
GYield	0.459	0.069	6.680	0.000
RPCDI	−0.057	0.012	−4.800	0.000
PFI	1.828	0.310	5.900	0.000

Dependent variable: GCST.

. The fixed effects model (FE) was used to analyze the impact of FUT on GCST in Guizhou Province, with the Hausman test supporting the fixed effects (p = 0.0000). The regression results showed that CLR, GYield, and PFI had significant positive effects on GCST (p < 0.01), while RPCDI had a significant adverse effect (p < 0.01). The effect of GPAP was not significant (p = 0.059). The within-group R² of the model was 56.28%, and the F-statistic (p = 0.000) indicated that the overall model had high explanatory power.

The regression results demonstrate that DFUT variables have significant effects on grain carbon sink transfer (GCST), reflecting both explicit and implicit transitions in farmland use. Compared with Zhou [27], who highlighted that farmland restructuring in agroforestry transition zones primarily improves carbon sequestration through landscape-level reconfiguration and ecological function compensation, this study provides micro-level evidence that specific DFUT indicators—such as CLR and PFI—can independently and significantly enhance GCST. The strong positive effect of CLR (coefficient = 5.136, p < 0.01) underscores the role of land area expansion in supporting surplus grain production and embedded carbon transfer, aligning with the “land scale elasticity” identified by Wu et al. but offering a more quantified relationship in cropland-dominant systems. Moreover, GPAP exerts a marginally positive effect (coefficient = 0.730, p = 0.059), which partially resonates with Wang [50], who argued that structural substitution in crop types often faces time-lag effects due to technological and institutional inertia. This supports the interpretation that, despite potential benefits, structural DFUT pathways require more extended gestation periods to manifest carbon sink outcomes at scale. PFI, representing ecological function transition, shows a substantial positive impact (coefficient = 1.828, p < 0.01). While Wang [50] suggested that excessive inputs may exacerbate agricultural carbon emissions in high-input zones, this study shows that in the karst mountain context, moderate input intensification may still contribute to GCST via enhanced biomass and surplus. This finding highlights a critical regional ecological threshold in DFUT strategy design and echoes Zhang [51], who emphasized that human interventions can amplify net ecosystem productivity when matched with local conditions. Furthermore, the negative coefficient of RPCDI (−0.057, p < 0.01) enriches the socio-economic dimension of carbon sink flow. It reveals that rising rural incomes, though beneficial for livelihoods, may indirectly reduce agricultural labor retention, thereby weakening grain output and carbon sink services.

4.3. The Threshold Effect of FUT on GCST

To verify the nonlinear impact of FUT on GCST, a threshold regression model was constructed using grain yield per unit area (GYield) as the threshold variable. The model identifies a statistically significant single-threshold effect (

Table A2. Threshold effect test results.

Threshold Variables	Threshold	Sum of Squares of the Residuals	Mean Square Error	F-Value	p-Value
GYield	single threshold	3.8464	0.0385	20.64	0.0033

), with a critical value of 2.0922 t/hm² (

Table 4. Threshold regression estimates.

Variable	Coefficient	Standard Error	T-Value	p-Value
CLR	−3.681751	0.9423495	−3.91	0.000
GPAP	−1.323921	0.4107055	−3.22	0.002
RPCDI	−0.5559079	0.0582662	−9.54	0.000
PFI	3.323067	0.3624083	9.17	0.000
Categorisation of threshold variables (_catc.FUT)
Interval 0 (GYield < 2.0922)	16.71146	1.929115	8.66	0.000
Interval 1 (2.0922 ≤ GYield)	15.89249	1.828462	8.69	0.000
constant term (math.)	−2.334795	0.4184752	−5.58	0.000

). When GYield < 2.0922, the influence of FUT on GCST is markedly strong (coefficient = 16.711, p < 0.01), suggesting that in low-yield regions, land use transitions—through cultivation expansion; structural adjustment; or moderate intensification—significantly promote surplus grain output and carbon sink transfer. When GYield ≥ 2.0922, the effect of FUT on GCST remains significantly positive (coefficient = 15.892, p < 0.01) but shows a marginal decline, which implies that in high-yield regions, even though carbon sink services continue to be transferred outward, the additional effect of FUT diminishes due to ecological carrying limits and the tapering of productivity growth.

The results confirm a stage-dependent mechanism: FUT enhances GCST more substantially in early stages of agricultural intensification, but its marginal effect weakens as production approaches ecological saturation. This finding aligns with Liu [52], who proposed a production possibility frontier (PPF)-based land optimization pathway. It underscores the importance of identifying the ecological productivity boundary for carbon sink regulation and further confirms the marginal saturation of carbon sink gains in high-yield areas. Meanwhile, Ling [53] emphasized the “ecological saturation point” of carbon sink enhancement in their study on global soil carbon thresholds, which echoes this research. Zhang [51] demonstrated that human interventions—such as farmland consolidation and precision fertilization—significantly enhanced the net ecosystem productivity (NEP) of China’s croplands. This study’s recognition of the carbon sink potential in low-intensity regions provides additional evidence that FUT-guided interventions can expand the functional boundaries of agricultural carbon sinks, offering a strategic direction for ecological restoration and carbon governance in underperforming mountainous regions.

4.4. Robustness Test

4.4.1. Excluding Special Years Test

Considering the structural impact of the extreme drought climate in Guizhou during 2010–2011 on the agricultural production system, this study excluded the observations from 2010 to 2011 from the sample to ensure the robustness of the model results. It reconstructed the panel dataset (N = 200 → 180). A fixed-effects model was re-estimated, and the coefficient directions of each variable were consistent with the baseline model (

Table A3. Robustness test model results.

Variable	Excluding Special Years Test	Endogeneity Test	Tobit Model
CLR	4.039 ***	5.751 ***	5.041 ***
GPAP	0.784 **	0.757 **	0.752 *
GYield	0.351 ***	0.556 ***	0.476 ***
RPCDI	−0.063 ***	−0.054 ***	−0.0530 ***
PFI	1.725 ***	1.746 ***	1.676 ***
cons	−2.986 ***	−4.201 ***	−3.657 ***
obs	114	108	120
Prob > F	0.000	0.000	0.000

Note. * p < 0.05, ** p < 0.01, *** p < 0.001.

), effectively verifying the robustness of the baseline regression results.

4.4.2. Endogeneity Test

To avoid endogeneity issues, RPCDI was used as an instrumental variable, lagged by two periods, and then reintroduced into the model for regression analysis. As shown in Table A3, the results remain consistent with the previous conclusions, proving the effectiveness of the instrumental variable selection and the absence of weak instrumental variable problems, thus confirming the robustness of this study’s findings.

4.4.3. Replace the Tobit Model

To verify the robustness of the fixed effects results, a panel Tobit model (xttobit) was employed, accounting for the truncated nature of GCST. The key variables—CLR; GYield; and PFI—remained significantly positive; while RPCDI maintained a significant adverse effect. GPAP became statistically significant at the 5% level. The consistency in sign and significance across models, along with firm panel-level heterogeneity (rho = 0.857), confirms the robustness of the main findings.

5. Discussion

5.1. Comparison with Existing Studies

This study contributes to the literature on agricultural land use transition and carbon accounting by introducing grain carbon sink transfer (GCST) as a framework for analyzing the spatial reallocation of ecological services. Unlike traditional research focused on carbon emissions from agri-food trade [54,55], GCST captures the flow of carbon sink benefits across regions, offering a complementary perspective on environmental equity, agroecosystem performance, and virtual carbon dynamics. The intensive use of production factors far beyond the ecological threshold and excessive cropland development accelerated the regional NEP decline [53]. Identifying a nonlinear threshold effect driven by GYield advances the theoretical understanding of FUT. This suggests a stage-dependent mechanism where land use change, productivity, and carbon sink services interact. The threshold yield value of 2.092 t/hm²—below the national average (~5.09 t/hm²)—can identify low-yield zones that benefit from targeted incentives; ecological compensation; or green technology subsidies to maintain both productivity and carbon sink function. In contrast, though still net exporters of carbon sink services, high-yield areas face ecological saturation risks and should focus on soil conservation, input optimization, and carbon monitoring, shifting from uniform food production targets to carbon-smart agriculture.

Drawing on international research, particularly studies from Latin America and Sub-Saharan Africa, can further broaden our perspective. In Latin America, agricultural expansion has driven deforestation, significantly impacting the region’s carbon sink capacity [56]. Large-scale commodity agriculture has led to deforestation in the Amazon region [57], resulting in the loss of carbon sinks and exacerbating greenhouse gas emissions. However, the implementation of the Payment for Ecosystem Services (PES) mechanism [58] has played a crucial role in promoting agricultural sustainability and ecological protection. Climate-Smart Agriculture (CSA) has become an effective response to the agricultural challenges posed by climate change in rural areas [59]. Sub-Saharan Africa faces severe challenges related to land degradation and low agricultural productivity [60,61]. Indigenous communities in the region often adopt climate-smart traditional agricultural practices, such as using organic fertilizers and mixed cropping, which significantly enhance the carbon sink potential of agricultural systems [62].

Understanding FUT will help improve regional grain carbon sink capacity by adjusting farmland utilization. To reduce grain carbon sink loss, our study proposes the following policy recommendations for the FUT: (1) Increase CLR to boost GCST, and prioritize intensity over area in farmland expansion. In stable zones, land consolidation should improve efficiency, while ecologically sensitive areas require a “negative list” and an occupancy-balance mechanism, linking development to carbon sink potential. On the recessive side, enhance GYield and Pesticide-Fertilizer Intensity through precision fertilization, eco-friendly pest control, and low-carbon technologies. Promote smart irrigation and drone protection to foster a cycle of income growth and carbon reduction. (2) Regional Compensation Systems. Zunyi and Bijie, as net carbon-sink exporters, should receive compensation for their contribution to carbon sequestration, while Guiyang, as a carbon sink recipient, must reduce import dependency by developing local food systems. Implementing a “grain carbon footprint” label and creating a provincial monitoring platform using remote sensing and big data will enable real-time tracking of carbon flow and facilitate differentiated governance. This monitoring system can help ensure equitable distribution of carbon sink services and encourage more sustainable agricultural practices. (3) Addressing Low-Yield Regions. In regions with low yields, improve soil health using methods like biochar and legume rotations to boost productivity and carbon sequestration. For high-yield areas, the focus should shift to precision agriculture and carbon-efficient practices. A “technology package + service network” model can aid this transition, supported by green financial products to incentivize sustainable intensification. This approach ensures that high-yielding areas continue to generate carbon sink services without compromising ecological integrity.

5.2. Limitations and Future Directions

Several limitations of this study are apparent. First, the current estimation of GCST relies on fixed carbon sink coefficients and proxy indicators for grain flows, which may not fully capture the dynamic nature and regional heterogeneity of carbon sequestration processes. Future research should incorporate ecological variables, such as remote sensing-derived data, changes in soil organic carbon, and farmland net primary productivity (NPP), to enable more precise quantification of carbon sinks. Second, this study identifies a significant yield threshold that affects the marginal impact of FUT on GCST, yet the underlying mechanisms remain underexplored. Further investigations are needed to examine how institutional settings, market factor allocation, and ecological constraints jointly shape transformation pathways and to uncover region-specific response patterns and their driving forces. Finally, as GCST represents the flow of carbon sink services mediated by grain trade, it currently does not account for the effects of non-grain crops, crop rotation systems, or integrated agricultural models. Future studies should broaden the scope of carbon sink carriers to comprehensively assess the carbon sequestration potential and spatial redistribution patterns under diverse agricultural land-use systems.

6. Conclusions

This study investigates the impact of FUT on GCST in Guizhou Province, with a focus on how different types of FUT influence the spatial flow of carbon sinks. Our findings indicate that increased grain production through both dominant and recessive transitions significantly enhances GCST. Regions with low grain yields can benefit from agricultural expansion, while those with higher yields should focus on improving quality to balance productivity with ecological limits.

The main contributions of this study are as follows: First, it constructs a comprehensive FUT indicator system, combining both dominant and recessive dimensions, and quantifies the spatial redistribution of carbon sink services driven by interregional grain trade. Second, it employs fixed-effects and threshold regression models to uncover the linear and nonlinear effects of FUT on GCST, identifying a critical yield threshold (2.092 t/hm²) that divides the positive and diminishing impacts of FUT on GCST. Third, it provides empirical evidence and region-specific policy recommendations to help coordinate the promotion of food security and carbon sink capacity in mountainous areas.

This research introduces an innovative framework for understanding the relationship between FUT and GCST, contributing valuable insights for both theoretical studies and practical applications in agricultural and ecological policy. The results suggest that enhancing the carbon sink capacity in low-yield regions through farmland use transition could significantly contribute to environmental sustainability, while high-yield areas need to focus on optimizing agricultural inputs and managing ecological limits to ensure continuous carbon sink service flow.