Abstract: Robust variogramis the basic guarantee of the effective spatial prediction using kriging methods. The existing variogram estimators behave poorly in practice. Some are not robust against the outliers in data; others are complex in algorithmin order to achieve robustness or hard to be used widely. The authors propose a simple variogramesti mator based on medianfiltering method, by taking into account correlation of neighbor hoodlags to improve its robustness against outliers. The robustness properties between the existing estimators and the proposed one are compared and analyzed. Experiments are carried out with data existing different types of outliers. The results show that the proposed variogram estimator is not only highly robust against outliers, but also easy to be used generally.

Key words: variogram, median filtering method, robust estimator, outlier, lag, noise

摘要： 稳健变差函数是克里金进行有效预测的根本保证, 而现有的计算方法存在以下几个问题:稳健性差、算法复杂、普适性差。为此, 提出了一种以中值滤波方法为基础, 兼顾邻域间相关性的高稳健、简单的变差函数计算方法。对新方法和现有方法的稳健性进行了分析和对比。最后, 使用存在特异值的数据对所提出的方法进行了检验。结果表明, 新方法不但具有高稳健性, 还具有普适性强且实现简单的优点。

关键词: 变差函数, 中值滤波, 稳健方法, 特异值, 滞后距, 噪声