Star catalogues hand out ages to several decimal places, and the numbers look as solid as a temperature reading. They are not. For decades the models behind them simply assumed how much helium a star holds, because nobody knew how to measure it. The EU-funded CartographY project set out to stop assuming, and in the process it showed how far precision can drift from accuracy — the difference between treating astrophysics as a method and treating it as a view.
Ask an astronomer how old a star is and the answer arrives with decimal places. It reads like a measurement, in the same way a thermometer reading does. In practice it is the output of a model, and a model is only as good as what was fed into it. Change one assumption at the input and the age at the output shifts by a stretch of time longer than the whole of human history. The figure does not look any less confident for it.
The awkward assumption sits near the very beginning. Helium is the second most abundant element in the Universe, and inside a star it is no footnote: it shapes the structure, the temperature, the brightness and the length of the entire life. Yet traditional stellar models did not measure it. They assumed it, borrowing a value or a relationship and applying it across the board. The measurement was missing because nobody knew how to take it.
This is where the two words part company. Precision describes how tightly a method repeats itself. Accuracy describes whether it lands on the truth. A pipeline built on a mistaken assumption can be exquisitely precise and consistently wrong, and it will quote a small error bar while it is at it.
The CartographY project, funded by the European Research Council and coordinated by Guy Davies at the University of Birmingham, was built around exactly that gap. Davies sums up the outcome as the ability to infer stellar ages "in a more honest and self-consistent way".
The way in is sound. Stars ring: waves travel through the interior and surface as tiny, regular changes in brightness. Reading those oscillations is asteroseismology, and it reaches places no telescope image can. "A main project activity was to map and measure internal stellar helium using asteroseismology," Davies says, describing it as the study of stellar oscillations. The seismic data did not work alone. It was combined with spectroscopy, which reads the chemistry of the surface, and with astrometry, which pins down positions and distances.
Comparing a star against theory means generating grids of stellar evolution models, and those grids grow enormous fast. Computing every point the slow way would take longer than the science can wait. So the team trained machine-learning emulators to stand in for the heavy calculation: the emulator learns what the physics produces and then returns an answer in a fraction of the time. Nothing about the physics is loosened. The arithmetic simply stops being the bottleneck.
The statistical layer is Bayesian hierarchical modelling, which allows whole populations of stars to be studied at once instead of one object at a time. Patterns invisible in a single star, such as how helium levels track other elements across different groups, come into view across a sample.
It also rewrites the rules for choosing between rival models. "Rather than picking the model that delivers the smallest formal uncertainty, it identifies which models are actually best supported by the observational data," Davies says. The smallest error bar stops being the prize.
Stars slow down as they age, which makes rotation look like a clock. The project found that current models of how rotation evolves are less trustworthy than the field had assumed, and a shaky model makes a shaky ruler. The response was not to abandon the idea but to test it: the same Bayesian framework is being turned on rotation theories to check them against data before anyone leans on them for an age.
An exoplanet is described through its host star, so an error in the star is inherited by the planet. The PLATO mission of the European Space Agency will deliver data on the parent stars of exoplanets, which puts a premium on stellar profiles that are accurate and not merely tidy. The consequences are concrete:
CartographY did not make stellar ages certain. It made them honest, by measuring a parameter that models used to assume and by carrying the remaining doubt through the mathematics rather than hiding it. That is a smaller claim than a confident headline number, and a far more useful one. For anyone weighing up astronomy as a degree, the lesson is about the workshop rather than the sky: much of the job is deciding how much a figure is actually worth.
5 questions · one minute · nothing saved
EN 
CS
DE
ES
FR
HU
IT
PL
PT
РУ
SK
TR
УК
AR
中文
