(runs ≥ 2 consecutive wet days)
(days/year)
(annual max mm)
timing shift
The prior finding here — no detectable trend in Saladita's annual rainfall totals — stands. But totals are a blunt instrument. A place could receive identical annual rainfall while becoming structurally drier in the dry months and more intense in the wet ones. That reorganization would matter for plants, the estuary, and anyone waiting for rain to stop or start.
This analysis runs seven structural tests on ERA5 daily precipitation 1979–2025. The question is whether the shape of the rainy year has changed, not just the total. Two tests return a trend significant at p<0.05: wet-spell length, which survives a Bonferroni correction for seven tests, and wet-run clustering, which does not. The single-day extreme is marginal (p=0.058). Dry-spell length, monsoon onset, monsoon end and cumulative-deficit days show no trend detected.
Revised 26 September 2026: rerun at the corrected grid cell 17.8375°N, 101.7606°W. The June 2026 version of this page used a cell 44 km to the south-east (17.5897°N, 101.4317°W — a different town).
Revised 29 September 2026: the same seven definitions were run on two independent daily rain products, the CPC gauge analysis and MERRA-2. Wet runs rise in all three products and only ERA5 is significant; the shorter wet spell appears in ERA5 alone. The ERA5 numbers on this page are unchanged. The comparison is in its own section.
Wet events are grouping more tightly.
We counted the number of independent wet runs per year — sequences of two or more consecutive days with at least 1 mm of rain. A higher count means rain is arriving in tighter, more distinct bursts rather than scattered single days.
+1.0 additional wet runs per decade (p=0.015, R²=0.12). Mean: ~18 wet runs per year.
This is significant at 0.05 but does not clear the Bonferroni threshold for seven tests (p<0.0071). Across the record, 1988 had the fewest wet runs (10) and 2015 the most (26). It is modest in absolute terms — about one additional burst of wet days every decade. Rain is not falling more often; it is falling more often in clumps.
The R² of 0.12 means that calendar year explains about 12% of the variance in wet-run count — the rest is ENSO, storm-track variability, and noise. The Bonferroni-corrected threshold for 7 tests at α=0.05 is p<0.0071; this finding does not clear it (p=0.015). Treat as indicative, not definitive.
Two of these five find a significant change.
Each metric computed per year, linear regression against year. Slopes expressed per decade.
In ERA5 the longest wet spell is 3.3 days shorter per decade (p=0.0008, R²=0.22); CPC and MERRA-2 do not show it. No trend is detected in the longest dry spell (−0.7 days per decade, p=0.83).
The wet-spell result is the cleanest regression in the set and the only one that survives correction for seven tests: the longest run of wet days averaged 21.4 days, reached 45 days in 1983 and 1984, and has not exceeded 18 in the 2020s. Together with the clustering count it describes one mechanism in ERA5 — the same water arriving in more, shorter bursts. The three-product comparison below does not confirm the shorter wet spell. The longest rainless stretch averaged 120 days and shows no trend.
+11.8 mm/decade annual maximum, p=0.058. Marginal, not significant.
The slope is in the direction a more clustered regime predicts — if rain concentrates into shorter bursts, daily extremes should rise — but it does not reach p<0.05, let alone the Bonferroni threshold (p<0.0071). The mean annual maximum is 99 mm; the record is 293 mm in 2016. The p99 daily-rain trend on the companion climate page is not significant either (+4.9 mm/decade, p=0.14).
No shift detected in when the monsoon arrives or leaves.
Onset defined as the first day after April 1 where the 7-day trailing average exceeds 5 mm/day; end as the last such day before November 1. Neither date shows a detectable trend.
Mean onset around June 20, mean end around October 20. No shift detected in either.
The onset moves +2.4 days per decade (p=0.18) and the end −1.3 days per decade (p=0.23); neither is distinguishable from no change. Inside the season ERA5 shows a shorter longest wet spell; two other rain products do not (see the three-product comparison).
A dry September, a wet start to October, and no sign yet of a late finish.
This year October has felt like September usually does. The numbers below put that in context. This note is dated and will go stale; the season-end definition is the one in the section above.
Typical September: ERA5 median, 1991–2025. Typical October: the local weather station’s median for the whole month, 2018–2025. October 1–10, 2026 is at least 94 mm (3.7 in).
September 2026 was the second-driest September since 1991.
In ERA5, only September 2023 (74 mm) was drier among the 36 Septembers from 1991 to 2026. October is normally the second-wettest month here, so a wet early October is the season doing what it usually does. Most of this year's early-October rain fell from October 7 to 10, including 2.0 in (51 mm) on October 9. The station total is a floor: the station is logged a few times a day, and a missed reading can undercount a day.
No. Early-October rain and September rain do not predict when the season ends.
Across 1991–2025, rain on October 1–5 is unrelated to the season-end date (Spearman −0.04, p=0.81), and so is September rain (−0.12, p=0.49). The median end date is October 20, and 12 of the 35 seasons ended on October 25 or later.
A strong El Niño raises the odds of heavy late-October rain, not of a later end.
NOAA's ENSO advisory (read October 10, 2026) says a “strong-to-very strong El Niño likely through January-March 2027.” In the three strong El Niño years in the ERA5 record since 1991, late October was much wetter than normal, but most of that rain came in a few days and the season did not reliably end late:
Oct 11–31: ERA5 rain total. Top 3 days: share of that total that fell in the wettest three consecutive days. End: season-end date as defined above. Normal: median, 1991–2025.
The seasonal model leans slightly wet. Three years is too few to call it.
The ECMWF seasonal forecast started October 10 (51 runs) gives a median of 89 mm for October 11–31 (10th to 90th percentile 46–124 mm) and 20 mm for November (6–50 mm). ERA5’s normals for those periods are 69 mm and 14 mm. These are different models, so the gap is a hint, not a forecast of a wet finish. The practical read: expect rain through most of October as in a normal year, with a higher-than-usual chance of one large late-October storm.
No structure trend holds across products.
The same definitions were run on two independent daily rain products for this coast: the NOAA CPC gauge analysis and NASA MERRA-2 corrected precipitation. Slopes per decade with 95% intervals from a 3-year block bootstrap.
Wet runs rise in all three products; only ERA5 is significant. The shorter wet spell is in ERA5 alone.
The wet-run count has the same sign in ERA5, CPC and MERRA-2, and the CPC and MERRA-2 intervals include zero. The longest wet spell shortens by 3.26 days per decade in ERA5; CPC gives −0.44 with an interval that spans zero, and MERRA-2 gives +1.97. Wet days a year and the heaviest day change sign between products. No rainfall-structure trend at this coast is robust across products. The ERA5 numbers on this page reproduce exactly and are kept as ERA5 results. There was no rain gauge at the break before November 2023, so none of the three is ground truth.
ERA5 is a ~31 km grid cell, not a rain gauge. Daily precipitation in ERA5 represents a spatial average over the grid box and is derived from model reanalysis assimilating observations, not from a local instrument. Point-scale rainfall events (tropical downpours, orographic enhancement) are smoothed. Absolute daily values are less reliable than the long-term trends.
Temporal autocorrelation reduces effective sample size. Annual metrics derived from daily data are weakly serially correlated. The effective n is probably 30–40 rather than 47. The wet-spell result (p=0.0008) is robust to this correction; wet-run clustering (p=0.015) is not guaranteed to be.
Multiple comparison: 7 tests at α=0.05. The naive Bonferroni threshold is p<0.0071. Wet-spell length (p=0.0008) clears it. Wet-run clustering (p=0.015) is significant at 0.05 but not after correction; treat it as supporting evidence for the same mechanism, not as a stand-alone finding. The single-day extreme (p=0.058) is marginal.
ENSO dominates year-to-year variance. El Niño years tend toward suppressed Guerrero monsoon; La Niña years toward wetter conditions. The long-term trends reported here are estimated over the full ENSO cycle. ENSO-conditional analysis was not run for this piece; the clustering signal survives the pooled regression but has not been tested separately within ENSO phases.
The null results are informative. No shift is detected in monsoon onset (p=0.18) or end (p=0.23), in cumulative-deficit days (+3.5 days/decade, p=0.30) or in the longest dry spell (p=0.83). Nothing here says the dry stretches are getting longer.
No trend is detected in the annual total. In ERA5 the structure shows some change: the longest wet spell is about three days shorter per decade, and wet days come in about one more separate run per decade. CPC and MERRA-2 do not reproduce the shorter wet spell, so no structure trend holds across products; wet runs rise in all three and only ERA5 is significant. The hardest day trends heavier, but not significantly, and the season's start and end show no shift.Synthesis · Revised 29 September 2026
Sources and method
Data: ERA5 daily precipitation sum at 17.8375°N, 101.7606°W via Open-Meteo archive API (archive-api.open-meteo.com/v1/archive?daily=precipitation_sum&models=era5), 1979–2025 (47 years, 17,167 daily values). Re-fetch annually.
Comparison products: NOAA CPC Global Unified Gauge-Based Analysis of Daily Precipitation (0.5°, via NOAA PSL) and NASA MERRA-2 bias-corrected precipitation (PRECTOTCORR, via NASA POWER), each at the cell nearest the site, 1979–2025 (MERRA-2 from 1981). A year enters when it holds at least 330 days of data. Slopes by least squares on annual values; intervals from a 3-year block bootstrap, 4,000 draws.
Metrics per year:
- Max dry-spell: longest run of consecutive days < 1 mm.
- Max wet-spell: longest run of consecutive days ≥ 1 mm.
- Wet-run count: number of independent runs of ≥ 2 consecutive wet days (≥ 1 mm/day).
- Annual max 1-day: maximum single-day total.
- Deficit days: days where 30-day trailing mean < 20% of long-term DOY climatology mean.
- Monsoon onset: first DOY after April 1 where 7-day trailing mean > 5 mm/day.
- Monsoon end: last DOY before November 1 where 7-day trailing mean > 5 mm/day.
Regression: scipy.stats.linregress (OLS) applied to each annual metric vs calendar year. p-values approximate — autocorrelation reduces effective n. Bonferroni threshold for 7 simultaneous tests: p<0.0071. Scripts: scripts/analyze_rainfall_regime.py → functions/api/_findings_rainfall_regime.js.
Analysis run: 28 September 2026 at the corrected cell. The June 2026 numbers on this page came from the 17.5897°N, 101.4317°W cell and are superseded.
Corrected 28 September 2026: the series is now requested as ERA5 itself (models=era5); without the default product's 2017 model splice, dry-spell length, the single-day extreme and monsoon end are no longer significant.