青海省表层土壤属性数字制图

Digital Mmapping of Topsoil Attributes in Qinghai Province

对于土壤景观复杂的大区域,样点往往较为稀疏,如何准确地进行土壤预测制图仍是一个需要研究的问题。本文以青海省为研究区,基于近年采集205个土系调查点数据,采用随机森林模型,分别建立了表层(0~20 cm)土壤全氮、有机碳、粉粒含量和pH四个基本土壤属性与环境协同变量(海拔、坡度、地形湿度指数、年降水量、年平均气温、归一化植被指数、地表温度和地表反射率)之间的定量关系模型,对该地区进行了土壤多要素预测制图,分析了影响土壤空间变异的控制性因素。交叉验证结果显示,全氮、有机碳、粉粒含量和pH的R^2分别是0.61、0.53、0.47和0.54,这说明随机森林模型可解释47%以上的土壤空间变异。表层土壤全氮和有机碳空间分布趋势东南高,西北低,pH呈现出相反的空间模式;粉粒含量东高西低,预测结果高值出现在柴达木盆地和南部玉树、果洛地区。环境变量的重要性分析表明,年降水量对表层土壤全氮、有机碳、pH空间分布模式具有控制性影响,夜间地表温度与表层土壤粉粒含量空间变异具有较强的协同关系。

In Tibetan Plateau, ecological environmental protection and agriculture - animal husbandry management urgently require the fine and accurate spatial distribution of soil information. However, there are only old and rough conventional soil maps at present. The data were derived from 205 soil survey samples of Qinghai Province in recent years. Quantitative models of four basic soil attributes and environmental synergistic variables in 0 ~ 20 cm topsoil were established with random forest method. The basic soil attributes included total nitrogen (TN), soil organic carbon (SOC), silt content and pH and the environmental synergistic variables mainly included altitude, slope, topographic wetness index, annual precipitation, annual average temperature, normalized vegetation index, land surface temperature and land surface reflectance. Through the quantitative model, the mapping of soil multi- factor prediction was carried out, and the control effect of soil- forming environmental factors on the spatial variability of soil properties was analyzed. The results showed that the spatial distribution of TN and SOC in the topsoil was higher in the southeast and lower in the northwest of Qinghai Province, while the pH showed an opposite spatial pattern. Silt content was higher in the east and lower in the west of Qinghai Province, and the high predicted value also appeared in the Qaidam Basin, south of Yushu and Guoluo areas. The cross validation results showed that the R2 of TN, SOC, silt content and pH were 0.61, 0.53, 0.47 and 0.54, respectively. The random forest models accounted for more than 47% of soil spatial variability. The annual precipitation controlled the spatial distribution patterns of TN, SOC and pH in the topsoil. And there was a strong correlation between the surface temperature at night and the spatial variation of the silt content in the topsoil.