Local level flood forecasting system using mathematical model incorporating WRF model predicted rainfall

Abstract

Bangladesh is one of the most flood prone countries in the world. To mitigate the recurrent flood damage, flood forecasting is used as an efficient non-structural flood induced disaster management system in the country. National level flood forecasting system is forecasted mainly for river station water level hydrograph and very coarse level inundation map whereas this study objective is to develop 5 days forecasted flood inundation map and hydrograph at house level flood information at Rowmari Upazilla of Kurigram district. This study area is surrounded by the mighty Brahmaputra River and flashy Jinjiram River. As a result flood occurs every year and destroys agricultural products of large areas, causes death, damage to property, environmental pollution and destruction to roads and bridges. A reliable forecast with longer lead time is a way of reducing the damages. In this study a weather prediction model (WRF) was coupled with a hydrologic model and a hydrodynamic model for predicting floods at Rowmari upazilla of Kurigram district. At first a HEC-HMS continuous hydrologic simulation model is developed for the Brahmaputra basin based on Soil Moisture Accounting (SMA) algorithm and excess rainfall was transformed to direct runoff using the Clark unit hydrograph technique. At the same time, Hydrodynamic Model HEC-RAS 4.1 is setup with geometry data and observed boundary data. This hydrodynamic model is simulated with unsteady condition and calibrated and validated with observed water level. WRF 3.2 weather model was configured and used to predict rainfall over the basin 120 hours into future. Output of the weather model is incorporated with calibrated and validated hydrologic model HEC-HMS 4.0 and simulated every day during monsoon to forecast discharge at Bahadurabd. This study has developed three mathematical relations between Bahadurabad station to other boundary of hydrodynamic model for forecast boundary generation. Then hydrodynamic model is simulated every day using forecast boundary to generate flood inundation map and forecast hydrograph at Rowmari Upazilla of Kurigram. The HEC-HMS application produced satisfactory performance taking into consideration lumped parameters. Also, among the optimized parameters; Maximum Soil Infiltration Rate, Surface Storage Capacity, Initial Surface Storage and Tension Zone Storage Capacity show the higher sensitive. The estimated NSE value for the calibration and validation period is 0.85 and 0.82. Hydrodynamic Model (HEC-RAS) performance during calibration and validation period in terms of R2 and NSE against observed water level data is found to nearly 1. The Manning's roughness coefficient (n) and the coefficient of expansion/contraction (k) are key parameters to calibrate of HEC-RAS model. Analysis of forecast performance indicates that the forecast for the first 3 days are good and next 2 days are average to poor according to BWDB guideline. This developed flood forecasting system is capable of predicting the inundated area of Rowmari Upazilla during a monsoon season.