How Scientists Measure Historical Temperatures Using Ice Cores

There's a number that has been sitting in climate datasets for decades: 280 parts per million. That's the CO₂ concentration in the atmosphere before we started burning coal and oil at industrial scale. Today we're past 420. If you want to know how scientists are so confident about that baseline, the answer is buried in ice. Literally.

Ice cores are the closest thing we have to a time machine. They don't just tell us what the atmosphere looked like thousands of years ago. They tell us how historical temperatures moved, decade by decade, ice layer by ice layer. And once you understand how that measurement actually works, you can't unsee it. The whole picture of past climate stops being a black box and becomes a physical process you can follow from snowflake to dataset.

I got interested in this after a conversation with a glaciologist who showed me a core segment the length of my forearm. She pointed at a faint grey band and said, "That's a volcanic eruption from 1783. The one that caused crop failures across Europe." The band was maybe two millimetres thick. I've been reading about ice core science ever since, and I want to walk you through how the temperature part actually works, because most explanations skip the interesting bit.

Why ice cores are the best thermometer we have for the deep past

Thermometers were invented in the 17th century. Reliable, standardized weather stations only became common in the 19th. So how do you know what the temperature was 50,000 years ago? You don't ask a thermometer. You ask something that recorded temperature indirectly, and you learn to read it.

These indirect records are called proxies. Tree rings are proxies. Coral skeletons are proxies. Ocean sediment is a proxy. Ice cores are a proxy, and they happen to be one of the best, for three reasons.

What makes ice cores different from other records

First, ice accumulates in layers. Every year, snow falls, compacts, and leaves a distinguishable band. That gives you a built-in calendar.

Second, ice traps air. As snow compacts into ice, bubbles of atmosphere get sealed inside. Those bubbles contain actual samples of ancient air, with the CO₂ and methane concentrations of the time. No other archive gives you that.

Third, the water molecules themselves carry a temperature signal. That's the part I want to spend real time on, because it's the piece most articles wave away with a sentence like "scientists use isotopes."

You can pull ice cores from Antarctica going back 800,000 years, and the oldest continuous record extends even further in some projects. Greenland cores go back over 100,000 years. That's not a small window onto the past. That's the whole span of multiple ice ages.

How scientists actually measure temperature from ice

Here's the mechanism, and it's genuinely elegant once it clicks.

Water is H₂O, but not all water is identical. A small fraction of water molecules contain a heavier oxygen atom, oxygen-18, instead of the common oxygen-16. Both exist in the atmosphere. The question is which one ends up in the snow that falls on a glacier.

Oxygen isotopes as a thermometer

The answer depends on temperature. When water evaporates from the ocean, it preferentially carries the lighter isotope. As that vapor moves toward the poles and cools, it condenses and falls as snow. The colder it gets, the more the heavier isotope gets left behind in the rain or snow that falls earlier in the journey, and the lighter isotope keeps traveling.

So the snow that lands on a Greenland glacier in a cold year has a lower ratio of ¹⁸O to ¹⁶O than snow that lands in a warmer year. Scientists measure that ratio, express it as δ¹⁸O (delta-eighteen-O), and read it as a temperature signal.

Cold year = more negative δ¹⁸O. Warm year = less negative. That's the whole core of the method.

Why it's not a perfect thermometer

I'll be honest, this is where it gets messier, and I think that's worth saying out loud. δ¹⁸O doesn't track temperature alone. It also depends on:

• How far the water traveled before it fell (distance from the ocean source matters)
• Altitude of the deposition site
• Seasonal patterns of snowfall (winter snow versus summer snow carries different signals)
• Changes in the source region's own isotope composition over long timescales

So a δ¹⁸O curve is not a direct temperature curve. It's a temperature curve with noise and confounding factors baked in. Glaciologists calibrate it against modern instrumental records, and they cross-check it against other proxies. A friend of mine who works on Antarctic cores once described the process to me as "reading a thermometer through frosted glass." You can do it. You just need to be careful.

That said, the broad signal is robust. When δ¹⁸O says the last glacial maximum was much colder than today, tree rings, ocean sediments, and pollen records all say the same thing. Independent proxies agreeing is how you know you're not fooling yourself.

How do scientists know how old ice cores are?

This is the question people ask me most often, and it's the right one to ask. A temperature curve is useless if you don't know what year each layer corresponds to.

The first tool is annual layer counting. In regions with strong seasonal contrast, each year of snowfall leaves a visible pattern: summer snow is slightly different from winter snow, sometimes in dust content, sometimes in chemistry, sometimes in density. You count the layers like tree rings. For the top few thousand years in Greenland, this works remarkably well.

Volcanic markers and radioactive dating

Volcanic eruptions leave ash and sulfate deposits across huge areas. A major eruption shows up as a distinct chemical spike in ice cores from different locations at exactly the same depth. If you know when that eruption happened from historical records, you have a fixed point in the chronology. The 1783 Laki eruption in Iceland is one example. So is the 1815 Tambora eruption, which caused the "year without a summer."

For older ice, scientists use radioactive isotopes as additional anchors. Beryllium-10 and chlorine-36 are produced by cosmic rays in the atmosphere and deposited in ice. Their concentrations vary with solar activity, and certain spikes have been linked to known events. Carbon-14 dating of trapped CO₂ can also help, though it has limits.

For the deepest, oldest ice, layer counting eventually stops working. The layers get compressed so thin that they're no longer distinguishable. At that point, scientists switch to ice flow models, which simulate how the ice has deformed and thinned over time. These models are calibrated against the layers you can still count, then extrapolated downward.

The result is a timescale with an uncertainty range at each depth. Not a single year, but a window. For recent ice, the uncertainty might be a few years. For ice 800,000 years old, it can be several thousand years. That's the honest version, and it's better than the confident-sounding version that pretends precision where there isn't any.

What do ice cores tell us about CO₂ and temperature?

This is where ice cores earn their reputation. The trapped air bubbles give you direct measurements of ancient atmospheric CO₂. The isotopes give you a temperature estimate. Line them up and you get a relationship that has held across multiple ice ages.

Period	CO₂ (ppm, approximate)	Global temperature relative to pre-industrial
Last glacial maximum (~20,000 years ago)	180–190	About 4–6°C colder
Mid-Holocene (~6,000 years ago)	~260–270	Roughly similar to pre-industrial, slightly warmer in some regions
Pre-industrial (~1750)	~280	Baseline
Today (2026)	~425	About 1.3–1.5°C warmer

The CO₂ numbers for older periods are direct measurements from bubbles. The temperature estimates are derived from isotopes and other proxies. The two curves track each other closely across the ice age cycles. When CO₂ goes up, temperature goes up. When it goes down, temperature follows.

There's a nuance here that gets twisted in online arguments. In the ice core record, CO₂ sometimes appears to rise slightly after temperature at the start of a deglaciation. That's because orbital changes trigger the initial warming, which then releases CO₂ from the ocean, which amplifies the warming. CO₂ is a feedback in that context, not the initial trigger. But the amplification is real, and it's a large part of why ice ages end as dramatically as they do.

Today, we're the ones adding CO₂ directly. The mechanism is the same. The difference is the source and the speed.

Was Greenland warmer 1000 years ago?

Yes, and this is one of those facts that gets used in ways it shouldn't.

Around 1000 CE, parts of Greenland were warm enough that Norse settlers established farms and grazing land. Ice core records from Greenland do show a warm period around that time, often called the Medieval Warm Period. So the claim is factually grounded. The problem is what people do with it.

The Medieval Warm Period was not globally synchronous. It was a regional phenomenon, strongest in the North Atlantic and parts of Europe. Other regions were cooler during the same centuries. When you average temperature estimates from around the world, the medieval period does not come out warmer than today. It comes out cooler.

You can find this in the ice core data itself if you look at multiple cores from different locations. Greenland's warmth at 1000 CE doesn't show up the same way in Antarctic cores from the same centuries. That's the giveaway. A regional warm patch is not the same as a global warm period, and conflating the two is a category error.

What do 97% of scientists believe is causing climate change?

The overwhelming majority of climate scientists attribute recent warming primarily to human greenhouse gas emissions, especially CO₂ from burning fossil fuels. The 97% figure comes from surveys and analyses of published climate literature, and multiple independent studies using different methods have arrived at similar numbers.

Ice cores support this conclusion directly. They show that atmospheric CO₂ is now higher than at any point in at least 800,000 years, and the rise has been extraordinarily fast compared to natural variations. The isotope signature of that added CO₂ points to fossil fuel combustion, not volcanoes or ocean outgassing. When you combine the ice core record with modern atmospheric measurements, the human fingerprint is hard to miss.

How do scientists analyse ice cores?

Pulling a core out of a glacier is only the beginning. The analysis is where the science actually happens, and it involves a surprising amount of chemistry and logistics.

From drill site to lab

Cores are cut into sections, packed in insulated boxes, and shipped to freezers at research facilities. They have to stay frozen the entire time. A single melt event can ruin a section. I've heard stories of cores arriving with cracked sections because of a handling mistake, and the data loss is permanent.

In the lab, the core is processed in a cold room, often at -20°C or colder. Scientists slice it lengthwise into slabs. One slab goes for isotope analysis. Another goes for chemistry. Another is archived for future study, because analytical techniques keep improving and you don't want to destroy the only sample.

What actually gets measured

• δ¹⁸O and δD for temperature reconstruction
• CO₂ and methane from trapped air bubbles, released by crushing or melting the ice in a vacuum
• Dust concentration, which tells you about wind patterns and aridity in source regions
• Major ions like sulfate, chloride, and nitrate, which reflect volcanic activity, sea salt, and pollution
• Trace elements that can pin down specific events, like lead from Roman mining showing up in Greenland ice

Each measurement adds a layer to the picture. The temperature signal is one thread. The atmospheric composition is another. The dust and chemistry give context. Together they let scientists reconstruct not just how warm it was, but what the atmosphere was doing, how winds moved, and what was happening on the surface of the planet at the same time.

Why this matters more than most people realize

Ice cores are not just a curiosity. They're the reason we can say with confidence that current CO₂ levels are unprecedented in human history. They're the reason we know that temperature and CO₂ have moved together across ice age cycles. They're the reason the climate conversation has a baseline at all.

Every time someone says "the climate has always changed," the ice core record is the answer. Yes, it has. And now we know how, and why, and how fast. The ice told us.

The next time you see a graph of CO₂ over the last 800,000 years, look at the far right edge. That vertical spike at the end, the one that shoots up past everything else in the record by a wide margin? That's us. The ice cores don't editorialize. They just record what's in the air. And what's in the air right now is unlike anything the planet has seen in nearly a million years.