Validation of a database of mean uranium, thorium and potassium concentrations in rock samples of Portuguese geological units, generated of literature data

The European Atlas of Natural Radiation, recently published, contains a collection of maps of Europe showing the levels of natural sources of radiation. Among the lacunae of the Atlas are maps of U, Th and K concentrations in rocks due to lack of European-wide geochemical surveys of bedrock units. The objective of this paper is to investigate the usability of scattered geochemical data of rock samples for large-scale mapping of U, Th and K concentrations in geological units. For this purpose, geochemical data were compiled from literature sources to produce a geochemical database (LIT database) that includes 2817 entries of U, Th and K concentrations measured in rock samples of geological units outcropping in Portugal. Given the methodical heterogeneity within LIT database, the influence of the geochemical analysis techniques was assessed through a three-way analysis of variance (ANOVA) using geological units, geochemical analysis techniques and loss on ignition (LOI) as categorical variables. The percentage of variation explained by geological factors was large (>35%), while the percentage of variation explained by the geochemical analysis techniques and LOI was generally lower than 5%. The geological factors were the main source of variability in the data, followed by the error component which can be assumed to represent the true spatial variability of geochemical concentrations. The pairwise comparison of the least square (LS) means computed through the ANOVA for each geochemical analysis technique indicates that LIT database can be considered consistent within itself, thus, reliable. In order to validate the usability of literature data the terrestrial gamma dose rate (TGDR) calculated from LIT database (TGDRcalc) was compared to the TGDR displayed in the Radiometric Map of Portugal (TGDRobs). The correlation between TGDRcalc and TGDRobs was highly significant (p < 0.001) and the results of a paired sample t-test and Wilcoxon median tests indicate that the differences between the arithmetic means of TGDRcalc and TGDRobs were not statistically significant (p = 0.126 and p = 0.14, respectively). Distributions of TGDRcalc and TGDRobs were seemingly equal according to the Kolmogorov-Smirnov and Anderson-Darling tests. Although, systematic discrepancies between TGDRcalc and TGDRobs were observed for sedimentary rocks, the compatibility of the RMP and LIT databases can be considered acceptable, which implies that the estimation of the contents of terrestrial radionuclides using literature data for large-scale mapping of U, Th and K contents in geological units is reasonable.