It is tricky to predict precipitation and even trickier to quantify the value of a forecast. For forecast ranges of 36 hours, in general around 80 % of the MetGIS predictions answer the question correctly if there's going to be more than 1 mm of precipitation in the 33-36 h forecast interval.
A glimpse of the performance of MetGIS related to precipitation forecasts is given by the adjacent contingency table, which refers to precipitation predictions for the Austrian mountain station Rax (1546 m), not far from Vienna. The numbers in the diagonal of the matrix represent the occasions where the observation was exactly in the class of the forecast (56.2 % in this example), indicating a nearly perfect forecast. However, since forecasts can also be considered to be quite decent if they can be assigned to classes next to the diagonal (cases where for instance 8 mm rain were observed, but 11 mm predicted), the accuracy of the Rax precipitation forecast can be calculated as 88 %.
Generally it should be noted that the reliability of the precipitation forecasts certainly depends on location and season. In weather situations where precipitation principally falls in form of short, but sometimes heavy showers (in the tropics all year long, in mid-latitudes mainly in summer), the exact start and end of rain is very difficult to predict, as is the precise location of the downpour (there can be heavy rain in one village and dry conditions in a close-by location).
|Observed 24-hour Precipitation|
|< 0.1 mm||0.1 - 1 mm||1 - 10 mm||> 10 mm||Total|
|< 0.1 mm||319||23||10||0||352|
|0.1 - 1 mm||147||60||41||4||252|
|1 - 10 mm||119||70||220||58||467|
|> 10 mm||1||3||46||70||120|