Any suggestions on models and data would be best to predict clouds in the days ahead and day of the eclipse? HRRR? Total cloud?
Below is a related discussion I found. Curious how those rcommendations overlay with models in Flowx.
This is what THE EXPERT in eclipse weather Jay Anderson posted on SEML (Solar Eclipse Mailing List) today:
Model madness
Date: Thu, 07 Mar 2024 15:36:02 PST
All of the weather sites show the same cloud details for a given model, but some to a better job of showing it to you. Here’s my assessment:
Numerical models:
weather cod edu: tends to show a fairly stark map of forecast cloud cover that doesn’t distinguish well between partly cloudy and cloudy. however, the site probably updates more quickly than the others because it services storm chasers who require very fast access to new information. ECMWF cloud maps are not available on the site. It does show some models that others don’t, particularly the NAM NST, which shows cloud cover as a simulated infrared satellite image; it gives the best “feeling” of the cloud amount but is only available for 60 hours into the future. Because simulated IR also includes ground emission, you have to loop it and watch closely to distinguish between low cloud and ground. The high-res RAP model doesn’t show cloud. The HRRR updates every hour in the daytime (it’s for storm forecasting), but goes out only 48 hours.
weather us: similar to the College of DuPage (weather.cod), but with an odd depiction of cloud (black clouds, yellow for clear skies). The site does give access to the ECMWF model, generally regarded as the best of the long-range numerical forecasts. Site also has a high-resolution GFS .125 version of the GFS. It has too many models - three is enough, but if you are mobile, pick a spot where all or many agree that skies will be clear.
windy com: the best for cloud depiction and it has a cursor you can move around to get an actual measure of cloud amount. A limited set of models, but the important ones: ECMWF, GFS, ICON, NAM, HRRR. Easy to zoom in and out and you can get a spot forecast at any point by right-clicking and selecting “forecast for this location.” Windy allows you to compare spot forecasts from all of the models on one page.
pivotalweather com: another storm-chaser favourite. Similar to weather.us with 26 models to select from but many of the models don’t offer cloud cover (particularly the ECMWF). Also a strange cloud depiction: clear areas are white; blue areas are cloudy. Mouse cursor shows cloud amount as you move across the map. Clicking brings up a sounding, if you know how to use them. Pivotal weather has one model the others don’t: the GEFS, particularly cloud cover from the GEFS. This model is an ensemble - it averages a parameter across 31 “runs” of the GFS model, giving an output that is the average of those runs. Each one of the model runs is initialized slightly differently to represent unknown or poorly measured observations at the start of the model calculations. It is the average of a range of opinions on the future. The GEFS goes out for 15 days (it goes much longer, but I haven’t found a site that shows the next 15 days.) I have no idea how useful it is but the idea is nice. Some ensemble forecasts use multiple models.
There are other sites, but I suggest you explore these four and pick the one you like. Pick (a) model(s) you like.
Some hints:
-
try the sites out now and pick one to use rather than doing it on eclipse morning. Some work better on cell phones than others. Save the links.
-
Experiment: pick a model and a forecast at a particular time in the future, go back 12 hours and 24 hours (or more) in the model start time and see how the forecast has changed between the current model run and the other one.
-
don’t use many models
-
Experiment: check a forecast for several days from now against the satellite picture when that day arrives.
-
Models can be hydrostatic and non-hydrostatic. The ECMWF is non-hydrostatic, the GFS is hydrostatic, and so on. Non-hydrostatic models are generally thought to be best for convective clouds, hydrostatic for system cloud, but this is probably no longer completely true (I’m a bit out of date on model characteristics).
-
Models have different resolutions: ECMWF, 9 km; GFS, 22 km; NAM, 5 km; HRRR, 3 km; ICON, 13 km. Etc. Smaller is generally thought better (?) but is computationally expensive and so only the short-range, limited-area models have really high resolution. It makes a difference in rough terrain.
So, now you know. Go crazy 
Jay
Jay Anderson’s recognized expertise in eclipse climatology and eclipse weather forecast consultation is such that he gets a free trip to every total solar eclipse anywhere in the world with TravelQuest (free except that he usually has to be a tour leader, if Travel Quest has multiple tours).