Estimation of Carbon Sequestration Capacity of Cultivated Land Based on Improved CASA-CGC Model—A Case Study of Anhui Province

Abstract

Quantifying carbon sequestration in cultivated land ecosystems is essential for achieving carbon neutrality and ensuring food security, yet current models often fail to capture the complex interactions between crop phenology and environmental factors at regional scales. This paper proposed an improved CASA-CGC model that couples crop phenological parameters with photosynthetic physiological processes, enabling precise carbon sink accounting at the growth cycle scale of cultivated land ecosystems. Results indicate that the carbon sequestration capacity of cultivated land in the province significantly increased from 2010 to 2022, with an average increase of 163.04 g C m⁻², and the spatial pattern showed a centralized evolution characteristic. Model validation showed that the accuracy of the CASA-CGC model is significantly better than traditional methods. Compared with remote sensing inversion products and 93 ground measurement point data, the improved CASA-CGC model increased the R² by 0.155 and reduced the RMSE by 4.19 compared with the tr-CASA model. The innovative introduction of the GeoDetector model reveals that the nonlinear interaction between natural and human factors dominates the carbon sequestration process (accounting for 60%), with the interaction effect between altitude and cropping system configuration being the strongest (q = 0.312), confirming that humans can significantly amplify the potential of natural carbon sinks by optimizing cropping systems.

1. Introduction

Farmland is not only the foundation of food production, but also an important carbon sink system that plays a crucial role in ensuring national food security and regional sustainable development [1,2]. The No. 1 central document [3] issued on 23 February 2025 focuses on “ensuring the double guarantee and double improvement in the quantity and quality of cultivated land”, and makes an important deployment from the aspects of strictly guarding the quantity, improving the quality, and controlling the use, which points out the direction for the protection of cultivated land. As an important component of terrestrial ecosystems, cultivated land ecosystems not only provide resources such as food and fiber for humans, but also play a critical role in the carbon cycle [4]. Under the dual challenges of global climate change and food security, China has put forward major strategic goals of achieving a carbon peak before 2030 and carbon neutrality before 2060. Studying the carbon sequestration capacity of farmland ecosystems during different crop growth cycles not only helps achieve agricultural carbon neutrality goals but also provides a theoretical basis for scientific management and sustainable utilization of farmland resources, which is of great practical significance for ensuring food security and addressing climate change [5].

The study of carbon sequestration in agricultural ecosystems began in the 1980s, with the systematic assessment of global soil organic carbon pools by Post et al. laying the foundation for this field [6]. Subsequently, as the issue of climate change became more prominent, the research focus gradually shifted to the carbon cycling process of terrestrial ecosystems. Early work mainly focused on the dynamics of soil organic carbon and its influencing factors, and the Century model [6] and RothC model [7] developed during this period provided important tools for simulating the carbon cycle in cultivated land; West and Post confirmed that conservation tillage can significantly increase soil carbon storage [8]. In the 21st century, Lal further clarified the important role of arable land in carbon sequestration [9], while Smith et al. began to explore the impact of climate change on the carbon sequestration capacity of farmland [10]. Since the 2010s, the widespread application of remote sensing technology and model simulation has promoted the development of regional and even global-scale carbon stock assessment, and the economic value of agricultural carbon sinks has gradually been valued [11]. It is worth noting that different crops have significant differences in carbon sequestration capacity due to their biological characteristics, growth cycles, and management methods [12]. Therefore, studying the carbon sequestration capacity of farmland ecosystems during different crop growth cycles is of great significance for revealing agricultural carbon cycling mechanisms, optimizing farmland management, and promoting sustainable agricultural development.

With the advancement of remote sensing (RS) technology, accurate monitoring and estimation of carbon sequestration in farmland ecosystems have become increasingly feasible [13,14]. While many studies focus on model applications, such as Agro-C, CANDY, CENTURY, CEVSA, DNDC, and DSSAT, few offer systematic evaluations, especially in quantifying the accuracy and uncertainty of different models under specific cropping systems and management practices [15,16,17]. Moreover, existing assessments often rely on limited spatial data, lacking comprehensive national- or global-scale analysis, and seldom address the interaction between carbon sequestration, farming measures, and climate change. In this context, the CASA model stands out as a widely applied tool for regional carbon sink estimation due to its reliance on RS-driven light use efficiency [18]. Although traditional CASA models have limitations in their response to differences in crop types and growth stages in terms of photosynthetic parameters, introducing growth stage-specific parameters and reconstructing the photosynthetic module can effectively improve their accuracy in simulating carbon absorption and release processes in farmland, providing a reliable tool for multi-scale carbon sink assessment [19].

Although significant progress has been made in the study of carbon sequestration capacity in farmland ecosystems both domestically and internationally [20], there are still shortcomings: existing research mostly focuses on a single crop or growth stage, lacking systematic comparisons of the entire growth cycle of different crops [21]. The model has limitations in simulating the interaction between crops, soil, and climate systems [22]. The data relies heavily on point or regional scales and lacks systematic evaluation on a larger scale; Moreover, there is insufficient exploration of the interaction mechanism between carbon sequestration assessment, farmland management, and climate change [23]. Therefore, this study takes Anhui Province, an important grain production base in China, as the research area, integrates multi-source remote sensing and ground observation data, improves the photosynthesis module of the CASA model, introduces crop growth period parameters, constructs the CASA-GCC model, achieves high-precision simulation of carbon processes throughout the entire growth cycle of different crops, and compares and verifies its performance better than traditional CASA models; Further utilizing geographic detectors, the interactive effects of farmland management measures and climate change on carbon sequestration capacity were analyzed, providing zoning and differentiated optimization measures and scientific basis for improving farmland carbon sequestration capacity and low-carbon agricultural development.

2. Materials and Methods

2.1. Study Area

Anhui Province is located in southeastern China, in the Yangtze River and Huai River basins, with a span of 114°54′~119°37′ E, 29°41′~34°38′ N (Figure 1). It has a superior geographical location and is an important hub connecting East China and Central China. The terrain of the province is diverse, including the Huaibei Plain, Jianghuai Hills, and the mountainous areas of southern Anhui, which are the main agricultural areas. Due to the geographical boundary between the Qinling Mountains and the Huai River spanning the entire province, ecological elements such as climate, biology, and soil exhibit obvious vertical and horizontal transition characteristics [23]. It belongs to the transitional zone between warm temperate and subtropical regions, with a mild climate and moderate rainfall. The annual average temperature is 14–17 °C, the average precipitation is 773–1670 mm, with more in the south and less in the north. The suitable climate conditions in the region provide favorable natural conditions for local agricultural production. Anhui Province is an important grain production base in China, with a total cultivated land area of approximately 55,466 km². The main grain crops include rice, wheat, and corn [24,25]. The Anhui Statistical Yearbook (2011–2023) showed that during this period, the proportion of main grain crops planted in Anhui Province has shown a steady upward trend [26].

2.2. Data Collection and Processing

This study involves a wide range of data, including crop distribution and phenology data, remote sensing inversion data, meteorological data, terrain data, etc. The relevant data sources and resolutions are shown in

Table 1. Main data sources and their resolutions.

Data	Spatial Resolution	Source	URL
Crop distribution data	1 km	The National Ecological Data Center	https://nesdc.org.cn/
Phenological data	1 km
Solar radiation data	1 km	The Geographic Data Sharing Infrastructure, global resources data cloud	www.gis5g.com
Annual maximum Normalized difference vegetation index (NDVI)	30 m	The Resources and Environmental Science and Data Center	http://www.resdc.cn/
Rainfall and temperature data	30 m
Elevation and geomorphic type data	90 m
Crop maturity distribution data	1 km
Potential and actual evapotranspiration data	1 km	The National Qinghai–Tibet Plateau Science Data Center	https://www.tpdc.ac.cn/

.

2.3. Methods

2.3.1. Theoretical Basis for Model Improvement

The improvement in the CASA model is based on the principle that crop carbon sequestration is governed by the dynamic coupling of phenology and photosynthetic physiology. Traditional CASA models, which use constant parameters, fail to capture the stage-specific physiological changes (e.g., in canopy structure and photosynthetic efficiency) that occur as crops develop. Therefore, we propose that incorporating growth-stage-specific parameters is essential to mechanistically simulate carbon uptake dynamics. The CASA-CGC model implements this by dynamically adjusting key parameters according to crop phenological calendars (

Table 2. The ${\mathsf{\epsilon }}_{\mathrm{m}\mathrm{a}\mathrm{x}}$ of Major Crops in Anhui Province (Suggested value).

Types of Crops	${\mathsf{\epsilon }}_{\mathbf{m}\mathbf{a}\mathbf{x}}$(gC·MJ−1)
Wheat	0.7
Corn	1.1
Rice	0.9

), enabling a more accurate and physiologically realistic estimation of regional carbon sequestration.

2.3.2. Calculation of Carbon Sequestration Capacity of Cultivated Land

Net Ecosystem Productivity (NEP) Estimation Model

Guided by the theoretical framework presented in Section 2.3.1, the following methodologies were employed to implement and validate the improved CASA-CGC model. The net primary productivity NPP of cultivated land within different crops was estimated, and the net ecosystem productivity NEP of cultivated land is obtained by subtracting the respiratory consumption of heterotrophic organisms, which represents the carbon sequestration capacity of cultivated land ecosystems [27]. When calculating this index, it is necessary to adhere to the principle of spatiotemporal correspondence between crop type, phenology, and cultivated land distribution. The calculation formula is as follows: Equations (1) and (2):

$$\mathrm{N}\mathrm{E}\mathrm{P}\left(\mathrm{x},\mathrm{t}\right) = \mathrm{N}\mathrm{P}\mathrm{P}\left(\mathrm{x},\mathrm{t}\right) - {\mathrm{R}}_{\mathrm{h}}(\mathrm{x},\mathrm{t})$$

(1)

$$\mathrm{N}\mathrm{P}\mathrm{P}\left(\mathrm{x},\mathrm{t}\right) = \mathrm{A}\mathrm{P}\mathrm{A}\mathrm{R}\left(\mathrm{x},\mathrm{t}\right) \times \mathsf{\epsilon }(\mathrm{x},\mathrm{t})$$

(2)

where $\mathrm{N}\mathrm{E}\mathrm{P}(\mathrm{x},\mathrm{t})$ is the net ecosystem productivity of vegetation for pixel x in month t (gCm⁻² month⁻¹), representing the carbon sequestration capacity index of cultivated land. $\mathrm{N}\mathrm{P}\mathrm{P}(\mathrm{x},\mathrm{t})$ is the net primary productivity of vegetation for pixel x in month t (gCm⁻² month⁻¹). ${\mathrm{R}}_{\mathrm{h}}(\mathrm{x},\mathrm{t})$ is the heterotrophic respiration consumption factor (soil respiration) of pixels in month t (gCm⁻² month⁻¹). $\mathrm{A}\mathrm{P}\mathrm{A}\mathrm{R}(\mathrm{x},\mathrm{t})$ is the photosynthetically active radiation absorbed by pixel x in month t (MJm⁻² month⁻¹). $\mathsf{\epsilon }(\mathrm{x},\mathrm{t})$ is the actual light energy utilization rate of pixel x in month t (gC·MJ⁻¹).

(1)

Photosynthetically active radiation absorbed by vegetation ($\mathrm{A}\mathrm{P}\mathrm{A}\mathrm{R}$)

The total solar radiation and physiological characteristics of vegetation determine the photosynthetically active radiation absorbed by vegetation [28], which should be calculated according to Formula (3):

$$\mathrm{A}\mathrm{P}\mathrm{A}\mathrm{R}\left(\mathrm{x},\mathrm{t}\right) = \mathrm{S}\mathrm{O}\mathrm{L}\left(\mathrm{x},\mathrm{t}\right) \times \mathrm{F}\mathrm{P}\mathrm{A}\mathrm{R}\left(\mathrm{x},\mathrm{t}\right) \times 0.5$$

(3)

where $\mathrm{S}\mathrm{O}\mathrm{L}(\mathrm{x},\mathrm{t})$ is the total solar radiation of pixel x in the t-th month (MJm⁻² month⁻¹), $\mathrm{F}\mathrm{P}\mathrm{A}\mathrm{R}(\mathrm{x},\mathrm{t})$ is the proportion of photosynthetically active radiation absorption by vegetation, the constant 0.5 is the proportion of solar effective radiation (wavelength 0.38–0.71) that vegetation can utilize to the total solar radiation.

It is generally believed that vegetation type and coverage determine the proportion of incident photosynthetically active radiation absorbed by the vegetation layer. The differences in crop season, growth period, and crop maturity in different regions can all affect the size of FPAR values [29]. The NDVI can effectively reflect vegetation types and growth conditions, and there is a linear correlation between NDVI and FPAR. It can be calculated using NDVI and should be calculated according to Formula (4):

$$\mathrm{F}\mathrm{P}\mathrm{A}\mathrm{R}\left(\mathrm{x},\mathrm{t}\right) = {\displaystyle \frac{(\mathrm{N}\mathrm{D}\mathrm{V}\mathrm{I}\left(\mathrm{x},\mathrm{t}\right) - {\mathrm{N}\mathrm{D}\mathrm{V}\mathrm{I}}_{\mathrm{i},\mathrm{m}\mathrm{i}\mathrm{n}})({\mathrm{F}\mathrm{P}\mathrm{A}\mathrm{R}}_{\mathrm{m}\mathrm{a}\mathrm{x}} - {\mathrm{F}\mathrm{P}\mathrm{A}\mathrm{R}}_{\mathrm{m}\mathrm{i}\mathrm{n}})}{({\mathrm{N}\mathrm{D}\mathrm{V}\mathrm{I}}_{\mathrm{i},\mathrm{m}\mathrm{a}\mathrm{x}} - {\mathrm{N}\mathrm{D}\mathrm{V}\mathrm{I}}_{\mathrm{i},\mathrm{m}\mathrm{i}\mathrm{n}})}} + {\mathrm{F}\mathrm{P}\mathrm{A}\mathrm{R}}_{\mathrm{m}\mathrm{i}\mathrm{n}}$$

(4)

where ${\mathrm{N}\mathrm{D}\mathrm{V}\mathrm{I}}_{\mathrm{i},\mathrm{m}\mathrm{i}\mathrm{n}}$ is the minimum NDVI value for the i-th vegetation type, ${\mathrm{N}\mathrm{D}\mathrm{V}\mathrm{I}}_{\mathrm{i},\mathrm{m}\mathrm{a}\mathrm{x}}$ is the maximum NDVI value of the i-th vegetation type, take values of 0.001 and 0.95, respectively.

(2)

Light energy utilization rate ($\mathsf{\epsilon }$)

The utilization rate of solar energy reflects the ability of vegetation to convert sunlight into organic matter [30]. The specific calculation process is to first determine the maximum light energy conversion value for different vegetation types, and then adjust it according to temperature and moisture conditions. It should be calculated according to Formula (5):

$$\mathsf{\epsilon }\left(\mathrm{x},\mathrm{t}\right) = {\mathrm{T}}_{\mathsf{\epsilon }}\left(\mathrm{x},\mathrm{t}\right) \times {\mathrm{W}}_{\mathsf{\epsilon }}\left(\mathrm{x},\mathrm{t}\right) \times {\mathsf{\epsilon }}_{\mathrm{m}\mathrm{a}\mathrm{x}}$$

(5)

where ${\mathrm{T}}_{\mathsf{\epsilon }}(\mathrm{x},\mathrm{t})$ is the temperature stress coefficient, ${\mathrm{W}}_{\mathsf{\epsilon }}(\mathrm{x},\mathrm{t})$ is the water stress coefficient, ${\mathsf{\epsilon }}_{\mathrm{m}\mathrm{a}\mathrm{x}}$ is the maximum light energy utilization efficiency of vegetation without any restrictions.

The temperature stress coefficient should be calculated according to Formula (6):

$${\mathrm{T}}_{\mathsf{\epsilon }}\left(\mathrm{x},\mathrm{t}\right) = {\mathrm{T}}_{\mathsf{\epsilon }1}\left(\mathrm{x},\mathrm{t}\right) \times {\mathrm{T}}_{\mathsf{\epsilon }2}(\mathrm{x},\mathrm{t})$$

(6)

$${\mathrm{T}}_{\mathsf{\epsilon }1}\left(\mathrm{x},\mathrm{t}\right) = 0.8+0.02{ \times \mathrm{T}}_{\mathrm{o}\mathrm{p}\mathrm{t}}\left(\mathrm{x}\right) - 0.0005 \times {\left[{\mathrm{T}}_{\mathrm{o}\mathrm{p}\mathrm{t}}\left(\mathrm{x}\right)\right]}^2$$

(7)

$${\mathrm{T}}_{\mathsf{\epsilon }2}\left(\mathrm{x},\mathrm{t}\right) = {\displaystyle \frac{0.1814}{\left\{1+{\mathrm{e}}^{0.2\times \left[{\mathrm{T}}_{\mathrm{o}\mathrm{p}\mathrm{t}}\left(\mathrm{x}\right)-10-\mathrm{T}(\mathrm{x},\mathrm{t})\right]}\right\}\times \left\{1+{\mathrm{e}}^{0.3\times \left[-{\mathrm{T}}_{\mathrm{o}\mathrm{p}\mathrm{t}}\left(\mathrm{x}\right)-10+\mathrm{T}(\mathrm{x},\mathrm{t})\right]}\right\}}}$$

(8)

where ${\mathrm{T}}_{\mathsf{\epsilon }1}(\mathrm{x},\mathrm{t})$, under low and high temperature conditions, vegetation photosynthesis is limited, leading to changes in the intrinsic physiological properties of plants and a decrease in vegetation NPP, when the average temperature of a month is ≤−10 °C, the value of ${\mathrm{T}}_{\mathsf{\epsilon }1}(\mathrm{x},\mathrm{t})$ is 0. ${\mathrm{T}}_{\mathsf{\epsilon }2}(\mathrm{x},\mathrm{t})$, when the temperature changes from the optimal temperature to high or low temperature, vegetation photosynthesis is limited, resulting in a gradual decrease in light energy utilization efficiency. ${\mathrm{T}}_{\mathrm{o}\mathrm{p}\mathrm{t}}\left(\mathrm{x}\right)$, the temperature that is most suitable for vegetation growth represents the temperature corresponding to the best vegetation growth condition, and take the temperature at the corresponding stage with the highest NDVI value for different crops as the optimal temperature.

The water stress coefficient refers to the impact of vegetation’s water conditions on its physiological characteristics, resulting in a decrease in light energy utilization efficiency. It is directly proportional to the content of effective water in the environment [30]. The range of values for water stress factors is 0.5–1 (from extreme drought to very humid). It should be calculated according to Formula (9):

$${\mathrm{W}}_{\mathsf{\epsilon }}\left(\mathrm{x},\mathrm{t}\right) = 0.5 + {\displaystyle \frac{0.5\times \mathrm{E}(\mathrm{x},\mathrm{t})}{{\mathrm{E}}_{\mathrm{p}}(\mathrm{x},\mathrm{t})}}$$

(9)

where $\mathrm{E}(\mathrm{x},\mathrm{t})$ is the actual evapotranspiration of the region (mm), ${\mathrm{E}}_{\mathrm{p}}(\mathrm{x},\mathrm{t})$ is the potential evapotranspiration of the region (mm). When $\mathrm{E}(\mathrm{x},\mathrm{t})$ ≤ ${\mathrm{E}}_{\mathrm{p}}(\mathrm{x},\mathrm{t})$, the actual evapotranspiration value is taken as $\mathrm{E}(\mathrm{x},\mathrm{t})$, when $\mathrm{E}(\mathrm{x},\mathrm{t})>(\mathrm{x},\mathrm{t})$, the actual evapotranspiration value is taken as ${\mathrm{E}}_{\mathrm{p}}(\mathrm{x},\mathrm{t})$.

The utilization ratio of photosynthetically active radiation by vegetation, without any restrictive factors, is the ${\mathsf{\epsilon }}_{\mathrm{m}\mathrm{a}\mathrm{x}}$ by vegetation [31]. The maximum solar energy utilization efficiency of vegetation varies depending on the physiological and ecological characteristics of different vegetation types, and is often obtained through simulation methods. Zhu et al. simulated the ${\mathsf{\epsilon }}_{\mathrm{m}\mathrm{a}\mathrm{x}}$ of different vegetation types [32], but did not subdivide the different crop types in cultivated land, and the value may vary in different regions. Furthermore, by consulting literature [33,34,35], the ${\mathsf{\epsilon }}_{\mathrm{m}\mathrm{a}\mathrm{x}}$ of major crops in Anhui Province was obtained (Table 2). Consistent with our theoretical basis, the most significant modification to the traditional CASA model was the introduction of growth-stage-specific parameters. Specifically, the maximum light use efficiency (${\mathsf{\epsilon }}_{\mathrm{m}\mathrm{a}\mathrm{x}}$) and the optimal temperature were defined not merely by crop type, but by the distinct physiological status of each key growth stage (

Table 3. Growth Cycle of Major Crops in Anhui Province.

Crops	Different Crop Growth Stages
Wheat	Sowing-overwintering	Overwintering	Greening stage	Jointing stage	Heading and flowering	Maturation stage
	Late October to late December	Early January to early February	Mid February to late February	Early March to late March	Early April to early May	Mid May to early June
Corn	Seedling emergence stage	Jointing stage	Boot stage	Tasseling stage	Milk ripening stage	Maturation stage
	Mid June to early July	Mid July	Late July to early August	Mid August	Late August to mid September	Late September
Rice	Greening stage	Early tillering stage	Late tillering stage	Boot stage	Heading stage	Milk ripening stage	Yellow ripening stage
	Early June	Mid to late June	Early July	Mid July	late July to early August	Mid to late August	Early to mid September

), thereby mechanistically capturing the temporal variability in carbon sequestration capacity.

R_h Estimation Model

The respiratory consumption of ${\mathrm{R}}_{\mathrm{h}}$ by heterotrophic organisms is mainly due to the respiration consumption of microorganisms in the soil, and temperature and precipitation are the two important factors affecting soil microbial respiration [36]. It should be calculated according to Formula (10):

$${\mathrm{R}}_{\mathrm{h}}\left(\mathrm{x},\mathrm{t}\right) = 0.22 \times \left[\exp \left(0.0912 \times T\left(x,t\right)\right) + \mathrm{l}\mathrm{n}\left(0.3145 \times \mathrm{P}\left(\mathrm{x},\mathrm{t}\right) + 1\right)\right] \times 30 \times 46.5\%$$

(10)

where $\mathrm{T}(\mathrm{x},\mathrm{t})$ is the average temperature (°C) of pixel x in month t, $\mathrm{P}(\mathrm{x},\mathrm{t})$ is the precipitation amount (mm) of pixel x in month t.

Determine the Growth Period of Different Crops

Based on the annual phenological dataset of the three major grain crops in China downloaded from the National Ecological Science Data Center, summarize the main growth cycles of different grain crops (Table 3).

Accuracy Comparison of Estimation Models

To verify the estimation accuracy of Tr-CASA and CASA-CGC models in farmland ecosystems, this study used 93 matched sample data for statistical analysis. These sample data are sourced from the study of [20,22,36,37,38,39], which includes ground observations of the region and further obtains remote sensing inversion products (MOD17A2H, spatial resolution 500 m, temporal resolution 8 d). Based on the R programming language (https://www.r-project.org/ accessed on 5 October 2025), estimated values of two models at corresponding sample positions were extracted, and the performance of the models was quantitatively evaluated by calculating the coefficient of determination (R²) and root mean square error (RMSE). Among them, R² is used to measure the strength of the linear relationship between model simulation values and observed values, while RMSE is used to evaluate the absolute error level of estimated values.

2.3.3. One-Way ANOVA Method

Box plots were created using Origin software (https://www.originlab.com/ accessed on 8 October 2025) to illustrate the impact of individual factors on crop carbon sequestration capacity [31]. Additionally, one-way ANOVA and multiple comparisons (p < 0.05) were employed to test the significance of differences in the effects of various factors on soil-crop carbon sequestration capacity.

2.3.4. Random Forest Model

The Random Forest model, an ensemble algorithm based on classification and regression trees [40], was employed to evaluate the relative importance of factors affecting soil carbon sequestration capacity [38]. This method improves prediction accuracy while being robust to overfitting, multicollinearity, and imbalanced data. We used the %IncMSE index to quantify variable importance, with higher values indicating stronger influence.

2.3.5. Geographical Detector

Quantitative analysis of the single and multi factor interactions of factors such as vegetation coverage, topography, climate change, and farmland management measures on carbon sequestration capacity based on a geographic detector model (

Table 4. The main influencing factors of crop farmland carbon sequestration capacity.

Influence Aspect	Evaluation Factors
Natural environmental factors	Elevation
	Landform type
	Temperature
	Rainfall
	Vegetation coverage
Human social factors	Crop type
	Crop maturity

). The geographic detector model is a set of statistical methods for detecting spatial heterogeneity and revealing its underlying driving forces [41,42]. The core idea is based on the assumption that if an independent variable has a significant impact on a dependent variable, then the spatial distribution of that independent variable and the dependent variable should be similar (Figure 2). One unique advantage of a geographic detector is that it detects the interaction between two factors and the dependent variable. The general method for identifying the interaction is to add the product term of the two factors to the regression model and test its statistical significance [43].

By comparing the q-values of different factors A ∩ B and A + B, the types of interactions can be classified into five categories [44]: nonlinear weakening, single factor nonlinear weakening, double factor enhancement, independent and nonlinear enhancement (

Table 5. Interaction types and judgment criteria in the geographic detector model.

Illustration	Judgment Criteria		Interaction Type
	q(A ∩ B) < Min(q(A),q(B))		nonlinear weakening
	Min(q(A),q(B)) < q(A ∩ B) < Max(q(A),q(B))		single factor nonlinear weakening
	q(A ∩ B) > Max(q(A),q(B))		double factor enhancement
	q(A ∩ B) = q(A) + q(B)		independent
	q(A ∩ B) > q(A) + q(B)		nonlinear enhancement
Min(q(A),q(B))               Max(q(A),q(B))		q(A) + q(B)                   q(A ∩ B)

).

3. Results

3.1. Calculation Results of Carbon Sequestration Capacity of Cultivated Land During Different Crop Growth Cycles

3.1.1. Carbon Sequestration Capacity of Different Crops from 2010 to 2022

The three major grain crops exhibited distinct spatial distributions across Anhui Province, which formed the basis for our subsequent carbon sequestration analysis (Figure 3). Wheat (40,879 km²) and corn (16,899 km²) were predominantly cultivated in the northern regions, while rice (34,104 km²) was primarily scattered throughout the central and southern parts of the province (Figure 3).

A clear increasing trend in carbon sequestration capacity was observed for all three crops from 2010 to 2022 (Figure 3). The average values for wheat, corn, and rice increased from 16.25, 121.81, and 126.32 g C m⁻² in 2010 to 25.68, 148.21, and 216.28 g C m⁻² in 2022, respectively. Spatially, the centers of high carbon sequestration exhibited a notable shift towards the central part of the province over this period. For instance, wheat’s high-value zone broadened around counties like Linquan and Funan, while corn and rice hotspots showed a northward and centralizing trend, respectively. This spatial reorganization resulted in a pronounced convergence of carbon sequestration hotspots in central Anhui.

3.1.2. The Variation in Carbon Sequestration Capacity of Different Crops from 2010 to 2022

Analysis of the changes in carbon sequestration between 2010 and 2022 demonstrated a widespread increasing trend, though with significant spatial heterogeneity (Figure 4a–c). The increase ranged from −198.17 to 115.07 g C m⁻² for wheat, −276.64 to 236.40 g C m⁻² for corn, and −283.38 to 606.54 g C m⁻² for rice. The majority of planting areas (68.75% for wheat, 89.97% for corn, and 78.36% for rice) showing enhanced carbon sinks. This overall upward trend occurred despite localized decreases in specific counties, reflecting a complex but positive spatial redistribution of carbon sequestration over the past decade.

3.1.3. Comparison of Carbon Sequestration Results Between Tr-CASA Model and CASA-CGC Model for Different Crops

The improved CASA-CGC model demonstrated superior performance over the traditional Tr-CASA model in simulating carbon sinks for wheat, corn, and rice in 2022 (Figure 3d–f vs. Figure 5a–c). The Tr-CASA model produced unrealistic value ranges with extensive negative carbon sink estimates across all crops, including wheat (−173.30 to 176.70 g C m⁻²), corn (–167.89 to 454.56 g C m⁻²), and rice (–183.53 to 1098.54 g C m⁻²). In contrast, the CASA-CGC model generated more constrained and physically plausible results, significantly reducing the prevalence of negative values and narrowing the value ranges for wheat (–158.18 to 134.13 g C m⁻²) and corn (–132.05 to 232.06 g C m⁻²). Moreover, the spatial distribution of high-value areas simulated by the CASA-CGC model, particularly for corn and rice, showed better agreement with actual crop planting patterns. Further spatial consistency testing was conducted, and the results showed that the CASA-CGC model achieved significantly improved spatial autocorrelation (Moran’s I: 0.58–0.66) and significantly reduced estimated dispersion (CV: 35.4% −42.3%) in all crops, quantitatively proving that its simulation results have better stability and rationality in spatial distribution compared to the Tr CASA model (Moran’s I: 0.49–0.57; CV: 43.8–51.5%). Overall, the CASA-CGC model demonstrated higher stability and rationality in simulating carbon sinks for the three crops, particularly in reducing abnormally low values (negative carbon sinks). This indicates that by introducing more refined crop growth dynamics and environmental response mechanisms, the CASA-CGC model can more accurately reflect the spatial variability characteristics of crop carbon sinks at the regional scale, providing a more reliable model tool for agricultural carbon sink assessment (

Table 6. Comparison of Spatial Statistical Indicators for Carbon Sink Simulation between Tr-CASA and CASA-CGC Models.

Indicator	Crop Type	Tr-CASA Model	CASA-CGC Model	Improvement
Moran’s I	Wheat	0.49	0.58	+18.4%
	Corn	0.52	0.61	+17.3%
	Rice	0.57	0.66	+15.8%
Coefficient of Variation (CV)	Wheat	51.5%	42.3%	−17.9%
	Corn	48.2%	38.7%	−19.7%
	Rice	43.8%	35.4%	−19.2%

).

Based on multi-source validation data (remote sensing products and 93 ground measurements), we comprehensively evaluated the performance of the traditional Tr-CASA model and the improved CASA-CGC model. Quantitative results showed that against remote sensing products, the CASA-CGC model achieved an R² of 0.79 (a 27.4% increase over the Tr-CASA model’s 0.62) and reduced the RMSE to 14.29 (a 23.3% decrease from 18.63). When validated with ground measurements, it also exhibited strong performance, with an R² of 0.72 (24.1% higher than 0.58) and an RMSE of 12.67 (24.2% lower than 16.71) (

Table 7. Accuracy verification of carbon sequestration estimation models for different cultivated lands.

Carbon Sequestration Estimation Models	Tr-CASA Model		CASA-CGC Model
	R2	RMSE	R2	RMSE
Remote sensing inversion products	0.62	18.63	0.79	14.29
Ground-measured	0.58	16.71	0.72	12.67

).

This marked improvement is likely due to the incorporation of crop growth dynamics and optimized environmental response mechanisms in the CASA-CGC model, enabling it to better capture the spatiotemporal variability of carbon sinks across different crops. The consistent performance gain across both validation methods confirms the robustness and general applicability of the model improvements. In summary, validation results confirm that the CASA-CGC model significantly improves the accuracy of regional farmland carbon sink estimation through mechanistic optimizations. The model demonstrates strong performance suitable for regional-scale assessments, with accuracy metrics representing a notable advancement over existing studies while maintaining computational efficiency. Nevertheless, certain areas of persistent underestimation suggest that further refinement of soil respiration modules and incorporation of management practice data could yield additional improvements.

3.2. Analysis of the Impact of Single Factors on Crop Carbon Sequestration Capacity

3.2.1. Analysis of the Impact of Natural Environmental Factors on Crop Carbon Sequestration Capacity

The study area exhibits an elevational gradient ranging from −181 to 1806 m, generally decreasing from south to north. As shown in Figure 6, carbon sequestration across crops demonstrated a nonlinear response to elevation, characterized by an initial decrease followed by an increase, with interspecific differences becoming most pronounced at mid-elevation gradients (p < 0.05). Mean annual temperature (10.01–18.78 °C) showed a weak negative correlation with carbon sequestration, though crop-type differences were more significant under higher temperature conditions (p < 0.05). Similarly, annual precipitation (200–778 mm) did not exert a strong overall influence, yet interspecific variability in carbon sequestration increased markedly under higher rainfall regimes (p < 0.05). In contrast, vegetation coverage was positively correlated with carbon sequestration across all crops, reflecting the central role of canopy photosynthesis in carbon accumulation. Geomorphologically, the area is dominated by low-altitude plains (45.12%), mid-altitude plains (22.31%), gently undulating low mountains (12.52%), and high-altitude plains (7.70%). Significant differences in carbon sequestration were observed among landform types (p < 0.05), with the highest values occurring in low-altitude plains, indicating superior suitability of flat terrain for crop growth and carbon storage [2].

3.2.2. Analysis of the Impact of Human Social Factors on Crop Carbon Sequestration Capacity

Based on the Anhui Provincial Statistical Yearbook, wheat in Anhui is predominantly cultivated as a single crop, whereas rice and corn are managed under double-cropping systems in certain regions. Spatially, single-cropping systems are concentrated in the northern part of the study area, while double-cropping is more prevalent in the south. Statistical analysis focusing on corn and rice (Figure 7) revealed significant differences in carbon sequestration between these crops under different cropping systems (p < 0.05). This variation can likely be attributed to the extended growing period in double-cropping systems, which enhances carbon sequestration potential in cultivated land [25].

3.3. Interaction Effects of Influencing Factors on Crop Carbon Sequestration Capacity

3.3.1. Ranking of Impact Factor Importance

The relative importance of factors influencing carbon sequestration capacity is shown in Figure 8. Elevation, cropping system, and vegetation cover emerged as the three most influential factors, indicating their dominant roles in controlling carbon sequestration dynamics. Crop type and rainfall exhibited moderate impacts, while landform type and temperature showed relatively lower influence. Among natural factors, elevation and vegetation cover demonstrated particularly strong effects, whereas the cropping system represented the most significant factor among human activity-related variables.

3.3.2. The Interaction Effect Between Influencing Factors

The interactions between natural environmental and human social factors primarily influence crop carbon sequestration capacity through two patterns: two-factor enhancement and nonlinear enhancement, with the latter being dominant (60% vs. 40%). Two-factor enhancement mainly occurs between elevation, temperature, vegetation coverage, crop type, and cropping system, whereas interactions involving landform type and rainfall with human factors predominantly exhibit nonlinear enhancement.

As quantified by the interaction detector model (

Table 8. Index of the effect of interaction between natural environmental and human social factors on crop carbon sequestration capacity.

Evaluation Factors	Elevation	Landform Type	Temperature	Rainfall	Vegetation Coverage
Crop type	0.289 ↑	0.254 ↑↑	0.291 ↑↑	0.297 ↑↑	0.293 ↑
Crop maturity	0.312 ↑	0.240 ↑↑	0.303 ↑	0.285 ↑↑	0.281 ↑↑

Note: ↑ denotes dual-factor enhancement, while ↑↑ denotes nonlinear enhancement.

), the elevation ∩ cropping system interaction exerted the strongest effect (q = 0.312) on carbon sequestration capacity. In contrast, landform type ∩ crop type (q = 0.254) and landform type ∩ cropping system (q = 0.240) showed relatively weaker influences. Overall, interactions involving elevation, climatic factors, and vegetation coverage with crop type and cropping system consistently demonstrated significant effects on carbon sequestration capacity.

4. Discussion

4.1. Accuracy of the Carbon Sequestration Capacity Accounting Model for Cultivated Land

This study cross-validated the simulation results of the CASA model through multi-source data and independent methods to ensure the reliability of the accounting of farmland carbon sequestration capacity [45,46]. The validation results demonstrate that our improved CASA model effectively addresses several limitations in existing carbon sequestration assessment methods. While previous studies have reported challenges in accurately simulating crop-specific carbon dynamics [19,21], our model achieves significantly higher accuracy (R² > 0.72) through the incorporation of growth-stage-specific parameters. This represents a notable advancement over traditional approaches that often apply uniform parameters across diverse agricultural landscapes [23]. Compared to conventional CASA implementations, our localized calibration of key parameters, particularly the maximum light use efficiency (ε_max), has proven crucial for reliable carbon sink estimation [47]. This finding aligns with recent research emphasizing the importance of parameter localization in ecological modeling [45], yet our study provides novel insights into its specific impact on agricultural carbon sequestration assessment. However, several limitations warrant consideration. The model’s performance under extreme climate conditions remains uncertain, and its dependence on remote sensing data introduces potential uncertainties during cloud-prone periods. Furthermore, while our results show good agreement with soil organic carbon datasets [48], the representation of certain management practices requires further refinement.

4.2. Factors Affecting the Carbon Sequestration Capacity of Cultivated Land

4.2.1. The Impact of Various Regional Influencing Factors on the Carbon Sequestration Capacity of Cultivated Land

This study systematically reveals the dominant role of altitude, planting system, and vegetation cover in the carbon sequestration capacity of cultivated land, and its impact mechanism presents new complexity compared to existing research. Unlike the previous conclusion that carbon sequestration capacity decreases monotonically with altitude [13,49], the nonlinear law of “first decreasing and then increasing” discovered in this study reveals the replacement of dominant limiting factors under different altitude gradients: in the transition zone from low altitude plains to hills, soil erosion and strong human interference lead to carbon loss [50,51]; In the mid to high altitude areas, although the heat conditions are limited, the weakening of human interference and the combination of cold and humid environments promote the accumulation of organic carbon [32,52,53]. This discovery deepens our understanding of the carbon cycling mechanism in farmland in the mountainous plain transition zone.

In terms of planting system, the carbon sequestration capacity of double season crops is increased by 35–40% compared to single season crops, which is significantly higher than the 15–25% reported in earlier studies [54,55]. This difference may be attributed to the optimized crop combination and refined management measures adopted in this study area, which achieved more sustainable carbon input by extending photosynthesis time and increasing biomass returning to the field [56]. Consistent with the findings in reference [55], double cropping enhances carbon sequestration capacity through a dual pathway of increasing the vegetation carbon pool and strengthening the soil carbon pool. However, this study further quantified the contribution rate of this synergistic effect.

Of particular note is that the correlation between vegetation coverage and carbon sequestration capacity observed in this study (R² = 0.91) indicates that in complex agricultural landscapes, vegetation coverage not only promotes carbon fixation by enhancing photosynthesis, but also effectively inhibits carbon loss through mechanisms such as canopy interception and reduced runoff [57,58]. Complementing the conclusion of reference [30], this study confirms that the role of vegetation cover in preventing and controlling soil erosion and its contribution to maintaining soil carbon pools have been underestimated in the past.

These findings collectively indicate that enhancing the carbon sequestration capacity of arable land requires a coordinated consideration of the synergistic effects of terrain constraints, planting systems, and vegetation management. Compared with the universal recommendations proposed in reference [59,60], this study emphasizes differentiated strategies based on regional characteristics: soil and water conservation projects should be implemented in low mountain and hilly areas, reasonable multi-cropping should be promoted in areas with sufficient light and heat, and high vegetation coverage should be maintained through optimized field management. This system optimization path provides new ideas for the precise enhancement of regional farmland carbon sinks (Figure 9 and Figure 10).

4.2.2. The Impact of Interactions Among Regional Influencing Factors on the Carbon Sequestration Capacity of Cultivated Land

This study reveals the key mechanism affecting the carbon sequestration capacity of farmland. the interaction of multiple factors rather than a single factor dominates the carbon sequestration process, with nonlinear enhancement accounting for 60%, significantly higher than the 40–50% level typically reported in previous studies [61,62]. This discovery confirms that the synergistic effect between natural and human factors plays an important role in the formation of carbon sinks beyond simple superposition.

Specifically, the dual factor enhancement effect is mainly reflected between environmental factors, such as altitude and temperature, and human management measures such as crop type and planting system, which is consistent with the conclusion of the literature [21]. However, this study further quantified the strength of this coupling effect. Of particular note is the nonlinear enhancement effect (accounting for 40%) between relatively stable natural backgrounds, such as terrain and rainfall, and human activities, which reveals the background dependence of human activity effects. For example, consistent with the findings in reference [63], rice cultivation in plain areas forms strong carbon sinks by creating anaerobic environments; In hilly areas, similar farming activities lead to carbon loss due to intensified soil erosion [1]. This phenomenon of “same cause, different effects” highlights the regulatory role of natural substrates in the effects of human activities [40].

In terms of interaction intensity, the interaction between altitude and planting system is the strongest (q = 0.312), which goes beyond the traditional understanding that climate factors dominate [39]. Research has shown that the key to determining carbon sequestration capacity is not altitude itself, but the planting system selected based on thermal conditions [64]. Low to medium altitude areas have become key areas for promoting multi-cropping and achieving high carbon input due to their suitable hydrothermal conditions. However, the principles of high altitude and low altitude are limited by low temperature and high-intensity human interference, respectively. This discovery provides a new perspective for optimizing regional planting layout [63].

In contrast, the interaction between terrain and crop types is relatively weak (q = 0.254), which may be due to the established pattern of suitable land and planting in long-term practice [65]. This relatively stable configuration suggests that we can further enhance carbon sequestration potential in specific regions by fine-tuning existing models [47].

Based on these findings, we suggest that future farmland management strategies should shift from single-factor regulation to multi-factor collaborative optimization. In hilly and mountainous areas, priority should be given to implementing soil and water conservation technologies to stabilize the negative effects of the nonlinear relationship between terrain and human activities [66]. In plain areas, it is necessary to strengthen the positive nonlinear enhancement effect of “rainfall management” by improving irrigation facilities and maintaining water drought rotation [67]. This differentiated system optimization path provides a new theoretical basis and practical direction for achieving precise enhancement of carbon sequestration in farmland.

4.3. Implications and Limitations

This study provides effective methodology and scientific insights for regional farmland carbon sink assessment by improving the CASA model and combining it with geographic detectors [36,37]. Of course, there are still some areas that can be further deepened in this study. For example, the model’s representation of certain agricultural management measures (such as precise irrigation and fertilization) is not yet direct enough. In the future, efforts can be made to integrate higher precision agricultural management data to improve simulation accuracy. However, due to limitations in data acquisition, direct observation and simulation of the long-term dynamics of soil carbon pools still need to be strengthened. In addition, the framework of this model can be considered to be combined with more diverse agricultural ecological zone data in the future to test and enhance its universality. These directions will be the focus of future research that deserves attention.

5. Conclusions

Based on the improved CASA-CGC model, this study systematically evaluated the carbon sequestration capacity of cultivated land during the growth cycle of major grain crops in Anhui Province from 2010 to 2022, revealing the mechanisms of natural and human factors. Research has shown that the carbon sequestration capacity of cultivated land exhibits significant spatiotemporal differentiation characteristics; the carbon sequestration capacity of the three major grain crops all show an increasing trend, and spatially exhibits an evolutionary pattern of clustering towards the central region. The evolution of this spatial pattern reflects the effective matching of agricultural management measures with regional natural conditions, especially in the central region, which exhibits higher carbon sequestration efficiency due to its transitional climate characteristics and optimized planting systems. Among them, rice has the highest average carbon sequestration capacity and increase, followed by corn and wheat. Through geographic detector analysis, it was found that altitude, planting system, and vegetation cover are the dominant factors affecting the carbon sequestration capacity of regional arable land, and the interaction between factors is dominated by nonlinear enhancement (accounting for 60%). This discovery breaks through the cognitive limitations of traditional single-factor management and reveals the key role of multi-factor synergy in carbon sink formation. Especially, the interaction between altitude and planting system is the strongest (q = 0.312), proving that humans can significantly enhance the natural carbon sink potential by optimizing planting systems. The improved CASA-CGC model shows high accuracy in simulating regional NPP and carbon sequestration capacity (R² > 0.72, RMSE < 14.3), providing a reliable tool for regional-scale assessment of cultivated land carbon sequestration. However, further improvement is needed in terms of the applicability of the model under extreme climate conditions and the quantitative characterization of management measures in this study. Based on the above findings, we suggest that future agricultural management policies should fully consider the coupling relationship between “natural social” factors, adopt differentiated and systematic management strategies, avoid a “one size fits all” approach, and achieve synergistic improvement in farmland productivity and carbon sequestration capacity.