Predict Wind - Which Models do you use?

[Franziska]

Originally Posted by SailingHarmonie

PredictWind shows the outputs from all the major weather models that are out there and presents the raw outputs from the models. There really are no others. Predict wind is not "accurate" or not, it just reports what the major weather service models are saying. It is exactly the same data that everybody else has.



People who complain about the wind always being stronger than forecast usually just don't understand the presentation of the data. The "wind speed" forecast is a 2 minute average wind speed. So literally HALF THE TIME the wind is blowing harder than that! So, yes, when you are trimming sails you can expect winds stronger than the "forecast" half the time. Even wind gusts are averaged, just over shorter periods, typically 3 seconds. And how many people actually look at the gust forecasts and use that in their evaluation of the conditions?

Funny you mention the gusts forecast. I usually this instead of the normal wind forecast and closer to reality sustained.

Saying this for the Med and Iberian Atlantic coast.

Sent from my MHA-L29 using Cruisers Sailing Forum mobile app

[Jamme]

Originally Posted by SailingHarmonie

. . . when the models all agree on a forecast trend, then I have confidence they are all likely to be close to correct, and picking a "favorite" doesn't matter.

If they differ in ways that are important to me, then I know that something between one of them and all of them are wrong.

Couldn’t agree more. The most used PW tool for me is the “tables” feature, where I can quickly and easily compare all models in one single page; see example below.
What I have been noticing is that, often, models disagree more on “when” than “what” — one model may predict a front to arrive earlier than others, so numbers may wildly disagree at a specific “snapshot” time, but not so much so if you read and understand the whole “movie” — throughout, say, 24-48 hrs.

[Simi 60]

Originally Posted by Moana Bella

For Australian coastal, I trust the European model the most, and GFS the least.
I found on the Western Australia coast, they all underestimate the wind by about 10-15 knots. Probably because there is always lots of wind there. But on the Southern and East coast, I have found them to be very good inside of a week.

The BOM who have more Meteorology knowledge than you or I use the ACCESS modelling, with their own input on top

[wholybee]

Originally Posted by markxengineerin

If I were a better programmer I would
-take a snapshot of the models each day
-keep a 10 day history
-measure the average accuracy of the different models at 1 day, 2 days, 3 days, etc and for different areas. This could be roughly done by using the "current 12 hour" prediction as Truth, but would be even better if there were an easy way to compare to actual measured
-Make an app or website to continuously monitor/share the results

Maybe someone has done it, or someone here could. It would be a fun project and powerful amount of data relatively easy to get.

Predictwind has done it. It is the validation function available on the web app, but not in the offshore app.

[wholybee]

Originally Posted by SailingHarmonie

PredictWind shows the outputs from all the major weather models that are out there and presents the raw outputs from the models. There really are no others. Predict wind is not "accurate" or not, it just reports what the major weather service models are saying. It is exactly the same data that everybody else has.

People who complain about the wind always being stronger than forecast usually just don't understand the presentation of the data. The "wind speed" forecast is a 2 minute average wind speed. So literally HALF THE TIME the wind is blowing harder than that! So, yes, when you are trimming sails you can expect winds stronger than the "forecast" half the time. Even wind gusts are averaged, just over shorter periods, typically 3 seconds. And how many people actually look at the gust forecasts and use that in their evaluation of the conditions?

It isn't just a 2 minute average. It is much worse. It is an average of 1-3 hours, over 81 square kilometers. (more or less depending on the model) The higher resolutions you see (both in distance and time) are interpolated in the software.

That is the most important thing people seem to miss about the way the models work. Withing that timeframe and square space, the wind speed and direction will vary dramatically from the models.

It is a mistake to try and follow a route or very specific advice from any modeling or routing software. It is just a tool to give you something to think about. Predictwind (or almost any model) are very reliable in determining if a planned route is SAFE, or helping decide the departure day. But they should not be used to determine the exact route, or the departure hour.

[rweeton]

My wife and I use all the models. We sort of average them out in our heads. Unless one is radically different it doesn't seem to make much difference.

[hlev00]

Originally Posted by GreenWave

Based on your note above it sounds like you are so-so on PW. Is there something else you use and like better?

I use a variety depending on the circumstances. For offshore and route planning i like PW especially the ability to get forecasts over Iridium Go. I also use the CAPE forecasts on long passages offshore or coastal in the summer months. I just don’t find their enhanced forecasts more useful. In general their interface is pretty well done.

For my coastal cruising in New England I like the free version of PW but my go-to is SailFlow as I get wind, some human analysis and current observations from a lot of bouys. I couple that with Wunderground. I also reference Windy when I drop off of PW pro. They have Euro for free. They have tons of other data. One downside is that you now have to sign up for the Pro version to get access more that five days out and I don’t know what else as I haven’t bothered with Windy since before then.

[HankOnthewater]

I agree with Sailingharmonie in post #6

Originally Posted by Jamme

Couldn’t agree more. The most used PW tool for me is the “tables” feature, where I can quickly and easily compare all models in one single page; see example below.
What I have been noticing is that, often, models disagree more on “when” than “what” — one model may predict a front to arrive earlier than others, so numbers may wildly disagree at a specific “snapshot” time, but not so much so if you read and understand the whole “movie” — throughout, say, 24-48 hrs.
Attachment 278279

I agree with Jamme about arriving fronts, either delayed or sooner than forecast. The way I do this is to look at fronts coming my direction (in West Australia) nearly all from a westerly direction.
1. I look about 8 to 10 days out, I use mostly a subscribed Windy for that. I note what the prediction is for fronts to arrive at a certain point, and if the airpressure is rising or falling (lows are deepening or not).
2. I follow these fronts once daily: arrival, direction of travel (and with that, changes of wind direction), deepening or not (and with that changes of wind speed).
3. With about 4 days out, even if the models adapted their forecast (of front arrival time) over the last 4-6 days, I found that a front that is speeding up, will also speed up in the last few days, and still arrive before the forecasted time.

Generally, if all models agree then the likelihood of the forecast being close to correct is high. If nearly all models disagree, then it is time to be conservative and pick the worst, or second worse forecast. Or better, use own developed intuition and knowledge.

In regards to course planning, I agree with the above posters to stay as close as possible to a straight line. I certainly would not go out on a limb (on a long tack one way) just because some software in the cloud made a forecast.

In regards to models, I look at all of them (fairly time consuming), and if I only look at one, it is generally the Euro model (ECMWF).
If I am not at home in west Oz, then weather gazing is becoming more time consuming for me.

[Tedd]

Rather than look at a bunch of different models, the approach I've taken is to gain experience using one model and comparing it to actual reports and on-the-water experience. I've chosen the ECMWF model, more or less arbitrarily, but the point is not so much which model to use as it is always using the same model, so that I get to understand how that model relates to the waters I sail in.

I took this approach based on my experience building CFD and thermal models in a completely unrelated environment. What I found with my own models was that it was valuable to use the same modeling tool and a consistent set of modeling assumptions and practices. Over time that helped me build more accurate models. I believe the same principle applies here: If I always use the same model (which, presumably, is developed using consistent assumptions and modeling practices) I will learn to understand it more rapidly than if I muck about with more than one model.

[Bill_Giles]

I generally use the default GFS but often glance at the others. The ECMWF is also pretty standard but I think that the UK Met Office is always very conservative and may often be one Beaufort scale number up. I fancy that this is a hang over from the Michael Fish forecast.

However, I like to look at the UK Met Office surface pressure diagrams which puts it into context and you can more accurately judge what is ahead.

Notwithstanding any of this these forecasts are generally accurate as to the coming winds except that the time of arrival is flexible more than a day or two ahead.

[billreilly]

For those cruisers who don't want to pay for PredictWind, you can get worldwide sailing weather forecasts for free on PassageWeather.

https://www.passageweather.com

[SV Coronado]

Use the PredictWind Model Validation tool. It will show the model performance of each model in your desired sailing location.

https://help.predictwind.com/en/arti...alidation-tool

Also, use your own eyes! Look at the weather outside and the forecast from several days prior. Was the forecast fairly accurate or not?

As many others have stated, no model is the absolute best model over time. They will come and go as conditions change.

[nofacey]

Over and beyond the multiple weather models, Predictwind offers accounts with both higher resolution (noticeably different results when dealing with tradewinds over land masses…..like Windward island chain in the Caribbean) and ocean current gribs. I agree with post on going rhumbline vs PW’s fastest course, especially on multi week passages - but deviating to catch a couple knots of current was totally worthwhile.

[TPG]

Recently did Cape Town to Grenada and we found GFS slightly more accurate than Euro. A few days all the models were off, but I found myself relying on the GFS model more than others.

[wholybee]

Another important point is "How do you define accurate?"

When I look at the forecast, I read it (for example) as "Later today the will will increase by 20% and turn more North." When read this way, the results of all the models are very accurate and much better than "At 2pm the wind will be 24kts from 300 degrees."

The former is plenty useful for planning and knowing what to expect. The later will leave you frustrated when it isn't exactly perfect.