Explaining uncertainty in a broad subject like geology is usually done in very different terms for each input, or type of observation. If you want to be systematic you can often consider three categories – or dimensions – of reliability; which are precision, accuracy and trueness. These are slightly artificial concepts, and the first two often have some degree of over-lap. 

Precision

We can sometimes be precise about precision. There is precision as in the 2 sigma (standard deviation) peak derived from a mass spectrometer, or a GPS site location on a geological map. Often there is additional fuzziness to such precision; a strontium isotopic ratio has a 2 sigma measurement peak, but the sea-water reference curve has addition imprecision that is often overlooked. Some samples contain genuine vitrinite particles and their reflectivity can be measured as a histogram, with a mean and a standard deviation error bar. In contrast, other samples, such as those from paralic facies, might contain a wider range of vitrinite macerals, including perhydrous vitrinite, with slightly suppressed reflectivity. This can be precisely measured but its meaning is significantly different.

Accuracy

 I generally consider accuracy as being the close clustering of repeat observations. If a gun-man shoots at a wall with a rifle the bullet holes are small, and a tight cluster of repeat shots/holes indicates high accuracy. In contrast, low precision is like the broad blast of a shotgun, but even these blasts can be accurately clustered in a small area by a marksman. A lab can measure vitrinite reflectance (Vr) or strontium (Sr) isotopes over a few hundred meters of section and get similar repeat values and, as a result, be happy their technique is considered accurate.

Trueness; the big picture

In both the examples of Vr thermal maturity and Sr isotopic dating the first test of data is, do values increase downhole? Both should do so. What if they do not? 

In ODP site 768 in the Sulu Sea the following K/Ar radiometric data from Bellon and Rangin (1991) is from below 320m in core 769C, in the section of volcanic lapilli tuffs.

Sample	Calc age	40Ar R 10-7 cm3/g	36Ar 10-9 cm3/g	%40Ar R	K2O (%)
7R-1, 118-120 cm	20.29 ± 0.81	7.70	9.45	42.9	1.17
	20.83 ± 0.72	7.90	2.84	48.3	1.17
9R-3, 110-113 cm	19.48 ± 0.52	10.54	2.47	59	1.67
	20.14 ± 0.48	10.90	1.89	66.1	1.67
9R-4, 49-50 cm	14.25 ± 0.79	8.02	5.43	33.2	1.74
	15.07 ± 0.93	8.49	6.51	30.5	1.74

There are reasonable amounts of K₂O for K/Ar dating, good precision error bars, and each sample analysed twice seems to have accuracy, but the youngest ages are in the deepest sample. In a stratified tuff this is an indication that there are problems with trueness.

I have published examples of such problems (see example of Sr dating in this paper, copy of Figure 4 shown below), and to their credit Vahrenkamp (1996, 1998; on Sarawak limestones) and Zeiza et al. (2012; on Java limestone) reported times where there was no increase in Sr ages downhole over hundreds of meters, and considerable variation in each Sr measurement / age.  The general age is about right, compared to biostratigraphic control, but there are important questions about what the data means, especially if an “average” age is given, with significant digits to one decimal place.

Care in reporting is required

One can cast doubt over many proxy observations in geology. There is even accidental or deliberate blurring of inconvenient data that does not fit, a process which I call “data munging” (a computing term, with the backronym mash until no good). This is over-emphasising some imprecision, or downgrading work, for instance as “old fashioned”, like an ad hominem attack in debates. In the Figure above the original biostratigraphy was derided as “old-school” and less sophisticated than the modern technique of Sr dating, which appeared to be both linear and of reported “high precision” (the small measurement error bars).

On the other hand one could try and tabulate every aspect of data quality until a report is beyond the length allowed in publications, and is tedious and pedantic to read. What is needed in good reporting is to prioritise critical aspects that might impact a deduction, and the qualities of this data should be reported to the reader. For instance, the age diagnostic planktonic foraminifera Orbulina is very hard to positively identify in thin section, unless luck presents a near perfect axial section. The genus Sphaeroidinellopsis has a much less precise evolution datum – it is rarely mentioned in zonal schemes for this reason. It gradually increases in abundance over several million years around the same time as the famous Orbulina datum (base Middle Miocene). Sphaeroidinellopsis however has an extremely distinct wall structure and can even be recognised in a random thin section from a single example. So, if the Middle Miocene or younger age of a hard planktonic marl is an important input to the hypothesis, then it is worthwhile to photograph your candidate Orbulina marker to put minds at rest that this is true**. On the other hand, if you have specimens of Sphaeroidinellopsis a comment such as “… with the distinctive cortex layer visible” will probably suffice (although a photo is nice). Simply stating “on the presence of planktonic foraminifera of Middle Miocene or younger age” is insufficient if the sample is known to be a hard marl.

Modern geological reports have lost this quality of perspective on reliability. Hard science insist on error bars as part of the observation process. Geology has gone soft and thinks all data is equal. It is not.

This is probably the first of many such posts on this subject, as new examples come to light. For now just realise proxy data in geology has several dimensions of reliability. We need to constantly try and improve our documentation of these dimensions, in order to increase our observational resolution and reliability. Changing the quality of age, facies or other data can have major knock-on effects on the ranking of alternative hypotheses. There are always alternative hypotheses in a data poor, non-experimental science.

** Sadly, twenty years ago such caution was not usually necessary as most micropalaeontologists were better trained and more careful. In the past few decades I have seen so much cut-price, rushed work by inexperienced people that now I really like to see statements of care and caution, to re-assure me that the work is robust. This is the basic concept in recognising data is never plain or simple; Miocene rocks are not colour-coded Yellow and Pliocene rocks are not Green. Argue a case. Convince, as neatly as possible that you have been aware of the pitfalls and have taken steps to avoid them. If in doubt, add a photo or some other cross-check.

References

Bellon, H.R., Rangin, C., 1991. Geochemistry and isotopic dating of the Cenozoic volcanic arc sequences around the Celebes and Sulu seas. In E. A. Silver, C. Rangin, & M. T. von Breymann (Eds.), Proceedings of the Ocean Drilling Program Scientific Results, Vol 124, p. 321-338.

Vahrenkamp, V.C., 1996. Growth and Demise of the Miocene Central Luconia Carbonate Province: Implications for Regional Geology and Reservoir Production Behaviour. Shell unpublished report

Vahrenkamp, V.C., 1998. Miocene carbonates of the Luconia province, offshore Sarawak: implications for regional geology and reservoir properties from strontium-isotope stratigraphy. Geological Society Malaysia Bulletin 42, 1-13

Zeiza, A., Simaeys, V., S., Musgrove, F., Sekti, R., Hakiki, F., 2012. The impact of differential subsidence rates in shallow water carbonate reservoir quality: an example from the East Java basin, Indonesia. Proceedings from Proceedings Indonesian Petroleum Association Convention: 36, 1–13