Spatial variability and hydro/geochemical profiling of the elemental composition of mineral deposits and drip water from caves using unsupervised chemometric modelling

Caves are complex systems characterized by high variability in trace element signatures that simple geochemical analysis cannot detect since useful information hidden in the dataset may be lost. Sufficiently robust unsupervised statistical approaches may provide these significant signatures for cave hydro/geochemical profiling. The present study focused primarily on quantifying the relevant contributions of elements and their relationships to the variability of the hydro/geochemical profiles of speleothems, sediments and drip water within a cave using Principal Component Analysis (PCA) after Varimax rotation and two-way joining analysis as a clustering method. This prospective study was conducted on the Ursilor (Bears') show cave (Romania). The contents of Ca, Mg, Sr, Al, P, S, Fe and Zn in the mineral deposits were determined via X-ray fluorescence, while the content of these elements in drip water was determined via inductively coupled plasma-optical emission/mass spectrometry. The contents of HCO3-, Cl-, F-, NO3-, and SO42-; pH; electrical conductivity; and organic and inorganic carbon fractions were also determined. In the case of speleothems, the determinations were performed on randomly selected spots, reflecting the composition at various growth times. The PCA revealed that the geochemical profile of speleothems can be described by four principal components (PCs), represented by 71% for stalactites and 82/ 92% by three PCs for stalagmites/sediments. The strong signatures of Ca, Mg/Ca, K, and detrital elements (Al, Si, P, S) in stalagmites/sediments in proportions of 41/46%; K, Fe and Zn 30/32%; and Sr and Sr/Ca 11/15% were retained by PCA. The geochemical profile of the stalactites was described by Mg (35%), P, Zn, and Sr (21%); Al and Si (15%); and S, Ca and Mg/Ca (10%). Reducing the dimensionality of the data matrix to alkaline earth elements, two PCs described 95% of the mineral deposits, including the strong signatures of Mg/Ca and Sr/Ca, exceeding 50% of the variability in the first PC. Stalactites and stalagmites were primarily grouped based on their geochemical profile and less so on location, yielding two clusters/four subclusters of stalagmites. The PCA plot did not show a separation between stalactites and stalagmites, as their geochemical profiles differ significantly only in terms of Fe concentration. The hydrochemical profile of drip water was described by 4 PCs at a percentage of 95%: (i) alkaline elements, alkaline earth elements, anions and organic carbon and (ii) trace elements and DOC. The profile of Ca2+-HCO3- imprint drip water was described by two PCs at proportions of 66% (Ca, Mg, Na, K, Mg/Ca, Sr/Ca, Ba/Ca ratios, HCO3- and organic carbon) and 74% by trace elements (Co, Ni, Ti, Fe, Mn, V and Si). The transition trace elements have very complex chemical behaviour and appear in drip water as Ca-/Si-rich inorganic/organic colloids and simple ions. Although complementary to the PCA, the two-way joining analysis in a heatmap did not offer relevant information in the description of the hydro/geochemical profile of the karst system compared to PCA. A study of a farmed calcite sample highlighted its difference in composition compared with stalagmites in terms of Mg, Si, K, Zn, Sr and Sr/Ca. Future development could include the application of PCA to seasonally collected farmed calcite and drip water samples to obtain hydro/geochemical profiles of stalagmites in the making to determine the seasonal leaching of metals from bedrock/soil by drip water, inclusion in stalagmites, and seasonal growth. The practical significance of this paper for trace element research is of great concern because it allows these recorded indicators to be connected to the ancient climate and environment of stalagmites.