How can we predict the climate for the next 50 years if we can’t predict the weather next week?

This was asked by a commenter, Skeptic517, on my blog yesterday.

that’s a good question, one that I suspect most climate activists cannot answer.

Sure enough, none of the respondents knew the answer.

So what do you think? Is long-term climate prediction hokum?

And now the answer.

Climate prediction is possible and reasonable because the nature of forecasting is very different from weather forecasting.

If you like technical terms, the weather forecast is an initial value problemwhile climate prediction is mostly a boundary value problem. I will explain this in a moment.

And the nature of the prediction is very different.

For weather prediction, we predict the exact state of the atmosphere at a given time. The peak in Spokane will be 68F next Thursday. The low center would be 978 hPa and located in central Iowa. Specific in time and space.

84 hour weather forecast for sea level pressure (solid line) and precipitation (shadow)

For climate projections, we are not forecasting surface weather maps for 4:00 a.m. on January 4, 2090. That would make no sense.

Instead, we forecast average conditions, often over a wider area. Will average spring temperatures in Washington State be warmer or cooler than current values? Will the 10-year winter average precipitation at the end of this century be greater than the recent 10-year average?

Forecast of annual mean surface air temperature change from late 20th century to mid-21st century

That’s a way of thinking.

The essential insight you need is to understand the different nature of weather versus climate prediction.

Weather forecasting: a matter of initial value

Weather prediction, a matter of initial value, begins with a comprehensive 3-D description of the atmosphere called initialization. Then, large supercomputers are used to solve equations that describe atmospheric physics in order to accurately predict the future state of the atmosphere at specific times.

Forecast accuracy diminishes over time, and in about two weeks, almost all predictability is lost….which MIT Professor Edward Lorenz has theoretically described.

Here is a graph of the loss of forecasting skill over time for US models (about 18,000 ft in the northern hemisphere). The blue color is for the top US global model, GFS. The kill forecast drops rapidly between 5 and 10 days.

Climate forecasting: a boundary value problem

Forecasting skills for specific weather features are lost after about 2 weeks because the atmosphere essentially loses its memory of the atmosphere’s original observed state.

In long-term climate forecasting, the main constraint is not the initial conditions, but the amount of radiation entering and leaving the atmosphere. If we know the amount of radiation entering and leaving the top of the atmosphere, climate models can generate realistic average climates for those conditions.

The amount of radiation emitted and absorbed by the atmosphere is very much controlled by the composition of the atmosphere…. which we have to assume (e.g. how much CO2, methane and particles are in the atmosphere).

Such predictions are only as good as our estimates of the amount of greenhouse gases in the atmosphere in 50 or 100 years. Great uncertainty! But we do the best we can.

So climate prediction does NOT make sense.

I’m leaving out some subtleties: e.g. the initial state might have some effect on the climate simulation. But you can try to work around that by running a set of many climate predictions, each starting out slightly differently.

Anyway, now you know the answer!