Hours without rain or thermal stress.

Two ways to ask the engine.

Forelore's audit engine answers two queries over the same 25-year climate floor dataset:

The Harbor Island case study above is the first query type, applied to a wedding date selection. The same engine serves vacation rentals, property buys, festivals, multi-venue portfolios, and destination comparisons.

Microclimate-aware. Not a regional average.

Forelore audits at 0.1° lat/lon resolution — each pixel is roughly 7 miles across, the size of a small coastal town. Two villas 12 miles apart on the same coast may have meaningfully different floors. The audit is venue-specific: your exact location, not a regional or city-level smoothed average.

Rain — the primary filter.

An outdoor event lives or dies on rain. Forelore uses NASA GPM IMERG Final V07 — the gold-standard satellite precipitation dataset, with 0.1° lat/lon resolution (~7 mi) and 30-minute temporal cadence. Each 30-minute slot counts as rain-free below 0.1 mm/h rain rate. No rain, no event lost.

Daytime rain only. The audit counts rain falling between 8am and 10pm local time, ignoring rain at 3 AM when nobody is outside. A pre-dawn downpour doesn't ruin a daytime ceremony — and traditional climate averages that lump night and day together mask the signal that actually matters for outdoor events.

Why satellite rain, not reanalysis? Validated against NOAA Stage IV gauge ground truth, ERA5 reanalysis exhibits a −2.3-hour drizzle bias in places like Miami — counting low-rate rain the gauges don't register. IMERG matches the gauge within ±0.4 hours. The choice of rain dataset alone can shift the audit by hours per day.

Then: strong thermal stress.

Once rain has cleared, what's left is thermal comfort. A 30°C day with full sun, no wind, and 80% humidity feels around 38°C to a person standing outside in formal clothing — the thermometer reads 30°C, the body feels 38°C. Forelore measures what your people actually feel, using the UTCI (Universal Thermal Climate Index) — the scientific standard for outdoor thermal stress.

UTCI uses 4 inputs (not just temperature).

UTCI consumes air temperature, mean radiant temperature (sun + ground + buildings), wind speed, and water vapor pressure (humidity). It accounts for all of them simultaneously using a 6th-order polynomial fit (Bröde et al. 2012, IJBM). Below −13°C or above +32°C UTCI, the body experiences strong thermal stress — those hours are excluded from "pleasant" in the audit. The UTCI dataset is Di Napoli et al. 2021 (DOI 10.1002/gdj3.102), published in Geoscience Data Journal.

The two 90s.

When Forelore reports a "high-confidence floor" of X pleasant hours, two numbers are both 90% — and they measure different things. How often a day clears the floor: at least 90% of the time (coverage). How sure we are of that: about 90% confidence. The floor is set at the level where a one-sided 90% confidence bound holds the ≥90% rate even given the sampling uncertainty of a 25-year record.

The inter-annual variability is measured directly from 25 years of observations, not modeled. We also checked the year-to-year autocorrelation in the data: for this rain-driven metric it is negligible, so the 25 years behave as roughly 25 independent samples — the floor isn't resting on one lucky decade.

What this means for you: the floor is a deliberately conservative number — the level an event-day clears at least 90% of the time, at 90% confidence. The audit's job is to hand you that defensible floor, not the average day — so you can stand behind the date.

ENSO sensitivity: our Jul 2000–Jun 2025 sample contains 3 La Niña / 5 El Niño / 17 Neutral years (long-run climatology ~25/30/45%). Bootstrap re-weighting to climatological ENSO frequencies moves the audited floor by ≤0.3 hours — the 2nd-lowest order statistic already absorbs the worst La Niña year in the sample directly.

Full statistical derivation (Weibull plotting position, autocorrelation test via Bretherton 1999 + Wilks 2011, empirical stationarity per Sun et al. 2018, ENSO robustness analysis, references to David & Nagaraja 2003 and IPCC AR6): read the methodology note.