基于遥感作物参考曲线的黄土高原苹果产区始花期监测  

Monitoring of Initial Flowering Period based on Remote Sensing Crop Reference Curve of Apple Producing Areas in Loess Plateau

作　　者：王梓霖 王钊 孙亮[3] 孙政 段梦琦 王永前[1] WANG Zilin;WANG Zhao;SUN Liang;SUN Zheng;DUAN Mengqi;WANG Yongqian(College of Resources and Environment,Chengdu University of Information Technology,Chengdu 610225,China;Agricultural Remote Sensing and Cash Crop Meteorological Service Center of Shaanxi Province,Xi'an 710016,China;Institute of Agricultural Resources and Regional Planning,Chinese Academy of Agricultural Sciences,Beijing 100081,China)

机构地区：[1]成都信息工程大学资源与环境学院,四川成都610225 [2]陕西省农业遥感与经济作物气象服务中心,陕西西安710016 [3]中国农业科学院农业资源与农业区划研究所,北京100081

出　　处：《遥感技术与应用》2024年第2期328-336,共9页Remote Sensing Technology and Application

基　　金：陕西省气象局地方项目“风云三号03批气象卫星工程地面应用系统区域特色遥感应用系统农田生态遥感应用系统”(FY-3(03)-AS-12.14)。

摘　　要：苹果种植业为陕西省提升农业、富裕农民的支柱型产业。黄土高原苹果产区每年都有花期防冻害和产量评估的需求,花期监测尤为重要,而常规的花期监测方法存在监测成本大和精度低的问题。采用一种基于遥感作物参考曲线重建的方法,介绍比较了该方法与其他方法应用于常见作物的精度,并使用地理国情普查地表覆盖分类数据识别的MODIS纯像元中提取一组NDVI时间序列,将其用于拟合类似Sentinel-2的观测值,将这种方法应用于黄土高原地区苹果作物上,根据苹果样点的归一化植被指数(NDVI)序列(作物参考曲线)与重建后的苹果NDVI序列在空间上的物候差异,绘制历史10 m高分辨率的始花期监测结果,实现了花期监测由点到面的转化。研究结果表明:2019~2021三年验证区域苹果始花期的探测结果平均值与地面数据绝对误差均在1~3 d之内,平均绝对误差为0.926 d,均方根误差为1.503 d。该方法可用于苹果花期高精度监测,预测结果可为实际生产中苹果花期防冻害和产量评估工作的开展提供参考,同时这种方法也可以用于研究其他作物的物候特征,如小麦、玉米等,有望为农业生产提供更准确的决策支持。The apple cultivation industry is a pivotal sector for elevating agriculture and enriching farmers in Shaanxi Province.The apple-producing regions on the Loess Plateau face annual demands for blossom period frost protection and yield assessment.Blossom period monitoring is particularly crucial,and conventional meth‐ods suffer from high monitoring costs and low accuracy.This study introduces a method based on reconstructing crop reference curves using remote sensing.It compares the accuracy of this method with others applied to com‐mon crops.Subsequently,a set of NDVI time series is extracted from MODIS pure pixels identified by geo‐graphic national survey land cover classification data.These time series are then used to fit observations similar to Sentinel-2.The method is applied to apple crops in the Loess Plateau region.Based on the spatial phenologi‐cal differences between the Normalized Difference Vegetation Index(NDVI)sequences of apple sample points(crop reference curve)and the reconstructed apple NDVI sequences,historical high-resolution onset blossom period monitoring results at 10 meters are generated,achieving the transformation of blossom period monitoring from point to area.The research results indicate that the detection results for the onset blossom period of apples in the validation region for the years 2019~2021 have an average absolute error within 1~3 days compared to ground data.The average absolute error is 0.926 days,and the root mean square error is 1.503 days.This meth‐od can be used for high-precision monitoring of the apple blossom period,and the predicted results can serve as a reference for frost protection and yield assessment in actual apple production.Additionally,this method can be applied to study the phenological characteristics of other crops such as wheat and maize,promising more accu‐rate decision support for agricultural production.

关 键 词：作物参考曲线 物候 苹果始花期探测 遥感 黄土高原 

分 类 号：TP79[自动化与计算机技术—检测技术与自动化装置]