Abstract
Almost all water resources projects require past record of streamflow data and longer the record, the better the decision that can be taken during design or operation stage. However, in most of the cases, a long record of streamflow data is not available and it becomes essential to synthetically generate sequence of streamflow those are statistically similar to the observed data. Models to generate such sequences are available for a single river (single-site) and for both river and its tributaries (multi-site); however, comparative studies of these models needs to be done, before implementation to actual system. This study deals with the comparison of the performances of single-site and multi-site, seasonal streamflow generation models, applied to an existing river with tributary across which reservoirs were constructed. Since cross-correlation structure of the flows in a river–tributary system plays an important role in the integrated operation of the reservoirs, multi-site models are developed, as the cross-correlation cannot be preserved by the single-site models. Performances of the developed single-site and multi-site models are compared in terms of mean, standard deviation, skewness, serial correlation and cross-correlation of the observed and the generated series. The results indicated that cross-correlations are well preserved by the multi-site models only, whereas other statistical parameters, except serial correlation, are well preserved by both the single-site and multi-site models.
Highlights
Synthetic generation of streamflow is one of the major areas in stochastic hydrology
A comparative study on the performances of single-site AR model and multi-site AR model, for synthetic generation of flows in an existing river and tributary is presented in this paper
Results indicate that regarding preservation of mean, standard deviation and serial correlation, both single-site models and multi-site models produce very good results with each distribution, for both the rivers
Summary
Synthetic generation of streamflow is one of the major areas in stochastic hydrology. Since the flow through a river is inherently stochastic, sufficient information about this flow is almost essential in either design or operation of any water resources project. Such information is usually retrieved from the observed records of flows. Any system designed with such limited data becomes shortsighted and inherits the risk of being inadequate for the unknown flow sequences that the system may experience in future. To deal with this issue of limited available data, usually a synthetic generation model is used
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have