Why maxwind_kph in the Daily Forecast Doesn’t Match Any Hourly wind_kph — and How to Find the Real Peak

If you’ve ever pulled day.maxwind_kph from a forecast response and then fetched the hourly breakdown for the same day, you’ve probably noticed they don’t agree. Sometimes the daily max is higher than any individual hour’s wind_kph. Sometimes it’s lower. Occasionally it matches one hour exactly, which almost feels like a coincidence.

It’s not a bug. It’s a consequence of how those two fields are computed from fundamentally different data. If you’re building anything that cares about peak wind — event cancellation logic, construction site safety alerts, outdoor scheduling — you need to know which one to trust and when.

Where daily maxwind_kph actually comes from

day.maxwind_kph is derived from model output at the daily aggregation layer. The models we ingest — HRRR, NAM, GFS — produce wind forecasts at their native time steps, which vary by model. HRRR runs at hourly resolution out to 18 hours, then 3-hourly. GFS goes out to 16 days but its native output is 3-hourly for the first 10 days, then 12-hourly beyond that. The daily max is computed across the full set of available model time steps for that calendar day, including steps that don’t surface as discrete hours in the hourly API response.

The hourly forecast response interpolates to give you a clean 24-point series. Interpolation works well for smooth fields like temperature. It’s not great for wind, because wind peaks are often narrow spikes — a frontal passage might push gusts hard for 90 minutes and then ease off. If that spike falls between two interpolated hourly points, it gets smoothed down in both adjacent hours, but it was still captured in the raw model step that fed maxwind_kph.

That’s the main reason daily max beats hourly max: the peak existed in the model at a sub-hourly or non-aligned step, got smoothed in the hourly interpolation, but was preserved in the daily aggregation.

When hourly max beats daily max

The reverse case is rarer but it does happen. When a station-based observation is blended into the hourly series — particularly a METAR reading near the request location — that observed gust can exceed what the model predicted for the same period. The daily aggregation is model-driven; the hourly series can be model-blended-with-obs. A measured gust that the model underforecast will appear in wind_kph at that hour but won’t necessarily push the daily max upward, depending on when the aggregation was computed relative to observation ingestion.

Practically: for future days, treat maxwind_kph as the more conservative upper bound. For today and yesterday, the hourly series may catch observationally-driven peaks the daily max missed.

Why gust_kph makes this messier

There’s a third number in play. The hourly response also includes gust_kph — the forecast or observed gust speed, typically the peak 3-second wind over that hour — as opposed to wind_kph, which is sustained (10-minute mean at 10m, following WMO convention). The daily response has maxwind_kph but no explicit daily max gust field.

maxwind_kph at the daily level corresponds more closely to max sustained wind than to max gust, though the mapping isn’t always clean depending on which model run contributed that value. If you’re trying to flag conditions where gusts exceed a threshold — 60 kph for an outdoor event risk model, say — you can’t rely on maxwind_kph alone. You need to scan gust_kph across the hourly array for that day.

A concrete extraction pattern

Here’s the approach that actually works for peak wind detection across a day, using the forecast endpoint with days=3 and hourly data:

// Pseudocode — adapt to your language/HTTP client
const response = await weatherapi.get('/forecast.json', {
  q: location,
  days: 3,
  hour: null  // get full hourly arrays
});

for (const day of response.forecast.forecastday) {
  const hours = day.hour;

  const maxSustained = Math.max(...hours.map(h => h.wind_kph));
  const maxGust      = Math.max(...hours.map(h => h.gust_kph));
  const reportedDayMax = day.day.maxwind_kph;

  // Use the higher of the two sustained-wind figures
  const effectiveMax = Math.max(maxSustained, reportedDayMax);

  // For gust-sensitive applications, maxGust is the number you care about
  console.log({
    date: day.date,
    effectiveMaxSustained: effectiveMax,
    maxGust: maxGust
  });
}

Taking Math.max(maxSustained, reportedDayMax) gives you a ceiling that’s robust to both the interpolation-flattening problem and the obs-blending problem. You’re not double-counting — you’re taking the highest value each source gives you.

For gust-sensitive logic, maxGust from the hourly scan is the number to threshold against. Don’t conflate it with maxwind_kph.

The 3-hourly gap beyond day 3

Past roughly the 48-hour mark, HRRR drops out and you’re mostly on NAM and GFS. GFS at 3-hourly resolution means the hourly interpolation is spanning larger gaps, which amplifies the smoothing problem. A wind spike that’s 2 hours wide in real atmosphere might span a single 3-hour GFS step — it’ll appear as a mild elevation across those three interpolated hours rather than a clear peak. maxwind_kph at the daily level is typically more reliable for days 4–7 than trying to infer the peak from the hourly series, precisely because it was computed on the raw model steps before interpolation flattened them.

This isn’t a defect you can route around. It’s an inherent trade-off of offering a uniform hourly interface across model resolutions that aren’t all hourly. Weight your confidence in hourly-derived peaks lower as you move further out, and fall back to the daily max as your upper bound.

One thing worth flagging before you ship

If you’re building a strict safety-threshold system — something where exceeding X kph triggers an operational hold — don’t set your trigger at the exact model-forecast level. Models routinely underforecast local wind peaks in terrain-channeled locations: coastal inlets, mountain passes, urban street canyons. maxwind_kph at 55 kph in complex terrain probably deserves a trigger threshold several kph lower in your app. The API gives you the number; it doesn’t know how exposed your specific asset is.

The question worth sitting with: which direction hurts more to be wrong? Missing a real peak, or false-alerting on a calm day? That answer should determine whether you bias toward the hourly max, the daily max, or the gust field. They’re not interchangeable, and the difference will find you eventually.

Scroll to Top