ارزیابی کارایی مدل هیبریدی GRU-LSTM در پیش بینی طوفان های گرد و غبار (مطالعه موردی: استان خوزستان)

نوع مقاله : مقاله پژوهشی

نویسندگان

1 دانشجوی دکتری گروه مهندسی آبیاری و آبادانی، پردیس کشاورزی و منابع طبیعی، دانشگاه تهران، کرج، ایران.

2 دانشیار گروه مهندسی آبیاری و آبادانی، پردیس کشاورزی و منابع طبیعی، دانشگاه تهران، کرج، ایران.

3 دانشیار دانشکده مهندسی عمران، پردیس دانشکده های فنی، دانشگاه تهران، تهران، ایران.

4 دانشجوی دکتری علوم و مهندسی آب، دانشکده کشاورزی و منابع طبیعی، دانشگاه ازاد اسلامی واحد اهواز، ایران.

چکیده

درک صحیح از وقوع طوفان‌های گرد و غبار در هر منطقه و آگاهی از تغییرات زمانی- مکانی این پدیده به مدیریت و کاهش خسارت‌های ناشی از گرد و غبار کمک شایانی می‌کند. در سال‌های اخیر، توسعه فرامدل‌ها و ترکیب آن‌ها با الگوریتم‌های بهینه‌سازی به منظور مدل‌سازی و پیش‌بینی متغیرهای آب و هوایی، مورد توجه زیادی قرار گرفته‌ است. از این رو در مطالعه حاضر، نوعی رویکرد ترکیبی به منظور پیش‌بینی فراوانی روزهای همراه با طوفان گرد و غبار (FDSD) در مقیاس فصلی پیشنهاد شده که در آن از ترکیب شبکه‌های عصبی LSTM و GRU استفاده می‌شود. در این پژوهش، عملکرد مدل هیبریدی پیشنهادی با شبکه عصبی مبتنی بر توابع پایه شعاعی (RBF) و ماشین بردار پشتیبان (SVM) مورد مقایسه قرار گرفته است. بدین منظور، از داده‌های ساعتی گرد و غبار و کدهای سازمان جهانی هواشناسی در مقیاس فصلی با طول دوره آماری ۳۰ ساله (2019-1990) در هفت ایستگاه سینوپتیک استان خوزستان استفاده شد. نتایج معیارهای ارزیابی در مرحله آموزش و آزمایش مدل‌ها نشان داد که مدل هیبریدی GRU-LSTM عملکرد بهتری نسبت به سایر مدل‌های مورد استفاده به منظور پیش‌بینی فراوانی روزهای همراه با طوفان گرد و غبار ارائه می نماید؛ به طوری که مدل هیبریدی پیشنهادی با ضریب همبستگی (0/988-0/905=R)، ریشه میانگین مربعات خطا (RMSE=0/313-0/402 day)، میانگین قدر مطلق خطا (MAE= 0/144-0/226 day) و ضریب نش‌-ساتکلیف (0/903-0/819=NS)، عملکرد بهتری نسبت به سایر مدل‌های مورد استفاده در پیش‌بینی شاخص FDSD داشته است. در مجموع با مقایسه مدل‌های مورد استفاده، روش هیبریدی GRU-LSTM بهترین عملکرد و بعد از آن مدل SVM بهترین نتیجه را ارائه نمود. لذا مدل هیبریدی پیشنهادی می­تواند به عنوان ابزاری مناسب جهت پیش­‌بینی شاخص FDSD و به تبع آن اتخاذ تصمیمات مدیریتی به منظور کاهش خسارات طوفان­‌های گرد و غبار، در منطقه مطالعاتی مورد استفاده قرار گیرد.

کلیدواژه‌ها

موضوعات


عنوان مقاله [English]

Evaluating the Performance of GRU-LSTM Hybrid Model in Predicting the Dust Storms Events (Case Study: Khuzestan Province in Southwest of Iran)

نویسندگان [English]

  • Mohammad Ansari Ghojghar 1
  • Shahab Araghinejad 2
  • Javad Bazrafshan 2
  • Banafsheh Zahraie 3
  • Ehsan Parsi 4
1 Ph.D. Candidate, Department of Irrigation and Reclamation Engineering, College of Agriculture and Natural Resources, University of Tehran, Karaj, Iran.
2 Associate Professor, Department of Irrigation and Reclamation Engineering, College of Agriculture and Natural Resources, University of Tehran, Karaj, Iran.
3 Associate Professor, School of Civil Engineering, College of Engineering, University of Tehran, Tehran, Iran.
4 Ph.D. Candidate, Department of Water Science and Engineering, Faculty of Agriculture and Natural Resources, Islamic Azad University Ahwaz, Ahwaz, Iran.
چکیده [English]

Understanding the frequency of dust storms in each area and being mindful of temporal-spatial variation of this event can help to monitor and reduce the damages induced by dust events. Due to the increasing development of metamodels and their combination with optimization algorithms used to model and predict hydrological variables, machine learning models due to high accuracy in forecasting, in the form of a black box, have received a lot of attention. Therefore, in the present study, a hybrid approach is proposed to predict the Frequency of Dust Storm Days (FDSD) on a seasonal scale, which uses a combination of Lang Short Term Memory (LSTM) and Gated Recurrent Unit (GRU) neural networks. In this study, the performance of the proposed hybrid model was compared with a neural network based on Radial Basis Functions (RBF) and Support Vector Machine (SVM). For this purpose, hourly dust data and codes of the World Meteorological Organization were used on a seasonal scale with a statistical period of 30 years (1990-2019) for seven synoptic stations in Khuzestan province. The results of the evaluation criteria in the training and testing Stages showed that the GRU-LSTM hybrid model offered better performance than other models used to predict the frequency of days with dust storms; The proposed hybrid model with correlation coefficient (R) of 0.905-0.988, Root Mean Square Error (RMSE) of 0.313-0.402 day, Mean Absolute Error (MAE) of 0.144-0.236 day, and Nash-Sutcliffe (NS) of 0.819-0.903 had better performance compared to the other models used in predicting the FDSD index. In general, comparing the models used in this study, the GRU-LSTM hybrid method and the SVM model, respectively, provided the best prediction skills. As a result, application of the proposed hybrid model can be used as a suitable tool to predict the FDSD index and adopting management decisions to reduce the dust storms damages in the study area.

کلیدواژه‌ها [English]

  • Prediction
  • Dust storm
  • SVM
  • GRU-LSTM method
  • Khuzestan province
Abdolshahnejad M, Khosravi H, Nazari Samani A A, Zehtabian G R, and Alambaigi M (2020) Determining the conceptual framework of dust risk based on evaluating resilience (Case study: Southwest of Iran). Strategic Research Journal of Agricultural Sciences and Natural Resources 5(1):33-44 (In Persian)
Ansari ghojghar M, Pourgholam-Amiji M, Bazrafshan J, Liaghat A, Araghinejad S (2020) Performance comparison of statistical, fuzzy and perceptron neural network models in forecasting dust storms in critical regions in Iran. Iranian Journal of Soil and Water Research 51(8):2051-2063
Araghinejad S (2013) Data-driven modeling: Using MATLAB® in water resources and environmental engineering (Vol. 67). Springer Science & Business Media
Araghinejad Sh, Ansari Ghojghar M, Pourgholam-Amiji Liaghat A, and Bazrafshan J (2018) The effect of climate fluctuation on frequency of dust storms in Iran. Desert Ecosystem Engineering Journal 7(21):13-32 (In Persian)
Araghinejad Sh, Ansari Ghojghar M, Pourgholam Amiji M, Babaeian I (2020) Modeling the relationship between dust storms and extreme and average temperature variables in the western half of Iran. Journal of Climate Research 8(31-32)
Baghbanan P, Ghavidel Y, Farajzadeh M (2020) Temporal long-term variations in the occurrence of dust storm daysin Iran. Meteorology and Atmospheric Physics (2020) 132:885–898
Basak D, Pal S, Patranabis D C (2007) Support vector regression. Neural Information Processing 11:203- 225
Cao R, Jiang W, Yuan L, Wang W, Lv Z, and Chen Z (2014) Inter-annual variations in vegetation and their response to climatic factors in the upper catchments of the Yellow River from 2000 to 2010. Journal of Geographical Sciences 24(6):963-979
Chen S, Cowan C F N, and Grant P M (1991) Orthogonal least squares learning algorithm for radial basis function networks. IEEE Transactions on Neural Networks 2(2):302-309
Dahiya S, Singh B, Gaur S, Garg V K, and Kushwaha H S (2007) Analysis of groundwater quality using fuzzy synthetic evaluation. Journal of Hazardous Materials 147(3):938-946
Goudie A S and Middleton N J (2006) Desert dust in the global system. Springer Science & Business Media
Hahnenberger M and Nikoul K (2014) Geomorphic and land cover identification of dust sources in the eastern Great Basin of Utah, U.S.A. Journal of Geomorphology 204(2):657-672
He Z, Wen X, Liu H, Du J A (2014) Comparative study of artificial neural network, adaptive neuro fuzzy inference system and support vector machine for forecasting river flow in the semiarid mountain region. Journal of Hydrology 509:379–386
Hamel L (2009) Knowledge discovery with support vector machines. Hoboken, N.J. John Wiley
Hussain D and Khan A A (2020) Machine learning techniques for monthly river flow forecasting of Hunza River, Pakistan. Earth Science Informatics, DOI: 10.1007/s12145-020-00450-z.
Jamalizadeh Tajabadi M R, Moghaddamnia A R, Piri J, and Ekhtesasi M R (2010) Application of artificial neural networks in dust storm prediction (Case study: Zabol city). Iranian journal of Range and Desert Research 17(2):205-220 (In Persian)
Karegar M E, Bodagh Jamali J, Ranjbar Saadat Abadi A, Moeenoddini M, and Goshtasb H (2017) Simulation and numerical analysis of severe dust storms Iran East. Journal of Spatial Analysis Environmental Hazards 3(4):101-119 (In Persian)
Kim D, Chin M, Kemp E M, Tao Z, Peters-Lidard C D, and Giroux P (2017) Development of high-resolution dynamic dust source function-A case study with a strong dust storm in a regional model. Atmospheric Environment 159:11-25
Liaghat A, Araghinejad Sh, Ansari Ghojghar M, Pourgholam Amiji M (1399) Comparing the performance of SARIMA and holt-winters time series models with artificial intelligence methods in forecasting dust storms (Case study: Sistan and Baluchistan Province.( Physical Geography Research Quarterly, DOI: 10.22059/jphgr.2021.303847.1007524.
Misra D, Oommen T, Agarwa A, Mishra S K, Thompson A M (2009) Application and analysis of supportvector machine based simulation for runoff andsediment yield. Biosystems Engineering 103(3):527–535
Mohammadi G H (2015) Analysis of atmospheric mechanisms in dust transport over west of Iran. Ph.D. Thesis, Tabriz University, 142 pp. (In Persian)
O’Loingsigh T, McTainsh G H, Tews E K, Strong C L, Leys J F, Shinkfield P, and Tapper N J (2014) The Dust Storm Index (DSI): A method for monitoring broadscale wind erosion using meteorological records. Aeolian Research 12:29-40
Shaiba H, Alaashoub N S, Alzahrani A A (2018) Applying machine learning methods for predicting sand storms. 2018 1st International Conference on Computer Applications & Information Security (ICCAIS), Riyadh, Saudi Arabia
Shaker Sureh F and Asadi E (2019) Meteorological and hydrological drought communication in Salmas Plain. Desert Ecosystem Engineering Journal 8(22):89-100 (In Persian)
Sobhani B, Salahi B, and Goldust A (2015) Study the dust and evaluation of its possibility prediction based on statistical methods and ANFIS model in Zabol University. Geography and Development Iranian Journal 13(38):123-138  (In Persian)
Vapnik VN (1995) The nature of statistical learning theory. Springer, New York
Zeinali B (2016) Investigation of frequency changes trend of days with dust storms in western half of Iran. Journal of Natural Environment Hazards 5(7):100-87 (In Persian)
Zhu YM, Lu XX, Zhou Y (2007) Suspended sediment flux modeling with artificial neural network: An example of the longchuanjiang river in the upper yangtze catchment. Geomorphology 84(4):111-125