The scattered showers foretold for late Sunday and Monday have evaporated from the model runs now, and so have been looking around for a computer model run SOMEWHERE that had rain foretold for us. Once again, I could find a rain prediction for us near the end of a 15 day model prediction period. If we got all the rain the models predict for us in the 10-15 day period, year after year as they seemingly do, we’d all be growing bananas and coffee.
Below is a model prediction that was, in a sense, painful to see, and then de-construct, in a manner of speaking. A tropical storm/hurricane forms in the model, one that hasn’t even formed in real life yet (that’s how great our models are)! Eventually it races up the Baja California coast then turns toward Tucson, even strengthening that bit as it moves over the Sea of Cortez/Gulf of California. The map below is valid for 5 PM AST, Sept. 29th.
But the planets of the upper level steering flow are not aligned correctly; that storm takes a sudden right turn into Hermosillo and never makes it to AZ. Dud number one last night’s in the model run, except maybe they need some rain around Hermosillo so its not a dud for them. You can see this sequence here.
Then, right after this missed tropical storm, a trough from the lower latitudes, rich with moisture, comes barreling across the Southwest with major rains indicated in AZ two days later, in early October. Great!
Except that this situation must be de-constructed by checking out the NOAA spaghetti factory outputs to see if the indication of a trough here in early October is a likely dud as well. I really hate checking, though, when the model run shows something I want to have happen.
The first map below shows the postiion of an upper level trough at 500 millibars as it was projected for October 2nd at 5 PM AST. Looks great over AZ doesn’t? Below this map is the map of the rain in the 12 hour period prior to the same time as the upper air map, denoted by the green area (of course). Bring it on, baby!
Now, let’s check….as we must do as scientists and junior cloud mavens. We can’t just take things at face value that far in advance, as you already know.
Below that rain map is the “spaghetti” plot resulting from some bad starting data which is deliberately input into the model as it begins its calculations. The model is run again and again with different bad data points at the outset several times over. These bad data points produce slightly different forecasts over time, forecasts of jet streams and such that get more different as time marches on in the model run. Often though, a weather system is so strong, so powerful, that these slight errors that are put into the model make little difference, even 10-15 days out. Other times, slight errors make huge differences.
Instead of, “garbage in, garbage out”, we get some incredibly important information about our model runs doing this, though it seems counterintuitive. Where the many lines from each model run are in chaos, that is, all over the place, forget whatever was predicted for that region; its very uncertain, “garbagy.”
But if those lines cluster your area, you’re real forecast, the single run produced by the model from, say, as in this case, from last night’s data for October 2nd, is much more likely to happen.
With these thoughts in mind, I then looked at the NOAA NWS spaghetti plots that tell us whether the ACTUAL model run showing a nice trough right over AZ is an outlier (read, is a “phony”). When I looked at the spaghetti plot, the wild display of blue contour lines in the eastern Pacfic and western US told me that last night’s run, with that great trough was a “phony”, an outlier model run, and that outlier run is indicated by the yellow line. You can see how that yellow line, dipping into AZ, is NOT accompanied by a bunch of blue lines also dipping to the south in this direction. Instead, they’re all over the eastern Pac and West Coast. So, no strong signal for a good rain on October 2nd, dang.
After this great explanation (hah!) you, too, will instantly see that the two maps produced from last night’s model run and shown here are not very likely to be observed, but then’s there’s probably going to be a college football game on TEEVEE somewhere anyway, so who care’s?! :}
What can you do with this information? Let’s say you’re watching TEEVEE again, and your favorite weather presenter comes on with his long range forecast and has rain for October 2nd. You can then turn to your TEEVEE viewing mates and inform them, with great confidence, “That’s not gonna happen!” They’ll be quite amazed that YOU could critique an honest-to-goodness, highly paid weather presenter, one that has fun everyday predicting weather and gets a huge salary for having that fun while I have fun but get nothing. Your friends might ask how you know that rain won’t happen. You answer, “I checked the spaghetti plots. It ain’t gonna happen. Let’s watch some fubball.”
Of course, that rain prediction could STILL happen, but the chances are very small.
The sad end.
1though the blue lines are not exactly the same contour height level)