[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"item-3324":3,"related-3324":46},{"id":4,"title":5,"url":6,"summary":7,"summary_zh":8,"content":8,"source_name":9,"source_url":8,"published_at":10,"category":11,"cover_url":8,"hotness":12,"is_selected":13,"score":14,"score_detail":15,"sources":22,"tags":24,"search_phrases":31,"slug":34,"view_count":35,"doi":8,"paper":36,"created_at":45},3324,"Diag-STFN：全球收获前作物产量预测的诊断时空多模态融合网络——覆盖38国玉米29国小麦（Ecological Informatics 2026）","https:\u002F\u002Fm2.mtmt.hu\u002Fapi\u002Fpublication\u002F37354974?&&labelLang=hun","《Ecological Informatics》2026年第96期：Zhuang等提出Diag-STFN——一种诊断时空多模态融合网络，用于全球收获前作物产量预测。该网络基于数据集特征选择模型结构，以确定是否需要时间趋势耦合和空间模块激活。在三种前置期（早、中、晚季）下，基于覆盖38国玉米和29国小麦的CY-Bench基准数据集进行评估。结果表明，所提方法在所有前置期均实现了两种作物的最低汇总NRMSE，并在MAPE和KGE等补充指标上保持领先。消融研究表明诊断模块选择提供了主要的性能提升；方差分解显示性能差异在国家之间大于模型之间。",null,"《Ecological Informatics》96 (2026) 103860","2026-09-17T00:00:00Z","论文",10,false,78,{"impact":16,"substance":17,"depth":16,"authority":18,"freshness":19,"relevant":20,"comment":21},18,22,14,6,1,"方法新颖、覆盖38国玉米与29国小麦的全球收获前产量预测研究，学术价值突出但产业落地尚早，适合作为前沿技术资讯收录。",[23],{"name":9,"url":6},[25,26,27,28,29,30],"智慧农业","农业人工智能","产量预测","小麦","玉米","遥感",[32,33],"Diag-STFN 作物产量预测","CY-Bench 玉米 小麦","Diag-STFN作物产量预测-3324",0,{"doi":8,"openalex_id":8,"authors":37,"venue":8,"cited_by_count":35,"oa_url":8,"card":38,"direction":42,"ingested_from":44},[],{"tldr":39,"method":40,"finding":41,"direction":42,"opportunity":43},"提出诊断式时空多模态融合网络Diag-STFN，实现全球收获前玉米小麦产量预测。","基于CY-Bench基准，按数据特征诊断选择时间趋势与空间模块，覆盖38国玉米2","各前置期均取得最低NRMSE，诊断模块选择贡献最大，国家间差异大于模型间差异。","农业人工智能与决策模型","可探索自适应诊断机制迁移至其他作物，并针对国家间差异开展区域化建模与不确定性量化。","agent","2026-09-24T00:04:02.684732Z",{"total":19,"page":20,"page_size":19,"items":47},[48,104,146,176,227,271],{"id":49,"title":50,"url":51,"summary":52,"summary_zh":53,"content":8,"source_name":54,"source_url":51,"published_at":55,"category":11,"cover_url":8,"hotness":12,"is_selected":13,"score":56,"score_detail":57,"sources":60,"tags":62,"search_phrases":63,"slug":66,"view_count":35,"doi":67,"paper":68,"created_at":103},2016,"Transfer learning integrated with SCOPE model for maize yield prediction based on spectral data for different spatial scales","https:\u002F\u002Fdoi.org\u002F10.1016\u002Fj.compag.2026.112381","Transfer learning integrated with SCOPE model for maize yield prediction based on spectral data for different spatial scales。Computers and Electronics in Agriculture","基于光谱数据的不同空间尺度玉米产量预测中，将迁移学习与SCOPE模型相结合","Computers and Electronics in Agriculture","2026-09-09T00:00:00Z",80,{"impact":16,"substance":17,"depth":16,"authority":18,"freshness":58,"relevant":20,"comment":59},8,"将迁移学习与SCOPE辐射传输模型结合，实现跨空间尺度光谱数据的玉米产量预测，方法新颖且对遥感估产有实用价值，值得进入每日精选。",[61],{"name":54,"url":51},[25,26,27,29,30],[64,65],"农业人工智能 产量预测 智慧农业 玉米","农业人工智能 产量预测","农业人工智能产量预测智慧农业玉米-2016","10.1016\u002Fj.compag.2026.112381",{"doi":67,"openalex_id":69,"authors":70,"venue":54,"cited_by_count":35,"oa_url":8,"card":96,"direction":100,"ingested_from":102},"W7211999297",[71,73,76,79,81,83,85,88,90,92,94],{"name":72,"orcid":8},"Ruomei Zhao",{"name":74,"orcid":75},"Yanling Zhang","https:\u002F\u002Forcid.org\u002F0000-0002-2323-2600",{"name":77,"orcid":78},"Weimin Guo","https:\u002F\u002Forcid.org\u002F0000-0003-4318-7154",{"name":80,"orcid":8},"Wei Liu",{"name":82,"orcid":8},"Long Zhao",{"name":84,"orcid":8},"Xiaoyuan Tian",{"name":86,"orcid":87},"Yelu Zeng","https:\u002F\u002Forcid.org\u002F0000-0003-4267-1841",{"name":89,"orcid":8},"Lulu An",{"name":91,"orcid":8},"Aiguo Wang",{"name":93,"orcid":8},"Qiang Xu",{"name":95,"orcid":8},"Hong Sun",{"tldr":97,"method":98,"finding":99,"direction":100,"opportunity":101},"结合迁移学习与SCOPE模型，利用不同空间尺度光谱数据预测玉米产量。","迁移学习与SCOPE辐射传输模型，多尺度光谱数据。","该方法能跨空间尺度提升玉米产量预测精度。","农业遥感与作物表型","可探索迁移学习在跨区域、跨年份及多作物产量预测中的泛化能力。","openalex","2026-09-10T23:30:01.419872Z",{"id":105,"title":106,"url":107,"summary":108,"summary_zh":109,"content":8,"source_name":110,"source_url":107,"published_at":111,"category":11,"cover_url":8,"hotness":12,"is_selected":13,"score":112,"score_detail":113,"sources":118,"tags":120,"search_phrases":122,"slug":125,"view_count":35,"doi":126,"paper":127,"created_at":145},3177,"A comparative analysis on maize yield prediction using sentinel 2A and Landsat 8 satellite image in Sundarganj, Gaibandha, Bangladesh","https:\u002F\u002Fdoi.org\u002F10.3329\u002Fbjar.v51i1.92530","Maize is an important cereal crops in Bangladesh. Over the last two decades, its cultivation has increased promisingly, especially in the Northern part of Bangladesh. The effective estimation of crop yields at a regional scale holds significant importance in facilitating decision-making within the agricultural sector, thereby ensuring grain security. The traditional ground-based measurement techniques suffer from inefficiencies, and there exists a need for a reliable, precise, and effective method for estimating regional crop yields. This study used remote sensing (RS) techniques for forecasting pre-harvest maize yield to improve the management system. Currently, the normalized difference vegetation index (NDVI) is widely used to predict crop yield including maize. However, the present study used Landsat 8 (~ 30 m) and Sentinel 2A (~ 10 m) high resolution data for 2018-2019 and 2019-2020 to predict maize yield based on the year 2020-2021 at Sundarganj Upazila in Gaibandha district. The single cloud free image acquisition date based on maximum NDVI for both satellite images was used for each maize growing period to develop a yield prediction model. A regression model was performed between NDVI values and 20 farmers field-level maize yields. The absolute mean error of prediction was about 10.30% for Landsat 8 and 6.70% for Sentinel 2A compared to the actual maize yield during 2020-2021. The study revealed that NDVI data extracted from Sentinel 2A high resolution satellite images can be successfully used to predict the maize yield with appreciable accuracy. Finally, this study has demonstrated the efficacy of combining multi-temporal remote sensing data for accurate maize yield estimation, aiding agricultural authorities and production enterprises in the timely formulation and refinement of cropping strategies and management policies for the ongoing season. Bangladesh J. Agril. Res. 51(1): 501-521, March 2026","玉米是孟加拉国重要的谷类作物。过去二十年间，其种植面积增长显著，尤其是在孟加拉国北部地区。在区域尺度上有效估算作物产量对于促进农业部门决策、进而保障粮食安全具有重要意义。传统的地面测量技术效率低下，亟需一种可靠、精确且有效的区域作物产量估算方法。本研究采用遥感（RS）技术预测收获前玉米产量，以改进管理体系。目前，归一化植被指数（NDVI）被广泛用于预测包括玉米在内的作物产量。然而，本研究利用Landsat 8（约30 m）和Sentinel 2A（约10 m）高分辨率数据，基于2018—2019年和2019—2020年的数据，对盖班达县孙达尔甘杰乌帕齐拉2020—2021年的玉米产量进行预测。在每个玉米生长期，选取两颗卫星影像中NDVI最大值对应的单幅无云影像获取日期，用于建立产量预测模型。对NDVI值与20户农民田块级玉米产量进行回归建模。与2020—2021年实际玉米产量相比，Landsat 8的绝对平均预测误差约为10.30%，Sentinel 2A约为6.70%。研究表明，利用Sentinel 2A高分辨率卫星影像提取的NDVI数据可成功用于预测玉米产量，且精度令人满意。最后，本研究证明了结合多时相遥感数据进行准确玉米产量估算的有效性，有助于农业主管部门和生产企业在当季及时制定和完善种植策略与管理政策。Bangladesh J. Agril. Res. 51(1): 501-521, March 2026","Bangladesh Journal of Agricultural Research","2026-09-21T00:00:00Z",65,{"impact":58,"substance":114,"depth":115,"authority":116,"freshness":58,"relevant":20,"comment":117},20,16,13,"基于Sentinel 2A与Landsat 8的玉米遥感估产对比研究，方法清晰、误差数据具体，对遥感估产有参考价值，但属区域小尺度研究，公共影响有限。",[119],{"name":110,"url":107},[25,27,29,30,121],"NDVI",[123,124],"Sentinel 2A Landsat 8 玉米产量预测","孟加拉国 Sundarganj 玉米遥感估产","Sentinel2ALandsat8玉米产量预测-3177","10.3329\u002Fbjar.v51i1.92530",{"doi":126,"openalex_id":128,"authors":129,"venue":110,"cited_by_count":35,"oa_url":107,"card":140,"direction":100,"ingested_from":102},"W7213918280",[130,132,134,136,138],{"name":131,"orcid":8},"N Mohammad",{"name":133,"orcid":8},"MA Islam",{"name":135,"orcid":8},"MG Mahboob",{"name":137,"orcid":8},"MM Rahman",{"name":139,"orcid":8},"I Ahmed",{"tldr":141,"method":142,"finding":143,"direction":100,"opportunity":144},"用Sentinel 2A与Landsat 8的NDVI回归模型预测孟加拉国玉米产量并比较精度。","基于最大NDVI单期影像与20个农户地块产量做回归，比较两种卫星。","Sentinel 2A预测绝对平均误差6.70%，优于Landsat 8的10.30%。","可探索多时相NDVI与机器学习融合，提升小农户尺度玉米估产精度与迁移性。","2026-09-22T23:30:23.551878Z",{"id":147,"title":148,"url":149,"summary":150,"summary_zh":8,"content":8,"source_name":151,"source_url":149,"published_at":152,"category":11,"cover_url":8,"hotness":12,"is_selected":13,"score":153,"score_detail":154,"sources":157,"tags":159,"search_phrases":161,"slug":164,"view_count":35,"doi":165,"paper":166,"created_at":175},3152,"An integrated IoT and machine learning framework for maize yield prediction and smart agriculture in Tanzania","https:\u002F\u002Fdoi.org\u002F10.1080\u002F23311932.2026.2725400","An integrated IoT and machine learning framework for maize yield prediction and smart agriculture in Tanzania。Cogent Food & Agriculture","Cogent Food & Agriculture","2026-09-22T00:00:00Z",69,{"impact":155,"substance":16,"depth":115,"authority":116,"freshness":12,"relevant":20,"comment":156},12,"论文提出物联网与机器学习融合的玉米产量预测框架，方法有新意但属区域性应用研究，影响力有限，时效性强。",[158],{"name":151,"url":149},[25,26,27,160,29],"物联网",[162,163],"坦桑尼亚 玉米 产量预测","物联网 机器学习 智慧农业","坦桑尼亚玉米产量预测-3152","10.1080\u002F23311932.2026.2725400",{"doi":165,"openalex_id":167,"authors":168,"venue":151,"cited_by_count":35,"oa_url":149,"card":8,"direction":174,"ingested_from":102},"W7213978365",[169,172],{"name":170,"orcid":171},"Alcardo Alex Barakabitze","https:\u002F\u002Forcid.org\u002F0000-0001-8960-8415",{"name":173,"orcid":8},"Yasinta Nzogera","智慧农业 \u002F 农业物联网","2026-09-22T23:30:10.178097Z",{"id":177,"title":178,"url":179,"summary":180,"summary_zh":181,"content":8,"source_name":54,"source_url":179,"published_at":182,"category":11,"cover_url":8,"hotness":12,"is_selected":13,"score":183,"score_detail":184,"sources":187,"tags":189,"search_phrases":191,"slug":194,"view_count":35,"doi":195,"paper":196,"created_at":226},3006,"Explainable spectral–image fusion multi-task learning for maize canopy biochemical and structural trait retrieval","https:\u002F\u002Fdoi.org\u002F10.1016\u002Fj.compag.2026.112450","Accurate and interpretable estimation of multiple maize canopy traits is important for proximal crop monitoring and high-throughput phenotyping. This study developed a spectral–image fusion multi-task learning framework (MM-MTL) for the simultaneous retrieval of chlorophyll index (Chl index), leaf area index (LAI), nitrogen balance index (NBI), and anthocyanin index (Anth index). Proximal hyperspectral observations were collected using a Specim IQ camera covering 400–1000 nm with 204 spectral bands. For each observation, an ROI-mean spectral vector was extracted to retain fine-grained canopy reflectance information, while a co-registered pseudo-RGB image derived from visible bands of the same hyperspectral cube preserved two-dimensional canopy structure and visible appearance. MM-MTL integrates spectral and image feature extraction, task-specific fusion, and uncertainty-weighted multi-task learning to jointly estimate the four traits. A total of 1,023 valid samples collected from nine field campaigns across the 2024 and 2025 growing seasons were used for model development and evaluation. Under plot-grouped five-fold cross-validation, MM-MTL achieved R 2 values of 0.872, 0.885, 0.722, and 0.807 for Chl index, LAI, NBI, and Anth index, respectively, and consistently outperformed the single-task, single-representation, and conventional regression baselines. Performance decreased under more challenging generalization settings, with R 2 values ranging from 0.543 to 0.680 under leave-one-campaign-out validation and from 0.425 to 0.640 under bidirectional cross-year validation. Ablation and post-hoc analyses further showed that preserving image spatial organization improved prediction, while task-wise routing, input-representation masking, wavelength perturbation, and cross-fold stability analyses revealed trait-dependent use of spectral and spatial information. These results demonstrate that complementary spectral and spatial representations derived from the same hyperspectral observation can improve multi-trait maize canopy retrieval, while transfer across acquisition campaigns, years, and field environments remains an important direction for further improvement.","准确且可解释地估算多个玉米冠层性状，对于近地作物监测和高通量表型分析具有重要意义。本研究开发了一种光谱—图像融合多任务学习框架（MM-MTL），用于同时反演叶绿素指数（Chl index）、叶面积指数（LAI）、氮平衡指数（NBI）和花青素指数（Anth index）。近地高光谱观测使用Specim IQ相机采集，覆盖400–1000 nm，共204个光谱波段。对于每次观测，提取ROI均值光谱向量以保留细粒度冠层反射率信息，同时从同一高光谱立方体的可见光波段生成配准的伪RGB图像，以保留二维冠层结构和可见外观。MM-MTL集成了光谱与图像特征提取、任务特定融合以及不确定性加权多任务学习，以联合估算这四种性状。研究使用2024年和2025年生长季9次田间试验采集的共计1,023个有效样本进行模型开发与评估。在按小区分组的五折交叉验证下，MM-MTL对Chl index、LAI、NBI和Anth index的R²分别为0.872、0.885、0.722和0.807，且持续优于单任务、单表征和传统回归基线。在更具挑战性的泛化设置下，模型性能有所下降，留一试验验证的R²范围为0.543–0.680，双向跨年验证的R²范围为0.425–0.640。消融分析和事后分析进一步表明，保留图像空间组织可提升预测性能，而任务路由、输入表征掩蔽、波长扰动和跨折稳定性分析揭示了光谱与空间信息的性状依赖性利用方式。这些结果表明，从同一高光谱观测中提取的互补光谱与空间表征可改善多性状玉米冠层反演，而跨采集试验、年份和田间环境的迁移仍是未来改进的重要方向。","2026-09-19T00:00:00Z",81,{"impact":16,"substance":17,"depth":16,"authority":18,"freshness":185,"relevant":20,"comment":186},9,"提出光谱-图像融合多任务学习框架，1023份样本跨两年验证，方法新颖且结论可靠，对作物高通量表型研究有实质参考价值。",[188],{"name":54,"url":179},[25,26,29,30,190],"高通量表型",[192,193],"玉米冠层 多任务学习 高光谱","Specim IQ 玉米 表型","玉米冠层多任务学习高光谱-3006","10.1016\u002Fj.compag.2026.112450",{"doi":195,"openalex_id":197,"authors":198,"venue":54,"cited_by_count":35,"oa_url":179,"card":221,"direction":100,"ingested_from":102},"W7213658705",[199,201,203,205,206,207,209,212,214,216,218],{"name":200,"orcid":8},"Penglei Zhang",{"name":202,"orcid":8},"Tianbo Hao",{"name":204,"orcid":8},"Zhuoyuan Zhao",{"name":95,"orcid":8},{"name":86,"orcid":87},{"name":208,"orcid":8},"Zheng Cui",{"name":210,"orcid":211},"Ta Na","https:\u002F\u002Forcid.org\u002F0000-0002-1348-5655",{"name":213,"orcid":8},"Lang Qiao",{"name":215,"orcid":8},"Durval Dourado Neto",{"name":217,"orcid":8},"Feng Yang",{"name":219,"orcid":220},"Jingzhu Wu","https:\u002F\u002Forcid.org\u002F0000-0002-8386-1038",{"tldr":222,"method":223,"finding":224,"direction":100,"opportunity":225},"提出光谱-图像融合多任务学习框架，同时反演玉米冠层四个生化与结构性状。","Specim IQ高光谱与伪RGB融合，不确定性加权多任务学习，1023样本交叉","融合模型精度优于单任务基线，但跨年份与跨环境泛化性能明显下降。","可研究跨年份\u002F跨环境域适应与迁移学习，提升多性状反演泛化能力。","2026-09-20T23:30:01.739142Z",{"id":228,"title":229,"url":230,"summary":231,"summary_zh":232,"content":8,"source_name":233,"source_url":230,"published_at":234,"category":11,"cover_url":8,"hotness":12,"is_selected":13,"score":56,"score_detail":235,"sources":237,"tags":239,"search_phrases":241,"slug":244,"view_count":35,"doi":245,"paper":246,"created_at":270},2666,"Automated Machine Learning-Driven UAV Remote Sensing for Accurate Winter Wheat Water Content Prediction","https:\u002F\u002Fdoi.org\u002F10.3390\u002Frs18183161","Crop water content is a critical indicator of crop growth status, and its efficient and accurate monitoring is essential for agricultural water resource management. Conventional methods for monitoring winter wheat water content, however, rely mainly on destructive sampling and are labor-intensive and time-consuming. To address these limitations, this study explored the potential of unmanned aerial vehicle (UAV) remote sensing for the rapid and accurate assessment of winter wheat water content. High-resolution canopy remote sensing images were acquired using UAVs equipped with multispectral (MS), RGB, and thermal infrared (TIR) cameras during the flowering and filling stages under six irrigation treatments. Ground-truth sampling data were integrated with the UAV-derived remote sensing data, and an automated machine learning (AutoML) framework—which automatically searches over a range of candidate algorithms and hyperparameters to select the optimal model—was employed to establish regression models for predicting winter wheat moisture content (MC). All models were evaluated using five-fold cross-validation. The results demonstrated that MC prediction performed best during the filling stage, with the TIR sensor achieving the highest accuracy (R2 = 0.812, MAE = 0.0204, RMSE = 0.0274). Compared with single-sensor approaches, multi-sensor fusion further improved predictive performance, achieving an R2 of 0.876, an MAE of 0.0191, and an RMSE of 0.0259 for MC prediction. These findings indicate that UAV-based multi-sensor remote sensing provides an effective means of monitoring winter wheat water content, facilitating timely assessment of crop growth status and optimized irrigation management. Moreover, the use of AutoML enables high-accuracy prediction with minimal human intervention, enhancing the precision of crop water monitoring and advancing precision agriculture.","作物含水量是反映作物生长状况的关键指标，对其进行高效、准确的监测对农业水资源管理至关重要。然而，传统冬小麦含水量监测方法主要依赖破坏性采样，费时费力。为解决这些局限，本研究探索了无人机（UAV）遥感在快速准确评估冬小麦含水量方面的潜力。在六种灌溉处理下，利用搭载多光谱（MS）、RGB和热红外（TIR）相机的无人机在开花期和灌浆期获取了高分辨率冠层遥感图像。将地面实测采样数据与无人机遥感数据相结合，采用自动化机器学习（AutoML）框架——该框架可在一系列候选算法和超参数中自动搜索以选择最优模型——建立预测冬小麦含水量（MC）的回归模型。所有模型均采用五折交叉验证进行评估。结果表明，灌浆期MC预测表现最佳，其中TIR传感器精度最高（R2 = 0.812，MAE = 0.0204，RMSE = 0.0274）。与单传感器方法相比，多传感器融合进一步提升了预测性能，MC预测的R2达到0.876，MAE为0.0191，RMSE为0.0259。这些发现表明，基于无人机的多传感器遥感为监测冬小麦含水量提供了有效手段，有助于及时评估作物生长状况并优化灌溉管理。此外，AutoML的使用使得在最少人工干预下实现高精度预测成为可能，提升了作物水分监测的精度，推动了精准农业发展。","Remote Sensing","2026-09-15T00:00:00Z",{"impact":16,"substance":17,"depth":16,"authority":18,"freshness":58,"relevant":20,"comment":236},"AutoML结合无人机多传感器遥感预测冬小麦含水量，方法新颖、数据扎实，对精准灌溉有实用价值，值得进入每日精选。",[238],{"name":233,"url":230},[25,26,28,30,240],"精准灌溉",[242,243],"农业人工智能 智慧农业 精准灌溉 小麦","农业人工智能 智慧农业","农业人工智能智慧农业精准灌溉小麦-2666","10.3390\u002Frs18183161",{"doi":245,"openalex_id":247,"authors":248,"venue":233,"cited_by_count":35,"oa_url":230,"card":265,"direction":100,"ingested_from":102},"W7213246708",[249,252,254,256,259,262],{"name":250,"orcid":251},"Fan Ding","https:\u002F\u002Forcid.org\u002F0000-0001-5482-8290",{"name":253,"orcid":8},"Qian Cheng",{"name":255,"orcid":8},"Fuyi Duan",{"name":257,"orcid":258},"Shuaipeng Fei","https:\u002F\u002Forcid.org\u002F0000-0002-8774-7929",{"name":260,"orcid":261},"Junjie Feng","https:\u002F\u002Forcid.org\u002F0000-0001-8900-2691",{"name":263,"orcid":264},"Zhen Chen","https:\u002F\u002Forcid.org\u002F0000-0002-2847-0042",{"tldr":266,"method":267,"finding":268,"direction":100,"opportunity":269},"用无人机多光谱、RGB和热红外遥感结合AutoML预测冬小麦含水量。","无人机多传感器影像与地面采样，AutoML自动选模型，五折交叉验证。","灌浆期热红外精度最高R²=0.812，多传感器融合提升至R²=0.876。","可探索AutoML与多时相\u002F多源卫星遥感融合，实现区域尺度作物水分精准监测。","2026-09-16T23:30:29.163192Z",{"id":272,"title":273,"url":274,"summary":275,"summary_zh":276,"content":8,"source_name":277,"source_url":274,"published_at":278,"category":11,"cover_url":8,"hotness":12,"is_selected":13,"score":279,"score_detail":280,"sources":283,"tags":285,"search_phrases":287,"slug":289,"view_count":35,"doi":290,"paper":291,"created_at":325},2505,"A continually evolving knowledge-guided deep learning framework for daily maize yield formation","https:\u002F\u002Fdoi.org\u002F10.1016\u002Fj.agsy.2026.104981","CONTEXT As multi-source agricultural datasets expand, accurate and interpretable simulation of crop yield formation is increasingly feasible and important for food security. Process-based models are widely used for yield forecasting but require laborious regional calibration. Artificial intelligence models learn physiological patterns from large datasets but lack physiological consistency and generalize poorly under data-scarce conditions. OBJECTIVE We developed AgroEvoDeep-Yield, a knowledge-guided model that encodes mechanistic understanding through its neural architecture, enabling explicit and interpretable simulation of daily maize yield formation from flowering to maturity. We further established a framework that supports continual model evolution as new data become available. METHODS The model was pretrained on APSIM-simulated grain number and daily grain biomass sequences to impose physiological consistency. Following pretraining, the model evolved through two continual learning stages. First, field-level continual learning using 796 field observations reduced biases inherited from APSIM. The model was then adapted to regional-scale multi-source inputs using long-term records from 945 meteorological stations and expert-constrained cultivar-parameter optimization. RESULTS AND CONCLUSIONS After field-level continual learning, AgroEvoDeep-Yield reduced grain number and yield simulation errors by 29% and 24%, respectively, relative to APSIM, and outperformed conventional machine learning models in out-of-region prediction. Region-level continual learning reduced yield RMSE to 1.09 t\u002Fha, and expert-constrained cultivar-parameter optimization further reduced it to 0.80 t\u002Fha. SIGNIFICANCE This study demonstrates how a knowledge-guided model can be repeatedly adapted as new data become available, providing a practical framework for continual model evolution and a basis for future agricultural decision support.","背景 随着多源农业数据集的不断扩展，准确且可解释地模拟作物产量形成过程日益可行，对粮食安全也愈发重要。基于过程的模型广泛用于产量预测，但需要耗费大量精力进行区域校准。人工智能模型能够从大规模数据集中学习生理模式，但缺乏生理一致性，在数据稀缺条件下泛化能力较差。目标 我们开发了AgroEvoDeep-Yield，这是一种知识引导模型，通过其神经网络架构编码机理认识，能够对玉米从开花到成熟的逐日产量形成过程进行显式且可解释的模拟。我们还建立了一个框架，以支持在新数据可用时模型的持续演化。方法 该模型在APSIM模拟的粒数和逐日籽粒生物量序列上进行预训练，以施加生理一致性约束。预训练后，模型通过两个持续学习阶段进行演化。首先，利用796个田间观测数据进行田块级持续学习，以减少继承自APSIM的偏差。随后，利用945个气象站的长期记录和专家约束的品种参数优化，将模型适配到区域尺度多源输入。结果与结论 经过田块级持续学习后，AgroEvoDeep-Yield相对于APSIM将粒数和产量模拟误差分别降低了29%和24%，并在区域外预测中优于传统机器学习模型。区域级持续学习将产量RMSE降至1.09 t\u002Fha，专家约束的品种参数优化进一步将其降至0.80 t\u002Fha。意义 本研究表明，知识引导模型可以在新数据可用时被反复适配，为模型持续演化提供了实用框架，并为未来农业决策支持奠定了基础。","Agricultural Systems","2026-09-14T00:00:00Z",85,{"impact":17,"substance":281,"depth":16,"authority":18,"freshness":58,"relevant":20,"comment":282},23,"知识引导的深度学习框架实现玉米逐日产量形成模拟，误差显著低于APSIM且支持持续学习，方法新颖、数据规模扎实，对智慧农业决策支持有参考价值。",[284],{"name":277,"url":274},[25,26,27,29,286],"作物模型",[288,65],"农业人工智能 产量预测 作物模型 智慧农业","农业人工智能产量预测作物模型智慧农业-2505","10.1016\u002Fj.agsy.2026.104981",{"doi":290,"openalex_id":292,"authors":293,"venue":277,"cited_by_count":35,"oa_url":274,"card":320,"direction":42,"ingested_from":102},"W7212532256",[294,296,298,301,304,307,309,312,314,317],{"name":295,"orcid":8},"Junji Ou",{"name":297,"orcid":8},"Wenyao Yan",{"name":299,"orcid":300},"Fangzheng Chen","https:\u002F\u002Forcid.org\u002F0009-0006-5061-6227",{"name":302,"orcid":303},"Tao Ye","https:\u002F\u002Forcid.org\u002F0000-0002-5037-8410",{"name":305,"orcid":306},"Ke Liu","https:\u002F\u002Forcid.org\u002F0000-0002-8343-0449",{"name":308,"orcid":8},"Matthew Tom Harrison",{"name":310,"orcid":311},"William D. Batchelor","https:\u002F\u002Forcid.org\u002F0000-0002-3881-6246",{"name":313,"orcid":8},"Yong Chen",{"name":315,"orcid":316},"Kelin Hu","https:\u002F\u002Forcid.org\u002F0000-0001-9321-0821",{"name":318,"orcid":319},"Puyu Feng","https:\u002F\u002Forcid.org\u002F0000-0003-4845-9876",{"tldr":321,"method":322,"finding":323,"direction":42,"opportunity":324},"提出知识引导的深度学习框架AgroEvoDeep-Yield，实现玉米逐日产量形成的可解释模拟与持续","基于APSIM模拟数据预训练，结合796个田间观测和945个气象站数据持续学习。","田间持续学习使粒数和产量误差比APSIM降低29%和24%，区域优化后RMSE降至0.80 t\u002Fha","可探索将知识引导与持续学习框架迁移至其他作物，并融合实时遥感与物联网数据实现动态产量预测。","2026-09-15T23:30:04.611750Z"]