---

作 者 ：连纲,郭旭东,傅伯杰,虎陈霞

期 刊 ：生态学报 2008年 28卷 3期 页码：946~954

关键词：土壤养分；空间变异；环境因子；空间预测；小流域；

Keywords:soil nutrients, spatial variation, environment indicators, spatial prediction,

摘 要 ：在黄土高原小流域尺度上，地形和土地利用是影响土壤属性变异的重要因素。在横山县朱家沟小流域采集111个表层土样，比较不同土地利用及景观位置土壤养分变异及分布特征，分析土壤养分与地形因素的关系并利用相关环境信息进行空间预测。结果表明，不同土地利用类型土壤养分存在显著性差异。有机质和全氮含量表现为坝地最高，灌木地最低，而全磷含量以梯田最高。在不同景观位置，沟平地有机质和全氮含量显著高于其他景观位置，而全磷差异不显著。土壤有机质与复合地形指数CTI、汇流动力SPI、沉积物运移指数STI显著负相关，全氮与沉积物运移指数STI显著负相关，而全磷只与坡度β显著负相关。多元线性逐步回归模型预测土壤有机质和全氮较好，而全磷预测结果较差；回归-克里格预测有效地减小了残差，与实测值较为接近，预测精度高于回归预测。

Abstract:The study on the spatial variability of soil properties is vital for sustainable land management. Both topography and land use are pivotal factors which affect the variability of soil properties on the catchment scale in the loess hilly area. This study analyzed the spatial variation of soil nutrients from different land use types and landscape positions, based on the data of 111 surface soil points (0~20cm) in the Zhujiagou catchment on the Loess Plateau. We measured soil organic matter (SOM), total nitrogen (TN), and total phosphorus (TP) and used correlation analyses to determine relationships between soil nutrients and terrain attributes. Finally, terrain attributes and land use types were used to predict the spatial distribution of the soil properties by using multiple-linear regression analysis and regression-kriging. The results showed that concentrations of these soil nutrients were very low in the surface soil, and the coefficients of variation for soil properties were moderate. Soil nutrients were significantly different among different land use types. Higher values of SOM and TN were found in check-dam farmland and lower values from shrub land. Significant differences among landscape positions were observed for SOM and TN, and concentrations of SOM and TN located in the flat valley position were higher than in other positions. There were negative correlations between SOM and compound topographic index (CTI), stream power index (SPI), and sediment transport index (STI). Similarly, TN has negative correlation with sediment transport index (STI), and a significant negative correlation was found between TP and slope (β). To some extent, correlations between these terrain attributes and soil properties reflect patterns of soil development caused by water flow through and over the landscape. From the regression models, we determined that variability of measured soil properties ranged from 13% to 51%. The regression model for TN had the highest R2 value, followed by SOM and TP. The regression models were relatively precise for the SOM and TN, but variation was large with a high smoothing effect on the predicted values. For TP, the predicted result was very poor. To further explain the variations, we combined step-wise regression with residuals interpolated using kriging. Results showed that regression-kriging can improve the accuracy of prediction.

---