Methodology

Core principles

Data sources

Sub-score formulas

Each sub-score is normalized to 0–100 before being combined. These are the exact thresholds running in production.

avg ∈ [18, 24] °C  → 100  (ideal)
avg ∈ [15, 27] °C  → 85
avg ∈ [10, 30] °C  → 70
avg ∈ [5, 32]  °C  → 55
otherwise          → 40

< 10 km/h  → 100  (Beaufort 0–2)
< 20 km/h  → 85   (Beaufort 3)
< 30 km/h  → 70   (Beaufort 4)
< 40 km/h  → 50   (Beaufort 5)
< 50 km/h  → 30   (Beaufort 6)
≥ 50 km/h  → 10   (Beaufort 7+)

score = 100 - precipitation_probability_max
if precipitation_sum > 10 mm  → score -= 20
else if precipitation_sum > 5 mm → score -= 10
score = max(0, score)

code ≤ 1   (clear / mainly clear) → 100
code ≤ 3   (partly cloudy)        → 80
code ≤ 48  (fog)                  → 60
code ≤ 57  (drizzle)              → 45
code ≤ 67  (rain)                 → 35
code ≤ 77  (snow)                 → 30
code ≤ 86  (showers)              → 25
code ≥ 95  (thunderstorm)         → 15

Composite scores

Three composite scores are derived from the four sub-scores using fixed weights:

temp   × 0.40
+ wind   × 0.20
+ precip × 0.30
+ code   × 0.10

temp   × 0.30
+ wind   × 0.30
+ precip × 0.30
+ code   × 0.10

code   × 0.40
+ wind   × 0.30
+ precip × 0.20
+ temp   × 0.10

Overall score = (Travel + Outdoor + Photography) / 3, rounded.

Golden Hour Photography Score

A dedicated formula for /golden-hour/[city] pages. Starts at 100 and applies graduated penalties:

Weather code:
  code = 0    → -2  (clear, no dramatic clouds)
  code ≤ 2    → +5  (partly cloudy — IDEAL for golden hour)
  code = 3    → -8  (overcast)
  code 45–48  → -18 (fog)
  code 51–57  → -25 (drizzle)
  code 61–67  → -35 (rain)
  code ≥ 80   → -40 (showers / storms)

Wind (segmented penalty, capped at -45):
  ≤ 12 km/h   → 0
  12–25 km/h  → 0.6 × excess
  25–40 km/h  → +1.0 × excess
  40–55 km/h  → +1.4 × excess
  > 55 km/h   → +1.8 × excess

Precipitation:
  if probability > 20%  → -(prob - 20) × 0.15

Temperature:
  if avg < 5°C or > 35°C → -3

final = clamp(0, 100, round(score))

Activity scoring

Each activity (hiking, beach, city, photography, skiing) has its own threshold profile. Penalties accumulate but are capped at 60 total points.

Penalties scale with how far conditions exceed each threshold. Activity-specific bonuses apply for ideal conditions (e.g., +8 for photography under clear skies, +8 for beach when 25–32 °C and clear).

Recommendation logic

Numerical scores convert to one of four labels — but weather codes can override the label downward:

Severe weather (rain, showers, thunderstorm):
  score ≥ 60  → "fair"
  otherwise   → "poor"

Mild bad weather (drizzle, fog):
  score ≥ 70  → "good"
  score ≥ 40  → "fair"
  otherwise   → "poor"

Snow:  treated as mild bad weather
Clear / partly cloudy:
  score ≥ 80  → "excellent"
  score ≥ 60  → "good"
  score ≥ 40  → "fair"
  otherwise   → "poor"

This is why a 78-point day with rain shows as "fair," not "good." The cap is intentional and prevents misleading recommendations.

Forecast reliability

Forecast confidence drops linearly with day index, with a floor at 60%:

reliability(day) = max(60, 95 - day_index × 5)

≥ 85%  → "Very High"  (days 1–2)
≥ 75%  → "High"       (days 3–4)
otherwise → "Medium"  (days 5+)

For "Most Stable," "Sunniest," and "Least Windy" rankings we deliberately limit the candidate set to days 1–7. Beyond that, sub-daily variation in the underlying models loses precision.

Limitations & honest disclosures

• •Daily resolution. Composite scores use daily aggregates. A morning thunderstorm followed by a clear afternoon may still show as "fair" overall.
• •No microclimate adjustments. Coastal vs. inland differences within the same city are not modelled.
• •No air quality. Pollution, pollen, and visibility from haze are not factored into scores.
• •Personal preference varies. Our "ideal" temperature band of 18–24 °C is calibrated for general travel comfort. Cold-weather travelers and beach-goers will weight differently.
• •Not for safety-critical decisions. For aviation, marine, or severe-weather planning, always consult official meteorological agencies.

Authorship & review