The Influence of Accurate Lag Time Estimation on the Performance of Stream Flow Data-driven Based Models

The timing of variables for stream flow prediction models remains a common problem, particularly with regards to stream flow data-driven based models. Accurate estimation of lag time (Lt), which is representative of time interval, is a potential solution to this problem. The main objective of this paper is to explore possibilities of improving the performance of stream flow data-driven based models by developing a new and simple hydrological approach to accurately estimate Lt between upstream and downstream stations in the Selangor River Basin. The Lt was estimated by means of several empirical formulas along with the new proposed approach that is based on observed hourly stream flow records and water level data. The estimated Lt was used to select the input and output variables for two Multiple Linear Regression models (MLR)—as examples of data-driven based models—to explore the capacity to enhance stream flow data-driven based model performance. An assessment of the two models’ result performance indicates that the MLR model, which is dependent on the proposed method, produced roughly 5 % overall improvement in root mean squared error (RMSE), mean absolute error (MAE) and mean absolute percentage error (MAPE). Not only do the results highlight how effective the proposed approach is in estimating Lt to improve the stream flow MLR model’s performance, but they also serve as a starting point to further study potential improvements from other types of data-driven based models. Copyright Springer Science+Business Media Dordrecht 2014