Developing a Taxonomic Soil Dataset from SSURGO for Hydrological and Water Quality Modeling

The file provided has two R scripts that can be used to aggregate SSURGO map units up to their subgroup taxonomic level. The script divides the soils listed in the SSURGO data into five classes: soils having fragipan, wet soils, wet soils having fragipan, soils having shallow bedrock (lithic), and soils that do not exhibit any of these four characteristics (general).The first script, 'fun_standerdise_layers.R’ extracts the soil mukey data from 'SoilDB' R package (Beaudette et al., 2021), connects it with the SSURGO soil data obtained from the SWAT website (https://swat.tamu.edu/data/), and standardizes the soil data with uniform soil layer depths considering depth-weighted average. The second script, 'create_taxonomy_soils.R' reads this data and first classifies it into taxonomic groups and then clusters it for each of the four hydrologic group within each taxonomic group by considering similarity in soil hydraulic conductivity, soil available water capacity, soil erosivity, and soil depth. A SWAT model developed using the taxonomic data was shown to show similar nutrient and hydrologic loads when compared with a SWAT model developed using SSURGO. The parameter distribution of both the models were also found to be similar. However, the taxonomic model reduced simulation time by half, as the taxonomic dataset mitigates boundary-based discontinuity issues that arise in soil surveys when compiling the SSURGO dataset.

本次提供的文件包含两个R脚本，可用于将土壤地理数据库（SSURGO）的制图单元聚合至其亚类分类学层级。这些脚本将SSURGO数据中收录的土壤划分为五类：含硬磐层的土壤、湿生土壤、含硬磐层的湿生土壤、具有浅基岩（岩性）的土壤，以及不具备上述四类特征的普通土壤。
首个脚本为`fun_standerdise_layers.R`，其从"SoilDB" R包（Beaudette等人，2021年）中提取土壤制图单元键（mukey）数据，并将其与从SWAT网站（https://swat.tamu.edu/data/）获取的SSURGO土壤数据进行关联，同时基于深度加权平均法，以统一的土层深度对土壤数据进行标准化处理。
第二个脚本为`create_taxonomy_soils.R`，其读取上述处理后的数据，首先将其划分为分类学组别，随后针对每个分类学组内的四个水文分组，基于土壤导水率、土壤有效持水量、土壤侵蚀力以及土层深度的相似性进行聚类分析。
基于分类学数据构建的土壤与水评估工具（SWAT）模型，与基于SSURGO数据构建的SWAT模型相比，其模拟的养分与水文负荷结果具有相似性。两款模型的参数分布也被证实具有相似性。但该分类学模型将模拟时长缩短了一半，原因在于分类学数据集可缓解在编制SSURGO数据集过程中，土壤调查所产生的基于边界的不连续性问题。