Convergence season: why summer sea-breeze days are an API showcase

It is the time of year when, on the right day, the whole south of England flies a line. A sea breeze pushes inland from two coasts, the two flows meet somewhere over the middle, and where they meet the air has nowhere to go but up. The result is a convergence line: a thin band of lift, a couple of kilometres wide, that can run a hundred kilometres or more, where you fly straight and stay up without circling. For a cross-country pilot it is the difference between a good day and a season-best one.

This is the kind of day that shows most clearly why a soaring-specific API exists. A general weather app can tell you the surface wind, the temperature, the chance of rain. It cannot tell you that two airmasses are going to collide over Wiltshire at three in the afternoon, because finding that needs a field most weather products simply do not compute.

Convek computes it. `horizontal_convergence_s1` is the kinematic convergence of the 925 hPa wind field - literally how hard the air is being squeezed together at that level, in units of inverse seconds. Positive means converging, which means rising. The mirror field `horizontal_divergence_s1` is the same thing with the sign flipped. There is a full explainer on how they are derived, but for reading summer days the numbers worth carrying in your head are these: background noise sits around 5e-5, a weak line shows around 1e-4, and a solid mid-afternoon sea-breeze front runs 3e-4 to 5e-4 or more.

The way to use it in convergence season is not to stare at the magnitude at one point - it is to look at the shape. Pull `horizontal_convergence_s1` over a bounding box through the grid endpoint for the afternoon hours, render it as a layer, and look for a line: a coherent band of convergence, roughly perpendicular to the wind, that holds its position across several forecast hours. That persistence is the tell. A momentary blob is just turbulence; a line that is still there an hour later is a street you can plan a flight around.

The divergence field earns its keep on the other kind of summer day - the blue, high-pressure, looks-promising-but-never-quite-goes day. Strong low-level divergence under high pressure means large-scale subsidence: air sinking from above and quietly capping the boundary layer no matter how good `wstar_ms` looks. If divergence over your launch is sitting high through the morning under a ridge of high pressure, treat the day rating with more suspicion than usual. Convergence tells you where to go; divergence tells you when not to bother driving.

This is the one signal that is genuinely hard to fake from generic weather data. Thermal strength you can approximate from stability indices. Cloudbase you can estimate. But a convergence line is a pattern that only exists in a model that resolved the two flows colliding, which is the whole reason for running a 4 km regional model instead of reskinning a global one. It is, in a real sense, the field that justifies the project. Go find a line this summer and tell me whether the model called it - the coverage page has the field list and the grid endpoint docs have the rest.