The Swinomish Channel

The one place in the library with no current station in it, where the stream lags the tide by hours, and where depth and air draft matter more than the current does.

Common questions

Can my boat get through the Swinomish Channel?

That is a question about depth rather than current. The channel is a dredged cut and it shoals between dredgings, so what matters is your draft against the dredged depth plus the height of tide at the moment you are in it. A boat that does not fit does not fit at slack either — waiting for the turn does not help, only a bigger tide does. Read the current Notice to Mariners and controlling-depth survey before committing a deep-draft boat.

Does the Swinomish Channel turn at high water at La Conner?

No, and this is the mistake that catches people. The channel is driven by the difference in water level between Skagit Bay at the south end and Padilla Bay at the north — not by the tide at either end — and the stream that difference produces lags the tide by about five and a quarter hours. Timing a transit off a La Conner tide table will be wrong by hours rather than minutes.

Which way does the current run in the Swinomish Channel?

North or south along the cut, and which one depends on whether Skagit Bay or Padilla Bay is standing higher at that moment rather than on whether the tide is rising or falling. Because the two ends belong to different tidal regimes, the channel can be setting different ways at its two ends at the same time, so a single direction for the whole channel is an average and not a promise about the water where you are.

When is the current strongest in the Swinomish Channel?

Near the equinoxes. The channel is driven by the twice-a-day part of the tide rather than the once-a-day part, and the twice-a-day tide is largest when the sun is over the equator — so the strong season is March to May and September to November, and the quiet months are December and January, and July and August. The channel is therefore at its weakest during the summer weeks most cruising boats use it. Two knots is rare in any season, happens only on a few days a year and always inside that strong season, and the stream is setting north when it does.

Is there a current station in the Swinomish Channel?

There is not. The nearest published current station is about four miles outside the entrance, which is why the stream here has to be computed rather than looked up. TidePush drives a physical model of the channel with the head between the two tide gauges and calibrates it against a 2006 USGS survey — the only current measurements ever made inside the cut.

How deep is the Swinomish Channel and will my boat fit?

It is a dredged federal cut with a twelve-foot project depth, so what decides whether you fit is the height of tide at the moment you are in it, added to that dredged depth, against your draft. A boat that does not fit does not fit at slack either — no departure time changes it, only a bigger tide. Note that the project depth is a design figure rather than a recent survey.

How much clearance is under the Rainbow Bridge at La Conner?

About seventy-five feet at the mean higher high water mark, and because clearance moves opposite to the tide it varies by roughly fourteen feet across the range — the tide that gives your keel the most water gives your mast the least. Cables have also been reported over the channel about three feet below the span; that figure is reported rather than surveyed, and for a tall rig it rather than the bridge is the limiting obstruction.

Why this channel is different from every other gate

Every other timed passage in TidePush is a published current station: an agency measured the water and predicts it, and the app reads that prediction. The Swinomish has no station in it at all. It is also the only place in the library where three different constraints bind at once — the stream, the depth under your keel, and the air over your mast — and where the second of those usually matters more than the first.

Why the channel flows at all, and why it runs late

In short: The channel connects two bays whose tides do not keep the same time. Water runs through it because one end is higher than the other, and that is why the current does not turn when the tide is high.

The channel is a cut between two bays that keep different time. Skagit Bay at the south end has the larger tidal range and runs later than Padilla Bay at the north. Water flows through the cut because one end stands higher than the other, so what drives the stream is the DIFFERENCE between the two — the head — rather than the tide at either.

That difference peaks well away from local high or low water, and the channel has enough water in it that the stream takes time to respond to the push and time to stop once moving. The result is a stream that lags the tide by around five and a quarter hours, which is why a tide table for La Conner is the wrong instrument for this decision.

It also follows that the two ends need not agree. The cut can be setting north at one end and south at the other, so where the app names a direction it is describing the channel’s average behaviour.

What a tide is made of

In short: A tide is a stack of regular waves added together, each produced by one astronomical motion and each with a period fixed forever. Learn how large each one is at a place and that place’s tide becomes arithmetic.

A tide chart looks irregular: two highs a day, uneven, and the whole pattern breathing over a fortnight. That messiness is an illusion of addition. Underneath, the tide is a set of perfectly regular waves laid on top of one another, each produced by a single astronomical motion, each with a period fixed by orbital mechanics rather than by local geography.

Measure a place for a few weeks and you can work out how large each wave is there and when it peaks. Those figures are the place’s harmonic constants, and once you have them its tide becomes arithmetic — computable for any date, past or future, with no measurements and no network. That is the only reason this channel can be modelled where no current station exists, and why the model does not run out the way a fetched forecast does.

Each wave carries a standard name. The letter says which body and which motion; the number says how many times a day it cycles.

M for moon, S for sun, N for the moon’s ellipse, K and O for the tilt off the equator. A trailing 2 means twice daily and 1 means once daily, which makes M4 and M6 four and six times a day — the shallow-water distortions that appear only over places like the Skagit flats.

Springs and neaps fall straight out of the first two rows, and are not a separate phenomenon at all. The moon’s wave takes twelve hours and twenty-five minutes; the sun’s takes exactly twelve. That small mismatch means the two slide against one another and line up again every fourteen and three-quarter days. Crests together, they reinforce, and you get springs: the largest ranges and the fastest streams. Crests opposed, they partly cancel, and you get neaps. The fortnightly breathing every tide table shows is those two waves beating against each other and nothing more. (“Spring” here is from spring forth rather than the season — there are springs in January.)

Which half of the tide this channel feels

In short: The Salish Sea has an unusually large once-a-day tide, and this channel is almost blind to it. The cut runs on the twice-a-day tide, because that is the part that arrives at its two ends unevenly.

The Salish Sea is a mixed tide: its once-a-day waves are unusually large, together carrying about as much as all the twice-a-day ones combined. That is why the two daily highs here are so unequal. If you were predicting the height of the water, they would be half your answer.

This channel does not run on height. It runs on the difference between its ends, and there the once-a-day waves very nearly cancel: they arrive at Skagit Bay and Padilla Bay at almost the same size and almost the same moment, so they lift the whole cut like a bathtub and drive little through it. The twice-a-day waves do not. They reach the Skagit end distinctly larger and a little later, and that mismatch is what tilts the surface.

Read the last column. A twice-a-day wave turns something like a fifth of its own height into head; a once-a-day wave turns almost none of it, however large the wave is. The lunisolar wave is among the biggest in the basin and it drives the channel less than waves a quarter its size, because it raises both bays together. Taken as groups, the once-a-day waves carry roughly the same quantity of tide as the twice-a-day ones and produce about an eighth as much head. The Swinomish is a twice-a-day channel sitting inside a mixed-tide sea, and it is largely blind to the half of the tide that makes this coastline distinctive. The proportions come from the 2006 survey’s own gauge analysis at the two ends of the cut.

How the current is worked out, and what it was checked against

In short: Nobody publishes a current forecast for this channel, so TidePush calculates one from the tide gauges at each end and checks it against the only survey ever done inside the cut, in 2006. It is a model, not a measurement.

TidePush synthesises the water level at each end from published harmonic constants, takes the difference, and drives a physical model of the channel with it: the head accelerates the water, friction resists it, and the balance between them sets both the peak speed and the lag. The constants are calibrated against a 2006 USGS survey, which is the only occasion the current in this channel has been measured. The model reproduces both the peak speed and the delay that survey recorded.

That survey is Grossman, Stevens, Gelfenbaum and Curran, Nearshore circulation and water column properties in the Skagit River Delta, northern Puget Sound, Washington, USGS Scientific Investigations Report 2007-5120. It was a juvenile salmon habitat study rather than anything written for boaters, and the current record in it is incidental to that purpose. Two acoustic Doppler profilers sat in the cut from late March to mid-May 2006 — seven weeks, and the only weeks anyone has publicly measured this water. That window falls inside the channel’s strong season, which is reassuring for the peak it caught and unhelpful in another way: nothing has ever checked the model against a summer or a winter.

It is worth knowing why the model reads slower than local knowledge does. Pilotage for this cut is usually given as two to three knots, and the model tops out below that. The two are describing different water. The figure the model was fitted to is the speed averaged through the whole depth of the channel; the same survey clocked the surface running faster, and fast enough on a spring flood to land squarely in the two-to-three range. The surface is what a skipper watching the bank sees. A hull feels something between the two, and the model is deliberately the conservative of the pair.

An earlier version of the app took its channel velocity from a NOAA hydrodynamic forecast, read at the grid element off La Conner. That seemed the obvious source until it was checked against the survey: the finest published element of that model is around a hundred metres and the maintained channel is thirty metres wide, so it could not have been resolving the cut. Its agreement with the survey was poor and the disagreement was structured rather than random, which is the signature of a model resolving the wrong thing rather than one that is merely noisy. It now contributes only as an optional correction to the water level at the ends, where its resolution is adequate.

THIS IS A MODEL AND THE APP SAYS SO. Being the only available answer is a reason for care, not confidence: the calibration rests on one survey from 2006, and nothing has measured the channel since. Treat the direction and the timing as good, and the exact speed as an estimate.

Two further things about that calibration are worth stating plainly. The report contains no current harmonic analysis at all — the constants it publishes are water level at each end, and the stream is inferred from the head through a fitted transfer function rather than measured wave by wave. And the fitted constants are not physical: they imply an effective channel length of about 20 km for a cut 11 km long, and an effective depth of about 25 m for water around 4 m deep. Both are absurd read as geometry. What they are absorbing is the storage behaviour of the Skagit and Padilla flats, which the model has no other way to represent. It works, and it is not a description of the channel.

When it runs hard, and when it does not

In short: Because the channel runs on the twice-a-day tide, it is strongest near the equinoxes and weakest near the solstices. Midsummer, when most boats transit, is its quiet season.

The two halves of the tide peak at opposite times of year. The once-a-day waves exist because the moon and sun sit off the equator, so they are strongest when that tilt is greatest — at the solstices, in June and December. The twice-a-day waves run the other way: the solar one is largest when the sun is over the equator, which is the equinoxes. A channel driven by the once-a-day tide would be at its fiercest in midsummer. This one does the opposite.

Across a dozen model years the pattern is consistent. The strong season is March to May and September to November; the quiet months are December and January, and July and August. The channel is at its most docile during precisely the weeks most cruising boats use it, which is convenient and worth not relying on.

Two knots is rare here in any season. Only a few days in a year reach it at all, every one of them inside that strong season, and through the summer the channel does not get there. When it does happen it is always northbound — the model has never produced a southbound peak that large. Skagit Bay stands very slightly higher than Padilla through most of the 2006 record, and that small standing difference biases the whole system toward a net push north.

The seasonal shape is a stronger claim than any individual day, and all of it is model output rather than measurement. Nothing has measured this channel outside one spring in 2006.

Depth matters more than current here

In short: Whether you fit depends on the tide at the moment you are actually in the channel. If your boat is too deep, no departure time fixes it — only a bigger tide.

On most gated passages the question is when to go. Here it is often whether to go at all. The cut carries a twelve-foot project depth, and the water you actually have is that depth plus the height of tide at the moment you are in the channel — not at the moment you cast off, which for a boat coming from Everett or Seattle can be several hours earlier and several feet different.

Two caveats belong with that figure. The project depth is a design target rather than a recent survey; dredging runs on a cycle and the bed shoals unevenly between rounds. And the model assumes a flat bed both along the channel and across it, where a real dredged cut shoals at its edges — so a boat off the centreline has less water beneath it than the number allows for. This is the largest unquantified assumption in the depth check, and it is stated rather than buried.

The Rainbow Bridge, and the cables below it

In short: The bridge stays put and the water moves, so a high tide that helps your keel is the same tide that squeezes your mast. Cables have been reported below the span, and for a tall rig those are the real limit.

Air draft is the mirror image of depth, and the same tide drives both in opposite directions. The span offers about seventy-five feet at the mean higher high water mark, and swings roughly fourteen feet across the tidal range. TidePush computes the clearance for the minute you pass beneath it and shows it on the card.

It does not check that figure against your rig, because it does not know your air draft, and it says so rather than implying a verdict it has not reached. The more serious point is that cables have been reported over the channel about three feet below the span. That report is not a survey, and for a tall rig the cables rather than the bridge are the real limit.

Plan a passage — back to TidePush →