Uncertainties of urban flood modeling: Influence of parameters for different underlying surfaces

Abstract

Rapid urbanization has dramatically increased the impermeable surface in urban area, which led to urban severe flooding and waterlogging in the world, especially in China and southeast Asia. There are more than 100 cities that suffered from urban flood every year since 2006, and more than 100 million citizens are involved in China. Urban flood mitigation is one of the most critical issues for both water administration and city management agency, in which urban flood modeling is vital and necessary. Whereas, there are relatively few data of waterlogging and runoff in urban area during flooding episodes to calibrate and validate the models, for there are usually few observation facilities installed in the cities. This paper used a combination of experiment and modelling to overcome the lack of reliable discharge data and be able to characterize the urban flooding problems in Xiamen Island, China. This paper simulated the urban flooding in Xiamen Island based on a hydrodynamic model coupled with hydrological model. The datasets of underlying surfaces were input to the model, including the terrain data, building plan, land use, etc. The uncertainty of the urban flood model was analyzed based on the generalized likelihood uncertainty estimation (GLUE) method with shuffled complex evolution Metropolis (SCEM-UA) sampling algorithm. The key parameters were evaluated by on-site experiment to reduce the uncertainties of the model, which could improve the accuracy of the model. If using the recommended parameter value range, the average relative error of flood depth was less than 27.2% at 90% confidence level. A typical rain pattern of 50 years return event was used for flood simulation. The results show that the main inundated areas (flooded depth more than 40 cm) are located in three groups: southeast to the Yundang Lake, around the Hubian Reservoir, along the Exhibition Road. The other inundated areas that less than 40 cm deep are scattered in some low-lying land of Xiamen Island. The main inundated areas simulated are consistent with the point survey of urban flooding, which verifies that the suggested model and the on-site experiment is effective and reliable for urban flood prediction.