AUTOMATING MINERAL IDENTIFICATION IN STALAGMITES USING HYPERSPECTRAL IMAGING AND MACHINE LEARNING

Stalagmites (secondary carbonate deposits in caves) can provide nearly continuous records of past climate in continental settings. We used hyperspectral imaging (HSI), a relatively less destructive technique, to identify mineralogy in stalagmites from different regions. This method was tested on a stalagmite sample, called Stalagmite MAJ-4, from Madagascar. Results obtained from this sample were cross-checked with a petrographic study of oversized thin sections and X-ray diffraction (XRD) analyses on selected intervals. A machine learning algorithm was then developed to automate mineral identification and classification across the sample.

To test the replicability of this method, we analyzed more stalagmite samples from Trapiá and Furna Nova caves in Brazil. Preliminary HSI results from these Brazilian stalagmites revealed distinct spectral signatures of various carbonate minerals. These signatures enable mineralogy interpretation by comparison with a reference spectral library, thereby potentially minimizing the need for traditional methods, i.e. XRD and petrography. HSI analysis is very useful for mineralogy identification, which could be used as a guide for further mineralogical and geochemical analysis, offering insights into environmental condition during the growth of stalagmites.