پیش‌بینی مدل مکانی سطح ایستابی با استفاده از تابع هایپربولیک تانژانت شبکه ی عصبی مطالعه ی موردی: دشت سرخون

نوع مقاله : علمی

نویسندگان

1 جغرافیا-علوم انسانی- دانشگاه زنجان-زنجان-ایران

2 دانشجوی دکتری، آب و هواشناسی، دانشگاه زنجان، زنجان، ایران

چکیده

در سال‌های اخیر سطح تراز آب‌های زیرزمینی در اثر تغییرات اقلیمی و همچنین شیوه و میزان بهره‌برداری از آن‌ها، روند نزولی داشته است. با توجه به افزایش تقاضای آب و افت شدید آب­های زیرزمینی، مدیریت پایدار این منابع از اهمیت شایانی برخوردار است. پیش‌بینی سطح ایستابی با استفاده از مدل‌های ریاضی و آماری می‌تواند کمک قابل‌توجهی به برنامه‌ریزی و تصمیم‌گیری‌های مناسب جهت تأمین آب در درازمدت، داشته باشد. در این مطالعه تلاش شده است تا سطح آب­های زیرزمینی با استفاده از شبکه­ی عصبی گرادیان دیسکنت و تابع انتقال Hyperbolic Tangent پیش‌بینی شود. مدل تابع انتقال Tanh با تعداد 40 نرون در لایه پنهان با ضریب همبستگی 99/0 و مجذور مربعات خطا 01/0 برای پیش‌بینی سطح ایستابی پیاده‌سازی شد. با تعمیم این مدل به ده چاه مشاهده­ای و برون­یابی در محیط سامانه­ی اطلاعات جغرافیایی، مدل مکانی پیوسته سطح ایستابی در دشت سرخون برای سال 1400 تخمین زده شد. نتایج نشان داد که سطح ایستابی در قسمت‌های غربی دشت با مقدار 98/72 متر بیشترین و در بخش شرقی دشت با توجه به تراکم جمعیتی بیشتر مقدار با 72/18 متر کمترین سطح ایستابی را خواهند داشت. با توجه به میزان خطای پایین مدل، می‌توان نتیجه گرفت که با اجرای این مدل در دیگر حوزه‌ها می‌توان پیش‌بینی صحیحی از سطح آب‌های زیرزمینی به دست آورد و در برنامه‌ریزی و مدیریت پایدار آب‌های زیرزمینی از آن استفاده نمود

کلیدواژه‌ها


عنوان مقاله [English]

Prediction of the Water Table Surface Model using the Hyperbolic Tangent Function of the Neural, Network Case Study: Sarkhoon Plain

نویسندگان [English]

  • masoud jalali 1
  • Mohamad Kamangar 2
  • Robab Razmi 2
1 geography-humanities-university of zanjan-zanjan-iran
2 - Ph.D, Student, Department of Humanities, Zanjan University, Zanjan, Iran
چکیده [English]

1- Introduction
In recent years, groundwater level has been descending due to climate change as well as method and use of them, especially in arid and semi - arid regions. According to the United Nations studies, Iran is considered one of the countries facing a shortage of water. In terms of climatic conditions, much of the country is but arid and semiarid regions. Water Table level control using observation wells are the main source of information to investigate the hydrological changes in these areas. Due to the recent drought and water shortages on a wide area of the country, the importance and sensitivity of groundwater management is increasing. Predicting the Water Table level using mathematical and statistical models can contribute significantly to proper planning and decisions to provide long - term water supply. In this study, the level of underground water using the gradient network and the transfer function of Tangent has been tried. Because of recent decades, neural network model studies show the high capability of this model in exploring the relationship between data and the recognition of patterns. Coppola and al (2003) investigated the possibility of predicting the level of 12 observation wells in different climatic situations, using artificial neural
 

 

networks, in an area near the Temba Bay of Florida. Their results showed that, in modeling of the waters of the limestone and karstic areas, neural networks performed appropriate performance. diacplous et al. (2005)conducted an investigation to predict 18 months of groundwater level to predict an underground water level in the Mesrar Valley in Crete, Greece. The results indicate that the lonenberg algorithm is the most appropriate model.
2- Methodology
The recent multi - year drought in the province of hormozgan has resulted in the aggravation of drought conditions and the imposition of many problems on water resources in particular in the underground reservoirs in particular. Sarkhoon Plain plain is one of the areas close to the provincial capital of hormozgan. In this paper, prediction of the spatial model of the Water Table plain of Sarkhoon Plain plain using artificial neural network method and Hyperbolic rule is used to investigate the fault level of this model. in this study, the data of ten observation wells during the 25 - year period of 1990 - 1387 to 1392 - 1392 of the regional water organization of hormozgan province have been used. Artificial neural networks are one of the computational methods that utilize the learning process using called Nero, by adjusting the weights, using the input - output samples that are available. This model is subsequently used to estimate the output value for the new data. The weight of the hidden layer and the output layer are changed so that the error rate is min. This error is represented as follows.
(1) E = 1 / 2 [(y - O) ^ 2]
The following algorithm is illustrated in order to train the neural network.
η > 0 and E > 0
After implementation of neural network algorithms with different neurons in matlab software, the results of predicting the water height of Sarkhoon Plain with Hyperbolic transformation functions were obtained.

 

 to determine the best spatial model of different levels of groundwater depth, the soil water models were used. in order to choose the best extrapolation method in this study, eight methods were used and finally the model that had the lowest fault was considered as optimal model.
3- Results
In this study, neural network model was implemented with different neurons to predict the level of groundwater level. After reviewing the evaluation criteria, the neural network model was selected as the top model with 40 neurons in the latent layer and with its extension to observation wells a spatial prediction model was obtained from groundwater level. The very low error and the high correlation of this model, from the results of the test data, shows its efficiency in predicting the level of groundwater level. Using this algorithm for data of ten wells, water height was predicted for twelve months of 1400 year. The results of this research have proved the superiority of neural networks to numerical models, This spatial model can be used to control the rate of water harvesting in different locations for sustainable water resources management, to determine the structure of input parameters of the neural network, the effects of drought periods and the effects of parameters such as rainfall, temperature and evapotranspiration in predicting groundwater levels.
4- Discussion and conclusion
In this study, neural network model was implemented with different neurons to predict the level of groundwater level. After reviewing the evaluation criteria, the neural network model was selected as the top model with 40 neurons in the latent layer and with its extension to observation wells, a spatial prediction model was obtained from groundwater level. The results of diacplous et al. (2005) showed the superiority of lonenberg neural network over other models that have sufficient layers of latent layers, while it seems that the use of multiple latent layers with multiple neurons in different models leads to error reduction and the choice of superior model selection. Toarimino, Chua

 
 and Sethi (2012) emphasize the short - term forecasts of groundwater fluctuations. They have used the parameters of precipitation, evaporation - evapotranspiration and water level in the neural network model. Despite the higher parameters, the absolute mean of their superior model error has been higher than the average model error of the present research. It is probably due to the low intensity of the hidden layer neurons as well as their short time ranges. The results of the study indicate that the use of a neural network algorithm with the number of static neurons cannot be a measure of the performance evaluation of a model. This spatial model can be used to control the rate of water Picked up in different locations for sustainable water resources management, to determine the structure of input parameters of the neural network, the effects of drought periods and the effects of parameters such as rainfall, temperature and evatranspiration in predicting groundwater levels.
 

کلیدواژه‌ها [English]

  • Water Table
  • Neural Networks
  • transfer function
  • Sarkhoon Plain
Adamowski, J., & Chan, H.F. (2011). A wavelet neural network conjunction model for groundwater level forecasting. Journal of Hydrology, 407(1–4), 28-40.
Anh. Quan Tran, Taniguchi. Kenji. (2018). Coupling dynamical and statistical downscaling for high-resolution rainfall forecasting: case study of the Red River Delta. Vietnam, https://doi.org/10.1186/s40645-018-0185-6.
Chang, F., Chang, L., Huangm C. (2016). Prediction of monthly regional groundwater levels through hybrid soft-computing techniques. Journal of Hydrology. Volume541, Part B, October 2016, 976-965.
Coppola, E., Szidarovszky, F., Poulton, M., & Charles, E. (2003). Artificial Neural Network Approach for Predicting Transient Water Levels in a Multilayered Groundwater System under Variable State, Pumping, and Climate Conditions. Journal of Hydrologic Engineering. 8(6), 348-360.  doi:10.1061/(ASCE)1084-0699(2003)8:6(348).
Daliakopoulos, I.N., Coulibaly, P., & Tsanis, I.K. (2005). Groundwater level forecasting using artificial neural networks. Journal of Hydrology. 309(1–4), 229-240. doi: http://dx.doi.org/10.1016/j.jhydrol.2004.12.001.
Magesh, N.S., Chandrasekar, N., & Soundranayagam, J.P. (2012). Delineation of groundwater potential zones in Theni district, Tamil Nadu, using remote sensing, GIS and MIF techniques. Geoscience Frontiers. 3(2), 189-196. doi: http://dx.doi.org/10.1016/j.gsf.2011.10.007.
Gazman, S., Paz, J., Target, M., (2017). The Use of NARX Neural Networks to Forecast Daily Groundwater Levels. Water Resources Management, 31(5), 1591–1603.
Mohanty, S., Jha, M, Kumar, A. and Sudheer, K,P. (2010). Artificial Neural Network Modeling for Groundwater Level Forecasting in a River Island of Eastern India. Water Resources Management. 24(9), 1845-1865. From: doi: 10,1007/s11269-009-9527-x.
Nadiri, A., Vahedi, F., & Moghadam, A. (2016). Groundwater level prediction using a supervised composite fuzzy logic model. Hydrogeomorphology. 6, 115-134.
Nadiri, A., Yosefade, Sayyed. (2017). Comparison of Artificial Neural Network Models, Fuzzy Logic and Adaptive Neuro Fuzzy Inference for Estimation of Hydraulic Conductivity of Maragheh-Bonab Plain Aquifer. Hydrogeomorphology. 10, 21-40.
Nayak, P., Satyaji Rao, Y.R., and Sudheer, K.P., (2006). Groundwater level forecasting in a shallow aquifer using artificial neural network approach. Water Resources Management. 2(1), 77-99.
Rajay, T., Poraslan, F. (2015). Temporal and spatial prediction of groundwater level in the referee plain. Hydrogeomorphology, 4,1-19.
Rajay, T.,  Zinivand, A., & Jafari, H. (2016). Groundwater level prediction of Sharifabad catchment of Qom using neural wavelet models. Journal of Applied Geographical Research. 16(42), 7-26.
Sethi, R.R., Kumar, A., Sharma, S.P., & Verma, H.C. (2010). Prediction of water table depth in a hard rock basin by using artificial neural network. International Journal of Water Resources and Environmental Engineering, 2(4), 95-102. 
Taormina, R., Chau, K.w., & Sethi, R. (2012). Artificial neural network simulation of hourly groundwater levels in a coastal aquifer system of the Venice lagoon. Engineering Applications of Artificial Intelligence. Vol.25, No.8, 1670-1676. doi: