View Single Post
  #6  
Old March 14th 17, 07:07 AM posted to rec.aviation.soaring
WaltWX[_2_]
external usenet poster
 
Posts: 310
Default SkySight forecast available for SW USA

Walt Rogers "WX" REVIEW of Skysight - Service announced today...
Western and Eastern U.S. domains

I first learned of Skysight in late 2016 from Jim Staniforth, who was impressed by Matthew Scutter's work. I browsed it just before writing my Soaring article published in February 2017 and included it a table of soaring web services. My first introduction came when briefing the U.S. Team in Benalla during several weeks in January.

Overall I'm very impressed. Matthew Scutter bases his work on Dr Jack's RASP software. It is running one of the later versions of the U.S. community WRF numerical prediction model at something like 3-4km resolution (comparable to NOAA/NWS's HRRR model - which also uses the very same model framework).. Matthew is forecasting for seven domains in the world: East, West Australia, Alice Springs, East and Western U.S., Europe and South Africa. A model run is produced before the soaring day starts... and in the evening for the next day, plus three additional soaring days. There is a total of four days forecast: Today, Day+1 (tomorrow), Day+2 and Day+3. Each model run is initialized by the U.S. Global Forecasting System (GFS) model for each of the days. It's amazing what one person can do with cloud computing resources! Over 2 terabytes of data are moving around each day! Matthew is ... or was....a Google software engineer with cloud infrastructure experience, skilled at managing a large deployment like this (also with several years of running RASP). From now on he will be working full time on the project (not counting the time flying his Discus 2A in 1000km flying weather. Matthew is Junior World Soaring champion and competed at Benalla 2017 for the Australian team.

My initial impressions are that Skysight is a major contender as a world class soaring forecast service. It is not as polished at TopMeteo, but can be expected to be constantly improved over time. Forecasts of thermal height (soarable height Hcrit) and thermal strength are comparable to all the services from my spot checks during the last several months. XCSkies covers soaring over the world using the U.S. GFS model and also performs quite well. Although, lately XCSkies has not shown any indications of new development.

Take one day as an example ... Mar 13, 2017 valid time 21GMT (2pm PDT). I have compared graphical forecasts for Skysight, XCSkies, DrJack and TopMeteo for the SW U.S. concentrating on a small region in my neighborhood around Mojave, California. Here, I've posted a folder of screen shots for thermal height (Hcrit) and thermal strength in a Dropbox folder for you to compa

bit.ly/SkysightCompare_20170317

My Comments comparing this one day... also based on a general impression of how all the services perform:

Skysight wins in the comparison of both time and spatial resolution (horizontal detail; 30minute time resolution). So far, thermal height (Hcrit) and thermal strength are comparable to DrJack with the NAM (to be expected)... even comparable to XCSkies when looking at their GFS and RAP output. Surprisingly, the XCSkies NAM output looks "hot"... too strong thermal strength. DrJack is of course much harder to read the fields and coarser in resolution. TopMeteo uses double digit thermal strengths on their charts. TopMeteo from my experience tends to run "conservative" on thermal heights/strengths... but not unreasonably so. All services except DrJack legacy provides 4-5 day forecasts.. but Skysight provides high detail (time and resolution) after day one. In todays numerical weather prediction world, it doesn't make any sense to operate at any higher resolutions than 3-4km... because the physical simulation of thermal prediction don't work well at higher detail.

XCSkies in my opinion has a slight edge in the UI department. It's easier to read model winds plotted as wind vectors over the colorized fields. Skysight's embedded streamline vectors over the color contoured graphics on top Google Maps gets washed out. I'm finding that I use the slider to change transparency a lot ... first to geolocate myself, then to clearly see the colorized fields and interpret them numerically. I like having some vector lines of reference like roads on XCSkies maps because I then don't have to constantly fiddle with transparency. Skysight's UI is not bad... a close second... but needs improvement.

Skysight is being actively developed by a young, and energetic software developer with loves to fly gliders. Quitting your day job to start such an ambitious undertaking is a big motivator. So, we'll be seeing and hearing a lot more from Matthew. Even though TopMeteo has a strong team of skilled modeling/science professionals who are also glider pilots... they had better keep an eye on Matthew's project.

Walt Rogers "Wx"