Franklin Paredes, Edilberto Guevara Pérez


Droughts occur when rainfalls diminish or cease for several days, months or years. In the last five years several meteorological droughts have occurred in Venezuela, impacting negatively water supply, hydro¬power and agriculture sectors. In order to provide institutions with tools to manage the water resources, a probabilistic model has been developed and validated to predict in advance the occurrence of meteorological droughts in the country using monthly series of 632 rainfall stations. The standardized precipitation index (SPI) was used to identify dry events of each rainfall series. A principal component analysis associated to a geographic information system was used to define geographically continuous homogeneous sub-regions (HS) for the values of SPI. For each HS a representative station was selected (reference station, RS). A lagged correlation analysis was applied to the SPI series of the RS and the corresponding series of anomaly indices of 10 macroclimatic variables (MV). The four MV with higher correlation in each RS were orga¬nized into three levels (–1, 0 and +1), using the quartiles Q2 and Q4 as values of truncation. The SPI series are expressed in four ranges: non-dry, moderately dry, severely dry and extremely dry. The conditional probability of occurrence of the four ranges of SPI was determined in every combination that can occur in the four VM best correlated. The resulting model in each RS was validated using the SPI series from 20 meteorological stations operated by the Servicio de Meteorología de la Fuerza Aérea Venezolana (Mete¬orological Service of the Venezuelan Air Force) which were not used in the development of the models. Results indicate that models detected the occurrence of ES with an accuracy ranging from 85.19 to 100%; the success is directly proportional to the length of records used in the development of the model. This methodology could be applied in any country that has long, continuous and homogeneous


Meteorological drought, Venezuela, forecast models.

Full Text: