大面积高寒山区土壤养分空间预测与管理分区

土壤养分空间分布是土壤质量监测与可持续管理的关键信息。在大面积高寒山区,土壤调查样本较为稀疏,准确地预测土壤养分的空间变异并建立合理的管理分区,具有重要科学意义。本文基于青海省土系调查的205个土壤剖面数据,利用随机森林模型结合环境因素变量(地形、气候、植被、遥感数据), 分别建立了表层(0-20cm)土壤全氮(TN)、全钾(TK)、全磷(TP)与之间的定量关系,对青海省土壤养分含量的空间分布进行了预测；在此基础上,结合全国土壤养分的分级标准,利用投影寻踪法,生成了土壤养分的管理分区。留一交叉验证结果显示,全氮(TN)、全钾(TK)、全磷(TP)空间预测的R_²分别是0.89、0.85、0.82,可解释土壤养分空间变异的80%以上,表明随机森林模型和环境因素变量结合可以在稀疏样本条件下有效地预测大面积高寒山区土壤养分空间变异。青海省土壤养分呈现东高西低的分布模式,土壤综合养分高等级 出现在南部的玉树、果洛、黄南和东部的湟水谷地地区；低等级主要分布在柴达木盆地、可可西里、海南州中南部；全省土壤综合养分分级均在中上等级以上 ,占全省面积是81%。分析发现,植被是影响青海省表层土壤养分全氮(TN)、全磷(TP)、全钾(TK)空间分布的主要环境因素,其中年降水、地表温度是影响青海省表层土壤全氮(TN)空间模式的重要因素；地表覆被、海拔和地表温度等环境因子对表层土壤全磷(TP)的空间变异占主导作用；年降水、昼夜温差是影响表层土壤全钾(TK)的空间模式的重要因素。

Spatial Prediction and Management Zoning of Soil Nutrients in Large-scale Alpine Mountainous Areas

The spatial distribution of soil nutrients is the key information for soil quality monitoring and sustainable management. In a large area of Alpine mountainous areas, soil survey samples are relatively sparse. How to accurately predict the spatial variation of soil nutrients is a problem worthy of study. Taking Qinghai Province as the study area, Based on the data of 205 sample points of soil series survey in Qinghai Province in recent years, using random forest model and environmental factor variables (terrain, climate, vegetation and remote sensing data), the quantitative relationships between surface (0-20cm) soil total nitrogen (TN), total potassium (TK) and total phosphorus (TP) were established respectively, and the spatial prediction of soil nutrient content in Qinghai Province was carried out; On this basis, the management zoning of soil nutrients was generated by using the projection pursuit method and the national soil nutrient classification standard. The cross validation results show that the R_² of spatial prediction of total nitrogen (TN), total potassium (TK) and total phosphorus (TP) are 0.89, 0.85 and 0.82 respectively. The model can explain more than 80% of the spatial variation of soil nutrients, indicating that the combination of random forest model and environmental factor variables can effectively predict the spatial variation of soil nutrients in large-area Alpine mountainous areas under the condition of sparse samples. The distribution pattern of soil nutrients in Qinghai Province is high in the East and low in the West. The high level of soil comprehensive nutrients appears in Yushu, Guoluo, Huangnan in the South and Huangshui Valley in the East; The lower grades are mainly distributed in Qaidam Basin, Hoh Xil and the south central part of Hainan prefecture; The soil nutrient classification of the whole province is above the middle and upper grades, accounting for 81% of the total area of the whole province. It is found that vegetation is the main environmental factor affecting the spatial distribution of total nitrogen (TN), total phosphorus (TP) and total potassium (TK) in topsoil in Qinghai Province, among which annual precipitation and surface temperature are important factors affecting the spatial model of total nitrogen (TN) in topsoil in Qinghai Province; The spatial variation of total phosphorus (TP) in topsoil was dominated by environmental factors such as surface cover, altitude and surface temperature; Annual precipitation and temperature difference between day and night are important factors affecting the spatial model of total potassium (TK) in topsoil..