Study On Flood Forecasting Method Based On Initial State Variable Correction

Flood forecasting is one of the important non-engineering measures for flood control,and accurate and real-time forecasting can reduce loss to the lowest.The most widely used hydrological model in flood forecasting is the conceptual model now,whose prediction errors are from 5 aspects of the model input,model structure,model parameters,the measured value and the initial state variables.When a flood forecasting model is selected,the model structure is determined,and the quality of data input and measured value are the same,so it is one of the effective ways to improve the accuracy of flood forecasting method by modifying the initial state variables to make it more consistent with the actual situation,but now there is no effective method to correct the initial state variables.To solve this problem,a flood forecasting method based on the modified Initial State Variable Correction(ISVC)is proposed in this paper,and different types of floods of the basin are selected to test the effect of ISVC method.The main content of the paper are as follow.(1)The necessity and feasibility of modifying the initial value of state variables.In this paper,the characteristics and shortcomings of the current error correction methods for flood forecasting are fully analyzed,and then it points out that there are some difficulties and limitations in revising model input,model structure,model parameters and measured values.At present,the value of initial state variables is got from the continuous calculus.In the process of calculating the values of state variables,the errors of input and the parameters of the model will inevitably lead to errors in the initial value of state variables after continuous calculating,especially for the first flood in the year or the flood after a long-time drought.The modification of the initial value of state variables is mainly based on manual experience,which has some shortcomings such as high professional requirements and strong subjectivity.Therefore,it is necessary to propose a more scientific and effective method for the modification of the initial value of state variables.Through the analysis of the influence of the initial value of state variable on the flood simulation of single flood and continuous flood,it is found that the more accurate the initial values of state variables are,the higher the accuracy of flood forecasting will be in flood forecasting.At the same time,the initial values of state variables have continuous and systematic impact on flood forecasting.Therefore,it is feasible to modify the initial values of state variables to improve the accuracy of flood forecasting.(2)The hydrological characteristics of Saitang basin are studied.The precipitation,discharge and evaporation data of the basin from 1992 to 2018 are collected and analyzed respectively.The precipitation trend is studied by Mann-Kendall trend test and MWP catastrophe test.The results show that in the past 30 years,the annual precipitation days in flood season have decreased significantly,and the maximum precipitation in Yangxi and Zhangzhuang station has increased significantly,but the other precipitation indices have no obvious change trend.The analysis of flood characteristics shows that the flood in Saitang basin is allover from April to July,and the average last time of a flood is 6.08 days,the average forecast period is 19.5 hours,the average runoff-rainfall ratio is 0.53,the flood shape has one single peak,and the flood process is relatively gentle.Evaporation characteristics analysis shows that the daily and monthly evaporation of the basin has no obvious change trend,and the annual evaporation has a significant downward trend.The paper analyses the influence of ten-day evaporation on flood forecasting.The results show that the calculated ten-day evaporation parameter instead of the measured evaporation is the feasible.When the data length of the evaporation data are more than 20 years,the calculated ten-day evaporation parameters are stable.(3)The principles,steps and characteristics of the IS VC method are explained systematically.The value of initial state variables is corrected by using measured and predicted flow in stationary period in the ISVC method.It has 7 steps to realize the correction.Firstly the value of the initial state variables is got by the continuous hydrological model to implement the flood forecasting.Then choose the stationary period and establish the error objective function by using the measured and calculated discharge in the stationary period.Next,to judge the adjustment direction of the initial value,give the range of the initial value,and some optimal algorithms are used to calibrate the optimal initial value of the state variable.Finally,these calibrated optimal initial state variables are used to carry out a new flood forecasting.The ISVC method does not make any changes to the hydrological model and the original forecasting method.And it has the characteristics of independence,universality and practicability.(4)The application of the ISVC method in Saitang basin and the results analysis.ISVC method can be applied to historical flood simulation and real-time flood forecasting.Saitang basin is chosen as the example,and the flood forecasting scheme of the Saitang basin is compiled with the Xinanjiang model,and the application effect of ISVC method is verified by using historical flood.After the correction of 17 floods,the forecasting effect of 12 floods is better,the improvement rate is 70.6%,and the qualified rate of the revised scheme is increased from 79.3%to 84.1%.The results show that ISVC method can effectively improve the accuracy of flood forecasting.Then the floods are divided into eight categories to analyze the possible changes after correction.Typical floods,such as the flood with better effect after correction and the flood with worse effect after correction,are selected to illustrate the application of ISVC method in detail.The influence of the selection of stationary period in ISVC method on the correction results is analyzed and discussed in this paper.The results show that the selection of stationary period is very important and has an important influence on the correction results.Setting different thresholds of stationary period and correcting them respectively,and then it shows that the thresholds setting has an important effect,and improper setting will lead to inadequate correction and over-correction.The similarities and differences between initial values of state variables and the model parameters are also analyzed in this paper.Through Monte-Carlo statistical experiment,the similarities and differences of initial values and their effects on the corrected results are pointed out.The method of choosing the best value from many times' calculations has effectively solved the adverse effects caused by the phenomenon of different initial values and the same effect.