Ask a cruiser when they are leaving for the tropics and the answer usually contains two entirely different claims, welded together so smoothly that nobody notices the seam.
"We're going in June, and we're watching for a window."
The first half is a claim about climate. The second is a claim about weather. Both can be correct, but they come from different sources, are known at different times, and answer different questions. The trouble starts when they blend into one feeling that the trip is on.
Climatology answers what is normally true here at this time of year. A forecast answers what will happen on Tuesday. Climatology describes the population of all Junes; a forecast describes one particular week. Neither converts into the other, and the failures that follow from trying are what this article is about.
What climatology is good for, and blind to
Climatology is the accumulated record: what the wind, the pressure patterns and the storm activity have historically done in a place in a given month. Pilot charts are its traditional expression; seasonal outlooks from meteorological services are the modern one.
It is superb at four things: choosing a season, choosing a route in the large, setting expectations for the kind of sailing a passage is, and defining the shoulders — where the reasonable part of the year begins and ends.
What it cannot do is say anything about a particular day. A pilot chart showing a month is generally settled is entirely compatible with the week you chose being awful. Averages are made of variation; the variation is hidden, not removed.
Its second blindness: it is a record of the past, and the past is not a fixed guide. Pair it with the current seasonal outlook from a national meteorological service, which is the thing that says how this year is expected to differ.
What a forecast is good for, and blind to
A forecast is a physics simulation: somebody measured the state of the atmosphere and ran the equations forward. It degrades with range because that measurement was imperfect — the atmosphere is not observed everywhere — and the model is itself an approximation. Both errors compound, so a forecast does not fail at a threshold. It loses resolution progressively, small features disappearing first and large patterns surviving longest.
At long range, then, a model may be broadly right about where the highs are and quite wrong about the trough that will spoil your Thursday. The chart looks smooth. That smoothness is not calm weather; it is the model running out of skill and returning something that resembles climatology — the trap, because it still looks like a forecast. How far ahead the output carries real skill is a question for the issuing service's published guidance, not for the number of days your app is willing to draw.
What a forecast cannot do is authorise a season. It has nothing to say about November. Every forecast is a statement about the near future of a decision whose frame was set months earlier.
The table to take to the chart table
| Climatology | Forecast | |
|---|---|---|
| Answers | What is normally true here in this month? | What will happen on these specific days? |
| Source | Pilot charts, historical records, seasonal outlooks | Model output, synoptic charts, professional forecasters |
| Decided when | Months to a year ahead | Days ahead, revisited daily |
| Governs | The season, the route in the large, the shape of the year | The departure, the tactics, the sail plan |
| Typical question | "Is October a reasonable month to head south?" | "Is Tuesday a reasonable day to leave?" |
| Blind to | Any particular day; how this year differs from the average | Anything beyond its skill horizon; the season as a whole |
| Characteristic failure | Treating an average as a promise | Treating a long-range run as certainty |
| Can never | Authorise a departure | Authorise a season |
| Should produce | A window of dates, and a deadline | A go / no-go, on one day, for one passage |
Read the bottom two rows again. A pilot chart can never authorise a departure. A forecast can never authorise a season.
The handover: where one takes over from the other
The two are not rivals. They operate in sequence, and knowing where the handover falls is the skill.
Months out — climatology only. You choose a season, a direction, rough dates and the hard deadlines the season must fit inside. No forecast exists for this period; any confidence about specific weather at this range is manufactured.
Weeks out — reconnaissance. Watch the synoptic charts, not to time a departure but to learn how the pattern is behaving this year. A crew that has watched for a fortnight reads the departure chart far better than one starting on the day.
Days out — the forecast takes over. Only now is a specific departure a real question. Climatology gave you the window; within it, the forecast decides.
At sea — the forecast, updated, plus your own instruments. The plan you left with has a shelf life.
The common structural error is letting a decision made in the first stage authorise an action in the third. "We always go in June" is a climatological statement being used as a departure decision. It is not one.
Long-range charts manufacture deadlines that do not exist
Every weather app shows you a fortnight, and the graphics are identical at every range: same icons, same arrows, same authoritative rendering. Nothing tells you that day two and day twelve are different kinds of claim. Day two is a forecast. Day twelve is a suggestion that will be replaced several times before it arrives.
The danger is not that people believe day twelve exactly. It is subtler: a long-range chart forms a picture in your head, and the picture persists after the data has been superseded. When the update shows something different, you experience it not as new information but as an annoying deviation from what you already knew.
The second failure mode produces genuinely bad departures. A chart showing an unpleasant feature arriving in ten days creates a deadline that does not really exist — we need to go now, before that comes through. It may not arrive, or may arrive elsewhere, later, or weaker. But it has already converted a patient decision into an urgent one, and urgency, not weather, is what puts boats to sea in conditions their crews would otherwise have declined.
Deadlines from a calendar are real. Deadlines from a long-range model are frequently fiction, and more persuasive precisely because they feel like seamanship rather than schedule pressure.
Working it on the New Zealand–tropics corridor
Our reference passage is the one we know best: New Zealand to Fiji and back, four consecutive round trips between 2023 and 2026 in a 2007 Fusion 40, refitted in 2022, NZ Category 1 certified, 1.1 metre draft. It is roughly 1,100 nautical miles and typically seven to ten days at sea. Boats generally run north around May–June and south around October–November, with the South Pacific cyclone season falling roughly November to April. Treat those dates as indicative and shifting: before building a plan on any month named here, check the current seasonal outlook from a national meteorological service and your own insurance conditions, which are usually the stricter of the two.
Notice what that does and does not do. It sets a window, and a deadline at the far end of it — climatology working correctly. It does not name a departure day, and it cannot. A boat that reads "October–November" as authorisation to sail on a particular Tuesday because the haul-out is booked has used a statement about all Novembers to decide one week.
Southbound shows the other half. Minerva Reef, mid-ocean, is commonly used as a staging point on the way south. A stopover does not shorten the passage; it splits one long commitment into two shorter ones, each timed on its own forecast — the purchase of an extra decision point, which is the most valuable currency in passage planning. We work the corridor season through in full in The Pacific Year.
The short version
Write your next passage's window and its departure decision on two separate pieces of paper. The window is a range of dates from climatology and the current seasonal outlook; the departure is a single day from a forecast. Never let one page argue for the other. Set a written rule about long-range charts; ours is that nothing beyond the near term may create urgency, only curiosity. And if the window closes with no acceptable forecast in it, do not lower the standard: the plan has failed, and you need the other plan you built in advance.
Keep reading: see the distinction applied to a real passage in our guide to the Fiji to New Zealand passage, then work out where the season ends with our South Pacific cyclone season plan. If you want the whole year in one place — the windows, the routes and what happens between the lines on the chart — that is The Pacific Year.