Reading the wind.
Checking your session…
Name the limit once. Get one email when the forecast crosses it.
An alert is a standing question: “tell me if the forecast at this site reaches this number in the next day.” You write the question once, from your own document, and The Wind Agent asks it for you around the clock. When the answer changes from no to yes, you get an email that names the window and the peak. When nothing has changed, you get nothing. This page explains how rules are written, how often they run, what the email contains and what the morning brief adds.
Writing a rule
- Metric. Mean wind speed or gust. Choose the one your document limits. Gusts matter for lifting, mean speed for spraying and drone work.
- Comparison. At or above the threshold, or at or below it.
- Threshold. The limit, in your unit.
- Label and source document. A short name such as “Tower crane out-of-service” and the document it comes from, so the email carries its own provenance.
- Look-ahead. How many hours ahead to look, from 1 to 72. The default is 24.
- Email on or off. Turn delivery off to keep the rule for the preview and the fleet board.
The Wind Agent does not suggest limits and holds no statutory ones. The number comes from you. If you are not sure where to start, read setting a limit from a document.
How rules are evaluated
A scheduled scan runs every 30 minutes. For each saved site with a rule it reads the modelled forecast and looks at every hour from now until the end of the look-ahead. Consecutive hours that breach the rule are merged into one window with a start, an end, a length in hours and the peak value and time. A single hour with missing data ends a window and is counted as unknown, never as calm. The result is one of four states:
- BREACH — at least one window.
- CLEAR — every hour evaluated, none breached.
- UNKNOWN — nothing breached but some hours had no data, or none could be evaluated.
- NO_RULE — the site has no rule to evaluate.
Alerts read the 10 m model level, where gust is defined. If a site works at a different height, the rule screen tells you so; the fleet board and the instrument show the height-matched value.
No repeat emails
Each fired window is recorded against the site and the UTC hour it starts, and every hour inside it is marked, with a seven-day memory. On the next scan, a window whose hours were already alerted is skipped. A storm that lasts into tomorrow does not email you every half hour; a second storm later in the look-ahead does.
What the email contains
- A subject line with the site, the rule label and the look-ahead, such as “Wind alert: Cork Docks — Crane gust limit forecast within 24 h”.
- The rule in words, and the look-ahead.
- Each breach window in site-local time with its length and the peak value and when it occurs.
- A link straight into the instrument at that site.
- An evidence line: the forecast source, when it was fetched, how many hours had no data, and the reminder that this is a modelled forecast and not an on-site measurement.
It is deliberately plain. There is no traffic-light colour that could be mistaken for a permission.
The morning brief
On Pro and above you can also receive one morning email covering up to 25 of your sites. It is sent at the hour you choose in your own time zone, 06:00 by default, once a day. Each site gets a lamp, the current wind at its working height, the worst hour of the day against your limit, the ensemble peak with its member count, the best window if there is one, whether the models agree, and the latest observation from the nearest station. Everything in it is computed, not written by a language model.
A worked example
A tower crane on a site in Cork has a 38 mph gust limit from the manufacturer's manual. You create a rule with metric gust, at or above 38 mph, look-ahead 36 hours. At 17:30 the scan finds gusts above 38 mph from 03:00 to 07:00 tomorrow, peaking at 44 mph at 05:00. You get one email. At 18:00, 18:30 and every scan after it, the same window is recognised and nothing is sent. At 21:00 the forecast adds a second window from 14:00 to 16:00 the day after. That one is new, so you get a second email. Throughout, you can press Preview now on the rule to see exactly what would be sent, without sending it.
What your plan includes
Email alerts and the morning brief start on Pro, which holds 25 saved sites. Ops adds the fleet board and public boards with 100 sites; Enterprise holds 500. Starter (3 sites) includes the instrument, records and the agent. Alerts are delivered by email only today; web push is planned.
Alert questions
How often are alert rules checked?
Every 30 minutes. A scheduled scan reads the forecast for each saved site that has a rule, evaluates the hours inside its look-ahead, and sends an email only when a breach window is new.
Will I get the same alert over and over?
No. Each breach window is recorded by site and start hour, and an hour that has already been alerted is not alerted again for seven days. A window that continues into later hours does not re-fire. A genuinely new window, later in the forecast, does.
Which height does an alert use?
The 10 m model level. The alert evaluator reads mean speed or gust at 10 m, and the rule screen says so when a site has a different working height. For the height-matched value, use the fleet board or the instrument.
What happens if forecast data is missing?
Missing hours are counted as UNKNOWN, break any open window, and are reported in the evidence line. They are never treated as calm. If every hour is evaluated and none breaches, the state is CLEAR; if any hour is unknown and none breaches, the state is UNKNOWN.
Can I test a rule without sending anything?
Yes. Preview now evaluates a rule immediately against the current forecast and shows the windows it would send. Nothing is emailed or recorded.
What does an alert tell me to do?
Nothing. It is a prompt to look. It lists when the forecast crosses your rule and how high it peaks, with the source and fetch time. Whether to stop work is decided under your own procedure.