Elsevier

Journal of Hydrology

Volumes 454–455, 6 August 2012, Pages 26-41
Journal of Hydrology

Prediction of monthly rainfall on homogeneous monsoon regions of India based on large scale circulation patterns using Genetic Programming

https://doi.org/10.1016/j.jhydrol.2012.05.033Get rights and content

Summary

Prediction of Indian Summer Monsoon Rainfall (ISMR) is of vital importance for Indian economy, and it has been remained a great challenge for hydro-meteorologists due to inherent complexities in the climatic systems. The Large-scale atmospheric circulation patterns from tropical Pacific Ocean (ENSO) and those from tropical Indian Ocean (EQUINOO) are established to influence the Indian Summer Monsoon Rainfall. The information of these two large scale atmospheric circulation patterns in terms of their indices is used to model the complex relationship between Indian Summer Monsoon Rainfall and the ENSO as well as EQUINOO indices. However, extracting the signal from such large-scale indices for modeling such complex systems is significantly difficult. Rainfall predictions have been done for ‘All India’ as one unit, as well as for five ‘homogeneous monsoon regions of India’, defined by Indian Institute of Tropical Meteorology. Recent ‘Artificial Intelligence’ tool ‘Genetic Programming’ (GP) has been employed for modeling such problem. The Genetic Programming approach is found to capture the complex relationship between the monthly Indian Summer Monsoon Rainfall and large scale atmospheric circulation pattern indices – ENSO and EQUINOO. Research findings of this study indicate that GP-derived monthly rainfall forecasting models, that use large-scale atmospheric circulation information are successful in prediction of All India Summer Monsoon Rainfall with correlation coefficient as good as 0.866, which may appears attractive for such a complex system. A separate analysis is carried out for All India Summer Monsoon rainfall for India as one unit, and five homogeneous monsoon regions, based on ENSO and EQUINOO indices of months of March, April and May only, performed at end of month of May. In this case, All India Summer Monsoon Rainfall could be predicted with 0.70 as correlation coefficient with somewhat lesser Correlation Coefficient (C.C.) values for different ‘homogeneous monsoon regions’.

Highlights

► The information of ENSO and EQUINOO is used to model monthly variation of Indian Summer Monsoon Rainfall. ► Separate analyses are carried out for All India rainfall and five homogeneous monsoon regions. ► ‘Genetic Programming’ (GP) has been employed for this modeling such problem and found to capture the complex relationship. ► Monthly variation of all-India monsoon rainfall is captured with correlation coefficient being 0.866. ► Influence of ENSO and EQUINOO on Rainfall is maximum for Central North-East India and minimum for Peninsular India.

Introduction

It is scientifically and mathematically challenging to use climate signals for the prediction of basin-scale hydrologic variables, because the climatic systems are very complex and physics of many systems is still not very clearly understood. The difficulties in modeling such complex systems are considerably reduced by the recent Artificial Intelligence tools like Artificial Neural Networks (ANNs); Genetic Algorithm (GA) based evolutionary optimizer and Genetic Programming (GP). Hence such AI tools are tried nowadays for modeling complex systems like basin-scale stream-flow forecasting using the information of large-scale atmospheric circulation phenomena.

Indian Summer Monsoon Rainfall is always found to vary annually leading to profound impacts on agriculture based Indian economy. Generally meteorological forecasts are generated for three timescales, viz. short-range (1–2 days ahead), medium-range (3–10 days ahead) and long-range forecasts for monthly and seasonal scales. In India, India Meteorological Department (IMD) generates the short and long-range predictions, whereas the National Centre for Medium Range Weather Forecasting (NCMRWF), New Delhi is responsible for the medium-range predictions.

Prediction of ISMR is having a long history. It started with the work of Sir Henry Blanford in 1886, which was entirely based on Himalayan snowfall. John Eliot used extra-Indian factors, viz. Pressure of Mauritus, Zanzibar and Seychelles in the monsoon forecast of 1896. Sir Gilbert Walker proposed statistical association for monsoon forecast. He systematically examined the relationship between Indian monsoon rainfall and global circulation parameters. He selected 28 predictors to issue forecast based on regression equation during the year 1906 (Jagannathan, 1960, Rao and Rama Moorthy, 1960, Rao, 1965). Most of the Walker’s predictors (except Himalayan snow accumulation) were signs of different facets of Southern Oscillation (Shukla and Paolino, 1983). Savur (1931) showed that 7 out of the 28 parameters had lost their significance in due course of time. Since then, extensive research work has been carried out on empirical seasonal forecasting of Indian Summer Monsoon Rainfall. Some of the noteworthy studies can be listed as following: Banerjee et al., 1978, Kung and Sharif, 1982, Bhalme et al., 1986, Gowariker et al., 1989, Gowariker et al., 1991, Parthasarathy et al., 1988, Parthasarathy et al., 1991, Parthasarathy et al., 1995, Krishna Kumar et al., 1995, Krishna Kumar et al., 1997, Rajeevan et al., 2004, etc. In spite of many efforts in the long range prediction of all-India summer monsoon rainfall (AISMR), it is felt that achieved success is not adequate and there is much scope to investigate new predictors and new methodologies of ISMR prediction.

Even though forecast for ‘All India Summer Monsoon Rainfall’ is available before every monsoon nowadays, still such forecasts have limited use due to significant variation in monsoon rainfall over the country in the same season. Hence it is felt that the rainfall forecasts can be more useful if issued at regional or sub-divisional scale. Hence this work attempts to develop models for issuing medium range forecasts of monthly monsoon rainfall from June through October, at all India level, as well as for five homogeneous monsoon regions of India. It is felt that such regional forecasts will have better utility than total ISMR due to availability of the information at smaller spatio-temporal scale.

Simultaneous variations of climatic conditions and hydrologic variables over widely separated regions on the surface of earth have long been discovered and noted by the meteorologists, world over. Such recurrent patterns are commonly referred to as “hydroclimatic teleconnection”. It is established that the natural variation of hydrologic variables is linked with these large-scale atmospheric circulation pattern through hydroclimatic teleconnection (Dracup and Kahya, 1994; Eltahir, 1996; Jain and Lall, 2001; Douglas et al., 2001; Ashok et al., 2001, Ashok et al., 2004; Marcella and Eltahir, 2008; Maity and Nagesh Kumar, 2008). Indian hydrometeorology is prominently influenced by two large-scale atmospheric circulation patterns. The first is El Niño-Southern Oscillation (ENSO) from tropical Pacific Ocean and second is the Equatorial Indian Ocean Oscillation (EQUINOO) from Indian Ocean. (Gadgil et al., 2004).

El Niño-Southern Oscillation (ENSO), which is a large-scale circulation pattern from tropical Pacific Ocean, is established to influence Indian Summer Monsoon Rainfall. Another large scale circulation pattern from Indian Ocean viz. Indian Ocean Dipole Mode (IOD) also influences the Indian Summer Monsoon rainfall (Saji et al., 1999).

Equatorial Indian Ocean Oscillation (EQUINOO) is the atmospheric component of the IOD mode (Gadgil et al., 2003, Gadgil et al., 2004). Gadgil et al. (2003) had shown that the Indian Summer Monsoon Rainfall is not only associated with ENSO, but also with EQUINOO. They suggest that one can scientifically predict the Indian Summer Monsoon Rainfall by knowing the prior EQUINOO status. Equatorial zonal wind index (EQWIN) is considered as an index of EQUINOO, which is defined as negative of the anomaly of the zonal component of surface wind in the equatorial Indian Ocean Region (60°E–90°E, 2.5°S–2.5°N). Weakening of ENSO-ISMR relationship is indicated by few researchers (Krishna Kumar et al., 1999). It is also established that ENSO-ISMR is modified by the influence of Indian Ocean Dipole (IOD) mode. Consideration of these two indices is found to give better results as compared to the analyses by researchers using just ENSO index (Gadgil et al., 2004). Thus, in this study, apart from ENSO index, EQUINOO index from Equatorial Indian Ocean is considered simultaneously, which is supposed to take care the temporal change in relationship between ENSO and ISMR.

Since nearly 80% of Indian Summer Monsoon Rainfall is due to the southwest monsoon, interaction between various oceans due to ENSO and EQUINOO regulates the amount and distribution of the rainfall over the sub continent. Such association is more prominent for the large aerial scale. It is also prominent for longer temporal scale (seasonal) or smaller temporal scale (monthly).

The search for a new methodology for predicting the All India Summer Monsoon Rainfall has been continued for a long time. The search is active for both, the long term as well as short term forecasts of Indian Summer Monsoon Rainfall. In recent years, an efficient Artificial Intelligence tool Genetic Programming has been used for modeling complex systems. Hence the Genetic Programming approach has been used to predict monthly Indian Summer Monsoon Rainfall over India in this study.

Section snippets

Objectives of the work

This work intends to develop models for medium range (1 month ahead) forecasts of monthly Indian Summer Monsoon rainfall for ‘All India’, as well as for five homogeneous monsoon regions of India, by using ENSO and EQUINOO indices as large scale atmospheric circulation information, with the help of Artificial Intelligence tool Genetic Programming. The study also deals with development of models for prediction of one time ISMR forecast on end of May, for All India Summer Monsoon Rainfall and five

Data

Sea surface temperature (SST) anomaly from the Niño 3.4 region (120°W–170°W, 5°S–5°N) is used as the ‘ENSO index’ in this study. Monthly sea surface temperature data from Niño 3.4 region for the period, January 1950 to December 2006, data are obtained from the website of the National Weather Service, Climate Prediction Centre of National Oceanic and Atmospheric Administration (NOAA) (http://www.cpc.noaa.gov/data/indices/). EQWIN, the negative of zonal wind anomaly over equatorial Indian Ocean

Methodology

Artificial Intelligence tools like Artificial Neural Networks, Support Vector Machines as well as Neuro-fuzzy Systems and data driven models have become popular tools for predictions in Water Resources related research problems (Drecourt, 1999; Giustolisi and Savic, 2006; Chau, 2006; Li et al., 2006; Lin et al., 2006; Shiri and Kisi, 2011; Muttil and Chau, 2007; Partal and Kisi, 2007). Many researchers have used Genetic Programming for modeling complex systems. (Babovic, 2000; Aytek and Kisi,

Results and discussions

As stated earlier, two separate analyses are carried out in this work. The first analysis uses real time monthly ENSO and EQUINOO indices of three previous monthly time steps for prediction of rainfall, which is nothing but medium range forecast with 2 weeks lead time by using monthly data of ENSO and EQUINOO in real time. The second analysis is one time analysis at the end of month of May, for computing monthly and hence monsoon rainfall in months of June through September at the end of May. It

Conclusions

Established research works indicate an association between the large-scale circulation pattern and hydrologic variables of large spatial and temporal scale. In this study, All India Summer Monsoon Rainfall as well as regional summer monsoon rainfall in India is investigated for possible influence of the large-scale circulation patterns on it. The ENSO and EQUINOO information is used as the large-scale input, which is established to be important for Indian hydroclimatology. Genetic Programming,

References (62)

  • K. Ashok et al.

    Individual and combined effect of ENSO and Indian Ocean Dipole on the Indian summer monsoon

    J. Clim.

    (2004)
  • V. Babovic

    Translating data to knowledge: a case study in sediment transport

    Comput.-Aided Civil Infrastruct. Eng.

    (2000)
  • V. Babovic

    Data mining in hydrology

    Hydrol. Process.

    (2005)
  • V. Babovic et al.

    The evolution of equation from hydraulic data, Part I: Theory

    J. Hydraul. Res.

    (1997)
  • V. Babovic et al.

    The evolution of equation from hydraulic data, Part II: Applications

    J. Hydraul. Res.

    (1997)
  • V. Babovic et al.

    Genetic programming as a model induction engine

    J. Hydroinform.

    (2000)
  • V. Babovic et al.

    Rainfall runoff modeling based on genetic programming

    Nord. Hydrol.

    (2002)
  • A.K. Banerjee et al.

    On foreshadowing southwest monsoon rainfall over India with mid-tropospheric circulation anomaly of April

    Indian J. Meteorol. Hydrol. Geophys.

    (1978)
  • M. Baptist et al.

    On inducing equations for vegetation resistance

    J. Hydraul. Res.

    (2007)
  • H.N. Bhalme et al.

    Forecasting of monsoon performance over India

    J. Climatol.

    (1986)
  • M. Brameier et al.

    Evolving teams of predictors with linear genetic programming

    Genetic Programming and Evolvable Machines

    (2001)
  • W.W. Douglas et al.

    The El Niño southern oscillation and long-range forecasting of flows in Ganges

    Int. J. Climatol.

    (2001)
  • J.A. Dracup et al.

    The relationship between U.S. streamflow and La Niña Events

    Water Resour. Res.

    (1994)
  • Drecourt, J.P., 1999. Application of neural networks and genetic programming to rainfall runoff modeling. Danish...
  • Drunpob, A., Chang, N.B., Beaman, M., 2005. Stream flow rate prediction using genetic programming model in a semi-arid...
  • E.A.B. Eltahir

    El Niño and the natural variability in the flow of the Nile River

    Water Resour. Res.

    (1996)
  • F.D. Francone

    Discipulus Owner’s Manual

    (1998)
  • S. Gadgil et al.

    Droughts of the Indian summer monsoon: role of clouds over the Indian Ocean

    Curr. Sci.

    (2003)
  • S. Gadgil et al.

    Extremes of the Indian summer monsoon rainfall: ENSO and equatorial Indian Ocean Oscillation

    Geograph. Res. Lett.

    (2004)
  • O. Giustolisi

    Using genetic programming to determine Chèzy resistance coefficient in corrugated channels

    J. Hydroinform., IWA-IAHR Publishing, UK

    (2004)
  • O. Giustolisi et al.

    A symbolic data-driven technique based on evolutionary polynomial regression

    J. Hydroinform., IWA-IAHR Publishing, UK

    (2006)
  • Cited by (83)

    • A multivariate EMD-LSTM model aided with Time Dependent Intrinsic Cross-Correlation for monthly rainfall prediction

      2022, Applied Soft Computing
      Citation Excerpt :

      Previously researchers used time series models instead of highly complicated physical models for rainfall predictions [1–4]. Later, machine learning (ML) based data driven models were introduced to handle the non-linear relationships between predictor variables and the rainfall [5–8]. Artificial Neural Networks (ANN) had been the most popular data driven method used for the prediction of dynamic, non-linear and non-stationary hydrological variables and a number of hybrid variants and improvisations involving ANN were proposed in the past [9].

    • Long-term simulation of daily rainfall across India: Performance of cumulus convection schemes in regional climate model during southwest and northeast monsoon

      2021, Atmospheric Research
      Citation Excerpt :

      Numerous studies have been conducted to understand the mechanism and physical processes of the Indian monsoon for last couple decades and still being continued (Shukla and Paolino, 1983; Webster et al., 1998; Goswami et al., 1998; Goswami and Ajaya Mohan, 2000; Lal et al., 2000; Singh et al., 2007; Maity et al., 2007; Rajeevan et al., 2008; Wang et al., 2015; Guhathakurta et al., 2015; Kakade and Kulkarni, 2016; Kishore et al., 2016; Devanand et al., 2018a; Dutta and Maity, 2018; Bhate and Kesarkar, 2019; Suman and Maity, 2020). Modelling of Indian monsoon is indeed a challenging task due to the lack of comprehensive understanding of the interactions between small scale processes and large scale oceanic/atmospheric circulations (Ajaya Mohan and Goswami, 2003; Maity et al., 2007; Kashid and Maity, 2012). General circulation models are mostly unable to capture the inter/intra seasonal variation of monsoon rainfall due to their coarse resolution and inability to simulate small scale physical processes (Gadgil and Sajani, 1998; Goddard et al., 2003), specially at finer spatial resolution (Krishna Kumar et al., 2005; Pattanaik and Kumar, 2010; Mohanty et al., 2018).

    • A synthesis of drought prediction research over India

      2021, Water Security
      Citation Excerpt :

      As noted earlier, the spatial heterogeneity of ISMR and its correlations with climatic teleconnections is one of the important factors responsible for low-skill forecasts. Owing to this, many recent studies have performed analysis at regional scale considering a set of homogenous regions in India [24–25,55,108,96]. However, a finer-scale analysis, without predefining homogenous regions based only on the regional climate, can be more valuable in finding the precursor regions and improving the forecasts.

    View all citing articles on Scopus
    View full text