Tuesday, August 8, 2017

Long-Range Eclipse Cloud Forecasts Are Pointless

Sometimes a thin layer of clouds is enough to ruin your day
Now that we're within t-minus 2 weeks, I'm already seeing weather forecasts for the eclipse on August 21st.

Don't waste your time reading them.

Even at relatively short ranges of a couple of days, predicting cloud cover is quite difficult, just ask anyone in Air Force Weather who has to provide detailed cloud cover forecasts for military operations.

Long ranges are even worse.   First, present-day computer models have very little skill at predicting the large-scale upper-level circulation at long lead times.  One common metric used to evaluate model skill is the 500-mb anomaly correlation, which assesses the accuracy of upper-level forecasts.  Based on experience, 60% is considered a minimum threshold for a "useful" forecast.  Although anomaly correlations have been steadily increasing over the past few decades, they remain well below 60% for a 10 day forecast (and are even worse at longer lead times).  An example is provided below for the ECMWF forecast system.

Source: ECMWF
Second, the ensemble spread is so large at these lead times that essentially you can't extract numbers for cloud-cover odds that are any better than climatology.  Below is the GEFS ensemble forecast for the morning of 21 August.  Some members have troughs near the west coast, but differ in amplitude and position.  One instead has a short-wave trough near Jackson Wyoming (lower right, GASP!).  Much different solutions.

Source: Penn State E-wall
Third, people like to infer cloud cover from a large-scale forecast (like those provided above), but this is actually quite difficult to do.  The odds of cloud cover might be somewhat higher downstream of an upper level trough and somewhat lower downstream of a ridge, but dry slots, topographic effects, convection, and other quasi-chaotic phenomenon are going to play a role, in some cases for the better (e.g., clearer) or for the worse (e.g., cloudier).

Finally, there is no medium range forecast model today that has sufficient resolution to explicitly predict cloud-scale processes.  At best, they resolve the processes contributing to layered clouds forced by large-scale processes.  However, during summer over the western U.S., we're dealing primarily with clouds driven by much smaller scale processes [stratus and stratocumulus (west coast), altostratus and altocumulus, cumulus and cumulonimbus, cirrus].   Imagine trying to forecast a west coast marine push or a monsoon thunderstorm with 2-week lead time and you see my point.  And don't assume that the eclipse path is too far north for monsoon influence.  Monsoon surges into the northern half of the western US happen.

Bottom line.  Keep calm and carry on.

No comments:

Post a Comment