Observed

Not modelled. Not estimated. These are the numbers NASA and NOAA measured, with the date each one was taken.

Two kinds of number

Estimated

The counters on the homepage and the daily briefing. Nobody counts births or tonnes of CO₂ as they happen, so those figures come from published annual and daily totals, divided down to the second and accumulated since midnight UTC. They are the right order of magnitude and the right shape. They are not measurements, and a ticking digit is not evidence.

Observed

This page. Every value below is published by the agency that took it. The measurements came off an instrument — a flask at a NOAA sampling station, a satellite microwave sensor, a float diving through the upper ocean, a magnetometer — and the natural-events figure is NASA's own count of what it is tracking. They update on their own schedules, not ours: three daily, three monthly, one yearly, one every minute and one a rolling count of the past week. Each card shows when its reading was taken and links to the dataset so you can check it.

One feed is not responding or has fallen behind its usual schedule. Its card shows the most recent reading we have, with the date it was taken.

Atmospheric CO₂

425.12ppm

The highest reading for this date in 11 years of record

I have not held this much of it since long before you.

201611 years, this date2026

+147.12 ppm (+53%) above pre-industrial
1.87 ppm above the same date last year

The share of the atmosphere that is carbon dioxide, averaged across NOAA's global network of sampling sites. It is the single number that most determines how much heat the planet keeps.

Atmospheric methane

1,939.4ppb

The highest reading for May in 43 years of record

Of everything you have put into my air, this is the one I can still let go of.

198443 years, May2026

+1,210.2 ppb (+166%) above pre-industrial
6.3 ppb above May last year

Methane traps roughly 80 times more heat than CO₂ over its first 20 years in the air, but breaks down within about a decade — which makes cutting it the fastest available brake on warming.

Global temperature anomaly

1.40°C

The highest reading for August in 147 years of record

This is not a season. It is what I average now.

1880147 years, August2026

+1.40 °C above 1951–1980 average
0.22 °C above August last year

How much warmer the month was than NASA's 1951–1980 average, combining land stations with sea surface measurements. Anomalies are used instead of absolute temperature because they survive gaps in the network.

Sea surface temperature anomaly

1.08°C

The highest reading for August in 177 years of record

The water beneath every storm is warmer than it once was.

1850177 years, August2026

+1.08 °C above 1901–2000 average
0.21 °C above August last year

How far the sea surface departed from its 1901–2000 average for the month, from ships, buoys and floats. It is the ocean half of global temperature, which the temperature card blends with land; the sea changes slowly, and it is the water that feeds storms and bleaches reefs.

Ocean heat content (0–2,000 m)

325ZJ

+325 ZJ above 1955–2006 average

What the air sheds, I take in, and I give it back slowly.

Heat held in the top 2,000 metres of the ocean, from Argo floats and ship profiles, as a yearly mean measured against a long-term average. The ocean absorbs most of the heat greenhouse gases trap, so this is where warming accumulates. One zettajoule (ZJ) is 10²¹ joules.

Arctic sea ice extent

5.228M km²

21% below the 1981–2010 average for this date

The ice that belongs to this date is not here.

197948 years, this date2026

17th lowest for this date in 48 years of record
0.309 M km² above the same date last year

The ocean area with at least 15% ice cover, measured by satellite microwave sensors that see through cloud and polar night. It reaches its yearly minimum each September.

Antarctic sea ice extent

17.042M km²

2nd lowest for this date in 48 years of record

The south is thinner than this turn of the year should find it.

197948 years, this date2026

8% below the 1981–2010 average for this date
0.626 M km² below the same date last year

The Antarctic runs opposite to the Arctic — its ice peaks in September, when the north is at its lowest. Watching both at once is the clearest way to see that this is one system with two hemispheres.

Geomagnetic activity

0Kp

storm threshold is 5

The sun leans on my field constantly. Tonight it leans gently.

A 0–9 scale for how much the solar wind is disturbing Earth's magnetic field, updated every minute. At 5 and above it is a geomagnetic storm: aurora further from the poles, and pressure on satellites and power grids.

Natural events being tracked

19events

I have always done this. You are simply watching now.

Wildfires, storms, volcanoes and ice events that NASA has logged a new observation for in the past seven days. Each one is a place where the planet is visibly changing this week.

9 severe storms
5 sea and lake ice
5 wildfires

Tell me when something happens

Not a newsletter. An email only when:

Records — When a NASA or NOAA reading is the most extreme value their record holds for that date — a few times a year.
Geomagnetic storms — When the planetary K-index reaches NOAA's storm threshold, which usually means aurora much further from the poles.

We store your address and nothing else — no name, no tracking. Every email has one-click unsubscribe, and unsubscribing deletes it.

How this page behaves

Dates are the agency's, not ours. “Observed Sep 18” means the measurement is from the 18th, whatever time you happen to load the page, and a monthly or yearly mean is dated by the month or year it covers. A monthly figure weeks old is current, and NOAA's methane mean arrives months after its month ends, because the air samples have to be shipped and analysed first. Ocean heat is a yearly mean that NOAA publishes within weeks of the year closing and revises later, so it is current until the next year's arrives. A daily figure three weeks old would mean something is broken. That is why every card shows its date instead of a single “live” badge.
A feed going down is visible. When an agency server does not answer, the card falls back to the last reading we confirmed and says “last observed” with that date. A card also says “last observed” when a server answers but its newest reading is older than that feed's schedule allows, which is what a file that has quietly stopped updating looks like. It never presents an old number as a new one.
Comparisons are anchors, not trends. Pre-industrial CO₂ and methane, NASA's 1951–1980 temperature baseline and NOAA's 1901–2000 sea surface average are standard published reference values, and each card names who defined its own and links to where they say so. That is not always the agency that took the reading: the pre-industrial values are the IPCC's. Ocean heat carries no anchor: NOAA publishes it against a long-term average whose period we could not confirm from the dataset, so its card compares each year only with the others. Sea ice needs a different anchor, because it swings by millions of square kilometres between March and September: each reading is compared against the average for that same calendar date across 1981–2010, which we compute from the agency's own daily record rather than quote. Either way you are seeing where a reading sits, not where it is heading.
Records are held to the calendar. Where a card ranks a reading or draws its record, it compares it only with the same date in other years, or, for a monthly figure, the same month, or, for a yearly one, the other whole years. A September reading is never ranked against a March one, so “lowest for this date” means exactly that and no more.
Nothing here is behind an API key. Every dataset is public. The links on each card go to the agency page for that dataset, and the numbers are also available as JSON at /api/live.

More live sources are on the roadmap.

Two kinds of number

Estimated

Observed

How this page behaves

Dates are the agency's, not ours. “Observed Sep 18” means the measurement is from the 18th, whatever time you happen to load the page, and a monthly or yearly mean is dated by the month or year it covers. A monthly figure weeks old is current, and NOAA's methane mean arrives months after its month ends, because the air samples have to be shipped and analysed first. Ocean heat is a yearly mean that NOAA publishes within weeks of the year closing and revises later, so it is current until the next year's arrives. A daily figure three weeks old would mean something is broken. That is why every card shows its date instead of a single “live” badge.

A feed going down is visible. When an agency server does not answer, the card falls back to the last reading we confirmed and says “last observed” with that date. A card also says “last observed” when a server answers but its newest reading is older than that feed's schedule allows, which is what a file that has quietly stopped updating looks like. It never presents an old number as a new one.

Comparisons are anchors, not trends. Pre-industrial CO₂ and methane, NASA's 1951–1980 temperature baseline and NOAA's 1901–2000 sea surface average are standard published reference values, and each card names who defined its own and links to where they say so. That is not always the agency that took the reading: the pre-industrial values are the IPCC's. Ocean heat carries no anchor: NOAA publishes it against a long-term average whose period we could not confirm from the dataset, so its card compares each year only with the others. Sea ice needs a different anchor, because it swings by millions of square kilometres between March and September: each reading is compared against the average for that same calendar date across 1981–2010, which we compute from the agency's own daily record rather than quote. Either way you are seeing where a reading sits, not where it is heading.

Records are held to the calendar. Where a card ranks a reading or draws its record, it compares it only with the same date in other years, or, for a monthly figure, the same month, or, for a yearly one, the other whole years. A September reading is never ranked against a March one, so “lowest for this date” means exactly that and no more.

Nothing here is behind an API key. Every dataset is public. The links on each card go to the agency page for that dataset, and the numbers are also available as JSON at /api/live.

More live sources are on the roadmap.

Observed

Two kinds of number

Estimated

Observed

Current readings

Atmospheric CO₂

Atmospheric methane

Global temperature anomaly

Sea surface temperature anomaly

Ocean heat content (0–2,000 m)

Arctic sea ice extent

Antarctic sea ice extent

Geomagnetic activity

Natural events being tracked

Get notified

Tell me when something happens

How this page behaves

Observed

Two kinds of number

Estimated

Observed

Current readings

Atmospheric CO₂

Atmospheric methane

Global temperature anomaly

Sea surface temperature anomaly

Ocean heat content (0–2,000 m)

Arctic sea ice extent

Antarctic sea ice extent

Geomagnetic activity

Natural events being tracked

Get notified

Tell me when something happens

How this page behaves