收藏设为首页 广告服务联系我们在线留言

干旱区地理 ›› 2022, Vol. 45 ›› Issue (3): 763-773.doi: 10.12118/j.issn.1000-6060.2021.347

• 水文与水资源 • 上一篇    下一篇

基于遥感影像的黄河内蒙古段冰-水分类研究

翟涌光1(),张鑫1,冀鸿兰1(),牟献友1,张宝森2   

  1. 1.内蒙古农业大学水利与土木建筑工程学院,内蒙古 呼和浩特 010018
    2.黄河水利委员会黄河水利科学研究院,河南 郑州 450003
  • 收稿日期:2021-08-03 修回日期:2021-10-10 出版日期:2022-05-25 发布日期:2022-05-31
  • 通讯作者: 冀鸿兰
  • 作者简介:翟涌光(1986-),男,副教授,主要从事环境遥感监测及遥感大数据处理研究. E-mail: ychia@imau.edu.cn
  • 基金资助:
    国家重点研发计划项目(2018YFC1508401);国家自然科学基金项目(51969020);国家自然科学基金项目(51569020);2018自治区应用技术研究与开发资金项目(201802104)

Ice-water classification in Inner Mongolia reach of the Yellow River based on remote sensing images

ZHAI Yongguang1(),ZHANG Xin1,JI Honglan1(),MOU Xianyou1,ZHANG Baosen2   

  1. 1. College of Water Conservancy and Civil Engineering, Inner Mongolia Agricultural University, Hohhot 010018, Inner mongolia, China
    2. Yellow River Institute of Hydraulic Research, Zhengzhou 450003, Henan, China
  • Received:2021-08-03 Revised:2021-10-10 Online:2022-05-25 Published:2022-05-31
  • Contact: Honglan JI

摘要:

及时获取凌汛期河冰和水体的空间分布特征,对于预测冰凌灾害、提高防凌信息化管理水平有重要意义。遥感技术是当前获取河冰和水体空间分布的最主要手段之一。但是,黄河水体有大量悬浮泥沙,这给基于遥感技术的高精度冰-水分类带来了挑战。以黄河内蒙古段为例,基于Landsat 8 OLI遥感影像数据,在利用归一化积雪指数(NDSI)及河道矢量数据排除无关地物的基础上,对比了近红外波段反射率值、归一化差异水体指数(NDWI)、归一化积雪指数(NDSI)、改进的归一化积雪指数(MNDSI)以及归一化差异未封冻水体指数(NDUWI)在黄河内蒙古段典型河道河冰、水体分类中的表现,计算各指标总体分类精度及Kappa系数并进行阈值稳定性分析。结果表明:在利用NDSI和高清历史影像排除河道外无关地物后,NDUWI在各子段影像中的总体分类精度和Kappa系数均达到90.00%及0.90以上,其河冰、水体最优区分阈值大体分布于阈值中值附近。研究结果可为凌汛期黄河冰凌监测方法的选取以及冰上爆破位置的拟定提供依据。

关键词: 河冰, 遥感指数, 阈值稳定性, 不确定性, Landsat 8, 黄河内蒙古段

Abstract:

Timely delivery of detailed information on the spatial distribution of river ice during ice-flood season is highly valuable for predicting, and improving communication on, ice disaster. Remote sensing technology provides a key method for obtaining the spatial distribution of river ice. However, the large amount of suspended sediment in the Yellow River represents a challenge to high-precision discrimination between ice and water based on remote sensing technology. Taking the Inner Mongolia reach of the Yellow River as an example, this study compares and evaluates the performance of five indices in the classification of river ice and water: near-infrared reflectance; normalized difference water index (NDWI); normalized difference snow index (NDSI); improved normalized snow index (MNDSI); and normalized difference unfrozen water index (NDUWI). The overall classification accuracy and Kappa coefficient (a measure of reliability) were calculated for each index, and the threshold stability of each index was analyzed. The results show that NDUWI achieves the highest accuracy and reliability (Kappa coefficient) in each studied subregion. The overall classification accuracy and Kappa coefficient of NDUWI are more than 90.0% and 0.90, respectively, and the optimal discrimination threshold between river ice and ice-free water is close to the median value. These results can provide a basis for the selection of ice monitoring methods and optimization of ice-blasting locations on the Yellow River during ice-flood season.

Key words: river ice, remote sensing index, threshold stability, uncertainty, Landsat 8, Inner Mongolia reach of the Yellow River