Comprehensive Analysis of Agricultural CEE: Efficiency Assessment, Mechanism Identification, and Policy Response — A Case Study of Anhui Province
农业 CEE 的综合分析：效率评估、机制识别与政策响应——以安徽省为例

Chenxin Ru¹, Sihan Wang¹, Zhiya Liang¹, Bingyue Liao¹, Lei Luo¹, Shaowei Ning^1,* , Bhesh Raj Thapa²
陈心如 ¹ ，王思涵 ¹ ，梁志亚 ¹ ，廖冰月 ¹ ，罗磊 ¹ ，宁绍伟 ^{1 ,} * ，巴什·拉杰·塔帕 ²

¹ College of Civil Engineering, Hefei University of Technology, Hefei 230009, China
¹ 合肥工业大学土木工程学院，中国安徽省合肥市 230009

² Universal Engineering and Science College, Pokhara University, Lalitpur 44700, Nepal
² 普尔大学普尔分校通用工程与科学学院，尼泊尔普尔市拉利特普尔区 44700

Abstract: We applied a multidimensional approach to analyze spatiotemporal patterns and driving mechanisms of agricultural carbon emission efficiency (CEE) in Anhui Province, providing evidence for region-specific carbon reduction policies. Using city-level panel data from 2010 to 2022, we analyzed spatial agglomeration patterns of agricultural CEE through global and local Moran's I indices and Local Indicators of Spatial Association (LISA) scatter plots. Agricultural CEE in Anhui Province showed consistent annual growth throughout the study period. Spatial heterogeneity analysis revealed a distinct geographical pattern—high in the south and low in the north, high in the west and low in the east—characterized by positive spatial autocorrelation. A random forest model weighted by SHapley Additive exPlanations (SHAP) values identified key driving factors, with agricultural plastic film use intensity emerging as the primary inhibiting factor of agricultural CEE. Spatial Durbin model (SDM) results showed that agricultural inputs—fertilizers, pesticides, plastic films, and irrigation—generate significant "spillover effects," impacting both local and neighboring regions with nonlinear spatial characteristics. Coupling coordination degree analysis indicated improving alignment between agricultural output value and CEE, though significant optimization potential remains. We established an efficiency assessment-mechanism identification-policy response framework, offering a geographically weighted empirical approach for advancing agricultural low-carbon transformation.
摘要：我们采用多维度方法分析了安徽省农业碳排放效率（CEE）的时空分布特征及驱动机制，为制定区域特色碳减排政策提供了依据。基于 2010 年至 2022 年的市级面板数据，我们通过全球和局部莫兰指数（Moran's I）以及空间关联局部指标（LISA）散点图，分析了农业 CEE 的空间聚合模式。安徽省农业 CEE 在研究期间呈现持续增长趋势。空间异质性分析揭示了明显的地理格局——南部高、北部低，西部高、东部低，且存在正空间自相关性。基于 SHapley Additive exPlanations（SHAP）值加权的随机森林模型识别出关键驱动因素，其中农业塑料薄膜使用强度被确认为农业 CEE 的主要抑制因素。空间杜宾模型（SDM）结果表明，农业投入品（化肥、农药、塑料薄膜和灌溉）产生显著的“溢出效应”，以非线性空间特征影响本地及邻近区域。耦合协调度分析表明，农业产出价值与 CEE 之间的协调性有所提升，但仍存在显著优化潜力。我们建立了效率评估-机制识别-政策响应框架，为推进农业低碳转型提供了基于地理加权的实证方法。

Graphic abstract:
图形摘要：

Keywords：Agricultural CEE; Anhui Province; Driving Factors; Spatial Analysis; Coupling Coordination Degree
关键词：农业 CEE；安徽省；驱动因素；空间分析；耦合协调程度

1. Introduction
1 引言

Since the 1980s, the increasing frequency of climate anomaly-induced natural disasters has elevated climate change to a central issue in international politics. Carbon emissions encompass the release of carbon dioxide (CO₂) and other greenhouse gases (methane, nitrous oxide, etc.) from human activities into the atmosphere, exacerbating global warming and resulting in climate anomalies and ecological degradation [1]. Research confirms carbon emissions as the primary driver of global warming, with fossil fuel combustion and land-use changes significantly elevating atmospheric carbon dioxide concentrations [2]. Global greenhouse gas emissions reached approximately 59.1 Gt CO2-eq in 2020 [3]. Agricultural activities contribute approximately 24% of total global carbon emissions, with the Asia-Pacific region generating over 40% of these agricultural emissions [4]. Agricultural food systems generate approximately one-third of global anthropogenic greenhouse gas (GHG) emissions, establishing agriculture as a critical carbon source [23]. Consequently, investigating the patterns and driving mechanisms of agricultural carbon emissions has become essential for optimizing agricultural practices and addressing global climate challenges.
自 20 世纪 80 年代以来，气候异常引发的自然灾害频发，使气候变化成为国际政治的核心议题。碳排放指人类活动向大气中释放二氧化碳（CO₂）及其他温室气体（如甲烷、一氧化二氮等），加剧全球变暖并导致气候异常和生态退化[1]。研究证实，碳排放是全球变暖的主要驱动因素，化石燃料燃烧和土地利用变化显著提升了大气中二氧化碳浓度[2]。2020 年全球温室气体排放量约为 59.1 亿吨二氧化碳当量[3]。农业活动约占全球碳排放量的 24%，其中亚太地区贡献了超过 40%的农业排放量[4]。农业食品系统约占全球人为温室气体（GHG）排放量的三分之一，使农业成为关键的碳源[23]。因此，研究农业碳排放的模式和驱动机制对于优化农业实践和应对全球气候挑战至关重要。

China leads global carbon dioxide emissions, contributing approximately 33% of the worldwide total [5], while simultaneously ranking among the world's largest agricultural producers, with the agricultural sector accounting for roughly 10% of China's GDP. The agricultural sector constitutes a significant source of carbon emissions in China, with the grain system alone—encompassing land-use changes, production processes, packaging, and waste management—generating 18 billion tons of CO2 equivalent in 2015, representing 34% of national emissions [6]. Anhui Province, situated in southeastern China along the middle and lower reaches of the Yangtze River, represents one of China's critical agricultural production regions. In 2021, Anhui's grain cultivation area reached 7.31×10⁶ hectares with total production of 4.09×10¹⁰ kg, both metrics ranking fourth nationally [24]. As one of only six provinces capable of consistently supplying commercial grain to other regions, Anhui plays a crucial role in ensuring national food security [25]. Anhui Province features diverse topography—including plains, hills, mountains, lakes, and rivers—creating an ideal representative case study for agricultural emissions research.
中国是全球二氧化碳排放量最大的国家，占全球总排放量的约 33%[5]，同时也是全球最大的农业生产国之一，农业部门占中国国内生产总值（GDP）的约 10%。农业部门是中国碳排放的重要来源，仅粮食生产系统——包括土地利用变化、生产过程、包装和废物管理——在 2015 年就产生了 18 亿吨二氧化碳当量，占全国排放量的 34%[6]。安徽省位于中国东南部的长江中下游地区，是中国重要的农业生产基地之一。2021 年，安徽省粮食播种面积达 7.31×10 ⁶ 公顷，总产量为 4.09×10 ¹⁰ 千克，两项指标均位列全国第四[24]。作为全国仅有的六个能够持续向其他地区供应商业粮食的省份之一，安徽省在保障国家粮食安全方面发挥着至关重要的作用[25]。安徽省地貌多样，包括平原、丘陵、山地、湖泊和河流，为农业排放研究提供了理想的典型案例。

Research on agricultural carbon emissions in China remains limited, with a particular scarcity of studies examining Anhui Province. Existing literature primarily addresses two dimensions: the spatiotemporal patterns and driving factors of agricultural carbon emissions [7-12], and the relationships between agricultural carbon emissions and various external factors. These investigations encompass the impact of agricultural production trusteeship on carbon reduction [13], forecasting of agricultural-pastoral carbon emissions [7], economic development influences [14,15], and food security or decoupling effects [16-19]. These studies span diverse agricultural regions including the Yellow River Basin, "Belt and Road" countries, and multiple provinces [11,20,21]. Agricultural carbon emissions demonstrate significant spatial heterogeneity due to variations in geographical scale and economic development, rendering absolute emission values inadequate for accurately assessing regional emission performance. Consequently, this study employs agricultural CEE—combining emission data with agricultural gross output value—as a more comprehensive indicator for evaluating regional agricultural emission performance. Previous research has examined the relationship between agricultural CEE and food security [18,19]as well as agricultural CEE's driving mechanisms [21]. Overall, existing agricultural CEE literature exhibits limitations in both geographical scope and analytical depth. Studies specifically addressing Anhui Province's agricultural CEE remain notably scarce. Furthermore, extant research on agricultural CEE driving factors lacks methodological comprehensiveness, either employing conventional algorithms to broadly analyze influence mechanisms [22] or focusing on inter-factor relationships [5]while overlooking systemic impacts.
中国农业碳排放研究尚处于起步阶段，其中安徽省的相关研究尤为匮乏。现有文献主要聚焦于两个维度：农业碳排放的时空分布特征及其驱动因素[7-12]，以及农业碳排放与各类外部因素之间的关联性。这些研究涵盖了农业生产承包制对碳减排的影响[13]、农业-牧业碳排放预测[7]、经济发展影响[14,15]以及粮食安全或脱钩效应[16-19]。这些研究涉及黄河流域、"一带一路"国家及多个省份[11,20,21]农业碳排放因地理尺度和经济发展水平的差异而呈现显著的空间异质性，导致绝对排放量无法准确评估区域排放绩效。因此，本研究采用农业碳排放强度（CEE）——将排放数据与农业总产值相结合——作为更全面的指标，用于评估区域农业排放绩效。先前研究已探讨了农业 CEE 与粮食安全的关系[18,19]以及农业 CEE 的驱动机制[21]。总体而言，现有农业 CEE 研究在地理范围和分析深度方面均存在局限性。针对安徽省农业 CEE 的专门研究仍较为匮乏。 此外，现有关于农业 CEE 驱动因素的研究在方法论上存在不全面性，要么采用传统算法进行广泛的影响机制分析[22]，要么聚焦于因素间关系[5]，而忽视了系统性影响。

This study addresses the regional research gap by examining spatial differentiation patterns of agricultural CEE in Anhui Province, while advancing both substantive knowledge and methodological approaches. Our methodology incorporates both global and local Moran's I indices with LISA scatter plots [51]to identify spatiotemporal patterns of agricultural CEE in Anhui from 2010 to 2022, followed by machine learning algorithms enhanced with SHAP interpretability analysis to quantify determinant variable impacts. To explain local anomalies—particularly the spatial "siphon effect"—we employ the SDM to analyze both direct and indirect effects of various agricultural activities on CEE. Additionally, we incorporate quadratic terms to assess potential nonlinear relationships within these impact mechanisms. Finally, we implement a coupling coordination model to evaluate Anhui Province's sustainable development potential, forming the basis for policy recommendations that simultaneously optimize agricultural economic development and carbon emission reduction. By contextualizing findings within Anhui's unique geographical and topographical characteristics, this study contributes to the broader understanding of agricultural CEE nationwide, potentially accelerating progress toward national carbon neutrality goals.
本研究通过考察安徽省农业碳排放强度（CEE）的空间差异化模式，填补了区域研究空白，同时在实质性知识和方法论方面均有所推进。我们的研究方法结合了全球和地方莫兰指数（Moran's I）与 LISA 散点图[51]，识别了 2010 年至 2022 年安徽省农业 CEE 的时空模式，随后采用增强型机器学习算法并结合 SHAP 可解释性分析，量化了决定性变量的影响。为解释局部异常现象——尤其是空间“虹吸效应”——我们采用 SDM 模型分析各类农业活动对碳排放的直接与间接影响。此外，引入二次项以评估这些影响机制中潜在的非线性关系。最后，我们构建了耦合协调模型，评估安徽省可持续发展的潜力，为同时优化农业经济发展和碳排放减少的政策建议提供依据。通过将研究结果置于安徽省独特的地理和地形特征背景下，本研究为全国农业碳排放强度变化的理解提供了新的视角，有望加速实现国家碳中和目标的进程。

This paper is structured as follows. Section 2 presents the data sources and methodological framework employed in this study. Section 3 details the empirical findings and analytical results. Section 4 discusses the key implications of our findings and proposes targeted policy recommendations. Section 5 concludes by synthesizing the research contributions and limitations.
本文结构如下。第二部分介绍了本研究采用的数据来源和方法论框架。第三部分详细阐述了实证研究结果和分析结论。第四部分讨论了研究发现的关键启示，并提出了针对性的政策建议。第五部分总结了研究贡献和局限性。

2. Study Area and Data
2 研究区域与数据

Anhui Province (116°–119°E, 30°–34°N) comprises 6 prefecture-level cities and 9 provincial directly administered county-level units, forming 25 administrative regions with 50 subordinate counties (cities, districts). Located in Eastern China within the Yangtze River Delta hinterland, Anhui Province straddles the Yangtze and Huai Rivers and encompasses three distinct geographical regions: the North China Plain, Jianghuai Hills, and Wannan Mountains, conferring both strategic geographical advantages and resource diversity [47]. Based on distinctive biophysical characteristics and hydrological systems (comprising the Huai, Yangtze, and Qiantang River basins), Anhui Province is conventionally segmented into three regions: Northern Anhui (Fuyang, Bozhou, Suzhou, Bengbu, Huainan), Central Anhui (Chuzhou, Hefei, Lu'an, Anqing), and Southern Anhui (Ma'anshan, Wuhu, Tongling, Chizhou, Xuancheng, Huangshan) [26]. Anhui Province occupies a transitional climatic zone between warm temperate and subtropical regions, with pronounced monsoon characteristics [48]. Northern Anhui, situated north of the Huai River, features predominantly plains topography with relatively arid conditions and wheat-based agricultural systems, exhibiting lower economic development indices; in contrast, the central and southern regions are characterized by hilly and mountainous terrain with higher humidity levels, supporting rice cultivation, tea production, and diverse agricultural commodities [27], correlating with more advanced economic development. Figure 1 illustrates the geographical location and elevation information of Anhui Province.
安徽省（东经 116°–119°，北纬 30°–34°）下辖 6 个地级市和 9 个省直辖县，共组成 25 个行政区，下设 50 个县（市、区）。安徽省位于中国东部长江三角洲腹地，横跨长江与淮河，涵盖三个截然不同的地理区域：华北平原、江淮丘陵和皖南山地，既具备战略地理优势，又拥有丰富的资源多样性[47]。基于独特的生物物理特征和水文系统（包括淮河、长江和钱塘江流域），安徽省传统上被划分为三个地区：皖北（阜阳、亳州、宿州、蚌埠、淮南）、皖中（滁州、合肥、六安、安庆）和皖南（马鞍山、芜湖、铜陵、池州、宣城、黄山）[26]。安徽省位于暖温带与亚热带气候过渡带，具有明显的季风气候特征[48]。安徽北部位于淮河以北，以平原地形为主，气候相对干燥，以小麦为主的农业生产体系，经济发展指标较低；而中部和南部地区则以丘陵和山地地形为主，湿度较高，适宜水稻种植、茶叶生产及多种农产品生产[27]，与较高的经济发展水平相吻合。图 1 展示了安徽省的地理位置及海拔信息。

Figure 1. Physiographic setting of the study region
图 1. 研究区域的自然地理环境

This study develops an integrated analytical framework utilizing multi-source heterogeneous data derived from the 'Anhui Statistical Yearbook,' 'China Agricultural Yearbook,' and statistical bulletins of 16 prefecture-level cities spanning 2010-2022. Missing data were addressed using cubic spline interpolation techniques, with robust validation through Pearson correlation coefficient testing (r>0.98, p<0.01) confirming high concordance between interpolated values and original observations. Data normalization was achieved through Z-score standardization to eliminate dimensional heterogeneity among variables. Spatial and statistical visualizations were generated using specialized analytical platforms including GeoDa, R, and ArcGIS.
本研究构建了一个综合分析框架，利用来自《安徽省统计年鉴》、《中国农业年鉴》以及 2010 年至 2022 年间 16 个地级市统计公报的多源异质数据。缺失数据通过三次样条插值技术进行处理，并通过皮尔逊相关系数检验（r>0.98，p<0.01）进行稳健验证，证实插值值与原始观测值之间具有高度一致性。通过 Z 分数标准化对数据进行归一化处理，以消除变量间的维度异质性。空间与统计可视化分析采用 GeoDa、R 和 ArcGIS 等专业分析平台生成。

3. Methods
3 种方法

Figure 2 shows the methodological flowchart of this study, with detailed explanations to be provided in the following subsections.
图 2 展示了本研究的方法流程图，具体说明将在以下各小节中详细阐述。

Figure 2. Methodology flowchart of this study.
图 2. 本研究的方法流程图。

3.1 Estimation of Agricultural Carbon Emissions
3.1 农业碳排放估算

Based on established research findings, we selected five indicators—effective irrigation area, fertilizer application, agricultural plastic film usage, agricultural diesel consumption, and pesticide usage—to estimate carbon emissions from agricultural production.
基于现有研究成果，我们选取了五个指标——有效灌溉面积、化肥施用量、农业塑料薄膜使用量、农业柴油消耗量和农药使用量——来估算农业生产过程中的碳排放量。

In equation (1), represents the total agricultural carbon emissions; represents the -th carbon source; and represents the carbon emission coefficient of this source, as detailed in Table 1.
在方程（1）中， 表示总农业碳排放量； 表示第 个碳源； 表示该碳源的碳排放系数，具体详见表 1。

Table 1. Agricultural Carbon Sources and Carbon Emission Coefficients
表 1. 农业碳源及碳排放系数

3.2 Calculation of Agricultural CEE
3.2 农业 CEE 的计算

This study defines agricultural CEE as the ratio of agricultural output value (10,000 yuan) to carbon emissions (kg). A higher agricultural CEE value indicates greater agricultural output value created per unit of carbon emitted, thus more accurately reflecting regional agricultural production efficiency under carbon constraints. The calculation formula is as follows:
本研究将农业碳排放效率（CEE）定义为农业产值（10,000 元）与碳排放量（千克）的比值。农业碳排放效率值越高，表明单位碳排放所创造的农业产值越高，从而更准确地反映碳约束下区域农业生产效率。计算公式如下：

In equation (2), represents agricultural CEE, represents agricultural output value, and represents carbon emissions.
在方程（2）中， 表示农业碳排放强度， 表示农业产值， 表示碳排放量。

3.3 Analytical Methods for Spatial Differentiation Patterns of Agricultural Carbon Emissions
3.3 农业碳排放空间差异化模式的分析方法

Moran's I index is an important spatial statistical measure used to determine the existence of spatial autocorrelation, thereby guiding the selection of appropriate spatial statistical methods [50]. To examine the overall spatial dependence of agricultural CEE, we employed the global Moran's I index to analyze the spatial differentiation characteristics of agricultural carbon emissions. The definition is as follows:
莫兰指数（Moran's I index）是一种重要的空间统计量，用于判断空间自相关性的存在，从而指导选择合适的空间统计方法[50]。为了考察农业碳排放的整体空间依赖性，我们采用了全球莫兰指数来分析农业碳排放的空间差异特征。其定义如下：

In equation (3) and (4), represents Moran's I; is the deviation of the attribute value from its mean, where is the agricultural CEE of a specific region and is the average agricultural CEE across all regions; is the spatial weight between elements and ; and is the aggregation of all spatial weights.
在方程（3）和（4）中， 表示莫兰指数； 是属性值与其平均值的偏差，其中 是特定地区的农业 CEE， 是所有地区农业 CEE 的平均值； 是元素 和 之间的空间权重； 是所有空间权重的聚合。

For Moran’s I, a standardized statistic is typically used to test the significance of spatial autocorrelation. The value range of is between [-1, +1], with values greater than 0 indicating positive spatial correlation, values less than 0 indicating negative correlation, and a value of 0 indicating no spatial autocorrelation. The calculation formula is:
对于莫兰指数（Morán’s I），通常使用标准化统计量 来检验空间自相关性的显著性。 的取值范围为[-1, +1]，其中大于 0 的值表示正空间相关性，小于 0 的值表示负相关性，而值为 0 则表示不存在空间自相关性。计算公式为：

In equation (5), and are the theoretical expected value and theoretical variance of , respectively. If the value is positive and statistically significant, it indicates a significant positive spatial correlation, meaning similar observation values tend to cluster spatially.
在方程（5）中， 和 分别表示 的理论期望值和理论方差。如果 的值为正且在统计上显著，则表明存在显著的正空间相关性，即相似的观测值倾向于在空间上聚集。

When global spatial correlation exists, the Local Moran's I index is used to identify spatial clustering relationships, with the calculation method as follows:
当存在全球空间相关性时，采用局部莫兰指数（Local Moran's I）来识别空间聚类关系，其计算方法如下：

Where, represents the local Moran's I index; , where represents the agricultural CEE of a th region . The calculation method of in Equation (6) is as follows:
其中， 表示局部莫兰指数； 表示第 个区域的农业 CEE，其中 表示第 个区域的农业 CEE。方程 (6) 中 的计算方法如下：

Using from Equation (6) as the x-axis and as the y-axis, the plane can be divided into four quadrants. By analyzing regions with confidence levels greater than 95% (exhibiting non-random distribution), specific spatial clustering relationships can be determined.
将方程（6）中的 作为 x 轴， 作为 y 轴，平面可被划分为四个象限。通过分析置信水平大于 95%的区域（表现出非随机分布），可确定具体的空间聚类关系。

3.4 SHAP Interpretability Analysis Model
3.4 SHAP 可解释性分析模型

3.4.1 Random Forest
3.4.1 随机森林

Random forest is an ensemble learning algorithm composed of multiple decision trees that enhances model prediction accuracy and robustness by integrating their predictive results [29]. After model training is completed, the feature importance assessment function quantifies the influence intensity of various factors on agricultural CEE. Feature importance is calculated using the following formula:
随机森林是一种由多个决策树组成的集成学习算法，通过整合各决策树的预测结果来提升模型预测的准确性和鲁棒性[29]。模型训练完成后，特征重要性评估函数用于量化各类因素对农业 CEE 的影响强度。特征重要性通过以下公式计算：

In equation (8), denotes the -th feature, represents the number of decision trees, and indicates the impurity reduction by feature at node .
在方程（8）中， 表示第 个特征， 表示决策树的数量， 表示特征 在节点 处的纯度降低量。

3.4.2 SHAP interpretability analysis
3.4.2 SHAP 可解释性分析

Based on the random forest model, SHAP interpretability analysis applies Shapley value theory [33] to quantify the specific contribution of each feature to individual predictions. The calculation formula for the SHAP value of feature is:
基于随机森林模型，SHAP 可解释性分析采用夏普利值理论[33]来量化每个特征对个体预测的具体贡献。特征 的 SHAP 值计算公式为：

Where is a feature subset, is the set of all features, is the model output on feature subset , and is the total number of features.
其中， 表示特征子集， 表示所有特征的集合， 表示模型在特征子集 上的输出， 表示特征的总数。

3.5 Spatial Econometric Model: SDM
3.5 空间计量经济学模型：SDM

Dependencies in spatial relationships occur not only in dependent variables but also in independent variables [45]. The calculation formula is:
空间关系中的依赖性不仅存在于因变量中，也存在于自变量中[45]。计算公式为：

Whererepresents the dependent variable, represents the independent variable, represents time, is the random error term, and are spatial correlation coefficients, represents the spatial lag coefficient, and is the number of regions.
其中， 表示因变量， 表示自变量， 表示时间， 表示随机误差项， 和 表示空间相关系数， 表示空间滞后系数， 表示区域数量。

3.6 Coupling Coordination Model
3.6 耦合协调模型

The coupling coordination model serves as a quantitative indicator to measure the coordination state between systems, exploring the degree of interaction and development level between systems [49]. Based on existing research, the following formulas are used to quantitatively measure the coupling coordination degree between agricultural output value and agricultural CEE in Anhui Province:
耦合协调模型作为定量指标，用于衡量系统之间的协调状态，探讨系统之间相互作用的程度及发展水平[49]。基于现有研究，采用以下公式定量衡量安徽省农业产值与农业 CEE 之间的耦合协调程度：

In these equations, represents the agricultural output value of various cities in Anhui Province; represents the agricultural CEE of various cities in Anhui Province; represents the coupling degree; represents the coordination index; andrepresents the coupling coordination degree, where , The closer is to 1, the higher the coupling coordination degree between agricultural output value and agricultural CEE across cities in Anhui Province.
在这些方程中， 表示安徽省各城市的农业产值； 表示安徽省各城市的农业 CEE； 表示耦合度； 表示协调指数； 表示耦合协调度，其中 ， 越接近 1，安徽省各城市农业产值与农业 CEE 之间的耦合协调度越高。

4. Results
4 结果

4.1 Geographic Information System (GIS) Calculation Results
4.1 地理信息系统（GIS）计算结果

4.1.1 Agricultural Production Factors
4.1.1 农业生产要素

Agricultural production input utilization in Anhui Province showed significant spatiotemporal differentiation between 2010 and 2022 (Table 2). Fertilizer application decreased from 3.198 to 2.718 million tons (annual reduction rate: 1.4%, p<0.05), while pesticide usage declined from 117,000 to 74,000 tons (annual reduction rate: 3.6%). Conversely, effective irrigation area expanded by 23.9%, and agricultural plastic film and diesel consumption increased by 26.7% and 8.7%, respectively. Spatial analysis (Figure 3) reveals that input factor intensity follows a high in the north, low in the south gradient. GIS kernel density estimation shows that consumption centers for fertilizers, plastic films, pesticides, and diesel shifted eastward by 38.2±5.6 kilometers during 2010-2022, while high-value irrigation areas expanded northward by 12.3%. These spatial shifts strongly correlate with regional economic development disparities and evolving agricultural production structures. Recent zero-growth initiatives for chemical inputs and policies promoting organic fertilizer substitution have contributed to declining fertilizer usage and reduced agricultural input redundancy rates [27]
安徽省农业生产投入利用在 2010 年至 2022 年间呈现显著的时空差异（表 2）。化肥施用量从 319.8 万吨减少至 271.8 万吨（年均减少率：1.4%，p<0.05），农药使用量从 11.7 万吨减少至 7.4 万吨（年均减少率：3.6%）。相反，有效灌溉面积扩大了 23.9%，农业塑料薄膜和柴油消耗量分别增加了 26.7%和 8.7%。空间分析（图 3）显示，投入因子强度呈现北高南低的梯度分布。GIS 核密度估算显示，2010 年至 2022 年间，化肥、塑料薄膜、农药和柴油的消费中心向东移动了 38.2±5.6 公里，而高价值灌溉区向北扩展了 12.3%。这些空间变化与区域经济发展差异及农业生产结构演变密切相关。近期实施的化学投入零增长政策及推广有机肥料替代的措施，已推动化肥使用量下降并降低农业投入冗余率[27]。.

Table 2. Consumption of Agricultural Production Factors in Anhui Province (2010-2022)
表 2. 安徽省农业生产要素消耗量（2010-2022）

Figure 3. Spatial variation of agricultural practices
图 3. 农业生产方式的空间分布

4.1.2 Analysis of Agricultural CEE
4.1.2 农业 CEE 分析

Annual change rates of agricultural CEE were calculated for cities across Anhui Province to identify temporal fluctuation patterns during the study period. As illustrated in Figure 4, developed urban centers such as Hefei and Wuhu showed robust annual growth rates of 3.5-4.2% in agricultural CEE, while northern Anhui cities including Huaibei and Bozhou exhibited more modest growth rates of 1.2-1.8%. Notably, Tongling experienced a significant 9.7% decrease in agricultural CEE during 2019, primarily due to the negative impacts of regional flooding disasters. During the policy-intensive implementation period (2018-2020), the provincial agricultural CEE growth rate accelerated to 2.8%, confirming the phased strengthening effect of targeted emission reduction policies.
安徽省各城市农业碳排放强度（CEE）的年变化率被计算出来，以识别研究期间的时空波动模式。如图 4 所示，合肥和芜湖等发达城市中心展现出 3.5%至 4.2%的强劲年增长率，而安徽北部城市如淮北和亳州则呈现 1.2%至 1.8%的较为温和的增长率。值得注意的是，2019 年铜陵市的农业 CEE 出现了显著下降，降幅达 9.7%，主要归因于区域性洪涝灾害的负面影响。在政策密集实施阶段（2018-2020 年），全省农业 CEE 增长率加速至 2.8%，这证实了针对性减排政策分阶段强化效应的有效性。

Figure 4. Trends in Agricultural CEE Changes Across Cities in Anhui Province (2010-2022)
图 4. 安徽省各城市农业 CEE 变化趋势（2010-2022）

Multi-period comparative analysis (Figure 5) reveals that agricultural CEE levels in central and southern cities of Anhui Province consistently exceed those in northern regions, with the overall provincial agricultural CEE showing significant spatial convergence patterns. With enhanced agricultural production conditions and improved economic circumstances, agricultural CEE has increased across all regions of Anhui Province, with the western region centered around Lu'an City showing the most significant improvement. This substantial efficiency improvement correlates strongly with Lu'an City's implementation of a comprehensive energy conservation, emission reduction, and low-carbon development action plan [37]. The observed inter-annual and spatial variations suggest that progressive industrial structure optimization and strengthened environmental pollution governance in Anhui Province have contributed to increasingly effective carbon emission control mechanisms [31]
多时期比较分析（图 5）显示，安徽省中部和南部城市的农业碳排放强度（CEE）水平始终高于北部地区，全省农业 CEE 呈现显著的空间趋同趋势。随着农业生产条件的改善和经济状况的提升，安徽省各地区农业 CEE 均有所提升，其中以六安为中心的西部地区改善最为显著。这一显著效率提升与六安市实施的综合节能减排和低碳发展行动计划[37]密切相关。观察到的年际和空间变异表明，安徽省逐步优化的产业结构和加强的环境污染治理措施，为碳排放控制机制的日益有效发挥提供了重要支撑[31]。.

Figure 5. Spatial Distribution Patterns of Agricultural CEE in Anhui Province (2010-2022)
图 5. 安徽省农业 CEE 的空间分布模式（2010-2022）

4.1.3 Spatial Autocorrelation Tests and Local Outlier Analysis
4.1.3 空间自相关性检验与局部异常值分析

Analysis utilizing the spatial adjacency weight matrix (Figure 6) indicates that agricultural CEE in Anhui Province demonstrates significant spatial autocorrelation (global Moran's I: 0.145, p<0.1, 2010-2022 average). The temporal evolution of spatial dependence exhibits a distinct three-stage pattern (Table 3):
基于空间邻接权重矩阵（图 6）的分析表明，安徽省农业 CEE 存在显著的空间自相关性（全球莫兰指数：0.145，p<0.1，2010-2022 年平均值）。空间依赖性的时空演变呈现出明显的三个阶段模式（表 3）：

Stage I—Agglomeration Intensification Period (2010-2014): Global Moran's I increased significantly from 0.065 to 0.221 (p<0.05), reflecting strengthened spatial association attributable to widespread agricultural mechanization initiatives.
第一阶段——聚合强化期（2010-2014）：全球莫兰指数显著提升，从 0.065 升至 0.221（p<0.05），这反映出空间关联性增强，主要归因于农业机械化措施的广泛推行。

Stage II—Disaster Disturbance Period (2015-2017): Global Moran's I decreased sharply by 42.6%, with spatial correlation substantially weakened. This pronounced fluctuation corresponds primarily to the province-wide natural disasters in Anhui during 2016, which resulted in extensive crop damage. The agricultural disaster-affected area exceeded 1.3 million hectares [52]
第二阶段——灾害扰动期（2015-2017 年）：全球莫兰指数（Moran's I）急剧下降 42.6%，空间相关性显著减弱。这一显著波动主要对应安徽省 2016 年发生的全省性自然灾害，导致大面积农作物受损。农业灾害受灾面积超过 130 万公顷[52]。

Stage III—Policy Reconstruction Period (2018-2022): The average Global Moran's I value stabilized at 0.137. During this period, Anhui Province implemented national agricultural sustainability initiatives by systematically reducing chemical fertilizer and pesticide applications, fostering coordinated development of agricultural CEE across regions that gradually strengthened correlation with geographical factors. Nevertheless, in recent years (2020-2022), Global Moran's I values have exhibited a modest declining trend. This trend corroborates the hypothesis that progressive technology diffusion increasingly attenuates traditional geographical constraints on agricultural efficiency performance.
第三阶段——政策重建期（2018-2022）：全球莫兰指数（Global Moran's I）的平均值稳定在 0.137。在此期间，安徽省通过系统性减少化肥和农药使用量，推动农业碳中和（CEE）在各地区协调发展，逐步增强与地理因素的相关性，落实了国家农业可持续发展政策。然而，近年来（2020-2022 年），全球莫兰指数呈现出轻微下降趋势。这一趋势印证了技术扩散的逐步推进正逐渐削弱传统地理因素对农业效率表现的制约作用。

Table 3. Moran's I Test Results (2010-2022)
表 3. 莫兰 I 系数检验结果（2010-2022）

*Significant at p < 0.05
*在 p < 0.05 水平上具有统计学意义。

Figure 6. Heatmap of Spatial Weight Matrix for Anhui Province
图 6. 安徽省空间权重矩阵的热力图

Local Moran's I tests were conducted to identify specific spatial clustering patterns of agricultural CEE within the study region. Figure 7 illustrates the local Moran's I spatial autocorrelation results across the provincial territory. Among the 16 cities analyzed, an average of 11 regions annually clustered within the first quadrant (high-high cluster) and third quadrant (low-low cluster), demonstrating statistically significant positive spatial autocorrelation. Concurrently, an average of 6 regions annually exhibited spatial outlier characteristics in the second quadrant (high-low cluster) and fourth quadrant (low-high cluster), indicating localized spatial heterogeneity.
局部莫兰指数（Moran's I）分析用于识别研究区域内农业 CEE 的特定空间聚类模式。图 7 展示了省级行政区范围内局部莫兰指数的空间自相关结果。在所分析的 16 个城市中，平均每年有 11 个地区聚类于第一象限（高-高聚类）和第三象限（低-低聚类），表明存在统计学上显著的正空间自相关。同时，平均每年有 6 个地区在第二象限（高-低聚类）和第四象限（低-高聚类）中表现出空间异常特征，表明存在局部空间异质性。

Figure 8 depicts the evolving spatial clustering typology of the 16 prefecture-level cities throughout the study period. Results demonstrate that the spatial high-value concentration of agricultural CEE in Anhui Province persistently remains anchored in the central and southern regions. Southern Anhui cities—particularly Huangshan and Xuancheng—form stable high-high clusters (H-H), constituting a core efficiency agglomeration zone. Hefei exhibits statistically significant high-low spatial anomalies (H-L cluster, z=2.31), indicative of a spatial "siphon effect" whereby this provincial capital's agglomeration of technological resources and innovation capacity induces talent and capital outflow from adjacent regions, generating a pronounced center-periphery gradient pattern. Temporally, spatial agglomeration phenomena in northern and western Anhui have progressively intensified, as evidenced by evolving spatial patterns in Bozhou and Bengbu. This spatial evolution suggests that cross-regional collaborative governance mechanisms could effectively mitigate existing development disparities in northern Anhui, potentially accelerating balanced agricultural low-carbon transformation trajectories throughout the province.
图 8 展示了研究期间 16 个地级市空间聚类类型的演变。研究结果表明，安徽省农业 CEE 的空间高价值集中现象始终稳定地集中在中部和南部地区。安徽南部城市——尤其是黄山和宣城——形成稳定的高-高聚类（H-H），构成一个核心效率聚集区。合肥市呈现出统计上显著的高低空间异常（H-L 聚类，z=2.31），表明该省会城市通过聚集技术资源和创新能力，引发人才和资本从周边地区外流，形成明显的中心-边缘梯度分布模式。从时间维度看，安徽北部和西部地区的空间聚合现象呈现逐步强化趋势，这在亳州和蚌埠的空间格局演变中得以体现。这种空间演变表明，跨区域协作治理机制有望有效缓解安徽北部地区现有发展不平衡问题，并可能加速全省农业低碳转型进程。

Figure 7. LISA Scatter Plot of Agricultural CEE in Anhui Province
图 7. 安徽省农业 CEE 的 LISA 散点图

Figure 8. Local Moran's Index Cluster Map of Anhui Province (2010-2022)
图 8. 安徽省局部莫兰指数聚类图（2010-2022）

4.2 Driving Factors and Their Interactive Relationships in Agricultural CEE
4.2 农业 CEE 中的驱动因素及其相互作用关系

4.2.1 SHAP Interpretability Analysis of Agricultural CEE Determinants
4.2.1 农业 CEE 决定因素的 SHAP 可解释性分析

We employed an explainable machine learning approach [35]using a random forest regression model integrated with SHAP value analysis [34]. This methodology calculates average absolute SHAP values for individual features, systematically quantifying their contributions to model predictions. We implemented two complementary visualization techniques: SHAP Summary Beeswarm Plot [32]and feature importance graphs, both illustrating SHAP value distributions across explanatory variables (Figure 9).
我们采用了一种可解释的机器学习方法[35]，结合随机森林回归模型与 SHAP 值分析[34]。该方法通过计算每个特征的平均绝对 SHAP 值，系统性地量化其对模型预测的贡献。我们实现了两种互补的可视化技术：SHAP 摘要蜂群图[32]和特征重要性图，两者均展示了解释变量中 SHAP 值的分布情况（图 9）。

Figure 9(a) depicts SHAP value distributions that quantify the impact of key agricultural factors on CEE. Five explanatory variables were incorporated as model inputs: fertilizer usage (x₁), effective irrigation area (x₂), agricultural plastic film usage (x₃), agricultural diesel consumption (x₄), and pesticide usage (x₅), with agricultural CEE as the response variable. Results show that agricultural plastic film usage (x₃) exerts a pronounced negative impact on agricultural CEE, exhibiting the widest SHAP value distribution range and indicating maximal model sensitivity to this factor. For pesticide usage (x₅), elevated application rates paradoxically correlate with enhanced agricultural CEE (positive SHAP values). Conversely, increased fertilizer application (x₁) shows a consistent negative association with efficiency (negative SHAP values), indicating diminishing returns at higher application rates.
图 9(a)展示了 SHAP 值分布，用于量化关键农业因素对 CEE 的影响。模型输入了五个解释变量：化肥使用量（x₁）、有效灌溉面积（x₂）、农业塑料薄膜使用量（x₃）、农业柴油消耗量（x₄）和农药使用量（x₅），农业 CEE 作为响应变量。结果显示，农业塑料薄膜使用量（x₃）对农业 CEE 具有显著负面影响，其 SHAP 值分布范围最广，表明模型对该因素的敏感性最高。对于农药使用量（x₅），施用量增加反而与农业 CEE 提升呈正相关（正 SHAP 值）。相反，化肥施用量增加（x₁）与效率呈一致的负相关关系（负 SHAP 值），表明施用量越高，效率越低。

SHAP values for agricultural diesel consumption (x₄) exhibit bidirectional distribution across both positive and negative domains, with high-magnitude values (darker points) distributed symmetrically along the spectrum. This pattern indicates that diesel consumption impacts on agricultural CEE are heterogeneous, enhancing efficiency in some contexts (positive SHAP values) while diminishing it in others (negative SHAP values). This symmetric distribution suggests diesel consumption effects show considerable contextual variability, likely influenced by interactions with other agricultural factors or regional characteristics. Conversely, SHAP values for effective irrigation area (x₂) predominantly cluster in the positive domain, indicating a consistent efficiency-enhancing effect on agricultural CEE, though with modest magnitude compared to other factors.
农业柴油消费（x₄）的 SHAP 值在正负两个域中呈现双向分布，高幅度值（深色点）沿光谱对称分布。这一模式表明，柴油消费对农业 CEE 的影响具有异质性，在某些情境下提升效率（正 SHAP 值），而在其他情境下降低效率（负 SHAP 值）。这种对称分布表明，柴油消费的影响存在显著的上下文变异性，可能受其他农业因素或区域特征的交互作用影响。相反，有效灌溉面积（x₂）的 SHAP 值主要集中在正值域，表明其对农业 CEE 具有一致的效率提升效果，尽管与其他因素相比，其影响幅度相对较小。

Figure 9(b) shows that agricultural plastic film usage (x₃) contributes the highest explanatory power to agricultural CEE variance, accounting for 34.3% of model prediction power, underscoring its central role in the agricultural carbon emission system. Pesticide usage (x₅) and fertilizer application (x₁) rank second and third in importance at 22.5% and 18.8% respectively, while agricultural diesel consumption (x₄) and effective irrigation area (x₂) show limited predictive contributions. Further SHAP dependency analysis reveals that increases in agricultural plastic film usage consistently exacerbate agricultural CEE deterioration, suggesting structural inefficiencies in current plastic film management practices.
图 9(b)显示，农业塑料薄膜使用量（x₃）对农业碳排放强度（CEE）变异性的解释能力最高，占模型预测能力的 34.3%，凸显其在农业碳排放系统中的核心作用。农药使用量（x₅）和化肥施用量（x₁）分别以 22.5%和 18.8%的占比位列第二和第三位，而农业柴油消耗量（x₄）和有效灌溉面积（x₂）的预测贡献有限。进一步的 SHAP 依赖性分析表明，农业塑料薄膜使用量的增加会持续加剧农业碳排放效率的恶化，这表明当前塑料薄膜管理实践中存在结构性低效问题。

Our findings suggest that optimizing agricultural CEE in Anhui Province requires prioritizing three interventions: controlling plastic film application intensity, reducing agrochemical inputs (particularly plastic films and fertilizers), and enhancing soil quality parameters including fertility and carbon sequestration capacity [36]. Despite the positive correlation between pesticide application and agricultural CEE, the ecological toxicity of excessive pesticide use requires implementing optimized protocols that balance productivity benefits with ecosystem preservation. Although diesel consumption and irrigation infrastructure show modest contributions to CEE variation, these factors warrant integration into regional agricultural policy frameworks for holistic system optimization.
我们的研究结果表明，优化安徽省农业碳排放效率（CEE）需优先实施三项干预措施：控制塑料薄膜使用强度、减少农药化肥投入（尤其是塑料薄膜和化肥），以及提升土壤质量参数（包括肥力和碳封存能力）[36]。尽管农药使用与农业碳排放效率呈正相关，但过量农药使用的生态毒性要求实施优化方案，以平衡生产效益与生态系统保护。尽管柴油消耗和灌溉基础设施对碳排放效率变异贡献有限，但这些因素仍需纳入区域农业政策框架，以实现系统优化。

Figure 9. Importance of Sample Feature Values for Model agricultural CEE
图 9. 样本特征值对农业 CEE 模型的重要性

4.2.2 SDM Assessment of Factor Interaction Effects
4.2.2 决策支持模型（SDM）对因素交互作用效应的评估

Previous research identified potential spatial "siphon effects" in agricultural resource distribution, necessitating the incorporation of spatial econometric techniques including Lagrange Multiplier (LM) tests for spatial lag variables and spatial autocorrelation coefficient (Rho(ρ)) analysis. Based on comprehensive diagnostic testing, the SDM was selected as the optimal specification for quantifying spatial interaction effects. Detailed estimation results and model diagnostics are presented in Table 4, while model goodness-of-fit indicators (pseudo-R², Akaike Information Criterion, Bayesian Information Criterion) are summarized in Table 5.
先前研究识别出农业资源分布中潜在的空间“虹吸效应”，这 necessitating 空间计量经济学技术，包括拉格朗日乘子（LM）检验用于空间滞后变量及空间自相关系数（Rho(ρ)）分析。基于全面诊断测试，SDM 被选定为量化空间交互作用效应的最优规格。详细的估计结果和模型诊断见表 4，而模型拟合优度指标（伪 R²、Akaike 信息准则、贝叶斯信息准则）总结于表 5。

Table 4. Parameter Estimates and Spatial Diagnostic Statistics from the SDM
表 4. SDM 的参数估计及空间诊断统计量

Notes: ***, **, *, · indicate statistical significance at the 0.1%, 1%, 5% and 10% levels, respectively.
注释：***、**、*、·分别表示统计显著性水平为 0.1%、1%、5%和 10%。

Table 5. Comparative Model Performance Metrics for the SDM
表 5. SDM 的比较模型性能指标

The SDM exhibits robust explanatory power with a pseudo-R² value of 0.7741, explaining approximately 77.41% of variance in agricultural CEE across the study region. The model demonstrates superior fit compared to ordinary linear specifications as evidenced by lower Akaike Information Criterion values, while the Lagrange Multiplier test for residual autocorrelation (p=0.57549) confirms the absence of significant spatial dependencies in model residuals, validating the SDM's effectiveness in capturing underlying spatial structures. Spatial autocorrelation analysis reveals a statistically significant spatial coefficient (Rho=0.60569, p<2.22e-16), confirming substantial spatial lag dependency in agricultural CEE distribution patterns. Robust statistical validation through both Wald test (p<2.22e-16) and Likelihood Ratio test (p=2.9976e-15) demonstrates the SDM's significant superiority over non-spatial specifications, empirically confirming that regional agricultural CEE is determined by complex interaction effects between local agricultural input factors and spillover influences from contiguous geographical units.
SDM 模型展现出强大的解释能力，伪 R² 值为 0.7741，解释了研究区域内农业 CEE 变异性的约 77.41%。与普通线性模型相比，该模型表现出更优的拟合效果，这体现在更低的 Akaike 信息准则值上。此外，残差自相关性拉格朗日乘子检验（p=0.57549）证实了模型残差中不存在显著的空间依赖性，这进一步验证了 SDM 在捕捉潜在空间结构方面的有效性。空间自相关分析显示，空间系数显著（Rho=0.60569，p<2.22e-16），证实农业 CEE 分布模式存在显著的空间滞后依赖性。通过沃尔德检验（p<2.22e-16）和似然比检验（p=2.9976e-15）的稳健统计验证，证明了 SDM 相较于非空间模型具有显著优越性，实证证实了区域农业 CEE 由本地农业投入因素与相邻地理单元溢出效应之间的复杂交互作用共同决定。

Local driving factor analysis revealed that fertilizer (β = -0.748, p < 2.53e-14) and pesticide usage (β = -0.335, p < 6.17e-06) significantly inhibit agricultural CEE, while agricultural plastic film (β = 0.783, p < 2.2e-16) and effective irrigation area (β = 0.622, p < 1.84e-14) significantly promote CEE by enhancing resource utilization efficiency. Notably, agricultural diesel impact on CEE did not pass the significance test (β = -0.151, p = 0.105), suggesting its effect may be influenced by other factors. Further analysis of spatial lag effects revealed that agricultural inputs in adjacent regions have asymmetric impacts on local CEE. Specifically, reduced pesticide (lag.pesticide: β = -0.771, p < 6.64e-06) and agricultural diesel (lag.agricultural diesel: β = -0.483, p = 0.032) usage in adjacent regions significantly enhances local CEE, indicating cross-regional environmental governance can alleviate carbon emission pressure through synergistic effects. However, agricultural plastic film usage in adjacent regions (lag.agricultural plastic film: β = -0.641, p = 0.008) exhibits a negative impact on local CEE, suggesting the need for optimized resource allocation through technology sharing
局部驱动因素分析表明，肥料（β = -0.748，p < 2.53e-14）和农药使用量（β = -0.335，p < 6.17e-06）显著抑制农业 CEE，而农业塑料薄膜（β = 0.783，p < 2.2e-16）和有效灌溉面积（β = 0.622，p < 1.84e-14）通过提升资源利用效率显著促进 CEE。值得注意的是，农业柴油对 CEE 的影响未通过显著性检验（β = -0.151，p = 0.105），表明其作用可能受其他因素影响。进一步的空间滞后效应分析表明，相邻地区的农业投入对当地 CEE 具有不对称影响。具体而言，相邻地区农药使用量减少（滞后农药：β = -0.771，p < 6.64e-06）和农业柴油使用量减少（滞后农业柴油：β = -0.483，p = 0.032）显著提升了当地 CEE，表明跨区域环境治理可通过协同效应缓解碳排放压力。然而，相邻地区农业塑料薄膜使用量（滞后农业塑料薄膜：β = -0.641，p = 0.008）对当地 CEE 产生负面影响，表明需通过技术共享优化资源配置。.

We then introduced quadratic terms to construct a spatial nonlinear model. Table 6 presents the testing results and Table 7 shows the goodness-of-fit indicators.
随后，我们引入二次项构建了空间非线性模型。表 6 展示了验证结果，表 7 列出了拟合优度指标。

Table 6. Nonlinear Model Test Results
表 6. 非线性模型测试结果

Table 7. Goodness-of-Fit Indicators of Nonlinearity
表 7. 非线性拟合优度指标

The integrated analysis of Tables 6-7 and Figure 10 reveals the following relationships between agricultural inputs and CEE: fertilizer (linear term = -0.748, quadratic term = 0.457, p < 3.16e-13) and pesticide (quadratic term = 0.056, p = 0.015) impacts on CEE exhibit U-shaped relationships, suggesting limited emission reduction benefits at low usage levels but significantly improved marginal benefits beyond threshold values; effective irrigation area displays an inverted U-shaped relationship (linear term = 0.622, quadratic term = -0.246, p < 7.48e-07), with excessive water usage leading to efficiency decline. Furthermore, spatial spillover effect analysis indicates that rational use of agricultural plastic film (lag.I(Agricultural plastic Film²) = 1.155, p < 2.2e-16) and agricultural diesel (lag.I(Agricultural diesel²) = 0.674, p < 7.84e-10) in adjacent regions can enhance local CEE through technology diffusion, while cross-regional competition for effective irrigation area (lag.I(Effective irrigation Area²) = -0.510, p < 1.88e-05) exacerbates efficiency losses, highlighting the importance of watershed collaborative management.
表 6-7 和图 10 的综合分析揭示了农业投入与 CEE 之间的以下关系：化肥（线性项= -0.748，二次项= 0.457，p < 3.16e-13）与农药（二次项= 0.056，p = 0.015）对 CEE 的影响呈现 U 型关系，表明在低使用水平下减排效益有限，但超过阈值后边际效益显著提升；有效灌溉面积则呈现倒 U 型关系（线性项=0.622，二次项=-0.246，p < 7.48e-07），过量用水导致效率下降。此外，空间溢出效应分析表明，农业塑料薄膜的合理使用（滞后.I(农业塑料薄膜²) = 1.155，p < 2.2e-16）和农业柴油（滞后.I(农业柴油²) = 0.674，p < 7.84e-10）在相邻地区可通过技术扩散提升当地 CEE，而跨区域对有效灌溉面积的竞争（滞后.I(有效灌溉面积²) = -0.510，p < 1.88e-05）则加剧效率损失，凸显了流域协同管理的重要性。

Figure 10. Nonlinear Relationships of Production Factors
图 10. 生产要素的非线性关系

4.3 Coupling Degree Analysis
4.3 耦合度分析

Building upon established methodological frameworks [39], this investigation systematically categorizes coupling coordination degree classifications (Table 6) and implements comprehensive temporal assessment of the synergistic relationships between agricultural output value and agricultural CEE across 16 prefecture-level administrative regions in Anhui Province during the 2010-2022 study period. This analytical approach facilitates the development of scientifically-grounded policy recommendations for simultaneously enhancing both agricultural CEE and agricultural economic productivity.
基于既有的方法论框架[39]，本研究系统性地对耦合协调程度分类进行分类（表 6），并针对安徽省 16 个地级行政区 2010-2022 年研究期间农业产值与农业 CEE 之间的协同关系，开展了全面的时序评估。该分析方法为制定科学依据的政策建议提供了依据，旨在同时提升农业 CEE 与农业经济生产力。

Table 8. Coupling Coordination Degree Classification System
表 8. 耦合协调度分类系统

Overall, the coupling coordination degree across the studied regions exhibited a progressive improvement trajectory throughout the 2010-2022 analytical period. Specifically, Huangshan City (coordination range: 0.67-0.72) and Chizhou City (coordination range: 0.57-0.65) maintained consistent high-quality coordination levels over extended temporal intervals, with ecological agricultural output values constituting 34.2% and 28.7% of total agricultural output respectively, and green-certified agricultural product coverage exceeding the 50% threshold; Tongling City exhibited a notable coupling coordination degree progression (0.68-0.84), which facilitated substantial optimization of agriculture-environment synergistic efficiency parameters, approaching the "excellent coordinated development" classification threshold. This progression reflects fundamental positive transitions in both industrial transformation processes and environmental protection mechanisms within the regional agricultural system. The coupling coordination metrics for Bengbu City and Chuzhou City demonstrated relatively elevated values during the 2018 reference period but exhibited significant temporal volatility across other measurement intervals, indicating potential instability in their agricultural development-environmental protection balance mechanisms.
总体而言，研究区域的耦合协调程度在 2010 年至 2022 年的分析期间呈现出逐步提升的趋势。具体而言，黄山市（协调范围：0.67-0.72）和池州市（协调范围：0.57-0.65）在较长时段内保持了高质量的协调水平，生态农业产值分别占农业总产值的 34.2%和 28.7%，绿色认证农产品覆盖率均超过 50%的阈值；铜陵市呈现显著的耦合协调度提升趋势（0.68-0.84），推动农业与环境协同效率参数实现显著优化，接近“优秀协调发展”分类阈值。这一趋势反映了该地区农业系统在产业转型进程与环境保护机制方面均发生根本性积极转变。蚌埠市和滁州市的耦合协调指标在 2018 年基准期内表现相对较高，但在其他测量区间内呈现显著的时序波动，表明其农业发展与环境保护平衡机制可能存在潜在不稳定性。

Figure 11. Spatiotemporal Dynamics of Coupling Coordination Degree
图 11. 耦合协调度的时空动态变化

The coupling coordination degree, functioning as a critical quantitative metric for measuring the equilibrium between urban agricultural economic output trajectories and agricultural CEE performance, serves as a fundamental parameter in comprehensive assessment frameworks for urban agricultural sustainable development capacity. With the progressive temporal evolution and systematic implementation of scientifically-informed policy intervention measures, carbon emission challenges associated with contemporary urbanization processes will potentially achieve effective mitigation and regulatory control. According to quantitative forecasting models developed by [40], the coordination coefficient of the integrated coupling system between new-paradigm urbanization and carbon emissions trajectories is projected to reach a highly coordinated developmental stage by 2032. Further empirical research by [38]substantiates that systematic enhancement of green innovation technological capabilities represents an effective strategic pathway for catalyzing comprehensive low-carbon transformation processes and facilitating multi-dimensional sustainable development outcomes. Future research directions could systematically investigate the differential impacts of diverse agricultural policy frameworks on coupling coordination degree dynamics, while simultaneously examining optimization pathways for enhancing agricultural sustainable development capacity through integrated technological innovation systems, multi-stakeholder social participation mechanisms, and adaptive governance structures.
耦合协调度作为衡量城市农业经济产出轨迹与农业碳排放与能效（CEE）绩效之间平衡的关键定量指标，是评估城市农业可持续发展能力综合框架中的基础参数。随着科学决策的政策干预措施的逐步实施和系统推进，当代城市化进程中碳排放挑战有望实现有效减缓与调控。根据[40]开发的定量预测模型，新范式城市化与碳排放轨迹之间集成耦合系统的协调系数预计将于 2032 年达到高度协调的发展阶段。[38]的进一步实证研究证实，系统性提升绿色创新技术能力是催化全面低碳转型过程、促进多维度可持续发展成果的有效战略路径。未来研究方向可系统性地探讨不同农业政策框架对耦合协调程度动态的差异化影响，同时研究通过集成技术创新体系、多利益相关方社会参与机制及适应性治理结构，提升农业可持续发展能力的优化路径。

5. Discussion
5 讨论

5.1 Spatiotemporal Characteristics of Agricultural CEE
5.1 农业 CEE 的时空特征

Agricultural production across Anhui Province is progressively transitioning toward low-carbon systems. Agricultural economic development serves as the predominant driver of carbon emission growth in this region [30]. Northern Anhui, a principal grain-producing region, is transitioning toward mechanized, large-scale agricultural systems; Central Anhui promotes diversified development through expanded livestock breeding and economic crop cultivation; Southern Anhui leverages its mountainous resources to cultivate high-value crops such as tea and fruit that generate greater economic returns.
安徽省的农业生产正逐步向低碳系统转型。农业经济发展是该地区碳排放增长的主要驱动力[30]。安徽北部作为主要粮食产区，正向机械化、大规模农业生产体系转型；安徽中部通过扩大畜牧业养殖和经济作物种植推动多元化发展；安徽南部则依托山地资源，发展茶叶、水果等高附加值作物，以实现更高的经济效益。

Agricultural CEE has shown consistent improvement throughout the study period. Western Anhui, with its abundant agricultural resources and strategic position along the Wanjiang Belt for industrial transfer from eastern regions, has experienced accelerated development of modern agriculture, becoming the most rapidly advancing agricultural zone in the province. Spatial autocorrelation analysis reveals that agricultural CEE in Anhui Province exhibits significant spatial dependence, though these clustering patterns have gradually weakened over time. The advancement of water-conservation practices and environmental protection technologies, combined with Anhui's implementation of energy conservation and emission reduction policies, has effectively promoted coordinated development between agricultural carbon mitigation and food industry sustainability.
农业碳排放强度（CEE）在研究期间呈现持续改善趋势。安徽西部凭借丰富的农业资源及在皖江带的战略位置（作为东部地区产业转移的承接地），现代农业发展加速，已成为全省农业发展最快的区域。空间自相关分析表明，安徽省农业碳排放强度存在显著的空间依赖性，但这些聚类模式随时间推移逐渐弱化。节水技术和环境保护技术的进步，加上安徽省实施的节能减排政策，有效促进了农业碳减排与食品工业可持续发展的协调发展。

5.2 Effects of Driving Factors
5.2 驱动因素的影响

Key factors influencing agricultural CEE in Anhui Province were identified through SHAP interpretability analysis and spatial effect analysis using the Durbin model, along with the interactive relationships between these factors. Reasonable control of agricultural film, fertilizer, and pesticide usage is critically important for improving agricultural CEE. Coupling degree analysis indicates that the coordination between agricultural output value and agricultural CEE in Anhui Province generally improved from 2010 to 2022, although significant differences exist among cities.
通过 SHAP 可解释性分析和 Durbin 模型空间效应分析，识别了安徽省农业碳排放强度（CEE）的关键影响因素，并揭示了这些因素之间的相互作用关系。合理控制农业薄膜、化肥和农药的使用对提升农业 CEE 至关重要。耦合度分析表明，2010 年至 2022 年间，安徽省农业产值与农业 CEE 之间的协调程度总体有所提升，但不同城市之间存在显著差异。

Old industrial cities such as Anqing and Huaibei face greater environmental pressure during economic transformation and demonstrate deficiencies in agricultural carbon emission management. These differences reflect both varying urban development models and disparities in agricultural sustainable development and environmental governance capabilities among cities. While Anhui Province has made progress in improving agricultural CEE, differentiated policies are still needed to better balance economic development and environmental protection objectives.
老工业城市如安庆和淮北在经济转型过程中面临更大的环境压力，且在农业碳排放管理方面存在不足。这些差异既反映了不同城市发展模式的差异，也体现了城市间农业可持续发展和环境治理能力的差距。虽然安徽省在提升农业碳排放管理水平方面取得了一定进展，但仍需制定差异化政策，以更好地平衡经济发展与环境保护目标。

5.3 Policy Recommendations
5.3 政策建议

In accordance with modern ecological civilization concepts and current development requirements, the primary task for high-quality agricultural development in Anhui Province remains the continuous advancement of agricultural industrial structure optimization and modernization to enhance competitiveness and achieve low-carbon development goals [41]. Fertilizers and agricultural films have significant negative impacts on agricultural CEE, making reduction in their usage a key priority for improvement. Governments should implement targeted measures to control agricultural non-point source (NPS) pollution and reduce soil contamination [42]. The inverted U-shaped relationship between effective irrigation area and agricultural CEE necessitates strengthened comprehensive assessment of water resources and agricultural production systems. In water-scarce regions, promoting efficient water-saving technologies to achieve precision irrigation is essential for sustainable agricultural development [43]
根据现代生态文明理念和当前发展要求，安徽省高质量农业发展的首要任务仍是持续推进农业产业结构优化和现代化，以提升竞争力并实现低碳发展目标[41]。化肥和农业薄膜对农业碳排放具有显著负面影响，因此减少其使用量是改善现状的关键优先事项。政府应采取针对性措施控制农业非点源污染（NPS）并减少土壤污染[42]。有效灌溉面积与农业碳排放之间存在倒 U 型关系，这要求加强水资源与农业生产系统的综合评估。在水资源匮乏地区，推广高效节水技术以实现精准灌溉是实现农业可持续发展的重要途径[43]。.

Given the spatial spillover effects of agricultural inputs, agricultural science and technology modernization remains fundamental to developing modern agriculture. Governments should create favorable conditions to attract highly educated talent to eco-agriculture industries, improve the educational level of agricultural workers, and establish solid foundations for organic agriculture development. Promoting technological exchange and cooperation among Northern, Central, and Southern Anhui regions is also critical for regional agricultural development [46]. Anqing and Huaibei should formulate comprehensive agricultural green transformation support plans to enhance sustainable development capabilities and promote the integration of agriculture, science, and technology [44]
鉴于农业投入的空间溢出效应，农业科学技术现代化仍是发展现代农业的根本所在。政府应创造有利条件吸引高素质人才投身生态农业产业，提升农业劳动者教育水平，为有机农业发展奠定坚实基础。促进安徽北部、中部和南部地区之间的技术交流与合作，对区域农业发展至关重要[46]。安庆和淮北应制定全面的农业绿色转型支持计划，提升可持续发展能力，促进农业、科学与技术的融合[44]。.

5.4 Limitations and Prospects
5.4 限制与展望

This study found that agricultural CEE in Anhui Province exhibits significant but fluctuating spatial autocorrelation, with global Moran's I average value of 0.145 and a coefficient of variation reaching 62.3%. Mechanism diagnostics reveal that traditional Queen first-order adjacency matrices have limitations in capturing complex geographical boundaries. Specifically, the transition zone between Northern Anhui plains and Southern Anhui mountains (such as Lu'an and Anqing) experiences distorted spatial weight allocation due to hydrological-geomorphological heterogeneity (plains accounting for <30% vs. mountains >65%). This distortion leads to anomalous quadrant values in LISA analysis, with second and fourth quadrants accounting for 31.2% of observations. This limitation may underestimate the spatial spillover effects of cities in the north-south transition zone. For example, the actual technology diffusion potential of Huainan City (identified as a Low-Low outlier) is not fully captured in the model.
本研究发现，安徽省农业 CEE 呈现显著但波动的空间自相关性，全球莫兰指数平均值为 0.145，系数变异值高达 62.3%。机制诊断表明，传统的女王第一阶邻接矩阵在捕捉复杂地理边界方面存在局限性。具体而言，安徽北部平原与南部山区的过渡地带（如六安和安庆）由于水文地貌异质性（平原占比<30% vs. 山区>65%），导致空间权重分配失真。这种扭曲导致 LISA 分析中异常的象限值，第二和第四象限占观测值的 31.2%。这一局限性可能低估了南北过渡区城市空间溢出效应。例如，淮南市（被识别为低-低异常值）的实际技术扩散潜力在模型中未得到充分反映。

Despite these limitations, this study systematically reveals the spatiotemporal differentiation patterns and multi-scale driving mechanisms of agricultural CEE in Anhui Province through the construction of an efficiency-coordination-policy three-dimensional analytical framework. The research conclusions provide empirical support for differentiated policy design in the "Anhui Province Agricultural Carbon Peak Action Plan." This research particularly offers scientific basis for establishing dynamic thresholds for fertilizer application in Northern Anhui (≤287kg per mu) and optimizing collaborative irrigation management in Southern Anhui mountainous areas through cross-city water rights trading mechanisms. Future research should focus on carbon-water coupling mechanisms at small watershed scales and the moderating role of farmer behavior in technology adoption. These investigations will advance agricultural low-carbon transformation from "efficiency improvement" toward comprehensive "system reconstruction."
尽管存在这些局限性，本研究通过构建效率-协调-政策三维分析框架，系统揭示了安徽省农业碳排放的时空差异化特征及多尺度驱动机制。研究结论为《安徽省农业碳达峰行动计划》的差异化政策设计提供了实证依据。本研究特别为安徽北部地区（≤287 公斤/亩）建立化肥施用动态阈值，以及通过跨市水权交易机制优化安徽南部山区灌溉管理提供了科学依据。未来研究应聚焦小流域尺度上的碳-水耦合机制，以及农民行为在技术采用中的调节作用。这些研究将推动农业低碳转型从“效率提升”向“系统重构”的全面转变。"Claude can make mistakes.
Please double-check responses.

6. Conclusion
6 结论

Based on agricultural data from 16 prefecture-level cities in Anhui Province from 2010 to 2022, this study systematically reveals the spatiotemporal differentiation patterns and multi-scale driving mechanisms of agricultural CEE by integrating spatial econometric analysis (global and local Moran's I indices, SDM) with the SHAP interpretability analysis model.
基于安徽省 16 个地级市 2010 年至 2022 年的农业数据，本研究通过整合空间计量经济学分析（全球和局部莫兰指数、SDM）与 SHAP 可解释性分析模型，系统揭示了农业碳排放强度（CEE）的时空差异化特征及多尺度驱动机制。

The study indicates that agricultural CEE in Anhui Province grew at an annual rate of 2.3%, with a declining north-south difference index and significant spatial convergence trends. Southern Anhui regions such as Huangshan and Xuancheng formed stable high-value clusters (High-High) due to ecological agriculture proportions exceeding the provincial average by 1.8 times, while Northern Anhui regions such as Bozhou and Suzhou, constrained by lower agricultural mechanization rates, remained in contiguous low-value areas (Low-Low). SHAP interpretability analysis results indicate that agricultural plastic film usage constitutes the most significant negative driving factor influencing agricultural CEE. Reasonable control of fertilizer and pesticide usage can significantly enhance agricultural CEE. Spatial spillover effects demonstrate that fertilizer and pesticide usage have significant inhibitory effects on agricultural CEE, while agricultural plastic film and effective irrigation area generate positive spillovers through technology diffusion mechanisms.
研究表明，安徽省农业 CEE 以年均 2.3%的增速发展，南北差异指数呈下降趋势，且空间趋同趋势显著。安徽南部地区如黄山、宣城等形成稳定的高价值集群（高-高），因生态农业比例较全省平均高出 1.8 倍，而安徽北部地区如亳州、宿州等受农业机械化率较低制约，仍处于相邻低价值区域（低-低）。SHAP 可解释性分析结果表明，农业塑料薄膜使用是影响农业 CEE 的最显著负面驱动因素。合理控制化肥和农药使用量可显著提升农业 CEE。空间溢出效应显示，化肥和农药使用对农业 CEE 具有显著抑制作用，而农业塑料薄膜和有效灌溉面积则通过技术扩散机制产生正向溢出效应。

Nonlinear threshold analysis further reveals that fertilizer and pesticide application exhibit U-shaped relationships with CEE, while effective irrigation area efficiency displays inverted U-shaped characteristics, highlighting the necessity for precise management and control strategies. Coupling coordination degree assessment shows that Huangshan and Chizhou consistently maintain excellent coordination levels due to carbon labeling systems. Tongling promotes coordination improvement through industrial emission reduction, while Bengbu and Chuzhou experience coordination value fluctuations influenced by climate variability.
非线性阈值分析进一步揭示，化肥和农药施用量与碳排放强度（CEE）呈现 U 型关系，而有效灌溉面积效率则表现出倒 U 型特征，强调了精准管理与控制策略的必要性。协调度评估结果表明，黄山和池州凭借碳标签体系始终保持较高的协调水平。铜陵通过工业排放削减推动协调水平提升，而蚌埠和滁州则受气候变异影响，协调值出现波动。

At the theoretical level, this study constructs an efficiency-spatial-coordination three-dimensional analytical framework, comprehensively evaluating agricultural CEE in Anhui Province and thoroughly analyzing interactions among key driving factors. This framework not only provides new perspectives for global agricultural carbon emission research but also offers empirical evidence for formulating agricultural carbon reduction policies in China. Future research should deepen carbon-water coupling mechanism studies at small watershed scales and explore the moderating role of farmer behavior in technology adoption, thereby advancing agricultural low-carbon transformation from efficiency optimization toward comprehensive system reconstruction.
在理论层面，本研究构建了效率-空间-协调三维分析框架，对安徽省农业碳排放效率进行了全面评估，并深入剖析了关键驱动因素之间的相互作用。该框架不仅为全球农业碳排放研究提供了新视角，还为制定中国农业碳减排政策提供了实证依据。未来研究应深化小流域尺度上的碳-水耦合机制研究，并探索农民行为在技术采用中的调节作用，从而推动农业低碳转型从效率优化向综合系统重构的转变。

Author Contributions
作者贡献

Chenxin Ru: Methodology, Writing- Original draft preparation; Sihan Wang: Supervision, Data Curation; Zhiyan Liang: Investigation, Visualization; Bingyue Liao: Resources, Visualization; Lei Luo: Data curation; Ning Shaowei: Conceptualization, Supervision, Writing – review and editing; Bhesh Raj Thapa: Software, Writing – review and editing
陈心如方法学、写作——初稿撰写；王思涵监督、数据整理；梁志燕：调查、可视化；廖冰月资源、可视化；罗磊数据整理；邵伟宁：概念化、监督、写作——审阅与编辑；Bhesh Raj Thapa：软件、写作——审阅与编辑.

Conflicts of Interest
利益冲突

The authors declare that they have no conflict of interest.
作者声明，他们之间不存在利益冲突。

Funding Statement
资金来源说明

This work was supported by National Training Program of Innovation and Entrepreneurship for Undergraduates (grant number S202410359412), National Natural Science Foundation of China (42271084), and Natural Science Foundation of Anhui Province (2208085US15).
本研究得到国家大学生创新创业培训计划（项目编号：S202410359412）、国家自然科学基金（42271084）和安徽省自然科学基金（2208085US15）的资助。

7. Reference
7 参考文献

Intergovernmental Panel on Climate Change. Climate change: the IPCC scientific assessment [Internet]. Cambridge: Cambridge University Press; 1990 [cited 2025 Apr 5]. Available from: https://www.ipcc.ch/report/climate-change-the-ipcc-1990-and-1992-assessments/
政府间气候变化专门委员会. 气候变化：政府间气候变化专门委员会科学评估报告 [互联网]. 剑桥：剑桥大学出版社；1990 [引用日期：2025 年 4 月 5 日]. 获取自：https://www.ipcc.ch/report/climate-change-the-ipcc-1990-and-1992-assessments/.

Intergovernmental Panel on Climate Change. Climate change 2013: the physical science basis. Contribution of Working Group I to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change [Internet]. Stocker TF, Qin D, Plattner GK, Tignor M, Allen SK, Boschung J, et al., editors. Cambridge: Cambridge University Press; 2013 [cited 2025 Apr 5]. Available from: https://www.ipcc.ch/report/ar5/wg1/
政府间气候变化专门委员会. 气候变化 2013：物理科学基础. 政府间气候变化专门委员会第五次评估报告第一工作组报告 [互联网].斯托克 TF, 秦 D, 普拉特纳 GK, 蒂格纳 M, 艾伦 SK, 博斯洪 J, 等, 编. 剑桥: 剑桥大学出版社; 2013 [引用日期: 2025 年 4 月 5 日]. 来自: https://www.ipcc.ch/report/ar5/wg1/.

United Nations Environment Programme, UNEP Copenhagen Climate Centre. Emissions gap report 2020 [Internet]. United Nations Environment Programme; 2020 [cited 2025 Apr 5]. Available from: https://www.unep.org/emissions-gap-report-2020
联合国环境规划署，联合国环境规划署哥本哈根气候中心. 2020 年温室气体排放差距报告 [互联网]. 联合国环境规划署; 2020 [引用日期: 2025 年 4 月 5 日]. 获取自: https://www.unep.org/emissions-gap-report-2020.

Food and Agriculture Organization of the United Nations. Greenhouse gas emissions from agrifood systems - global, regional and country trends, 2000-2022 [Internet]. FAOSTAT analytical brief series, no. 94. Rome: FAO; 2024 [cited 2025 Apr 5]. Available from: https://openknowledge.fao.org/handle/20.500.14283/cd3167en
联合国粮食及农业组织. 农业食品系统温室气体排放——全球、区域及国家趋势（2000-2022 年）[互联网].FAOSTAT 分析简报系列，第 94 号。罗马：粮农组织；2024 年[引用日期：2025 年 4 月 5 日]。可从以下链接获取：https://openknowledge.fao.org/handle/20.500.14283/cd3167en.

Yang X, Liu Y, Bezama A, Thrän D. Agricultural carbon emission efficiency and agricultural practices: implications for balancing carbon emissions reduction and agricultural productivity increment. Environ Dev. 2024;50:101004. https://doi.org/10.1016/j.envdev.2024.101004
杨 X，刘 Y，贝扎马 A，特兰 D. 农业碳排放效率与农业生产方式：平衡碳排放减少与农业生产力提升的启示. 环境与发展. 2024;50:101004. https://doi.org/10.1016/j.envdev.2024.101004.

National Development and Reform Commission of China. 2020 China's greenhouse gas emissions report [Internet]. 2021 [cited 2025 Apr 5]. Available from: https://www.ndrc.gov.cn/xxgk/jd/wsdwhfz/202106/t20210609_1282862.html
中国国家发展和改革委员会. 2020 年中国温室气体排放报告 [互联网]. 2021 [引用日期: 2025 年 4 月 5 日]. 获取自: https://www.ndrc.gov.cn/xxgk/jd/wsdwhfz/202106/t20210609_1282862.html.

Wei Z, Wei K, Liu J, Zhou Y. The relationship between agricultural and animal husbandry economic development and carbon emissions in Henan Province, the analysis of factors affecting carbon emissions, and carbon emissions prediction. Mar Pollut Bull. 2023;193:115134. https://doi.org/10.1016/j.marpolbul.2023.115134
魏志, 魏凯, 刘杰, 周勇. 河南省农业与畜牧业经济发展与碳排放的关系、影响碳排放的因素分析及碳排放预测. 海洋污染通报. 2023;193:115134. https://doi.org/10.1016/j.marpolbul.2023.115134.

Li J, Li S, Liu Q, Ding J. Agricultural carbon emission efficiency evaluation and influencing factors in Zhejiang province, China. Front Environ Sci. 2022;10:1005251. https://doi.org/10.3389/fenvs.2022.1005251
李杰, 李松, 刘青, 丁建. 中国浙江省农业碳排放效率评价及影响因素研究. 环境科学前沿. 2022;10:1005251. https://doi.org/10.3389/fenvs.2022.1005251.

Liu Y, Gao Y. Measurement and impactor analysis of agricultural carbon emission performance in Changjiang economic corridor. Alex Eng J. 2022;61:873-881. https://doi.org/10.1016/j.aej.2021.04.083
刘燕, 高燕. 长江经济带农业碳排放绩效测量与影响因素分析. 亚历克斯工程期刊. 2022;61:873-881. https://doi.org/10.1016/j.aej.2021.04.083

Shan T, Xia Y, Hu C, Zhang S, Zhang J, Xiao Y, Dan F. Analysis of regional agricultural carbon emission efficiency and influencing factors: Case study of Hubei Province in China. PLOS ONE [Internet]. 2022;17(4). https://doi.org/10.1371/journal.pone.0266172
山 T，夏 Y，胡 C，张 S，张 J，肖 Y，丹 F. 区域农业碳排放效率及影响因素分析：以中国湖北省为例. PLOS ONE [互联网]. 2022;17(4). https://doi.org/10.1371/journal.pone.0266172

Xiong C, Chen S, Xu L. Driving factors analysis of agricultural carbon emissions based on extended STIRPAT model of Jiangsu Province, China. Growth and Change [Internet]. 2020;51(3):1401-1416. https://doi.org/10.1111/grow.12384
熊超, 陈松, 徐丽. 基于扩展 STIRPAT 模型的江苏省农业碳排放驱动因素分析. 增长与变革 [互联网]. 2020;51(3):1401-1416. https://doi.org/10.1111/grow.12384

Wen C, Zheng J, Hu B, Lin Q. Study on the Spatiotemporal Evolution and Influencing Factors of Agricultural Carbon Emissions in the Counties of Zhejiang Province. International Journal of Environmental Research and Public Health [Internet]. 2023;20(2):189. https://doi.org/10.3390/ijerph20010189
文晨, 郑杰, 胡波, 林青. 浙江省各县农业碳排放的时空演变及影响因素研究. 环境与公共卫生国际期刊 [互联网]. 2023;20(2):189. https://doi.org/10.3390/ijerph20010189

Sun X, Guang S, Cao J, Zhu F, Liu J, Sriboonchitta S. Effect of Agricultural Production Trusteeship on Agricultural Carbon Emission Reduction. Agriculture [Internet]. 2023;13(7):1457. https://doi.org/10.3390/agriculture13071457
孙 X，广 S，曹 J，朱 F，刘 J，斯里邦奇塔 S. 农业生产托管对农业碳排放减排的影响. 农业 [互联网]. 2023;13(7):1457. https://doi.org/10.3390/agriculture13071457

Wang H, Wu Y, Zhu N. Measurement and spatio-temporal heterogeneity analysis of coupling coordination between development of digital economy and agricultural carbon emission performance. PLOS ONE [Internet]. 2024. https://doi.org/10.1371/journal.pone.0305231
王浩, 吴勇, 朱宁. 数字经济与农业碳排放绩效耦合协调的测量及时空异质性分析. PLOS ONE [互联网]. 2024. https://doi.org/10.1371/journal.pone.0305231

Liu X, Ye Y, Ge D, Wang Z, Liu B. Study on the Evolution and Trends of Agricultural Carbon Emission Intensity and Agricultural Economic Development Levels—Evidence from Jiangxi Province. Sustainability [Internet]. 2022;14(21):14265. https://doi.org/10.3390/su142114265
刘 X，叶 Y，葛 D，王 Z，刘 B. 农业碳排放强度与农业经济发展水平演变及趋势研究——以江西省为例. 可持续发展 [互联网]. 2022;14(21):14265. https://doi.org/10.3390/su142114265

Liang L, Wang H, Huai H, Tang X. Study of the Decoupling Patterns between Agricultural Development and Agricultural Carbon Emissions in Beijing Tianjin Hebei Region from 2000 to 2020. Land [Internet]. 2024;13(6):839. https://doi.org/10.3390/land13060839
梁丽, 王红, 怀红, 唐晓. 2000-2020 年京津冀地区农业发展与农业碳排放解耦模式研究. 土地 [互联网]. 2024;13(6):839. https://doi.org/10.3390/land13060839

Wu Y, Chen D, Luo M, Gao F, Li Z. The decoupling effect between net agricultural carbon emissions and economic growth based on LCA. Environment, Development and Sustainability [Internet]. 2023. https://doi.org/10.1007/s10668-023-04236-3
吴宇，陈东，罗敏，高飞，李志. 基于生命周期评估的农业净碳排放与经济增长的脱钩效应. 环境、发展与可持续性 [互联网]. 2023. https://doi.org/10.1007/s10668-023-04236-3

Liu A, Yang S. Study on the spatio-temporal coupling and drivers of agricultural carbon emission efficiency and food security. Frontiers in Environmental Science [Internet]. 2024;12:1503733. https://doi.org/10.3389/fenvs.2024.1503733
刘安, 杨松. 农业碳排放效率与粮食安全时空耦合及其驱动因素研究. 环境科学前沿 [互联网]. 2024;12:1503733. https://doi.org/10.3389/fenvs.2024.1503733

Cao Y, Ji X, Yao J, Xu N, Chen M, Yang X, Liu Z, Li Z, Mo F. Research on Coupling Coordination of Agricultural Carbon Emission Efficiency and Food Security in Hebei Province, China. Sustainability [Internet]. 2024;16(13):5306. https://doi.org/10.3390/su16135306
曹宇，季晓，姚军，徐宁，陈明，杨晓，刘志，李志，莫峰. 中国河北省农业碳排放效率与粮食安全耦合协调研究. 可持续发展 [互联网]. 2024;16(13):5306.https://doi.org/10.3390/su16135306

Jia L, Wang M, Yang S, Zhang F, Wang Y, Li P, Ma W, Sui S, Liu T, Wang M. Analysis of Agricultural Carbon Emissions and Carbon Sinks in the Yellow River Basin Based on LMDI and Tapio Decoupling Models. Sustainability [Internet]. 2024. https://doi.org/10.3390/su16010468
贾丽, 王敏, 杨松, 张峰, 王勇, 李鹏, 马伟, 隋松, 刘涛, 王敏. 基于 LMDI 和 Tapio 解耦模型分析黄河流域农业碳排放与碳汇. 可持续发展 [互联网]. 2024. https://doi.org/10.3390/su16010468

Shu Q, Su Y, Li H, Li F, Zhao Y, Du C. Study on the spatial structure and drivers of agricultural carbon emission efficiency in Belt and Road Initiative countries. Sustainability [Internet]. 2023;15(13):10720. https://doi.org/10.3390/su151310720
舒青，苏燕，李辉，李峰，赵燕，杜春. 一带一路沿线国家农业碳排放效率的空间结构与驱动因素研究. 可持续发展 [互联网]. 2023;15(13):10720. https://doi.org/10.3390/su151310720

Deng L, Yuan S, Bai P, Li H. Assessment of Agricultural Carbon Emissions in Xinjiang Based on Machine Learning Algorithms and Analysis of Driving Factors. Chinese Journal of Eco-Agriculture [Internet]. 2023;31(2):265-279. https://doi.org/10.12357/cjea.20220501
邓丽, 袁松, 白鹏, 李辉. 基于机器学习算法的新疆农业碳排放评估及驱动因素分析. 中国生态农业 [互联网]. 2023;31(2):265-279. https://doi.org/10.12357/cjea.20220501

Bai Z, Zhang N, Winiwarter W, Ma L. Decline in carbon emission intensity of global agriculture has stagnated recently. Proceedings of the National Academy of Sciences of the United States of America [Internet]. 2024;121(34). https://doi.org/10.1073/pnas.2317725121
白志，张宁，温尼瓦特，马立. 全球农业碳排放强度下降近期趋于停滞. 美国国家科学院院刊 [互联网]. 2024;121(34). https://doi.org/10.1073/pnas.2317725121

Anhui Provincial Bureau of Statistics. Anhui Statistical Yearbook 2022 [Internet]. 2022 [Accessed 2025.4.5]. http://tjj.ah.gov.cn/oldfiles/tjj/tjjweb/tjnj/2022/index.htm
安徽省统计局. 《安徽统计年鉴 2022》[互联网]. 2022 [2025 年 4 月 5 日访问]. http://tjj.ah.gov.cn/oldfiles/tjj/tjjweb/tjnj/2022/index.htm

Sun P, Zhang Q, Cheng C, Singh VP, Shi P. ENSO-induced drought hazards and wet spells and related agricultural losses across Anhui province, China. Natural Hazards [Internet]. 2017;89:963-983. https://doi.org/10.1007/s11069-017-3002-4
孙平，张强，程超，辛格·维杰普拉萨德，石平. 厄尔尼诺-南方涛动引发的干旱灾害与湿润期及其对安徽省农业损失的影响. 自然灾害 [互联网]. 2017;89:963-983. https://doi.org/10.1007/s11069-017-3002-4

Cui Y, Zhou Y, Jin J, Jiang S, Wu C, Ning S. Spatiotemporal characteristics and obstacle factors identification of agricultural drought disaster risk: A case study across Anhui Province, China. Agricultural Water Management [Internet]. https://doi.org/10.1016/j.agwat.2023.108554
崔宇，周宇，金杰，江松，吴晨，宁松. 农业干旱灾害风险的时空特征及影响因素识别：以中国安徽省为例. 农业水管理 [互联网]. https://doi.org/10.1016/j.agwat.2023.108554

Dong H, Zhou Y, Zhang Y, Chen T, Sun Z, Hou X, Liu Y. Measurement and temporal and spatial characteristics of agricultural eco-efficiency under climate change: a case study of Anhui, China. Frontiers in Sustainable Food Systems [Internet]. 2023;7. https://doi.org/10.3389/fsufs.2023.1255235
董红, 周勇, 张勇, 陈涛, 孙志, 侯晓, 刘勇. 气候变化背景下农业生态效率的测量及其时空特征: 以中国安徽省为例. 可持续食品系统前沿 [互联网]. 2023;7. https://doi.org/10.3389/fsufs.2023.1255235

Dubey A, Lal R. Carbon footprint and sustainability of agricultural production systems in Punjab, India, and Ohio, USA. Journal of Crop Improvement [Internet]. 2009;23(4):332-350. https://doi.org/10.1080/15427520902969906
杜贝 A, 拉尔 R. 印度旁遮普邦和美国俄亥俄州农业生产系统的碳足迹与可持续性. 作物改良杂志 [互联网]. 2009;23(4):332-350. https://doi.org/10.1080/15427520902969906

Zhao Y, Zhu W, Wei P, Fang P, Zhang X, Yan N, Liu W, Zhao H, Wu Q. Classification of Zambian grasslands using random forest feature importance selection during the optimal phenological period. Ecological Indicators [Internet]. 2022;(7):1-10. https://doi.org/10.1016/j.ecolind.2021.108529
赵宇，朱伟，魏鹏，方鹏，张晓，严宁，刘伟，赵浩，吴强. 基于随机森林特征重要性选择的赞比亚草地分类研究（最佳物候期）. 生态指标 [互联网]. 2022;(7):1-10.https://doi.org/10.1016/j.ecolind.2021.108529

Qi Y, Liu H, Zhao J, Zhang S, Zhang X, Zhang W, Wang Y, Xu J, Li J, Ding Y. Trends and driving forces of agricultural carbon emissions: A case study of Anhui, China. PLOS ONE [Internet]. 2024. https://doi.org/10.1371/journal.pone.0292523
齐宇，刘红，赵杰，张松，张晓，张伟，王勇，徐建，李军，丁勇. 农业碳排放的趋势与驱动因素：以中国安徽省为例. PLOS ONE [互联网]. 2024. https://doi.org/10.1371/journal.pone.0292523

Xu J. Study on spatiotemporal distribution characteristics and driving factors of carbon emission in Anhui Province. Scientific Reports [Internet]. 2023. https://doi.org/10.1038/s41598-023-41507-5
徐杰. 安徽省碳排放时空分布特征及驱动因素研究. 科学报告 [互联网]. 2023. https://doi.org/10.1038/s41598-023-41507-5

Peng ZH, Tian JH, Chen BH, Zhou HB, Bi H, He MX, Li MR, Zheng XY, Wang YW, Chong T, Li ZL. Development of machine learning prognostic models for overall survival of prostate cancer patients with lymph node-positive. Scientific Reports [Internet]. 2023;13(1):1-12. https://doi.org/10.1038/s41598-023-45804-x
彭志，田建华，陈柏华，周海波，毕红，何明，李明瑞，郑晓燕，王永伟，钟涛，李志良。淋巴结阳性前列腺癌患者总体生存率的机器学习预测模型开发. 科学报告 [互联网]. 2023;13(1):1-12. https://doi.org/10.1038/s41598-023-45804-x

Zheng Q, Wang Z, Zhou J, Lu J. Shap-CAM: Visual Explanations for Convolutional Neural Networks based on Shapley Value. arXiv - CS - Computer Vision and Pattern Recognition [Internet]. 2022. https://arxiv.org/abs/2208.03608
郑庆, 王志, 周杰, 卢杰. Shap-CAM: 基于夏普利值的卷积神经网络可视化解释. arXiv - 计算机科学 - 计算机视觉与模式识别 [互联网]. 2022. https://arxiv.org/abs/2208.03608

Lee YG, Oh JY, Kim DS, Kim GB. SHAP Value-Based Feature Importance Analysis for Short-Term Load Forecasting. Journal of Electrical Engineering & Technology [Internet]. 2022. https://doi.org/10.1007/s42835-022-01161-9
李永圭, 哦志宇, 金大善, 金光秉. 基于 SHAP 值的特征重要性分析在短期负荷预测中的应用. 电气工程与技术期刊 [互联网]. 2022. https://doi.org/10.1007/s42835-022-01161-9

Atzmueller M, Furnkranz J, Kliegr T, Schmid U. Explainable and interpretable machine learning and data mining. Data Mining and Knowledge Discovery [Internet]. 2024. https://doi.org/10.1007/s10618-024-01041-y
阿茨穆勒 M, 弗恩克兰茨 J, 克莱格 T, 施密德 U. 可解释和可解释的机器学习与数据挖掘. 数据挖掘与知识发现 [互联网]. 2024. https://doi.org/10.1007/s10618-024-01041-y

Hu JJ, Guo HC, Xue YY, Gao MT, Zhang SP, Tsang YF, et al. Using a mixture of microalgae, biochar, and organic manure to increase the capacity of soil to act as carbon sink. J Soils Sediments [Internet]. 2019 May 21;19(7):2971-82. https://doi.org/10.1007/s11368-019-02337-z
胡 JJ，郭 HC，薛 YY，高 MT，张 SP，曾 YF 等. 采用微藻、生物炭和有机肥料混合物提高土壤作为碳汇的容量. 土壤与沉积物 [互联网]. 2019 年 5 月 21 日;19(7):2971-82.https://doi.org/10.1007/s11368-019-02337-z.

Lei J. Efficient strategies on supply chain network optimization for industrial carbon emission reduction. arXiv [Internet]. 2024. https://doi.org/10.48550/arXiv.2404.16863
雷杰. 工业碳排放减排的供应链网络优化高效策略. arXiv [互联网]. 2024. https://doi.org/10.48550/arXiv.2404.16863

He Y, Wang Y, Quan C. Coupled Coordination and Drivers of Green Technology Innovation and Carbon Emission Efficiency. Sustainability [Internet]. 2024;16(7):2710. https://doi.org/10.3390/su16072710
何宇，王宇，全春. 绿色技术创新与碳排放效率的耦合协调及其驱动因素. 可持续发展 [互联网]. 2024;16(7):2710. https://doi.org/10.3390/su16072710

Wang SJ, Kong W, Ren L, Zhi DD. Research on misuses and modification of coupling coordination degree model in China. Journal of Natural Resources [Internet]. 2021;36(3):793. https://doi.org/10.31497/zrzyxb.20210319
王思杰, 孔伟, 任亮, 智德德. 中国耦合协调度模型滥用与改进研究. 自然资源学报 [互联网]. 2021;36(3):793. https://doi.org/10.31497/zrzyxb.20210319

Gou Y, Xing Q, Yang Z. Matching degree evaluation between new urbanization and carbon emission system in China: a case study of Anhui Province in China. Scientific Reports [Internet]. 2023. https://doi.org/10.1038/s41598-023-38971-4
郭宇，邢庆，杨志. 中国新城镇化与碳排放体系匹配度评估：以安徽省为例. 科学报告 [互联网]. 2023. https://doi.org/10.1038/s41598-023-38971-4

Liu F, Wang C, Zhang Y, Zhou S, Yang Y, Wu X, Hu F, Liu C. Data-driven analysis and evaluation of regional agriculture for high-quality development of Anhui Province in the Yangtze River Delta. Environmental Science and Pollution Research [Internet]. 2021;29:22490-22503. https://doi.org/10.1007/s11356-021-17358-w
刘峰, 王晨, 张勇, 周松, 杨阳, 吴晓, 胡峰, 刘晨. 基于数据驱动的安徽省长江三角洲地区农业发展分析与评价研究. 环境科学与污染研究 [互联网]. 2021;29:22490-22503.https://doi.org/10.1007/s11356-021-17358-w

Shaheen ME, Tawfik W, Mankoula AF, et al. Determination of heavy metal content and pollution indices in the agricultural soils using laser ablation inductively coupled plasma mass spectrometry. Environ Sci Pollut Res Int [Internet]. 2021. https://doi.org/10.1007/s11356-021-13215-y
沙欣 ME，塔维克 W，曼库拉 AF 等. 采用激光剥蚀电感耦合等离子体质谱法测定农田土壤中重金属含量及污染指标. 环境科学与污染研究国际版 [互联网]. 2021. https://doi.org/10.1007/s11356-021-13215-y

Hashem NM, Hassanein EM, Hocquette JF, Gonzalez-Bulnes A, Ahmed FA, Attia YA, Asiry KA. Agro-Livestock Farming System Sustainability during the COVID-19 Era: A Cross-Sectional Study on the Role of Information and Communication Technologies. Sustainability [Internet]. 2021;13(12):6521. https://doi.org/10.3390/su13126521
哈希姆 NM，哈桑恩 EM，霍克特 JF，冈萨雷斯-布尔内斯 A，阿赫迈德 FA，阿蒂亚 YA，阿西里 KA. 农业-畜牧业生产系统在 COVID-19 时代可持续性：信息和通信技术作用的横断面研究. 可持续性 [互联网]. 2021;13(12):6521.https://doi.org/10.3390/su13126521

Liu Y, Ma X, Shu L, Hancke GP, Abu-Mahfouz AM. From Industry 4.0 to Agriculture 4.0: Current Status, Enabling Technologies, and Research Challenges. IEEE Transactions on Industrial Informatics [Internet]. 2020;17(6):4322-4334. https://doi.org/10.1109/TII.2020.3003910
刘宇，马晓，舒亮，汉克 GP，阿布-马福兹 AM.从工业 4.0 到农业 4.0：现状、使能技术及研究挑战. 工业信息学交易[互联网]. 2020;17(6):4322-4334. https://doi.org/10.1109/TII.2020.3003910

Bekti RD, Rahayu A, Sutikno. Maximum likelihood estimation for spatial Durbin model. Journal of Mathematics and Statistics [Internet]. 2013;9(3):169-174. https://doi.org/10.3844/jmssp.2013.169.174
贝克蒂 RD，拉哈尤 A，苏蒂克诺. 空间杜宾模型的最大似然估计. 数学与统计学杂志 [互联网]. 2013;9(3):169-174. https://doi.org/10.3844/jmssp.2013.169.174

Zhou X, Ding D. Factors influencing farmers' willingness and behaviors in organic agriculture development: An empirical analysis based on survey data of farmers in Anhui Province. Sustainability [Internet]. 2022;14(22):14945. https://doi.org/10.3390/su142214945
周 X，丁 D. 有机农业发展中农民意愿与行为的影响因素：基于安徽省农民调查数据的实证分析. 可持续发展 [互联网]. 2022;14(22):14945. https://doi.org/10.3390/su142214945

Zhou Y, He X, Li S. Research on the coordination of fiscal structure and high-quality economic development: An empirical analysis with the example of Anhui province in China. PLOS ONE [Internet]. 2023;18(6). https://doi.org/10.1371/journal.pone.0282761
周宇，何晓，李松. 财政结构与高质量经济发展的协调研究：以中国安徽省为例的实证分析. PLOS ONE [互联网]. 2023;18(6). https://doi.org/10.1371/journal.pone.0282761

Mei Z, Li C, Zhao J, Zixuan L, Chen K, Huang X, Zhao Z. The temporal and spatial evolution characteristics and driving factors of ecosystem service bundles in Anhui Province, China. Land [Internet]. 2024;13(6):736. https://doi.org/10.3390/land13060736
梅紫，李晨，赵杰，李子轩，陈凯，黄晓，赵志. 中国安徽省生态系统服务功能时空演变特征及驱动因素研究. 土地 [互联网]. 2024;13(6):736. https://doi.org/10.3390/land13060736

Sui G, Wang H, Cai S, Cui W. Coupling coordination analysis of resources, economy, and ecology in the Yellow River Basin. Ecological Indicators [Internet]. 2023;156:111133. https://doi.org/10.1016/j.ecolind.2023.111133
隋刚, 王浩, 蔡松, 崔伟. 黄河流域资源、经济与生态耦合协调分析. 生态指标 [互联网]. 2023;156:111133. https://doi.org/10.1016/j.ecolind.2023.111133

Chen Y. Spatial autocorrelation equation based on Moran's index. Scientific Reports [Internet]. 2023;13(1):3946. https://doi.org/10.1038/s41598-023-45947-x
陈 Y. 基于莫兰指数的空间自相关方程. 科学报告 [互联网]. 2023;13(1):3946. https://doi.org/10.1038/s41598-023-45947-x

Anselin L. Local indicators of spatial association-LISA. Geogr Anal. 1995;27(2):93-115. https://doi.org/10.1111/j.1538-4632.1995.tb00338.x
安塞林 L. 空间关联的局部指标——LISA. 地理分析. 1995;27(2):93-115. https://doi.org/10.1111/j.1538-4632.1995.tb00338.x

Wang Q, Zhao M, Luo J, Liu W. Analysis of spatial and temporal variation of agricultural carbon emissions and influencing factors in Anhui Province. Chin Agric Sci Bull. 2024;40(16):104-111. https://doi.org/10.11924/j.issn.1000-6850.casb2023-0785
王青，赵明，罗杰，刘伟. 安徽省农业碳排放的时空变异及其影响因素分析. 中国农业科学通报. 2024;40(16):104-111.https://doi.org/10.11924/j.issn.1000-6850.casb2023-0785.