For an urbanizing watershed, terrain properties are in changing, and the parameters of a hydrological model employed for simulating and forecasting the watershed flood processes are also in changing accordingly, this is called model parameter dynamics. This study proposed a methodology that considers this parameter dynamics, and tested in an urbanizing watershed. It has been found that the parameter of physically based distributed hydrological model is LUC stationary.

There are no long-term reasonable rainfall data to build a hydrological model in karst river basins to a large extent. In this paper, the PERSIANN-CCS QPEs are employed to estimate the precipitation data as an attempt in the Liujiang karst river basin, 58 270 km², China. An improved method is proposed to revise the results of the PERSIANN-CCS QPEs. The post-processed PERSIANN-CCS QPE with a distributed hydrological model, the Liuxihe model, has a better performance in karst flood forecasting.

Quantitative precipitation forecast produced by the WRF model has a similar pattern to that estimated by rain gauges in a southern China large watershed, hydrological model parameters should be optimized with QPF produced by WRF, and simulating floods by coupling the WRF QPF with a distributed hydrological model provides a good reference for large watershed flood warning and could benefit the flood management communities due to its longer lead time.

The distributed hydrological model has not yet been applied in large watershed flood forecasting due to some limitations. By proposing a method for estimating channel cross section size with remote sensing data, employing the PSO algorithm optimize model parameters and running the model on high-performance supercomputer with parallel computation technique, this article successfully applied the Liuxihe model in a larger watershed flood forecasting in southern China at high resolution.