Caractérésation,variabilité et prévision de la sécheresse dans le Nord-Est Algérien

Abstract

Drought is a natural phenomenon with harmful effects on ecosystems and human related activities, mainly agriculture. The objective of this thesis is the analysis and prediction ofdrought in northeastern Algeria using monthly precipitation data of 123 rainfall stations from1960 to 2013 and respective interpolated temperature data from 1979 to 2013. Seven indicesof drought that use precipitation data as single input were compared; One was theStandardized Precipitation Index (SPI) which is more preform for all types of climate at various time scales. The Markovian forecasting models for the order I and II were used toestimate the different probabilities of drought appearance in the seven plains for both annual and seasonal time scales based on the results of the SPI-12 and SPI-3. The probability ofhaving two successive dry events in the southern plains was higher particularly on Batnawhere it reached 68%, and contrarily to the plains of Annaba (48%) and Setif (43%) located in a sub-humid region, where the probability was smaller. The results of the order II process were similar to the order I. To better identify and analyze the behavior of drought, otherindices that include other parameters than precipitation were used. The RDI index based on the ratio of precipitation to potential evapotranspiration was also used with SPI for the time scales of 3-, 6- and 12-months. Spatial and temporal patterns of droughts were obtained usingPrincipal Component Analysis in S-Mode with Varimax rotation applied to both SPI and RDI. For all time scales of indices, two main components were retained; the first one identified thenorthern region characterized by a humid and moist sub-humid climate, and explained morethan 40% and 43% of the drought spatial variability in SPI and RDI, respectively. The second component explained 27% and 28% for the same indices, representing the southern plainswith arid and semi-arid climate. The Modified Mann-Kendall test was used to assess trends ofthe RPC scores, showing generally not significant trends for decreasing drought occurrenceand severity in both identified drought sub-regions for all time scales.