1. Introduction

Water resources are highly influenced by the hydrologic cycle and play a role in the agriculture economic development. However, as it is shown by the intergovernmental panel on climate change report [1], the phenomenon of changing climate is set way to exacerbate an already

serious situation of water supply for various users. Agricultural production will be one of the sectors most vulnerable to climate change and variability. The water budget must now be shared with agriculture, urban use, industry, recreation and livestock watering; the future will be seeing an increasing competition for water. Spatial and temporal changes in precipitation and temperature patterns will have an impact on the viability of dry land farming and therefore necessitate irrigation where rainfall was previously adequate. Efficiency improvement in irrigation lies among the key strategies for saving more water and promote a sustainable intensification of agriculture when water scarcity becomes a major constraint to production [2]. Nevertheless, irrigation water for crops is globally the major consumptive use of water resources. Due to the above-mentioned challenges, it is important to improve the management of agricultural water, which would involve the accurate estimation of consumptive uses. One of the techniques is the measurement of evapotranspiration which is a major component of the hydrologic cycle. Evapotranspiration (ET) is an essential component of the water balance, and it is a significant consumptive witness of precipitation and water applied for irrigation of cropland [3]. ET can help for highly efficient management of water uses in agriculture and set up real water-saving systems.

Basically, ET includes two processes: One is evaporation and the second is transpiration. The latter is the process of removing water from vegetation or any other moisture containing living surface. Evapotranspiration includes two processes. During the plant growth, the water stored in the soil is taped and transferred in the atmosphere. Transpiration is the evaporation of water in the vascular system of plants through the leaf stomata when they open and close controlled by their guard cells. Based on this bio-physical process, transpiration involves a living organism and its tissues. ET is then the process, whereby water originating from a wide range of sources is transferred from the soil compartment and/or vegetation layer to the atmosphere. ET is the largest outgoing water flux from the Earth's surface and accurately quantifying ET is critical for the development of crop cultures in an increasing drier environment, and it can contribute to a greater understanding of a range of agricultural ecosystem processes. ET is particularly fundamental when dealing with water resource management issues such as irrigation water or water reserve management [4]. ET cannot be directly measured but it has to be estimated by monitoring the exchange of energy/water above the vegetated surface (remote sensing) or as a residual term of the hydrological balance.

Several methods are currently used to measure and estimate ET: One of them is the lysimeter method or soil water budget. That method may be accurate but lysimeters are expensive, and the extent of their measurement is localized (i.e., they provide data for a very small area compared to the field surface, so it can only be used in field locations). Another one uses micrometeorological data to compute ET. A widely used approach by these data is the FAO-24 and by extension the FAO-56 procedure, based on ET0 and Kc [5]. ET acquisition can be obtained with different instruments at the scale of: the leaf (porometer), an individual plant (i.e., sapflow, lysimeter), the field scale (i.e., field water balance, Bowen ratio, scintillometer) and the landscape scale (i.e., eddy correlation and catchment water balance) [6]. The flux measurement of micro-meteorological station can only represent the value in a point or a limited area (several meters to several hundred meters). However, a scintillometer can measure averaged sensible heat fluxes in a distance of 500m to 10 km, which is an average of time and space. The measurement scale of a scintillometer is matching to the grid scale of atmospheric model and the pixel scale of remote sensing, as a result. This advantage promotes the development of scintillometers in recent years [7].

The objective of this work is to compare two methods used for estimating ET rate: scintillometry and meteorological measurements with the FAO-PM56 model based on ET0 and Kc with the reference evapotranspiration for the crop (ET0) and the specific crop coefficient for the cultivation type at its stage development (Kc). Also, it is compared how the final result of ET, calculated with the two different measurements, can be more sensible to different environmental parameters. The sensitivity of the two methods is calculated and the influence of the main environmental parameters on the accurate values of ET.

In order to anticipate the necessary action for water management, a model to forecast ET based on an artificial neural network (ANN) is developed. In recent years, ANN models have become extremely popular for prediction and forecasting in a number [8, 9] of domains, including finance, power generation, medicine, water resources and environmental science [8, 9]. The evapotranspiration process calculated with the FAO-PM56 and scintillometry data is a nonlinear process. ANN models are quite appropriate for the simulation of ET leading to good results.

The predicted output final results are compared for two different input parameters of ET calculated with scintillometry and micro-meteorology data. The final goal of this study is to find which input data can be reliable to obtain ET forecast performances. Moreover, the optimal number of predicted days to obtain a correct final performance and the optimal number of input days of data to obtain a correct prediction are tested.
