Prediction of Streamflow Drought Using SPI and Markov Chain in Kharkheh's Basin

Drought is one of the most complex natural disasters in the world and it occurs whenever available water of a system wouldnt be enough for at least supplying one of the biological, economic and social requirements during a considerable period of time. Although there is not a universal definition of drought, drought can be defined in different disciplinary perspectives: meteorological drought, agricultural drought, hydrological drought and socioeconomic drought. Hydrological drought is defined as a significant decrease in the availability of water in all its forms such as streamflow, lake and reservoir level and groundwater level appearing in the land phase of the hydrological cycle. Hydrological droughts can have widespread impacts by reducing or eliminating water supplies, deteriorating water quality, restricting water for irrigation and causing crop failure, reducing power generation, disturbing riparian habitats, limiting recreation activities, and affecting a diversity of economic and social activities. The aim of this study is to assess and predict hydrological drought using SDI index and non-stationary Markov chain.
In this study, in order to determine hydrological drought state, for reference periods of 3, 6, 9 and 12 months in 4 hydrometric stations located on Kharkheh basin (Afarineh, Polchehr, Ghoorbaghestan and Polzal) over 1976-2009 (33 years), the Streamflow Drought Index (SDI) based on cumulative streamflow was used which is an index analogous to SPI. Four overlapping selected reference periods in each hydrological year are namely: October–December, October–March, October–June, and October–September (one complete hydrological year). Assuming that the underlying process possess the structure of a non-stationary Markov chain, after determining drought states, state transition probability was calculated in all reference periods and all stations.
While the correlations between SDI and SPI indices were poor, correlations between SDI values in different reference periods indicated that hydrological drought state in whole year can be predicted by October-March SDI index with high reliability. Another conclusion can be drawn from the present study is to use precipitation data instead of streamflow's in order to predict hydrological drought at the stations with high correlation between SDI and SPI indices.
The main features of the methodology are using a simple index called Streamflow Drought Index or SDI that is used to characterize severity of the hydrological drought for overlapping periods of 3, 6, 9 and 12 months (reference periods) within each hydrological year, considering five drought states, generating the matrix of state transition frequency for a selected pair of reference periods under the hypothesis of a Markov chain for the underlying state process, possibility of drought severity prediction in longer cumulative reference period using severity drought in prior reference periods and using of rainfall data instead of streamflow's in order to predict hydrological drought at the stations with a high correlation between SDI and SPI.