(amplitude, all months)
(peak-to-trough)
air tide (M2)
(September → January)
The ocean has tides, and so does the air. Sunlight heats the upper atmosphere, mostly the ozone and water vapour, once a day, and the atmosphere rings at twice that rate. Near the equator the result is a pressure wave that peaks at about 10 in the morning and 10 at night and bottoms out at about 4, every day, whatever the weather is doing. We fitted it to every reading of the local station barometer from November 2023 to September 2026, one calendar month at a time.
Two things came out. First, this site had the size wrong: the forecast and the station page were removing a fixed 3.3 mb daily swing, when the measured twice-daily swing is 2.0 mb in summer and 2.5 mb in winter. The forecast now uses the monthly table below. Second, the Moon is in the barometer too. Its air tide is about 0.05 mb, far below what one reading can resolve, but three years of hourly averages pull it out with the same timing each year.
About 1.25 mb in the dry season, 1.01 mb in the rains.
Amplitude of the 12-hour wave in each month. It is largest from December to March and smallest from June to August, and each year repeats the pattern: July was 1.03, 1.00 and 0.95 mb in 2024, 2025 and 2026.
| Month | 12-h amplitude (mb) | Peak-to-trough (mb) | 12-h peak (local h) | 8-h amplitude (mb) | 24-h amplitude (mb) |
|---|---|---|---|---|---|
| January | 1.245 | 2.49 | 10:29 | 0.227 | 0.69 |
| February | 1.245 | 2.49 | 10:38 | 0.165 | 0.86 |
| March | 1.242 | 2.48 | 10:38 | 0.087 | 0.78 |
| April | 1.137 | 2.27 | 10:39 | 0.071 | 0.81 |
| May | 1.079 | 2.16 | 10:43 | 0.144 | 0.81 |
| June | 1.012 | 2.02 | 10:41 | 0.121 | 0.62 |
| July | 1.005 | 2.01 | 10:36 | 0.090 | 0.76 |
| August | 1.038 | 2.08 | 10:41 | 0.070 | 0.89 |
| September | 1.222 | 2.44 | 10:32 | 0.018 | 0.65 |
| October | 1.210 | 2.42 | 10:17 | 0.117 | 0.58 |
| November | 1.210 | 2.42 | 10:15 | 0.191 | 0.73 |
| December | 1.260 | 2.52 | 10:16 | 0.217 | 0.73 |
The twice-daily swing is 23% larger in the dry season than in the rains, and its peak comes about half an hour earlier in October–December.
A seasonal fit to the 34 monthly amplitudes gives a mean of 1.157 mb, an annual term of 0.125 mb peaking about 1 January and a semiannual term of 0.043 mb (R² = 0.78, F(4,29) = 25, p = 5×10−9). Single months run from 0.92 mb (August 2026) to 1.31 mb (January 2025). The first maximum falls at 10:15 local from October to December and at 10:36–10:43 from February to August; the second is twelve hours later. The 8-hour harmonic does something stranger: 0.018 mb in September, 0.227 mb in January, more than tenfold, and its peak hour moves from about 02:30 in the dry season to about 07:30 in the wet. The 24-hour wave is 0.58–0.89 mb with its maximum between about 04:30 and 07:00, but it is noisier and has no clean season, because daily heating and the sea breeze push it around.
Until 27 September 2026 this site removed a 3.3 mb daily swing from the barometer. The measured 12-hour swing is 2.0–2.5 mb.
The old figure was probably the full daily range, which includes the 24-hour wave, rather than the 12-hour tide alone. The difference mattered in one place: the “pressure falling” checks on the forecast and the station page compare a 3-hour change with the expected tidal change for that hour, and the old curve's twice-daily swing was 30–65% too large, most of all in summer. They now use a month-by-month table of the 12-hour and 8-hour waves fitted to this station (amplitude and peak hour for each). The alarm thresholds, 3 mb and 6 mb of anomaly in 3 hours, are unchanged.
The Moon pulls on the air here too: 0.048 mb, in step every year.
| Year | Hours | M2 amplitude (mb) | M2 phase | S2 amplitude (mb) |
|---|---|---|---|---|
| 2024 | 8,635 | 0.051 | 255° | 1.138 |
| 2025 | 8,737 | 0.044 | 245° | 1.157 |
| 2026 (to Sep) | 4,709 | 0.050 | 237° | 1.146 |
| All | 22,081 | 0.048 | 248° | 1.148 |
A 12.42-hour pressure wave of 0.048 mb, about 1/24 of the Sun's pressure tide, with the same timing within 18° in each of three years.
Fitting the Sun's lines and the Moon's together (24-hour, 12-hour and 8-hour solar waves plus K1, P1, K2, M2 and N2) to the hourly pressure leaves an M2 wave of 0.048 mb. To check that it is not noise, the same fit was run 200 times with M2 replaced by a fake frequency between 1.80 and 2.20 cycles a day: the fakes have a median of 0.007 mb, a 99th percentile of 0.019 mb and a maximum of 0.027 mb, so M2 beats all 200 (p < 0.005). Spectral leakage from the much larger solar wave at the M2 frequency is only 0.006 mb. Averaged by lunar hour, the pressure runs from −0.046 to +0.050 mb over the lunar half-day. The size matches the textbook lunar air tide in the tropics, a few hundredths of a millibar. N2, the next lunar line, is 0.014 mb, inside the noise.
One barometer reading resolves 0.034 mb (0.001 inHg). The Moon's tide is found only by averaging about 20,000 hours.
Nobody will see it on a single day, and it does nothing to the weather. It is here as a check on the instrument and the method: if a home weather station can recover a 0.05 mb lunar signal with a stable phase, its record is clean enough to trust for the 1–2.5 mb solar tide above. The same line is absent in the station's temperature (0.048 °F, not significant).
What the barometer does not show.
- No weekly cycle. Pressure, temperature, humidity, sunshine and rain show nothing at 7 days (band false-alarm probability 0.6–1.0).
- No spring–neap cycle. The ocean's two-week tide rhythm does not reach the air: nothing at 14.77 or 13.66 days survives the number of frequencies tested.
- No lunar-month cycle. Nothing at 27.3–29.5 days (band false-alarm probability 0.22 in pressure). The Moon shows up in the air only as its twice-daily tug.
- A weak band near 4 days (twin peaks at 3.93 and 4.07 days, about 0.17 mb each, band false-alarm probability 0.01), the period of tropical easterly waves. But it does not strengthen in the rainy season as those waves would, so it is unconfirmed.
- A weak band near 48 days (0.55 mb, band false-alarm probability 0.02), in the range of the Madden–Julian Oscillation. Three years is too short to call it a period; unconfirmed.
One consumer weather station. Its barometer reads in steps of 0.001 inHg; its absolute calibration does not matter here because only the daily shape is fitted. The array was dead from 15 May to 18 July 2026 and those weeks are excluded, so May–July 2026 has fewer hours.
Rainy months are noisier. The residual scatter is 0.64–0.73 mb in the wet season against 0.45–0.52 mb in the dry, but standard errors stay near 0.01 mb, so the seasonal cycle is not a noise effect.
The forecast's pressure is a model's, not the station's. The surf forecast applies the same monthly tide to the weather model's surface pressure; the station page applies it to the station itself. The table is a local fit and is not meant for other coasts.
The air here breathes twice a day: up at half past ten, down at four, a little harder in the dry season than in the rains. The Moon joins in at a twenty-fourth of the strength, on its own clock. A falling barometer in the afternoon is the tide, not a storm.Synthesis · 27 September 2026
Sources and method
Data: the local Ambient Weather station's relative pressure, 297,356 five-minute readings from 9 November 2023 to 19 September 2026, local time UTC−6; dead-array window 15 May – 18 July 2026 removed.
- Solar tide: a 48-hour centred running mean was subtracted (it passes the 12- and 24-hour waves untouched), then a constant plus 24-, 12-, 8- and 6-hour harmonics were fitted by least squares to each calendar month (34 months with at least 3,000 readings) and to each calendar month with all years pooled. Peak hour = local time of the first maximum.
- Seasonal fit: the 34 monthly 12-hour amplitudes against annual and semiannual harmonics of day of year.
- Lunar tide: hourly means, the same 48-hour detrending, a joint least-squares fit of S1, K1, P1, S2, M2, N2, S3 and K2, per year and for the whole record; null distribution from 200 random frequencies in 1.80–2.20 cycles/day kept at least three frequency-resolution steps away from any fitted line.
- Periodograms: FFT of hourly series with gaps zero-filled after removing annual harmonics; background from a running median in log frequency; band false-alarm probabilities corrected for the number of independent frequencies in each band.
In the code: functions/api/_pressure_tide.js holds the monthly table (12-hour and 8-hour amplitude and peak hour) used by /api/forecast and /api/station.
Analysis run: 26–27 September 2026.
Cite as: “The air tide: the barometer's twice-daily tide at La Saladita, and the Moon's”, La Saladita Field Guide, lasaladita.com/findings/the-air-tide/, 2026-09-27.