How to use these maps

These maps may be used to assess whether an upcoming precipitation event may be unusual and therefore even heavier than predicted by the model or whether the model predicts such events too often.

For instance, the AVN model climatology for April and the corresponding CPC observations at the same resolution show that the AVN predicts 0.5" amounts in a 24-hour period too often over most of the country, except over a small region where these events actually occurred often. Thus, the model generally has poor skill at this time of year for this amount except in the areas getting repeatedly hit, and a forecaster may expect less precipitation. However, these data cannot distinguish between whether the model tends to produce the right number of systems but with too broad coverage or whether it makes good predictions of areal coverage but makes some systems that never actually form.

If the number of observed events of interest is small, the sample may be insufficient to draw any conclusions. Then it may be helpful to look at the previous and following months and the annual totals. The individual months are primarily useful for giving an indication of seasonal shifts, while the annual totals will give the most overall reliable results.

The kinds of questions these data can help with and an example of a rare-event forecast are given on the Usefulness/Motivation page. More detailed tips are given on the Comparison of forecasts and observations page.

Remember, the whole forecast scenario has to be evaluated. One cannot consistently make good forecasts by taking shortcuts such as blindly multiplying the model QPF by some factor.