Waterfront running routes for travellers

Water is the one landmark a visitor cannot misread. You may not know which way north is in Barcelona or Vancouver, but you always know the sea is on one side and the city is on the other. That single fact removes most of the navigation load from a run in a place you landed in yesterday.

Why a waterfront route is the safe choice in a strange city

A waterfront path removes two decisions. They are the two that go wrong first: which way you are facing, and where you turn. The water answers the first for the whole run, and a promenade that follows a shoreline answers the second by having very few places where a wrong choice is even available.

Route finding fails when you are tired on ground you do not know. An out and back along water removes that entirely, because the way home is the path you just ran in reverse.

You cannot get lost on the way back from a place you have already been.

• Orientation is free. The water is on the same side the whole way out and on the other side the whole way back.
• Junction count is usually low.
• You can stop the run at any point and still be somewhere legible, because the shoreline leads back, and there are fewer moments on it where you have to decide anything.
• Surfaces on a promenade tend to be continuous, so you are not switching between a path and a road every few hundred metres.

Exposure is the cost. Wind off open water is stronger than the same wind two streets inland, and it will be behind you one way and in your face the other. Plan to run into it first, while you are fresh.

How these routes were measured

Every route below is measured the same way. That is the only thing that makes a two kilometre harbour front and an eight kilometre river path comparable. Distance is the length of the published route line, not the length of the trail or park it sits on.

Surface is recorded as one of two values, paved or mixed, and nothing finer. Corner count is produced by one detector applied to every route, so the number means the same thing in Sydney as it does in Seoul.

The catch?

A gentle curve along a shoreline is not a corner, and neither is a short sidestep around a building that puts you straight back on your original heading. A bend only counts as a turn when the direction of travel changes by at least 55 degrees measured 35 m before and after the bend, and the same 55 degrees still has to hold when the bend is re-measured over a wider 56 m window.

Short harbour fronts

These are routes you repeat. You do not finish them. Both are lines and not loops, so treat the published distance as half of your smallest sensible run.

Run the Hong Kong harbourfront out and back and you have just over four kilometres with two corners on the way out and the same two coming home. Twice out and back is eight kilometres on a path you learned in the first ten minutes.

That is a better first run abroad than an ambitious loop you have to think about.

Middle distance promenades

This is the useful band. Every route here is a line and not a loop, and every one of them is a normal easy run in one direction, or a solid session out and back. The corner counts vary far more than the distances do, and the corner count is what decides how much attention the route will ask of you.

Zero corners across the whole 3.45 km, and the Paris quays are the extreme case. A route with no turns is a fact about the shape of the path, not a gap in the data, and the app says so instead of staying quiet about it. On a route like that you read "This course has no turn guidance. Follow the route on the map." That is deliberately different from "Turn guidance could not be loaded for this course, so this run has no turn cues.", which means something went wrong.

The Seoul stream path is close behind with a single corner.

At the other end, the Thames Path stretch and the two San Francisco routes each carry eight or nine corners in under four and a half kilometres. Divide one by the other and that is a corner roughly every 420 m on the Thames Path, every 440 m at Crissy Field and every 540 m along the Embarcadero, against nothing at all to remember on the Seine. How those corners get called out is covered in city running routes with turn cues, and the London and San Francisco sets in London running routes and San Francisco running routes.

Long riverside stretches

Past seven kilometres, shape matters as much as length. Two of these return you to where you started and two do not.

The two loops are closed circuits, so seven kilometres is the whole job and you finish where your bag is. Sydney's circuit matches the length its council publishes: Canada Bay Council describes the Bay Run as a 7km circuit around Iron Cove that hugs the shoreline, fetched 2 August 2026, and the measured route line comes out at 7.09 km.

Lines are different animals. The Hudson line is 8.12 km one way, so running it end to end and back is over sixteen kilometres. Unless you have a way back, treat the published distance as the outward leg and pick your own turning point instead of the end of the line. The New York routes are compared side by side in New York running routes.

Out and back, and where a line ends

Eleven of the thirteen waterfront routes above are lines, not loops, and a line has an end you have to deal with. Decide before you start whether the end of the route is your finish or your halfway point, because deciding it at kilometre eight with tired legs and no transit card is how a good run becomes a bad afternoon.

1. Turn on time, not on distanceGive yourself a clock limit for the outward leg and turn when it runs out, whatever the distance says. Wind and crowds make the two legs take different amounts of time even on the same path.
2. Check what is at the far endA promenade can end at a station, a ferry pier or nothing at all. If it ends at nothing, the return leg is not optional and you should have planned for it before you set off.
3. Expect no turn help on the way backIf you run a curated route in reverse, turn cues switch off completely instead of being mirrored, and the app says so: "You are running this course in reverse, so turn cues are off." A bear left is not a bear right when read backwards, and a wrong instruction at a junction is worse than silence.
4. Know that nothing is going to reroute youWhen you leave the route the app tells you how far off you are and keeps recording. Finding your way back is up to you, which on a waterfront usually means walking to the water and turning the way you came.

What open water does to a signal

Two causes, and a shoreline has both. GPS.gov lists satellite signal blockage due to buildings, bridges and trees, and signals reflected off buildings or walls, which it calls multipath among the causes of reduced accuracy, and states that GPS enabled smartphones are typically accurate to within a 4.9 m radius under open sky, and that the performance worsens near buildings, bridges and trees, fetched 2 August 2026.

So a shoreline path gives you an open sky on one side and, in many cities, a wall of buildings or the underside of a bridge on the other. A track can therefore look ragged in a covered section and clean in an open one within the same run.

Neither condition is unusual and neither is a defect in your phone.

What an app does when the signal actually goes is the part that matters. When GPS drops mid-route, Runflake stops judging the course rather than declaring you off it, and says so: "No usable signal, so the course is not being checked. Your run keeps recording." A gap of up to 5 seconds between usable fixes does not reset the off-route evidence, and turn cues are not pushed forward while course judging is suspended, so a signal gap cannot make the app skip past a corner.

Nor is the off-route corridor a fixed distance. It is 2.5 times the phone's own reported accuracy, with a floor of 45 m and a ceiling of 80 m, so it widens exactly when the position is least trustworthy. That floor was chosen in simulation, and the project's own configuration notes call it provisional until real-device measurements exist.

Where the route data comes from

The route shapes come from OpenStreetMap and are credited to OpenStreetMap contributors under the ODbL licence. OpenStreetMap's own copyright and licence page states that the data is licensed under the Open Data Commons Open Database License and that anyone using it must credit OpenStreetMap and make clear that the data is available under that licence, fetched 2 August 2026.

Turn instructions are computed from each route's own shape by Runflake's own turn detector before the route is published, not written by hand and not taken from a navigation provider. That has an honest consequence: where the mapped line has a defect, the detector can produce a corner that a runner standing there would not call a corner. It also means the counts in the tables above are directly comparable, because one rule produced all of them. Turn distances are rounded down before being spoken, to the nearest 50 m (10 m below 100 m) in metric and the nearest 50 ft in imperial, so you never hear an odd number like "In 117 meters". Rounding down rather than up matters at a corner, because a number that arrives early makes people turn early.

A published route is a stretch of a trail, not the trail. The Chicago Park District describes the Lakefront Trail as running from Ardmore Avenue in the north to 71st Street in the south, separated in 2018 into an 18 mile bike trail and an 18.5 mile pedestrian trail, fetched 2 August 2026, while the route above is a 5.10 km slice of it. Singapore's National Parks Board describes East Coast Park as 180 hectares stretching over 13km, fetched 2 August 2026, against a 4.07 km published route.

Neither is a contradiction. They are answers to different questions, and before a run in a city you do not know, the one you want is the second one.