Precipitation Forecast And Runoff Simulation Of Hongwei Farm In Sanjiang Plain

Based on the precipitation data from 1967-2015 in the Sanjiang Plain Red Guard Farm area and surrounding meteorological stations,this paper analyzes the temporal characteristics of short-term precipitation in the basin,and establishes an ARIMAMarkov combined forecasting model to predict the precipitation in the study area from2010 to 2015.The predicted results are used as the basic input data of the meteorological database in the SWAT hydrological model basic database,determine its applicable parameters,use SWAT-CUP to adjust the parameters,and compare and analyze the measured runoff with the simulated runoff.This shows that the SWAT hydrological model has a good performance in the Sanjiang Plain The adaptability and high accuracy can provide a scientific basis for the change law of hydrology and water resources in the Sanjiang Plain,and provide a scientific reference for local runoff forecasting.The main contents are as follows:(1)Through the Hongwei Station in the basin,the annual and monthly precipitation data from 1967-2015 at the surrounding four weather stations in Tongjiang,Raohe and Fuyuan,using trend analysis,moving average,and MANN-KENDALL non-parametric test methods Analyze and calculate the inter-annual,intra-annual and periodic variation characteristics of precipitation in the Hongwei Farm area.From 1967 to 2015,the annual precipitation at Hongwei Station,Tongjiang Station,and Raohe Station was relatively stable,with a slight upward trend.The precipitation at Fuyuan Station showed an obvious decreasing trend year by year,and Hongwei Station,Tongjiang Station and Raohe Station The precipitation in 5 years and 15 years shows cyclical changes,and Fuyuan Station shows periodic changes on the time scales of 9 and 20 years.Changes during the year:Fuyuan Station showed a downward trend year by year except for summer(June,July,August),Hongwei Station,Tongjiang Station and Raohe Station in spring(March,April,May),summer(June,July,August),autumn(September,October,November)and winter(December,January,February of the following year),the precipitation change trend is stable without significant fluctuations.(2)Precipitation forecast in the study area.Apply the ARIMA-Markov combined forecasting model,that is,after using the ARIMA model to predict the annual precipitation,then apply the Markov model to narrow the forecast range and obtain the forecast value.(3)Establish the SWAT model and establish the meteorological database,land use database and soil database required by the model.According to the DEM data of Hongwei Station,Tongjiang Station,Raohe Station and Fuyuan Station in and around Hongwei Farm in Sanjiang Plain,the hydrological analysis module of Arc GIS software is used to extract the river network water system,and combine the underlying surface data of the study area.The study area is divided into 179 hydrological response units(HRU).Analyze the sensitivity of 13 parameters that affect runoff.(4)Use the SWAT-CUP program to calibrate and verify the model parameters.The model parameter calibration is based on the monthly average runoff data of the Caizuizi Hydrological Station in the study area from 2010 to 2013.The accuracy verification selects the Caizuizi Hydrological Station 2014-Monthly average runoff data in 2015.The results showed that the deterministic coefficient R2 during the calibration period was 0.70,and the efficiency coefficient(NS)was 0.72;the determinant coefficient R2 during the verification period was 0.75,and the efficiency coefficient(NS)was 0.73.(5)The calibrated model parameters remain unchanged.Replace the measured precipitation from the meteorological data Red Guard station with the precipitation predicted by the ARIMA-Markov combined forecasting model.Compare the accuracy of the input measured value and the predicted value to illustrate the ARIMA-Markov combination.The prediction model has high prediction accuracy.Among them,the decisive coefficient R2 is 0.69,and the efficiency coefficient(NS)is 0.71.Although the simulation result of the input predicted value is lower than the simulation result of the input measured value,both of them have reached the model evaluation standard and the accuracy is high.