The Canadian Coast Guard cameras, from Seymour Narrows to the Canso Canal

At 09:06:34 this morning a self-unloader was crossing the frame of the Curve 1 camera, which stands out in Lake St. Pierre with nothing around it but water. Red hull, the boom lying down the length of the deck, wake running back towards the flat Quebec shore on the horizon. That camera belongs to the Canadian Coast Guard, and the file holding the ship was written at 13:12:45 UTC, six minutes after the stamp burnt into the picture.

The Coast Guard runs 45 of these. Its camera dashboard listed 45 records when I loaded it at 13:09 UTC on 2026-10-17, under the heading "Ice Monitoring System - Web Cameras", in four sectors: Atlantic, St. Lawrence, Great Lakes and Pacific. The page says plainly that the cameras are intended for the Coast Guard's operational use and that the images are offered to the public as a courtesy. 47 of them sit in our catalogue and 44 were reading live when I counted the source table at 13:09 UTC today.

Nothing in the set was placed for the view. Each one is pointed at a spot where a ship has to get something right: a dredged cut, a pair of range structures, a lock, a bridge pier, a tide race. That is exactly what makes them worth opening, because a camera chosen by someone who needs to see a specific thing usually shows you that thing better than a camera chosen for a postcard.

The ones to open first

• Curve 1, Lake St. Pierre sits on the dredged channel where it bends across the lake, and Pointe du Lac upriver is listed on the dashboard as a front range structure. The water is close under the lens at both, which is why a ship fills a quarter of the frame rather than sitting on the horizon.
• Lévis is the one to send to someone who has never looked at a camera page. It faces the Quebec City waterfront across the river: the Citadelle on the cliff, the Château Frontenac above the terrace, the hillside in full colour this week, and at 09:00:19 EDT a cruise ship lit up along the quay with a second white hull behind it and two ferries out in the stream.
• Point Edward looks up the St Clair River at both spans of the Blue Water Bridge, with the American and Canadian flags on poles in the foreground. Ships here are held to 12 statute miles per hour, 10.4 knots, from Fort Gratiot Light down to the St. Clair Flats, so a laker coming under the bridge takes its time and you get a long look at it.
• Livingstone Channel watches the cut west of Bois Blanc Island in the Detroit River, which is restricted to downbound vessels. Everything in the frame at 08:29 EDT was low and golden: the scrub on the dyke still in leaf, the water silver, a light structure out at the vanishing point.
• Sault Ste. Marie gives you the Canadian lock from above, with the International Bridge behind it and a work site in between. Tower cranes stood over the ground beside the chamber at 08:58:29 EDT and the trees along the left bank were already orange.
• Canso Canal is the prettiest lock in the set and the only one where the road swings out of the way. The chamber runs from the bottom of the frame to a swing bridge on the left, there is a small white light tower on the point, and the hills across the Strait of Canso were under flat grey cloud this morning.
• Seymour Narrows is the one with a timetable, and it has its own section below.

What the file times tell you

The frames come with two clocks, and the gap between them is the useful part. Each picture carries the site name and a local timestamp burnt into the top corners, and the file it arrives in carries a Last-Modified header. On Point Edward today the header moved from 13:07:22 to 13:12:24 UTC whilst I watched, a clean five minutes, and the stamp inside the frame ran about ten minutes behind that. Lévis and Sault Ste. Marie did the same thing in the same five minutes.

Not all of them run at that rate. Livingstone Channel was still serving its 12:57:27 file at 13:20, twenty three minutes old, and Sailors Encampment was on the same slower cycle. So if you are waiting for a ship, open two cameras: one five minute site to tell you the light is good, one slow site for the thing you actually want to see. The dashboard prints the last file update next to every camera, which is the fastest way to tell which cycle a site is on today.

A few of them are weather, not scenery. Fort Amherst at the Narrows of St John's was behind a lens full of raindrops this morning, grey sea, nothing of the harbour entrance at all, and Sailors Encampment on the St Marys River was in fog so thick the frame was a flat grey card with a compass bearing in the corner. Both tell you something true about the morning. One of the Great Lakes sites, Port Lambton, still shows a file from 2025-03-17 on the dashboard, so it is the one to skip.

The camera with a timetable

Seymour Narrows is the tide race between Vancouver Island and Quadra Island, and the current there is predicted to the minute by the Canadian Hydrographic Service. Today at station 08108, which is at 50.133333, -125.35 and so within a few hundred metres of the camera, the predictions run: slack at 03:54 Pacific, maximum flood of 7.6 knots at 07:26, slack again at 11:12, maximum ebb of 5.8 knots at 14:23, slack at 18:01.

The flood was the big one and it ran in the dark. Sunrise at those coordinates was 07:45:45 Pacific, nineteen minutes after the peak, which is why the 05:40 frame I pulled was monochrome with the lights of the far shore smeared across the water. The one to watch today is the afternoon ebb at 14:23, in full daylight, with the whole of the narrows pulling north.

That is the pattern worth knowing: a current station and a camera at the same coordinates give you the time to open the page. I am going to pair the two for every camera in these posts that has a current or tide station near it, because the predictions are published and the camera is free, and guessing at the moment wastes both.

The thing I cannot get from the Coast Guard's data is which way each camera is actually looking. Several frames carry a bearing in the corner, S W on Sailors Encampment and S on Livingstone Channel, but it is burnt into the image rather than published in a field, so it cannot be read. If you know of a public dataset that gives a camera's bearing as a number, tell me, and I will point every arrow on the map the right way.