Mean Error (Bias) of QPF input
We might be curious about the nature of the greater errors in our precipitation forecasts. Here is the Mean Error (ME) of the QPF ensemble mean. ME is a measure of bias where a value of zero would indicate no bias.
When all data are taken together, there is a slight positive bias. But for increasingly large accumulation thresholds, the ME indicates a strong negative bias. So these verification scores suggest a slight overforecasting tendency for light precipitation and a strong underforecasting tendency for heavy precipitation.
Once again we see the transition from the QPF-based ensembles to the climatology-based ensembles. This is where the tendency for a rapid increase in negative bias of non-zero precipitation amounts is apparent.
