近20年华北粮食主产区冬小麦种植面积遥感快速提取研究

Study on rapid extraction of winter wheat planting area in the main grain production regions of North China using remote sensing over the recent 20 years

关键词:  遥感  MODIS  冬小麦  华北  粮食主产区

Key words:  remote sensing  MODIS  winter wheat  North China  the main grain production regions

摘要:

【目的】 基于遥感数据，研究快速提取华北粮食主产区近20年（2001—2020年）冬小麦种植面积的方法，生成准确的长时间序列冬小麦面积遥感产品，为政府决策机构和科研单位的工作提供数据支持。【方法】 文章基于经过滤波重构的MODIS植被指数产品，分析了研究区不同纬度下冬小麦在整个生长季中的时序特征，考虑到不同区域冬小麦物候差异，提出了一种关键生长季时序NDVI曲线匹配的方法，在无样本的条件下，快速提取冬小麦面积。通过使用统计年鉴进行面积验证，并结合目视解译的样本和高分辨率数据哨兵2号提取的结果，计算混淆矩阵并进行精度评价。【结果】 与2001—2018年的统计年鉴数据对比，平均相对误差为16.1%；与目视解译和哨兵2号分类结果中的6 459个采样点的精度评价相比，总体精度达到87.4%，kappa系数为0.61。【结论】 根据冬小麦的物候特征，通过提取NDVI的时序特征并采用时序NDVI曲线匹配算法，可以快速准确地提取华北粮食主产区冬小麦的种植面积和分布情况。

Abstract:

【Purpose】 Based on remote sensing data，this study aims to explore a method for rapidly extracting the winter wheat planting area in the main grain production regions of North China over the past 20 years（2001-2020），and generate accurate long-term time series winter wheat acreage remote sensing products to provide data support for government decision-making institutions and research institutions.【Method】 Based on the filtered and reconstructed MODIS vegetation index products，this article analyzed the temporal characteristics of winter wheat in different latitudes of the study area throughout the entire growth season. Taking into account the phenological differences of winter wheat in different regions，a method based on the matching of key growth season temporal NDVI curves was proposed to rapidly extract the winter wheat area under the condition of no samples. The extracted area was validated using statistical yearbooks，and accuracy assessment was conducted by combining visual interpretation samples and high-resolution data from Sentinel-2，calculating the confusion matrix.【Result】 Compared with the statistical yearbook data from 2001 to 2018，the average relative error was 16.1%. Compared with the accuracy evaluation of 6，459 sampling points from visual interpretation and Sentinel-2 classification results，the overall accuracy reached 87.1%，with a kappa coefficient of 0.65.【Conclusion】 Based on the phenological characteristics of winter wheat，the rapid and accurate extraction of winter wheat planting area and distribution in the main grain producing regions of North China can be achieved by extracting the temporal features of NDVI and employing a temporal NDVI curve matching algorithm.