Improving runoff prediction by data assimilation in HBV hydrological model for upper euphrates basin

Advancing technology and increasing world population are valid reasons for using natural resources effectively. The necessity of efficient water resources management highlights the hydrology science. Improvements in modelling and forecasting studies contribute to optimal operation of hydraulic structures, decreased risk of flooding and drought and increased hydropower generation. In this study, Karasu Basin, which is a headwater of Euphrates River, is selected as a pilot region. Firstly, HBV hydrological model is calibrated and validated for the years 2002-2008 and 2009-2013 respectively and daily runoff values are forecasted for 2015. Data Assimilation (DA) technique, which is commonly used in atmosphere, meteorology and hydrology science in the last decade, is used to improve the forecast results. One of the 4-Dimensional variational (4D-VAR) methods, Moving Horizon Estimation (MHE) is selected among the variety of data assimilation algorithms. HBV model, which is integrated into Delft-FEWS platform, is configured to run with MHE. The model inputs and states are assigned as objective function variables utilized in DA application. Recently improved satellite technology products of MODIS and MSG-SEVIRI snow covered area and SSMI/S snow water equivalent are used in DA after preprocessing. Model initial states are updated by DA application and then short and medium range (2 to 9 days lead time) runoff forecasting is done for 2009-2013 water years with perfect forecast data sets. In addition, during 2015 water year snowmelt period, real time runoff forecasting is conducted using Numerical Weather Prediction and data assimilation approach. The results show that DA applications provide significant improvement on the performance of streamflow forecasts. Moreover, utilization of satellite snow products in DA applications increase the consistency of forecasted internal model variables compared to the observed snow data. Since the study includes up-to-date satellite snow products through a data assimilation method in real time forecasting which results in improved lead time runoff forecast accuracy, this could be considered as a pioneer application for operational hydrology in Turkey.