稀疏观测数据的空间内插方法的分析与比较

ANALYSIS AND COMPARATION TO THE METHODS OF SPATIAL INTERPOLATION WITH SPARSE OBSERVING DATUM

依靠稀疏样本数据描述空间指标在地理空间上的连续分布是实践中常常遇到的问题，内插是解决问题的基本方法。以云南省为实验区，15年来6～8月的均温为空间描述指标，在90m×90m的分辨率水平上，分别应用不同方法进行内插处理，对内插结果进行了分析和比较。结果表明关联函数法内插效果最好，较好地体现了云南北低南高、西高东低的总体气温变化规律，同时又体现了河谷地带的干热特点。如果无法建立关系函数，则使用克里金插值效果较好，反距离内插法次之，趋势面分析和泰森多边形内插法效果最差。趋势面分析中的高阶多项式内插不优于低阶多项式。

It is a usual problem in practice that some spatial indexes are described for their continuous distribution in geographic space by sparse observing datum. And interpolation is a basic way to solve the problem. That study based on Yunnan province uses different interpolations＇ methods to deal with the average tempreture from June to August for their continuous distribution in geographic space with 90 m resolution. Then the results were analyzed and compared, which reveal that incidence function interpolation has the best effect. It shows the regulation of temperature change in Yunnan Province, which is low temperatures in north, high temperatures in south, and high temperatures in east and low temperatures in west. And also it presents the temperature characteristics of the dry and hot gorges. If it is impossible to build an incidence function, then Kriging interpolation is the second, inverse distance weighted interpolation is the third, polynomial surface interpolations and Thiessen polygon interpolation is the worst choice. Polynomial surface interpolations also show that the higher power of polynomial interpolation is not better than the lower.