Analysis of SPI drought class transitions using loglinear models

A total period of 67 years of standardized precipitation index (SPI) data sets were divided into three periods of 22/23 years and a loglinear modeling approach has been used to investigate differences relative to drought class transitions among these three periods. The study was applied to several locations in Alentejo region, southern Portugal, and four drought severity classes were considered. The drought class transitions were computed for the three periods to form a 3-dimensional contingency table. The application of loglinear modeling to these data allowed the comparison of the three periods in terms of probabilities of transition between drought classes in order to detect a possible trend in time evolution of droughts which could be related to climate change. Results show that the drought behavior for the first and last periods is similar, both showing worse drought events than the second. If just the second and third periods were compared one could conclude that droughts were aggravating and easily this behavior could be attributed to climate change, supporting the common assumption that a trend for progressive aggravation of drought occurrence exists. Therefore, results are more consistent with the existence of a long-term natural periodicity; however, this hypothesis should be tested using longer time series.