So, I am unsure how I should interpret the situation when the majority (let’s suppose 3/4) of my data aligns above the LMWL, while a smaller part (suppose 1/4) aligns below the LMWL.

I conduct research at a riverbank filtration site, where the water being exploited is a mixture of surface water from the river and native groundwater. I have data from the river, the production wells, and a piezometer outside the river’s influence, representing landside (native) groundwater.

To briefly sketch the background of the study area - the aquifer is shallow (up to 10 m below ground level) and consists of alluvial gravels, so groundwater flow (vertical and horizontal) is rapid. The river from which the water is abstracted is fed by meltwater (melting snow falling in the mountains) for a considerable part of the year - note on the schematic figure that some river points are strongly depleted, and some resemble native groundwater (baseflow recharge period in the summer (?)).

From what I have read in the literature, citing, for example, the book ‘Environmental Isotopes in Hydrogeology’ (Clark and Fritz, 2013): „Most meteoric and subsurface processes shift the d18O-d2H signature of waters to a position below the local meteoric water line. It is rare to find precipitation or groundwater that plots above the line, i.e. showing a deuterium excess or 18O depletion.” However, this is precisely the situation in my data, and I am unsure how to interpret it.

I found item https://linkinghub.elsevier.com/retrieve/pii/S2214581818300132in which there is a sentence “Groundwater (mainly spring water) plots on or above the LMWL for the UR (Fig. 4a), indicating that infiltration occurs without change to the isotope composition of precipitation”. However, I do not know whether this would be a sufficient inference for my data.

What is certain is that for most of my points, relative to the LMWL, there is an enrichment of d2H and a depletion of d18O - but, again, I recall - not for all of them (see schematic figure).

The post has already gotten long, so I’ll end here - but I’m happy to answer additional questions in the comments. I would be grateful for any suggestions. Maybe someone had a similar issue. Obviously, I cannot share the raw data as it will be used to publish the paper, that's why I did this simple schematic plot.