县域土壤养分离群样点检测及其合理采样数研究

为进一步研究县域耕地土壤养分质量评价,增加采样数据的准确性和减少采样分析成本,本试验以白水县为研究区域,对2226个采样点的土壤样品中有效磷、碱解氮和速效钾采用四分位数法和局部Moran’s I方法进行全局和局部离群样点的去除,并对去除后的采样点进行空间特征分析。然后采用经典统计学(Cochran公式)方法和地统计学(普通克里金)方法对县域土壤养分合理采样数目进行讨论,比较两种方法各自的优越性。结果表明:碱解氮和速效钾均检测出3个全局离散样点,有效磷全局离群样点数目为7,三种养分局部离群样点检测数目分别为89、90和92;三种土壤养分均表现出很强的空间相关性,变异系数随着离群样点去除而降低;对于土壤养分大体概况研究,Cochran方法确定样本较为适合,而由克里金插值法确定的合理采样数目更能精确地反映局部土壤养分空间状况。

Outlier identification and reasonable sampling number of soil nutrient at county level

The aim of our experiment was to provide more accurate datasets and less cost of sampling and analysis for evaluating the nutrient quality of arable land at a county level.The Quartile method and the local Moran’s I method were used to identify the global and local outlier of AP,AN and AK in Baishui County.Spatial characteristics were analyzed after removing the outlier.The classical statistics method(Cochran formula) and geo-statistics method(Kriging) were used to calculate the reasonable sampling number of soil nutrient.The results showed that 3 global outliers were detected in AN and AK,the number of global outlier in AP was 7.The local outliers in AN,AK and AP were 89,90 and 92,respectively.Three of the soil nutrient performed significant spatial correlation.The coefficient of variation(CV) fell down after outliers were removed.Cochran method was more suitable for the research of the general trend of soil nutrient while the reasonable sampling number calculated by Kriging method could present the local spatial characteristics of the soil nutrient more accurately.