Spatial Stochastic Forecasting of Snow Cover Using First Order Markov Chain

Evaluation of snow storage is of high importance in water balance studies and optimum operation of water resources in arid and semi-arid regions like Iran. Particularly in the river basins nearby Zagrous Mountains where surface water flows mainly consist of spring runoffs, stochastic forecasting of the snow storage at the end of the year is necessary. In this study stochastic forecasting of snow in river basins of the Karkheh, Dez, Karun and some parts of the Marun was investigated using the first order Markov Chain model. Snow cover data retrieved from NOAA-AVHRR satellite images between 1989 and 2004 were applied as inputs to the model. Two possible states were defined for each snow cover map including existence (1) and non-existence (0) of snow. Through applying the Markov Chain process, snow cover maps of the study area were predicted for March 2000 to 2004. Results show that stochastic forecasts of snow cover properly consist with satellite derived maximum snow cover maps.So that, not only the area of snow covered lands was successfully estimated, but also the exact location of the snow or dry covers was appropriately predicted in more than 80% of the pixels. The performance of the model was assessed using contingency tables and three measures including: Probability of Detection, False Alarm Ratio and Critical Success Index. Results reveal the promising capability of the first order Markov Chain model to forecast snow covered area.