Elsevier

Cognitive Systems Research

Volume 70, December 2021, Pages 109-116
Cognitive Systems Research

Cognitive computing models for estimation of reference evapotranspiration: A review

https://doi.org/10.1016/j.cogsys.2021.07.012Get rights and content

Highlights

  • Irrigation practices can be advanced by the aid of cognitive computing models.

  • The cognitive computing model outperforms emprical methods over estimation of reference evapotranspiration (ET0).

  • ANN, SVM and Genetic programming play crucial role in ET0 prediction.

  • Second order neural network (SONN) is the most promising approach for prediction of ET0.

Abstract

Irrigation practices can be advanced by the aid of cognitive computing models. Repeated droughts, population expansion and the impact of global warming collectively impose rigorous restrictions over irrigation practices. Reference evapotranspiration (ET0) is a vital factor to predict the crop water requirements based on climate data. There are many techniques available for the prediction of ET0. An efficient ET0 prediction model plays an important role in irrigation system to increase water productivity. In the present study, a review has been carried out over cognitive computing models used for the estimation of ET0. Review exhibits that artificial neural network (ANN) approach outperforms support vector machine (SVM) and genetic programming (GP). Second order neural network (SONN) is the most promising approach among ANN models.

Introduction

Economic advancement and increasing global population expand the demand for agricultural production. The global food requirement will increase about 60% by the year 2050. Globally, it is assessed that agricultural activity consumes about 70% of the gross water (Provenzano & Sinobas, 2014). The global irrigated land occupies only 16% of cultivatable area (Playán et al., 2014). The arid and semi-arid regions occupy 36% of the global land (Safriel et al., 2005) and it is predicted that drought hazard will further increase due to global warming (Alcamo et al., 2007, Arnell et al., 2011). Water productivity (WP) is the ratio between crop yield and total water use (Pereira et al., 2002). Water consumed by plants is less than 65% of supplied water (Chartzoulakis & Bertaki, 2015). Irrigation automation is challenging due to the spatial and temporal variability of soil, weather and plant data (Dabach et al., 2011, Soulis and Elmaloglou, 2018). The evolution of first generation irrigation technology started with multi-client electronic hydrants for utilization on the dispensation network. The second generation irrigation technology was variable frequency pump. The microirrigation method was the third generation in irrigation technology wherein WP was increased but marginally installed due to high initial investment. The sub–surface drip irrigation (SDI) was the fourth generation in irrigation technology that was invented to solve the issues of surface drip irrigation, specifically for the elimination of emitter clogging issue. The fifth generation in irrigation technology was deficit irrigation that was invented to supply reduced amount of water without affecting the yield based on crop growth stage (Levidow et al., 2014, Kang et al., 2017). The cognitive computing-based intelligent irrigation system is the current trend to provide economical and efficient models for agricultural water management (Haider et al., 2007, Torres-Rua et al., 2012, Behmann et al., 2014, Kamilaris et al., 2017, Niu et al., 2017, Chlingaryan et al., 2018, Kamilaris and Prenafeta-Boldú, 2018).

The rest of the paper structure is described as follows. Section 2 exhibits the most widely used empirical method for the estimation of ET0. The crop water requirements prediction model is highlighted in Section 3. The cognitive computing methods used for ET0 prediction are discussed in Section 4. The summary of the review is described in Section 5. The conclusion part is described in Section 6. Finally, cognitive computing-based irrigation system (CCIS) is outlined in Section 7 as future work.

Section snippets

Reference evapotranspiration (ET0) estimation model

The ET0 is an important metric to understand crop water requirements and obtain a satisfactory yield (Temesgen et al., 2005). The ET0 plays a vital role in the computation of irrigation water requirements, estimation of crop evapotranspiration (ETc) and to design soil water balance models. Weather data, such as air temperature (AT), wind speed (WS), solar radiation (SR), sunshine hours (SS), relative humidity (RH), rainfall (RF), air pressure (AP) and vapour pressure (VP) are used as key input

Crop water requirements prediction model

Crop water requirements play a crucial role in irrigation systems for the determination of water productivity (WP), wherein ETc is determined using ET0 and crop coefficient (Kc). The Kc indicates the actual plant transpiration. The crop water requirements prediction model based on crop growth stage and weather data are highlighted in Eq. (2) (Allen et al., 2005).Crop water requirement=Kc*ET0

Cognitive computing models for estimation of ET0

Cognitive computing is a domain of computer science that imitates the phenomenon of human brain (Gocić et al., 2015). Few aspects, such as consciousness and cognition, are key features of cognitive computing techniques. Cognitive computing methods exploit resistance for uncertainty and imprecision, ensure conformity and robustness and offer economical solutions (Keskin & Terzi, 2006). Cognitive computing techniques have been used for different agricultural applications to build intelligent and

Performance metrics for cognitive computing models

The key performance factors in cognitive computing models are local minima, convergence rate, over fitting and generalization potential. The review exhibited that MLFF-NN and GP models got blocked in local minima easily and that ANFIS + CSA and ENN models can be used to overcome this issue and predict ET0. The ANFIS + CSA, MLFF-NN and SONN models delay in convergence rate and NARX and ENN models can be used to address this issue and predict ET0. The MLFF-NN and GEP models are prone to over

Conclusion

The review of cognitive computing models for ET0 prediction was conducted to determine the most promising approach for irrigation automation. The survey firstly outlines the neural network approaches for reference evapotranspiration (ET0) prediction. The statistical performance analysis of neural network models exhibited that second order neural network (SONN) is the most promising method for ET0 prediction, considering the balance between RMSE and R2values.

Secondly, SVM approaches are outlined

Future irrigation system

Cognitive computing-based intelligent irrigation system (CCIS) can be designed, which offers a balance between high water productivity and economy. The CCIS approach preserves the benefit of low initial investment of traditional surface irrigation system with higher WP through the employment of supervised, unsupervised, semi-supervised and reinforcement learning techniques. The ET0 prediction plays a crucial role in CCIS to predict crop water requirements accurately.

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Acknowledgments

The authors express their gratefulness to the Rev. Fr. Augustine Paimpallil, CMI, Principal, Christ Institute, Bengaluru and Daisy Agnes, English Professor, for the manuscript’s proofreading, language improvement and style improvement.

References (64)

  • O. Kisi

    Pan evaporation modeling using least square support vector machine, multivariate adaptive regression splines and M5 model tree

    Journal of Hydrology

    (2015)
  • L. Levidow et al.

    Improving water-efficient irrigation: Prospects and difficulties of innovative practices

    Agricultural Water Management

    (2014)
  • G. Mountrakis et al.

    Support vector machines in remote sensing: A review

    ISPRS Journal of Photogrammetry and Remote Sensing

    (2011)
  • L.S. Pereira et al.

    Irrigation management under water scarcity

    Agricultural Water Management

    (2002)
  • A. Prieto et al.

    Advances in artificial neural networks and machine learning

    Neurocomputing

    (2013)
  • A. Prieto et al.

    Neural networks: An overview of early research, current frameworks and new challenges

    Neurocomputing

    (2016)
  • M. Ravansalar et al.

    A wavelet–linear genetic programming model for sodium (Na+) concentration forecasting in rivers

    Journal of Hydrology

    (2016)
  • M. Shehab et al.

    A survey on applications and variants of the cuckoo search algorithm

    Applied Soft Computing

    (2017)
  • K.X. Soulis et al.

    Optimum soil water content sensors placement for surface drip irrigation scheduling in layered soils

    Computers and Electronics in Agriculture

    (2018)
  • A. Wunsch et al.

    Forecasting groundwater levels using nonlinear autoregressive networks with exogenous input (NARX)

    Journal of Hydrology

    (2018)
  • A. Ab. Ghani et al.

    Gene-expression programming for sediment transport in sewer pipe systems

    Journal of Pipeline Systems Engineering and Practice

    (2011)
  • H.Z. Abyaneh et al.

    Performance evaluation of ANN and ANFIS models for estimating garlic crop evapotranspiration

    Journal of Irrigation and Drainage Engineering

    (2011)
  • S. Adamala et al.

    Evapotranspiration modeling using second-order neural networks

    Journal of Hydrologic Engineering

    (2014)
  • J. Alcamo et al.

    Future long-term changes in global water resources driven by socio-economic and climatic changes

    Hydrological Sciences Journal

    (2007)
  • M. Alizamir et al.

    Modelling reference evapotranspiration by combining neuro-fuzzy and evolutionary strategies

    Acta Geophysica

    (2020)
  • R.G. Allen et al.

    Crop evapotranspiration-Guidelines for computing crop water requirements-FAO Irrigation and drainage paper 56

    Fao, Rome

    (1998)
  • R.G. Allen et al.

    FAO-56 dual crop coefficient method for estimating evaporation from soil and application extensions

    Journal of irrigation and drainage engineering

    (2005)
  • R.G. Allen et al.

    FAO-24 reference evapotranspiration factors

    Journal of Irrigation and Drainage Engineering

    (1991)
  • J. Behmann et al.

    A review of advanced machine learning methods for the detection of biotic stress in precision crop protection

    Precision Agriculture

    (2014)
  • C.M. Burt et al.

    Evaporation Research: Review and Interpretation

    Journal of Irrigation and Drainage Engineering

    (2005)
  • D. Chen

    Daily reference evapotranspiration estimation based on least squares support vector machines

    Computer and Computing Technologies in Agriculture V

    (2012)
  • S. Dabach et al.

    Numerical investigation of irrigation scheduling based on soil water status

    Irrigation Science

    (2011)
  • Cited by (25)

    • Forecasting weekly reference evapotranspiration using Auto Encoder Decoder Bidirectional LSTM model hybridized with a Boruta-CatBoost input optimizer

      2022, Computers and Electronics in Agriculture
      Citation Excerpt :

      Nonetheless, the ETo computations using the FAO-56 PM approach served as the sole basis for estimating this input–output relationship. The topic has sparked a flurry of studies into the development of various forms of ML algorithms (Faskari et al., 2022; Naganna et al., 2020), as seen by the recent reviews in this domain lately (Chia et al., 2020; Dong et al., 2021; Hebbalaguppae Krishnashetty et al., 2021; Kumar et al., 2012; Mokari et al., 2022). Some examples of these computing models include adaptive neuro-fuzzy inference (Tao et al., 2018), artificial neural network (Sanikhani et al., 2018), gene expression programming (Muhammad et al., 2021), multivariate adaptive regression spline (Adnan et al., 2021), hybrid dynamic evolving neural fuzzy inference system (Ye et al., 2022), extreme learning machine (Granata, 2019), support vector machine (Ferreira et al., 2019), XGBoost algorithm (Han et al., 2019), gradient boosting decision tree (Zhao et al., 2021), and several version of hybrid ML algorithms tuned using nature-inspired algorithms (Roy et al., 2021; Ruiming and Shijie, 2020; Tikhamarine et al., 2019).

    View all citing articles on Scopus
    View full text