Elsevier

Journal of Hydrology

Volume 555, December 2017, Pages 397-406
Journal of Hydrology

Research papers
A binary genetic programing model for teleconnection identification between global sea surface temperature and local maximum monthly rainfall events

https://doi.org/10.1016/j.jhydrol.2017.10.039Get rights and content

Highlights

  • We developed a new model (BGP) for teleconnection studies.

  • The BGP classifies monthly rainfall events.

  • The BGP can forecast rainfall classes based on global SST variations.

  • The BGP exhibits teleconnection signals with explicit mathematical expressions.

Abstract

The effectiveness of genetic programming (GP) for solving regression problems in hydrology has been recognized in recent studies. However, its capability to solve classification problems has not been sufficiently explored so far. This study develops and applies a novel classification-forecasting model, namely Binary GP (BGP), for teleconnection studies between sea surface temperature (SST) variations and maximum monthly rainfall (MMR) events. The BGP integrates certain types of data pre-processing and post-processing methods with conventional GP engine to enhance its ability to solve both regression and classification problems simultaneously. The model was trained and tested using SST series of Black Sea, Mediterranean Sea, and Red Sea as potential predictors as well as classified MMR events at two locations in Iran as predictand. Skill of the model was measured in regard to different rainfall thresholds and SST lags and compared to that of the hybrid decision tree-association rule (DTAR) model available in the literature. The results indicated that the proposed model can identify potential teleconnection signals of surrounding seas beneficial to long-term forecasting of the occurrence of the classified MMR events.

Introduction

There is an increasing interest in the hydrological community to evaluate the characteristics of extreme weather events (Danandeh Mehr and Kahya 2017b). The social and environmental impacts of these events make it necessary to improve our knowledge of their nature and develop forecasting techniques on a long-term basis. Recent studies in different regions have shown that large-scale ocean-atmospheric indices provide significant predictive information about extreme events (e.g., Kahya and Dracup, 1993, Stahl and Demuth, 1999, Rucong et al., 2001, Camberlin et al., 2001, Rowell, 2003, Ghasemi and Khalili, 2006, Segele et al., 2009, Nazemosadat and Ghaedamini, 2010, Chun et al., 2013, Danandeh Mehr et al., 2014, Elsanabary and Gan, 2015, Degefu and Bewket, 2017). For example, Rucong et al. (2001) indicated significant relationship between sea surface temperature (SST) of the Pacific Ocean and summer monsoon in Mid-Eastern China. Rowell (2003) showed that the anomalies of Mediterranean SSTs have a considerable impact on wet season precipitation over the Sahel, Africa. Stefan et al. (2004) showed that during persistent positive phase of North Atlantic Oscillation (NAO), meteorological and hydrological drought conditions are expected in southern Romania. Revadekar and Kulkarni (2008) studied the impact of the El Niño-Southern Oscillation (ENSO) on India's winter extreme rainfall events and demonstrated that ENSO index can be used for the estimation of extreme rainfall events in the region, four to six months in advance. More recently, using a simple correlation analysis, Degefu and Bewket (2017) reported that that ENSO and the Indian Ocean Dipole (IOD) are the major anomalies affecting extreme floods in Omo-Ghibe River, Ethiopia. There are also some evidence reporting strong teleconnection between ocean-atmospheric patterns and climate systems across Iran (e.g., Nazemosadat and Ghasemi, 2004, Ghasemi and Khalili, 2006, Nazemosadat and Ghaedamini, 2010, Sabziparvar, et al., 2011, Hosseinzadeh Talaee et al., 2014). For instance, the effects of ENSO on the 49-year precipitation distribution in Iran was studied by Nazemosadat et al. (2006). The results revealed that around the mid-1970s, the precipitation has significantly enhanced in southwest and north of Iran as a result of consistent increase in frequency and intensity of El Niño events. Nazemosadat and Ghaedamini (2010) demonstrated strong linkage between Madden–Julian oscillation (MJO) and daily, monthly, and seasonal precipitation over Southern Iran and Arabian Peninsula. Sabziparvar et al. (2011) studied the impacts of different ENSO phases on reference evapotranspiration variability and demonstrated significant correlations between ENSO events and seasonal evapotranspiration variations in 54% of their study sites.

A variety of traditional approaches such as auto-regressive moving average, linear and nonlinear regression/correlation were used in the aforementioned studies to identify the potential linkages and feedbacks between large-scale patterns and terrestrial precipitation variability. However, these methods are not robust enough to characterize the complex nonlinear nature of teleconnection signals (Chang et al., 2017a). Therefore, the outcomes of such studies are not used for operational forecasting. To cope with the problem of nonlinearity, data-driven techniques such as artificial neural networks (ANN), extreme learning machine (ELM), association rules (AR), and genetic programing (GP) are suggested in recent studies (e.g., Mwale and Gan, 2005, Mwale and Gan, 2010, Kashid and Maity, 2012, Danandeh Mehr et al., 2014, Chang, et al., 2017b). For example, linear GP was used by Kashid et al. (2010) to predict gridded multi-site weekly rainfall from ENSO indices, Equatorial Indian Ocean Oscillation (EQUINOO) indices, Outgoing Longwave Radiation (OLR), and lag rainfall at grid points, over the Mahanadi catchment, India. The authors demonstrated that the model outputs can be used satisfactorily to predict streamflow in the catchment. Satisfactory application of the wavelet-ANN-GA models for long-term rainfall forecasting at southern Africa and upper Blue Nile Basin were reported by Mwale et al., 2007, Elsanabary and Gan, 2014, respectively. Our review concerning the application of data-driven techniques in teleconnection studies across Iran showed only a few studies. For example, Meidani and Araghinejad (2014) investigated the relation between the streamflow and precipitation variability with anomalies of the Mediterranean SSTs over southwest of Iran and indicated superiority of SSTs series to some other climate indices as the potential predictors. More recently, Nourani et al. (2017) suggested a new hybrid decision tree-association rule (DTAR) model to determine the dominant predictors and monitor the maximum monthly precipitation events. The authors showed that SST variation over Red Sea, Mediterranean Sea, and Black Sea influence the monthly precipitation amounts in North West of Iran. The authors also demonstrated that geographical location of stations and the distribution of precipitation data affect the measures of the rules and forecasting outcomes.

With the motivation of implementing power of teleconnections signals to forecast extreme rainfall events on a long-term basis, the main objectives of this study is to develop a new GP-based model, termed binary-GP (BGP), so that SST signals are explicitly linked to classified extreme rainfall events. The proposed BGP model is trained and tested using maximum monthly rainfall (MMR) records of two rain gauges stations at North West of Iran and SST series of the Black Sea, Mediterranean Sea, and Red Sea. Although GP, as supervised machine learning method, has already been implemented for regression-based teleconnection studies (e.g., Kashid and Maity 2012), to our knowledge the potential of GP to solve binary classification problems in hydrology has never been explored. In this sense, a particular attention is given to the data pre- and post-processing techniques as well as evaluation criteria required for the verification of the proposed BGP model. Given an extra attempt, skill of the proposed model is compared with those of the hybrid DTAR model previously reported by Nourani et al. (2017). From a hydrological standpoint, while earlier works categorized climatic indices (i.e., predictors) within rigid bounds (e.g., Tadesse, et al., 2004, Dhanya and Kumar, 2009), the present study investigates teleconnection between climatic indices and target variable using flexible thresholds (here three rainfall thresholds). Inasmuch as the proposed model explicitly links the large-scale indices to the basin-scale results, it is motivating to be used in practice.

Section snippets

Study area and data

In this study, the monthly maximum rainfall (here after MMR) records of two rain gauge stations of Iran including Tabriz (38.05°N, 46.17°E) and Kermanshah (34.21°N, 47.90°E) Stations as well as SSTs of the surrounding seas comprising Black Sea, Mediterranean Sea, and Red Sea are used (Fig. 1). Both stations are located in the northwest of the country with the elevation about 1350 m above sea level. The main reasons behind the selection of these gauges are their locations and length of available

Results and discussion

The DCCFs between MMR and SST series at each rain gauge station for values of i = 0, ±1, ±2, …, ±24 have been presented in Fig. 5. The figure also provides corresponding 95% confidence for each function. These are helpful for identifying optimum lags of the detrended SST series of three surrounding seas that might be useful predictors of MMRt. The figure shows significant positive correlation at lags 4–6 with annual periodicity at both stations. Therefore, four to six month lag might be an

Concluding remarks

This paper introduced the development and application of a new classification-forecasting model, namely BGP for teleconnection studies between oceanic and station-based hydroclimatological variables. It is a three-phase explicit model that integrates the certain types of data pre-processing and post-processing approaches with monolithic GP engine to enhance the ability of GP to solve both classification and regression problems, simultaneously. In the proposed model, SST series at different

Acknowledgements

This research has been carried out as a part of a postdoctoral research project at University of Tabriz funded by the Iran's National Elites Foundation (BMN). The authors gratefully acknowledge Technology Affairs of University of Tabriz for their tremendous help during the research. The authors also thank Iran Meteorological Organization and Earth System Research Laboratory, NOAA for providing the data used in this study.

References (50)

  • M. Ravansalar et al.

    Wavelet-linear genetic programming: a new approach for modeling monthly streamflow

    J. Hydrol.

    (2017)
  • M.B. Richman et al.

    Attribution and prediction of maximum temperature extremes in SE Australia

    Procedia Comput. Sci.

    (2014)
  • M. Shoaib et al.

    Runoff forecasting using hybrid wavelet gene expression programming (WGEP) approach

    J. Hydrolo

    (2015)
  • M. Sokolova et al.

    A systematic analysis of performance measures for classification tasks

    Inf. Process. Manage.

    (2009)
  • U. Bhowan et al.

    Genetic programming for classification with unbalanced data

  • P. Camberlin et al.

    Seasonality and atmospheric dynamics of the teleconnection between African rainfall and tropical sea-surface temperature: Atlantic vs ENSO

    Int. J. Climatol.

    (2001)
  • N.-B. Chang et al.

    The impact of global unknown teleconnection patterns on terrestrial precipitation across North and Central America

    Atmos. Res.

    (2017)
  • K.P. Chun et al.

    Prediction of the impact of climate change on drought: an evaluation of six UK catchments using two stochastic approaches

    Hydrol. Process.

    (2013)
  • A. Danandeh Mehr et al.

    Climate change impacts on catchment-scale extreme rainfall variability: case study of Rize Province, Turkey

    J. Hydrol. Eng.

    (2017)
  • M.A. Degefu et al.

    Variability, trends, and teleconnections of stream flows with large-scale climate signals in the Omo-Ghibe River Basin, Ethiopia

    Environ. Monitor. Assess.

    (2017)
  • C.T. Dhanya et al.

    Data mining for evolution of association rules for droughts and floods in India using climate inputs

    J. Geophys. Res. Atmos.

    (2009)
  • M.H. Elsanabary et al.

    Wavelet analysis of seasonal rainfall variability of the upper blue Nile basin, its teleconnection to global sea surface temperature, and its forecasting by an artificial neural network

    Mon. Weather Rev.

    (2014)
  • A.R. Ghasemi et al.

    The influence of the Arctic oscillation on winter temperatures in Iran

    Theoret. Appl. Climatol.

    (2006)
  • J. Han et al.

    Data mining: concepts and techniques

    (2006)
  • P. Hosseinzadeh Talaee et al.

    Hydrological drought in the west of Iran and possible association with large-scale atmospheric circulation patterns

    Hydrol. Process.

    (2014)
  • Cited by (33)

    • Z-numbers based novel method for assessing groundwater specific vulnerability

      2023, Engineering Applications of Artificial Intelligence
    • Factors affecting the presence of Arctic charr in streams based on a jittered binary genetic programming model

      2022, Ecological Indicators
      Citation Excerpt :

      Additionally, we compared the results of the novel JBGP approach with those of BGP, classic DT, and state-of-the-art jittered DT (JDT) approaches, as the benchmarks. The BGP (Danandeh Mehr et al., 2017) is an emerging ML method that applies monolithic GP to classify the results attained via explicit regression models for a given process. DT is a frequently used classification algorithms in practice, and JDT is a hybrid DT model that uses DT to classify jittered data.

    • Fuzzy-based large-scale teleconnection modeling of monthly precipitation

      2022, Handbook of HydroInformatics: Volume II: Advanced Machine Learning Techniques
    • Metrics of the water performance engineering modeling

      2022, Handbook of HydroInformatics: Volume II: Advanced Machine Learning Techniques
    • Historical changes in hydroclimatic extreme events over Iran

      2022, Climate Impacts on Extreme Weather: Current to Future Changes on a Local to Global Scale
    View all citing articles on Scopus
    View full text