| 133 | 0 | 42 |
| 下载次数 | 被引频次 | 阅读次数 |
及时准确地预测冬小麦产量对保障国家粮食安全、促进农业可持续发展与实施乡村振兴战略具有重要意义。传统模型难以揭示时序动态特征与产量的非线性复杂关系，本文融合时序遥感、气象与土壤多源数据，构建LSTM-Attention-XGBoost集成学习框架，实现河南省县域尺度冬小麦产量预测，并结合SHAP(SHapley Additive exPlanations)方法解析特征驱动机制与空间分异特征。结果表明：所构建模型对冬小麦产量估算精度(R2=0.875,RMSE=0.430 t/hm2,MAE=0.340 t/hm2)相比LSTM、LSTM-Attention和LSTM-XGBoost分别提升68.7%、15.3%和18.9%;特征贡献呈现“土壤属性—植被指数—气象要素”的重要性排序；模型在拔节—抽穗阶段达到最佳预测性能(R2=0.918,RMSE=0.348 t/hm2,MAE=0.302 t/hm2),可提前64 d实现高精度估产；县域尺度产量空间异质性显著，高产区主要位于豫东、豫北平原，中低产区集中于豫中平原向豫南丘陵及西部山地过渡带，整体格局多年相对稳定，过渡带年际波动明显。本文研究框架可提升多源时序数据条件下的冬小麦产量预测能力，并从模型解释拓展至空间格局解析，为区域尺度粮食管理、预警与精准农业决策提供支持。
Abstract：Timely and accurate prediction of winter wheat yield is crucial for ensuring national food security, promoting sustainable agricultural development, and supporting rural revitalization strategies.To address the limitations of traditional models in capturing temporal dynamics and the complex nonlinear relationships of winter wheat yield, this study integrates multi-source data, including time-series remote sensing, meteorological variables, and soil properties, and develops an ensemble learning framework based on LSTM-Attention-XGBoost for county-level winter wheat yield prediction in Henan Province.Furthermore, SHapley Additive exPlanations(SHAP) is employed to interpret feature contributions and reveal the underlying driving mechanisms as well as the spatial differentiation of yield.The findings are as follows.(1) The proposed model achieves the highest accuracy among all compared models, with an R2 of 0.875,RMSE of 0.430 t/hm2,and MAE of 0.340 t/hm2,representing improvements of 68.7%,15.3%,and 18.9% over the LSTM,LSTM-Attention, and LSTM-XGBoost models, respectively.(2) The feature contributions follow a clear hierarchy of "soil attributes > vegetation indices > meteorological factors".(3) The model achieves optimal predictive performance during the jointing-heading stage(R2= 0.918,RMSE = 0.348 t/hm2,MAE = 0.302 t/hm2),enabling high-accuracy yield estimation approximately 64 days prior to harvest.(4) The winter wheat yield exhibits significant spatial heterogeneity at the county scale, with high-yield areas mainly distributed in the eastern and northern plains of Henan Province, while medium-and low-yield areas are concentrated in the transitional zones extending from the central plains to the southern hilly areas and the western mountainous areas.The overall spatial pattern remains relatively stable across years, with notable interannual variability in the transitional zones.Overall, the proposed framework enhances the predictive capability for winter wheat yield under multi-source temporal data conditions and extends model interpretation to spatial pattern analysis, thereby providing effective support for regional-scale grain management, early warning, and precision agricultural decision-making.
[1] Huang Xianda,Huang Jianxi,Li Xuecao,et al.Early mapping of winter wheat in Henan province of China using time series of Sentinel-2 data[J].GIScience & Remote Sensing,2022,59(1):1534-1549.
[2] 杨亚东，段丁丁，巨章宏，等.基于FAO-GAEZ模型的中国马铃薯县域种植空间布局优化[J].地理研究，2022,41(12):3352-3363.
[3] 刘海猛，卢佳祎，王成新，等.基于栅格尺度的黄河流域人地关系时空耦合分析[J].生态学报，2024,44(15):6499-6512.
[4] 苏伟锋，王华.区域环境污染对社会道德失范评价的影响：基于具身认知视角的实证研究[J].地理与地理信息科学，2024,40(2):81-88.
[5] 胡伟，刘嘉宇，周琛.基于多源数据融合的城市细分地类识别与可解释性分析[J].地理与地理信息科学，2025,41(6):26-32.
[6] 王来刚，郑国清，郭燕，等.融合多源时空数据的冬小麦产量预测模型研究[J].农业机械学报，2022,53(1):198-204.
[7] Chen Lili,Zhao Hongsheng,Song Ge,et al.Optimization of cultivated land pattern for achieving cultivated land system security:a case study in Heilongjiang Province,China[J].Land Use Policy,2021,108:105589.DOI:10.1016/j.landusepol.2021.105589.
[8] Ding Tianran,Steubing B,Achten W M J.Coupling optimization with territorial LCA to support agricultural land-use planning[J].Journal of Environmental Management,2023,328:116946.DOI:10.1016/j.jenvman.2022.116946.
[9] 刘彦随，冯巍仑，李裕瑞.现代农业地理工程与农业高质量发展：以黄土丘陵沟壑区为例[J].地理学报，2020,75(10):2029-2046.
[10] 屠爽爽，简代飞，龙花楼，等.广西主要农作物生产格局演变特征与机制研究[J].地理学报，2022,77(9):2322-2337.
[11] 岳海旺，魏建伟，刘朋程，等.基于AMMI模型和GGE双标图对玉米品种产量稳定性和适应性选择评价[J].中国农业大学学报，2024,29(9):24-37.
[12] Desloires J,Ienco D,Botrel A.Out-of-year corn yield prediction at field-scale using Sentinel-2 satellite imagery and machine learning methods[J].Computers and Electronics in Agriculture,2023,209:107807.DOI:10.1016/j.compag.2023.107807.
[13] 朱炯，杜鑫，李强子，等.关键时相长势—环境和景观特征对河北省县级尺度冬小麦单产估算精度影响分析[J].遥感学报，2022,26(7):1354-1367.
[14] Tian Huiren,Wang Pengxin,Tansey K,et al.A deep learning framework under attention mechanism for wheat yield estimation using remotely sensed indices in the Guanzhong Plain,PR China[J].International Journal of Applied Earth Observations and Geoinformation,2021,102:102375.DOI:10.1016/j.jag.2021.102375.
[15] 王旭，刘波，陈正超，等.基于多源数据和LSTM模型的县域冬小麦估产[J].农业现代化研究，2023,44(6):1117-1126.
[16] Zhang Haiyang,Zhang Yao,Liu Kaidi,et al.Winter wheat yield prediction using integrated Landsat 8 and Sentinel-2 vegetation index time-series data and machine learning algorithms[J].Computers and Electronics in Agriculture,2023,213:108250.DOI:10.1016/j.compag.2023.108250.
[17] 刘月峰，刘世峰，张振荣.基于StaMaLSTM和多源数据的玉米产量预测[J].中国农机化学报，2025,46(9):81-90.
[18] Sajid S S,Shahhosseini M,Huber I,et al.County-scale crop yield prediction by integrating crop simulation with machine learning models[J].Frontiers in Plant Science,2022,13:1000224.DOI:10.3389/fpls.2022.1000224.
[19] Zhong Renhai,Zhu Yue,Wang Xuhui,et al.Detect and attribute the extreme maize yield losses based on spatio-temporal deep learning[J].Fundamental Research,2023,3(6):951-959.
[20] Ma Yuchi,Zhang Zhou,Kang Yanghui,et al.Corn yield prediction and uncertainty analysis based on remotely sensed variables using a Bayesian neural network approach[J].Remote Sensing of Environment,2021,259:112408.DOI:10.1016/j.rse.2021.112408.
[21] Tiruneh G A,Meshesha D T,Adgo E,et al.Mapping crop yield spatial variability using Sentinel-2 vegetation indices in Ethiopia[J].Arabian Journal of Geosciences,2023,16(11):631.DOI:10.1007/s12517-023-11754-x.
[22] Li Yuanchao,Zeng Hongwei,Zhang Miao,et al.A county-level soybean yield prediction framework coupled with XGBoost and multidimensional feature engineering[J].International Journal of Applied Earth Observation and Geoinformation,2023,118:103269.DOI:10.1016/j.jag.2023.103269.
[23] 朱志畅，葛焱，臧晶荣，等.基于无人机图像和SHAP特征筛选的小麦田间产量预测方法研究[J].麦类作物学报，2025,45(2):264-274.
[24] Zhu Guocan,Zhao Chenxuan,Zhou Lili,et al.Winter wheat yield prediction at a county scale using time series variation features of remote sensing spectra and machine learning[J].European Journal of Agronomy,2025,170:127751.DOI:10.1016/j.eja.2025.127751.
[25] Dong Jie,Pang Zhouye,Fu Yangyang,et al.Annual winter wheat mapping dataset in China from 2001 to 2020[J].Scientific Data,2024,11(1):1218.DOI:10.1038/s41597-024-04065-7.
[26] Liu Man,He Wei,Zhang Hongyan.CN_wheat10:a 10 m resolution dataset of spring and winter wheat distribution in China (2018-2024) derived from time-series remote sensing[J].Earth System Science Data,2025,18(1):465-491.
[27] Konduri V S,Vandal T J,Ganguly S,et al.Data science for weather impacts on crop yield[J].Frontiers in Sustainable Food Systems,2020,4:52.DOI:10.3389/fsufs.2020.00052.
[28] Hu Jinlong,Miao Chiyuan,Su Jiajia,et al.An upgraded high-precision gridded precipitation dataset for the Chinese mainland considering spatial autocorrelation and covariates[J].Earth System Science Data,2025,17(8):3987-4004.
[29] Hengl T,Mendes De Jesus J,Heuvelink G B M,et al.SoilGrids250m:global gridded soil information based on machine learning[J].PLoS One,2017,12(2):e0169748.DOI:10.1371/journal.pone.0169748.
[30] 高崇，陈国坤，段兴武，等.基于对象级间隙填充与SG滤波的Sentinel-2 NDVI时间序列重建[J].农业工程学报，2025,41(10):212-220.
[31] Defalque G,Santos R,Bungenstab D,et al.Machine learning models for dry matter and biomass estimates on cattle grazing systems[J].Computers and Electronics in Agriculture,2024,216:108520.DOI:10.1016/j.compag.2023.108520.
[32] Chandel N S,Jat D,Chakraborty S K,et al.Deep learning assisted real-time nitrogen stress detection for variable rate fertilizer applicator in wheat crop[J].Computers and Electronics in Agriculture,2025,237(Part A):110545.DOI:10.1016/j.compag.2025.110545.
[33] Zhao Lei,Um D,Nowka K,et al.Cotton yield prediction utilizing unmanned aerial vehicles (UAV) and Bayesian neural networks[J].Computers and Electronics in Agriculture,2024,226:109415.DOI:10.1016/j.compag.2024.109415.
[34] Zhou Chengquan,Yang Guijun,Liang Dong,et al.Recognizing black point in wheat kernels and determining its extent using multidimensional feature extraction and a naive Bayes classifier[J].Computers and Electronics in Agriculture,2021,180:105919.DOI:10.1016/j.compag.2020.105919.
[35] 左璐，孙雷刚，鲁军景，等.石家庄市地表热环境时空演变及其影响因素分析[J].地理与地理信息科学，2025,41(3):119-128.
[36] Shafiee S,Mroz T,Burud I,et al.Evaluation of UAV multispectral cameras for yield and biomass prediction in wheat under different sun elevation angles and phenological stages[J].Computers and Electronics in Agriculture,2023,210:107874.DOI:10.1016/j.compag.2023.107874.
[37] 吴子豪，邹心颖，张杨，等.基于XGBoost-Geoshapley模型的典型黑土区土壤有机碳局部障碍因子识别及优化[J].农业工程学报，2025,41(20):258-268.
基本信息:
中图分类号:P237;S512.11
引用信息:
[1]刘涛,张苏,方志祥,等.基于LSTM-Attention-XGBoost的河南省冬小麦产量预测与空间格局分析[J].地理与地理信息科学,2026,42(04):27-35.
基金信息:
河南省自然科学基金面上项目(252300421849); 河南省高校科技创新人才支持计划项目(26HASTIT018)
2026-07-13
2026-07-13
2026-07-13