INDAL 1995-96: Spatial Statistics
The quality or the
concentration of (AlO, SiO) of the Radhanagari bauxite deposit was
predicted via kriging. INDAL was using a custom-built FORTRAN based
software for kriging, which was yielding unacceptably high prediction error.
Since the software offered only a limited choice in the variogram model
selection, MATHEMATICA programs doing Exploratory Data Analysis, suggesting an
appropriate variogram model, were written first. It turned out that the
wave-effect variogram model, as opposed to the spherical one, suggested by the
software, fitted the data better. Since the software did not have a provision
for fitting the wave-effect model, C programs were written fitting the
wave-effect model and also for the eventual kriging. Thus although the
prediction error was reduced considerably, it was still high (due to the noise
in the data) for their purpose, and was reported accordingly.