بررسي كارايي توابع هدف مبتني بر منحني تداوم جريان و معيارهاي ارزيابي آماري در واسنجي مدل‌هاي هيدرولوژيك

Evaluation the efficiency of flow duration curve based and statistical criteria objective functions in calibrating hydrological model

در ﻣﺪل ﻫﺎي ﻫﯿﺪروﻟﻮژﯾﮏ ﺑﻪ ﻣﻨﻈﻮر ﻣﺪل ﺳﺎزي ﺑﻬﺘﺮ ﻓﺮاﯾﻨﺪ ﺑﺎرش رواﻧﺎب ﻧﯿﺎز اﺳﺖ ﺗﺎ ﻣﺪل ﺑﺎ اﺳﺘﻔﺎده از داده ﻫﺎي ﻣﺸﺎﻫﺪاﺗﯽ واﺳﻨﺠﯽ ﺷﻮد. در ﻓﺮاﯾﻨﺪ واﺳﻨﺠﯽ ﻣﺪل ﻫﺎي ﻫﯿﺪروﻟﻮژﯾﮏ، ﻋﻼوه ﺑﺮ ﮐﯿﻔﯿﺖ داده ﻫﺎي ﻣﺸﺎﻫﺪات و اﻟﮕﻮرﯾﺘﻢ ﺑﻬﯿﻨﻪ ﺳﺎزي ﺗﺎﺑﻊ ﻫﺪف ﻧﯿﺰ ﺑﺮ ﮐﺎراﯾﯽ ﻣﺪل ﻣﻮﺛﺮ اﺳﺖ. در اﻏﻠﺐ ﭘﮋوﻫﺶ ﻫﺎ، ﻣﻌﯿﺎرﻫﺎي آﻣﺎري ﻣﺜﻞ NSE و RMSE ﺑﻪﻋﻨﻮان ﺗﺎﺑﻊ ﻫﺪف در ﻓﺮاﯾﻨﺪ واﺳﻨﺠﯽ ﻣﺪل ﻫﺎي ﻫﯿﺪروﻟﻮژﯾﮏ اﺳﺘﻔﺎده ﻣﯽ ﺷﻮﻧﺪ. ﺑﺎ ﺗﻮﺟﻪ ﺑﻪ ﺳﺎﺧﺘﺎر ﻣﺪل و رواﺑﻂ ﻣﻮرد اﺳﺘﻔﺎده در ﻫﺮ ﯾﮏ از ﻣﻌﯿﺎرﻫﺎي ارزﯾﺎﺑﯽ، ﻫﺮ ﯾﮏ از آنﻫﺎ در ﺷﺒﯿﻪ ﺳﺎزي ﺑﺨﺸﯽ از ﻫﯿﺪروﮔﺮاف ﮐﺎراﯾﯽ ﻣﻨﺎﺳﺒﯽ دارﻧﺪ. ﯾﮑﯽ از ﭘﺎراﻣﺘﺮﻫﺎي ﻣﻬﻢ ﻫﺮ ﺣﻮﺿﻪ ﮐﻪ ﺑﻪ ﻧﻮﻋﯽ ﻧﺸﺎن دﻫﻨﺪه ﻋﮑﺲ اﻟﻌﻤﻞ ﺣﻮﺿﻪ ﺑﺮاي ﻣﻘﺎدﯾﺮ دﺑﯽ ﻣﺨﺘﻠﻒ ﻣﯽ ﺑﺎﺷﺪ، ﻣﻨﺤﻨﯽ ﺗﺪاوم ﺟﺮﯾﺎن اﺳﺖ. در اﯾﻦ ﭘﮋوﻫﺶ، اﻗﺪام ﺑﻪ ﺑﺮرﺳﯽ و ﻣﻘﺎﯾﺴﻪ ﮐﺎراﯾﯽ ﺗﻮاﺑﻊ ﻫﺪف ﻣﺒﺘﻨﯽ ﺑﺮ ﻣﻨﺤﻨﯽ ﺗﺪاوم ﺟﺮﯾﺎن و ﺗﻮاﺑﻊ ﻫﺪف آﻣﺎري در ﺑﻬﯿﻨﻪ ﺳﺎزي ﭘﺎراﻣﺘﺮﻫﺎي ﻣﺪل ﻫﯿﺪروﻟﻮژﯾﮑﯽ HBV در ﺣﻮﺿﻪ زﯾﺎرت اﺳﺘﺎن ﮔﻠﺴﺘﺎن ﺷﺪ. ﭘﺲ از وارد ﮐﺮدن داده ﻫﺎي ورودي ﺑﻪ ﻣﺪل، ﺑﺎ اﺳﺘﻔﺎده از اﻟﮕﻮرﯾﺘﻢ DDS ﻣﺪل ﺑﺮاي ﻫﺮ ﺗﺎﺑﻊ ﻫﺪف ﺑﻪ ﺗﻌﺪاد 100 ﻣﺮﺗﺒﻪ ﺑﻬﯿﻨﻪ ﺷﺪ. ﭘﺲ از ﺑﻬﯿﻨﻪ ﺳﺎزي ﭘﺎراﻣﺘﺮﻫﺎي ﻣﺪل، اﯾﻦ ﭘﺎراﻣﺘﺮﻫﺎ ﺑﻪ ﻣﺪل ﻣﻌﺮﻓﯽ و ﻣﻘﺎدﯾﺮ دﺑﯽ ﺑﺮاي دوره ﻫﺎي واﺳﻨﺠﯽ و اﻋﺘﺒﺎرﺳﻨﺠﯽ ﺑﺮآورد ﺷﺪ. ﻧﺘﺎﯾﺞ ﻧﺸﺎن داد، ﻣﻌﯿﺎرﻫﺎﯾﯽ ﻣﺎﻧﻨﺪ NSE و KGE در ﺑﺮآورد داده ﻫﺎي ﭘﯿﮏ، ﻣﻌﯿﺎرﻫﺎﯾﯽ از ﻗﺒﯿﻞ RMSE و MAE در ﺑﺮآورد داده ﻫﺎي ﻣﺘﻮﺳﻂ و ﻣﻌﯿﺎرﻫﺎي ﻣﺒﺘﻨﯽ ﺑﺮ ﻣﻨﺤﻨﯽ ﺗﺪاوم ﺟﺮﯾﺎن در ﺑﺮآورد دﺑﯽ ﻫﺎي ﮐﻤﯿﻨﻪ ﻋﻤﻠﮑﺮد ﺑﻬﺘﺮي داﺷﺘﻪ اﻧﺪ. در ﺷﺒﯿﻪ ﺳﺎزي ﺑﺨﺶ ﻫﺎي ﻣﺨﺘﻠﻒ ﻫﯿﺪروﮔﺮاف دﺑﯽ روازﻧﻪ روش ﻣﺒﺘﻨﯽ ﺑﺮ ﻣﻨﺤﻨﯽ ﺗﺪاوم ﺟﺮﯾﺎن ﺑﻬﺘﺮﯾﻦ ﻋﻤﻠﮑﺮد را داﺷﺖ، ﻣﻌﯿﺎرﻫﺎي RMSE و MAE در رﺗﺒﻪ ﺑﻌﺪي ﻗﺮار داﺷﺘﻪ و ﻣﻌﯿﺎرﻫﺎي NSE و KGE ﻋﻤﻠﮑﺮد ﻣﻨﺎﺳﺒﯽ ﻧﺪاﺷﺘﻨﺪ.

In hydrological models, in order to better model the runoff process, it is necessary to calibrate the model using observational data. In the process of calibration of hydrological models, in addition to the quality of observation data and the optimization algorithm, the objective function also affects the efficiency of the model. In most studies, statistical criteria such as NSE and RMSE are used as objective functions in the calibration process of hydrological models. Given the structure of the model and the relationships used in each of the evaluation criteria, each of them has good performance in simulating a part of the hydrograph. One of the important parameters of each basin, which is a kind of basin reaction indicator for different discharge values, is the Flow Continuity Curve (FDC). In this study, the efficiency of objective functions based on flow continuity curve and statistical objective functions in optimizing the parameters of the HBV hydrological model in Ziyarat Watershed of Golestan Province was investigated and compared. After introducing input data to model using DDS algorithm, model was calibrated 100 times for each objective function. When model was calibrated, using optimized parameter sets model output for calibration and validation period was obtained. Results showed that criteria such as NSE and KGE have better performance in predicting high flows, criteria such as RMSE and AME predicted moderate flow discharge better and criteria based on FDC had better performance in predicting low flows. In prediction different parts of hydrograph FDC objective function has the best performance, RMSE and MAE were in sound order and NSE and KGE did not have suitable performance.