Then how do we approach forecasting extreme rainfall , Probabilistically?
Krzysztofowicz approach
- relies on forecaster experience and verification to calibrate forecasts
Ensemble approach
- using a coarser mesh, how do you assess risk of very heavy rainfall? Needs to be run at sufficient resolution to explicitly predict convection. Need to perturb physics.
- if 3 of ten members predict rain what does it mean?
MOS or other statistical approaches (MOS, LAMP). Provides a probability of reaching a certain threshold at a given point based on statistical relationships with various observed and model forecast meteorological parameters.
- Well calibrated especially for lighter thresholds (the ones that are observed most)
- for heaviest thresholds will always have very low probabilities
What about other approaches. For example the probability of receiving 4 inches or 5 inches somewhere within a predefined circle or ellipse.