SWAT模型参数自动率定的改进与应用

Parameters Auto Calibration with Modified SWAT Model

SWAT2005版本中新增了LH-OAT灵敏度分析模块和SCE-UA自动校准模块。在进行参数敏感性分析、参数率定时,计算耗费的时间随着子流域个数的增加而增加。以北京市为例,流域面积为16800km2,共划分了138个子流域,对连续5a的日过程进行模拟大概需要1min。用SWAT模型进行参数敏感性分析、参数自动率定时,需要用不同的参数来多次进行模拟,例如对北京SWAT模型的10个参数进行率定,模型要重复运行5000次,大概需要4d的时间。如果这个过程中出现错误,那么还需要重新进行参数率定。因此,这种参数率定方法对大流域来说是不合适的,不能令人满意的。针对这一问题,本文提出一种快速率定SWAT模型参数的方法,即:只对部分子流域进行模拟,如只模拟目标水文站控制的上游的几个子流域。这样一来,参数自动率定需要的时间就会大大减少。在把该方法应用到北京市的SWAT模型参数率定时,对选定的前辛庄站的5a的日径流过程进行率定时总共用时不超过30min,而且率定效果较好,这充分说明了该方法的高效性。该方法推广后可实现对分布式水文模型中多个水文站的径流过程进行快速率定。

LH-OAT analysis module and SCE-UA auto calibration module have been added to SWAT 2005.During the process of parameter sensitivity analysis and calibration,the computing time increases with the number of subbasins.For a watershed with an area of 16800 km2,subbasins of 138 for Beijing,the time is about 1 minute for a 5 yeas simulation of SWAT model.Because the parameters auto calibration program of SWAT model tries many times of different parameters,the model was rerun about 5000 times for 10 parameters used in Beijing,which will cost about 4 days.If any mistake occurs in this process,it will need to redo the calibration work.Therefore this calibration method is ill-suited and unacceptable for large basins.An idea is put forwarded and realized in the paper,that is we can only simulation some subbasins,such as the only subbasins flow to target hydro station.Then with only simulation of these several subbasins,the time used for parameters auto calibration will be greatly reduced.With this method applied in parameter calibration in Beijing Swat model,it takes no more than 30 minutes to calibrate the five years' day runoff of Qianxinzhuang hydro station and the results are preferably.The calibration of runoff process for a number of hydro stations in distributed hydrological model can be rapidly done through the generalization of this method.