The tide gauge

A water level station stands on a pier, wharf or piling over water that stays deep enough at low tide. Its sensors are over or in the water, its electronics are in a shelter or a nearby building, and a set of benchmarks is set in the ground on shore. San Francisco's station was installed in 1854. Some 27 of the stations belong to the core network of the Global Sea Level Observing System.

The figure shows the radar and the acoustic gauge side by side, as at Port Townsend during CO-OPS's 2008 tests of the radar, with a bubbler as the backup: CO-OPS's specification lists a primary sensor and a backup for each kind of station.

The primary sensor CO-OPS installs today is a microwave radar (its specification names the YSI WaterLog H-3611 and Nile radars). It hangs on an arm off the pier, aimed straight down, and times a 26 GHz pulse to the water and back. The beam spreads in a 10° cone: from 8.5 m up, as in a test at Duck, North Carolina, it covers a patch about 1.5 m across. CO-OPS began switching the network's stations from acoustic sensors to radar in 2012.

No part of it touches the water, and radio waves, unlike sound, travel through air at a speed that barely changes with the air's temperature. The radar sees the waves as well as the tide, and anything that passes under it (a boat, a bird, floating debris) can return a false echo. The top of its flange is the leveling point, the mark that surveyors tie to the benchmarks.

In 2017, most stations' primary sensor was still an Aquatrak acoustic gauge, and the current specification still lists it. Its head sends a pulse of sound down a half-inch PVC sounding tube to the water inside and times the echo. A calibration point in the tube, a fixed distance below the head, returns an echo of its own, which corrects the reading for the speed of sound in the air in the tube; two temperature sensors in the tube also help. The leveling point is the top shoulder of the collar on the calibration tube.

The sounding tube stands in a 6-inch PVC protective well strapped to a piling, open to the sea through a 2-inch double-cone orifice at its base. The well protects the tube but does little to damp the waves. An acoustic gauge's value is the mean of 181 one-second readings, three minutes centered on each tenth of an hour.

The backup is usually a pressure gauge: a bubbler. A pump in the shelter pushes a slow, steady flow of gas down a tube to an orifice fixed to a piling well below the lowest tide, where it bubbles out. The pressure needed to push the gas out is the weight of the water above, which the station converts into a height using the water's density. CO-OPS's orifice is a red brass pipe 2 inches across and 8 inches long, over two parallel plates 6¼ inches across that smooth the flow of water past it. A Sutron constant-flow bubbler, one of the approved backups, averages 72 readings taken five seconds apart into each six-minute value.

Other stations use a pressure sensor in the water, or a second radar. On the Great Lakes, where there's almost no tide, the primary sensor may instead be a float in a stilling well, read by a shaft angle encoder, to ±0.006 m.

The radar measures the water level once a second. Each six-minute value is the mean of the 360 readings centered on a tenth of an hour (from three minutes before to three minutes after 04:18, for example). Readings more than three standard deviations from the mean, such as a false echo off a gull, are discarded and the mean is recalculated; the standard deviation and the number of readings discarded are reported with it. Averaging over six minutes removes the waves, whose crests and troughs come and go every few seconds.

CO-OPS states each value as accurate to ±0.02 m against the station's datum, and a monthly mean to ±0.005 m, with a resolution of 1 mm. Stations that warn of tsunamis also send one-minute means.

A data collection platform (CO-OPS's documents name Sutron's) reads the sensors, calculates the means and sends the values to shore. Every station must have a GOES transmitter, which sends to NOAA's weather satellites in geostationary orbit over the equator, so its antenna points at one of them; some stations also send by internet modem. The station runs on a 12-volt battery charged by a solar panel.

CO-OPS's sensors are set up to transmit at six-minute intervals; its 2022 report to the Global Sea Level Observing System describes the data arriving in an hourly transmission. The data are published once they pass automatic checks; the verified record comes later.

A water level means nothing unless the sensor stays in place, so every long-term station has a network of ten benchmarks, marks set in the ground nearby, at least three of them of the most stable kinds. Surveyors measure the height difference between the sensor's leveling point and the benchmarks (a survey called leveling) to second-order, class I precision when a station is installed, again two to six months later, then about once a year or as scheduled, and after storms and earthquakes. The primary benchmark, within 500 m of the sensor, is included in every survey. Every five years, a benchmark (preferably the primary) is occupied with GNSS for four hours, to tie the station to the national reference frame.

The sensor measures from a station datum, a local, arbitrary zero that never changes. From 19 years of readings, the National Tidal Datum Epoch (1983–2001 for now; CO-OPS proposes to release datums based on 2002–2020 in 2029), CO-OPS calculates the tidal datums: mean lower low water (MLLW), the average of each day's lower low water, which is the zero of charts and of Bluegraph's tide readings; mean sea level (MSL); and mean higher high water (MHHW). At Port Townsend, MSL is 1.52 m above MLLW and MHHW is 2.60 m. On the Great Lakes, levels are given above low water datum, on the International Great Lakes Datum of 1985.

From a long record at a station (at least 30 days, and a year to resolve all of them), CO-OPS calculates the amplitude and timing of 37 harmonic constituents, the regular rhythms of the tide that follow the moon and sun. At Port Townsend, the largest are K1 (0.77 m, about a day), M2 (0.68 m, about half a day) and O1 (0.45 m). Summing them gives the predicted tide at any time. It is the astronomical tide alone: it knows nothing of wind, pressure, river flow or rain, so it is a calculation, not a forecast.

The difference between the measured water level and the predicted tide is the residual: what wind, pressure, rivers and the sea added to the tide of the moon and sun. It is how storm surges show up.

Some stations also carry weather sensors; Port Townsend has wind, air and water temperature, and pressure. Wind is a two-minute mean taken before each tenth of an hour, and the gust is the highest five-second mean in the six minutes before. Air and water temperature and pressure are means of 20 readings over two minutes, accurate to ±0.2 °C and ±0.5 hPa.

At each tide gauge, Bluegraph shows the measured six-minute water level as TIDE: above MLLW on the coasts, and above low water datum on the Great Lakes and the St. Lawrence. At the 193 tidal gauges, it also shows CO-OPS's predicted tide and the residual it derives from the two. After 60 days, the charts keep one reading an hour. The Sea level pages are built on the residuals. Where a station has weather sensors, their readings reach Bluegraph through NOAA's National Data Buoy Center. Port Townsend is PTWW1.