【摘要】 径流预测是开展水土流失监测和预报的基础,针对不同地区筛选出适宜的径流预测模型,可以为更好地开展流域水土保持工作提供便利。本研究基于SCS模型原理,探讨适合鲁中南低山丘陵区不同土地利用方式预测径流的方法,以药乡小流域2012—2015年观测到的径流、降雨数据为基础,选择该流域内4种土地利用方式(坡耕地、梯田、裸地、草地)的径流小区作为研究对象,以粒子群优化算法(PSO)作为参数优化的方法,对标准SCS模型及其修正模型(MS模型)进行参数率定,并将2016年的实测径流等资料用于模型验证,并基于TOPSIS法建立模型评价体系定量评价不同模型的应用效果。结果表明:1)标准SCS-CN模型在4种土地利用方式中的应用效果均不好,各项评价指标都较大程度偏离; 2)使用1stOpt软件并结合粒子群优化算法对标准SCS及MS模型进行优化,形成的SCS-CNLes模型和MSLes模型,二者参数值均具有一定的有效性,在率定期和验证期内,SCS-CNLes模型及MSLes模型应用结果都较好,各项模型评价指标均高于标准SCS-CN模型; 3)通过建立TOPSIS综合评价体系进行分析,MSLes模型在梯田中的应用效果最佳,模型合格率为100%,NSE值为0. 70,决定系数R~2=0. 77,RMSE=0. 87,可以较好应用于实际径流预测。研究成果可为鲁中南低山丘陵区地表径流预测模型的选择提供技术参考。更多还原

【Abstract】 [Background]Runoff prediction is the basis of soil and water loss monitoring and prediction,and selecting suitable runoff prediction models for different regions can provide convenience for better soil and water conservation work in the river basin. [Methods] Based on the principle of SCS model,this study discussed the method of forecasting runoff for different land use types in the mid-southern hilly region of Shandong province. This study selected the runoff plots of four land use types( sloping farmland,level terrace,bare land and grassland) in the small watershed as the research object. Based on the data of runoff and rainfall observed in 2012-2015 years in the Yaoxiang small watershed,the initial loss rate and runoff curve number of the standard SCS-CN model and its modified form( MS model) in this area were calibrated by using the particle swarm optimization( PSO),then the optimized parameters were taken into the original model for runoff prediction,and the application results of initial model was compared with the optimized model. The measured runoff and other data in 2016 year were used to verify the models. Then the pass percent of model,the Nash-Sutcliffe efficiency coefficient( NSE),RMSE and the coefficient of determination( R~2) were selected for evaluation indicators,and the model evaluation system was established based on the principle of TOPSIS method to quantitatively evaluate the application effect of different models. [Results]The application effect of standard SCS-CN model in four land use types was not applicable,and all evaluation indexes were deviated to a greater degree. The particle swarm optimization in 1 stOpt software was used to optimize the standard SCS-CN model and MS model,forming the SCS-CNLesmodel and MSLesmodel,and the initial loss rate and runoff curve number in the optimized model had certain validity. In the model rate period and the model validation period,the application results of SCS-CNLesmodel and MSLesmodel were all fine,and the evaluation indexes of each model were higher than that of standard SCS-CN model. The TOPSIS comprehensive evaluation system was established for analysis. The result showed that the application effect of the MSLesmodel in the level terrace was the best,the qualified rate of the model was 100%,the NSE value was 0. 70,the coefficient of determination was 0. 77,and the RMSE value was 0. 87. And followed by was the SCS-CNLesmodel in the grassland,the qualified rate of the model also was 100%,the NSE value was 0. 53,the coefficient of determination was 0. 83,and the RMSE value was 1. 03.[Conclusions]The results show that MSLesmodel may be applied to forecast the actual runoff to some extent in the level terrace in the mid-southern hilly region of Shandong province,and relevant results can provide theoretical reference for follow-up research of runoff prediction in this area.更多还原