What do AGREE and SPLIT mean on the map?

On the map in the instrument at /app, the top left corner has a small chip. It reads AGREE or SPLIT with a dot, followed by a count such as 3/3 models. It does not show a wind speed. It follows the hour you have selected on the time strip, so scrub to another hour and it can change. If there is not enough model data for that hour the chip shows a dash instead of either word. Cab mode does not have this chip. See reading the live map for the rest of the map.

The chip looks at three deterministic forecasts for that hour: ICON, GFS and the ECMWF model. It uses their 10 m wind only, so it does not change when you change your working height. It measures how far apart their wind speeds are and how far apart their directions are, and it also lowers the score a little for each hour of lead time, about 0.12 of a point per hour. The result is a score from 38 to 94. The panel calls it an uncalibrated model spread indicator. It is a measure of disagreement, not a probability and not a verdict on your limit.

The chip shows AGREE when the score is 60 or more, and SPLIT when it is below 60. The cut-off is fixed and you cannot change it. SPLIT means the three models are far enough apart, in speed, direction or both, that you should not lean on the single primary number. AGREE means the models are close, but a close forecast can still be wrong, and it says nothing about your limit. The same score also feeds the Model spread tick on the agreement spine, which uses a stricter cut-off of 70. A score of 65 reads AGREE on the chip and shows a warning tick on the spine. See the agreement spine ticks.

The count shows how many of the three models gave both a speed and a direction for that hour. With 3/3, all three were used. With 2/3 the score is still worked out from two models, so AGREE or SPLIT is based on less. With fewer than 2 there is no score and the chip shows a dash, and the spine's model tick says there is insufficient model coverage. A dash does not mean calm. It means the comparison could not be made. See unknown and missing data.

Because the score drops with lead time, hours several days ahead will show SPLIT more readily than hours in the next day, even for similar weather. Fast-moving fronts and coastal or hilly sites also tend to pull the models apart, since each one places the timing or the local detail slightly differently. This is common and does not mean the instrument is broken. Use it as a reason to look at the spread rather than a single line. The learning page why forecasts disagree explains the causes in more detail.

Open the exceedance fan in the inspector and look at the width of the band between P10 and P90 around your limit. Then read the observation line to see what the nearest station is reporting against what the model said for the same hour. Check the model chips under the readout, which list the primary model and the ensembles. If the primary line and the P50 of the ensemble sit on opposite sides of your limit, the spine's model tick turns ember and says split. See when models disagree and read the exceedance fan.