DEM分辨率对山洪淹没模拟影响

DEM是表征地形状况的重要指标,在山洪淹没模拟中起着重要作用。以2016年7月19日发生在河北省石家庄市井陉县西部山区的特大暴雨山洪事件为研究对象,利用HEC-HMS水文模型和FLO-2D水动力模型,设置5种DEM分辨率方案(30、20、15、10、5 m),从淹没范围和淹没水深两方面分析了DEM分辨率对山洪淹没模拟的影响,并从模型运行时间角度对其运行效率进行了评估。研究表明:DEM分辨率越高,则淹没范围越小,平均淹没水深越大,即模型模拟精度越高,山洪事件的淹没范围及水深越接近于实际调查情况(DEM分辨率为30和5 m时,淹没范围及淹没深度的模拟精度分别为0.56和0.41、0.76和0.79);随着DEM分辨率的提高,尤其从10 m提高至5 m时,模拟结果差异减少,模拟精度的增幅也减小,淹没范围和水深的精度增幅分别为2.70%和3.94%;DEM分辨率越高,则模型运行时间越长,模型运行效率越低,5 m DEM方案下的运行时间为30 m DEM方案的700倍。综合考虑模拟精度及模拟时间,10 m分辨率的DEM对山洪淹没模拟、风险评估及预警预报等更为适用。

Effects of DEM resolution on flash flood inundation simulation

Digital Elevation Model (DEM), as an important tool for characterizing the terrain features, plays a major role in the simulation of flash floods. Taking the rainfall triggered flash flood event occurred in Guiquan Catchment located in the western mountainous area of Jingxing County, Shijiazhuang, Hebei Province on July 19, 2016 as the research object, we adopted the HEC-HMS model and FLO-2D model to investigate the influence of DEM resolution on flash flood inundation simulation from aspects of flooded area and flood depth by setting five resolution schemes, which are 30 m, 20 m, 15 m, 10 m and 5 m. Then the operational efficiency of the models were evaluated according to their respective runtime. The results show that a higher resolution of DEM leads to the decrease of flooded area and the increase of average flood depth, which in turn enhances the accuracy of the model simulation. That is, the higher the DEM resolution, the closer to the observed data the simulated flooded area and flood depth get. The accuracy of flooded area and flood depth by 30 m DEM and 5 m DEM was 0.56 and 0.41, 0.76 and 0.79, respectively. As the DEM grid size decreased, the simulation accuracy increased so slowly that the difference of the simulation results (i.e., the flooded area and flood depth) between 5 m DEM and 10 m DEM was excessively small. The accuracy of flooded area and flood depth increased only by 2.70% and 3.94%, respectively. However, with the decrease of the DEM gird size, the runtime of FLO-2D increased sharply. The runtime under the 5 m DEM was 700 times longer than that under the 30 m DEM, which made the efficiency of the model much lower. Our analyses suggest that for many flash floods, a 10 m DEM presents a rational compromise between accuracy and runtime for simulating the flash flood inundation and it is more suitable for the risk assessment, early warning and forecasting of flash floods.