Characterisation and prediction of meteorological drought using stochastic models in the semi-arid Chéliff–Zahrez basin (Algeria)

Abstract

Study regionNorth Algeria. Study focusThe semi-arid to arid Chéliff–Zahrez basin faced several droughts with severe impacts on agriculture due to the high temporal and spatial distribution of rainfall. We explored the potential of the Standardized Precipitation Index (SPI), Markov chain models, the Drought Index and time series modelling to characterize meteorological drought. Time series of annual precipitation (1960–2010) from 65 meteorological stations across the basin were used. The basin was subdivided into five subbasins to account for spatial variability. New hydrological insights for the regionsThe analysis of the Standardized Precipitation Index showed few droughts in the period 1960–1970, whereas in the 1990s a multi-year drought occurred with SPIs as low as −2 (extremely dry) in many subbasins. The Markov chain analysis learnt that the probability of having two consecutive drought years appears to be higher in the southern subbasins. The Drought Index derived from transition probabilities indicates that the southern and the southwestern parts of the Chéliff–Zahrez basin are most drought prone. Time series modelling was applied to compute the SPI for different return periods (6‐17 years). Eleven models were tested and it appeared that the Asymmetric Power Autoregressive Conditional Heteroskedasticity (APARCH) approach was best performing based on several information criteria. For a return period of 17 years, the SPI is lower than −1.5 (severely dry) in many subbasins.