Evaluation of the climate outlook forum's seasonal precipitation forecasts of southeast South America during 1998-2002

GUILLERMO J. BERRI; PABLO L. ANTICO; LISA GODDARD

INTERNATIONAL JOURNAL OF CLIMATOLOGY

Wiley

Lugar: Chichester; Año: 2005 vol. 25 p. 365 - 377

0899-8418

The series of 16 Climate Outlook Forums (COFs) seasonal precipitation forecasts for Southeast South America since December 1997 are verified. The COF forecasts consists of a probabilistic tercile distribution of precipitation for the upcoming 3-months period, over a domain in South America between 20°S and 40°S, to the east of the Andes Mountains. The COFs forecast is the result of the consensus agreement between atmosphere model predictions, physically based statistical model predictions, results of diagnostic analysis and published research on climate variability over the region, and expert interpretation of this information in the context of the current situation. The validation is conducted by means of Hit Score (HS) that expresses the percentage of forecasts verifying in the dominant predicted category, and the Ranked Probability Skill Score (RPSS), which measures the difference between the probability distribution of the forecasts and the observed categories. The HS reveals that over most of the region the categorical agreement is low and close to 33%, which is the result that would be obtained by a 3-category random forecast, and only sparse regions achieve a categorical agreement between 40% and 50%. The regions with positive RPSS (i.e., forecasts better than climatology and potentially useful for applications), represent a 30% of the total area considered by the COFs. There are significantly large regions with negative RPSS i.e., forecasts worse than climatology. The result of the COF forecasts is compared with that of the IRI Net Assessment forecasts, the only source of physically based seasonal forecasts regularly available at the COFs discussions. There is general coincidence in the regions with positive and negative skill. Although one might expect the consensus of the COF forecasts to improve upon the skill of the IRI forecast, this was not the case during this period.