Before wide-spread of computers and invention of sample planning methods with random sampling, people took samples on the "most typical" points of territory, and then draw isolines by hand, taking into account type of vegetation, soil etc, known to them from first-hand field experience.
This give better results than regular sampling grids - less samples and more information about actual field of concentration of some element.
Being taught in the University these old pre-computer era methods, I thought about using them in GIS.
Idea of method is quite simple: we repesent qualitive difference between areas as set of boolean variables - indicators, and compute influence of each of them using least squares method.
This is my first paper ever published in the magazine. It was published in Vestink MGU, Geography Series in 1991. (PDF, Russian).
This is more recent paper, written together with N.Kosheleva (1997) (PDF, English)