基于PEST的HSPF模型径流模拟优化

摘要

[Background] In recent years,water cycle and water resource change have become more complicated due to the influence of climate change and human activities.Hydrological model is an approximate simulation of complex hydrological phenomena,which can realize the quantitative study of runoff.And parameter calibration is one of the most important steps in hydrological modeling.[Methods] In order to improve the calibration efficiency and simulation effect of the hydrological model,Hydrological Simulation Program in Fortran (HSPF) used extensively for assessing water quantity issues was selected in this paper.The PEST automatic calibration model with single objective function (squared error of daily flow) and multi-objective functions (squared error of daily flow,squared error of monthly flow,squared error of exceedance flow time) were applied in HSPF model,and the squared error of daily flow was improved by separating wet and dry seasons.These calibration methods were compared with Shanmei Reservoir watershed,southeast China.[Results] 1) Multi-objective calibration performed better with respect to goodness-of-fit measures (NSE,R2,RMSE and PBIAS) than single objective calibration.In multi-objective calibration,the values of the NSE were above 0.79,the values of the R2 were above 0.80,the values of the PBIAS were between-10％ and 0％,and the values of the RMSE were ＜30 m3/s during the calibration and validation.2) In terms of the results of two multi-objective calibration methods,when the squared error of daily flows was improved by separating wet and dry seasons,the simulation results of daily and monthly runoff were significantly improved during the dry season.Taking daily runoff simulation for example,the values of the NSE increased from-0.04 and 0.43 to 0.25 and 0.54,the values of the R2 increased from 0.37 and 0.59 to 0.47 and 0.64,the values of the PBIAS increased from-33.57％ and-22.72％ to-24.52％ and-16.79％,and the values of the RMSE decreased from 9.54 m3/s and 8.89 m3/s to 8.06 m3/s and 7.96 m3/s during the calibration and validation,respectively.[Conclusions] The objective function has a great influence in the automatic calibration of parameters.Multi-objective calibration performed better with respect to the trend of runoff,total runoff,and flow duration curves than single objective calibration.When the squared error of daily flow is improved by separating wet and dry seasons,the problem of dry flow was neglected in traditional multi-objective calibration was solved and the simulation results of daily and monthly runoff were significantly improved during the dry season.%参数率定是水文模型构建的关键步骤之一,为了提高模型的率定效率和模拟效果,针对HSPF模型,以晋江山美水库流域为例,将PEST应用于日径流模型的率定,分别采用单目标函数和多目标函数率定参数,并对日流量偏差这一目标函数进行改进,划分为丰水期和枯水期的日流量偏差,对比分析了单目标与多目标率定结果的优劣及多目标函数改进前与改进后率定结果的差异.结果表明:1)相比单目标率定,多目标率定的NSE、R2、RMSE和PBIAS等结果均得到不同程度的提高.2)2种多目标率定的结果,在对日流量偏差改进后,枯水期日径流的模拟结果得到显著提高,率定期和验证期的NSE分别由-0.04和0.43提升至0.25和0.54,R2分别由0.37和0.59提升至0.47和0.64,PBIAS分别由-33.57％和-22.72％提升至-24.52％和-16.79％,RMSE分别由9.54和8.89m3/s降低至8.06和7.96 m3/s.因此,相比单目标率定,多目标率定能更好的把握径流变化趋势、径流总量和流量保证率的特征;并且将日流量偏差划分丰、枯水期之后,研究区的枯水期径流模拟效果得到显著提高.