The hardest weather call at an outdoor event isn't whether the storm hits. It's whether it hits during the act, during the encore, during the egress, or after the lot has cleared. The standard hourly forecast — 30% chance of precipitation between 6pm and 7pm — is operationally useless. Sometime in an hour-wide window means everything; it means nothing.
AccuWeather MinuteCast is the product that closes the gap. Minute-by-minute precipitation forecasts for the next 120 minutes, refreshed continuously. For an outdoor festival, those 120 minutes are usually the exact window between "we still have time to call this" and "we've already lost the option to call this." This post is on how to use it.
What MinuteCast is
AccuWeather MinuteCast is a per-call API at $0.05–$0.25 per call depending on volume tier. For a single venue running 100 scans a day across an event weekend, the cost is a rounding error — somewhere between $5 and $25 per event-day. For a multi-venue festival operator running continuous monitoring on six sites, you're still measured in tens of dollars per day.
The product returns, for a single lat-long coordinate, the precipitation type and intensity for each of the next 120 minutes. Light rain at minute 14, heavy rain at minute 22, no precipitation minutes 23–48, light rain at minute 49, and so on. The forecast updates with every call, so a re-poll fifteen minutes later returns a fresh 120-minute window.
Why hourly forecasts fail
The structural problem with hourly forecasts is that they're optimized for the general public, not for operations. If there's a 30% chance of precipitation 6–7pm, the public-facing version that's right says "carry an umbrella." The operational version requires knowing whether the precipitation is at 6:05, 6:30, or 6:55 — and whether it's drizzle or a 25-minute downpour — because each maps to a different decision.
At 6:05, with the gates opening at 6:00 and the act starting at 7:30, you have time to hold the doors, run an updated nowcast with your consulting meteorologist, and decide on a weather hold without affecting the act. At 6:55, you've already committed the inflow. At 6:30, you have a 40-minute decision window, and the difference between a 5-minute drizzle and a 25-minute downpour is the difference between continuing the show and starting the weather-hold protocol.
The hourly forecast cannot distinguish between these. MinuteCast can.
The 120-minute window in practice
Walk through a single decision tree.
It's a festival, single-day, headliner at 9pm. Gates closed at 6pm. You're at 7pm. MinuteCast at 7pm returns: clear minutes 0–43, light rain minutes 44–58, clear minutes 59–120.
That's a 14-minute window of light rain starting at 7:44pm. The headliner's not on stage; the support act is. Light rain is operationally a non-event for most outdoor festivals — the stage and the crowd both handle it. The decision: continue, monitor the next refresh.
Re-poll at 7:30pm. MinuteCast now returns: clear 0–14, heavy rain 15–67, light rain 68–120.
The forecast has worsened. Heavy rain starting at 7:45pm, lasting 52 minutes. That's now operational. The act on stage gets paused at 7:45; the crowd will start moving for cover; egress paths around the stage get congested; medical events spike. Re-coordinate with the stage manager, prepare the weather-hold protocol, communicate the option to the headliner's management.
Re-poll at 7:45pm with the rain now starting. MinuteCast: heavy rain 0–35, light rain 36–62, clear 63–120.
Now you have the full picture. The heavy band is 35 minutes. The headliner is at 9pm; the band clears by 8:20pm; the egress paths drain by 8:30pm. The decision is to weather-hold the support act, restart at 8:30pm with a compressed support set, and bring the headliner on at the scheduled 9pm. You've kept the show, kept the audience, kept the artist relations, and stayed within the operational envelope.
That sequence is impossible with hourly forecasts. It is routine with MinuteCast.
Limitations to be honest about
MinuteCast is not lightning forecasting. Lightning is a separate AccuWeather product line and a separate contract. If you're running an outdoor event with any lightning risk, wire both — they're complementary, not substitutable.
MinuteCast is also point-location, not areal. A venue with a footprint larger than half a kilometer should poll multiple points around the footprint rather than the centroid. Storms move. Edge-of-footprint precipitation is operationally different from centroid precipitation. Most festival operators we work with poll three to five points across the site.
International coverage is strong but uneven. North America, Western Europe, Japan, and Australia are well-covered. South America, Eastern Europe, and most of Asia are usable but worth validating against your local meteorological consultant's nowcasts.
How we wire it in
MinuteCast is a BYOK environment-pillar integration in SignalGuard. Customer provisions the AccuWeather contract (developer.accuweather.com — the API quota tier you want depends on event volume), drops the API key into /integrations, and from that point forward every scan on the venue includes a MinuteCast strip showing the next 120 minutes of precipitation.
The strip lives in the Environment pillar of the scan brief. When MinuteCast indicates precipitation in a window that overlaps with a scheduled event time (pulled from Ticketmaster Discovery, also wired), the synthesis layer surfaces the overlap explicitly — "Heavy rain forecast 19:45–20:20, overlaps with scheduled support act 19:30–20:00."
Severity weighting depends on intensity and venue type. A drizzle at a covered amphitheater is Low. A heavy band at a fully open festival site is High. The composite score reflects the venue context, not just the weather signal in isolation.
The cost-per-decision math
A typical mid-sized outdoor festival runs roughly $200,000 in operational cost per event-day. A wrong weather call — pulling the headliner unnecessarily, or failing to pull and seeing crowd injury — is measured in the millions, before insurance. MinuteCast costs $5–$25 per event-day to run continuously.
The return on a single avoided wrong call is roughly five orders of magnitude. Which is why the question for most outdoor festival operators isn't whether to wire MinuteCast. It's why they don't have it wired already.
To wire it, the integration is on /integrations. To see what a baseline scan looks like before you wire MinuteCast, /scan runs free without it.