中国高分辨率国家土壤信息网格基本属性数据集（2010-2018）

土壤是人类生存和发展的基础，多个联合国可持续发展目标（SDGs）与土壤资源利用和管理直接相关。然而，全球和我国现有土壤信息大多源于历史土壤调查，较为粗略、陈旧，不能满足应对粮食安全、水资源紧缺、土地退化和气候变化等全球和区域性问题的需要。中国疆域辽阔，土壤景观复杂多样，人为活动强烈，建立高精度土壤信息网格在科学上和应用上均有重要意义。基于近年“我国土系调查与《中国土系志》编制项目”获得的5000多个代表性土壤剖面样点，采用预测性土壤制图范式，利用地理信息与遥感技术对成土环境条件进行精细刻画和空间分析，研发自适应深度函数拟合方法，集成先进的集合式机器学习方法，在高性能并行计算环境下生成了我国系列土壤属性（土壤有机碳、PH值、全氮、全磷、全钾、阳离子交换量、砾石含量（>2mm），砂粒、粉粒、粘粒、土壤质地类型、容重、土体厚度等）高分辨率三维栅格分布图，并估算了不确定性的空间分布。与现有土壤图和相关土壤数据集相比，本研究结果大幅提高了现有制图的准确性和精细度，并提供了空间预测的不确定性信息，更好地表征了我国土壤属性的空间变异特征。该工作初步构建了我国第1版高分辨率国家土壤信息网格，也是对全球数字土壤制图计划（GlobalSoilMap.net）的重要贡献，预期在土壤资源、农业、水文、生态、气候、环境等领域有广泛的应用前景，如土壤监测与管理、土壤功能评价、陆面过程模拟和法庭土壤物证溯源等。

Soil serves as the foundation for human survival and development. Multiple United Nations Sustainable Development Goals (SDGs) are directly associated with soil resource utilization and management. However, most existing soil information globally and in China originates from historical soil surveys, which are coarse and outdated, failing to meet the demands of addressing global and regional issues such as food security, water scarcity, land degradation, and climate change. China has a vast territory with complex and diverse soil landscapes and intensive human activities; thus, establishing a high-resolution soil information grid holds great scientific and practical significance. Based on over 5,000 representative soil profile samples collected from the recent "China Soil Series Survey and Compilation of the Soil Series of China" project, this study adopted the predictive soil mapping paradigm, utilized geographic information and remote sensing technologies to precisely characterize and spatially analyze soil-forming environmental conditions, developed an adaptive deep function fitting method, integrated advanced ensemble machine learning approaches, and generated high-resolution 3D raster maps of a series of soil properties across China, including soil organic carbon, pH value, total nitrogen, total phosphorus, total potassium, cation exchange capacity (CEC), gravel content (>2mm), sand content, silt content, clay content, soil texture type, bulk density, and soil thickness, under a high-performance parallel computing environment. Additionally, the spatial distribution of prediction uncertainty was estimated. Compared with existing soil maps and related soil datasets, the findings of this study substantially improve the accuracy and resolution of current soil mapping, provide uncertainty information for spatial predictions, and better characterize the spatial variability of soil properties in China. This work preliminarily constructs the first version of China's high-resolution national soil information grid, and also makes a significant contribution to the Global Digital Soil Mapping Initiative (GlobalSoilMap.net). It is expected to have broad application prospects in fields including soil resources, agriculture, hydrology, ecology, climate, and the environment, such as soil monitoring and management, soil function evaluation, land surface process simulation, and forensic soil evidence traceability.

搜集汇总

背景与挑战

背景概述

该数据集是中国高分辨率国家土壤信息网格基本属性数据集，覆盖2010年至2018年，基于5000多个土壤剖面样点，采用预测性土壤制图和机器学习方法，生成了土壤有机碳、pH、全氮等多种属性的三维栅格分布图，空间分辨率为250米和1公里。其特点是大幅提高了土壤属性制图的准确性和精细度，并提供了不确定性信息，适用于土壤资源管理、农业、水文、生态和气候等领域的研究与应用。