ارزیابی روش کنترل گروهی داده‌ها (GMDH) و سیستم استنتاج فازی-عصبی (ANFIS) در پیش‌بینی خشکسالی در چند نمونه اقلیمی مختلف

نوع مقاله : مقاله پژوهشی

نویسندگان

1 دانشجوی کارشناسی ارشد مهندسی منابع آب؛ بخش مهندسی آب؛ دانشکده کشاورزی؛ دانشگاه شهید باهنر کرمان

2 استادیار بخش مهندسی آب؛ دانشکده کشاورزی؛ دانشگاه شهید باهنر کرمان؛ کرمان؛ ایران

3 دانشیار بخش مهندسی آب؛ دانشکده کشاورزی؛ دانشگاه شهید باهنر کرمان؛ کرمان

چکیده

خشکسالی پدیده‌ای است که احتمال وقوع آن در همه نقاط کره زمین و با هر شرایط اقلیمی وجود دارد. پیش‌بینی خشکسالی می‌تواند نقش مهمی در مدیریت منابع آبی و بهره‌برداری بهینه از آن‌ها ایفا کند. در این مطالعه، برای پیش‌بینی خشکسالی، کاربرد دو روش هوشمند سیستم استنتاج فازی-عصبی (ANFIS) و کنترل گروهی داده‌ها (GMDH) چند نمونه اقلیمی مختلف ایران مورد ارزیابی قرار گرفته است. به این منظور از  شاخص بارش استاندارد شده (SPI) در سه مقیاس 6،3 و 12 ماهه استفاده شد. آمار و اطلاعات بارندگی طی یک دوره 20 ساله (2015-1996) در 7 ایستگاه سینوپتیک ایران با اقلیم‌های متفاوت بکار گرفته شد و جهت بررسی عملکرد مدل‌ها از سه معیار ریشه میانگین مربعات خطا (RMSE)، ضریب تبیین (R2) و میانگین قدرمطلق خطا (MAE) استفاده شد. نتایج نشان داد که در روش ANFIS مقدار ضریب تبیین در کمترین حالت مربوط به SPI سه ماهه (SPI-3) با 59/0 و بیشترین آن در  SPI دوازده ماهه (SPI-12) با مقدار 97/0 می‌باشد. در روش GMDH، مقادیر ضریب تبیین در هر سه مقیاس SPI و در تمامی اقلیم‌ها بین 90/0 تا 99/0 بدست آمد که نشان‌دهنده دقت قابل قبول این مدل بود. . همچنین نتایج حاکی از عملکرد مناسب SPI در مقیاس دوازده ماهه بودند. . در واقع، بهبود عملکرد مدل‌های ساخته شده با افزایش مقیاس زمانی محاسبه SPI، رابطه مستقیمی دارد. در نهایت نتایج مربوط به مقایسه مقادیر مشاهداتی و پیش‌بینی شده‌ی هر سه مقیاس زمانی با استفاده از روش GMDH در تمامی اقلیم‌ها نشان داد که پیش‌بینی خشکسالی با این روش قابل اطمینان و امکان استفاده از این روش برای پیش‌بینی‌های آتی میسر می‌باشد. بطور کلی نتایج تولید شده توسط هردو روش ANFIS و GMDH دارای دقت قابل قبولی می‌باشند اما پاسخ‌های بدست آمده از روش GMDH بهتر بوده و به عنوان مدل برتر در پیش‌بینی خشکسالی در این پژوهش معرفی می‌گردد

کلیدواژه‌ها

عنوان مقاله [English]

Drought forecasting using Group Method of Data Handling (GMDH) and Adaptive Neural-Fuzzy Inference System (ANFIS) in different climates

نویسندگان [English]

  • Habibeh Helmi 1
  • Bahram Bakhtiari 2
  • Korosh Qaderi 3
1 M. Sc. student in Water Resources Engineering, Water Engineering Department, Faculty of Agriculture, Shahid Bahonar University of Kerman, Kerman, Iran
2 Assistant Professor, Water Engineering Department, Faculty of Agriculture, Shahid Bahonar University of Kerman, Kerman, Iran
3 Associate Professor, Water Engineering Department, Faculty of Agriculture, Shahid Bahonar University of Kerman, Kerman, Iran
چکیده [English]

The results of model evaluation during training and testing demonstrated significant accuracy differences between various models. Results showed that in the ANFIS, minimum of R2 in SPI-3 was 0.59, in hyper-humid climates (Ramsar and Bandar-e-Anzali) and maximum of R2 in SPI-3 was 0.78, in hyper-arid climate (Zahedan) and humid climate (Yasuj). Also minimum of R2 in SPI-6 was 0.75, in semi-arid climate (Hamedan) and maximum of R2 in SPI-6 was 0.87, in hyper-arid and arid climates (Zahedan and Mashhad). In SPI-12, minimum of R2 was 0.88, in hype-arid and semi-arid climates (Zahedan and Hamedan) and minimum of R2 was 0.97 in arid climates (Mashhad). Also, results of ANFIS showed that membership functions type and climates type don't have effect on ANFIS performance and when model is using precipitation in two delay step and SPI in 3 delay step, it has acceptable and high accuracy results. In the GMDH, R2 is between 0.91-0.99 in all three SPI scales (SPI-3, SPI-6 and SPI-12) and in all climates which it indicates the acceptable accuracy of this model. In order to evaluate the results of GMDH models, the best models related to M4 and M9 that input variables are {SPI(t-1), SPI(t-2), SPI(t-3), SPI(t-4), SPI(t-5)} and {SPI(t-1), SPI(t-2), SPI(t-3), SPI(t-4), SPI(t-5), P(t-1), P(t-2)}. RMSE values indicated that it increases when climate type is changing. Hyper-humid and humid climates have RMSE more than other climates. It related to precipitation effect in models performance. M5 and M6 models that use just precipitation in the previous months have low performance in drought forecasting. Also results indicate that SPI is appropriate for 12-month scale. In fact, the performance of the models has direct relationship with the increasing of the SPI time scale. Finally, The results of the comparison of observed and calculated values of three SPI scales (SPI-3, SPI-6 and SPI-12) using the GMDH model in all climates showed that drought forecasting is reliable when this method used and it'll use possibility for future drought forecasting. In general, the results are accurate when using ANFIS and GMDH but the performance of the GMDH model is better than other model. Also, execution speed and GMDH calculations are far more than the ANFIS. Finally, in this study, GMDH propose as the best model for drought forecasting

کلیدواژه‌ها [English]

  • Forecasting
  • Drought
  • GMDH
  • ANFIS

مراجع

  1.  

    1. Azmathulla, M., Najafi, M., Hazi, A. 2013. Group Method of Data Handling to predict scour depth around bridge piers. Neural Computing and Application. 23(7), 2107-2112
    2. Aqil, M., Kita, I., Yano, A., Nishiyama, S. 2007. Analysis and prediction of flow from local source in a river basin using a Neuro–fuzzy modeling tool. Journal of Environmental Management. 85, 215-223
    3. Bacanli, U., Firat, M., Dikbas, F. 2009. Adaptive Neuro-Fuzzy Inference System for drought forecasting. Stochastic Environmental Research and Risk Assessment. 23, 1143-1154
    4. Bordi, I., Sutera, A. 2004. Drought variability and its climatic implications, Global and Planetary Change. 40(2), 115-127
    5. Chua, L.H.C., Wong, T.S.W. 2010. Improving event-based rainfall–runoff modeling usinga combined artificial neural network–kinematic wave approach. Journal of Hydrology. 390(1), 92-107
    6. David, J., Rappoport, P.N. 1974. The Use of Time Series Analysis techniques in forecasting meteorological drought. Bulletin of the American Meteorological Society, MonthlyWeather Review. 102(2), 176-180
    7. Djerbouai, S., Souage, D. 2016. Drought Forecasting Using Neural Networks, Wavelet Neural Networks, and Stochastic Models: Case of the Algerois Basin in North Algeria. Water Resources Management. 30, 2445-2464
    8. Edwards, D.C., McKee, T.B. 1997. Characteristics of 20th century drought in the United States at multiple time scales. Climatology Report. Colorado State University Fort Collins Colorado
    9. Fathabadi, A., Gholami,  H., Salajeghe ,  A., Azaivand, H., Khosravi, H. 2009. Drought forecasting using neural network and stochastic models. American Eurasian Network for Scientific Information 3(2), 137-146
    10. Hoseinzadeh, M., Qaderi, K., Ahmadi, M.M. 2015. Stage-discharge relationship modeling in river using artificial neural networks (ANN) and group method of data handeling (GMDH) methods (Case study: Schuylkill river). Journal of water and soil Conservation. 23(2), 171-184
    11. Hurst, H. 1951. Long Term Storage Capacity of Reservoirs, Transactions of the American Society of Civil Engineers. 116, 770-799
    12. Ivakhnenko, A.G. 1968. The Group Method of Data Handling-a Rival of the Method of Stochastic Approximation, Soviet Automatic Control c/c of Avtomatika. 1, 43-55
    13. Jang, J.S.R. 1993. ANFIS adaptive network based fuzzy inference system, IEEE Transactions on Systems, Man and Cybernetics. 3, 665–683
    14. Kamali, A., Moradi,  M., Moradi, N. 2015. Application of several artificial intelligence models and ARIMAX model for forecasting drought using the Standardized Precipitation Index. Journal of Environment Science Technology. 2 (12), 1201-1210
    15. Karamooz, M., Rasooli, K., Nazif, S. 2006. Completion of drought combined index and its prediction using neural networks. Proceedings of second conference of water resource management
    16. Kim, T., Valdes, J.B. 2003. Nonlinear model for drought forecasting based on a conjunction of wavelet transforms and neural networks. Journal of Hydrological Engineering. 8(6), 319-328
    17. McKee, T.B., Doesken, N.J. 1993. The relationship of drought frequency and duration to time scales. Eight Conference on Applied Climatology. Anaheim, CA, American Meteorological Society
    18. Mishra, A.K., Desai, V.R., 2005. Drought forecasting using stochastic models. Stochastic Environmental Research and Risk Assessment. 19, 326-339
    19. Morgan, G.C. 1998. Fuzzy logic Routlendge Encyclopedia of Philosophy. 3first edition Craig,E. Routledge, London
    20. Nguyen, V.,  Li1, O., Nguyen, L. 2017. Drought forecasting using ANFIS- a case study in drought prone area of Vietnam. The International Society of Paddy and Water Environment Engineering. 15(3), 605-616
    21. Özger, M., Mishra, A.K. 2012. Long lead time drought forecasting using wavelet and fuzzy logic combination model: a case study in Texas. Journal of Hydrometeoroly. 13(1), 284-297
    22. Qaderi, K., Arab, D.R., Teshnelab, M., Ghazagh, A. 2009. Intelligent Operation Modeling of Reservoirs Using Group Method of Data Handling (GMDH). Iran Water Resources Research. 6(3), 55-67
    23. Qaderi, K., Eivani, Z., Ahmadi, M.M. 2016. Estimation of Suspended Sediment Load Concentration in River Systemusing Group Method of Data Handling (GMDH). Journal of Watershed Management Research. 7(13), 218-229
    24. Rahimi, J., Ebrahimpour, M., Khalili, A. 2012. Spatial changes of Extended De Martonne climatic zones affected by climate change in Iran. Theoretical and Applied Climatology. 112, 409-418
    25. Rawls, W.J., Pachepsky, Y.A., Ritchie, J.C., Sobecki, T.M., Bloodworth, h. 2003. Effect of soil organic carbon on soil water retention. Journal of Water and Soil Resources. 116(1), 61-76
    26. Ravinesh, C.D., Ozgur, K., Vijay, P.S. 2016. Drought forecasting in eastern Australia using multivariate adaptive regression spline least square support vector machine and M5Tree model. Atmospheric Research. 184, 149-175
    27. Samadianfard, S., Asadi,  E. 2017.  Prediction of SPI drought index using support vector and multiple linear regressions. Journal of Water and Soil Resources. 6(4), 1-16
    28. Sumsudin, R., Ismail, S., Shabri, A. 2010. River flow forecasting: A Hybrid Model of Self Organizing Maps and Least Square Suport Vector Machine. Stochastic Environmental Research and Risk Assessment. 7,  8179-8212
    29. Santos, C., Morais, B., Silva, B. 2009. Drought forecast using an artificial neural network for three hydrological zones in San Francisco River basin Brazil. Proceedings of Symposium HS.2 at the Joint Convention of The International Association of Hydrological Sciences (IAHS) and The International Association of Hydrogeologists. 333, 302-312
    30. Sharma, B.R., Smakhtin, V.U. 2004. Potential of water harvesting as a strategic tool for drought mitigation. International Water Management Institue. 29(1), 112-118

    Soleimanikia, F. 2007. Gasoline prices modeling and forecasting in the Exchange Singapore using a genetic algorithm based neural network (GMDH). Master's thesis, Faculty of Economics of Tehran Unniversity

پژوهش های اقلیم شناسی
دوره 1397، شماره 35
اسفند 1397
صفحه 1-18

فایل ها

هم رسانی

ارجاع به این مقاله

آمار