Fig. 3: Regional PM10 patterns of correct and incorrect local dust event forecasts.
From: Deep multi-task learning for early warnings of dust events implemented for the Middle East
Comparison of the composite averages of regional PM10 fields between dust event samples that were correctly (left column), and incorrectly (right column) classified in the 24-h forecast. The first three rows represent periods in which the meteorological input is available for training. The bottom row represents the time of the event itself, which is unavailable for training.
