Are forecasts beyond a few days any good? Mine are (laugh). More seriously, though, and perhaps not surprisingly, the answer is, ‘it depends’.

Specifically, it depends on what’s being forecasted. Some weather patterns, such El Nino-Southern Oscillation and the Madden-Julian Oscillation, can be skillfully predicted weeks …if not months… in advance. Yet many others behave unexpectedly after only a matter of days (or less).

What? Say that again?

How is it possible that outlooks for weekly-averaged temperatures can be relatively accurate a month in advance, while most of us probably don’t have much confidence in forecasts for the timing and duration of rain showers tomorrow?

The answer stems from the fact that larger and more-slowly evolving weather systems, such as the ones associated with our heat waves and cold spells, are more predictable at longer lead times than are the day to day variations that control our rain showers. These systems last longer and trigger relatively gradual meteorological changes that can be more easily seen in the data ahead of time.

Longer-range prediction systems, to the extent the atmosphere’s dynamics allow it, unwaveringly stare at those slowly evolving (low frequency) processes with relatively little regard for the rapidly developing (higher frequency) stuff, like our rain showers. They intentionally try to ignore systems with day-to-day timescales, because skillfully predicting those many days in advance is basically impossible.

Skill scores (anomaly correlation coefficients) from a collection of weather models as a function of lead time. The two distinct groups of curves refer to different classes of weather phenomena. (NOAA/EMC)

You can think of the predictions shown as very general forecasts for the jet stream over the entire Northern Hemisphere. The scale on the left is one measure of how good the forecasts are (higher number is better). Different lines refer to different weather models. And the scale on the bottom refers to how many days in advance the forecasts are targeting.

With these models (and these are the global weather models like GFS, ECMWF, etc.), forecasts for smaller swirlies in the jet stream, encircled in red, are less accurately predicted at every given lead time than the bigger ones (encircled in green); i.e., their curves are everywhere lower and crash to the bottom more quickly. In fact they generally fall below 0.5 after about 5 days. For reference, under realistic circumstances, a score of just 0.5 will be achieved by using the climate average for the day. Thus, a forecast with this score (or worse) is useless.

Related: AccuWeather debuts 25 day weather forecasts

The point to take away from all of this is … the farther out in time you want your forecast, the less detail you should expect. In other words, a detailed snowfall accumulation prediction 5 days in advance is probably not reliable information, yet a forecast for colder and snowier than normal conditions in two weeks might very well be.

Related: Weather forecast accuracy versus skill: skill is what matters

So with that view in mind, let’s look at what big weather systems are coming down the road in the next 1-2 weeks, shall we?

As mentioned in last week’s post, 90°F heat will sweep all across Kansas tomorrow just as the northeast and Mid-Atlantic recovers from a blast of air with near winter-like magnitude. But those warmer signals won’t last long. And it now appears that a sustained warm-up for locations east of the Rockies might have to wait until early May.

High-altitude temperatures and winds expected late Saturday, from the GFS ensemble mean. Red shading indicates anomalous ridging/warm-air aloft. Black arrow identifies the jet stream. (Penn State)

Though some of the Kansas heat will very briefly expand to the Carolinas on Saturday, the setup shown above will allow another round of very chilly air to dive southward into the Great Lakes late this week and eventually into the East early next week (black arrow). Temperatures are expected to once again dip well-below normal for many locations east of the Rockies for several days. In fact, there is a growing model consensus for some snowflakes over the weekend in the Northern Plains.

Hmmm … can anyone say payback for last month?