واسنجی مدل هیدرولوژیکی Wetspa با استفاده از الگوریتم‌های بهینه‌سازی چندهدفه عنکبوت اجتماعی و عنکبوت بیوه سیاه

مدل ‌های بارش رواناب مفهومی، از جمله ابزارهای ساده و در عین حال کارآمد در مدل سازی هیدرولوژیکی است. این مدل‌ ها با درنظرگرفتن اطلاعات ورودی از قبیل بارش، تبخیر تعرق، دمای اندازه ‌گیری شده و اطلاعات توپوگرافی حوضه، سیستم جریان را با استفاده از روابط پیچیده ریاضی شبیه سازی می کند. در این مقاله، از مدل هیدرولوژیکی توزیعی wetspa برای شبیه سازی رواناب حوضه زشک استفاده شد. این پژوهش، قابلیت الگوریتم های بهینه سازی عنکبوت اجتماعی و عنکبوت اجتماعی بیوه سیاه را در واسنجی مدل هیدرولوژیکی wetspa به منظور شبیه سازی بارش رواناب حوضه زشک بیان می کند. از الگوریتم های بهینه سازی بالا به صورت چند هدفه برای واسنجی یازده پارامتر سراسری مدل wetspa استفاده شد. در این تحقیق از معیار نش ساتکلیف و نش ساتکلیف لگاریتمی نیز به عنوان تابع هدف استفاده شد تا به وسیله آنها، عملکرد مدل در پیش بینی دبی های حداکثری و حداقلی بهبود یابد. نتایج نشان داد که هر دو الگوریتم عنکبوت اجتماعی (sso) و عنکبوت بیوه سیاه (bwo) به ترتیب با ضریب رگرسیون 0.71 و 0.76، عملکردهای مناسبی را در واسنجی مدل از خود نشان دادند. مقدار شاخص rmse در دوره واسنجی نیز به طور متوسط برابر با 123.6 و 160.1 بود. همچنین، تجزیه و تحلیل حساسیت پارامترهای موثر نشان داد که 10k (ضریب رواناب سطحی) با 36% تاثیر بر مقدار دبی جریان، حساس ترین پارامتر سراسری مدل wetspa بود.

Calibration of WetSpa Model using Black Widow Spider and Social Spider Multi-objective Optimization Algorithms

IntroductionBy developing GIS and remote sensing technology, the widespread access possibility and local distribution of hydrological management parameters and variables have become practical. Runoff rainfall modeling is always an important and continuous need for practical issues in the fields of water resources evaluation, flood forecasting, engineering canals designing, and many other goals (Bone, 2001). Calculation of runoffrainfall results has been made practical and operationalized by using GIS techniques and a distributed hydrological model. The WetSpa runoffrainfall model is a hydrologicaldistribution model that was developed in Brussels in 1997 (Wange et al, 1997) and also, had been used in various research and executive projects by developing in various models. This hydrological model has the capability of performing simulations at the pixel level and because of this, it provides the possibility of using aerial and satellite images with accurate information measured at the basin level, which has a local distribution (Hooshyarypor at el, 1397). Due to the high uncertainties of hydrological parameters, the calibration model is one of the most important part of modeling whose optimization techniques are mainly are used for such purposes. There are various methods for optimization, sensitivity analysis, and also evaluation of uncertainty of models. Accordingly, this paper rsquo;s purpose is to calibrate the WetSpa hydrological model with a multiobjective optimization approach that uses the Social Spider Algorithm (SSA) and the Black Widow Spider (BWO) techniques. By this point of view, to achieve a reliable prediction, in addition to the usual discharges, the model must be able to predict high and low discharges accurately (including maximum and minimum), so the objective functions are selected in a way that the best match between observational and computational values can be achieved during the Vasanji process.Materials and MethodsWetSpa ModelWetSpa is a continuous local and temporal model in which all of the simulations are conducted continuously. The WetSpa model displays the water and energy balance for each calculation cell, considers rainfall processes, vegetation, snowmelt, wetting, infiltration, evapotranspiration, leakage, surface runoff, wall flow, and groundwater flow. The hydrological system simulated by this model consists of four layers: vegetation, soil surface, root zone, and saturated groundwater table.The Optimization algorithmsThe Social Spider Optimization (SSO) algorithm is a new optimization approach which is proposed in 2013 by Kause et al. Another optimization approach used in this paper is the Black Widow Spider (BWO) optimization algorithm. During the day, the black widow spider is out of sight and is mostly nocturnal, and rotates its network during the night. Generally, the widow spends most of her adult life on the same site (Andrite and Banta, 2002).Objective functions and model evaluationIn this paper, to evaluate the model, five statistical indices have been used of correlation coefficient (r), Root Mean Square Error (RMSE), mean absolute error (MAE), NashSutcliffe index, and NashSutcliffe logarithmic index. To increase the accuracy of the model in predicting the minimum and maximum flows, two alogarithm NashSutcliffe and NashSutcliffe logarithmic criteria were used.Zashk basin and model dataZashk Basin is located in Khorasan Razavi province and in the west of Mashhad with an area of 65.56 square kilometers.Result and DiscussionIn this part of the paper, the calibration results of the WetSpa model using social spider and black widow spider algorithms are presented. The problem decision variables are the 11 global parameters illustrated in Table 1. Objective functions must be selected in a way that at the end of the calibration process the best match between the observed and computational values can be obtained. Some of these functions give more weight in high flows, while others consider more weight in low flows and have more emphasis on them. In this regard, NashSutcliffe (NS) criterion and its logarithmic form (NSLog) have been used. In this paper, the results obtained based on these two functions are illustrated. After determining all the necessary local networks in basin modeling, precipitation, evaporation, temperature, and discharge information from 2009 to 2011 were used to calibrate the model and 2012 to 2014 data were used to validate the results. Each of the optimization algorithms was conducted with an initial population size of 100 people and over 100 generations (number of times the optimization model was executed). As it was observed, the answers obtained by BWO are better than the answers of SSO due to its higher NS values. According to the obtained results, NS and NSLog values are generally from 2.3 to 0.75 and from 0.165 to 0.01, respectively. This problem illustrates that the calibrated model has been more successful in simulating low discharges. In fact, since the model has been executed continuously, the simulation results in 2008 are considered as the WarmUp period.ConclusionThis paper calibrated the WetSpa distributed rainfallrunoff model. To calibrate the model, two evolutionary optimization algorithms of social spider (SSO) and black widow spider (BWO) were used in the Zashk basin of Mashhad. The results obtained from the usage of the model in this basin illustrated the satisfactory capability of these algorithms in calibrating the WetSpa model. Comparing the results obtained from the SSO and BWO algorithms illustrated that in the multiobjective calibration problem, the BWO algorithm was slightly more successful than the SSO model. Also, the results indicated that the simulation quality of low discharges was higher than the high ones. The reason for this problem, firstly, can be due to the lower abundance of high discharges in the evaluated data set and secondly, the relatively slow response of the model to changes in hydrological conditions in flood conditions in the catchment basin, because the temporary groundwater storage coefficient (K5) allocated a large amount for itself that can directly affect the reduction of surface runoff. On the other hand, the weakness of the results in predicting some low discharges can be related to the optimal value of groundwater recession coefficient (K2), which has taken a small amount in the calibration process. The results of this study illustrated that the effect of the surface runoff coefficient on the model results is much greater than other parameters (about 36%).