Numerical correlation of many multidimensional geological records
Abstract
Korelacja numeryczna wielowymiarowych danych dla kilku profili geologicznych
It is frequent task to correlate profiles or cores basing on different measurements performed on the series of samples. The difficulty arises when there are many profiles and none is the main or reference one. The reason is that the number of possible correlations grows exponentially with the number of profiles. To resolve the problem a Monte Carlo method is adopted here, what makes it very probable to discover the best correlations in a reasonable amount of computing time. The quality of a correlation is measured by a metric of dissimilarity of the samples. The final result, given in graphical form, has a form of lines connecting correlative samples from different profiles. The number of lines (correlations across profiles) is user-defined and can vary from one to dozens. The number of profiles, samples, and variables depends only on the computational resources. Large problems need longer computation times to achieve stable results.
It is frequent task to correlate profiles or cores basing on different measurements performed on the series of samples. The difficulty arises when there are many profiles and none is the main or reference one. The reason is that the number of possible correlations grows exponentially with the number of profiles. To resolve the problem a Monte Carlo method is adopted here, what makes it very probable to discover the best correlations in a reasonable amount of computing time. The quality of a correlation is measured by a metric of dissimilarity of the samples. The final result, given in graphical form, has a form of lines connecting correlative samples from different profiles. The number of lines (correlations across profiles) is user-defined and can vary from one to dozens. The number of profiles, samples, and variables depends only on the computational resources. Large problems need longer computation times to achieve stable results.