Freshwater supplies around the world are under significant pressure as a result of increasing consumption and pollution (Steffen et al., 2015, Mekonnen and Hoekstra, 2016). The agriculture consumes the most water, of about 92 percent of global total use of water (Hoekstra and Mekonnen, 2012). In Egypt, the serious problem facing the water supply system is the limited water resources and water shortage (Mohie El Din and Moussa, 2016). As the result of climate change conditions and rapid increase in the population, agricultural water resources are decreasing in regions worldwide especially in the semi-arid and arid zones (Farg et al., 2012). Water safety as well as food security links irrigated agriculture (Gennari et al., 2019). A number of tests have been carried out to enhance water usage and crop yield efficiency in irrigated agriculture for saving water in irrigated agriculture to achieve water management sustainability (Ward and Pulido-Velazquez, 2008). Many indicators are available for evaluating the sustainability of water and food, for example water footprint, water shortages, and crop water productivity (Mekonnen and Hoekstra, 2011, Liu et al., 2009). The water footprint (WF), an indication for direct and indirect usage of water, is a metric for determining how much freshwater a product consumes during its life cycle. Its components include green, blue and gray water (Hoekstra and Mekonnen, 2012). The precipitation water absorbed by plants is called green WF rather than runoff. Water intake from rivers, reservoirs, and groundwater is represented by blue WF, while freshwater resources used to assimilate pollutants are represented by grey WF. WF in agriculture has been extensively researched using a variety of crops and areas (Hoekstra and Mekonnen, 2012).
Water footprint (WF) research focused mainly on lowering the world average use of freshwater (Lovarelli et al., 2016). Local conditions, geographical area, atmosphere, and technology are all considered by the WF (Huang et al., 2012, Zhuo et al., 2016, Tuninetti et al., 2017). Water footprint concept as volumetric water-use indicator, can be estimated by dividing the actual evapotranspiration over the crop yield (Chapagain and Hoekstra, 2008). The first parameter, potential crop evapotranspiration, is essential in water footprint calculations. Several mathematical methods are used to assess the reference evapotranspiration (ETo). However, the Food and Agriculture Organization's (FAO) FAO-56 Penman-Monteith method (Allen RG, 1998) is more effective than others such as (Tabari et al., 2013, Gavilan et al., 2007) because of the strong basic physics and reasonable connections (Landeras et al., 2008). However, it requires comprehensive of weather parameters (Hobbins, 2016, Maroufpoor et al., 2020). Although this method is the most reliable, A wide range of spatiotemporal characteristics is needed (maximum and lowest air temperatures, wind speed, solar radiation and vapor pressure deficit), frequently inappropriate in many poor nations due to a lack of meteorological stations and weather data records (Abdullah et al., 2015, Almorox et al., 2015, Dadaser-Celik et al., 2016). While the reference evapotranspiration with the Penman-Monteith equation is assessed by (Mokhtar et al., 2020a, Mokhtar et al., 2020b), is most common for water footprint calculations (Chico et al., 2013, Hoekstra et al., 2009, Manzardo et al., 2014). Although these works produced good results, it consumes much time, cost, data, and effort. In the meantime, empirical methods for evapotranspiration estimate were widely applied with fewer meteorological parameters become more in use in areas in which complete climatic variables are absent, for instance models based on temperature, mass transfer and radiation, so that's the significant restriction for the application of Penman-Monteith model over the world (Feng et al., 2017, Feng, 2018).
Furthermore, evapotranspiration is affected by a variety of meteorological variables, making it difficult to deal with dynamic and nonlinear relationships between independent and dependent variables. As a result, developing empirical models that taking into account all of these complex processes are a big challenging (Wu et al., 2019). These datasets are not easily accessible and/or questionable for most areas throughout the world and notably in poor nations (Yamaç and Todorovic, 2020). Because of the highest performance among nonlinear input-output connections in the model (Xiao et al., 2019), Machine learning methods for the description of complex hydrological processes have been used by (Wu et al., 2019, Feng et al., 2019, Mokhtar et al., 2021b, Yaseen et al., 2018), including ETo estimation (Wang et al., 2017, Kisi and Sanikhani, 2015, Jovic et al., 2018), Machine learning established as an artificial intelligence (AI) discipline, involves algorithms that capture relevant information from vast data and utilize it for self-learning purposes to make accurate calculations or predictions (Saha and Manickavasagan, 2021). During the last decades, in the domain of water sciences and technologies the use of various machine learning technology has considerable relevance like artificial neural networks (ANN) (Landeras et al., 2008, Antonopoulos and Antonopoulos, 2017), support vector machines (SVM) (Shiri et al., 2014), fuzzy logic models, neuro-fuzzy models, support vector machines, random forest (Elbeltagi et al.), and k-Nearest Neighbor (k-NN) (Heddam, 2014, Rehman et al., 2019). Machine learning approaches were widely used to predict reference evaporation, actual of the water and to predict water resources variables, hydrological cycles, management of water resources, water quality prevision and storage activities (Elbeltagi et al., 2021b, Elbeltagi et al., 2021c, Mokhtar et al., 2021a). Goyal, 2014 employed ANN and other machine learning approaches such as least squares support vector regression (LS-SVR) and fussy logic to estimate regular evaporation in subtropical climates (Goyal et al., 2014). Laaboudi, 2012 examined the effectiveness of the use of artificial neural networks to evaluate ETo using incomplete meteorological data (Laaboudi et al., 2012). Algorithms for machine learning were utilized successfully to provide solutions to the problems related to potato cultivation in farmland, such as predict the potential of leaf water as suggested by (Zakaluk and Sri Ranjan, 2006), root development modeling (Delgoda et al., 2016), tuber growth (Fortin et al., 2010), the ETo estimation (Tabari et al., 2013, Sabziparvar and Tabari, 2010, Yamaç and Todorovic, 2020).
Therefore, the objective of this research is to 1) develop and compare between four machine learning (SVR, RF, XGB and ANN) over three potato’s governments and 2) select the best model in the best scenario, achieves high precision and low error in forecasting potato blue WF. This study can therefore provide an innovative modeling method that will enhance efforts to tackle the WFP forecast, which in turn will help mitigation strategies like water usage policies and food safety development plans.