欢迎访问《草业学报》官方网站,今天是 分享到:

草业学报 ›› 2022, Vol. 31 ›› Issue (10): 1-17.DOI: 10.11686/cyxb2021386

• 研究论文 •    

基于MODIS数据与机器学习的青藏高原草地地上生物量研究

金哲人(), 冯琦胜(), 王瑞泾, 梁天刚   

  1. 兰州大学草地农业生态系统国家重点实验室,兰州大学农业农村部草牧业创新重点实验室,兰州大学草地农业教育部工程研究中心,兰州大学草地农业科技学院,甘肃 兰州 730020
  • 收稿日期:2021-10-28 修回日期:2022-01-10 出版日期:2022-10-20 发布日期:2022-09-14
  • 通讯作者: 冯琦胜
  • 作者简介:E-mail: fengqsh@lzu.edu.cn
    金哲人(1997-),男,回族,河南郑州人,在读硕士。E-mail: Jinzhr20@lzu.edu.cn
  • 基金资助:
    国家重点研发计划(2019YFC0507701);国家自然科学基金(31672484);中国工程院咨询研究项目(2021-HZ-5);财政部和农业农村部:国家现代农业产业技术体系和兰州大学中央高校基本科研业务费专项资金(lzujbky-2021-kb13)

A study of grassland aboveground biomass on the Tibetan Plateau using MODIS data and machine learning

Zhe-ren JIN(), Qi-sheng FENG(), Rui-jing WANG, Tian-gang LIANG   

  1. State Key Laboratory of Grassland Agro-ecosystems,Key Laboratory of Grassland Livestock Industry Innovation,Ministry of Agriculture and Rural Affairs,Engineering Research Center of Grassland Industry,Ministry of Education,College of Pastoral Agriculture Science and Technology,Lanzhou University,Lanzhou 730020,China
  • Received:2021-10-28 Revised:2022-01-10 Online:2022-10-20 Published:2022-09-14
  • Contact: Qi-sheng FENG

摘要:

青藏高原位于我国西部,又被称为“世界第三极”,对我国和世界的生态以及气候变化影响显著。为了评估2000-2020年青藏高原草地地上生物量(aboveground biomass,AGB)的变化情况,本研究采用多种机器学习方法结合MCD43A4产品数据模拟了草地地上生物量,并对该区域草地地上生物量的时空特征进行分析。结果表明:1)构建的机器学习模型中,Rborist模型精度最高,基于筛选后变量的R2 达到0.6484。“prec_05”、“prec_06”、“tp_12”、“NDPI”、“prec_04”、“tmax_01”、“prec_08”、“prec_12”这8个变量与生物量相关;2)青藏高原东南部的生物量要高于西北部,呈现由东南向西北递减趋势;3)2000-2020年间青藏高原草地生物量稳步增长,整体向好发展。青藏高原61.38%的草地变化趋势不具有可持续性,4.67%的草地持续性轻微恶化,持续性明显恶化的区域占比1.19%,呈稳定或恢复趋势的区域占比32.76%。

关键词: 植被指数, 机器学习, 草地地上生物量, 时空分布, 青藏高原

Abstract:

The Tibetan Plateau, often referred to as “the third pole of the world”, is located in the Southwest of China, and makes significant contribution to ecology and climate change in China and around the word. Our study evaluated the change in aboveground biomass (AGB) on the Tibetan Plateau from 2000 to 2020. We used multiple machine learning methods combined with MCD43A4 product data to simulate the aboveground biomass and analyzed the temporal and spatial characteristics of AGB in this region. The main results were as follows: 1) Among the constructed machine learning models, Rborist model demonstrated the highest accuracy, with an R2 of 0.6484 based on screened variables, and eight variables were found to be highly correlated with biomass: precipitation in May, precipitation in June, average temperature in December, normalized difference phenology index (NDPI), precipitation in April, maximum temperature in January, precipitation in August and precipitation in December; 2) AGB in the southeast of the Tibetan Plateau was higher than that in the northwest, with a decreasing trend from the southeast to northwest; 3) AGB on the Tibetan Plateau increased steadily from 2000 to 2020, with an overall positive movement. However, 61.38% of Tibetan Plateau grasslands showed a trend of unsustainability, 4.67% showed a slight deterioration trend, 1.19% showed a significant deterioration trend, and 32.76% were stable or showed a recovery trend.

Key words: vegetation index, machine learning, aboveground biomass, spatial and temporal distribution, Tibetan Plateau