祁连山地甘肃臭草斑块土壤水分的空间自相关分析

Spatial autocorrelation analysis on soil moisture of Melica przewalskyi patch in a degraded alpine grassland of Qilian Mountains,Northwest China

用传统统计学方法探讨土壤水分空间分布的前提条件是用于研究的数据在统计上必须独立且均匀分布，但土壤水分在空间上一般存在空间自相关，这些空间自相关包含一些有用信息。本文利用Moran，系数研究了祁连山北坡甘肃臭草（Melicaprzewalskyi）单优种群斑块浅层剖面（0—30cm）土壤水分空间自相关关系、空间相关尺度，建立了各层土壤水分影响因子的线性回归模型和空间自回归模型，并比较了2种模型的分析结果。结果表明：各层土壤水分均具有空间正相关性与空间集聚特征，10～20cm层土壤水分的空间集聚特征较0-10cm和20～30cm土层更为明显；土壤水分的Moran I系数随着分析间隔距离的增大而减小，在〈4m的范围内各层土壤水分均存在正的空间自相关关系。影响甘肃臭草浅层剖面土壤水分空间分布的因素在不同土层不尽相同；空间自回归模型的LIK值和尺。的值比线性回归模型的值要大，从而显示出空间自回归模型的解释能力要优于经典线性回归模型．．

A prerequisite in using conventional statistical methods, such as regression models in investigating spatial distribution of soil moisture, is that the data regarding soil moisture should be statistically independent and identically distributed. However, soil moisture generally exists with spatial autocorrelation to some degree, which contains some useful information. In this paper, the spatial autocorrelation analysis of soil moisture in Melica przewalskyi patch was investigated based on Moran＇ s I index on the north slope of the Qilian Mountains. Moran＇ s I was applied to de- scribe spatial autocorrelation of soil moisture, and analyze the scales of spatial autocorrelation. Meanwhile, standard multiple linear regression model and spatial autoregressive model of soil moisture were constructed. The results showed that distribution of surface soil moisture all dis- played spatial autocorrelation characteristics. In addition, the spatial aggregation characteristics of the 20-30 cm depth were higher than that of the 0-10 and 10-20 cm depths. It was found that the Moran＇ s I decreased with the increase of the scale of spatial analysis. The spatial auto- correlation of surface soil moisture resulted from different soil depths. At the 10-20 cm depth, the community height and Melica przewalskyi coverage had significant effects on the spatial auto- correlation, while at the 20-30 cm depth, the Stipa krylovii coverage and community height sig- nificantly affected the spatial autocorrelation. Our analysis showed that spatial autoregressive model was better than the standard multiple linear regression model due to the spatial autocorrela- tion exerting more impact on the latter one.