February 14, 2012

The Longest Forecast

What are the longest regular numerical weather forecasts made by the National Weather Service and what do they show for this spring?

 First, a definition---a numerical weather forecast begins with a description of the atmosphere at one time and then using a computer model of the atmosphere to predict the exact state of the atmosphere at future times.  Like the temperature in Cleveland at 2 PM three days from now.  Or the exact position and pressure of a low pressure center in four days.   This compares with predicting climate--by which we the temporal mean or average state of the atmosphere over some location or region.

At the UW we run the WRF model out to 180 h (7.5 days) in weather forecasting mode.
The National Weather Service runs the GFS global model (which have shown on this blog several times) out to 16 days.  The European Center runs their high-resolution global model out to 10 days.  

So why go so far out in time when forecast skill really drops after 5-7 days?  The only reason is that in some situations useful forecast skill can hold out 7-14 days or more.

The major weather prediction entities (National Weather Service, European Center, UKMET office, and others) also regularly run global forecasts out to SIX MONTHS. To run that long they have to use both atmosphere AND ocean models, coupling them together.

So if forecast skill fades after roughly one week, why do they do this?  The reason is that the forecast centers hope that some average aspects of the weather are predictable out several months.  For example, some aspects of the future atmospheric conditions are driven by sea surface temperatures, which change relatively slowly...perhaps that might provide some skill.  In fact, that is exactly what we are doing subjectively when we talk about El Nino and La Nino weather....that the temperatures of the tropical Pacific drive weather in the midlatitudes (and elsewhere) for the months ahead.   Or perhaps the slowly changing snowpack over Asia might perturb that atmosphere during the next few months in a coherent way...offering some predictability. 

OK, lets check out the goods!  Here are the monthly forecasts from the National Weather Service Coupled Forecast System Version 2 (CFSv2) for the next 6 months.   This figure shows the surface temperature difference from normal (the anomaly) in degrees C.
A bit cooler than normal for March over us, but April is above normal and the remainder of the spring and summer are normal or above normal.  Wonderful news if true.

I can't show you the ECMWF seasonal forecasts (they charge!), but I can show you the long-term predictions from the excellent UKMET office coupled seasonal prediction system.  Here are the forecasts of surface temperature for the early spring and early summer for an ensemble (average) of a number of forecasts they made:
Very warm over the eastern and central U.S. and warmer than normal over us.  Time to check out the sales of sun tan lotion!  But before you get too excited, I should note that the skill of these forecasts is very modest.  But the consistency of the various systems is encouraging and provides added confidence.    My gut feeling from these forecasts and the persistence of the West Coast ridge pattern, is that this will be a far better spring than during the last two years.


  1. I think this a great post about these seasonal forecasts. While a warm spring might be great for the Pac NW, I noticed that the models were predicting normal or above normal for the rest of the U.S. My biggest concern for this is Texas. While above average temp doesn't always mean dry, its entirely possible they could see another hot and dry summer based on those forecasts. That being said, it could also be really hot and wet. I just know they don't need the drought to keep going.

    Thinking about the Pac NW, whether its in the models or not, I'm hoping for one last winter storm (down to the valley floors) before we get our normal spring. I DO hope we get a normal spring though.

  2. I'm not concluding what this spring will bring until I see the Taiwanese model.

    Leaving a remark tonight to acknowledge a comment you left two posts ago. What looked exciting a few days ago, now is more of a consistant flat ridge. Just like you said.

    I learned my lesson. Never happen again, promise. But hey, models are hinting on a return to la nina weather the week of Feb 24th-28th :)

  3. Anything better than last spring would be wonderful. That degree of horribleness extinguished any memory I had of the previous spring.

  4. It would be interesting to see how the seasonal forecasts for the last say 5 years compared to the actual average temperatures. Maybe "An exercise left for the student"...

    Averages can be deceiving. We are planning a trip to Disneyland in California this summer. I wondered if the temperatures would be much different between July and August. The average temperatures were about the same but, interestingly the highest record temperatures were in July! So, while the average in August might be a bit higher, the probability of hitting one of those really high temps is less.

  5. Ok, so we see the mean prediction. I assume that's the mean of a bunch of runs of the physical model. But the mean by itself is rather meaningless - do they ever show the variance, or some meaningful derivative of it (Bayesian credible interval, perhaps?).

    Otherwise I don't get a gut feeling on whether we are seeing signal or noise.

  6. Cliff, just thought I would let you know that your new posts are no longer showing up in my Google home feed (they stopped a couple weeks ago). If you are showing a drop in traffic, that may be why.
    If this has happened to anyone else, let me know how you fixed it.

  7. What a difference a ferry ride makes... I really notice on this set of charts how the dividing line between above and below normal seems to run just inland of us out here in the San Juan's. While most of the state moves from slightly below normal to slightly above normal, the islands never quite get there.

    I am also curious as to why the NWS seasonal outlooks still show us (not just the islands, but much of WA) as clearly below normal in the next three iterations, not going to normal until the May-June-July chart.

    We have bulb plants and many trees already starting to bloom, but snow in the forecast for the hilltops this weekend... I'm hoping the plants have inside information...

  8. The following situation reminds me of the question that starts with, "If a tree falls in the forest..."

    So, if a storm bombs in the eastern pacific, and it never hits the coast, was there really a storm?

    I saw this in this morning's forecast discussion and was compelled to look at the model forecast maps:


    While the links last:

    1600 PST thursday nothing happening http://www.atmos.washington.edu/mm5rt/data/current_gfs/load.cgi?images_d1/wgsfc.33.0000.gif
    0100 PST friday a little something http://www.atmos.washington.edu/mm5rt/data/current_gfs/load.cgi?images_d1/wgsfc.45.0000.gif
    0700 PST friday 50 to 60 knot gusts about 400 miles offshore
    1300 PST friday 70+ knot gusts incrementally closer but well offshore
    1900 PST friday field of 70+ gusts diminishing, incrementally closer again but still 100s of miles offshore http://www.atmos.washington.edu/mm5rt/data/current_gfs/load.cgi?images_d1/wgsfc.63.0000.gif
    1300 PST saturday small field of 50 to 60 knot gusts to sw of low approaching the coast

    A Canadian Buoy near the coordinates mentioned above (46036 South Nomad) http://www.ndbc.noaa.gov/station_page.php?station=46036 shows a drop in pressure from 0100 PST to 1400 PST (today) of 0.38 inches.

    So here is my question, since the models develop this storm offshore and then weaken it to the point where there is minor coastal impact at most, is it an issue? Models are great forecast tools, but at times they seem to only approximate future conditions. What if the models are wrong about the timing, location, and speed of the low enough that the coast sees the full or near full brunt of the storm. That could be hurricane force gusts tomorrow. Why the apparent confidence that couldn't possibly be the case?

  9. Correction: That could be hurricane force gusts Friday.

  10. It's interesting looking at the CFS and UKMET forecasts made last fall for this winter. Both of them expected "typical" La Nina weather for the Northwest (cooler/wetter than normal). So it wasn't just the human forecasts that were wrong, which partially rely on analogs (i.e. this is what happened in past La Nina winters).

    But the climate models were also wrong. And there's no La Nina "equation" in these models. For short-range forecasts, we often blame the model busts on poor analysis of initial conditions. But in these models, we actually have good analyses of the global Sea-Surface Temperatures (e.g. cold equatorial Pacific). Obviously there are still aspects of our ocean-atmosphere circulation that we just don't yet understand.

  11. Weather forecasting in true sense is the application of science and technology to forecast the atmospheric state for a given place or area for upcoming time.7 day forecast


Please make sure your comments are civil. Name calling and personal attacks are not appropriate.

More Wildfire Misinformation at the Seattle Times

 The Seattle Times continues to shamelessly exaggerate and hype the regional effects of climate change. This week they really went overboard...