地理科学进展 ›› 2022, Vol. 41 ›› Issue (12): 2327-2341.doi: 10.18306/dlkxjz.2022.12.011

• 研究论文 • 上一篇    下一篇

青海省土壤有机碳估算及其不确定性分析

周沛芳1,2,3(), 周涛1,2,3,*(), 刘霞1,2,3, 张亚杰1,2,3, 徐艺心1,2,3, 罗惠1,2,3, 于佩鑫1,2,3, 张靖宙1,2,3   

  1. 1.北京师范大学地表过程与资源生态国家重点实验室,北京 100875
    2.北京师范大学环境演变与自然灾害教育部重点实验室,北京 100875
    3.北京师范大学地理科学学部灾害风险科学研究院,北京 100875
  • 收稿日期:2022-03-07 修回日期:2022-10-25 出版日期:2022-12-28 发布日期:2022-12-31
  • 通讯作者: *周涛(1972— ),男,湖南冷水江人,教授,博士生导师,主要研究方向为全球变化与植被响应、陆地生态系统碳循环。E-mail: tzhou@bnu.edu.cn
  • 作者简介:周沛芳(1997— ),湖南常德人,硕士生,主要研究方向为气候变化及生态环境响应。E-mail: pfzhou@mail.bnu.edu.cn
  • 基金资助:
    第二次青藏高原综合科学考察研究项目(2019QZKK0405);国家自然科学基金项目(42277206)

Estimation of soil organic carbon and its uncertainty in Qinghai Province

ZHOU Peifang1,2,3(), ZHOU Tao1,2,3,*(), LIU Xia1,2,3, ZHANG Yajie1,2,3, XU Yixin1,2,3, LUO Hui1,2,3, YU Peixin1,2,3, ZHANG Jingzhou1,2,3   

  1. 1. State Key Laboratory of Earth Surface Processes and Resource Ecology, Beijing Normal University, Beijing 100875, China
    2. Key Laboratory of Environmental Change and Natural Disaster of Ministry of Education, Faculty of Geographical Science, Beijing NormalUniversity, Beijing 100875, China
    3. Institute of Disaster Risk Science, Faculty of Geographical Science, Beijing Normal University, Beijing 100875, China
  • Received:2022-03-07 Revised:2022-10-25 Online:2022-12-28 Published:2022-12-31
  • Supported by:
    The Second Tibetan Plateau Scientific Expedition and Research Program(2019QZKK0405);National Natural Science Foundation of China(42277206)

摘要:

土壤有机碳对区域碳平衡起着关键性的作用,量化其空间格局及动态变化是准确评估生态系统碳汇潜力的基础。然而,不同土壤有机碳估算方法和不同样本得出的结果存在非常大的差异和不确定性,尤其是地形复杂、对气候变化敏感的青藏高原地区。为定量评估不同方法估算的土壤有机碳密度空间分布格局在青藏高原地区的差异,论文以青海省为研究区,收集整理了青海省806个土壤有机碳密度采样点数据,基于气候、植被、地形和土壤等多种解释变量,采用逐步回归、反距离权重插值、普通克里格插值和随机森林模型4种不同的方法,对青海省表层(0~30 cm)土壤有机碳密度空间分布及其影响因素进行了探究。结果表明,归一化植被指数、光合有效辐射、总氮、年均温、海拔、年降水量和净初级生产力是土壤有机碳密度估算的重要变量;尽管4种方法所估算的青海省土壤有机碳密度的均值较为接近,处于5.14~5.62 kg C·m-2之间,但其变化范围存在较大差异,分别为0.17~23.25、0.34~46.61、0.56~35.08和0.62~24.85 kg C·m-2;4种方法模拟结果的均方根误差分别为3.93、3.37、3.48和3.19 kg C·m-2,平均标准差分别为0.12、0.51、0.61和0.27 kg C·m-2,其中随机森林模型的结果较为稳定且精度较高,也更能准确反映青海省土壤有机碳的空间分布格局。比较发现,现有的土壤有机碳产品(SoilGrids250m 2.0和HWSD v1.2)在反映青海省土壤有机碳的分布方面还存在较大差异,相对而言,SoilGrids250m 2.0产品的土壤有机碳和随机森林模拟结果比较接近。

关键词: 土壤有机碳, 随机森林模型, 空间插值, 回归模型, 青海省

Abstract:

Soil organic carbon plays a critical role in regional carbon balance. Quantifying the spatial pattern and dynamic changes of soil organic carbon can lay a solid foundation for a realistic assessment of the ecosystem carbon sink potential. Different methods and samples for the estimation of soil organic carbon can lead to dramatic differences and uncertainties in the estimated results, particularly in the Qinghai-Tibet Plateau area with complex terrain and sensitive to climate changes. In order to quantitatively assess differences of the soil organic carbon density spatial distribution pattern obtained by different methods in the Qinghai-Tibet Plateau area, this study used Qinghai Province as its research area, and collected data from 806 soil organic carbon density sampling sites in the province. Explanatory variables, including climate, vegetation, terrain, and soil conditions, and four different methods, including stepwise regression, inverse distance weighted interpolation, ordinary kriging interpolation, and random forest model, were used to investigate the surface (0-30 cm) soil organic carbon density spatial distribution in Qinghai Province and its influencing factors. The results suggest that the normalized difference vegetation index, photosynthetically active radiation, total nitrogen, annual mean temperature, elevation, annual precipitation, and net primary productivity are the major factors that influence the soil organic carbon density estimation. The average of the soil organic carbon density in Qinghai Province estimated by the four methods ranges from 5.14 to 5.62 kg C·m-2. But their variation range is significantly different, being 0.17-23.25, 0.34-46.61, 0.56-35.08, and 0.62-24.85 kg C·m-2, respectively. Simulation precision assessment revealed that the root-mean-square error of the results obtained by the four methods is 3.93, 3.37, 3.48, and 3.19 kg C·m-2, and their mean standard deviation is 0.12, 0.51, 0.61, and 0.27 kg C·m-2, respectively. Among the results, those obtained by the random forest model are relatively stable and of high precision, and can more accurately reflect the spatial distribution pattern of the soil organic carbon in Qinghai Province. The current products—SoilGrids250m 2.0 and HWSD v1.2—show relatively large differences in reflecting the distribution of the soil organic carbon in Qinghai Province. Comparatively, among these two soil carbon products, the results obtained by SoilGrids250m 2.0 and the random forest simulation results are more similar to each other.

Key words: soil organic carbon, random forest model, spatial interpolation, regression model, Qinghai Province