Soil and Landscape Grid Australia-Wide 3D Soil Property Maps (3" resolution) - Release 1

The Soil Facility produced a range of digital soil attribute products. Each product contains six digital soil attribute maps, and their upper and lower confidence limits, representing the soil attribute at six depths: 0-5cm, 5-15cm, 15-30cm, 30-60cm, 60-100cm and 100-200cm. These depths are consistent with the specifications of the GlobalSoilMap.net project (http://www.globalsoilmap.net/). The digital soil attribute maps are in raster format at a resolution of 3 arc sec (~90 x 90 m pixels). \n\nAttributes included: \nAvailable Water Capacity; \nBulk Density - Whole Earth; \nClay;\nEffective Cation Exchange Capacity;\npH - CaCl2; \nSilt; \nSand; \nTotal Nitrogen; \nTotal Phosphorus.\n\nPeriod (temporal coverage; approximately): 1950-2013; \nSpatial resolution: 3 arc seconds (approx 90m); \nTotal number of gridded maps for this attribute: 18; \nNumber of pixels with coverage per layer: 2007M (49200 * 40800); \nTotal size before compression: about 8GB; \nTotal size after compression: about 4GB; \nData license : Creative Commons Attribution 3.0 (CC By); \nTarget data standard: GlobalSoilMap specifications; \nFormat: GeoTIFF.\nLineage: The digital soil attribute maps and their uncertainties were generated by harmonising different sources of soil data collected from point locations and using a 3-dimensional spatial modelling technique. \n\nSoil inventory:\nThe national soil site data originates from two sources:\n\n(i) A set collated with the assistance of all the Australian State and Territory soil agencies (Searle, 2014). The individual State soil databases were combined into a single database adhering to the NatSoil Site Schema (Jacquier et al., 2012). This database contains morphological and laboratory data for all the soil profiles publicly available within existing agency databases in 2013.\n\n(ii) Spectroscopic estimates of the soil attributes with the Australian visible–near infrared database (Viscarra Rossel and Webster, 2012) on soil samples collected for the National Geochemical Survey of Australia (NGSA) (de Caritat, P & Cooper, M, 2011).\n\nHarmonisation to standard depths:\nData for each soil attribute, for all depths that were present in the inventory, was extracted and harmonised to the six standard depths using two different methods. When there were data from more than two depths, a mass preserving spline (Bishop et al., 1999) was fitted to derive the standard depths. When only two depths were present we used the imputation method described by Clifford et al. (2014).\n\nSpatial modelling:\nThe digital soil maps were generated by a 3-dimensional data mining-kriging approach with Monte Carlo resampling to produce estimates of uncertainty. The approach uses statistical relationships between the observed soil attributes at point locations and continuous values of more than 40 environmental covariates (including remote sensing, climatic data, a digital elevation model and terrain derivatives, gamma radiometrics and other geophysical data), and kriging of their residuals. The Cubist data mining software (Rulequest Research., 2008) implemented in the software R (R Core Team, 2013) was used for the data mining and the gstat package (Pebesma, 2004) was used for the geostatistical modelling. These hybrid models produce quantitative estimates of soil properties. Uncertainties in both parts of the model were quantified and expressed as the 90% confidence limits. Descriptions of the approach are given in Viscarra Rossel et al. (2015a); Viscarra Rossel and Chen (2011) and Viscarra Rossel, (2011).

土壤设施（Soil Facility）产出了一系列数字土壤属性产品。每套产品包含6幅数字土壤属性图及其上下置信限，分别对应6个深度层级的土壤属性：0-5cm、5-15cm、15-30cm、30-60cm、60-100cm及100-200cm。该深度分层标准与GlobalSoilMap.net项目（http://www.globalsoilmap.net/）的规范完全一致。所有数字土壤属性图均采用栅格（raster format）格式，分辨率为3角秒（约90×90米像素）。

包含的土壤属性如下：
有效含水量（Available Water Capacity）；
全土容重（Bulk Density - Whole Earth）；
黏粒（Clay）；
有效阳离子交换量（Effective Cation Exchange Capacity）；
氯化钙浸提pH值（pH - CaCl2）；
粉粒（Silt）；
砂粒（Sand）；
全氮（Total Nitrogen）；
全磷（Total Phosphorus）。

时间覆盖范围（近似）：1950年-2013年；
空间分辨率：3角秒（约90米）；
单属性栅格图总数量：18幅；
单图层有效像素数：2007百万（49200×40800）；
压缩前总容量：约8GB；
压缩后总容量：约4GB；
数据许可协议：知识共享署名3.0（CC BY）；
目标数据标准：GlobalSoilMap规范；
数据格式：GeoTIFF（GeoTIFF）格式。

数据溯源：
本数据集的数字土壤属性图及其不确定性信息，通过整合多源点位土壤观测数据，并采用三维空间建模技术生成。

土壤数据源清单：
国家级土壤点位数据来源于两个渠道：
（一）由澳大利亚各州及领地土壤机构协助整理的数据集（Searle, 2014）。各州市的土壤数据库被整合为遵循NatSoil点位模式（NatSoil Site Schema）的统一数据库，该库包含2013年各机构公开数据库中所有土壤剖面的形态学与实验室分析数据。
（二）基于澳大利亚国家地球化学调查（NGSA）采集的土壤样本，通过澳大利亚可见-近红外光谱数据库（Viscarra Rossel和Webster, 2012）得到的土壤属性光谱估算值（de Caritat, P & Cooper, M, 2011）。

标准深度归一化处理：
针对清单中所有深度层级的各土壤属性数据，我们采用两种方法将其归一化至6个标准深度：当数据覆盖深度多于2层时，通过拟合保质量样条函数（mass preserving spline, Bishop等, 1999）推导得到标准深度的属性值；当仅覆盖2层深度时，则采用Clifford等（2014）提出的插补方法。

空间建模：
本数据集的数字土壤图通过结合三维数据挖掘-克里金（kriging）法与蒙特卡洛重采样（Monte Carlo resampling）技术生成，以实现不确定性估算。该方法首先建立点位观测土壤属性与40余种环境协变量（包括遥感数据、气象数据、数字高程模型及其地形衍生变量、伽马辐射测量数据及其他地球物理数据）之间的统计关系，随后对模型残差进行克里金插值。数据挖掘环节采用R语言（R Core Team, 2013）环境下的Cubist数据挖掘软件（Rulequest Research., 2008），地质统计建模则使用gstat工具包（Pebesma, 2004）。该混合模型可输出土壤属性的定量估算值，模型两部分的不确定性均被量化并以90%置信限表示。本方法的详细说明可见Viscarra Rossel等（2015a）、Viscarra Rossel与Chen（2011）及Viscarra Rossel（2011）的相关文献。