资源科学 ›› 2021, Vol. 43 ›› Issue (12): 2393-2402.doi: 10.18402/resci.2021.12.03

• “澜沧江—湄公河流域农业资源与环境”专栏 • 上一篇    下一篇

基于时序遥感的柬埔寨水稻种植时空格局监测

黄翀1,2()   

  1. 1.中国科学院地理科学与资源研究所,资源与环境信息系统国家重点实验室,北京 100101
    2.中国科学院黄河三角洲现代农业工程实验室,北京 100101
  • 收稿日期:2021-04-23 修回日期:2021-07-24 出版日期:2021-12-25 发布日期:2022-02-16
  • 作者简介:黄翀,男,安徽六安人,博士,副研究员,主要从事生态遥感研究。E-mail: huangch@lreis.ac.cn
  • 基金资助:
    澜沧江—湄公河合作专项;国家自然科学基金项目(41901309);国家自然科学基金项目(41801353)

Monitoring rice cropping system in Cambodia and its influencing factors using time series MODIS images

HUANG Chong1,2()   

  1. 1. State Key Laboratory of Resources and Environmental Information System, Institute of Geographic Sciences and Natural Resources Research, CAS, Beijing 100101, China
    2. Yellow River Delta Modern Agricultural Engineering Laboratory, CAS, Beijing 100101, China
  • Received:2021-04-23 Revised:2021-07-24 Online:2021-12-25 Published:2022-02-16

摘要:

柬埔寨可耕地资源丰富,温度适宜,水稻生产极具潜力。及时监测水稻种植时空格局对于区域水稻生产管理、灾害风险评估和粮食政策制定具有重要意义。传统的水稻遥感监测研究大多只提供年际尺度的水稻空间分布,缺乏年内尺度水稻种植与收获信息。本文首先利用一年内所有可获取的MODIS影像,构建基于像元的MODIS NDVI年时间序列曲线;然后,选取最大值、最小值、均值和标准差逐像元计算时序统计参数特征,采用FastDTW算法计算像元时序曲线与水稻参考时序曲线的相似性特征,将时序统计特征与时序曲线相似度特征相结合,利用随机森林分类器,通过机器学习进行监督分类,提取水稻熟制信息;最后,结合时序曲线提取水稻物候特征,生成水稻收获时间信息,并对水稻耕作类型进行识别。研究表明:①柬埔寨水稻种植主要集中在洞里萨湖周围的低地平原和南部的湄公河下游。尽管柬埔寨全年热量条件适宜,但水资源获取限制对柬埔寨水稻种植时空格局具有显著影响。②水稻熟制以单季稻为主,约占全年水稻种植面积的80%,且分布区域稳定;双季稻面积约占20%,年际种植空间分布变化较大。雨季稻是柬埔寨水稻的主要种植类型,种植面积约占全年水稻面积的70%左右,年际变化不大;旱季稻和前雨季稻面积约占30%,年际空间分布差异显著。③对2011年和2016年水稻种植模式分析可知,灌溉条件和洪水对柬埔寨水稻种植时空具有重要影响。本文通过对柬埔寨年内水稻种植时空格局的高精度监测,识别其主要影响因素,为制定因地制宜和有弹性的水稻种植制度、保障柬埔寨粮食安全提供借鉴。

关键词: 水稻, 动态时间规整, 种植制度, 遥感, 监测, 随机森林, 柬埔寨

Abstract:

Cambodia has abundant arable land resources, suitable temperature, and great potential for rice plantation. Timely acquisition of rice cropping system information is important for regional rice production management, disaster risk assessment, and food policy formulation. Most traditional rice remote sensing monitoring studies only provide spatial patterns of rice distribution at the interannual scale, and information on rice planting and harvesting at the intraannual scale is often lacking. In this study, first all available MODIS time series data in a year were used to construct an image-based MODIS NDVI annual time series curve; the maximum value, minimum value, mean value, and standard deviation were selected to calculate the image-by-image time series statistical parameter features, and the FastDTW algorithm was used to calculate the similarity features between the image-by-image time series curve and the rice reference time series curve, and then the time series statistical features were combined with the time series curve similarity features, and the rice maturation information was extracted by supervised classification through machine learning using a random forest classifier. Finally, the rice phenological features extracted from the time series curves were combined to generate rice harvest time information for the identification of rice cultivation types. The study showed that rice cultivation in Cambodia is mainly concentrated in the lowland plains around the Tonle Sap Lake and the lower Mekong River. Although the thermal conditions in Cambodia are suitable for rice cultivation throughout the year, water access constraints have a significant impact on the spatial and temporal patterns of rice cultivation in the country. The rice maturity mode indicates that production was dominated by single-season rice, which accounted for about 80% of the annual rice cultivation area and had a stable distribution area; double-season rice accounted for about 20% of the area and showed a large interannual variation in the spatial distribution of cultivation. Wet season rice was the main type of rice cultivation in the country, and the planted area accounted for about 70% of the annual rice area with little interannual variation; dry season rice and ex-rainy season rice accounted for about 30% of the area, with significant interannual variation in spatial distribution. The analysis of rice cropping patterns in 2011 and 2016 showed that irrigation conditions and flooding had important spatial and temporal effects on rice cultivation in Cambodia. This study identified the main influencing factors through high-precision monitoring of the interannual and intraannual spatial and temporal patterns of rice cultivation in Cambodia, which provide a reference for the development of a locally adapted and resilient rice cultivation system to ensure food security in the country.

Key words: paddy rice, dynamic time warping (DTW), cropping system, remote sensing, monitoring, random forest, Cambodia