欢迎访问《生态学杂志》官方网站,今天是 分享到:

生态学杂志

• 方法与技术 • 上一篇    下一篇

基于机器学习估算青藏高原多年冻土区草地净初级生产力

李传华1,2*,孙皓1,王玉涛1,曹红娟1,殷欢欢1,周敏1,朱同斌1   

  1. 1西北师范大学地理与环境科学学院, 兰州 730070; 2中国科学院寒区旱区环境与工程研究所冰冻圈科学国家重点实验室/青藏高原冰冻圈观测试验研究站, 兰州 730000)
  • 发布日期:2020-05-10

Estimation of grassland net primary productivity in permafrost of Qinghai-Tibet Plateau based on machine learning.

LI Chuan-hua1,2*, SUN Hao1, WANG Yu-tao1, CAO Hong-juan1, YIN Huan-huan1, ZHOU Min1, ZHU Tong-bin1   

  1. (1College of Geography and Environmental Science, Northwest Normal University, Lanzhou 730070, China; 2Cryosphere Research Station on the QinghaiTibetan Plateau, State Key Laboratory of Cryospheric Science, Northwest Institute of EcoEnvironment and Resources, Chinese Academy of Sciences, Lanzhou 730000, China).
  • Published:2020-05-10

摘要: 净初级生产力(NPP)的估算还存在很大的不确定性。本文利用机器学习算法(RF和RBF-ANN)估算了2002—2018年青藏高原多年冻土区草地NPP,分析了青藏高原多年冻土区草地NPP的时空格局、变化特征及其对气候因子的响应。结果表明:(1)机器学习估算结果可靠,简单易行。(2)青藏高原多年冻土区草地NPP表现为东南向西北逐渐递减的趋势;NPP总量为175.39 Tg C·a-1,单位面积均值为164.10 g C·m-2·a-1,呈波动上升的趋势。(3)青藏高原多年冻土区草地NPP增加的面积占20.49%;各草地类型的NPP增长幅度不一致,表现为高寒沼泽草甸>高寒草甸>高寒草原>高寒荒漠草原。(4)温度是青藏高原多年冻土区草地NPP变化的主导因子,降水的影响沿东南向西北逐渐减弱。

关键词: 冬小麦, 减产风险, WOFOST作物模型

Abstract: There is still a great uncertainty in the estimation of net primary productivity (NPP). In this study, machine learning algorithm (RF and RBF-ANN) was used to estimate the NPP of grassland in permafrost of Qinghai-Tibet Plateau from 2002 to 2018. We analyzed the temporal and spatial pattern, variation characteristics, and response of grassland NPP to climate factors in the permafrost of Qinghai-Tibet Plateau. The results showed that: (1) The estimation results of machine learning are reliable and simple. (2) In the permafrost of Qinghai-Tibet Plateau, NPP showed a decreasing trend from southeast to northwest. The total NPP was 175.39 Tg C·a-1, and the average NPP per unit area was 164.10 g C·m-2·a-1, showing a fluctuating trend. (3) The area with increased NPP accounted for 20.49% of the total area. The amplitudes of NPPincrease differed with grassland types, with an order of alpine wet meadow > alpine meadow > alpine steppe > alpine desert steppe. (4) Temperature was the dominant factor driving grassland NPP change in permafrost area of the Qinghai-Tibet Plateau. The influence of precipitation gradually weakened along the southeast to northwest.

Key words: yield reduction risk, WOFOST crop model, winter wheat.