Original papers
Evolutionary algorithm for reference evapotranspiration analysis

https://doi.org/10.1016/j.compag.2018.04.003Get rights and content

Highlights

  • Estimation of the reference evapotranspiration (ET0).

  • Weather parameters influence on reference evapotranspiration (ET0).

  • The accuracy of the evolutionary algorothm.

Abstract

Evapotranspiration of important indicator for management and planning of water resources. It is essential to analyze the evapotranspiration in order to improve water resources planning. The main goal of the study was to analyze the evapotranspiration based on several input parameters. It is important to estimate the influence of the input parameters on the evapotranspiration. For such a purpose evolutionary algorithm was applied. The algorithm applied in this article has space solution of genetic programs. Therefore this methodology is known as genetic programming. The input parameters in the model are monthly minimum and maximum air temperatures, sunshine hours, actual vapour pressure, minimum and maximum relative humidity and wind speed. Results presented in this study could be used for practical application of water resources planning and management based on the input parameters influence on the evapotranspiration.

Introduction

Evapotranspiration (ET) represents a complex and non-linear process natural which represent the vital component of the hydrologic cycle and therefore accurate estimation of the ET is very necessary for water resources management, catchment water balance and irrigation systems. Reference evapotranspiration (ET0) is often used to estimate actual ET. There are different methods and approaches for the reference evapotranspiration. Each of the method has merits and drawbacks.

The single layer Penman-Monteith (PM) method is widely used method for the estimation of evapotranspiration (Srivastava et al., 2018). The accuracy of evapotranspiration estimate relies upon the quality of input weather data. According the results in article (Poon and Kinoshita, 2018) the largest decrease in ET is approximately 13–57 mm per month and is most prominent during the summer (April to September). The observed decrease in ET in article (Poon and Kinoshita, 2018) contributes to understanding of changes in water yield following wildfires, which is of interest for accurately modeling and predicting hydrologic processes in semi-arid landscapes. Optimal Interpolation scheme to generate a reference evapotranspiration was used in article (Tomas-Burguera et al., 2018), forcing meteorological variables, and their respective error variance where, a sensitivity analysis of observational uncertainties and network densification suggests the existence of a trade-off between quantity and quality of observations. Precipitation and reference evapotranspiration are key parameters in hydro-meteorological studies and used for agricultural planning, irrigation system design and management. Precipitation and evaporative demand are expected to be alter under climate change and affect the sustainable development (Pandey and Khare, 2017). Regression analysis in article (Straatmann et al., 2018) showed a close relationship between ETo measured with atmometers and weather station across locations and years. The Penman-Monteith FAO-56 equation requires the complete climatic records for estimating reference evapotranspiration (ETo) Mattar, 2018. Reference evapotranspiration has a significant role in agricultural water management. Commonly, FAO56 Penman-Monteith as standard method is used to calculate ET0 (Khanmohammadi et al., 2017). The main objective of the research (Mehdizadeh et al., 2017) was to investigate the performance of empirical equations and soft computing approaches including gene expression programming (GEP), two types of support vector machine (SVM) namely SVM-polynomial (SVM-Poly) and SVM-radial basis function (SVM-RBF), as well as multivariate adaptive regression splines (MARS) in estimating monthly mean reference evapotranspiration (ETo) in Iran and it was found that the MARS and SVM-RBF methods generally performed better than GEP and SVM-Poly. Overall, the performance of the MARS and SVM-RBF was better than used empirical equations (Mehdizadeh et al., 2017). The results in article (Feng et al., 2017) showed that during the past 60 years, annual sunshine hours, relative humidity, wind speed and precipitation decreased while temperature increased. ET0 decreased at a rate of −1.5 mm per decade, or −0.4, −0.7, −0.3, and −0.1 mm per decade in spring, summer, autumn, and winter, respectively (Feng et al., 2017). The main objective of study (Čadro et al., 2017) was to validate and determine, compared to the FAO-PM method, a suitable and reliable alternative ET0 equations that are requiring less input data and have a simple calculation procedure, with a special focus on Thornthwaite and Turc as methods previously often used in BiH. Relative contributions of climatic variables to ET0 were temporally unstable and varied considerably in the nine agricultural regions and the whole China; spatial distribution for relative contribution of various climatic variables showed significant diversity among various agricultural regions (Wang et al., 2017). In study (Gao et al., 2017) the daily reference evapotranspiration (ETref) was estimated through the Penman-Monteith method at 15 meteorological station where it was found that in annual time scale, solar radiation to be the most dominant variable influencing ETref, however, in seasonal time scale, average air temperature, maximum air temperature, relative humidity were the most dominant factors in spring, summer and winter, and autumn, respectively. Accurate estimation of reference evapotranspiration is vital to hydrological and ecological processes. The FAO-56 Penman–Monteith (PM) model has the higher accuracy for ET0 estimation, but it requires many meteorological inputs that are not commonly available. Therefore an ideal method is needed requiring as minimal as possible input data variables without affecting the accuracy of estimation (Feng et al., 2017). The study (Pandey et al., 2016) planned to identify a suitable alternative to the FAO-56 Penman-Monteith (FAO56PM) equation for calculating reference evapotranspiration (ET0) from chosen temperature and radiation based models utilizing monthly meteorological data from 30 destinations. The artificial neural networks (ANN) and the empirical methods of Priestley-Taylor, Makkink, Hargreaves and mass transfer were used to estimate the reference evapotranspiration with daily meteorological data (Antonopoulos and Antonopoulos, 2017). Artificial intelligence methods have potential for reference evapotranspiration estimation since the methods could handle complex and nonlinear data.

In this study reference evapotranspiration is analyzed based on several input parameters which represent weather parameters. There is need to estimated which weather parameters has the highest influence for the reference evapotranspiration. Also there is need to analyze the influence of the combinations of the input parameters on the reference evapotranspiration. The algorithm applied in this article represent one type of artificial intelligence method. The algorithm is based on evolutionary computation or genetic computation. The algorithm in this study space solution of genetic programs hence the algorithm is known as genetic programming (GP). The main aim of this study is to investigate the accuracy of GP approaches for ET0 modeling.

Section snippets

Data used

In the investigation meteorological data were used for the evapotranspiration analysis. The data include minimum (Tmin) and maximum (Tmax) air temperatures, sunshine hours (n), actual vapour pressure (VP), minimum (RHmin) and maximum (RHmax) relative humidity and wind speed. Since the evapotranspiration and humidity have pronounced inertia additional inputs are considered as well. The inputs are time lagged evapotranspiration, and time lagged humidity. Data were obtained from the World Bank

Results

To estimate reference evapotranspiration, two statistical indicators are used for comparison between real and predicted values of reference evapotranspiration. Table 3 shows the summary results of the reference evapotranspiration prediction based on the eight analyzed combinations. Based on the results on can see the best prediction for combination-8 according to the both statistical indicators. The worst prediction result is observed for conmbination-1. One can note three group of results for

Conclusion

In this study, reference evapotranspiration is analyzed by evolutionary algorithm, namely, genetic programming technique. As the input parameters monthly minimum and maximum air temperatures, sunshine hours, actual vapour pressure, minimum and maximum relative humidity and wind speed are used. Since the evapotranspiration and humidity have pronounced inertia additional inputs are considered as well. The inputs are time lagged evapotranspiration, and time lagged humidity. Genetic programming is

Cited by (46)

  • Estimation of actual evapotranspiration: A novel hybrid method based on remote sensing and artificial intelligence

    2022, Journal of Hydrology
    Citation Excerpt :

    However, the large number of parameters and the variety of units used for these parameters are cited as limitations of this equation (Talebmorad et al., 2020; Nourani et al., 2020; Chen et al., 2020). Researchers have thus tried to simplify ETo estimation by using fewer meteorological parameters in the form of AI methods (Shiri et al., 2014a; Gocic et al., 2015; Patil and Deka, 2016; Yassin et al., 2016; Antonopoulos and Antonopoulos, 2017; Feng et al., 2017a; Jovic et al., 2018; Zhang et al., 2018; Reis et al., 2019; Saggi and Jain, 2019; Petkovic et al., 2020; Mohammadi and Mehdizadeh, 2020; Roy et al., 2020; Ahmadi et al., 2021). Feng et al. (2017b) evaluated the performance of two AI models in estimating ETo using temperature as the only input in China.

  • Transfer learning augmented enhanced memory network models for reference evapotranspiration estimation

    2022, Knowledge-Based Systems
    Citation Excerpt :

    However, these models lack at capturing non-linear fluctuations of the data. In recent decades, ANN and SVM models [36,42,45,46] have been found to very accurate at predicting future evapotranspiration values. Furthermore, the most recent studies in this prediction domain have shown that the Deep Neural Network (DNN) [23,44,47] model has achieved the best prediction accuracy as compared to other AI-based models.

  • Development of reference evapotranspiration equations using an artificial intelligence-based function discovery method under the humid climate of Northeast India

    2020, Computers and Electronics in Agriculture
    Citation Excerpt :

    It has been noted that GP and GEP can be successfully applied to simulate complex hydrological processes with high precision (Martí et al., 2015; Razaq et al., 2016; Jing et al., 2019). Similarly, several studies have examined the predictive accuracy of GP and GEP in the modelling of ET0 (Parasuraman et al., 2007; Shiri, 2017; Jovic et al., 2018; Shiri et al., 2019). In this study, symbolic regression based ET0 formulae that are equivalent to the FAO56-P-M is designed and implemented using routinely measured meteorological parameters (n, Vw, Rh, Tmax, and Tmin), and two geographical parameters (Alt, and extraterrestrial radiation, Ra).

View all citing articles on Scopus
View full text