A lot of people and customers are used to interpret the radar. Therefore it is really useful to have an expected radar for the first few hours.

Expected radar images are not only used by forecasters but as input for automatic applications as well. In the past expected radar images were produced by advecting the pixels of the radar scan with the steering level wind. The default steering level is set to 700hPa.

This technique works well in about 90% of the cases. However, there are significant cases where the algorithms are severely flawed:

1.      Showers or thunder storms track much different from the prevailing wind direction and speed. This is a well known feature of severe storms or clusters of storms such as an MCS or MCC. They tend to have their own circulation and move with speeds and directions that are independent of the ‘steering’ wind.

2.      Precipitation areas in the radar domain are driven by winds at various steering heights. This mainly happens in large radar domains. For instance, warm frontal rain at 700 hPa in the south and showers at 850 hPa in the north.

3.      Interaction of precipitation with high orography. This happens everywhere with significant topography, say hills and mountains over 500 meters, but especially over 1000 meters. Currently showers dissipate going down the height gradient, but this does not seem to be sufficient.

The new expected radar algorithm uses object recognition. This results in much more realistic expected radar images.

Below is an example of the real observed radar (top), the old expected radar (middle) and the new expected radar (bottom). When the background is dark, it’s an observed image, color means expected image.