Monday, June 10, 2019

Does SkySight Work?

SkySight is a soaring forecast website that claims high resolution (both spatially and temporally) and accuracy forecast.  It is available for an annual subscription of $80.  Because it makes predictions for specific locations on an hourly bases it is possible to check the validity of its model.

Last Friday (6/7/19) was the best soaring day this spring.  I took Moshe's OLC trace and compared it to the SkySights prediction of thermal height.

Here is what I found:

Here is his trace superimposed on the SkySight (SS) prediction for how far an 18 meter could fly that day.  It is a daily prediction and does not change as you scroll the time bar at the top (ignore the hour).  SS rightly predicted it was going to be a good day.  Moshe in his 13.5M glider scored 494 OLC points (322 actual km flown).

At 12:55 he thermaled up to 6,660'.  SS predicted the top of climb (toc) to be 7,444'.

 At 13:12 he thermaled up to 6,861'.  SS predicted the top of climb (toc) to be 7,251'.

At 13:45 he thermaled up to 8,102'.  SS predicted the top of climb (toc) to be 6,923'.
At 14:33 he thermaled up to 9,042'.  SS predicted the top of climb (toc) to be 7,677'.

 At 15:17 he thermaled up to 6,700'.  SS predicted the top of climb (toc) to be 7,710'.  When you right click on SS it gives you the prediction for that spot.  In this case it was 7710' but the color would indicate that the toc should be 8-9,000'.  

At 16:05 he thermaled up to 9,811'.  The high point of his flight.  SS predicted the top of climb (toc) to be 1,345' and the color would indicate a predicted toc between 5-6,000'.  Again, the predicted value was different if you right clicked vs. looking at the colors.  By 4 pm, SS was predicting the lift to be spottier which seems correct but it did not accurately predict which spots were best.

He then flew to Twin Mountain Airport with out finding much lift.  He struggled there until he finally work back up to 6,672'.  SS predicted poor lift along his route and that the toc would only be 3,379'.  SS predicted that if he had flown over Mt. Washington and Twin Mountains (the Mountain not the Airport) he would have done much better.  I think he avoided the high country because it showed signs of over development.  Was there better lift in the mountains?  We will have to look at JD's and DW's traces who were there at the same time (or BU, NT, and EA who came thru a little later).

His final climb was at Moosilauke at 17:45 where he got to 8,682'. SS predicted 6,594'.

In conclusion, I have not made an conclusions about the usefulness of SS to me.  When its booming everywhere I think making accurate predictions is easy.  SS makes prediction about how the lift is going to die at the end of the day.  If it can predict where the lift will be at 18:30, that would be truly useful.


Moshe said...

Thanks for analyzing my flight, Tim! (Why did you choose mine?) My glider fits in the 13.5m class, but its actual wing span is 12.6m, and best glide ratio is about 32 or 33 at about 48 knots.

On the first leg the clouds were not working, the lift I did find did not get me anywhere near cloudbase, and there was lots of sink every time I left lift. Which means there was better lift somewhere, but I had trouble locating it, under the constraint of keeping within glide range of the few safe spots. Thus very slow leg. Second leg was high and easy and fast with a tailwind. The third and longest leg was slow due to more headwind than forecast.

Regarding why I didn't follow the high ground instead of the foot of the mountains near Twin Mtn: I've often gotten stuck low for a while at Gorham or at Twin Mountain. I was planning to follow a more direct route from the Mahoosucs (turnpoint was Goose Eye Mountain) to Franconia, away from the mountains, because forecasts I looked at in the morning (or night before) indicated better lift there (perhaps those models could "see" the overdevelopment). But as I was proceeding West I was drawn further south by the way the clouds looked, first to near Gorham (which was OK) and then near Twin Mountain (which was not so good). Probably could have climbed up over Mt. Washington, but that would have taken extra time and the clouds there didn't look better. And with my glider, even from 9000 feet, I can't glide all that far into a 15-knot headwind, so needed a more continuous line of good clouds.

Moshe said...

PS: I was flying a triangle task pre-declared before takeoff, so if I chose a less-than-ideal task WRT the actual conditions, I was stuck with it. I could of course deviate from the straight edges of the triangle to use better lift. But on the first leg I had limited options and the way the clouds looked turned out not helpful, nor did they look better anywhere nearby. In retrospect I probably would have been better off flying the same task in the opposite direction. (Check out RR's flight that day.)

Rick said...

Good analysis, Tim. Thanks for the report, Moshe. I'd like to see more of both.

SkySight seems pretty good to me, but as Evan says, it doesn't give any hint of the uncertainty caused by noise in the data.

It does a pretty good job of forecasting which way to go, I find.