LIMA, S. L.; http://lattes.cnpq.br/9724080345845333; LIMA, Santana Lívia de.
Résumé:
Drought is a climatic phenomenon that affects numerous regions around the world,
associated with water scarcity for prolonged periods, representing one of the most
complex and least understood natural hazards, assessing the risks caused by drought is
difficult as there is no globally accepted method to measure and qualify their impacts.
The effects of drought can be more drastic in regions susceptible to water scarcity, such
as semiarid regions, where the state of Rio Grande do Norte is inserted, for which it has
been suffering from impacts caused by the deficit of precipitation for several consecutive
years. Therefore, this work aims to analyze drought events in the state of Rio Grande do
Norte from 1950 to 2018 based on data from the SPEI drought index. In this study, the
SPEI index was used in different time scales, together with the multivariate statistics and
forecasting model. Multivariate statistics were applied in order to define homogeneous
drought regions. The results show the efficiency of the cluster analysis when identifying
regions with similar drought characteristics, where five groups were determined divided
by micro-regions, the SPEI-3 values reflect complex variations in drought conditions
according to their classification. The highest concentration of drought, according to its
duration, occurred in group G4 with the greatest presence of moderate drought, whereas
severe drought was identified in groups G1 and G5 and the extreme in groups G3 and G5,
with greater intensity, however with a shorter duration, so G4 was considered the driest
group and G1 the most humid. It was also possible to see the distribution of hydrological
drought for the two groups mentioned above as the most humid and driest, where G1
stood out for having a higher frequency of wet periods, whereas G4 showed a greater
evolution for dry episodes, increasing the number of episodes of hydrological drought.
Through the forecast made for groups G1 and G4, it was shown to be efficient, since both
groups had significant r2 values, and G4 had the best value (0.9061), the predicted series
followed the same pattern both of the validated and the observed, thus considering the
ARIMA model satisfactory for the groups analyzed, however, adjustments must be made
to minimize errors and improve the quality of the results, as it cannot be guaranteed that
there are ideal forecasting models for this variable.