[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"item-4311":3,"related-4311":110},{"id":4,"title":5,"url":6,"summary":7,"summary_zh":8,"content":9,"source_name":10,"source_url":6,"published_at":11,"category":12,"cover_url":9,"hotness":13,"is_selected":14,"score":15,"score_detail":16,"sources":24,"tags":26,"primary_topic":9,"secondary_topics":32,"classification":33,"search_phrases":80,"slug":83,"view_count":84,"doi":85,"paper":86,"created_at":109},4311,"Cross-Regional Digital Soil Mapping of OpenLandMap-Derived USDA Soil Texture Classes Using Remote Sensing and Machine Learning","https:\u002F\u002Fdoi.org\u002F10.3390\u002Fagriculture16192147","Soil is a fundamental environmental resource that supports food production and sustainable agricultural systems. Soil’s texture is a key soil property that affects its capacity to hold water, infiltration rate, fertility, erosion risk, and crop productivity. Traditionally, it is determined using hydrometer or pipette analysis. These laboratory procedures are costly and time-consuming, typically requiring approximately two working days, which limits their practicality for large-scale assessments. This paper proposes a cross-regional digital soil texture classification system based on remote sensing data and machine learning in Malaysia and Pakistan. The two countries were chosen because they represent contrasting environmental conditions: Malaysia has humid tropical conditions, and Pakistan has arid and semi-arid conditions. The OpenLandMap United States Department of Agriculture (USDA) soil texture classes served as the reference labels, and the spectral bands from Landsat, together with spectral indices and Shuttle Radar Topography Mission (SRTM) elevation, were used as predictor variables. The three ensemble models, including Random Forest (RF), XGBoost, and CatBoost, were evaluated for multi-class classification of USDA soil texture. The main Landsat-based Malaysia–Pakistan dataset produced the best results, with RF achieving the highest accuracy of 71.45%, XGBoost 71.31%, and CatBoost 67.33%. Additional experiments using 2025 Landsat 8\u002F9 and Sentinel-2 imagery achieved lower accuracies of 63.00% and 56.10%, respectively, while five-fold spatial block cross-validation produced a mean accuracy of 59.27 ± 1.84%. To strengthen dataset reliability, representative OpenLandMap-derived soil texture classes were externally validated using PCRWR sand–silt–clay fractions plotted on the USDA soil texture triangle and independent Malaysian soil resource references. The results of the feature importance analysis showed that elevation, Bare Soil Index (BSI), Normalized Difference Water Index (NDWI), and Normalized Difference Vegetation Index (NDVI) were important predictors for classifying OpenLandMap-derived USDA soil texture classes. These show that the Landsat-based spectral and terrain features, when used in conjunction with ensemble learning, could assist cross-regional soil texture classification for the USDA. However, class imbalance, spectral overlap between soil texture classes, temporal variability, and resolution mismatch between satellite predictors and soil labels remain important challenges. The findings of this study provide a useful foundation for large-scale soil texture mapping and future spatial validation across different environmental regions.","土壤是支撑粮食生产和可持续农业系统的基础环境资源。土壤质地是影响土壤持水能力、入渗速率、肥力、侵蚀风险和作物生产力的关键土壤属性。传统上，土壤质地采用比重计法或吸管法测定。这些实验室方法成本高、耗时长，通常需要约两个工作日，限制了其在大规模评估中的实用性。本文提出了一种基于遥感数据和机器学习的跨区域数字土壤质地分类系统，研究区域为马来西亚和巴基斯坦。选择这两个国家是因为它们代表了截然不同的环境条件：马来西亚属于湿润热带条件，巴基斯坦属于干旱和半干旱条件。以OpenLandMap提供的美国农业部（USDA）土壤质地类别作为参考标签，以Landsat光谱波段、光谱指数和航天飞机雷达地形测绘任务（SRTM）高程数据作为预测变量。评估了三种集成模型，包括随机森林（RF）、XGBoost和CatBoost，用于USDA土壤质地的多类分类。基于Landsat的马来西亚—巴基斯坦主数据集取得了最佳结果，RF达到最高精度71.45%，XGBoost为71.31%，CatBoost为67.33%。使用2025年Landsat 8\u002F9和Sentinel-2影像的额外实验分别取得了较低的精度，为63.00%和56.10%，而五折空间块交叉验证产生的平均精度为59.27 ± 1.84%。为增强数据集可靠性，使用PCRWR砂—粉—黏粒比例绘制在USDA土壤质地三角图上，并结合独立的马来西亚土壤资源参考资料，对具有代表性的OpenLandMap衍生土壤质地类别进行了外部验证。特征重要性分析结果表明，高程、裸土指数（BSI）、归一化差异水体指数（NDWI）和归一化差异植被指数（NDVI）是分类OpenLandMap衍生USDA土壤质地类别的重要预测因子。这些结果表明，基于Landsat的光谱和地形特征与集成学习结合使用时，可辅助USDA的跨区域土壤质地分类。然而，类别不平衡、土壤质地类别之间的光谱重叠、时间变异性以及卫星预测因子与s之间的分辨率不匹配",null,"Agriculture","2026-10-05T00:00:00Z","论文",10,false,77,{"impact":17,"substance":18,"depth":19,"authority":20,"freshness":21,"relevant":22,"comment":23},16,21,18,13,9,1,"跨区域遥感与集成学习土壤质地分类研究，方法扎实、结论有参考价值，但精度有限，属细分领域进展。",[25],{"name":10,"url":6},[27,28,29,30,31],"数字农业","智慧农业","机器学习","遥感","土壤制图",[],{"status":34,"locked":14,"version":35,"method":36,"evidence":37},"needs_review","2026-10-22.2","rules",{"candidates":38,"agricultural_evidence":42,"reason":79},[39,48,54,60,67,73],{"code":40,"score":41,"supported":42,"title":43,"abstract":45,"tags":47},"H2",15,true,[44],"soil",[46,44],"土壤",[31],{"code":49,"score":41,"supported":42,"title":50,"abstract":52,"tags":53},"I2",[51],"machine learning",[29,51],[29],{"code":55,"score":41,"supported":42,"title":56,"abstract":58,"tags":59},"I3",[57],"remote sensing",[30,57],[30],{"code":61,"score":62,"supported":42,"title":63,"abstract":64,"tags":66},"H7",3,[],[65],"干旱",[],{"code":68,"score":62,"supported":42,"title":69,"abstract":70,"tags":72},"H8",[],[71],"可持续农业",[],{"code":74,"score":75,"supported":14,"title":76,"abstract":77,"tags":78},"I1",2,[],[],[27,28],"多个产业证据相当，需要确认主归属",[81,82],"OpenLandMap USDA 土壤质地","Landsat 土壤质地 机器学习","OpenLandMapUSDA土壤质地-4311",0,"10.3390\u002Fagriculture16192147",{"doi":85,"openalex_id":87,"authors":88,"venue":10,"cited_by_count":84,"oa_url":6,"card":102,"direction":106,"ingested_from":108},"W7220411242",[89,91,93,96,99],{"name":90,"orcid":9},"Umair Maqsood",{"name":92,"orcid":9},"Fatima Iftikhar",{"name":94,"orcid":95},"Kashif Sattar","https:\u002F\u002Forcid.org\u002F0000-0001-8620-9419",{"name":97,"orcid":98},"Iftikhar A. Ahmad","https:\u002F\u002Forcid.org\u002F0000-0003-3719-2387",{"name":100,"orcid":101},"Madini Obad Alassafi","https:\u002F\u002Forcid.org\u002F0000-0001-9919-8368",{"tldr":103,"method":104,"finding":105,"direction":106,"opportunity":107},"用Landsat光谱与地形特征结合集成学习，实现马来西亚与巴基斯坦跨区域USDA土壤质地分类。","Landsat光谱波段、光谱指数与SRTM高程，RF\u002FXGBoost\u002FCatBo","RF精度最高71.45%，高程、BSI、NDWI、NDVI为关键预测因子，跨区域可行但精度受限。","农业遥感与作物表型","可探索多源遥感融合与类别不平衡校正，提升跨气候区土壤质地制图精度与迁移能力。","openalex","2026-10-06T23:30:20.282911Z",{"total":111,"page":22,"page_size":111,"items":112},6,[113,227,312,372,428,480],{"id":114,"title":115,"url":116,"summary":117,"summary_zh":118,"content":9,"source_name":119,"source_url":116,"published_at":120,"category":12,"cover_url":9,"hotness":13,"is_selected":14,"score":121,"score_detail":122,"sources":127,"tags":129,"primary_topic":132,"secondary_topics":136,"classification":142,"search_phrases":179,"slug":182,"view_count":84,"doi":183,"paper":184,"created_at":226},5102,"Integrating Proximal and Remote Sensing to Improve Characterization and Mapping of Soil Organic Carbon Stocks under Contrasting Land Uses","https:\u002F\u002Fdoi.org\u002F10.1007\u002Fs42729-026-03629-3","Abstract Soils are fundamental to support ecosystem functions. Therefore, understanding soil organic carbon stock (SCS) as a strategic environmental component for climate regulation, sustainable agricultural practices, and maximizing fertilizer application is essential. This study aimed to predict and map the SCS at Palmital Farm, Minas Gerais state, using land use, parent materials, soil classes, proximal and remote sensing data as predictor variables. A total of 76 soil samples were collected in the surface (A) and subsurface (B) horizons using a regular grid design (200 × 200 m). Carbon content was determined using the Walkley–Black method. Prediction models were created via Random Forest algorithm and validated through leave-one-out cross-validation. Environmental and anthropogenic influences on SCS were analyzed statistically and spatially. The highest mean SCS values were found at the surface horizon under native vegetation and pasture areas (~ 33 Mg ha⁻¹), due to continuous organic matter deposition and lack of disturbance. Charnockite-derived soils showed higher SCS (31.24 Mg ha⁻¹, surface). Among soil classes, Argissolos stood out reaching the highest SCS (35.86 Mg ha⁻¹). Remote sensing data and terrain attributes yielded the best SCS prediction (R²=0.43). The main factors influencing SCS were land use\u002Fmanagement and soil texture. Also, strong correlations were observed between vegetation indices, some terrain attributes and SCS. This approach, combining multiple variables, indicated a moderate potential to achieve accurate SCS predictions using these data, demonstrating the high variability and complexity of SCS spatial variation in detailed-scale mapping.","摘要 土壤是支撑生态系统功能的基础。因此，理解土壤有机碳储量（SCS）作为气候调节、可持续农业实践和最大化肥料施用的战略性环境组成部分至关重要。本研究旨在利用土地利用、成土母质、土壤类别、近端和遥感数据作为预测变量，对米纳斯吉拉斯州Palmital农场的SCS进行预测和制图。采用规则网格设计（200 × 200 m），在表层（A）和亚表层（B）共采集76个土壤样品。碳含量采用Walkley–Black法测定。通过随机森林算法建立预测模型，并采用留一交叉验证进行验证。对环境和人为因素对SCS的影响进行了统计和空间分析。表层中原生植被和牧草地利用方式下的SCS均值最高（约33 Mg ha⁻¹），这归因于有机质的持续沉积和缺乏扰动。源自紫苏花岗岩的土壤表现出较高的SCS（31.24 Mg ha⁻¹，表层）。在土壤类别中，Argissolos表现突出，达到最高SCS（35.86 Mg ha⁻¹）。遥感数据和地形属性获得了最佳的SCS预测效果（R²=0.43）。影响SCS的主要因素为土地利用\u002F管理和土壤质地。此外，植被指数、部分地形属性与SCS之间观察到较强的相关性。这种结合多变量的方法表明，利用这些数据实现准确SCS预测具有中等潜力，同时也揭示了详细尺度制图中SCS空间变异的高变异性和复杂性。","Journal of soil science and plant nutrition","2026-10-14T00:00:00Z",67,{"impact":123,"substance":124,"depth":125,"authority":20,"freshness":21,"relevant":22,"comment":126},8,20,17,"该研究结合近地与遥感数据及随机森林算法预测土壤有机碳储量，方法新颖、数据扎实，对农业信息化与精准农业有参考价值，但属区域性案例研究，影响范围有限。",[128],{"name":119,"url":116},[27,130,131,30,31],"随机森林","土壤有机碳",{"code":40,"slug":44,"name":133,"path":134,"parent_code":135},"土壤健康与改良","\u002Ftopic\u002Fenvironment\u002Fsoil","H",[137],{"code":55,"slug":138,"name":139,"path":140,"parent_code":141},"remote-sensing","遥感与农情监测","\u002Ftopic\u002Fdigital-agriculture\u002Fremote-sensing","I",{"status":143,"locked":14,"version":35,"method":36,"evidence":144},"classified",{"candidates":145,"agricultural_evidence":42},[146,150,154,161,165,171,175],{"code":40,"score":17,"supported":42,"title":147,"abstract":148,"tags":149},[44],[46,44],[31,131],{"code":55,"score":41,"supported":42,"title":151,"abstract":152,"tags":153},[57],[30,57],[30],{"code":155,"score":111,"supported":42,"title":156,"abstract":157,"tags":160},"F3",[],[158,159],"肥料","fertilizer",[],{"code":68,"score":62,"supported":42,"title":162,"abstract":163,"tags":164},[],[71],[],{"code":166,"score":62,"supported":42,"title":167,"abstract":168,"tags":170},"I4",[],[169],"空间分析",[],{"code":74,"score":22,"supported":14,"title":172,"abstract":173,"tags":174},[],[],[27],{"code":49,"score":22,"supported":14,"title":176,"abstract":177,"tags":178},[],[],[130],[180,181],"Palmital Farm 土壤有机碳","Minas Gerais 遥感 土壤碳","PalmitalFarm土壤有机碳-5102","10.1007\u002Fs42729-026-03629-3",{"doi":183,"openalex_id":185,"authors":186,"venue":119,"cited_by_count":84,"oa_url":220,"card":221,"direction":106,"ingested_from":108},"W7222335269",[187,190,193,196,199,202,205,208,211,214,217],{"name":188,"orcid":189},"Leonardo Augusto de Sousa","https:\u002F\u002Forcid.org\u002F0009-0003-1560-5600",{"name":191,"orcid":192},"Fernanda Almeida Bócoli","https:\u002F\u002Forcid.org\u002F0000-0001-5476-6886",{"name":194,"orcid":195},"Eduane José de Pádua","https:\u002F\u002Forcid.org\u002F0009-0003-1782-2756",{"name":197,"orcid":198},"Luiza Maria Pereira Pierangeli","https:\u002F\u002Forcid.org\u002F0000-0002-9218-3718",{"name":200,"orcid":201},"Junior César Avanzi","https:\u002F\u002Forcid.org\u002F0000-0003-2455-0325",{"name":203,"orcid":204},"Bruno Montoani Silva","https:\u002F\u002Forcid.org\u002F0000-0002-8240-8987",{"name":206,"orcid":207},"Carlos Victor Lima Albuquerque","https:\u002F\u002Forcid.org\u002F0009-0009-4597-4443",{"name":209,"orcid":210},"Luiz Roberto Guimarães Guilherme","https:\u002F\u002Forcid.org\u002F0000-0002-5387-6028",{"name":212,"orcid":213},"Marco Aurélio Carbone Carneiro","https:\u002F\u002Forcid.org\u002F0000-0003-4349-3071",{"name":215,"orcid":216},"Nilton Curi","https:\u002F\u002Forcid.org\u002F0000-0002-2604-0866",{"name":218,"orcid":219},"Sérgio Henrique Godinho Silva","https:\u002F\u002Forcid.org\u002F0000-0003-2750-5976","https:\u002F\u002Flink.springer.com\u002Fcontent\u002Fpdf\u002F10.1007\u002Fs42729-026-03629-3.pdf",{"tldr":222,"method":223,"finding":224,"direction":106,"opportunity":225},"结合近地与遥感数据及随机森林预测并制图巴西农场不同土地利用下的土壤有机碳储量。","76个样点、Walkley-Black法测碳，随机森林建模，留一交叉验证。","遥感与地形因子预测最佳（R²=0.43），土地利用和土壤质地主导碳储量空间变异。","可探索多源遥感与深度学习方法提升小尺度土壤碳制图精度，并耦合管理措施情景。","2026-10-16T23:30:31.433840Z",{"id":228,"title":229,"url":230,"summary":231,"summary_zh":232,"content":9,"source_name":233,"source_url":230,"published_at":234,"category":12,"cover_url":9,"hotness":13,"is_selected":14,"score":235,"score_detail":236,"sources":240,"tags":242,"primary_topic":245,"secondary_topics":249,"classification":256,"search_phrases":290,"slug":293,"view_count":62,"doi":294,"paper":295,"created_at":311},4693,"Scalable, cloud-native decision support system for Variable Rate Application (VRA) based on multi-year NDRE time series in arable precision agriculture","https:\u002F\u002Fdoi.org\u002F10.5281\u002Fzenodo.23282433","This paper presents an automated, cloud-native geospatial Decision Support System (DSS) that generates variable rate prescription maps directly from Sentinel-2 satellite observations. Designed as an end-to-end arable field VRA architecture, the system operates completely autonomously with minimal input, requiring only the field boundary polygon, the crop phenology class (winter or spring crop), and the multi-year observation window. The computation engine streams Cloud-Optimized GeoTIFF (COG) rasters via asynchronous cloud connectivity through the Microsoft Planetary Computer (MPC) SpatioTemporal Asset Catalog (STAC) Application Programming Interface (API), enforcing data quality through multi-stage Scene Classification Layer (SCL) filtering. Instead of the rapidly saturating Normalized Difference Vegetation Index (NDVI), the model relies on the Normalized Difference Red Edge (NDRE) index, which accurately captures canopy chlorophyll and green biomass dynamics. For each native 10×10 m grid cell, the system extracts the 95th percentile (P95) peak value within a crop-specific reproductive phenological window, eliminating single-day snapshot noise while validating winter or spring growth trajectories via automated phenological ratio calculations. The decision and prescription engine is built upon two core agronomic safety pillars: An exponential empirical Bayesian stability shrinkage that dampens unstable, high-variance zones toward the safe field average, preventing the over-application of inputs on drought-prone areas. A trigonometric Horn slope attenuation derived from digital elevation models that systematically reduces application rates on erosion-prone steep slopes. Cell-level rates are distributed within the target tolerance range using a rank-allocation algorithm coupled with a Hamilton quota (largest remainder method) correction, guaranteeing strict integer-based machine compatibility and zero-sum conservation of the total field input budget. The technology was validated on a 139.03 ha commercial production field across four reference seasons (2022, 2023, 2024, and 2026) for Variable Rate Seeding (VRS) in winter wheat (Triticum aestivum L., 230 kg\u002Fha baseline seed rate) in a rolling hilly landscape in Hungary, benchmarked against conventional uniform application and conservative multi-year VRS planning. The primary quantitative findings are summarized below: · Failure of conventional uniform application: Across the 14,250 cells, the flat 230 kg\u002Fha rate was optimal on only 1.1% of the field area. · Input reallocation: The model reallocated 1,633.7 kg of seed (and 1,167.1 kg compared to conservative VRS) from unproductive areas into high-capacity zones (Figure 1.). Across 98.9% of the field, conventional uniform management generated systematic under- or over-seeding, inducing yield losses or wasted operational expenditure. · Drought mitigation: In temporally unstable zones subject to extreme weather risk (covering 33.6% of the field, 46.69 ha), Bayesian stability shrinkage reduced seed allocation by 495 kg compared to uniform seeding. Establishing lower canopy densities in these vulnerable zones mitigates intra-crop water competition and prevents early-season forced ripening (Lollato et al., 2017). · Erosion control: On steep slopes exceeding 12% inclination (accounting for 9.75% of the field, 13.56 ha), the Horn topographic attenuation engine reduced seed input by 553 kg relative to the fixed 230 kg\u002Fha baseline. This intervention minimizes surface runoff damage and agrochemical leaching. · Economic return: Spatially optimizing plant density based on pedological and topographic constraints delivers substantial yield gains. Reallocating seed across the field—without any expenditure on supplemental fertilizer—produces a projected yield increase of +0.20 to +0.40 t\u002Fha (+28 to +56 tons of grain across the evaluated field) (Kindred et al., 2016). At a baseline wheat price of €192\u002Ft (70,000 HUF\u002Ft), this generates an estimated net value of €5,376 to €10,752 (1.96 to 3.92M HUF) on this single 139 ha field.","本文提出了一种自动化、云原生的地理空间决策支持系统（DSS），可直接从Sentinel-2卫星观测数据生成变量施肥处方图。该系统设计为端到端的耕地变量施肥（VRA）架构，完全自主运行且输入需求极低，仅需田块边界多边形、作物物候类别（冬作或春作）以及多年观测窗口。计算引擎通过Microsoft Planetary Computer（MPC）时空资产目录（STAC）应用程序接口（API）以异步云连接方式流式传输云优化GeoTIFF（COG）栅格数据，并通过多阶段场景分类层（SCL）滤波确保数据质量。该模型不采用易快速饱和的归一化差异植被指数（NDVI），而是依赖归一化差异红边指数（NDRE），后者能够准确捕捉冠层叶绿素和绿色生物量动态。对于每个原生10×10 m网格单元，系统提取作物特定生殖物候窗口内的第95百分位数（P95）峰值，消除单日快照噪声，同时通过自动化物候比率计算验证冬作或春作生长轨迹。决策与处方引擎建立在两个核心农艺安全支柱之上：指数经验贝叶斯稳定性收缩，将不稳定的高方差区域向安全的田块平均值收缩，防止在易旱区域过量施用投入品；基于数字高程模型推导的三角函数Horn坡度衰减，系统性地降低易侵蚀陡坡上的施用量。单元级施用量通过秩分配算法结合Hamilton配额（最大余数法）校正，在目标容差范围内分配，确保严格的基于整数的机器兼容性和田块总投入预算的零和守恒。该技术已在匈牙利起伏丘陵地貌的139.03 ha商业化生产田块上，针对冬小麦（Triticum aestivum L.，基准播种量230 kg\u002Fha）的变量播种（VRS），在四个参考季节（2022、2023、2024和2026年）进行了验证，并以常规均一施用和保守的多年VRS规划为基准进行比较。主要定量结果总结如下：· 常规均一施用的失败：在14,250个单元中，统一的230 kg\u002Fha施","Zenodo (CERN European Organization for Nuclear Research)","2026-10-10T00:00:00Z",83,{"impact":19,"substance":237,"depth":238,"authority":20,"freshness":13,"relevant":22,"comment":239},23,19,"基于多年度NDRE时序的云原生变量施肥决策系统，方法新颖、验证数据扎实，对精准农业信息化有较高参考价值。",[241],{"name":233,"url":230},[27,28,243,244,30],"变量施肥","精准农业",{"code":49,"slug":246,"name":247,"path":248,"parent_code":141},"agricultural-ai","农业AI与大模型","\u002Ftopic\u002Fdigital-agriculture\u002Fagricultural-ai",[250],{"code":251,"slug":252,"name":253,"path":254,"parent_code":255},"A2","wheat","小麦","\u002Ftopic\u002Ffield-crops\u002Fwheat","A",{"status":143,"locked":14,"version":35,"method":36,"evidence":257},{"candidates":258,"agricultural_evidence":42},[259,265,270,275,281,286],{"code":74,"score":260,"supported":42,"title":261,"abstract":263,"tags":264},11,[262],"precision agriculture",[],[27,28,244],{"code":49,"score":260,"supported":42,"title":266,"abstract":268,"tags":269},[267],"decision support",[267],[],{"code":251,"score":21,"supported":42,"title":271,"abstract":272,"tags":274},[],[253,273,252],"winter wheat",[],{"code":155,"score":276,"supported":42,"title":277,"abstract":278,"tags":280},7,[],[279,159],"施肥",[243],{"code":61,"score":62,"supported":42,"title":282,"abstract":283,"tags":285},[],[284],"drought",[],{"code":55,"score":22,"supported":14,"title":287,"abstract":288,"tags":289},[],[],[30],[291,292],"Sentinel-2 NDRE 变量施肥","云原生 决策支持系统 精准农业","Sentinel-2NDRE变量施肥-4693","10.5281\u002Fzenodo.23282433",{"doi":294,"openalex_id":296,"authors":297,"venue":233,"cited_by_count":84,"oa_url":230,"card":305,"direction":310,"ingested_from":108},"W7221365319",[298,300,302],{"name":299,"orcid":9},"Zoltán Szenek",{"name":301,"orcid":9},"Donát Szám",{"name":303,"orcid":304},"Gábor Milics","https:\u002F\u002Forcid.org\u002F0000-0002-4499-8322",{"tldr":306,"method":307,"finding":308,"direction":106,"opportunity":309},"构建云原生决策支持系统，用多年NDRE时序自动生成精准农业变量施肥\u002F播种处方图。","Sentinel-2 NDRE、COG\u002FSTAC云流式处理、贝叶斯收缩与坡度衰减","在匈牙利139公顷冬小麦田验证，系统优于传统均一施用与保守多年VRS规划。","可探索将NDRE时序DSS扩展至多作物、多区域及实时物联网数据融合的通用处方框架。","数字乡村与农业信息化","2026-10-11T23:30:36.790037Z",{"id":313,"title":314,"url":315,"summary":316,"summary_zh":9,"content":9,"source_name":317,"source_url":9,"published_at":9,"category":12,"cover_url":9,"hotness":13,"is_selected":14,"score":318,"score_detail":319,"sources":322,"tags":324,"primary_topic":327,"secondary_topics":331,"classification":332,"search_phrases":358,"slug":361,"view_count":75,"doi":9,"paper":362,"created_at":371},4477,"玉米生产精准农业系统综述：技术、应用与产量优化（2020—2026）","https:\u002F\u002Fm2.mtmt.hu\u002Fapi\u002Fpublication\u002F37500368","依据PRISMA 2020指南，系统检索Scopus与Web of Science共464篇文献，最终纳入129项研究；样本构成为田间试验（28.6%）、遥感与精准农业整合研究（25.5%）、机器学习应用（19.4%）、气候知情精准农业策略（23.4%）和土壤退化研究（3.1%）；研究表明精准农业技术通过定点管理和数据驱动决策支持，提升了玉米生产力、养分利用效率、水利用效率和产量预测准确性；指出采用精准农业仍受制于实施成本高、技术复杂、数据有限等因素。","Precision Agriculture in Maize Production Systematic Review 2026",74,{"impact":19,"substance":320,"depth":19,"authority":20,"freshness":62,"relevant":22,"comment":321},22,"基于PRISMA 2020的系统综述，纳入129项研究，数据规模与方法规范性强，对玉米精准农业技术路径与瓶颈有实质参考价值，但发布日期不明、时效性偏弱。",[323],{"name":317,"url":315},[28,325,29,326,244,30],"产量预测","玉米",{"code":328,"slug":329,"name":326,"path":330,"parent_code":255},"A3","maize","\u002Ftopic\u002Ffield-crops\u002Fmaize",[],{"status":143,"locked":14,"version":35,"method":36,"evidence":333},{"candidates":334,"agricultural_evidence":42},[335,339,344,349,354],{"code":74,"score":20,"supported":42,"title":336,"abstract":337,"tags":338},[244],[244],[28,244],{"code":328,"score":340,"supported":42,"title":341,"abstract":342,"tags":343},12,[326],[326],[326],{"code":55,"score":345,"supported":42,"title":346,"abstract":347,"tags":348},5,[],[30],[325,30],{"code":49,"score":350,"supported":42,"title":351,"abstract":352,"tags":353},4,[],[29],[29],{"code":40,"score":62,"supported":42,"title":355,"abstract":356,"tags":357},[],[46],[],[359,360],"玉米 精准农业 系统综述","玉米 遥感 机器学习 产量","玉米精准农业系统综述-4477",{"doi":9,"openalex_id":9,"authors":363,"venue":9,"cited_by_count":84,"oa_url":9,"card":364,"direction":368,"ingested_from":370},[],{"tldr":365,"method":366,"finding":367,"direction":368,"opportunity":369},"系统综述2020—2026年玉米精准农业技术、应用与产量优化研究。","按PRISMA 2020检索Scopus与WoS，纳入129项研究做分类分析。","精准农业通过定点管理与数据决策提升玉米产量、养分和水利用效率。","智慧农业 \u002F 农业物联网","成本高、技术复杂与数据不足制约推广，低成本轻量化方案与本地化数据共享是研究空白。","agent","2026-10-09T00:05:02.451957Z",{"id":373,"title":374,"url":375,"summary":376,"summary_zh":377,"content":9,"source_name":378,"source_url":375,"published_at":11,"category":12,"cover_url":9,"hotness":13,"is_selected":14,"score":379,"score_detail":380,"sources":383,"tags":385,"primary_topic":9,"secondary_topics":387,"classification":388,"search_phrases":408,"slug":411,"view_count":75,"doi":412,"paper":413,"created_at":427},4287,"The effect of train-test data splitting ratio on yield prediction from multispectral UAV imagery: effective reduction of training ratio with algorithm selection for single- and multi-date models","https:\u002F\u002Fdoi.org\u002F10.1007\u002Fs11119-026-10464-0","Abstract Context Optimizing the ratio of data required for training spectral grain yield (GY) prediction models is critical for efficient sampling of reference data, but has been rarely addressed so far. Given the high costs for collecting ground truth data, standard data splitting ratios which use the majority of the data for training and the rest for testing the models should be minimized, but in turn may affect prediction accuracy. Methods Therefore, this study evaluated 11 different train-test data splitting ratios (TSR), ranging from using 5% to 95% of the data for training, and the remaining portion as test sets. Models with six different machine learning algorithms were compared in winter wheat breeding trials conducted with each several thousand plots in two locations in Germany over a period of four years. The input data consisted of the (unmanned aerial vehicle) UAV-based NDRE index data from individual measurements dates as well as multi-date combinations. Results The results indicate that GY prediction remained relatively stable when decreasing TSR to about 0.30. Conversely, multi-date models tended to benefit more from higher TSR than single-date models. Support vector machine and random forest algorithms demonstrated relative advantage both for higher TSR and multi-date models, whereas partial least squares and ridge regression were the best algorithms for lowest TSR-values. Furthermore, analysis via repeated data splitting revealed minimum R² variability at a TSR of 0.30, but substantially increased variance at higher TSR values. Conclusions It is concluded that decreasing TSR while considering algorithm selection can potentially reduce costs without compromising prediction accuracy, therefore making spectral phenotyping methods more accessible and ready-to-use, whereas the absolute number of data points requires additional examination.","摘要 背景 优化光谱谷物产量（grain yield, GY）预测模型所需数据的训练-测试比例对于高效采集参考数据至关重要，但迄今鲜有研究涉及。鉴于获取地面真实数据的高昂成本，应尽量减少使用大部分数据进行模型训练、其余部分用于测试的标准数据划分比例，但这反过来可能影响预测精度。方法 因此，本研究评估了11种不同的训练-测试数据划分比例（train-test data splitting ratio, TSR），训练数据占比从5%到95%不等，剩余部分作为测试集。在德国两个地点、为期四年、每个试验包含数千个小区的冬小麦育种试验中，比较了采用六种不同机器学习算法的模型。输入数据包括来自单次测量日期的基于无人机（unmanned aerial vehicle, UAV）的NDRE指数数据以及多日期组合数据。结果 结果表明，当TSR降低至约0.30时，GY预测仍保持相对稳定。相反，多日期模型比单日期模型更倾向于从较高的TSR中获益。支持向量机（support vector machine）和随机森林（random forest）算法在较高TSR和多日期模型中均表现出相对优势，而偏最小二乘回归（partial least squares）和岭回归（ridge regression）在最低TSR值下表现最佳。此外，通过重复数据划分分析发现，TSR为0.30时R²变异性最小，但在较高TSR值下方差显著增大。结论 研究表明，在考虑算法选择的前提下降低TSR有可能在不影响预测精度的情况下减少成本，从而使光谱表型分析方法更加易于获取和实用，但数据点的绝对数量仍需进一步研究。","Precision Agriculture",75,{"impact":340,"substance":320,"depth":19,"authority":381,"freshness":21,"relevant":22,"comment":382},14,"基于四年多点小麦育种试验系统评估训练集比例与算法选择对无人机多光谱产量预测的影响，方法扎实、结论实用，对降低表型数据采集成本有直接参考价值。",[384],{"name":378,"url":375},[28,325,29,386,30],"小麦育种",[],{"status":34,"locked":14,"version":35,"method":36,"evidence":389},{"candidates":390,"agricultural_evidence":42,"reason":79},[391,395,400,404],{"code":251,"score":13,"supported":42,"title":392,"abstract":393,"tags":394},[],[253,273,252],[386],{"code":55,"score":13,"supported":42,"title":396,"abstract":398,"tags":399},[397],"multispectral",[],[325,30],{"code":49,"score":276,"supported":42,"title":401,"abstract":402,"tags":403},[],[29,51],[29],{"code":74,"score":22,"supported":14,"title":405,"abstract":406,"tags":407},[],[],[28],[409,410],"无人机 多光谱 小麦 产量预测","冬小麦 育种 遥感 机器学习","无人机多光谱小麦产量预测-4287","10.1007\u002Fs11119-026-10464-0",{"doi":412,"openalex_id":414,"authors":415,"venue":378,"cited_by_count":84,"oa_url":375,"card":422,"direction":106,"ingested_from":108},"W7220450104",[416,419],{"name":417,"orcid":418},"Lukas Prey","https:\u002F\u002Forcid.org\u002F0000-0001-6287-4596",{"name":420,"orcid":421},"Patrick Noack","https:\u002F\u002Forcid.org\u002F0000-0002-0226-6609",{"tldr":423,"method":424,"finding":425,"direction":106,"opportunity":426},"研究不同训练-测试数据划分比例对无人机多光谱小麦产量预测的影响，发现可降低训练比例。","基于德国四年两地点冬小麦育种试验，用UAV NDRE指数和六种机器学习算法比较1","训练比例降至约0.30时预测精度稳定，多时相模型更受益于高比例，SVM和随机森林较优。","可探索在更低训练比例下结合主动学习或迁移学习，进一步降低地面采样成本并验证跨地点泛化性。","2026-10-06T23:30:02.110164Z",{"id":429,"title":430,"url":431,"summary":432,"summary_zh":9,"content":9,"source_name":433,"source_url":9,"published_at":434,"category":12,"cover_url":9,"hotness":13,"is_selected":14,"score":435,"score_detail":436,"sources":438,"tags":440,"primary_topic":443,"secondary_topics":444,"classification":446,"search_phrases":467,"slug":470,"view_count":350,"doi":9,"paper":471,"created_at":479},4286,"Applications of Artificial Intelligence in Soil Characterization and Agriculture: A Systematic Review","https:\u002F\u002Fm2.mtmt.hu\u002Fapi\u002Fpublication\u002F37386408","Navarro Rubio等系统综述2018-2026年196项AI土壤特性与农业应用研究，遵循PRISMA 2020框架。综述发现研究明显向数据驱动方法转变，机器学习和深度学习模型主导近期研究；随机森林、支持向量机、梯度提升方法、人工神经网络、卷积神经网络、长短期记忆架构最常被报告；主要数据源包括原位传感、实验室测量、遥感影像和环境协变量，常通过多源数据融合框架集成；树集成模型在不同土壤特性上提供稳健性能。","Agronomy 2026 16(13):1241 DOI:10.3390\u002Fagronomy16131241","2026-10-02T00:00:00Z",78,{"impact":19,"substance":320,"depth":19,"authority":381,"freshness":111,"relevant":22,"comment":437},"PRISMA框架下196项研究的系统综述，方法学扎实、结论可靠，对农业AI土壤应用具较高参考价值。",[439],{"name":433,"url":431},[28,441,29,30,442],"农业人工智能","土壤监测",{"code":49,"slug":246,"name":247,"path":248,"parent_code":141},[445],{"code":40,"slug":44,"name":133,"path":134,"parent_code":135},{"status":143,"locked":14,"version":35,"method":36,"evidence":447},{"candidates":448,"agricultural_evidence":42},[449,455,459,463],{"code":49,"score":17,"supported":42,"title":450,"abstract":452,"tags":454},[451],"artificial intelligence",[453,29],"深度学习",[441,29],{"code":40,"score":260,"supported":42,"title":456,"abstract":457,"tags":458},[44],[46],[],{"code":55,"score":350,"supported":42,"title":460,"abstract":461,"tags":462},[],[30],[30],{"code":74,"score":22,"supported":14,"title":464,"abstract":465,"tags":466},[],[],[28],[468,469],"AI 土壤特性 系统综述","机器学习 土壤 遥感","AI土壤特性系统综述-4286",{"doi":9,"openalex_id":9,"authors":472,"venue":9,"cited_by_count":84,"oa_url":9,"card":473,"direction":477,"ingested_from":370},[],{"tldr":474,"method":475,"finding":476,"direction":477,"opportunity":478},"系统综述2018-2026年196项AI用于土壤特性与农业的研究。","PRISMA 2020框架，分析机器学习与深度学习模型及多源数据融合。","研究转向数据驱动，树集成模型在不同土壤特性上表现稳健。","农业人工智能与决策模型","可探索可解释AI与多源数据融合在土壤特性预测中的泛化能力。","2026-10-06T00:05:04.976398Z",{"id":481,"title":482,"url":483,"summary":484,"summary_zh":485,"content":9,"source_name":486,"source_url":483,"published_at":434,"category":12,"cover_url":9,"hotness":487,"is_selected":14,"score":488,"score_detail":489,"sources":491,"tags":495,"primary_topic":497,"secondary_topics":498,"classification":499,"search_phrases":528,"slug":531,"view_count":75,"doi":532,"paper":533,"created_at":543},4238,"Comparative Performance of Machine Learning and Deep Learning Algorithms for Agricultural Land Mapping in Anambra State, Nigeria","https:\u002F\u002Fdoi.org\u002F10.5281\u002Fzenodo.23106955","Abstract - Accurate delineation of agricultural land is fundamental to agricultural planning, food-security assessment, land-resource management, and monitoring of land-use transformation. However, the heterogeneous and fragmented character of agricultural landscapes in southeastern Nigeria presents substantial challenges for satellite-image classification. This study comparatively evaluated Random Forest (RF), Support Vector Machine (SVM), Artificial Neural Network (ANN), and Convolutional Neural Network (CNN) for agricultural land mapping in Anambra State, Nigeria. Remotely sensed imagery was classified into waterbody, dense forest, agricultural land, built-up area, and bare ground using the four algorithms. Model performance was evaluated using 210 validation samples and multiple confusion-matrix metrics, including sensitivity, precision, specificity, negative predictive value, overall accuracy, balanced accuracy, and Matthew’s correlation coefficient. Agricultural land estimates varied from 1,000.26 km² (21.78%) for SVM to 1,099.12 km² (23.93%) for ANN. SVM achieved the strongest overall performance, recording 88.5% sensitivity, 85.2% precision, 75.0% specificity, 83.3% overall accuracy, 81.7% balanced accuracy, and a Matthews correlation coefficient of 0.643. CNN produced the lowest overall accuracy of 74.3%. The findings demonstrate that increased model complexity does not necessarily translate into superior classification performance and establish SVM as an effective approach for regional agricultural land delineation under heterogeneous tropical landscape conditions. Keywords: agricultural land mapping; machine learning; deep learning; Support Vector Machine; Random Forest; artificial neural network; convolutional neural network; remote sensing","摘要——准确划定农业用地是农业规划、粮食安全评估、土地资源管理和土地利用变化监测的基础。然而，尼日利亚东南部农业景观的异质性和破碎化特征给卫星图像分类带来了巨大挑战。本研究比较评估了随机森林（RF）、支持向量机（SVM）、人工神经网络（ANN）和卷积神经网络（CNN）在尼日利亚阿南布拉州农业用地制图中的表现。利用这四种算法将遥感影像分类为水体、密林、农业用地、建成区和裸地。使用210个验证样本和多种混淆矩阵指标评估模型性能，包括灵敏度、精确率、特异度、阴性预测值、总体精度、平衡精度和马修斯相关系数。农业用地估计值从SVM的1,000.26 km²（21.78%）到ANN的1,099.12 km²（23.93%）不等。SVM取得了最强的总体表现，灵敏度为88.5%，精确率为85.2%，特异度为75.0%，总体精度为83.3%，平衡精度为81.7%，马修斯相关系数为0.643。CNN的总体精度最低，为74.3%。研究结果表明，模型复杂度的增加并不必然带来更优的分类性能，并确立了SVM作为异质性热带景观条件下区域农业用地划定的有效方法。关键词：农业用地制图；机器学习；深度学习；支持向量机；随机森林；人工神经网络；卷积神经网络；遥感","International Journal of Emerging Trends in Engineering and Development",25,61,{"impact":123,"substance":124,"depth":17,"authority":123,"freshness":21,"relevant":22,"comment":490},"尼日利亚区域案例研究，方法对比扎实但影响范围限于学术圈，可作为遥感与农业AI主题的参考素材。",[492,493],{"name":486,"url":483},{"name":486,"url":494},"https:\u002F\u002Fdoi.org\u002F10.5281\u002Fzenodo.23106954",[28,441,29,30,496],"农地制图",{"code":49,"slug":246,"name":247,"path":248,"parent_code":141},[],{"status":143,"locked":14,"version":35,"method":36,"evidence":500},{"candidates":501,"agricultural_evidence":42},[502,508,512,518,524],{"code":49,"score":503,"supported":42,"title":504,"abstract":506,"tags":507},27,[51,505],"deep learning",[453,29,51,505],[441,29],{"code":55,"score":276,"supported":42,"title":509,"abstract":510,"tags":511},[],[30,57],[30],{"code":513,"score":62,"supported":42,"title":514,"abstract":515,"tags":517},"A8",[],[516],"粮食安全",[],{"code":519,"score":62,"supported":42,"title":520,"abstract":521,"tags":523},"J1",[],[522],"农业规划",[],{"code":74,"score":22,"supported":14,"title":525,"abstract":526,"tags":527},[],[],[28],[529,530],"农业人工智能 农地制图 智慧农业 机器学习","农业人工智能 农地制图","农业人工智能农地制图智慧农业机器学习-4238","10.5281\u002Fzenodo.23106955",{"doi":532,"openalex_id":534,"authors":535,"venue":486,"cited_by_count":84,"oa_url":483,"card":538,"direction":106,"ingested_from":108},"W7219591984",[536],{"name":537,"orcid":9},"J.I. Igbokwe, C.E Ejikeme, J.O and Obi",{"tldr":539,"method":540,"finding":541,"direction":106,"opportunity":542},"比较RF、SVM、ANN、CNN在尼日利亚农业用地制图中的表现","遥感影像分类，210个验证样本，多种混淆矩阵指标评估","SVM总体精度最高（83.3%），CNN最低（74.3%），模型复杂度不保证性能","在碎片化热带景观中，可探索轻量级集成学习或迁移学习以平衡精度与效率","2026-10-05T23:30:27.987196Z"]