Comparing Three Hydrological Models for Flash Flood Simulations in 13 Humid and Semi-humid Mountainous Catchments

Abstract

Flash flood disaster ranks top among all the natural hazards across the world due to its high frequency, severity and fatality. However, flash flood simulation is still challenging in small and medium-sized catchments with complex orography, flashy hydrological responses and poor observations. Three distributed hydrological models, i.e., TOP-Model, HEC and CNFF, are selected to simulate flash floods in seven humid and six semi-humid catchments in China, with consideration of water balance (RER), peak flow rate (REQ) and its occurrence time (TP), hydrograph variation (SNSE) and model uncertainty. Influences of five catchment attributes are further investigated on individual model performances. All three models perform satisfactorily in humid catchments, but less satisfactorily in semi-humid catchments. Water balance is well obtained by CNFF, followed by HEC and TOPModel. Peak flow rate and its occurrence time are most accurately captured by CNFF and HEC, respectively. Hydrograph variations are well reproduced by HEC and CNFF. TOPModel performs well for picking peak flow and hydrograph variation in humid catchments. Uncertainty interval is narrowest for HEC with average relative interval length at 95% confidence level being 0.78 ~ 2.53. Most observations are bracketed by uncertainty intervals for TOPModel (64.79% ~ 91.91% of total). Three model performance indices (i.e., RER, REQ, and SNSE) are mainly affected by drainage area and forest ratio across humid and semi-humid catchments, while TP performance is mainly affected by mean slope in humid catchments.

Predicting Flood Event Class Using a Novel Class Membership Function and Hydrological Modeling

“…Additionally, our predictions were superior to those of the Xinanjiang model in the Yellow (mean NSE : 0.54 vs. 0.44) and Huaihe (mean NSE : 0.53 vs. 0.48) River Basins; the HEC, TOPModel and China Flash Flood (CNFF) models in the Songliao (mean NSE : 0.58 vs. 0.49, 0.48, and 0.34) and Peal (mean NSE : 0.71 vs. 0.54, 0.34, and 0.34) River Basins. Our predictions were inferior to the HEC model in the Yangtze (mean NSE : 0.64 vs. 0.81) and Southeast (mean NSE : 0.66 vs. 0.82) River Basins (Wang et al., 2019; Zhai, Guo, Liu, et al., 2021; Zhai, Guo, & Zhang, 2021). This is because more flood events and more stations were considered in our study, for example, 844 events at 38 stations in the Yangtze River Basin and 90 events at five stations in the Southeast River Basin, which are 10.3 and 1.6 fold greater than the events predicted by the HEC model, respectively.…”

“…Under these circumstances, the prediction improvements were not considerable when using the class-based calibration strategy. (Wang et al, 2019;Zhai, Guo, Liu, et al, 2021;. This is because more flood events and more stations were considered in our study, for example, 844 events at 38 stations in the Yangtze River Basin and 90 events at five stations in the Southeast River Basin, which are 10.3 and 1.6 fold greater than the events predicted by the HEC model, respectively.…”

New insights on measuring soil water content in arduino plataform

“…Therefore, it is necessary to monitor the variation of water content in the soil, making it possible to correctly manage its use (Berger et al, 2020;Braz et al, 2020). Furthermore, at the level of river basin management, several important hydrological processes such as surface runoff, infiltration rate, erosion, groundwater recharge must be continuously monitored (He et al, 2019;Reichert et al, 2020) as disasters such as flash floods that rank first among all natural hazards worldwide due to their high frequency, severity, and mortality (Varlas et al, 2018;Zhai et al, 2021). However, hydrological modeling appears to be an effective tool for simulating the basin's responses to intense rainfall (Zhang et al, 2020) but for this to be possible, it must be determined when and in what situations a given monitoring area will be affected by a of flood.…”