9:45 AM - 10:07 AM
[AHW26-04] Analysis of the impact of differences in geology and flood magnitude on model applicability using a vertical quasi-2D surface-subsurface flow model
★Invited Papers
Keywords:Richards' equation, rainfall-runoff model, quasi-2D model, geology
Flood forecasting requires a rainfall-runoff model that can make highly accurate predictions regardless of catchment characteristics or flood scale. We have developed a quasi-two-dimensional surface-subsurface flow model (hereafter, quasi-2D model) that physically represents saturated-unsaturated flows in the soil based on the vertical two-dimensional Richards' equation. This model simplifies the modeling by approximating the hydraulic gradient in the slope direction with the slope gradient. The quasi-2D model can accurately simulate rainfall runoff phenomena in the hillslope soil layer at a detailed resolution even at the river basin scale. In our previous study, the quasi-2D model was applied to the upstream area of the Kamo River in Kyoto, Japan. It was confirmed that the model achieved good reproducibility without parameter tuning by using observed soil properties.
In this study, the quasi-2D model was applied to catchments ranging in size from several tens to hundreds of square kilometers with varying catchment characteristics in Japan. Rainfall-runoff simulations were performed using consistent settings, including soil layer thickness and soil parameters. The differences in reproducibility between catchments and the relationship with geology, as well as the impact of flood magnitude on the applicability of the model, were analyzed.
Figure 1 shows the simulated and observed hydrographs for several catchments with different geology. In catchments with a high proportion of granite, the tendency for the model to overestimate discharge was more pronounced than in other catchments. It is known that granite weathering forms a thick weathering layer of sandy soil, and the permeability and storage effect of the bedrock would be high. Since the current model does not consider infiltration into the bedrock or loss, the computed discharge in catchments with such characteristics is likely to be overestimated, and the model would have low reproducibility.
Figure 2 shows a comparison of the simulation results and observed data for flood events of different magnitudes in the Takizawa Dam catchment, which showed particularly high reproducibility in Figure 1. Looking at the computed results for the soil layer thickness D=1.0 m, the difference between the observed and simulated values expands in the direction of underestimation during the recession part with increasing total precipitation. This trend was observed in almost all catchments, although the degree of difference varied from catchment to catchment. In addition, when the soil layer thickness was increased to 1.6 m, the computed discharge changed in a way that delayed the recession, improving the accuracy in some cases and worsening it in others. This suggests that the difference in reproducibility between events is difficult to explain only by the runoff mechanism through the hillslope surface soil layer considered in the current model, and that it may be due to other runoff processes, specifically relatively slow runoff from the groundwater aquifer.
Therefore, infiltration into the bedrock and outflow from the groundwater aquifer are important factors in the flood runoff process in mountainous regions, as well as water flow in the hillslope soil layer. In the future, we would like to increase the number of verification catchments to analyze in more detail the differences in model applicability due to geological characteristics, and realize a highly robust rainfall-runoff model by incorporating groundwater flow mechanisms into the model.
In this study, the quasi-2D model was applied to catchments ranging in size from several tens to hundreds of square kilometers with varying catchment characteristics in Japan. Rainfall-runoff simulations were performed using consistent settings, including soil layer thickness and soil parameters. The differences in reproducibility between catchments and the relationship with geology, as well as the impact of flood magnitude on the applicability of the model, were analyzed.
Figure 1 shows the simulated and observed hydrographs for several catchments with different geology. In catchments with a high proportion of granite, the tendency for the model to overestimate discharge was more pronounced than in other catchments. It is known that granite weathering forms a thick weathering layer of sandy soil, and the permeability and storage effect of the bedrock would be high. Since the current model does not consider infiltration into the bedrock or loss, the computed discharge in catchments with such characteristics is likely to be overestimated, and the model would have low reproducibility.
Figure 2 shows a comparison of the simulation results and observed data for flood events of different magnitudes in the Takizawa Dam catchment, which showed particularly high reproducibility in Figure 1. Looking at the computed results for the soil layer thickness D=1.0 m, the difference between the observed and simulated values expands in the direction of underestimation during the recession part with increasing total precipitation. This trend was observed in almost all catchments, although the degree of difference varied from catchment to catchment. In addition, when the soil layer thickness was increased to 1.6 m, the computed discharge changed in a way that delayed the recession, improving the accuracy in some cases and worsening it in others. This suggests that the difference in reproducibility between events is difficult to explain only by the runoff mechanism through the hillslope surface soil layer considered in the current model, and that it may be due to other runoff processes, specifically relatively slow runoff from the groundwater aquifer.
Therefore, infiltration into the bedrock and outflow from the groundwater aquifer are important factors in the flood runoff process in mountainous regions, as well as water flow in the hillslope soil layer. In the future, we would like to increase the number of verification catchments to analyze in more detail the differences in model applicability due to geological characteristics, and realize a highly robust rainfall-runoff model by incorporating groundwater flow mechanisms into the model.