A Representativeness Assessment of the Angell–Korshover 63-Station Network Sampling Based on Reanalysis Temperature Data

Abstract

Global climate observations from ground stations require an evaluation of the effectiveness of a station network, which is often an assessment of the geometric distribution of [Formula: see text] points on a sphere. The representativeness of the Angell–Korshover 63-station network (AK-network) is assessed in this paper. It is shown that AK-network can effectively sample the January global average temperature data of the NCEP/NCAR Reanalysis from 1948 to 2015 when estimating inter-decadal variations, but it has large uncertainties for estimating linear trends. This paper describes a method for the assessment, and also includes an iterative numerical algorithm used to search for the locations of 63 uniformly distributed stations, named U63. The results of AK-63 and U63 are compared. The Appendix explains a problem of searching for the optimal distribution of [Formula: see text] points on a unit sphere in three-dimensional space under the condition of the maximum sum of the mutual distances among the points. The core R code for finding U63 is included. The R code can generate various interesting configurations for different [Formula: see text], among which one is particularly surprising: The configuration of 20 points is not a dodecahedron although the configurations for [Formula: see text], and 12 are tetrahedron, octahedron, cube, and icosahedron, respectively.