Starlink Will Download the Weather. Here's Why I'm Paying for PredictWind Professional.
The four GRIB lines I won't untie from Galápagos without have to come from a workflow that still works when Starlink is dark. I'm buying PredictWind Professional from Bend. The boat can wait.
Tropical Storm Dolly spun up in the Atlantic on Thursday, 40 knots, a thousand-plus miles east of the Leewards. We're in Bend. The kids are not watching the NHC cone, and I'm not going to pretend a short-lived tropical storm is our problem.
It is a reminder. In July I wrote down four GRIB lines I will not untie from Puerto Ayora without. Those lines have to come from a workflow that still works when Starlink is dark, when the dish is scheduled off for the night watch, and when the only uplink left is the Iridium GO! exec I hope stays in its Pelican case. I already specced the radio. This note is the software.
The four lines haven't moved
The Galápagos–Nuku Hiva rehearsal is 2,965 nautical miles. Sailed the way you sail with four kids aboard — reefed at night, no spinnaker after dark — I plan on 180 miles a day and call it eighteen with the slow start. The go/no-go is not a vibe. It is four lines:
Southeast trades of 15–20 knots showing at 5°S, 95°W and holding across the first 500 miles on a seven-day ECMWF run. ITCZ north of about 2°N, with no southward sag in the 96-hour outlook. No organized convective cluster west of the islands on the satellite loop. Combined seas under 2.5 meters.
If those are true, we go, even if it's a Tuesday and the provisioning list has holes. Waiting for perfect on an 18-day passage is a category error. The software does not get to move those goalposts. It has to produce them, on a laptop in Oregon and on a satellite modem in the middle of the Pacific, and it has to keep producing them when the models disagree.
What a family boat actually needs from a forecast stack
Agreement is the signal. One model being optimistic is not a window. I want ECMWF, GFS, and UKMO on the same screen, plus PredictWind's own PWG and PWE runs, and I want to see where they stop overlapping. The first three days out of Galápagos are the whole problem — light air, leftover ITCZ squalls, a current with opinions. After that it's the most reliable downwind conveyor on earth. The stack has to be good at the ugly beginning, not the easy middle.
Two tools, used in order. Departure planning answers which day. Weather routing answers which track. I already think in departure windows (March through May, French Polynesia's cyclone season in the rearview). What I have not had is a way to put four candidate Tuesdays next to each other and read wind, waves, CAPE, and rain down the same 2,965-mile rhumb that I will not actually sail, because you don't rhumb-line this one. Drop southwest to 3–5°S first, then bend west. The routing has to know that, which means it has to know the boat.
Polars for a 50-foot cat we don't own yet will be wrong. That's fine. Wrong and consistent beats a gut feel that changes with coffee. I'll start with a conservative cruising polar — 180 miles a day is the planning number, not the brochure number — and tighten it when we have a hull and some miles. The point of running it now is to make the four lines a weekly habit, not a theory I dust off in Panama.
The piece I would not skip with kids aboard is wave modelling. PredictWind's Professional tier will show roll, vertical acceleration, and slamming along the route, not just a wind arrow. Their own threshold is blunt: a forecast roll of 4 is an uncomfortable, possibly dangerous passage even when the wind averages 20 knots. With four kids below, sea state is the actual limit. Wind is what people talk about on the dock.
The other Professional-only item that matters on this route is ocean current. Panama to Galápagos is one of the named cases in their docs, alongside the Gulf Stream and Gibraltar. Standard will route with currents in the calculation and hide them. Professional lets you see them. I want to see them. A family boat making that approach on a schedule is not the place to take the current as a rounding error.
Standard is the cruising default. I'm not buying the default.
PredictWind's own pages push Standard for cruising boats: $249 a year, or $99 every three months. Professional is $499 a year, or $249 every three months, with a seven-day trial on both. The extra $250 is high-resolution GRIBs, the current maps, the boat-motion numbers, AIS in the Offshore app, and the ability to see what the router already knew.
I am buying Professional. Not because more tiles on a map feel serious. Because the two things that actually change a go/no-go with children aboard — sea state, and current on the Panama–Galápagos and Galápagos–Marquesas legs — sit on that side of the line.
Put $499 next to the rest of the comms stack I published this week. Starlink is roughly $2,400 a year after the first. The Iridium GO! exec, averaged, is $1,500–$2,000. The forecast subscription is the cheap layer. It is also the layer I can start using from Bend, this week, without a boat, a dish, or a SIM.
PredictWind launched DataHub Pro on Wednesday, a $599 onboard hub: tracking, AI polars, over-the-horizon AIS, a failover between Starlink, LTE, and Iridium. That is a boat purchase. It wants NMEA 2000, a hull, and a Professional subscription feeding it. I am not buying a hub to sit in a drawer in Oregon. I'll look at it when we have instruments. The forecast subscription is the thing that has a job today.
How this actually runs on passage
The workflow splits the same way the comms stack does.
On Starlink, in the morning weather-and-email block I already budgeted power for: full website, all the models, departure planner, the works. Dish on, files down, dish can go back to sleep. That is 95% of the days.
On the GO! exec, at least one day in three, on purpose: PredictWind's Offshore app, which is built for this. Compressed GRIBs, routing, tracking, departure planning, over Iridium. Slow. Fine. The backup is not an emergency phone. It is a complete, independent weather workflow. If I only ever run it after the dish has failed, I will not be fast enough at it on the day it matters.
The four lines stay the four lines. If Professional's router wants to leave on a day that fails them, we don't leave. The app is a way to see the lines clearly, not a way to negotiate with them.
What I'm buying this week
The next step is the Professional forecast subscription, from here, so the Galápagos rehearsal stops living in a spreadsheet and starts living in the same app I'll use on passage:
No hardware. No DataHub. No second weather service "just to compare." One subscription, the Professional tier, exercised on a laptop in Bend until it is boring. Boring is the goal. The first time I download a GRIB over the GO! exec should not be 200 miles west of Isla Isabela with four kids asking if we can swim.
Disclosure: that PredictWind link is an affiliate link. If you subscribe through it I may earn a commission. It does not change the price, and it is the only forecast stack I am actually speccing.
