Detrital Zircons Can Give False Geological Ages
Zircons are a gold standard for dating. They can yield ages that are statistically significant, but geologically meaningless.
How confident are geologists in the ages of formations they study? The story often told is that radiometric dating produces dates that are super-reliable, because lab rates of radioactive decay don’t lie. The part of the story not told, though, is that many sources of bias can creep in. What constitutes a good sample? How many samples must be collected to converge on a reliable date? How far and wide should samples be collected? What should the geologist do with anomalous samples? Does statistical convergence necessarily translate into geological convergence? Can ‘reliable’ statistical dates be way off?
In a new paper in Geology, six geologists tested a widely-used technique of gathering detrital zircons: crystals of zirconium that often contain uranium (U) and its daughter product lead (Pb), found in debris piles below a slope. They tested the assumptions in their samples collected in Venezuela. Their title hints of troubles ahead: “Use and abuse of detrital zircon U-Pb geochronology—A case from the Río Orinoco delta, eastern Venezuela.” The Abstract warns:
Advancements in mass spectrometry methods over the past two decades have allowed for rapid measurement of (U-Th)/Pb isotopes for geochronologic applications, a tool that has deeply influenced the way sediment provenance and paleo-tectonic reconstructions are approached. Geochronology-based studies of sediment provenance typically rely on dating n ≈ 100–150 single detrital zircon crystals from individual samples, where the sample age distributions are used to make inferences about the parent age distributions, make qualitative geologic interpretations, and/or perform quantitative intersample comparisons. Most efforts to quantitatively compare detrital zircon age spectra make use of non-parametric dissimilarity statistics.
“The fascinating impressiveness of rigorous mathematical analysis, with its atmosphere of precision and elegance, should not blind us to the defects of the premises that condition the whole process.”—Thomas C. Chamberlin
Here, we use laser ablation–inductively coupled plasma–mass spectrometry (LA-ICP-MS) U-Pb detrital zircon moderate-n (n ≈ 100) and large-n (n ≈ 1000) results from unconsolidated fluvial sediments of the Río Orinoco delta, eastern Venezuela, to highlight the concealed pitfalls of making geological interpretations based on quantitative comparisons of U-Pb age distributions alone. Three samples analyzed at large n, selected from contrastingly different mean sediment grain sizes along the active channel of the Río Orinoco, yield large dissimilarities amongst their age spectra, resulting in the misleading conclusion that these were likely not sourced from the same parent distribution. We demonstrate that statistically significant differences amongst detrital zircon samples derived from the same (integrated) source region can be introduced by the dynamics of sediment transport, which may in turn lead to erroneous geologic interpretations arising from the inaccurate assumptions that, at present, condition the quantitative treatment of detrital zircon data.
In short, this means that geologists can falsely interpret ages because of poor assumptions. They begin with the following quote that points out the danger of trusting math without thinking:
“The fascinating impressiveness of rigorous mathematical analysis, with its atmosphere of precision and elegance, should not blind us to the defects of the premises that condition the whole process.”—Thomas C. Chamberlin (Chamberlin, 1899).
How common is the method of dating they describe?
Most sediment provenance studies that apply quantitative detrital zircon U-Pb geochronology rely on the assumption that direct intersample comparisons using dissimilarity metrics allows for determination of similar and/or distinct sediment sources (e.g., Saylor and Sundell, 2016; Vermeesch, 2018, and references therein). There are, however, two inconvenient and commonly ignored issues in geochronology-based sediment provenance: (1) detrital zircon U-Pb ages can cluster by grain size ; and (2) grain-size sorting is strongly influenced by the dynamics of sediment transport.
Most studies, they say, ignore these “inconvenient and commonly ignored issues” with the source data. Factors other than age (i.e., sediment transport) can skew the interpretation. Geologists, in actual practice, do attempt to minimize error. But do they recognize all the possible sources of error? How good is the presumption of innocence?
To determine the provenance of sediments, geologists are often tasked with making geological interpretations based on comparison of samples collected tens to thousands of kilometers apart, sometimes across terrane boundaries and modern or ancient ocean basins (e.g., Rainbird et al., 1992). One presumption that has led workers astray in detrital mineral geochronology is the belief that if a sufficient number of age observations are made in a sample, then the measured distribution of ages will approach the “true” distribution present in a given sedimentary system with sufficient accuracy to make geologically sound interpretations (e.g., Pullen et al., 2014). This premise remains ignorant of how detrital minerals fractionate within sedimentary environments, and also presumes that biases introduced through sample selection and processing are negligible (Sircombe and Stern, 2002; Sláma and Košler, 2012).
Geologists might assume they have a good sample, and conclude that biases are negligible, but be unaware of other sources of bias. The authors seek to point out one source of bias and teach their colleagues that “statistically significant” is not necessarily the same as “geologically significant.”
Here are some warnings about bias and interpretation in the paper:
- …our results demonstrate that physico-mechanical biasing (e.g., size and morphological sorting) can influence the distribution of ages found amongst samples deposited in contrasting energy environments within the same sedimentary system, thus affecting the accuracy of quantitative provenance interpretations built exclusively upon detrital zircon U-Pb dates.
- Although the tests described above—particularly for our large-n data set—can be considered statistically accurate, they are unlikely to result in geologically accurate conclusions….
- ….continuing to ignore statistically significant sources of natural systematic bias will only result in formulation of systematically misleading geological hypotheses.
The authors warn that their findings in Venezuela are going to make some of their colleagues uncomfortable. Simply gathering a lot of data, they say, is not enough to yield trustworthy conclusions:
The experiments shown here pose the uncomfortable problem: If the questions we formulate in such a simple experiment from an active fluvial system fail to be accurately answered using current data evaluation tools, what confidence do we have in our quantitative detrital zircon provenance interpretations from the ancient geological record? It is becoming increasingly clear that
continuing to ignore statistically significant sources of natural systematic bias will only result in formulation of systematically misleading geological hypotheses.
in order to resolve this, continuing to ignore statistically significant sources of natural systematic bias will only result in formulation of systematically misleading geological hypotheses. simply dating more crystals per sample is not the only answer. Although larger values of n are evidently fundamental for improving representativity and statistical accuracy (Pullen et al., 2014; Zhang et al., 2016; Nie et al., 2018), Thus, in order to improve the quantitative use and application of geochronologic-based detrital mineral provenance approaches, we must better our understanding of how zircon crystals pass through, and fractionate from, sedimentary systems (Hietpas et al., 2011), and use that understanding to inform our statistical treatments accordingly. This is particularly critical as the number of publications including a component of detrital zircon provenance continues to grow, because this is inevitably driving the community toward a “big data” approach that will (by necessity) continue to increase its reliance on robust statistical treatments.
They conclude that geologists need to be aware of their assumptions and not just assume that lots of data and statistical analysis will prevent bias. They end with four suggestions for future studies.
Now they tell us! How long have geologists been publishing papers about formations dated with detrital zircons? How long have they been unaware of the natural bias introduced by sediment transport? These six geologists ran a test with a “simple experiment” and found a source of bias that “most studies” have ignored. “Most studies” waltz down the primrose path of assumption, thinking they give the public reliable dates. Those dates, and the geological interpretations of what happened in the unobservable past, could be fraught with error. Will all those old papers be corrected now? Will anyone notice this paper and take corrective action from now on?
What we want to ask, now that this paper has shown a significant “unknown” in geological sample collection and testing, is what other unknowns are still out there? Here it is 2018. Radiometric dating is a century old. What “unknown unknowns” are still out there, to be discovered in future years, that will cast doubt on what geologists have been telling us? Do any of them really know how old their samples are?