Flood modeling for the Neka catchment in Mazandaran province, Northern Iran

Flood damage is a very serious disastrous problem in northern Iran; specifically at the Neka river basin. Contemporaneous data and information, as well as predictions and pre-emptive warnings on impending flood threats has become a prominent focus in flood protection and alleviation in this province. Loss of life along with extensive destruction to properties and infrastructure affected by large floods and inundations is what instigated this movement to strengthen flood preventative measures and systems. An integration between a distributed hydrological model (WETSPA) and a hydraulic GIS base model HEC-RAS, is the focus of this study to simulate and predict flooding in the basin. This is done by simulating flooding, thereby making it possible to identify precursors and patterns in flooding and also making flood predictions for this basin a more exact science. A model primary function of WETSPA is to provide data input to the hydrologic model by replicating and predicting the Neka river basin’s flow in its upstream sub-catchments. Using data of stream flow of all seasons from 1986 to1999, the hydrologic model was calibrated and then validated for the 2002 to 2004 flood seasons. It has proven to be accurate with the hydrographical predictions showing a baseline of 0.87 on the Nash-Sutcliffe efficiency for high flow and square correlation coefficient 0.974 and modifiedcorrelation coefficient 0.747 with an aggregated measure of 0.85, the performance is deemedexcellent forthe validation phase. The HEC-RAS hydraulic model, on the other hand, imitates flow and water profiles in the Neka river basin’s downstream flood plain. Hydrograph phases studied during the flood seasons of 1986-1999 and from 2002-2004 were used to calibrate and verify the hydraulic model respectively. Simulations of peak flood stages and hydrographs’ evaluations are congruent with studies and observations, with the former showing errors of below 0.107m.A 3D flood mapping for the flood plain is then constructed from the projected water profiles. The assimilation of the hydrological model (WETSPA) withthe hydraulic (HEC-RAS) model is found to be suitable in gauging flood threats on top of flood forecasting at the Neka river basin from the research findingsobtained. This is done by developing an effective flood prediction method,thereby producing a system for the management of its processes.The application and integration of the aforementioned models were done by incorporating data handling utilities and interfaces from this process management system. Collected stream flow data from calibrations cycles (flood seasons 1986 to 1999) and verification cycles (flood seasons 2002 to 2004) in the catchments were compared to the simulation results, and then used for the calibration and verification of the WETSPA model, which was performed specifically for upstream watersheds. Flow volumes, concentration time and peak discharges as well as hydrographs predictions reflected in the results established that the model produces good assessments and high accuracy estimate. Using the WETSPA model to calculate upstream inputs and employed as a boundary conditions, the HEC-RAS model is applied in the downstream flood plain of the river basin. The predicted results show that the system can effectively perform WetSpa and HECRAS calculations and forecast flood water levels. Nash-Sutcliffe effectiveness (CR3) is more than 0.92 along with elevated levels of water which were created with some effectiveness (CR5) of 0.94 for the validation period. Next the multilayer perceptron (MLP) Artificial Neural Networks (ANN) and WETSPA models were tested and compared in a comparative study of daily flow estimates. Results of the study basically showed that both models are capable and practical in calculating and estimating the hydrologic response. To determine their efficiency and practicality in hydrologic predictions, both models were used to construct flow predictions at the Neka basin. Thesquare correlation coefficient ( R2)and root mean square error (RMSE) for the WETSPA model was in a range of 0.964 to 0.974and 0.00184 to 0.0021 mm/h respectively while that for the ANN model was in a range from 0.917 - 0.960and 0.0022-0.0040 mm/h. It was shown that WETSPA was more accurate than ANN atshort-term (2002-2004).However, the ANN model showsgreater R2 and the lower error RMSE for forecasting at long-term (1986-1999).

Preview PDF FK 2011 101R.pdf Download (1MB) | Preview

Actions (login required)

View Item