The radioactive isotope of hydrogen, known as tritium (3H), is very often used as a dating tool in hydrogeological studies, since it enters the water cycle as part of the water molecule through precipitation. However, the assessment of groundwater transit times and recharge often requires knowing the local historical records of tritium levels in precipitation during the previous seven decades, or the tritium in precipitation (TIP) time series. Here, we compiled all tritium records in precipitation in Greece, with the majority of stations showing sporadic measurements, with the aim of reconstructing a TIP for groundwater dating purposes. The monitoring station of Vienna proved to be more reliable for filling the gaps in the time series than the Ottawa station. Conventional methods to fill the TIP gaps, such as the correlation method (CM), were compared with more advanced machine learning tools, such as the Convolutional Neural Networks (CNN). The comparison showed that the artificial Intelligence Method (AIM) performed best, due to its ability to capture complex nonlinear trends that are usually inherent in real-world data. The preliminary qualitative assessment of groundwater tritium data from Greece, in comparison to the TIP, showed the occurrence of groundwaters of mean transit times ranging from a few years to decades. Better incorporation of groundwater transit times and recharge rates into the study of aquifer systems is essential for developing strategies for sustainable water management in Greece and worldwide.
Read full abstract