Investigation of Markov Chain Uncertainty for Forecasting Hydrological Status Based on Meteorological Status

Examining and quantifying the degree of uncertainty in the prediction results of the models is the most important step before using the results of the models in water resources decisions making. In this research, the Markov chain technique was used to predict the hydrological status of the basin according to the amount of precipitation in the previous time step. The present study aims to determine the degree of uncertainty of prediction using the confidence interval of the probability of occurrence of events in different meteorological and hydrological conditions. To assess the uncertainty of the predictions by the 2D Markov chain transfer probability matrix, 20 periods with similar characteristics to historical conditions were simulated by the Monte Carlo method. To determine the uncertainty in estimating the hydroclimatology transfer matrix components, the nonparametric method of ratios for large samples and the exact method based on the sign test for the median of the predicted matrix components were used. The results of the long-term hydroclimatology matrix showed that the hydrological conditions tended to remain normal. The results of uncertainty analysis also indicated that the sign test, as a non-parametric method, in assessing the uncertainty of the Markov Chain acts better, and because the range of probabilities of transfer from different meteorological states to hydrological states is almost the same for all stations, the study basin is considered meteorologically and hydrologically homogeneous.