between 06:00 and 09:00
in that window, Oct–Mar
(share of all lows in the year)
(K1+O1) / (M2+S2)
Surfers carry a tide rule from other coasts: get up early, catch the low, ride the push. At La Saladita that rule does not fit the clock. We took the site's own tide model — the harmonic fit to the University of Hawaii Sea Level Center's Zihuatanejo gauge that has driven the forecast since 24 September 2026 — and asked it one plain question for every day from 1 October 2026 to 30 September 2027: at what local hour do the highs and lows fall, and what does the engine call the tide at 5, 6, 7 … 13 o'clock?
The answer is a clock, not a calendar. Lows fall at about 1–4 in the morning and 1–4 in the afternoon in every month of the year. Highs fall at about 7–10 in the morning and 7–10 at night. The lunar part of the tide moves the heights around a lot — which of the two lows is the lower one, whether there is one low or two, how big the day's range is — but it barely moves the hours, because at this gauge the twice-daily tide is mostly solar and the solar tide keeps clock time.
The practical consequences are simple. A dawn session between October and March is on a high tide on 71–100% of days and on a rising tide on almost all of the rest. A low tide between 6 and 9 am happens on six mornings in the year, all of them from May to August. And the engine's rule that the inside section “connects on a dropping mid tide” can only be satisfied from mid-morning on — 10 to 12 o'clock in the dry season — which is exactly when the wind turns (see the wind clock). The same year run one year earlier (October 2025 – September 2026) gives the same answers within a few points; the differences are listed in the tables.
Two low bands, two high bands, in every month.
Every high and low the model predicts for the year, binned by local hour. The tide's lows sit in two three-hour bands and its highs in two more; the six hours from 05:00 to 11:00 hold 2.5% of the year's 681 lows.
| Local hours | Lows (% of all lows) | Highs (% of all highs) |
|---|---|---|
| 00–03h | 24.1% | 1.2% |
| 03–06h | 23.1% | 1.5% |
| 06–09h | 0.9% | 32.4% |
| 09–12h | 1.6% | 14.2% |
| 12–15h | 28.5% | 1.5% |
| 15–18h | 18.8% | 2.1% |
| 18–21h | 0.6% | 34.0% |
| 21–24h | 2.5% | 13.2% |
tidePhaseAt: the top 20% of the surrounding 14-hour rise-and-fall) is high at that hour. The rust box is the 06–09h dawn window. The dark band sits on dawn from September to March, slides to mid-morning in April and May, and thins out in June and July — the months when the day's single diurnal high drifts across the morning. Hover a cell for the full four-way split.A low tide between 06:00 and 09:00 falls on 6 of 365 mornings (1.6%). Between September and April it does not happen at all.
The six are 9 May, 7–8 June, 6–7 July and 5 August 2027 — the days around the summer lunar standstill when the day collapses to a single low and that low drifts through the morning. The check year (October 2025 – September 2026) gives five such mornings (20 April, 19 May, 17–18 June, 17 July 2026), 1.4%. Over the whole year the engine labels the tide dropping-mid at 07–09h on 3% of hour-slots and low on 3%; it labels it high on 58% of 07:00 slots, 77% of 08:00 slots and 77% of 09:00 slots, and rising-mid on most of the rest.
October to March: high or rising. June and July: the only months with a real spread.
“Low between 06–09h” and “high between 06–09h” count days on which an extreme of that kind falls inside the window. The phase columns pool the engine's hourly label over the 07:00, 08:00 and 09:00 slots (and 10:00–12:00 for the late-morning column). The last two columns repeat the first two for the check year.
| Month | Low between 06–09h | High between 06–09h | High at 07–09h | Rising at 07–09h | Dropping at 07–09h | Dropping at 10–12h | Low 06–09h (check year) | High 06–09h (check year) |
|---|---|---|---|---|---|---|---|---|
| Jan | 0% | 87% | 94% | 3% | 3% | 73% | 0% | 94% |
| Feb | 0% | 86% | 82% | 14% | 4% | 63% | 0% | 79% |
| Mar | 0% | 71% | 74% | 17% | 9% | 50% | 0% | 65% |
| Apr | 0% | 60% | 53% | 32% | 11% | 33% | 3% | 50% |
| May | 3% | 32% | 43% | 40% | 9% | 26% | 3% | 36% |
| Jun | 7% | 10% | 36% | 50% | 3% | 16% | 7% | 23% |
| Jul | 7% | 10% | 39% | 52% | 2% | 16% | 3% | 13% |
| Aug | 3% | 32% | 61% | 34% | 1% | 28% | 0% | 36% |
| Sep | 0% | 60% | 83% | 16% | 1% | 61% | 0% | 63% |
| Oct | 0% | 100% | 96% | 1% | 3% | 79% | 0% | 100% |
| Nov | 0% | 90% | 93% | 0% | 7% | 89% | 0% | 93% |
| Dec | 0% | 90% | 95% | 0% | 5% | 85% | 0% | 90% |
| Year | 1.6% | 61% | 71% | 22% | 5% | 51% | 1.4% | 62% |
The dropping mid tide is a 10-to-12 o'clock tide from September to March, and hardly a morning tide at all in June and July.
In November the engine calls the 10–12h slots dropping-mid 89% of the time; in December 85%, October 79%, January 73%. Over the same months it calls the 07–09h slots dropping-mid on 3–7% of slots. In June and July the high itself has slid to late morning (the 10–12h slots are high 50–61% of the time), so the dropping tide moves to early afternoon. The check year moves individual cells by up to 13 points (March high-at-dawn 65% against 71%, June 23% against 10%) but never moves a month across the line between “dawn is high” and “dawn is not”.
A mixed tide whose twice-daily half keeps solar time.
A tide is a sum of cosines with fixed periods. The four that matter here, as averages over the 18.6-year lunar cycle: K1 (23.93 h, luni-solar declinational) 0.129 m, O1 (25.82 h, lunar) 0.099 m, S2 (12.00 h, solar) 0.117 m, M2 (12.42 h, lunar) 0.062 m. The conventional form factor F = (K1+O1)/(M2+S2) is 1.27 on those averages and 1.43 in 2026–27, when the lunar cycle swells the once-a-day constituents: on the Courtier scale both are “mixed, mainly semidiurnal” (the boundary to “mainly diurnal” is 1.5), and the two bands are in fact about equal — 0.29 m of once-a-day amplitude (K1+O1+P1+Q1) against 0.25 m of twice-a-day (M2+S2+N2+K2). In 2026–27, 16 percent of days have only one low.
What sets the clock is the unusual make-up of the twice-daily band. At most ocean coasts M2 is two to three times S2. Here S2 is nearly twice M2: 60% of the semidiurnal amplitude is solar. A solar cosine repeats every 12.00 hours exactly, so its lows fall at the same local hours on every day of the year — from the fitted phase, 02:34 and 14:34. The lunar M2 drifts 50 minutes a day around the clock, but at 0.062 m it can only pull the solar lows an hour or two either way, not relocate them. That is the 01–04h and 13–16h band in the histogram above.
The once-a-day band is what varies. K1 loses four minutes a day against the clock, so its high circles the 24 hours once a year; O1 loses 1.8 hours a day, so it circles once a fortnight. Together they decide which of the two solar lows is deeper and how big the day is (large near the moon's maximum declination, small near zero), and in May–July their combined high lands in the late morning, which is when the day sometimes collapses to a single low and one of those lows can land at dawn. In the dry season the combined diurnal high sits on the morning solar high, and dawn is a high tide almost without exception.
Compared with Acapulco, the unusual part of the Zihuatanejo tide is a weak Moon tide (M2), not a strong Sun tide.
The Zihuatanejo gauge (2011–2013) and the Acapulco gauge (2007–2011) were each fitted with the 18.6-year lunar corrections applied, so the two records can be compared across their different years. The once-daily tides agree between the gauges, and their clocks match (r = 0.992 over the 53 days the records overlap). In the twice-daily band, Zihuatanejo's M2 is only 0.35 of Acapulco's (95% range 0.34–0.37), while its neighbors are much closer to Acapulco's: N2 0.67, and the small NU2 and L2 about 0.9. The Sun's tide S2 and its close partner K2 are both larger at Zihuatanejo (1.43 and 1.29). Local ocean tide structure fits the M2 deficit best; a harbor-gauge effect is also possible; in every 30-day window of the gauge record, S2 exceeds M2.
A small tide: about half a meter a day, and near the top of its 18.6-year cycle.
The daily range averages 0.51 m; one day in ten reaches 0.71 m and the largest day of the year 0.85 m. A range of 0.6 m or more, the level of the forecast's “large tidal range” note (advice only, it does not change the score), comes on 28% of days in the primary year and 27% in the check year, most of them from January to March and July to September. These heights are from the tide model refitted with the 18.6-year lunar cycle (below); the last two columns show the forecast's original model for comparison.
| Month | Mean range (m) | Median (m) | p90 (m) | Days ≥ 0.6 m | One-low days | Mean range (check year) | Original model: mean range | Original model: days ≥ 0.6 m |
|---|---|---|---|---|---|---|---|---|
| Jan | 0.53 | 0.55 | 0.74 | 39% | 13% | 0.55 | 0.49 | 32% |
| Feb | 0.54 | 0.54 | 0.79 | 39% | 0% | 0.54 | 0.48 | 32% |
| Mar | 0.55 | 0.56 | 0.74 | 42% | 0% | 0.53 | 0.48 | 26% |
| Apr | 0.51 | 0.52 | 0.64 | 17% | 13% | 0.52 | 0.45 | 10% |
| May | 0.48 | 0.49 | 0.62 | 16% | 29% | 0.51 | 0.45 | 0% |
| Jun | 0.49 | 0.51 | 0.63 | 17% | 40% | 0.50 | 0.47 | 3% |
| Jul | 0.52 | 0.53 | 0.74 | 42% | 16% | 0.50 | 0.48 | 32% |
| Aug | 0.51 | 0.52 | 0.73 | 42% | 3% | 0.48 | 0.46 | 29% |
| Sep | 0.49 | 0.48 | 0.70 | 33% | 3% | 0.51 | 0.43 | 23% |
| Oct | 0.53 | 0.53 | 0.69 | 19% | 16% | 0.53 | 0.47 | 13% |
| Nov | 0.51 | 0.50 | 0.63 | 20% | 27% | 0.50 | 0.46 | 0% |
| Dec | 0.49 | 0.48 | 0.67 | 16% | 29% | 0.52 | 0.47 | 16% |
| Year | 0.51 | 0.51 | 0.71 | 28% | 16% | 0.52 | 0.47 | 18% |
The original model was fitted to 2011–13, near the low point of the 18.6-year cycle for the once-a-day tide, and carried no correction for it, so in 2026–27 its daily ranges run about 9% small. The refitted model includes the cycle and matches the Zihuatanejo gauge in years it was not fitted to.
The moon's orbit precesses over 18.61 years and modulates the once-a-day constituents strongly. The 2011–13 record behind the original model sat at a node longitude of about 246°, where Schureman's factors give fK1 = 0.96 and fO1 = 0.93; for the middle of the primary year N ≈ 318°, fK1 = 1.09 and fO1 = 1.15. Rescaling the original fit by those factors alone lifts the mean range from 0.47 to 0.50 m, about two-thirds of the way. The refitted model keeps the same 33 constituents, applies the nodal factors, and is fitted to 67,111 hours of the IOC Zihuatanejo gauge, 2013–2026. Tested on each of eight years left out of the fit in turn (2016–18 and 2022–26), its tidal amplitude is 1.00 of the gauge's (95% range 0.997–1.005) against 0.95 for the original model, its hourly error is 2.3 cm against 4.3 cm, and its highs and lows fall at the gauge's times (median offset 0 minutes; original model 6 minutes early). In 2026–27 the original model's daily range runs about 9% small, 0.47 m against 0.51 m. An independent open global tide model, EOT20, comes out at 1.02 of the gauge's amplitude in the same test. The clock does not move: the two models give the same phase label at 91% of hours, and the refitted one also has no dawn low from September to April.
The engine's best inside-section tide is a late-morning tide.
inside_section: optimal swell + dropping-mid → “connects”. From October to March that combination is essentially only available between 10 and 12 o'clock.
The rule is right about the water; it is the clock that is unforgiving. The hourly label is dropping-mid at 07–09h on 3% of slots year-round, and at 10–12h on 50–89% of slots from September to March. Those are also the hours in which the morning land breeze gives way to the onshore west or southwest wind (the pattern the score grades against, and the subject of the wind clock). So on this coast the “connects” inside and the clean wind rarely overlap in the dry season; the dawn call is a high or rising tide on the outside point, and the inside is a mid-morning gamble against the wind. In June and July the high slides to late morning and the dropping tide to early afternoon, after the wind has already turned.
Do not plan around a dawn low here. Plan around the wind, and take the tide as it comes: high at first light from October to March, low around lunchtime, high again at dinner.
The model describes the water level's clock, not how the wave breaks.
This is a model of the Zihuatanejo gauge, 25 km down the coast, not a gauge at La Saladita. Fit RMS 6.4 cm on a tide with 18.6 cm standard deviation; heights within about 20%. Extreme times agree within about 15 minutes with a published table for Zihuatanejo checked on 28 September 2026. A second public table disagreed by up to 1.7 hours and lists several spurious extremes a day, so it is not used as a check. Any systematic offset between the gauge and the point is unmeasured; it would shift the clock by minutes, not hours.
Two models. The clock tables (dawn lows and highs, phase by hour, where the lows fall) use the forecast's original model, because those are the labels the forecast shows; its highs and lows fall within about 6 minutes of the gauge. Heights and ranges use the refitted model with nodal corrections. The gauge is an IOC real-time station whose data are not quality-controlled by the provider.
“Dawn” is a fixed 06:00–09:00 window (start inclusive, end exclusive). Sunrise here runs from about 06:10 in June to 07:15 in January, so the window covers first light to mid-morning in every month rather than tracking the sun. Widening it to 05–10h adds five May–August lows and one 05-o'clock low in each of December and January; 06–09h itself stays empty from September to April.
The phase label is the engine's, with its thresholds. tidePhaseAt calls a level “high” when it is in the top 20% of the range seen in the surrounding ±7 hours and “low” in the bottom 20%; the mid labels follow the sign of the next 30 minutes. On the small-range days of a mixed tide these labels can flip on a few centimeters; the extreme-time counts (the low/high-between-06–09 columns) do not depend on the thresholds.
One year, and one check year. The clock repeats annually because it is solar; the lunar modulation (which low is lower, the range) shifts by about 11 days a year and the nodal cycle over 18.6 years. Percentages for a 28–31 day month move by a few points from year to year, as the check-year columns show.
Here the moon decides how much the tide moves and the sun decides when. The lows keep to one to four, morning and afternoon, twelve months a year. Dawn is a high tide from October to March, a rising one most other mornings, and a low one six mornings a year. The dropping tide the inside wants comes at ten or eleven — with the wind.Synthesis · 26 September 2026
Sources and method
Tide model: functions/api/_tide_model.js — 33 harmonic constituents fitted by least squares to the UHSLC research-quality hourly record for Zihuatanejo (station 687, 17.637°N 101.558°W, 3 March 2011 – 10 July 2013, 13,206 valid hours), heights above MLLW, no nodal correction. The same tideExtremes() and tidePhaseAt() the forecast engine calls for its phase labels; used here for the clock tables. Heights and ranges: functions/api/_tide_model_nodal.js — the same 33 constituents refitted by least squares with Schureman nodal factors to the IOC Zihuatanejo float gauge (hourly, May 2013 – September 2026, 67,111 hours after quality control, one offset per month), long-period constituents kept from the original fit, heights on the same MLLW datum; checked by leaving out one year at a time (eight years).
Windows: primary 1 October 2026 – 30 September 2027 (365 days); check 1 October 2025 – 30 September 2026. Local calendar days in America/Mexico_City (UTC−6, no daylight time). Every instant passed to the model is a true UTC millisecond.
- Extremes: highs and lows per local day from a one-minute scan; “low/high between 06–09h” = at least one extreme of that type with 06:00 ≤ t < 09:00.
- Hourly phase:
tidePhaseAtat each of the 24 local hours, tallied per month (28–31 days × 24 slots); pooled shares reported for 07–09h and 10–12h. - Hour histograms: all lows and highs in the year by local hour; “clusters” are runs of hours each holding ≥ 5% of the year's extremes.
- Daily range: highest minus lowest extreme within the local day (the engine's
tideSwingsFromModeldefinition), from the refitted model; share ≥ 0.6 m. Node factors fK1 = 1.006 + 0.115 cos N − 0.009 cos 2N, fO1 = 1.009 + 0.187 cos N − 0.015 cos 2N, fM2 = 1.0004 − 0.0373 cos N and their phase terms, N from the Meeus mean-node polynomial. - Form factor: F = (K1+O1)/(M2+S2) from the fitted amplitudes; classification on the Courtier scale (<0.25 semidiurnal, 0.25–1.5 mixed mainly semidiurnal, 1.5–3 mixed mainly diurnal, >3 diurnal). F = 1.19 from the original fit, 1.27 from the refitted cycle averages, 1.43 with the 2026–27 node factors applied; all under 1.5.
Script and artifact: scripts/analyze_tide_clock.mjs → functions/api/_findings_tide_clock.js (export FINDINGS_TIDE_CLOCK: every table on this page, the 24 × 12 phase strips for all four labels, the six dawn-low dates, and generated_at).
Analysis run: 26 September 2026.
Cite as: “The tide clock: why low tide at dawn almost never happens at La Saladita”, La Saladita Field Guide, lasaladita.com/findings/the-tide-clock/, 2026-09-26.