Research On Rainfall-Runoff Modeling Of QingJiang Valley

Flood is one of the most primary natural calamities worldwide, so how to reduce the side affect of flooding has been paid much attention throughout the world. The accurate runoff prediction and flood simulation help to overcome the flood and reduce the disasters lost. Rainfall-runoff prediction is a very valid approach to enhance accuracy of flood prediction and prolong the duration of prediction. It is very important not only in theory but also in practice to reasearch the physical forming reason of the rainfall and to search for new theories and methods about models of the rainfall-runoff prediction. Through "Digital Qingjiang" project, this paper focuses on the related subjects of rainfall-runoff prediction; meanwhile, Qingjiang rainfall-runoff prediction is discussed in detail.Firstly, the existent researches on rainfall-runoff models are carefully reviewed. After analyzing the fundamental principles of rainfall, runoff and evapotranspiration subsystems in water cycling, a conceptional rainfall-runoff modeling that can be applied to the Qingjiang valley is proposed. The whole valley has been divided into seven sub-valleys in the model, which could separately do the rainfall-runoff prediction.Secondly, the primary premise of the rainfall-runoff model is to process the collected raw information. By referring the tessellation weight-interaction calculating technology based on DEM valley boundary, the control acreage and weight coefficients of each rainfall station are calculated, and then the average rainfall amount can be obtained. This practice has made it possible that the space-time interpolation method can be used for those rainfall stations that lack such information.Then, theoretic basis, mathematic models and basic algorithms of Xinanjiang model are fully discussed in this paper. Due to the difficulty of automatically optimizing parameters, the layered parameter optimization and genetic algorithm to adjust all the parameters are put forward. After building up the model of Sanshuiyuan Xinanjiang model for Qingjiang valley, a novel method of estimating parameters to the model is presented. Simulating and practical results show that the model of Sanshuiyuan Xinanjiang can fit in Qingjiang valley quite well and estimation parameters of the model by using genetic algorithm is not only reasonable but also very helpful in increasing the precision of the prediction.In addition, by analyzing the structure of Tank model, a Tank model is proposed, which associats with the characteristics of Qingjiang valley. Modeling and test are carried out in a group of series tanks separately in seven sub-valleys. After emending the parameters to every single tank using the hybrid genetic algorithm, the general runoff by way of adding all the runoff in every tank in the sub-valley has been acquired.Finally, the implement effect of the Xinanjiang model and Tank model applied in the Qingjiang valley are compared. By comparing the advantages and disadvantages of each model as well as their applied circumstances, both two models stand as valuable examples for further researches in the rainfall-runoff modeling.