Please wait a minute...
Journal of Arid Land  2020, Vol. 12 Issue (5): 854-864    DOI: 10.1007/s40333-020-0097-3
Research article     
Precipitation forecasting by large-scale climate indices and machine learning techniques
University of Tehran, Tehran 1417466191, Iran
Download: HTML     PDF(406KB)
Export: BibTeX | EndNote (RIS)      


Global warming is one of the most complicated challenges of our time causing considerable tension on our societies and on the environment. The impacts of global warming are felt unprecedentedly in a wide variety of ways from shifting weather patterns that threatens food production, to rising sea levels that deteriorates the risk of catastrophic flooding. Among all aspects related to global warming, there is a growing concern on water resource management. This field is targeted at preventing future water crisis threatening human beings. The very first stage in such management is to recognize the prospective climate parameters influencing the future water resource conditions. Numerous prediction models, methods and tools, in this case, have been developed and applied so far. In line with trend, the current study intends to compare three optimization algorithms on the platform of a multilayer perceptron (MLP) network to explore any meaningful connection between large-scale climate indices (LSCIs) and precipitation in the capital of Iran, a country which is located in an arid and semi-arid region and suffers from severe water scarcity caused by mismanagement over years and intensified by global warming. This situation has propelled a great deal of population to immigrate towards more developed cities within the country especially towards Tehran. Therefore, the current and future environmental conditions of this city especially its water supply conditions are of great importance. To tackle this complication an outlook for the future precipitation should be provided and appropriate forecasting trajectories compatible with this region's characteristics should be developed. To this end, the present study investigates three training methods namely backpropagation (BP), genetic algorithms (GAs), and particle swarm optimization (PSO) algorithms on a MLP platform. Two frameworks distinguished by their input compositions are denoted in this study: Concurrent Model Framework (CMF) and Integrated Model Framework (IMF). Through these two frameworks, 13 cases are generated: 12 cases within CMF, each of which contains all selected LSCIs in the same lead-times, and one case within IMF that is constituted from the combination of the most correlated LSCIs with Tehran precipitation in each lead-time. Following the evaluation of all model performances through related statistical tests, Taylor diagram is implemented to make comparison among the final selected models in all three optimization algorithms, the best of which is found to be MLP-PSO in IMF.

Key wordsbackpropagation      genetic algorithms      machine learning      multilayer perceptron      particle swarm optimization      Taylor diagram     
Received: 03 December 2019      Published: 10 September 2020
Corresponding Authors: MALEKIAN Arash     E-mail:
About author: *Corresponding author: Arash MALEKIAN (E-mail:
Cite this article:

Mehdi GHOLAMI ROSTAM, Seyyed Javad SADATINEJAD, Arash MALEKIAN. Precipitation forecasting by large-scale climate indices and machine learning techniques. Journal of Arid Land, 2020, 12(5): 854-864.

URL:     OR

Row Index Row Index
1 The Pacific/North American Pattern (PNA) 19 Arctic Oscillation (AO)
2 North Atlantic Oscillation (NAO) 20 Antarctic Oscillation (AAO)
3 West Pacific Pattern (WP) 21 Southern Oscillation Index (SOI)
4 North Pacific Pattern (NP) 22 Central Indian Precipitation
5 East Pacific Pattern (EP) 23 Northeast Brazil Rainfall Anomaly
6 Pacific Decadal Oscillation (PDO) 24 Tropical Northern Atlantic (TNA)
7 Eastern Pacific Oscillation (EPO) 25 Tropical Southern Atlantic (TSA)
8 North Oscillation Index (NOI) 26 Atlantic Meridional Mode (AMM)
9 El Nino - Southern Oscillation (ENSO) 27 Atlantic Multi-decadal Oscillation (AMO)
10 Multivariate ENSO Index (MEI) 28 Western Hemisphere Warm Pool (WHWP)
11 Extreme Eastern Tropical Pacific SST (Nino 1+2) 29 North Tropical Atlantic SST Index (NTA)
12 Central Tropical Pacific SST (Nino 4) 30 Oceanic NINO Index (ONI)
13 East Central Tropical Pacific SST (Nino 3.4) 31 Trans Nino Index (TNI)
14 Sahel Standardized Rainfall 32 Pacific Warm pool (PWP)
15 Eastern Asia/ Western Russia (EA/WR) 33 Indian Ocean Dipole (IOD)
16 Caribbean Index (CAR) 34 Solar Flux
17 Bivariate ENSO Time series (BEST) 35 Monthly totals Atlantic hurricanes and named tropical storms
18 Quasi-Biennial Oscillation (QBO) 36 North Sea-Caspian Pattern (NCP)
Table 1 Initial large-scale climate indices (National Oceanic and Atmospheric Administration (2015)
Fig. 1 Schematic study trajectory diagram. As can be seen, after providing the inputs of two frameworks (i.e., concurrent model framework and integrated model framework), three methods, namely, backpropagation-based (BP-based) multilayer perceptron (MLP), genetic algorithm-based (GA-based) MLP and particle swarm optimization-based (PSO-based) MLP were targeted.
Fig. 2 Schematic topology of a three-layer MLP (Multilayer perceptron) applied in this study. In the input layer for each input, one neuron should be considered; the number of the neurons of the hidden layer in each model was selected based on trial and error and represented in its relating section; the number of the neurons in the output layer, as can be seen, is the modeled precipitation measure for each month.
Fig. 3 Diagram of statistical comparison among the models. The reference point on x-axis is the standard deviation of the observed dataset and the radial distance from this point (dashed contour) represents the RMSE (root mean square error) of the model; the dotted contours, which are the radial distance from the origin, shows the standard deviation of the simulated sets. It is clear that the most precise model is the one closer to the reference point (Taylor, 2001).
Fig. 5 Performance of all generated cases in three optimization methods. The 0-11 of x-axis shows the framework of CMF (concurrent model framework). IMF, integrated model framework. MAE, mean absolute error; RMSE, root mean square error.
Time scale Model framework Lead-time (month) Z
Monthly CMF 3 0.84
Monthly CMF 8 0.62
Monthly IMF - 0.56
Table 2 Z-test for selected cases in each method
Method Model framework Lead-time (month) Training Validation Test
(mm) (mm) (mm) (mm) (mm) (mm)
BP-based MLP CMF 3 18.53 12.10 20.43 14.90 19.37 12.38
GA-based MLP CMF 8 19.34 14.61 20.36 14.26 17.39 13.59
PSO-based MLP IMF - 21.21 15.12 19.17 13.69 18.54 12.97
Table 3 Selected models and their RMSE (root mean square error) and MAE (mean absolute error) under training, validation and test steps
Fig. 6 Result of Taylor diagram. As can be seen, the performance of all cases are in an approximately same range; however, the MLP-PSO is the closest case to the reference point.
[1]   Abbot J, Marohasy J. 2018. Forecasting of medium-term rainfall using artificial neural networks: case studies from eastern Australia. doi: 10.5772/intechopen.72619.[2017-12-27].
[2]   Abdi H, Williams L J. 2010. Principal component analysis. Wiley Interdisciplinary Reviews: Computational Statistics, 2: 433-459.
[3]   Allan R J, Beard G S, Close A, et al. 1996. Mean sea level pressure indices of the El Nino-Southern oscillation: relevance to stream discharge in south-eastern Australia. Divisional report. Canberra: CSIRO Division of Water Resources, 96/1.
[4]   Araghinejad S, Meidani E. 2013. A review of climate signals as predictors of long-term hydro-climatic variability. Climate Variability. doi:
[5]   Arvin A. 2015. Relationship between El-Nino-southern oscillation (ENSO) and total ozone variations in Iran. Geography and Development Iranian Journal, 12: 165-180. (In Farsi)
[6]   Ashrafi K, Shafiepour M, Ghasemi L, et al. 2012. Prediction of climate change induced temperature rise in regional scale using neural network. International Journal of Environmental Research, 6(3): 677-688.
[7]   Bensingh R J, Machavaram R, Boopathy S R, et al. 2019. Injection molding process optimization of a bi-aspheric lens using hybrid artificial neural networks (ANNs) and particle swarm optimization (PSO). Measurement, 134: 359-374.
[8]   Bjerknes J. 1969. Atmospheric teleconnections from the equatorial Pacific. Monthly Weather Review, 97(3): 163-172.
[9]   Brandimarte L, di Baldassarre G, Bruni G, et al. 2011. Relation between the North-Atlantic oscillation and hydroclimatic conditions in Mediterranean areas. Water Resource Management, 25: 1269-1279.
[10]   Bratton D, Kennedy J. 2007. Defining a standard for particle swarm optimization. Proceedings of the 2007 IEEE Swarm Intelligence Symposium, 120-127, doi: 10.1109/SIS.2007.368035.
[11]   Canon J, Gonzalez J, Valdez J. 2007. Precipitation in the Colorado River basin and its low frequency associations with PDO and ENSO signals. Journal of Hydrology, 333(2-4): 252-264.
[12]   Choubin B, Khalighi-Sigaroodi S, Malekian A, et al. 2014. Drought forecasting in a semi-arid watershed using climate signals: a neuro-fuzzy modeling approach. Journal of Mountain Science, 11: 1593-1605.
[13]   Degefu M A, Bewket W. 2017. Variability, trends, and teleconnections of stream flows with large-scale climate signals in the Omo-Ghibe river basin, Ethiopia. Environmental Monitoring and Assessment, 189(4): 142.
pmid: 28258340
[14]   Garro B, Vazquez R. 2015. Designing artificial neural networks using particle swarm optimization algorithms. Computational Intelligence and Neuroscience, ID 369298.
doi: 10.1155/2020/8825197 pmid: 33082776
[15]   Gaughan A E, Waylen P R. 2012. Spatial and temporal precipitation variability in the Okvango-Kwando-Zambezi catchment, southern Africa. Journal of Arid Environments, 82: 19-30.
[16]   Gerkaninezhad M Z, Bazrafshan O. 2018. Impact of climatic signals on the wet and dry season precipitation (case study: Persian Gulf and Oman Sea watersheds). Journal of the Earth and Space Physics, 44: 333-349. (In Farsi)
[17]   Ghazal R, Ardeshir A, Zahedi Rad I, 2014. Climate change and storm-water management strategies in Tehran. Procedia Engineering, 89: 780-787.
[18]   Gong D, Ho C. 2003. Detection of large-scale climate signals in spring vegetation index (normalized difference vegetation index) over the Northern Hemisphere. Journal of Geophysical Research, 108(D16): 4498.
[19]   Hatzaki M, Flocas H, Asimakopoulos D, et al. 2007. The eastern Mediterranean teleconnection pattern. International Journal of Climatology, 27(6): 727-737.
[20]   Hidalgo H, Dracup J. 2003. ENSO and PDO effects on hydroclimatic variations of the upper Colorado River basin. Journal of Hydrometeorology, 4(1): 5-23.
[21]   Hurrell J W. 1995. Decadal trends in the north Atlantic oscillation: regional temperatures and precipitation. Science, 269(5224): 676-679.
doi: 10.1126/science.269.5224.676 pmid: 17758812
[22]   Jiang M, Luo Y, Yang S. 2007. Particle swarm optimization-stochastic trajectory analysis and parameter selection. In: Felix T S C, Tiwari M K. Swarm Intelligence, Focus on Ant and Particle Swarm Optimization. doi: 10.5772/5104.
[23]   Jolliffe I T. 2002. Principal Component Analysis (2nd ed.). New York: Springer, 2.
[24]   Jones P, Jonsson T, Wheeler D. 1997. Extension to the north Atlantic oscillation using early instrumental pressure observations from Gibraltar and south-west Iceland. International Journal of Climatology, 17(13): 1433-1450.
[25]   Kaiser H. 1960. The application of electronic computers to factor analysis. Educational and Psychological Measurements, 20: 141-151.
[26]   Kampichler C, van Turnhout C, Devictor V, et al. 2012. Large-scale changes in community composition: determining land use and climate change signals. PLoS ONE, 7(4): e35272.
[27]   Karabok M, Kahya E, Karaca M. 2005. The influences of the Southern and North Atlantic oscillations on climatic surface variables in Turkey. Hydrological Processes, 19(6): 1185-1211.
[28]   Kriesel D. 2007. A Brief Introduction to Neural Networks. [2020-01-20]. >.
[29]   Mann P. 1997. Introductory Statistics (3rd ed.). New York: Wiley, 405.
[30]   Matyasovszky I. 2003. The relationship between NAO and temperature in Hungary and its nonlinear connection with ENSO. Theoretical and Applied Climatology, 74: 69-75.
doi: 10.1007/s00704-002-0697-1
[31]   Mitchell M. 1996. An Introduction to Genetic Algorithms. Cambridge: MIT Press, 3.
[32]   Nigam S, Shen H. 1993. Structure of oceanic and atmospheric low-frequency variability over the tropical Pacific and Indian Oceans. Journal of Climatology, 6(4): 657-676.
[33]   Oldenberg-van G, Burgers G, Tank A. 2000. On the El-Nino teleconnection to spring precipitation in Europe. International Journal of Climatology, 20(5): 565-574.
[34]   Ouyang R, Liu W, Fu G, et al. 2014. Linkages between ENSO/PDO signals and precipitation, stream-flow in China during the last 100 years. Hydrology and Earth System Science, 18(9): 3651-3661.
[35]   Pasini G. 2017. Principal component analysis for stock portfolio management. International Journal of Pure and Applied Mathematics, 115(1): 153-167.
[36]   Popescu M, Balas V, Popescu L, et al. 2009. Multilayer perceptron and neural networks. WSEAS Transactions on Circuits and Systems, 8(7): 579-588.
[37]   Pozo-Vazquez D, Gamiz-Fortis S R, Tovar-Pescador J, et al. 2005. El Nino-Southern oscillation events and associated European winter precipitation anomalies. International Journal of Climatology, 25(1): 17-31.
[38]   Prabhu M V, Karthikeyan R. 2018. Comparative studies on modelling and optimization of hydrodynamic parameters on inverse fluidized bed reactor using ANN-GA and RSM. Alexandria Engineering Journal, 57(4): 3019-3032.
[39]   Qui S, Chen B, Wang R, et al. 2018. Atmospheric dispersion prediction and source estimation of hazardous gas using artificial neural network, particle swarm optimization and expectation maximization. Atmospheric Environment, 178: 159-163.
[40]   Rosenblatt F. 1961. Principles of Neurodynamics: Perceptron and the Theory of Brain Mechanisms. Washington DC: Spartan Books, 245.
[41]   Saji N, Goswami B, Vinayachandran P, et al. 1999. A dipole mode in the tropical Indian Ocean. Nature, 401: 360-363.
[42]   Santos J, Corte J, Leite S. 2005. Weather regimes and their connection to the winter rainfall in Portugal. International Journal of Climatology, 25(1): 33-50.
[43]   Seiffert U. 2001. Multiple layer Perceptron training using genetic algorithms. Bruges: Proceedings of European Symposium on Artificial Neural Networks, 159-164.
[44]   Srinivasan D, Seow T H. 2003. Particle swarm inspired evolutionary algorithm (PS-EA) for multi-objective optimization problem. Acta Biomaterialia, 4: 2292-2297.
[45]   Taylor K E. 2001. Summarizing multiple aspects of model performance in a single diagram. Journal of Geophysical Research, 106(D7): 7183-7192.
[46]   Tyson P. 1987. Climate change and variability in southern Africa. The Quarterly Journal of Royal Meteorological Society, 8: 552-562.
[47]   Wallace J, Gutzler D. 1981. Teleconnections in the geopotential height field during the northern hemisphere winter. Monthly Weather Review, 109(4): 748-812.
[48]   Whitley D. 1994. A genetic algorithm tutorial. Statistics and Computing, 4: 65-85.
[49]   Wittan I H, Frank E. 2005. Data Mining: Practical Machine Learning Tools and Techniques (2nd ed.). San Francisco: Elsevier, 29.
[50]   Xu L, Chen N, Zhang X. 2018. A comparison of large-scale climate signals and the North American multi-model ensemble (NMME) for drought prediction in China. Journal of Hydrology, 557: 378-390.
[51]   Xu Z, Hou Z, Han Y, et al. 2016. A diagram for evaluating multiple aspects of model performance in simulating vector fields. Geoscientific Model Development, 9: 4365-4380.
[52]   Zahraei B, Karamouz M. 2004. Seasonal precipitation prediction using large scale climate signals. Salt Lake City: World Water and Environmental Resources Congress.
No related articles found!