Monday, June 27, 2011

Why Some Forecasts Don't Work Out and What We Can Do About It!

Today was a generally warm, cloudy day with occasional light sprinkles and showers, quite a contrast with what was predicted only a few days ago. The high today at Sea-Tac was 69F with a trace a rain. More rain fell over the south Sound and the coast.

Lets take a look at the National Weather Service forecasts for Sea Tac and vicinity provided on various recent days around 4 PM:

Friday: MONDAY...MOSTLY SUNNY. HIGHS IN THE 70S
Saturday: MONDAY...MOSTLY SUNNY. HIGHS MAINLY IN THE 70S. LIGHT WIND. Sea-Tac: 79F, 10% chance of rain (they mean measurable rain--at least .01 inch)
Sunday:MONDAY...MOSTLY CLOUDY. HIGHS MAINLY IN THE 70S. LIGHT WIND. Sea-Tac 75F, 10%

and this morning (Monday) at 4 AM the NWS was going for clouds with 75F at Sea Tac and 10% chance of rain.

We ended up with a cloudy day, with some sprinkles over most of the area and light rain in some locations. Upper 60s and low 70s.

So what changed? As I will discuss below, this was a difficult forecast, with the region right on the edge of big rain gradient...and the forecast models had quite a bit of uncertainty.

Here is the 24-h precipitation forecast from the UW prediction system starting at 5 AM this morning (and thus ending 5 AM on Tuesday morning). Not too bad. Shows the heavier showers on the coast and over SW Washington. But this is a very short forecast and we expect it to be excellent.


Now here is the forecast for the same period (24 h ending 5 AM on Tuesday) that was run at 5 AM Saturday. Big difference, the precipitation was well off the coast.

And here is the forecast starting Friday morning: same thing.

Now what changed? The flow field aloft was just subtly shifted between these forecasts..and that made all the difference between sunny and mid to upper 70s and cloudy, lower 70s/upper 60s with sprinkles. Lets take a look.

Here is the forecast from Friday for the upper level (500 mb) flow. See that low right off our coast? Pretty unusual for this time of the year.


And here is the same upper level flow from this morning's forecast (which should be very close to reality). Can you see the difference? REALLY subtle.

This is an example of a very difficult forecast situation in which the outcome is very sensitive to small differences in the forecast flow field.

So how can we deal with such a uncertain forecast? Well, meteorologists now have a new tool--ensemble forecasts: running many forecasts each starting a little differently. Half of the forecast show rain, half not...50% chance of rain! Having a variety of different forecasts...all very reasonable..makes it more likely that the forecast system will capture an upcoming change. Here is an ensemble forecast starting Friday afternoon at 5 PM, showing the 12-h probablility of precipitation during the day on Monday:

This guidance gave a real heads up that there would be a significant chance of showers on Monday. This is the future. And forecasts will be more useful because of it.

6 comments:

Rich M. said...

Cliff,

Thanks for the good explanation of how small changes (on a synoptic scale) can make big changes in the weather we perceive. Rainfall was less than a few hundredths in most places but that is a real difference from sunshine and mid to upper 70s.

I really appreciate the UW mesoscale models. Having forecast for this area for almost 36 years, it is remarkable how much the tools have improved from the old barotropic model (and the leeches in the mason jar) we had when I started! But the close calls where a little change results in a big percieved change are still difficult.

One comment on the ensemble from last Friday that you showed. Although that did argue for a chance of precip, subsequent runs all the way through Monday morning's run were pulling the precip slowly westward away from Puget Sound taking the 10% chance you see in Friday's run down to essentially 0 (holding the precip chance to just off the coast) only 12 hours out. ~ Rich Marriott

Jeff E said...

What is the best way to tell if the 10 day forecast is a "tough call" or if we are in the middle of a long term pattern that will hold up?

Is it more consistent in the summer or is always a difficult forecast situation?

The reason I am asking is trying to plan for an outdoor event on July 9th and need to know. Rent the tent or not?

JewelyaZ said...

@Jeff, I say "rent the tent" though I am not an expert like Cliff.

If it rains, your guests will be dry.

If it's sunny, your guests can escape the sun.

WIN!

Johannes Rexx said...

So weather forecasters can now conveniently preform a sensitivity analysis on the weather models they run, to generate a family of results. If there are large differences amongst these results then they know that the forecast will necessarily be expressed as a probability of rain for example. If the family of simulation results are fairly consistent then a very tight forecast is possible.

So the chaotic weather system may in fact be more chaotic some days and less chaotic on other days, depending on the initial atmospheric conditions. The more chaotic initial conditions results in only short term accurate forecasts while the less chaotic initial state allows for longer term accurate forecasts.

Do I understand this right?

Cliff Mass Weather Blog said...

Johannes...very insightful comments...you are correct...cliff

eprman said...

Cliff,

I don't know that I think saying that the forecast of rain is 50% is an improvement. When I see that the chance of rain is 50% I think that the weather will be generally overcast and cool but may or may not rain in my area. I don't expect sunny weather. In the situation you describe the weather is either going to be cloudy with or without rain, or it will be mostly sunny. How would you communicate to people that there are two different outcomes possible?

The thing I think is missing in forecasts is some indication of the confidence in the forecast. Depending on the weather situation the forecast can either an almost certainty or very chancy. When the weather is in a rapid period of chang the forecasts seem to be the least reliable.