基于Sentinel-2遥感影像的作物信息提取与需水量分析研究  

Crop Information Extraction and Water Demand Analysis Based on Sentinel-2 Remote Sensing Image

作　　者：钱鑫 李培显 谢宏全 郜薇薇 王杨 刘付程 QIAN Xin;LI Pei-xian;XIE Hong-quan;GAO Wei-wei;WANG Yang;LIU Fu-chen(School of Marine Technology and Geomatics,Jiangsu Ocean University,Lianyungang 222005,Jiangsu Province,China;Shandong Jintian Mapping Geographic Information Co.,Ltd,Jinan 250000,China)

机构地区：[1]江苏海洋大学海洋技术与测绘学院,江苏连云港222005 [2]山东金田测绘地理信息有限公司,济南250000

出　　处：《节水灌溉》2022年第5期33-38,46,共7页Water Saving Irrigation

基　　金：国家自然科学基金项目(41976187);江苏海洋大学“测绘科学与技术”重点学科。

摘　　要：气候变化扰动的水资源稀缺性使得农业可持续水资源利用面临危机,节水农业是现代社会农业转型的主要方式,更是可持续发展的根本出路。黑河流域是我国第二大内陆河,中游绿洲是我国的重要粮食生产基地,了解黑河流域的作物种植信息与作物需水特征,对于指导干旱半干旱区农业高效用水及国家粮食安全具有重要意义。基于Google Earth Engine(GEE)云平台利用Sentinel-2影像,结合黑河流域作物的物候特征选取了6-9月的影像数据,根据作物的物候特征重要性完成了特征优选。研究运用随机森林、支持向量机、决策树及投票法分类器完成了作物的识别分类与结果对比。最后,研究通过CROPWAT模型估算了黑河流域作物的需水量与灌溉用水量。研究结果表明:①GEE能够快速完成影像数据的去云、特征构建等预处理;②基于决策树分类器的土地分类结果精度达到82.5%,平均Kappa系数为0.73;估算了作物各个时期所需水量及灌溉用水量。构建的分类体系及作物的需水量估算为精准化管理、灌溉控制系统提供了一种新思路。The scarcity of water resources disturbed by climate change makes the sustainable utilization of water resources in agriculture face a crisis. Water-saving agriculture is the main way of agricultural transformation in modern society and the fundamental way out of sustainable development. Heihe River Basin is the second largest inland river in China, and the oasis in the middle reaches is an important food production base in China. Understanding the crop planting information and crop water demand characteristics in Heihe River Basin is of great significance for guiding agricultural efficient water use in arid and semi-arid areas and national food security. Based on Google Earth engine(GEE) cloud platform, by using sentinel-2 image and combined with the phenological characteristics of crops in Heihe River Basin to select the image data from June to September, the feature optimization was completed according to the importance of crop phenological characteristics. Random forest, support vector machine, decision tree and voting classifier were used to classify crops and compare the results.Finally, the CROPWAT model was used to estimate the crop water requirement and irrigation water consumption in the Heihe River Basin.The results show that:①GEE can quickly complete the preprocessing of image data, such as cloud removal and feature construction;②The accuracy of land classification results based on decision tree classifier is 82.5%, and the average Kappa coefficient is 0.73. The water demand and irrigation water consumption of crops in each period were estimated. The constructed classification system and crop water demand estimation provide a new idea for accurate management and irrigation control system.

关 键 词：遥感影像 GEE云平台 Sentinel-2 CROPWAT 作物需水量估算 灌溉用水量估算 作物识别分类 

分 类 号：S274[农业科学—农业水土工程] P237[农业科学—农业工程]