Since reliable modern records of climate only began in the 1880s, proxies provide a means for scientists to determine climatic patterns before record-keeping began, though it appears that the the distribution of proxy records for global average temperatures, just like the instrumental record, is strongly non-uniform, with more records in the northern hemisphere.

How accurate are climate proxies in giving us a clear picture of global average temperatures throughout Earth history?

  • $\begingroup$ What's the source for the distribution of proxy records being non-uniform? That would surprise me... $\endgroup$ Commented May 6, 2014 at 17:37
  • 1
    $\begingroup$ @EnergyNumbers Historical records since 1880. Meaning, compare the proxies to the records as a baseline. $\endgroup$
    – Richard
    Commented May 6, 2014 at 18:17
  • 1
    $\begingroup$ @Richard aren't those the data that the proxies are calibrated with? $\endgroup$
    – 410 gone
    Commented May 6, 2014 at 18:25
  • 2
    $\begingroup$ But climate's been really stable over the past 120 years - compared with the timescale we're looking at proxies over - so cross-calibration is of pretty limited use... $\endgroup$
    – kaberett
    Commented May 6, 2014 at 18:25
  • 2
    $\begingroup$ The limits of my understanding of this area are "crap, forams are probably not as good a record as we've been hoping for the past thirty years", but that doesn't count as an answer ;) $\endgroup$
    – kaberett
    Commented May 6, 2014 at 19:36

2 Answers 2


Overall, any individual proxy is not a reliable indicator. It's the combination of multiple proxies that provides a clearer picture.

For example, tree ring growth has been correlated with temperatures. However, it's not without it's anomalies:

...from the middle of the 20th century tree ring growth was less than might have been expected from the temperature record... ([source]


If we were to only look at tree-rings, we could not possibly trust on them as a reliable source, since they have known problems.

We can also use coral as a indicator of past climate. However, it's also somewhat questionable:

However, long (multi-centure) records are rare, and the possible influence of non-climatic influences has not yet been confidently established. source

Ice-cores are not immune to skepticism:

An indisputable interpretation of ice-core oxygen isotopes in terms of atmospheric temperature variability, moreover, remains elusive, and precise annual dating can be difficult.

However, if we can combine these three indicates along with other indicators (pollen, earth bores, etc), we can increase the reliability of our understanding. For example, if most indicators show a cold year, we can more reliability say it was a cold year.

Combining these proxies together helps us to gain a clearer picture of the past environment.

Overall Reliability

However, with that said, the amount of uncertainty with the multi-proxy method is not small.

At the Workshop on Mathematics in the Geosciences, Blake McShane presented a new algorithm for estimating paleoclimate. This algorithm provided a more reliable estimate by increasing the uncertainty of the predictions. That's an important point that I will restate:

Through his talk, he showed that his method was more reliable because it was less certain.

Even that wide uncertainty level, he questioned:

Indeed, this should make us increase our level of uncertainty (indeed perhaps our wide intervals are in fact optimistically too narrow)!


In the end, we can look at proxies for guidelines and trends, but they cannot be used for reliable readings.


The reality is that we don’t know how reliable proxies are. All of them have shown deficiencies. Corals have shown they don’t fix 18O like a time capsule, ice cores show that they’re not static (things migrate through the ice) and tree rings show a cooling trend in the late 20th century. Aggregating a bunch of proxies on the assumption that these errors cancel is unproven and a huge leap of faith. Anyone who claims proxies can detect global temperature to anything near what a thermometer can detect is either crazy or lying to you. I wish it weren’t so, but that’s reality. There are too many unknown errors and since we don’t actually have a record of temperature prior to 1850, we have nothing to compare it against.

  • 1
    $\begingroup$ References\evidence for some of your claims might help you to convince :-) $\endgroup$ Commented Aug 12, 2021 at 6:38

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.