Global Soil Inorganic Carbon Database

Data repository for the paper "Size, distribution and vulnerability of the global soil inorganic carbon (Huang et al. 2024, <i>Science, </i>DOI: 10.1126/science.adi7918)". The global soil inorganic carbon (SIC) database provides high-resolution (30 arc-seconds) global gridded SIC data for three soil depths (0-30cm, 30-100cm, 100-200cm). SIC is expressed as carbonate equivalent in fine earth (i.e., ≤ 2 mm), in unit of gC/kg(soil).This dataset includes six gridded data files in NetCDF format. Three files provide the estimations of present-day SIC for three soil depths. The other three files contain values of the standard deviations of each estimation based on machine learning models built during 10-fold cross validation. These data are generated through linking measured SIC content (N &gt; 200,000) to spatially explicit environment covariates by using the machine learning models (two-part models: binary classification and regression model) (Materials and Methods).There are two csv files documenting the compiled measured SIC at different locations across the globe. Site-level measurements were collected from the ISRIC World Soil Information Service (WoSIS) with standardised soil profile data, supplemented by additional databases from National Soil Surveys, resamplings and new measurements from China, from literature via “China National Knowledge Infrastructure” (https://www.cnki.net/) and “Web of Science” (http://www.isiknowledge.com/), from additional samples collected in South-West Germany, Egypt and Russia. University of Hohenheim holds the ownership of the measured data with the label "Germany_xxx". A small portion of entries of measured samples located in northeast China are left empty due to data sharing policies. For users need this part of data, please address to Dr. Songbai Hong ([e-mail]). The gridded NetCDF data is likely to be less accurate in observation limited regions compared to regions with enough measurements. Caution is advised when applying this data in such areas. Variable names follow the conventions used in WoSIS snapshot - September 2019. The study of "Significant loss of soil inorganic carbon at the continental scale" (National Science Review, Volume 9, Issue 2, February 2022, nwab120) and Supplementary Table S12 provide literature lists from which additional measurements were extracted.This archive also hosts programming codes for reading compiled data, training predictive models, generating NetCDF files or creating visual displays.

本数据集为论文《全球土壤无机碳的规模、分布与脆弱性》（Huang等，2024，《Science》，DOI: 10.1126/science.adi7918）配套的数据仓库。
全球土壤无机碳（soil inorganic carbon, SIC）数据库提供了三个土壤深度层级（0~30cm、30~100cm、100~200cm）的高分辨率（30角秒）全球格网化SIC数据。SIC以细粒组分（即粒径≤2mm）的碳酸盐当量形式表征，单位为gC/kg（土壤）。
本数据集包含6个NetCDF格式的格网数据文件。其中3个文件为当前状态下三个土壤深度的SIC估算结果，另外3个文件则对应基于10折交叉验证中构建的机器学习模型所得到的各估算结果的标准差。
上述数据通过机器学习模型（两阶段模型：二分类模型与回归模型），将实测SIC含量（样本量N>200000）与空间显性环境协变量关联生成（详见材料与方法）。
数据集附带2个CSV格式文件，记录了全球不同点位的整合实测SIC数据。点位实测数据源自ISRIC世界土壤信息服务（ISRIC World Soil Information Service, WoSIS）的标准化土壤剖面数据，补充数据来源包括：各国土壤调查数据集、中国区域的重采样与新增实测数据，以及通过《中国知网（China National Knowledge Infrastructure）》（https://www.cnki.net/）和《Web of Science》（http://www.isiknowledge.com/）检索文献获取的相关内容，另包含德国西南部、埃及与俄罗斯地区采集的额外样品。
霍恩海姆大学拥有标签为"Germany_xxx"的实测数据所有权。由于数据共享政策限制，中国东北地区的部分实测样品条目为空。若用户需要获取该部分数据，请联系洪松白博士（[e-mail]）。
相较于测量数据充足的区域，本数据集的格网NetCDF数据在观测数据有限的区域精度相对较低，在这类区域应用该数据时需谨慎。
变量命名遵循WoSIS 2019年9月快照版本的规范。此前发表的研究《大陆尺度下土壤无机碳的显著流失》（《国家科学评论》，2022年2月，第9卷第2期，nwab120）及其补充表S12，提供了本次研究额外提取实测数据的文献列表。
本数据集存档还包含用于读取整合数据、训练预测模型、生成NetCDF文件或创建可视化展示的编程代码。

搜集汇总

背景与挑战

背景概述

Global Soil Inorganic Carbon Database是一个高分辨率全球土壤无机碳数据集，覆盖三个不同土壤深度，包含六个NetCDF文件和两个CSV文件，数据通过机器学习模型生成。数据集适用于碳封存科学和无机地球化学领域的研究，但需注意在观测有限区域的准确性可能较低。