Astronomically reliable.
Meteorologically blind.

01 · The method

Harmonic analysis, in plain terms

Tide prediction does not work by calculating the moon's gravity at your beach. The real ocean is far too complicated for that. It works by measurement and pattern-fitting.

1. Measure. Install a tide gauge at the port and record the water level continuously, for years.
2. Decompose. The astronomical forcing is made of cycles with known, fixed periods: the twice-daily lunar cycle, the twice-daily solar cycle, the once-daily cycles, the fortnightly and annual ones, and so on. Each is called a constituent. Analysis works out how strongly each constituent shows up in that particular record, and how much it lags the astronomy.
3. Store. The result is a list of amplitudes and phases, one pair per constituent. That list is the port's tidal fingerprint. The Australian Hydrographic Office holds the main 22 constituents for standard and secondary ports.
4. Project. Because the periods are astronomical, they are known indefinitely into the future. Add the constituents back together for any future date and you have a prediction.

The elegance is that the physics of the local basin, the shape of the harbour, the friction and the resonance are all baked into the fitted amplitudes and phases. You never have to model them explicitly. You just have to measure long enough.

02 · The catch in step 1

Why the length of the record matters

Two constituents with similar periods can only be told apart if you observe for long enough that they drift out of step with each other. That sets a floor on how long a record has to be.

• About a month separates the main daily and twice-daily constituents crudely.
• About a year is generally needed to resolve the main set properly, including the annual and semi-annual terms.
• About 18.6 years is the full nodal cycle, a slow wobble in the moon's orbital plane that modulates the constituents. Long analyses either span it or correct for it.

This is also why datums like Lowest and Highest Astronomical Tide are quoted against a stated determination period. Datums and epochs.

The practical consequence: a genuine prediction for a place requires a long gauge record at that place. Places without one get derived predictions, which is a different and weaker product.

03 · The limitation

What a tide prediction cannot know

Harmonic analysis extracts the part of the water level that repeats on astronomical cycles. Everything that does not repeat on an astronomical cycle is, by construction, left out.

That means the prediction assumes average meteorological conditions. It contains no information about the weather on the day, because there is no way for it to. The weather in March 2028 is not encoded in a lunar cycle.

The difference between the predicted level and the level that actually occurs is called the residual, or the non-tidal component, and it comes from three main sources.

Barometric pressure

The atmosphere presses down on the sea. Low pressure lets the surface rise; high pressure pushes it down. The rule of thumb is roughly one centimetre of sea level for each hectopascal of pressure difference from the average, so the swing between a deep low and a strong high can be a few tens of centimetres.

Wind setup

Sustained wind blowing onshore piles water against the coast; offshore wind drags it away. The effect grows with the length of water the wind blows over and is much larger in shallow water, so gulfs, bays and broad shelves are the worst affected.

Storm surge

The two combined under an intense system. This is the dangerous case, it is what causes coastal inundation in a tropical cyclone, and it is forecast separately by official storm tide warning services rather than being derivable from a tide table.

04 · Scale

How big is the difference, and when does it matter?

On an ordinary day at an ordinary Australian port the residual is usually small compared with the tidal range, and the prediction is close. That is why tide tables work.

It stops being small when:

• The tidal range is small and the weather is not. Where the range is under a metre, a few tens of centimetres of surge is a large fraction of the whole signal. In a large-range port the same surge is proportionally trivial.
• The water is shallow and broad. Wind setup is far more effective over a shallow shelf or inside a gulf than off a steep coast.
• There is a system offshore. A tropical low, an east coast low or a deep frontal system can dominate the astronomical signal entirely.
• You are working to a small margin. If your plan has 200 millimetres of clearance in it, the residual is not a rounding error, it is the whole question.

The rule worth carrying: a tide prediction tells you what the astronomy will do. Before relying on it, look at the weather, and if there is a system about, treat the prediction as the baseline rather than the answer.

05 · Derived predictions

Secondary ports carry extra error

Most Australian locations do not have their own long gauge record. Their predictions are derived from a nearby standard port by applying published time and height differences.

Those differences are typically constants, or simple factors, derived from a comparatively short period of observation. They cannot capture everything that varies: the corrections may work well at spring tides and less well at neaps, or well in one season and less well in another.

So a secondary port prediction inherits all the uncertainty of the standard port's prediction and adds its own. The official tables are open about which ports are which, and about what corrections have been applied.

The problem arises with third-party sites that present a derived prediction as if it were a measured one, without naming the reference port. How to tell the difference, and why it matters.

06 · Two different products

A modelled sea level is not a tide prediction

This is the distinction that this site is in an unusual position to explain, because we publish one of these and not the other.

Atmosphr's ocean conditions panel shows a modelled sea level from a global numerical ocean model. The panel carries this line, and it is there deliberately:

Modelled sea level is relative to global mean sea level, not a local tide datum.

Here is what separates the two products.

Each is strong exactly where the other is weak. The harmonic prediction knows the place but not the weather. The model knows the weather but not the place.

The failure mode is reading one as the other. A modelled sea level of 0.6 metres does not mean the tide is 0.6 metres, because the zero is different and the number includes things a tide height does not. Providers of this data say so themselves: the model vendor's own documentation states that accuracy is limited in coastal areas and that the data is not suitable for coastal navigation. We repeat that on the panel rather than burying it.

07 · Practical

Where to get observed sea level, rather than predicted

If what you want is what the water is actually doing right now, you want gauge observations, not predictions and not a model.

• Bureau of Meteorology tide gauge observations, which for many ports publish observed against predicted, so you can see the residual directly. This comparison is the single most instructive thing on this page if you have never looked at it.
• State authority gauge networks, which often carry near-real-time water level for ports and estuaries.
• The Australian Baseline Sea Level Monitoring Project, for long-term high-quality records.
• The Australian Ocean Data Network, for archived observational data.

Watching observed and predicted side by side during a windy week is the fastest way to develop a realistic sense of how much to trust a tide table on any given day.

Related guides

• Tide datums explained
• King tides in Australia
• Why most places get two high tides a day, and where they do not
• Tides and tidal currents are not the same thing