[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"topic-SHAP":3},{"name":4,"kind":5,"tokens":6,"total":7,"page":8,"page_size":9,"items":10},"SHAP","tag",[4],4,1,100,[11,87,134,182],{"id":12,"title":13,"url":14,"summary":15,"summary_zh":16,"content":17,"source_name":18,"source_url":14,"published_at":19,"category":20,"cover_url":17,"hotness":21,"is_selected":22,"score":23,"score_detail":24,"sources":26,"tags":28,"primary_topic":17,"secondary_topics":32,"classification":33,"search_phrases":62,"slug":65,"view_count":7,"doi":66,"paper":67,"created_at":86},3798,"Debris Flow Susceptibility Mapping Using Explainable Deep Learning in Eastern Hindukush Pakistan","https:\u002F\u002Fdoi.org\u002F10.1002\u002Fesp.70419","ABSTRACT Debris flows are among the most destructive geomorphological hazards in mountainous regions, causing damage to human settlements, infrastructure and ecosystems. The Eastern Hindu Kush region of Pakistan is particularly vulnerable due to steep topography, intense monsoonal precipitation, active tectonics and increasing anthropogenic disturbance. This study develops an integrated debris flow susceptibility assessment framework by combining convolutional neural networks (CNN) with SHapley Additive exPlanations (SHAP). A total of 13 conditioning factors representing topographic, hydrological, geological, environmental and anthropogenic characteristics were prepared using GIS and remote sensing datasets at a spatial resolution of 30 × 30 m. The CNN model demonstrated strong predictive performance, achieving an area under the curve (AUC) of 0.94, overall accuracy (ACC) of 0.86, precision of 0.84, recall of 0.88, an F1‐score of 0.86 and a Kappa coefficient of 0.71. The resulting susceptibility map identified approximately 26.3% of the study area as high to very high susceptibility zones, predominantly concentrated along steep river valleys and highly dissected mountainous terrain. The susceptibility map was further validated by debris flow density analysis, which showed hazard density increasing from 0 events\u002Fkm 2 in very low zones to 0.048 events\u002Fkm 2 in very high zones, with most historical debris flow events concentrated within the very high susceptibility class. SHAP analysis identified slope, NDVI, rainfall, distance to roads and distance to streams as the most influential conditioning factors controlling debris flow occurrence. The proposed CNN–SHAP framework provides a robust and interpretable approach for regional‐scale debris flow susceptibility assessment, with applicability to similar mountainous environments. These findings provide valuable information for hazard mitigation, infrastructure planning and land‐use management in data‐scarce mountain regions.","泥石流是山区最具破坏性的地貌灾害之一，会对人类聚居地、基础设施和生态系统造成损害。巴基斯坦东兴都库什地区由于地形陡峭、季风降水强烈、构造活动活跃以及人为干扰日益加剧，尤为脆弱。本研究通过将卷积神经网络（CNN）与SHapley加法解释（SHAP）相结合，构建了一个综合性的泥石流易发性评估框架。利用GIS和遥感数据集，以30×30 m的空间分辨率制备了代表地形、水文、地质、环境和人为特征的13个 conditioning factors（孕灾因子）。CNN模型表现出较强的预测性能，曲线下面积（AUC）达到0.94，总体精度（ACC）为0.86，精确率为0.84，召回率为0.88，F1分数为0.86，Kappa系数为0.71。所得易发性图识别出约26.3%的研究区域为高至极高易发性区，主要集中于陡峭的河谷和高度切割的山地地形。易发性图进一步通过泥石流密度分析得到验证，结果显示灾害密度从极低区的0 events\u002Fkm²增加到极高区的0.048 events\u002Fkm²，且大多数历史泥石流事件集中在极高易发性类别内。SHAP分析识别出坡度、NDVI、降雨量、距道路距离和距河流距离是控制泥石流发生的最具影响力的孕灾因子。所提出的CNN–SHAP框架为区域尺度泥石流易发性评估提供了一种稳健且可解释的方法，适用于类似的山区环境。这些发现为数据稀缺山区的灾害减缓、基础设施规划和土地利用管理提供了有价值的信息。",null,"Earth Surface Processes and Landforms","2026-09-27T00:00:00Z","论文",10,false,0,{"impact":23,"substance":23,"depth":23,"authority":23,"freshness":23,"relevant":23,"comment":25},"泥石流易发性遥感与深度学习研究，属地质灾害领域，与三农、农业信息化无直接关联，不建议进入每日精选。",[27],{"name":18,"url":14},[29,4,30,31],"深度学习","遥感监测","地质灾害",[],{"status":34,"locked":22,"version":35,"method":36,"evidence":37},"needs_review","2026-10-22.2","rules",{"candidates":38,"agricultural_evidence":22,"reason":61},[39,46,54],{"code":40,"score":21,"supported":41,"title":42,"abstract":44,"tags":45},"I2",true,[43],"deep learning",[],[4,29],{"code":47,"score":48,"supported":41,"title":49,"abstract":50,"tags":53},"I3",7,[],[51,52],"遥感","remote sensing",[30],{"code":55,"score":56,"supported":41,"title":57,"abstract":58,"tags":60},"I4",3,[],[59],"GIS",[],"农业相关性需要复核",[63,64],"东兴都库什 泥石流 易发性","CNN SHAP 泥石流","东兴都库什泥石流易发性-3798","10.1002\u002Fesp.70419",{"doi":66,"openalex_id":68,"authors":69,"venue":18,"cited_by_count":23,"oa_url":78,"card":79,"direction":83,"ingested_from":85},"W7214538761",[70,73,75],{"name":71,"orcid":72},"Kashif Ullah","https:\u002F\u002Forcid.org\u002F0000-0003-2880-0977",{"name":74,"orcid":17},"Xi Chen",{"name":76,"orcid":77},"Tie Liu","https:\u002F\u002Forcid.org\u002F0000-0002-6879-4818","https:\u002F\u002Fonlinelibrary.wiley.com\u002Fdoi\u002Fpdfdirect\u002F10.1002\u002Fesp.70419",{"tldr":80,"method":81,"finding":82,"direction":83,"opportunity":84},"用CNN结合SHAP绘制巴基斯坦东兴都库什泥石流易发性图并解释关键因子。","CNN模型+SHAP，基于GIS与遥感13个30米条件因子。","高易发区占26.3%，坡度、NDVI、降雨、道路与河流距离最关键。","农业遥感与作物表型","可将CNN-SHAP框架迁移至山区农业灾害风险与耕地安全评估，弥补数据稀缺区研究空白。","openalex","2026-09-29T23:30:30.418175Z",{"id":88,"title":89,"url":90,"summary":91,"summary_zh":17,"content":92,"source_name":93,"source_url":17,"published_at":94,"category":20,"cover_url":17,"hotness":21,"is_selected":22,"score":95,"score_detail":96,"sources":103,"tags":105,"primary_topic":109,"secondary_topics":114,"classification":115,"search_phrases":129,"slug":132,"view_count":56,"doi":17,"paper":17,"created_at":133},684,"数农融合发展水平的测度、评价及影响因素分析:基于可解释机器学习模型(统计与决策 2026年第8期)","https:\u002F\u002Fguoyanwang.clcn.net.cn\u002Fwebui\u002Fdocview.aspx?chnid=3630&docid=8274913&leafid=14100","为破解当前我国数农融合不充分与区域发展不平衡难题,文章基于企业工商登记数据,从数农融合规模与渗透率两个维度构建数农融合发展水平评价指标体系,采用CRITIC法测度我国2000-2023年31个省份的数农融合发展水平,运用核密度估计、Dagum基尼系数、随机森林模型、SHAP方法等分析数农融合发展水平的区域差异及关键影响因素。研究发现我国数农融合水平持续上升且增速逐步加快,东部地区领先、西部地区相对偏低。邮政业务总量、农村人力资本、农林牧渔业总产值及财政涉农支出占比对数农融合发展水平具有显著正向影响。","**[摘要]**为破解当前我国数农融合不充分与区域发展不平衡的难题，助力农业高质量发展，文章基于企业工商登记数据，从数农融合规模与渗透率两个维度构建数农融合发展水平评价指标体系，采用CRITIC法测度我国2000—2023年31个省份的数农融合发展水平，运用核密度估计、Dagum基尼系数、随机森林模型、SHAP方法等分析数农融合发展水平的区域差异及关键影响因素。研究结果表明：我国数农融合水平持续上升且增速逐步加快；区域层面，东部地区的数农融合发展水平领先，西部地区相对偏低，四大地区均保持增长态势但增速分化明显；核密度曲线由单峰分布向多峰分布演变；数农融合水平的总体差异逐渐缩小，区域间差异是总体差异的主要来源，东部地区的区域内差异最大，东部与东北地区的区域间差异最大。影响因素分析结果显示，邮政业务总量、农村人力资本、农林牧渔业总产值及财政涉农支出占比对数农融合发展水平具有显著的正向影响；而农村投递线路长度与农村金融机构营业网点就业人数则在一定程度上负向影响数农融合发展水平。\n\n**[关键字]**数农融合发展水平，CRITIC法，机器学习，SHAP可解释性方法","新疆财经大学经济学院 潘癸邑 陈闻君 2026-08-14","2026-08-17T00:00:00Z",82,{"impact":97,"substance":98,"depth":99,"authority":100,"freshness":101,"relevant":8,"comment":102},18,22,20,14,8,"基于机器学习模型对数农融合进行测度与影响因素分析，方法新颖，数据详实，对农业信息化研究有较高参考价值。",[104],{"name":93,"url":90},[106,4,107,108],"机器学习","数农融合","面板数据",{"code":40,"slug":110,"name":111,"path":112,"parent_code":113},"agricultural-ai","农业AI与大模型","\u002Ftopic\u002Fdigital-agriculture\u002Fagricultural-ai","I",[],{"status":116,"locked":22,"version":35,"method":36,"evidence":117},"classified",{"candidates":118,"agricultural_evidence":41},[119,123],{"code":40,"score":21,"supported":41,"title":120,"abstract":121,"tags":122},[106],[],[4,106],{"code":124,"score":56,"supported":41,"title":125,"abstract":126,"tags":128},"D",[],[127],"渔业",[],[130,131],"数农融合 机器学习 面板数据 SHAP","数农融合 机器学习","数农融合机器学习面板数据SHAP-684","2026-08-17T23:54:56.281496Z",{"id":135,"title":136,"url":137,"summary":138,"summary_zh":17,"content":139,"source_name":140,"source_url":17,"published_at":141,"category":20,"cover_url":17,"hotness":21,"is_selected":22,"score":142,"score_detail":143,"sources":147,"tags":149,"primary_topic":153,"secondary_topics":157,"classification":158,"search_phrases":176,"slug":179,"view_count":7,"doi":180,"paper":17,"created_at":181},469,"Portable Hyperspectral Imaging Combined with Interpretable PLS-DA-SHAP for Geographical Origin Traceability and Authenticity Discrimination of Poria cocos (Fushen)","https:\u002F\u002Fwww.frontiersin.org\u002Fjournals\u002Fsustainable-food-systems\u002Farticles\u002F10.3389\u002Ffsufs.2026.1891249\u002Fabstract","本文提出便携式高光谱成像结合可解释PLS-DA-SHAP方法,用于茯苓(Fushen,Poria cocos)的地理来源溯源与真实性鉴别。研究展示了如何将机器学习与SHAP可解释性相结合,应用于中药材产地溯源。原文标题:Portable hyperspectral imaging combined with interpretable PLS-DA-SHAP for geographical origin traceability and authenticity discrimination of Poria cocos (Schw.) Wolf (Fushen),作者:Xing W、Zhang Y、Wei X、Nie S、Cheng X、Cai M、Yang Q、Wang Y,发表于Frontiers in Sustainable Food Systems 2026年第10卷doi:10.3389\u002Ffsufs.2026.1891249,2026年7月15日接收,8月7日发表。本研究为基于光谱成像的农产品\u002F中药材产地溯源提供了可解释AI方法范式。","## Abstract\n\n**Introduction:**\n\n_Poria cocos_ (Schw.) Wolf (_Fushen_) is a fungus that is edible and has medicinal value. The quality of the fungus varies with geographical origin and its commercial substitution by Chuanshen.\n\n**Methods:**\n\nThis study developed a rapid, non-destructive approach for the geographical origin traceability and authenticity discrimination of _Fushen_ using portable hyperspectral imaging (HSI; 400–1,000 nm) combined with interpretable machine learning. The discriminatory potential of four _Fushen_ regions was assessed. Partial least-squares discriminant analysis (PLS-DA), support vector machine, and multilayer perceptron were compared, while the successive projections algorithm and SHapley Additive exPlanations (SHAP) algorithms were applied for feature wavelength selection and model interpretation.\n\n**Results:**\n\nThe PLS-DA-SHAP strategy exhibited the best overall performance. After SHAP-based wavelength selection, PLS-DA combined with first derivative (D1) preprocessing achieved 92.33% accuracy for geographical origin traceability, whereas PLS-DA using raw spectra achieved 100% accuracy for authenticity discrimination. The outer layer of white _Fushen_ (M2) provided the highest regional discrimination accuracy of 100%. Wavelengths of 893 nm, 896 nm, and 881 and 423 nm were identified as key contributors to geographical origin traceability, authenticity discrimination using the All dataset, and authenticity discrimination using M2, respectively.\n\n**Discussion:**\n\nThese findings can support the use of portable HSI as a practical tool for on-site quality assessment of _Fushen_ and related edible and medicinal materials.\n\n## 1 Introduction\n\n_Poria cocos_ (Schw.) Wolf (_Fushen_) is a saprotrophic fungus of the Polyporaceae family. It commonly parasitizes the roots of Pinus species and forms dense white sclerotia that naturally enclose the growing portions of pine roots. This material is commonly referred to as _Fushen_ in China (). _Fushen_ is distributed mainly in Asia, America, Oceania, and parts of Africa, with China serving as the major cultivator (). _Fushen_ is commonly regarded as one of the representative species with both medicinal and edible properties. In addition to recording its therapeutic properties, ancient medical texts have emphasized its nutritional value and suitability for long-term consumption (). Therefore, _Fushen_ has a long history of use in the food industry, with a wide range of products such as porridge, pastries, herbal teas, and tonic soups, which are well accepted by consumers and reflect its edible value (). Modern studies have shown that _Fushen_ is rich in bioactive constituents, including polysaccharides and triterpenoids, and exhibits sleep-improving and anxiolytic effects, indicating its promising potential as a functional food ingredient (; ).\n\n_Fushen_ has a long cultivation history in China and is widely produced in Anhui, Yunnan, Hubei, and Hunan. Its quality is affected by production area conditions (e.g., soil, climate, and cultivation practices), which may lead to variations in key functional components (; ). In addition, counterfeit _Fushen_ products that closely resemble genuine products are present in the market and are difficult to identify visually. These counterfeit products, commonly known as Chuanshen (CS), are produced by artificially inserting pine roots into the sclerotium during the later growth stage of _Poria cocos_ (PC) (; ). In genuine _Fushen_, the lignocellulose of pine roots undergoes biological transformation with the mycelium of PC during development of the sclerotia, resulting in changes in the active components. In CS, the symbiotic period between pine roots and PC is relatively short, contributing to the marked compositional differences from genuine _Fushen_ (). Previous studies have also shown that _Fushen_ blocks can be divided into four regions, namely _Fushen_ wood, the inner layer of white _Fushen_, the outer layer of white _Fushen_, and skin _Fushen_, each with distinct chemical compositions (; ). However, whether genuine and counterfeit products differ across these regions and whether such regional differences can be effectively used for counterfeit discrimination remain unclear. Therefore, a reliable detection method is required to identify the geographical origin, authenticity, and regional characteristics of _Fushen_, thereby supporting product quality control.\n\nTraditional food detection methods mainly rely on analytical techniques such as liquid chromatography–mass spectrometry (; ), gas chromatography–mass spectrometry (), and ultra-high-performance liquid chromatography–quadrupole time-of-flight mass spectrometry (; ; ). Although these methods provide high analytical accuracy, they require complex procedures, long analysis times, high costs, and destructive pretreatment, which limit their suitability for rapid market screening (; ). As a rapid analytical technique, near-infrared spectroscopy still faces limitations related to low sensitivity, dependence on chemometric models, and insufficient model interpretability ().\n\nHyperspectral imaging (HSI) integrates spectral analysis with imaging acquisition and has become an effective, rapid, and non-destructive detection technique. Without requiring complex chemical pretreatment, HSI can simultaneously capture the spatial and spectral characteristics of samples. Moreover, the integration of HSI with machine learning algorithms has further improved both accuracy and efficiency, and this combined approach has now been widely adopted in food safety applications (; ; ). However, conventional HSI systems are generally bulky and require controlled operating environments, limiting their suitability for portable deployment and on-site detection. Conversely, portable spectral instruments are compact and more suitable for field-based applications. In recent years, devices such as SPECIM IQ have been successfully used for rapid and non-destructive on-site analysis, including saffron origin identification () and quantitative detection of rice adulteration (), demonstrating their practical value in food quality evaluation. Nevertheless, portable hyperspectral datasets are affected by high dimensionality, limited computational efficiency, and insufficient model interpretability. To address these constraints, the successive projections algorithm (SPA) and SHapley Additive exPlanations (SHAP) can be employed for feature wavelength selection. By removing redundant variables and retaining informative wavelengths, these methods enable dimensionality reduction and improve model efficiency. In addition, SHAP can be integrated with different machine learning models to quantify the marginal contribution of each wavelength and generate global importance rankings, thereby enhancing the interpretability of model decisions (; ).\n\nOn this basis, portable HSI combined with interpretable machine learning was used in this study to support the geographical origin traceability and authenticity discrimination of _Fushen_, with the aim of providing a technical reference for rapid and non-destructive detection of functional food materials. The specific objectives were as follows: (1) to acquire HSI data from _Fushen_ samples collected from different geographical origins, as well as from genuine and counterfeit products, and to divide the genuine and counterfeit samples into four regions; (2) to compare the classification performance of three algorithms, including support vector machine (SVM), partial least-squares discriminant analysis (PLS-DA), and multilayer perceptron (MLP), under six preprocessing methods; and (3) to use SPA and SHAP for characteristic wavelength selection, dimensionality reduction, classification efficiency improvement, and interpretation of the relationship between wavelengths and samples. The methodological workflow is illustrated in [Figure 1](https:\u002F\u002Fwww.frontiersin.org\u002Fjournals\u002Fsustainable-food-systems\u002Farticles\u002F10.3389\u002Ffsufs.2026.1891249\u002Fabstract#fig1). For abbreviations see [Supplementary Table S1](https:\u002F\u002Fwww.frontiersin.org\u002Fjournals\u002Fsustainable-food-systems\u002Farticles\u002F10.3389\u002Ffsufs.2026.1891249\u002Fabstract#sec24).\n\nFigure 1\n\n## 2 Materials and methods\n\n### 2.1 Sample collection\n\n_Fushen_ samples were collected from five authentic production areas in China, namely Jinzhai, Anhui (AJ); Yuexi, Anhui (AY); Hubei (HB); Hunan (HN); and Yunnan (YN). Four independent biological replicates were obtained from each production area, with each replicate derived from different farmer fields. Five blocks were randomly selected from each replicate, yielding 20 blocks per production area and a total of 100 authentic _Fushen_ blocks across the five areas. The counterfeit CS samples were collected from Jinzhai County, Anhui Province, China, with six independent biological replicates and five blocks randomly selected from each replicate, totaling 30 CS blocks. All samples were authenticated as genuine _Fushen_ or counterfeit CS by Associate Professor Yang Qingshan from Anhui University of Chinese Medicine. After collection, the outer skin was removed, and _Fushen_ blocks of similar size were selected for subsequent analysis. Representative sample images are shown in [Figure 1](https:\u002F\u002Fwww.frontiersin.org\u002Fjournals\u002Fsustainable-food-systems\u002Farticles\u002F10.3389\u002Ffsufs.2026.1891249\u002Fabstract#fig1), Step 1 Samples collection, and detailed sample information is provided in [Supplementary Table S2](https:\u002F\u002Fwww.frontiersin.org\u002Fjournals\u002Fsustainable-food-systems\u002Farticles\u002F10.3389\u002Ffsufs.2026.1891249\u002Fabstract#sec24).\n\n### 2.2 Data collection using the HSI system\n\nHyperspectral images were acquired using a portable hyperspectral camera (SPECIM IQ, SPECIM Spectral Imaging Ltd., Oulu, Finland) by scanning one surface of each _Fushen_ block ([Figure 1](https:\u002F\u002Fwww.frontiersin.org\u002Fjournals\u002Fsustainable-food-systems\u002Farticles\u002F10.3389\u002Ffsufs.2026.1891249\u002Fabstract#fig1), Step 2 Hyperspectral image acquisition). The camera was fitted with a visible–near-infrared lens covering 400–1,000 nm and provided 204 bands, a spatial resolution of 512 × 512 pixels, a spectral resolution of 7 nm, a sampling interval of approximately 2.94 nm, and band-division capability (). A white reference plate with 90% reflectance was used for reflectance correction during image acquisition.\n\nImages were captured while the samples were stationary. To reduce background interference, the samples and white reference plate were placed on a black background, as described in the literature (). The SPECIM IQ instrument was mounted on a tripod, and the lens was positioned vertically approximately 40 cm above the samples. Because all images were collected outdoors under sunlight, the simultaneous mode was adopted to minimize the effect of illumination variation, following previous research (). The integration and recording times were set to 2 ms and 32 s, respectively, to ensure consistent imaging conditions during HSI data collection. Pixel-level reflectance correction was performed using the black-white reference correction formula reported in the literature (). The corrected hyperspectral images were transferred to a computer for subsequent region of interest (ROI) extraction.\n\n### 2.3 Spectral data extraction from ROIs\n\nThe ROIs were manually delineated using Insight 1.3.1 software provided by the instrument manufacturer for subsequent analysis. For geographical origin traceability, ROIs were extracted from the complete regions of samples from five geographical origins (20 blocks per area, 100 blocks in total), with 15 ROIs per block, yielding 300 ROIs per area and 1,500 ROIs overall. For authenticity discrimination (All), ROIs were extracted from the complete regions of 30 authentic and 30 counterfeit CS blocks, with 10 ROIs per block, resulting in 300 ROIs for authentic samples and 300 for counterfeit samples, totaling 600 ROIs.\n\nGiven the differences in bioactive components among the four regions of _Fushen_, and following previous studies and morphological characteristics (), the authentic and counterfeit CS samples were divided from the inside outward into four regions: _Fushen_ wood (the internal host pine root, abbreviated as FSM), the inner layer of white _Fushen_ (the white sclerotial part near the host pine root, abbreviated as M1), the outer layer of white _Fushen_ (the white part near the outer region, abbreviated as M2), and the peripheral sclerotium (the outermost sclerotial part after peeling, abbreviated as M3), since all samples had their outer skin removed at collection. The division is illustrated in [Supplementary Figure S1](https:\u002F\u002Fwww.frontiersin.org\u002Fjournals\u002Fsustainable-food-systems\u002Farticles\u002F10.3389\u002Ffsufs.2026.1891249\u002Fabstract#sec24) and [Figure 1](https:\u002F\u002Fwww.frontiersin.org\u002Fjournals\u002Fsustainable-food-systems\u002Farticles\u002F10.3389\u002Ffsufs.2026.1891249\u002Fabstract#fig1), Step 2 Hyperspectral image acquisition. For each region, 10 ROIs were extracted per block from 30 authentic and 30 counterfeit CS blocks, yielding 300 ROIs per region for each category. Across the four regions, a total of 1,200 ROIs from authentic samples and 1,200 from counterfeit samples were obtained, amounting to 2,400 ROIs in total. These data were used to evaluate the spectral differences and the discriminatory ability of each region in authenticity discrimination.\n\nBased on the above strategy, the mean reflectance within each ROI was extracted as spectral information. Subsequently, the three independent datasets were each randomly divided into training and test sets at an 8:2 ratio.\n\n### 2.4 Spectral preprocessing\n\nSeveral preprocessing methods were applied to the raw spectral data to reduce the background noise and physical interference generated during HSI acquisition. These methods included multiplicative scatter correction (MSC), first derivative (D1), second derivative (D2), standard normal variate (SNV), Savitzky–Golay smoothing (SG), and wavelet transform (Wave).\n\n### 2.5 Machine learning classification model\n\nSVM is a supervised learning algorithm with a strong generalization ability. It separates classes by constructing an optimal hyperplane in the feature space, thereby maximizing the margin between categories. For linearly inseparable samples, kernel functions are used to map the original data into a high-dimensional feature space in which linear separation can be achieved (; ). The model classification performance is mainly controlled by the penalty factor and the kernel parameter. In this study, the model parameters were optimized using a grid search combined with 10-fold cross-validation. The candidate values for the penalty factor were 5, 10, 15, 50, and 100, whereas the kernel parameter was searched over 1e-7, 1e-6, 1e-5, 1e-4, 1e-3, 1e-2, 1e-1, and 1. Three kernel functions, namely linear, polynomial, and radial basis function kernels, were systematically compared to determine the optimal model configuration and improve classification performance.\n\nPLS-DA is a supervised pattern recognition method derived from partial least-squares regression. By extracting latent variables with maximum covariance with class labels, PLS-DA enables dimensionality reduction and classification to be achieved simultaneously while balancing the retention of discriminant information with the mitigation of high-dimensional effects (). Its classification performance depends on the selected number of PLS latent components, defined as n_components. In this study, the number of PLS latent components was searched from 1 to 21, and the optimal value was determined using 10-fold cross-validation.\n\nMLP is an artificial neural network with a relatively simple architecture that can learn from sample data and effectively represent classification features (). In this study, the MLP consisted of two hidden layers, with the number of neurons in each layer ranging from 50 to 500. The optimal parameters were determined by grid search. The rectified linear unit function was used as the activation function for hidden layers. The limited-memory Broyden–Fletcher–Goldfarb–Shanno algorithm was used for optimization. The regularization coefficient was set to 0.0001, and the maximum number of iterations was set to 10,000 to ensure sufficient model convergence.\n\n### 2.6 Feature wavelength selection algorithm\n\nTo address the high dimensionality, collinearity, and redundancy of hyperspectral data, SPA and SHAP were used to extract characteristic bands. Through dimensionality reduction and the removal of irrelevant variables, these methods were applied to simplify the models and improve prediction performance.\n\nSPA is based on vector-space projection. It identifies wavelength subsets with low collinearity and high information content by sequentially projecting wavelength vectors onto the orthogonal complement space of the remaining vectors. The optimal wavelength combination was selected according to the root mean square error of cross-validation, thereby achieving spectral dimensionality reduction (; ). To improve the stability and reliability of the selected wavelengths, 10-fold cross-validation was applied, and automatic scaling was used in the preprocessing parameters.\n\nSHAP is an interpretability method derived from Shapley value theory in cooperative game theory (). Each input feature is regarded as a participant in a cooperative game, and its Shapley value is calculated from the weighted average of its marginal contributions across all possible feature subset combinations. The resulting value quantifies the contribution of each feature to the model prediction. The specific calculation formula is provided in .\n\n### 2.7 Model evaluation and software\n\nA confusion matrix was used for model evaluation, with accuracy, precision, recall, and F1-score as the main metrics for assessing classification performance. The corresponding calculation formula is provided in . All analyses were conducted on a Windows 11 platform equipped with an AMD Ryzen 9 9955HX CPU, 32 GB of memory, and an NVIDIA RTX 5060 Laptop GPU. Data processing and model construction were performed using Python 3.12 (64-bit) and PyCharm Community 2024.1. All scripts were developed by our research group.\n\n## 3 Results\n\n### 3.1 Spectral curve analysis of samples\n\nThe mean raw spectral curves of _Fushen_ samples from different geographical origins are presented in [Figure 2A](https:\u002F\u002Fwww.frontiersin.org\u002Fjournals\u002Fsustainable-food-systems\u002Farticles\u002F10.3389\u002Ffsufs.2026.1891249\u002Fabstract#fig2). Across all origins, the reflectance increased over the 400–900 nm range, reached a maximum at approximately 850–900 nm, and then declined gradually between 900 and 1,000 nm. This pattern indicates an overall spectral trend characterized by an initial increase, followed by a slow decrease. Over the entire wavelength range, the reflectance of the AY and AJ samples was higher than that of the other samples. The HB samples displayed intermediate reflectance, whereas the YN and HN samples had relatively lower reflectance. For genuine _Fushen_, the four regional spectra showed a similar overall pattern, with the reflectance increasing and decreasing ([Figure 2B](https:\u002F\u002Fwww.frontiersin.org\u002Fjournals\u002Fsustainable-food-systems\u002Farticles\u002F10.3389\u002Ffsufs.2026.1891249\u002Fabstract#fig2)). However, an evident regional variation was observed, with the reflectance increasing from the inner to the outer region in the order FSM \u003C M1 \u003C M2 \u003C M3. Notably, M2 exhibited a reflectance profile similar to that of All. For counterfeit _Fushen_, the spectra of M1, M2, and M3 were highly consistent with that of All, whereas FSM showed a different spectral profile ([Figure 2C](https:\u002F\u002Fwww.frontiersin.org\u002Fjournals\u002Fsustainable-food-systems\u002Farticles\u002F10.3389\u002Ffsufs.2026.1891249\u002Fabstract#fig2)). M2 had the highest reflectance among the four counterfeit regions. In both genuine and counterfeit samples, FSM showed low reflectance and similar curve patterns. Although the spectral reflectance varied among s","Frontiers in Sustainable Food Systems 2026-08-07","2026-08-07T00:00:00Z",67,{"impact":144,"substance":99,"depth":97,"authority":144,"freshness":145,"relevant":8,"comment":146},12,5,"研究利用便携式高光谱成像结合可解释机器学习实现茯苓产地溯源与真伪鉴别，方法新颖，数据详实，对中药材质量监管有参考价值，但属细分领域进展，影响范围有限。",[148],{"name":140,"url":137},[4,150,151,152],"高光谱成像","产地溯源","中药材",{"code":47,"slug":154,"name":155,"path":156,"parent_code":113},"remote-sensing","遥感与农情监测","\u002Ftopic\u002Fdigital-agriculture\u002Fremote-sensing",[],{"status":116,"locked":22,"version":35,"method":36,"evidence":159},{"candidates":160,"agricultural_evidence":41},[161,167,172],{"code":47,"score":162,"supported":41,"title":163,"abstract":165,"tags":166},11,[164],"hyperspectral",[164],[],{"code":168,"score":7,"supported":41,"title":169,"abstract":170,"tags":171},"B6",[],[152],[152],{"code":40,"score":7,"supported":41,"title":173,"abstract":174,"tags":175},[],[106],[4],[177,178],"高光谱成像 产地溯源 中药材 SHAP","高光谱成像 产地溯源","高光谱成像产地溯源中药材SHAP-469","10.3389\u002Ffsufs.2026.1891249\u002Fabstract","2026-08-12T23:57:24.335667Z",{"id":183,"title":184,"url":185,"summary":186,"summary_zh":17,"content":187,"source_name":188,"source_url":17,"published_at":189,"category":20,"cover_url":17,"hotness":21,"is_selected":22,"score":190,"score_detail":191,"sources":193,"tags":195,"primary_topic":200,"secondary_topics":206,"classification":208,"search_phrases":230,"slug":233,"view_count":56,"doi":17,"paper":17,"created_at":234},380,"农户数字能力对低碳农业采纳的影响——基于陕黔川10057户猕猴桃种植户的机器学习分析","https:\u002F\u002Fwww.mdpi.com\u002F2071-1050\u002F18\u002F16\u002F8088","基于陕西、贵州、四川三省10057名猕猴桃种植户的微观调查数据,西安建筑科技大学向雯与西北农林大学高建中等于Sustainability发表论文。以农户对低碳农业技术的初始与持续采纳为因变量,构建涵盖数字接入、数字获取、数字处理和数字共享4个维度的数字能力综合评价体系。比较Lasso回归、极端梯度提升、随机森林和反向传播神经网络等多种实证模型,结果显示随机森林模型表现最优。采用SHAP方法定量评估各变量边际贡献,研究发现农户数字能力显著正向促进低碳农业技术的初始与持续采纳,且具有明显的维度异质性:数字接入和获取能力是初始采纳的核心驱动因素,数字处理和共享能力对持续采纳起决定性作用。","## 1. Introduction\n\nIn light of the ongoing decline in atmospheric quality and the momentous deterioration of the ecological environment, the pursuit of low-carbon development has emerged as an imperative response to these challenges. The sustainable development model of low-carbon development, which is characterised by energy conservation, reduced consumption, and pollution reduction, has become a crucial avenue for addressing environmental issues [[1](https:\u002F\u002Fwww.mdpi.com\u002F2071-1050\u002F18\u002F16\u002F8088#B1-sustainability-18-08088),[2](https:\u002F\u002Fwww.mdpi.com\u002F2071-1050\u002F18\u002F16\u002F8088#B2-sustainability-18-08088),[3](https:\u002F\u002Fwww.mdpi.com\u002F2071-1050\u002F18\u002F16\u002F8088#B3-sustainability-18-08088),[4](https:\u002F\u002Fwww.mdpi.com\u002F2071-1050\u002F18\u002F16\u002F8088#B4-sustainability-18-08088)]. In response to the specific conditions of its own development, China has put forth proposals for carbon peaking and carbon neutrality programmes. In 2020, China formally proposed a double carbon goal, committing to the international community to peak its carbon dioxide emissions by 2030 and achieve carbon neutrality by 2060 [[5](https:\u002F\u002Fwww.mdpi.com\u002F2071-1050\u002F18\u002F16\u002F8088#B5-sustainability-18-08088)]. It is widely acknowledged that agriculture represents a significant global source of greenhouse gas emissions. Consequently, it has become evident that farmland has an integral part to play in achieving the double carbon goal [[6](https:\u002F\u002Fwww.mdpi.com\u002F2071-1050\u002F18\u002F16\u002F8088#B6-sustainability-18-08088),[7](https:\u002F\u002Fwww.mdpi.com\u002F2071-1050\u002F18\u002F16\u002F8088#B7-sustainability-18-08088)].\n\nNevertheless, for an extended period, the direct cost of agricultural production has been the sole factor considered, with the cost of resource and environmental losses incurred during the production process remaining unaccounted for. The excessive and inefficient use of agricultural resources, such as pesticides and chemical fertilisers, in agricultural production has a direct impact on the quality and safety of agricultural products, the ecological environment, and human health. The negative externalities of this practice are becoming increasingly apparent [[8](https:\u002F\u002Fwww.mdpi.com\u002F2071-1050\u002F18\u002F16\u002F8088#B8-sustainability-18-08088),[9](https:\u002F\u002Fwww.mdpi.com\u002F2071-1050\u002F18\u002F16\u002F8088#B9-sustainability-18-08088),[10](https:\u002F\u002Fwww.mdpi.com\u002F2071-1050\u002F18\u002F16\u002F8088#B10-sustainability-18-08088)].\n\nOver the past few decades, China’s traditional agricultural production and cultivation patterns have undergone significant expansion, resulting in a considerable accumulation of high carbon emissions and a multitude of ecological damage issues [[11](https:\u002F\u002Fwww.mdpi.com\u002F2071-1050\u002F18\u002F16\u002F8088#B11-sustainability-18-08088),[12](https:\u002F\u002Fwww.mdpi.com\u002F2071-1050\u002F18\u002F16\u002F8088#B12-sustainability-18-08088),[13](https:\u002F\u002Fwww.mdpi.com\u002F2071-1050\u002F18\u002F16\u002F8088#B13-sustainability-18-08088)]. In recent years, the central No. 1 document has consistently emphasised the necessity of promoting environmentally sustainable agricultural practices. This entails accelerating the promotion and application of agricultural input reduction and efficiency technology, integrating water and fertiliser management, and establishing and enhancing the collection and utilisation of straw, agricultural film, pesticide packaging waste, livestock and poultry manure, and other agricultural waste collection and treatment systems. Furthermore, General Secretary Xi Jinping emphasised the necessity of developing ecological and low-carbon agricultural practices, asserting that environmentally sustainable practices are the foundation of modern agriculture.\n\nGreen and low-carbon agriculture has the potential to save energy and reduce emissions, facilitate the sequestration of carbon in an ecological manner, enhance the utilisation of agricultural resources, produce safe and high-quality agricultural products, and provide robust support for the realisation of China’s double carbon target [[14](https:\u002F\u002Fwww.mdpi.com\u002F2071-1050\u002F18\u002F16\u002F8088#B14-sustainability-18-08088),[15](https:\u002F\u002Fwww.mdpi.com\u002F2071-1050\u002F18\u002F16\u002F8088#B15-sustainability-18-08088)]. The foundation of low-carbon agricultural production is the utilisation of cutting-edge low-carbon agricultural technology [[16](https:\u002F\u002Fwww.mdpi.com\u002F2071-1050\u002F18\u002F16\u002F8088#B16-sustainability-18-08088),[17](https:\u002F\u002Fwww.mdpi.com\u002F2071-1050\u002F18\u002F16\u002F8088#B17-sustainability-18-08088)]. It is for this reason that research and development, promotion, and application of low-carbon agricultural technology have become pivotal measures to achieve carbon sequestration and emission reduction in agricultural production. The contemporary understanding of low-carbon agricultural technology is primarily concerned with the minimisation of inputs, including energy, fertilisers, pesticides and water, in agricultural production. This approach ultimately aims to curtail greenhouse gas emissions from agricultural sources. In particular, low-carbon agricultural technology encompasses a range of techniques and practices, including formula fertilisation technology, commercial organic fertiliser application, farm fertiliser application, no-till technology, and water-saving irrigation technology.\n\nThe implementation of the household contract responsibility system in China has resulted in the farming household becoming the primary entity engaged in production and management activities within rural areas, and is thus the most fundamental decision-making unit. The uptake of low-carbon agricultural technology by farmers is contingent upon the comprehensive promotion and implementation of ecological low-carbon agriculture in China [[18](https:\u002F\u002Fwww.mdpi.com\u002F2071-1050\u002F18\u002F16\u002F8088#B18-sustainability-18-08088)]. The decision to adopt technology is a behavioural choice made by an individual, influenced by the opinions of others and based on an evaluation of the advantages offered by the technology and their own suitability for it [[19](https:\u002F\u002Fwww.mdpi.com\u002F2071-1050\u002F18\u002F16\u002F8088#B19-sustainability-18-08088)]. The dissemination and uptake of low-carbon agricultural technology can be conceptualised as a process of technology diffusion [[20](https:\u002F\u002Fwww.mdpi.com\u002F2071-1050\u002F18\u002F16\u002F8088#B20-sustainability-18-08088)]. The process can be divided into two distinct stages: the initial stage concerns farmers making the decision to adopt low-carbon agricultural technology; the subsequent stage pertains to farmers deciding to continue the behaviour after adopting low-carbon agricultural technology [[21](https:\u002F\u002Fwww.mdpi.com\u002F2071-1050\u002F18\u002F16\u002F8088#B21-sustainability-18-08088)]. In accordance with the conventional perspective on economics, a considerable number of scholars have examined the internal and external elements influencing the initial adoption of low-carbon agricultural technology by farmers at both the theoretical and empirical levels. Internal factors include gender, age, health, education, peer effects, ecological awareness, household income, non-farm labour migration, time preference, and risk perception [[22](https:\u002F\u002Fwww.mdpi.com\u002F2071-1050\u002F18\u002F16\u002F8088#B22-sustainability-18-08088),[23](https:\u002F\u002Fwww.mdpi.com\u002F2071-1050\u002F18\u002F16\u002F8088#B23-sustainability-18-08088),[24](https:\u002F\u002Fwww.mdpi.com\u002F2071-1050\u002F18\u002F16\u002F8088#B24-sustainability-18-08088),[25](https:\u002F\u002Fwww.mdpi.com\u002F2071-1050\u002F18\u002F16\u002F8088#B25-sustainability-18-08088),[26](https:\u002F\u002Fwww.mdpi.com\u002F2071-1050\u002F18\u002F16\u002F8088#B26-sustainability-18-08088),[27](https:\u002F\u002Fwww.mdpi.com\u002F2071-1050\u002F18\u002F16\u002F8088#B27-sustainability-18-08088),[28](https:\u002F\u002Fwww.mdpi.com\u002F2071-1050\u002F18\u002F16\u002F8088#B28-sustainability-18-08088)]. The adoption of low-carbon agricultural technology by farmers is influenced by several external factors, including training, social capital, market incentives, government regulation, subsidies, agricultural insurance, participation in cooperatives, and machinery outsourcing services [[29](https:\u002F\u002Fwww.mdpi.com\u002F2071-1050\u002F18\u002F16\u002F8088#B29-sustainability-18-08088),[30](https:\u002F\u002Fwww.mdpi.com\u002F2071-1050\u002F18\u002F16\u002F8088#B30-sustainability-18-08088),[31](https:\u002F\u002Fwww.mdpi.com\u002F2071-1050\u002F18\u002F16\u002F8088#B31-sustainability-18-08088),[32](https:\u002F\u002Fwww.mdpi.com\u002F2071-1050\u002F18\u002F16\u002F8088#B32-sustainability-18-08088),[33](https:\u002F\u002Fwww.mdpi.com\u002F2071-1050\u002F18\u002F16\u002F8088#B33-sustainability-18-08088),[34](https:\u002F\u002Fwww.mdpi.com\u002F2071-1050\u002F18\u002F16\u002F8088#B34-sustainability-18-08088)]. A review of the literature reveals a paucity of studies investigating the sustained adoption of low-carbon agricultural technology by farmers. Many scholars have employed sustained use modelling to examine the influence of various factors, including technology cost and efficacy, technology complexity, government technology promotion, and policy subsidies, on the sustained adoption of low-carbon agricultural technology by farmers [[35](https:\u002F\u002Fwww.mdpi.com\u002F2071-1050\u002F18\u002F16\u002F8088#B35-sustainability-18-08088),[36](https:\u002F\u002Fwww.mdpi.com\u002F2071-1050\u002F18\u002F16\u002F8088#B36-sustainability-18-08088),[37](https:\u002F\u002Fwww.mdpi.com\u002F2071-1050\u002F18\u002F16\u002F8088#B37-sustainability-18-08088)]. A consensus was reached on the fundamental premise that the continued adoption of low-carbon agricultural technology by farmers represents an extension of the technology adoption process. This extension is based on the evaluation of the actual results of technology adoption in relation to individual expectations of the effects of low-carbon agricultural technology. It is frequently observed that the performance of a technology is at its lowest during the initial adoption phase. It is only through continued utilisation of the technology that its intrinsic value can be fully appreciated. Despite the ascendance of green agriculture in China in the context of production and development, a discrepancy persists between the initial and sustained adoption of green and low-carbon agricultural technologies by farmers. Nevertheless, the majority of existing studies focus on the initial decision-making process surrounding the adoption of low-carbon technology, neglecting to consider the dynamic characteristics of technology adoption behaviour across different stages. Furthermore, there is a tendency to conflate initial adoption with sustained adoption.\n\nThe digital economy is particularly adept at optimising the low-carbon agricultural industry landscape [[38](https:\u002F\u002Fwww.mdpi.com\u002F2071-1050\u002F18\u002F16\u002F8088#B38-sustainability-18-08088)]. The advent of the digital economy has given rise to a plethora of novel technologies, industries, and business models. Moreover, the digitisation of knowledge and information has become a primary factor in the production process [[39](https:\u002F\u002Fwww.mdpi.com\u002F2071-1050\u002F18\u002F16\u002F8088#B39-sustainability-18-08088)]. The integration of modern information technology with agriculture, including the internet, the Internet of Things, cloud computing, big data, and artificial intelligence, enables real-time collection of agricultural data, real-time monitoring of the agricultural ecological environment, precise input of agricultural production factors, quantitative decision-making in the agricultural production process, and intelligent regulation and control of agricultural production. These functions collectively constitute a new low-carbon agricultural production model. It has been demonstrated that the digital economy plays a notable role in the implementation of low-carbon agricultural technology by farmers [[40](https:\u002F\u002Fwww.mdpi.com\u002F2071-1050\u002F18\u002F16\u002F8088#B40-sustainability-18-08088)]. Concurrently, the digital economy has considerable expectations regarding the digital capability of farmers. Those with superior digital abilities are more adept at navigating digital technology and are more likely to receive a premium for their skills in the context of the advancement of the digital economy.\n\nThe disparate digital capability of farmers gives rise to variations in the digital divide and the digital dividend. This raises the question of whether digital capability has a significant positive effect on the initial adoption of low-carbon agricultural technology by farmers. Does digital capability also have a significant positive effect on the sustained adoption of low-carbon agricultural technology? Given the established link between initial and sustained adoption of technology, it is pertinent to inquire whether the factors influencing the initial behaviour of farmers in the adoption of low-carbon agricultural technology also exert an influence on their sustained adoption behaviour. This study proposes an innovative perspective to study the gap between the initial and sustained adoption of existing low-carbon agricultural technology based on a comprehensive analysis of the preceding findings. The objective of this study is to achieve an accurate match between low-carbon agricultural technology and farmers.\n\nFurthermore, it is worthwhile to mention that the conventional methodology utilised in agricultural policy research typically entails the gathering of data, the application of analytical models to elucidate the interrelationships between variables, and the formulation of policy recommendations. A review of the existing literature reveals that the majority of studies have formulated hypotheses and conducted regression analyses, primarily utilising econometric methods such as binary probit, propensity score matching, and double difference methods. Nevertheless, the inherent bias in establishing linear relationships between indicators in explanatory modelling results in insufficient discriminatory precision, which may lead to a wide range of policy recommendations that are disparate. The conventional techniques employed in statistical analysis are gradually being shown to have certain inherent limitations.\n\nIn recent years, machine learning methods have been employed extensively in agri-environmental and economic research due to their ability to effectively address complex non-linear problems. The field of machine learning is predicated on the analysis and learning of the mapping of existing data to expert classification models. Techniques for processing data and information are developed through algorithms that facilitate the stable and reliable classification of unknown categorical data. In the context of artificial intelligence and big data technology, the computational analysis method of machine learning can be applied to discover data laws or implicit models from multi-dimensional data that are difficult to perceive and discover. The computational analysis method offers several advantages, including high accuracy, rapid training speed, and straightforward implementation and interpretation. The method enhances the existing research paradigms and is extensively employed in diverse classification and regression tasks [[41](https:\u002F\u002Fwww.mdpi.com\u002F2071-1050\u002F18\u002F16\u002F8088#B41-sustainability-18-08088)].\n\nIn light of the aforementioned evidence, this study proceeded to devise and administer a questionnaire survey comprising 10,057 kiwifruit farmers situated in the provinces of Shaanxi, Guizhou, and Sichuan, which collectively represent the three leading kiwifruit-producing regions in China. The current status and challenges associated with the initial and sustained adoption of low-carbon agricultural technology by kiwifruit growers were further elucidated. The ordered logit model, which has been demonstrated to offer a robust explanatory framework for behavioural decisions, was selected as a reference for the analysis of the results obtained from other machine learning methods. The dataset was divided into training and testing sets based on the results of the experimental analysis, which employed a range of machine learning techniques. The optimal model is selected on the basis of the expected goodness of fit. Subsequently, the SHAP value interpretive method is implemented to ascertain the contribution rate of each digital capability dimension to the initial and sustained adoption of low-carbon agricultural technology by farmers, thereby facilitating the effective identification of the most significant influencing factors. The influence of digital capability and its constituent elements on the initial and sustained adoption of low-carbon agricultural technology by farmers is subjected to rigorous analysis, and recommendations for pertinent policy measures are put forth.\n\nThe contributions of this study are mainly reflected in the following three aspects: (1) Research Theme: This study investigates the characteristics of a sample of farmers and their initial and sustained adoption of low-carbon agricultural technology through micro-field research. In accordance with the findings of static technology adoption studies, further consideration is given to the continued adoption of technology. Accordingly, the objective of this study is to address the existing gap in the literature on the sustained adoption of low-carbon agricultural technology by farmers. (2) Research Content: This study provides a comprehensive examination of the composition of the digital capability of farmers, with the utilisation of factor analysis to construct and measure the digital capability index from four dimensions: digital access capability, digital acquisition capability, digital processing capability, and digital sharing capability. This offers a reference paradigm for the construction of indicators for subsequent related studies. Concurrently, it furnishes novel evidence to elucidate the influence of the digital capability of farmers and their multifaceted dimensions on the initial and sustained adoption of low-carbon agricultural technology by farmers. (3) Research Method: This study offers a comprehensive integration of the strengths inherent to both traditional econometric research methods and machine learning methods. In order to more accurately capture the impact of digital capability on the initial and sustained adoption of low-carbon agricultural technology by farmers, and to provide more scientific policy recommendations for the decision-making community, an ordered logit model, a Lasso regression method, a random forest method, an XGBoost model, and a BP neural network model were used.\n\n## 2. Theoretical Analysis and Research Hypotheses\n\n### 2.1. Theoretical Analysis Framework\n\nThe behaviours of farmers in the adoption of green and low-carbon agricultural technologies can be divided into two progressive stages: initial trial and long-term sustained application. However, there exist significant discrepancies in the behavioural constraint mechanisms between the two stages [[42](https:\u002F\u002Fwww.mdpi.com\u002F2071-1050\u002F18\u002F16\u002F8088#B42-sustainability-18-08088)]. The former stage is primarily restricted by decision-making uncertainty arising from insufficient information, whereas the latter stage hinges on perceived technical benefits and long-term production preferences. As a core endowment enabling farmers to adapt to the digital transformation of modern agriculture, digital capability can systematically resolve practical dilemmas hindering the adoption of low-carbon technology by farmers through two pathways: alleviating endogenous decision-making constraints and optimising exogenous production mechanisms. Digital capability effectively drives the behavioural transition of farmers from passive wait-and-see to proactive trials, and from one-off usage to regular, in-depth implementation. On the one hand, grounded in the bounded rationality decision-making theory and farmer behaviour theory, comprehensive digital capability integrated by multi-dimensional capability including digital access, information acquisition, data processing, and information sharing can break the ingrained cognitive habits shaped by the traditional farming experiences of farmers, expand supply channels of agricultural technical information, correct farmers’ biased risk perceptions of low-carbon technology, and reshape their long-term income expectations. From the endogenous cognitive dimension, it thereby facilitates farmers’ acceptance of innovative low-carbon production modes. On the other hand, drawing upon the sustained technology application theory, a sound d","Sustainability 2026-08-08","2026-08-08T00:00:00Z",75,{"impact":97,"substance":99,"depth":97,"authority":144,"freshness":48,"relevant":8,"comment":192},"基于大规模农户数据的机器学习研究，探讨数字能力对低碳农业采纳的影响，方法新颖，数据量大，结论可靠，对农业信息化与低碳发展有参考价值。",[194],{"name":188,"url":185},[196,197,4,198,199],"数字能力","低碳农业","随机森林","Sustainability",{"code":201,"slug":202,"name":203,"path":204,"parent_code":205},"B2","fruit","果树","\u002Ftopic\u002Fhorticulture\u002Ffruit","B",[207],{"code":40,"slug":110,"name":111,"path":112,"parent_code":113},{"status":116,"locked":22,"version":35,"method":36,"evidence":209},{"candidates":210,"agricultural_evidence":41},[211,216,220,225],{"code":201,"score":162,"supported":41,"title":212,"abstract":214,"tags":215},[213],"猕猴桃",[213],[],{"code":40,"score":21,"supported":41,"title":217,"abstract":218,"tags":219},[106],[],[4,198],{"code":221,"score":8,"supported":22,"title":222,"abstract":223,"tags":224},"H4",[],[],[197],{"code":226,"score":8,"supported":22,"title":227,"abstract":228,"tags":229},"I6",[],[],[196],[231,232],"低碳农业 数字能力 随机森林 Sustainability","低碳农业 数字能力","低碳农业数字能力随机森林Sustainability-380","2026-08-10T23:54:46.360428Z"]