What Is the Difference Between Weather and Climate?

Weather changes by the hour and is driven by the movement of air masses, fronts and pressure systems across the surface. It is chaotic in the technical sense: a small difference in conditions today grows into a large difference in a week, which is why forecasts are good for a few days and useless beyond about two weeks. Nobody can say whether it will rain in London on a given day next August.

Climate is the statistics of weather: the average, the range and the frequency of extremes at a place, conventionally calculated over thirty years. It is driven by things that change slowly, the amount of sunlight a latitude receives, the distance from the sea, the height above it, the ocean currents offshore and the composition of the atmosphere. Nobody can say what next 15 August will bring in London, but anyone can say it will be warmer than 15 February, and that is a statement about climate.

A warming climate is a shift in the whole distribution of weather, not a ban on cold days. If the average temperature rises by one degree, cold days still happen; they become somewhat rarer and hot days become somewhat more common, and the record highs start to outnumber the record lows. In the United States, record daily highs now beat record lows by about two to one, where in a stable climate the ratio would be even. A single cold week is a sample of one from a distribution that has moved, and it cannot show the movement.

The same logic runs the other way. A heatwave is not proof of climate change either, though scientists can now calculate, using models run with and without the extra greenhouse gases, how much more likely a particular heatwave was made by warming. The 2021 heat in the Pacific Northwest of North America, they concluded, would have been virtually impossible without it. Attribution of single events is a statement about probability, which is a statement about climate.

The thermometer record from weather stations goes back to the mid-nineteenth century, with ships' logs filling in the oceans, and it is checked by satellites since 1979 and by measurements of ocean heat, sea level, glaciers and the dates of spring. Before instruments, climate is read from proxies: tree rings, whose width records each summer's growth; ice cores, whose layers trap air bubbles and record temperature in the ratio of oxygen isotopes; ocean sediments; and corals. Together they reconstruct the last several hundred thousand years and show that the current rate of warming is faster than anything in that record. The differences in one table:

• •Weather: hours to days; climate: decades to millennia
• •Weather: a measurement; climate: a statistic of many measurements
• •Weather: predictable for about a week; climate: predictable for decades, because it depends on slow causes
• •Weather: what to wear today; climate: what to build, plant and insure for

The two questions need different tools and different experts. Meteorologists forecast weather by tracking the atmosphere's current state forward; climatologists project climate by working out how the energy balance of the planet responds to changes in the sun, the oceans and the atmosphere, and the second kind of prediction is in some ways easier, in the way that predicting a hot summer in the Sahara is easier than predicting rain next Tuesday. It also matters for decisions. A farmer plants for the climate and harvests in the weather; a city builds its drains for the climate it expects and floods in the weather it gets; and the gap between the two is where the damage of a changing climate arrives, in weather that the infrastructure was never designed for.

Between the weather forecast and the climate projection sits a middle kind of prediction, the seasonal outlook. Weather services issue forecasts saying whether the coming three months are likely to be warmer, colder, wetter or drier than usual. They cannot tell you about any particular day, but they can shift the odds, because some parts of the system change slowly enough to carry a signal forward. Sea surface temperatures, soil moisture and snow cover all hold a memory for weeks or months, and the state of the tropical Pacific, El Niño or La Niña, is the most useful of them.

These outlooks are expressed as probabilities, such as a sixty percent chance of a warmer than normal winter, and their skill varies by region. They work best in the tropics and in places strongly tied to Pacific conditions, and less well in western Europe, where the weather is dominated by the changeable Atlantic. The word normal is itself a climate statistic that moves. Weather services compare today's weather with thirty-year averages that are updated every ten years, so the current reference period, 1991 to 2020, is already warmer than the one it replaced.

The reason forecasts fail after a couple of weeks was found almost by accident. In 1961 the meteorologist Edward Lorenz, at the Massachusetts Institute of Technology, was running a simple computer model of the atmosphere. Wanting to look at one run again, he restarted it partway through by typing in numbers from a printout, which showed values to three decimal places rather than the six the computer stored. The new run tracked the old one at first, then drifted, and before long bore no resemblance to it. A difference of less than one part in a thousand had grown into completely different weather.

Lorenz published the finding in 1963, and a later talk of his asked whether the flap of a butterfly's wings in Brazil could set off a tornado in Texas, giving the idea its popular name. Forecasters now use it deliberately. Instead of running one forecast, they run dozens from slightly different starting conditions, a set called an ensemble. When the runs agree, confidence is high; when they spread apart, the forecast says so. The same reasoning explains why climate projections remain possible: you cannot follow a single path through a chaotic system, but you can describe the range of paths it tends to take.

When a forecaster says a month was warmer than average, the average usually refers to a climate normal, an official set of figures calculated over thirty years. The World Meteorological Organization recommends the period 1991 to 2020 as the current standard, replacing the earlier 1961 to 1990 period that many reports still use for long-term comparisons.

Using a fixed thirty-year window smooths out the ups and downs of individual years, which is exactly the difference between weather and climate. It also means that as the climate warms, the averages themselves shift, so a temperature described as normal today would have counted as warm against the older baseline. Checking which normal a report uses is a simple way to make sense of claims about records and averages.

Weather is the atmosphere's condition at a place and moment, chaotic and forecastable only for days; climate is the long-run pattern of that weather, driven by slow causes and predictable for decades. A warming climate shifts the whole distribution of weather rather than abolishing cold days, so no single storm or heatwave proves or disproves it, and the difference is the difference between what to wear and what to build.