The northeastern depressed region of Bangladesh is highly susceptible to recurrent flash flooding due to excessive rainfall over these areas and in the upstream hilly regions. Two such severe pre-monsoon flash flood events occurred in 2016 and 2017. This research attempts to forecast both flash flood events using a coupled atmospheric-hydrological numerical weather prediction (NWP) model, namely the weather research and forecasting (WRF) model. The ARW (Advanced Research WRF) model is able to predict the rainfall over these areas with a lead time of 91 h. However, the discharge and water level are overestimated by the WRF-Hydro model. The model predicts a flash flood with a lag of approximately 12 h with respect to the highest amount of rainfall. The overall performances of the models were satisfactory. The two parameters, rainfall and subsequent discharge, which are required for delineation of lag time, were almost precisely simulated. Simulated values also had fewer errors, justified by the root mean squared error (RMSE) and mean absolute error values. The Nash–Sutcliffe efficiency criterion scores for model-derived discharge were close to 1.0, and the RMSE-observation standard deviation ratio scores were less than 0.5. This finding proves that the NWP models could be considered for forecasting flash flood events over selected areas of Bangladesh.

由于这些地区和上游丘陵地区降雨过多，孟加拉国东北部的低洼地区极易遭受反复发生的山洪暴发。2016 年和 2017 年发生了两次这样的严重季风前山洪事件。本研究试图使用耦合的大气水文数值天气预报 (NWP) 模型，即天气研究和预报 (WRF) 模型来预测这两次山洪事件。ARW（高级研究 WRF）模型能够以 91 小时的提前期预测这些地区的降雨量。然而，WRF-Hydro 模型高估了流量和水位。该模型预测山洪暴发与最高降雨量的滞后时间约为 12 小时。模型的整体表现令人满意。两个参数，几乎精确地模拟了描绘滞后时间所需的降雨和随后的排放。模拟值的误差也较少，这由均方根误差 (RMSE) 和平均绝对误差值证明。模型导出的排放的 Nash-Sutcliffe 效率标准得分接近 1.0，RMSE 观察标准偏差比率得分小于 0.5。这一发现证明，可以考虑使用 NWP 模型来预测孟加拉国选定地区的山洪暴发事件。且RMSE-观察标准差比值小于0.5。这一发现证明，可以考虑使用 NWP 模型来预测孟加拉国选定地区的山洪暴发事件。且RMSE-观察标准差比值小于0.5。这一发现证明，可以考虑使用 NWP 模型来预测孟加拉国选定地区的山洪暴发事件。