Predict Wind?

[Hippocampus]

Peter you come to the table with considerable knowledge and understand wave reflection, refraction and dispersal. You understand how waves interact and will bend around a headland. Computer files may give information about wave interaction in open waters but commonly little insight to local coastal phenomena. Have found some even after years of cruising continue to not account for wind v wave, compression zones, down drafting from land contours, inlet phenomenon or other hyperlocal situations. I don’t think you can blame Windy, PredictWind or any of the common products for that lack of understanding. Think all but the most casual boaters should take the 1/2-1 hour to read and understand wave physics. Also the few hours to understand weather genesis and why any presentation of computer files will have their limitations. You don’t care about regional weather. You very much care about the weather directly over your head and the sea state under your boat.

 

[Shifty]

I believe Predict Wind runs their own weather model, as well as presenting many others. In particular their PWE and PWG models. These are their own model run with ECMWF and GFS initialization respectively. All of these weather models are run using a grid of cells, it is a finite element numeric type of computation, initialized by current observed conditions. Most of the models are using a fairly course grid and many don't include the effects of terrain, all to save computational time which is huge in these models. Big grids are OK in the open ocean where things don't change too fast, not so good inshore.

The big advantage of Predict Wind's models becomes apparent when there is terrain (they do consider terrain) and particularly where they run a 1km grid vs the 8km grid the others run. The areas served by this are limited because of the very high computational time, and you have to pay the subscription rate to get it ($29/yr is the normal rate). However the accuracy is very noticeably better, at least here in SF Bay and in the PNW where conditions can be very local and greatly influenced by terrain. You can check on the PW website coverage map for the areas served, 1km in small red squares.

In using any of these services, be sure to compare several different models. Often they will agree for today or the next couple, less often out 5 or more days. The agreement or disagreement you see in the models gives you a big clue about the reliability of the forecast: if everyone says it's going to blow in 5 days, it probably is!

Agree with DDW here. I’ve been using PredictWind for a few years and found it to be an excellent tool. Around Puget Sound, we’ve found the PWE and PWG models to be quite accurate. We never rely exclusively on the app though. We’ll typically compare it with NOAA forecasts and our own knowledge of how wind combines with currents and land. I’ve also used professional weather routers and found the 1km and 8km models to be on par.

It’s a great tool to start learning.

We just used it to plan a weather window during the big blow that came through in late December. We were at Alderbrook and it looked like the dock was going to get nailed, so we headed north after watching PredictWind and consulting several other sources. Hood Canal was sheltered through most of the day and we found one of the few calm ports at Ludlow to ride out the night.

We’re the white dot in this picture.

One note, the wave forecasts aren’t particularly useful for small coastal cruisers. Anything under five feet shows up as purple (calm). So, you really need to combine the wind speed (and gusts), duration, direction, and current to plan the passage.

 

[bags]

user question on predict wind

I have a friend cruising in the N Atlantic and he sent me a string so I could check his location:


https://forecast.predictwind.com/tracking/display/SV_name_1


It's those underscores I think that are wrong. I've tried ?,% and other chars but I can't get it... anyone familiar with this nomenclature?


Thanks much!

 

[Delfin]

How good is Predict Wind?


Are their predictive algorithms better than the other guys?



We really don't need current weather, we can read the stations, we need good forecasts... that's the hard job.



They are certainly quite pricey. Worth it?


And worth it for us intercostal guys, where crossing Tampa Bay is a big deal?

We're using PW in the Pacific, and it has some pluses and major minuses. Pluses - they daily prepare their own amalgam of different prediction models, and rank the accuracy of different models across different time slices. Negatives, and this comes from someone who has developed complex software apps and written the documentation for same. They have no user's manual, so you have to watch hours of poorly produced videos to learn anything other than intuitive basics. I asked them about this and their response as why no manual is that they were constantly making changes. This approach says they know little about software dev, since the standard practice is a release schedule, say every six months with mods to the user's manual released on the same date. It lacks some fundamental tools, like ability to measure distances or plot a multi turn course.


Their models seem very accurate, at least where we're voyaging.



Definitely not worth the pro version, but the standard is worth it for long trips.

 

[DDW]

I think your view of software quality and documentation is out of date. These days, you release beta software, have customers test it, all feedback and documentation is handled by the community on a forum (or some bad Youtube video). Even very large concerns are doing this, billion dollar enterprises. Sad that, but it is where we are. The SQA and Tech Pubs departments of old are long gone.

 

[Delfin]

I think your view of software quality and documentation is out of date. These days, you release beta software, have customers test it, all feedback and documentation is handled by the community on a forum (or some bad Youtube video). Even very large concerns are doing this, billion dollar enterprises. Sad that, but it is where we are. The SQA and Tech Pubs departments of old are long gone.

Perhaps, although having just sold my dev company to an insurance company I don't think I'm that out of date. Nor do I know of any software company that releases a beta version as a production release for the simple reason that beta versions are for testing with a small group of curated users, and finding out the crap the QA propeller heads missed. Granted, decent documentation is getting rarer, but an absence of any is a mark of people who need to go back to dev school. Check out the 160 page users guide to Coastal Explorer as an example of how it should be done. CE is arguably not much more complex than PW, but CE knows how to do the job.

 

[DDW]

I agree with your view of the Way It Should Be, but my experience in a wide range of recent software is far from that Way. Especially for stuff that is given away or sold cheaply. In professional level products, you can still find SQA, documentation, and support, but it is getting rarer even there.

When I say release beta software, they don't call it that, but there are glaring, elementary errors that would have been caught by the slightest bit of testing. Products from Microsoft, Apple, etc. What we used to call beta software. Yes, I've been in the business. Coastal Explorer is a medium priced package, that might support some level of quality. PW is given away for free, or a very modest subscription.