GOES-R cloud base altitudes and the challenge of multi-layer clouds

A new product in this year's Winter Experiment is the ACHA cloud base altitudes generated by the ACHA groups. Ceiling and visibility is a significant hazard to aviation, particularly in places like San Francisco, Chicago, and other busy hubs. While having observations of ceiling is immensely valuable to aviation forecasters, actually using satellite to estimate bases of the cloud that are obscured in visible or other imagery presents a unique challenge. This morning in the Southeast U.S. proved that.

Below is a snapshot of the ceiling heights at 1255 UTC overlaid with observed flight rule conditions:

20160223 1255 UTC GOES-R cloud base altitudes overlaid with observed flight rule conditions (where pink=LIFR, red=IFR, blue=MVFR, white=VFR). The dashed lines indicate areas of interest during the experiment today

Note in particular the area around Charlotte. To the east, observations indicated LIFR conditions extending along the coast from Georgia to southern Virginia (pink dashed line). IFR conditions were reported along the Appalachians further west (red dashed lines), with MVFR beyond that (cyan dashed lines). In the region of LIFR, ceilings of >500 feet were reported.

Interestingly, though, cloud base altitudes in the area of LIFR conditions were showing 12,000+ foot ceilings, with 10,000 - 12,000 feet further inland in the IFR areas. Various airports were in fact indicating the >500 foot ceilings and the GOES-R IFR probabilities were high in the area (below).

20160223 1245 UTC IFR probabilities overlaid with observed flight rule conditions

Note though, the smooth look to the probabilities. This indicates the presence of higher clouds obscuring the satellite data. A visible image at the same time confirmed the presence of a higher cloud layer.

20160223 1245 UTC visible image

Again, this case shows the challenges of estimating cloud base altitudes when multiple cloud layers are present. For single layer clouds (i.e. the are over Alabama) the algorithm has shown promise. However, in cases like this, it appears that the algorithm estimated the base of the top layer of clouds instead of the layer beneath.