Convergence lines: reading `horizontal_convergence_s1`
Convergence lines are the long-distance pilot's quiet superpower - thin streets of forced lift you can fly along without ever circling. Here is what the new convergence and divergence fields actually represent and how to read them.
If you have ever flown a sea breeze front in summer, or run for hours along a line of cumulus that just kept going, you have flown a convergence line. Convergence lines are places where two airmasses are forced together horizontally and have nowhere to go but up. The result is a long, narrow band of lift, often only a couple of kilometres wide but tens or hundreds of kilometres long, where you can fly straight and stay up without circling. For cross-country pilots they are the difference between a 100 km day and a 400 km one.
Until the grid endpoint launch, the Convek API had a single scalar convergence proxy (`convergence_ms`) that combined modelled vertical velocity with a directional change between 925 and 850 hPa. It was useful as an input to the day-rating score but was not a clean field to look at on a map. The grid release added two proper kinematic diagnostics alongside it: `horizontal_convergence_s1` and `horizontal_divergence_s1`, computed directly from the 925 hPa wind field on the model grid. This post is what those new fields actually mean.
The maths is the same maths every meteorology textbook starts with. Horizontal divergence is `du/dx + dv/dy`, where `u` and `v` are the eastward and northward components of the wind at 925 hPa. Where that sum is positive, air is spreading out and (by mass continuity) sinking from above. Where it is negative, air is being squeezed together and rising. `horizontal_divergence_s1` carries the raw signed quantity. `horizontal_convergence_s1` is just its negative, so positive values always mean rising air. Both have units of inverse seconds, written `1/s`. The `_s1` suffix in the field name is shorthand for that. The wind components are lightly smoothed with a 3x3 box filter before differencing, to suppress single-cell numerical noise without flattening real lines.
Numbers worth memorising. In our 4 km UK output, background noise sits around `5e-5 1/s` either side of zero - that is just synoptic-scale wind variation, not a feature. A weakly organised convergence line shows up around `1e-4 1/s`. A solid sea breeze front in mid-afternoon is `3e-4` to `5e-4 1/s`, sometimes more on a sharp day. Anything above `1e-3 1/s` is a strong, well-defined boundary, the kind that produces a visible cloud street if there is enough moisture. The exact threshold for a flyable line depends on cloudbase and wind shear above the boundary, but as a rule of thumb, anything `2e-4 1/s` or stronger and not aligned with the wind is worth a look.
What pilots are actually looking for in the field is not the magnitude on its own, it is the shape. A point of strong convergence on its own is a vertical core, useful but no different to a thermal. A line of moderate convergence, oriented roughly perpendicular to your intended track and stable over a few forecast hours, is the gold. The grid endpoint exists partly so apps can render the field as a coloured layer over a map, where the line structure is immediately obvious and the magnitude becomes secondary information.
The mirror field, `horizontal_divergence_s1`, is the one most people skip and probably should not. Strong positive divergence at 925 hPa indicates large-scale subsidence in the lower atmosphere. That is exactly the regime where a forecast can look promising on `wstar_ms` alone but the inversion never breaks, because the descending air is suppressing PBL growth from above. If `horizontal_divergence_s1` over your launch area is sitting at `3e-4 1/s` or above through the morning, treat the day rating with more scepticism than usual. Divergence is the field that tells you when not to drive.
Where convergence lines come from, and what to expect at 925 hPa. Four common patterns. Frontal convergence at the leading edge of a cold front - well-resolved at 4 km, but usually accompanied by enough cloud to stop you flying it. Synoptic and trough-line convergence in the larger flow, which is what 925 hPa shows most cleanly. Mountain-induced convergence where deep flows split around a ridge and rejoin downwind. And the partial signature of sea breezes - a UK sea breeze front itself is often shallow enough to sit below 925 hPa, but the convergence at the top of the sea-breeze head, where it pushes inland and lifts the larger airmass, does imprint on this field once the breeze is well-developed. If you want the very shallow surface signature instead, derive your own divergence from the surface wind layer that the grid endpoint also exposes - we publish 925 hPa as the primary diagnostic because it is the cleaner kinematic field, but the choice is yours.
Caveats worth being honest about. At 4 km grid spacing with a 3x3 smoother and centred differences, the effective baseline for the gradient is around 8 km. Convergence features narrower than that get spread across two cells. Real frontal and sea-breeze boundaries can be 2 to 3 km wide, so the model tells you a line is there but underestimates its sharpness. The 1 km nesting work on the roadmap will help. Until then, treat `horizontal_convergence_s1` as a 'where to look' field rather than an 'exactly here' field. Forecast hour matters too: convergence patterns build through the afternoon, so the 12Z forecast for 15:00 local is doing a lot more work than the 06Z forecast for 09:00 local.
How to fetch it. Both fields are available at any single coordinate via `/v1/site` and across the day via `/v1/forecast`, returned alongside `wstar_ms`, `cloudbase_agl_ft`, and the rest. For map use the new grid endpoint, which lets you ask for `horizontal_convergence_s1` over a bounded box for a chosen hour and render the line structure directly. The full field list is on the coverage page and the WRF model page describes how the diagnostics are computed.
Convergence is one of the few weather signals that genuinely makes a soaring forecast more useful than a generic weather forecast. Most fields tell you something about a single point in space and time. Convergence is fundamentally a field that only makes sense as a pattern. That is the field this update was about, and it is one of the things a soaring API can do that a general one cannot.