Jos — 14-Day Forecast

Best window (highest confidence, then grade, then length): Sat 7 Nov 26 0001–0800L (Marginal, 9 h) ☾, confidence 54%.

Learned corrections (METAR-verified): skill weighting — pooled all-field scores for 206; bias corrections — metcheck wind +4.7 kt, jma wind +4.2 kt, meteoblue wind +3.6 kt, meteoblue QNH −3.0 hPa, openmeteo dewpoint +3.0 °C, openmeteo QNH −3.0 hPa (+67 more); hour-of-day bias (local time, shrunk toward the all-hours bias) — jma morning wind +4.7 kt, metcheck morning wind +4.7 kt, metcheck afternoon wind +4.6 kt, meteoblue morning wind +3.8 kt, jma afternoon wind +3.8 kt, meteoblue afternoon wind +3.3 kt (+284 more); cloud base estimated as 410 ft per °C of dewpoint spread for sources that report no ceiling (≥60% cloud cover; flagged estimated, weighted ×0.5); probability calibration — rain chance: meteoblue, openmeteo, foreca, metcheck, openweather, visualcrossing, Tomorrow.io, wttrin, ecmwf, gfs, icon, ukmo, gem, ecmwfens; storm: openmeteo, metno, foreca, openweather, visualcrossing, Tomorrow.io, wttrin, ecmwf, gfs, icon, meteofrance, aviationweather, ukmo, jma, gem, ecmwfens, aifs (each source's stated probability mapped to its verified frequency by an isotonic fit on ≥200 pairs, proven on the holdout); plausibility bounds from ERA5 climatology 2016-2025 (mean ± 5 sd; screening only, not observed truth); skill blend 1 auto-tuned over 8464 holdout pairs (consensus beat the best single source, icon, by 17%); correction gates (newest-7-day holdout; pooled skill is always on) — regional skill: off this week (did not improve holdout); place skill: off this week (did not improve holdout); lead-time skill: off this week (did not improve holdout); small-sample skill shrinkage: off this week (did not improve holdout); bias correction: on (improved holdout by 24.1%); place bias: on (improved holdout by 12.0%); hour-of-day bias: on (improved holdout by 4.9%); QNH bias: on (improved holdout by 58.2%); gust estimates: off this week (did not improve holdout); cloud-base coefficient: off this week (insufficient holdout: 75 of 200 items); recent-form skill: off this week (did not improve holdout); probability calibration: on (improved holdout by 19.5%); nowcast anchor horizon: off this week (did not improve holdout); dust-visibility estimate: off this week (did not improve holdout); fog risk: off this week (did not improve holdout).

Outlook

Today — Thu 5 Nov 26

A Excellent · B Good · C Fair · D Marginal (flyable) · E Poor · F Very Poor · G Unacceptable · (worst condition in each period)

Confidence: 51% — data for 23 of 24 hours · first light 0554L · last light 1822L · Moon 15 % illuminated, rises 02:58, sets 15:18.

Tomorrow — Fri 6 Nov 26

A Excellent · B Good · C Fair · D Marginal (flyable) · E Poor · F Very Poor · G Unacceptable · (worst condition in each period)

Confidence: 54% — data for 24 of 24 hours · first light 0554L · last light 1821L · Moon 8 % illuminated, rises 03:46, sets 15:57.

Extended Outlook

A Excellent · B Good · C Fair · D Marginal (flyable) · E Poor · F Very Poor · G Unacceptable · (worst condition in each period)

Sat 7 Nov 26

Confidence: 54% — data for 24 of 24 hours · first light 0554L · last light 1821L · Moon 4 % illuminated, rises 04:34, sets 16:37.

Sun 8 Nov 26

Confidence: 54% — data for 24 of 24 hours · first light 0554L · last light 1821L · Moon 1 % illuminated, rises 05:24, sets 17:19.

Mon 9 Nov 26

Confidence: 54% — data for 24 of 24 hours · first light 0555L · last light 1821L · Moon 0 % illuminated, rises 06:14, sets 18:04.

Tue 10 Nov 26

Confidence: 5% — data for 2 of 24 hours · first light 0555L · last light 1821L · Moon 2 % illuminated, rises 07:06, sets 18:52.

TAF

No TAF available for this location.

Imagery

Source panel

Each source's own raw values, min–max over the next 6 hours from the issue hour (before any weighting or bias correction). — means the source does not publish that metric in the window. Wide gaps between rows are real disagreement; the forecast above weights sources by measured skill.

Reading this report

Grades: each hour carries one grade, set by its WORST factor. The bands below are the thresholds in force for this report (Settings); a value in a band earns that grade, and "-" means the factor never produces that grade. Cloud is graded on the worse of cover and ceiling. The Driver column names the limiting factor with its value; open it in the web report for the boundary and sources.

Flight windows: a window is at least 2 consecutive hours graded Marginal or better; an hour with insufficient data never counts toward one. ☾ marks a window outside first-to-last light.

Time: hours are local time (WAT, UTC+1) unless suffixed Z, which is UTC (TAF, METAR and issue times). Local midnight is labelled 0001.

Units: temperature °C · wind and gusts knots · visibility metres (km in text) · cloud ceiling feet · cloud cover % · rain chance and thunderstorm probability % · QNH hPa (altimeter setting at the field) · CAPE J/kg. A figure like "28 (26–30)°C" gives the p10–p90 range; "spread" marks disagreement between sources.

Symbols: ▲ better than the pool · ▽ worse (only when the 95% interval excludes the pool) · ≈ not distinguishable from the pool · unmarked = unproven (accuracy is measured against METAR and SYNOP observations). ⚠ on a remark means a low-confidence hour; a ⚠ paragraph on the first page is a convective-warning or data-age advisory. ⚡ is lightning observed by satellite near the field (©EUMETSAT).

Hourly glyphs: the wind arrow points the way the air moves (a north wind is ↓, an east wind ←); cloud cover shows as ☀ ◔ ◑ ◕ ☁ (clear, few, scattered, broken, overcast); a shaded hour label is wholly before first light or after last light.

Terms: "No data" - no source forecast the hour, or a core group (cloud, visibility, wind) is missing and the hour could still be flyable; missing data is never given a good grade. "Low-confidence" - few or weak sources behind the load-bearing metrics; treat the grade as indicative. "Insufficient data" - too little to grade at all. "Estimated" (est.) - the value is derived, not reported by a source (dust visibility, cloud base, gusts); it is declared where used and weighted at half. "Derived" - an hour interpolated from a coarser (3-hourly or daily) source, weighted at half and never used to score accuracy.

Data provenance

Corrections applied (detailed under Learned corrections on the first page): skill weighting, bias correction, hour-of-day bias, cloud-base estimate, probability calibration, climatology bounds, skill blend.

How accurate have we been

Measured against observations (METAR/SYNOP), jos only. Verification table computed 2026-11-04. Mean absolute error (MAE): lower is better; "beats persistence" compares the best source with "it will stay as it is now" on the same hours.

Learned bias (forecast minus observed; positive = sources run high), mean over scored sources, by local time of day. These are corrected for before sources are averaged.

Not yet available: Scores are decayed over the verification window (half-life 14 days); separate last-7-day and last-30-day windows are not stored. Bias is the current learned value; the history needed for a trend over time is not stored. Flight-window hit / bust / miss is not yet recorded.