The Offshore Voyaging Reference Site

Weather Analysis—Part 2, Strategic

In the last chapter I shared how I analyze the weather to plan cruises for four to five days out from forecast time—tactical analysis.

In this one I’m going to move on to looking out as much as two weeks (strategic analysis) and how we can use that process to plan safer and more fun cruises.

Login to continue reading (scroll down)

Inline Feedbacks
View all comments
Michael Jack

Great stuff, John. I am setup with the tools and started doing the process daily thanks to the first 3 articles. I hope to have this totally imbedded in my head by the beginning of the next season (it can take that long). I had to laugh at your description of North Sea and North Atlantic weather as most of what I read is by folks in your part of the world or the West Coast of the US none of which I recognize especially around the lovely neat progression of fronts. This is what today’s surface pressure chart looks like in my cruising area. It’s a hydra-headed monster (thank god I am not on board even moored):

Mike Evans

hi John, interesting article, thanks, do you think the tools in PredictWind allow you to do enough analysis for this type of passage planning?

I find it easy to use.


John, great advice. I put the starlink on the boat last summer for my cruise into the straight of Belle Isle. The weather was dissapointing but being able to easily look 15 days out at the trends allowed me to be much more strategic than in years past. I use TZ Navigator Professional and because of the download speed the area of my full Grib downloads i could watch weather windows develop and warch the three Hurricanes carefully finally making the dash fron Port au Basque thru the Bras dOr Lunenberg to Portland comfortably with the wind aft except for Lunenberg to Cape sable to escape before Lee entered the Gulf of Maine. Starlink is a disruptive technology. There is one feature of TZ Navigator that is worth nebtioning. There is the 15 day metragram wind speed gusts direction, wave heights, wave direction, précipitation, cloud cover.barimetric presdurec Wow! All untouched by the human mind! Beware!

This two week information pannel on the screen is associated with a little orange dot that one can move around with the cursor. The metrogram for the two weeks then reflects the weather for that particular location.

For a person who circumnavigated in the 1970s with only a baromeer eyeballs and SSB for weather this is absolutely astounding. I had to laugh the weather people asked us what the weather was before they made the forecast.

As you pointed out unless you are looking for more wind aft of the beam or pressed to vacate somewhere that is going to get really nasty there is no excuse to get hammered any more.

Besides I am getting too old to feel exuberation from being in a washing machine.

Matt Marsh

The whole idea of being able to run a global weather simulation on a supercomputer is, itself, relatively recent. So is the network that allows that model to be initialized with millions of points of properly calibrated satellite, ODAS, and ground station data.
It wasn’t that long ago that “oh, a ship at approximately L…. Lo…. radioed in 30 knots gusting 40 from the southeast, barometer 995 and falling, temperature 17°C” was the only initialization point for a forecast spanning 50,000 square miles.

I want to emphasize the importance of the initialization data in these models. The physics and mathematics behind GFS and ECMWF are a *very* close match to reality. However, multiphysics models in general are very sensitive to initialization and boundary conditions. In an area where ground stations and ODAS buoys are sparse, you need to make assumptions about how to fill in the gaps, and small changes in these assumptions often lead to big changes in the result.

I find considerable value in looking at ensemble runs and consecutive runs for long-term strategic purposes, as long as enough bandwidth to get them is available. If there is relatively little spread between different runs, that implies a relatively high confidence in the result. Significant differences between consecutive model runs, or between different runs in an ensemble, strongly imply that there’s some chaos or uncertainty in the initial conditions, leading to a low confidence in the results and a high probability of worse-than-expected conditions at any given point and time far in the future.

And I definitely would not want to be without the human-made charts. Model runs alone are, as John has pointed out many times, dangerous to rely on without skilled human interpretation, and the good old fronts-and-isobars charts are an essential part of that.