Premium
Flood forecasting for River Mekong with data‐based models
Author(s) -
Shahzad Khurram M.,
Plate Erich J.
Publication year - 2014
Publication title -
water resources research
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.863
H-Index - 217
eISSN - 1944-7973
pISSN - 0043-1397
DOI - 10.1002/2013wr015072
Subject(s) - flood forecasting , surface runoff , flood myth , environmental science , calibration , index (typography) , mekong river , meteorology , discharge , hydrology (agriculture) , computer science , statistics , mathematics , geology , geography , drainage basin , archaeology , geotechnical engineering , structural basin , ecology , paleontology , cartography , world wide web , biology
In many regions of the world, the task of flood forecasting is made difficult because only a limited database is available for generating a suitable forecast model. This paper demonstrates that in such cases parsimonious data‐based hydrological models for flood forecasting can be developed if the special conditions of climate and topography are used to advantage. As an example, the middle reach of River Mekong in South East Asia is considered, where a database of discharges from seven gaging stations on the river and 31 rainfall stations on the subcatchments between gaging stations is available for model calibration. Special conditions existing for River Mekong are identified and used in developing first a network connecting all discharge gages and then models for forecasting discharge increments between gaging stations. Our final forecast model (Model 3) is a linear combination of two structurally different basic models: a model (Model 1) using linear regressions for forecasting discharge increments, and a model (Model 2) using rainfall‐runoff models. Although the model based on linear regressions works reasonably well for short times, better results are obtained with rainfall‐runoff modeling. However, forecast accuracy of Model 2 is limited by the quality of rainfall forecasts. For best results, both models are combined by taking weighted averages to form Model 3. Model quality is assessed by means of both persistence index PI and standard deviation of forecast error.