Detrital zircon (dZ) U-Pb geochronology is a quickly developing field of geosciences; number of publications relying on dZ data is growing each year.

One of the most intriguing advances in dZ geochronology is studying our planet by manipulating big data from thousands of rock samples, of all ages, collected globally. For example, the largest, to our knowledge, compilation contains of about 200 thousand of dZ dates (Voice et al., 2011).

Apples need to be compared with apples, not with oranges. Kindergarten alumni know this.

This rule applies to dZr as well. However, how often do you see comparisons of different sets of dZr data that were filtered and analyzed differently? More than often' — should be an honest answer.


Let us examine a hypothetical example:

A researcher 'John' is interested in isolating all significant age populations in his sample that contains a wide variety of Mesozoic, Paleozoic and Precambrian grains. He also wants to compare his distribution to the one gathered by another researcher, 'Jane' from the same sedimentary formation. John is very careful not to filter out important populations, while Jane was particularly interested in young Mesozoic grains and did not care about older, mostly discordant populations. John used very weak discordant filters, while Jane used stringent ones. As a result, the two populations will not be statistically indistinguishable (which is a 'statistically-correct' expression to say that they will be, indeed, different).

In most cases raw data from ICPMS are reducted in either of the two data reduction packages: Iolite or Glitter. There, analyses are corrected for instrument drift and downhole fractionation; uncertainties are propagated; drill-throughs, zonal grains or other 'undesireables' are detected and such analyses omitted.

But what to do with reducted data? It still needs lots of filtering and computations. Non-conforming analyses (e.g., large degree of discordance, high analytical error, etc.) need to be filtered out; discordance need to be calculated (there are 2 way of doing it); 'best ages' need to be computed. Which age to use for the 'best age': ²⁰⁶Pb/²³⁸U or ²⁰⁷Pb/²⁰⁶Pb? When to switch between the systems?

Data also need to be plotted. Not once, not twice. Data need to be manipulated with, and immediately plotted, so that robustness of age populations are verified, and effects of any change are obvious.

In overwhelming amount of cases, MS-Excel is used for such 'games'. In best cases, some kind of macro's are used to help at least with filtering. In the worst: do it by hand, praying that you are attentive enough not to miss an analysis.

But, wait! What if you need to compare your data with someone else's? HOW do you do that? Should you use the same filtering that the other researcher did?