Study On Structuring ANN-and-TOPMODEL-based New Rainfall-Runoff Model

Natural disaster is one of the most vital problems faced by human beings. Recently, global extreme climate occurred frequently and flood caused by global warming raged all over the world. China lies at middle latitude, and flood here is centralized and intensive in flood season, and flood in mountains changes sharply. With the exacerbation of China’s water resources and environment crisis, it is urgent to realize the monitoring and control of regional water resources through hydrolog ical methods.As two mature runoff simulation tools. ANN and TOPMODEL have their own advantages. TOPMODEL is able to be used for the description of physical process but is difficult in capturing the nonlinear factors in runoff process, hence the powerful nonlinear processing capacity of ANN is integrated in the underground runoff generation plan of TOPMODEL, which is of simple structure and clear physical concept, in the essay, by taking early rainfall runoff information as the input and the daily initial groundwater level as output of ANN, to carry out parameter calibration for the entire integrated model by virtue of particle swarm algorithm, and thus to structure a new rainfall-runoff model which is not only of physical significance but also able to reflect the nonlinear relation in environment process.In order to evaluate the applicability of the new rainfall-runoff model under different scales, climates and ecological conditions of catchments, Youshuijie catchment, on the upstream of Hanjiang River watershed in warm and humid climatic region, Linyi catchment, the subcatchment of Huaihe River catchment in semi-humid and semi-arid climatic region, and Yingluoxia, subcatchment of Heihe River catchment in arid and semi-arid climatic region, are taken as the study regions to carry out rainfall-runoff simulation and relatively all-round test and evaluation to the new rainfall-runoff model, and make contrastive analysis with semi-distributed TOPMODEL. And the study result shows that:(1) In Youshuijie catchment in humid climatic region, the new rainfall-runoff model and TOPMODEL all have relatively satisfied results on runoff modeling and the coefficient of determination of the new rainfall-runoff model is higher than TOPMODEL.(2) In Linyi catchment, the performance of both of the two models in the region is worse than that in Youshuijie catchment, but the new rainfall-runoff model performs better than TOPMODEL.(3) In Yingluoxia, although the performance of TOPMODEL is far from satisfactory, but the new rainfall-runoff model has outstanding performance, and the coefficient of determination reaches over0.8, which shows the contribution of the new rainfall-runoff model in the low runoff region.In conclusion, the new rainfall-runoff model has relative outstanding performance under different catchments of different scales, climates and ecological conditions, it is always better than TOPMODEL, especially in in arid climatic region. The study shows that it is successful and feasible to integrate ANN into TOPMODEL, and the method icould be easilyextended to other hydrological models. In addition, since the new rainfall-runoff model is of simply structure, less parameters and high efficiency, it is easy to be coupled with large-scale climate mode and numerical weather prediction business mode, meanwhile it is also easy to be linked with the models of ecology, agroforestry, environment, soil and water conservation, etc.