Abstract
The problem of analyzing continuous and discontinuous meteorological surface parameters is investigated. Peculiarities of fitting mathematically a first and second degree surface to observational data are investigated from the standpoint of transforming and constraining the observed values and of the influence of observations as a function of distance. For two data density fields a comparison of results is made between the mathematical surfaces and a simplified procedure of selecting the nearest observation. Verification is based upon observed station values which are withheld from the surface fitting procedures. A significant improvement is derived by transforming the variables to ensure closer fitting in a particular range of the variable. For the area of influence under consideration, no appreciable differences were obtained by the incorporation of a distance weighting factor. It was found necessary to restrain the mathematical surfaces to ensure physical reality. Although the more complex mathematical surface does fit the large-scale nonlinear variations more accurately, its sensitivity to the distribution of data and smaller-scale fluctuations produced errors which were larger than those obtained using a linear surface.
