A prediction of future flows of ephemeral rivers by using stochastic modeling (AR autoregressive modeling)

Abstract

There are different flow prediction models such as Autoregressive models, Autoregressive moving average models, first-order autoregressive-moving average models, etc. The main purposes of this dissertation were to fit a model to represent a river flow data of 10 rivers in the Northern part of Cyprus. The modeling was built on the estimate of parameters, modeling the residuals, generating synthetic river flows, and checking for the goodness of fit to the monitored data. Finally, the findings were used to evaluate the synthetic series for future flow predictions. The study on available data demonstrated that the (AR) model was an efficient and reliable technique in which, the model identification technique was supplemented by the Akaike's information criterion (AIC) in order to decide the type and the order of the model. The Box-Pierce Porte Manteau test is used to check the dependency of residuals. it is recommended to generate stochastic modeling for the downstream drainage areas of the 10 rivers in which the surface geology totally changes and surface flow turns to be a subsurface flow due to the gravel and pebbles distributed all around the riverbeds.