Application of artificial neural network models in estimating nectarine crop yield under two-sided furrow irrigation

Due to the lack of rainfall, Iran is one of the arid countries in the world where most irrigation systems are done as surface irrigation. Due to the high costs of pressurized irrigation systems, improvement and modification of surface irrigation methods such as land leveling, the correct choice of irrigation method, proper design and thus increase efficiency is significant. If surface irrigation is properly designed and implemented, it is one of the most suitable methods for farmers due to the lack of complex equipment and devices. Researchers use artificial neural networks to simulate and estimate parameters such as weekly evaporation rate, daily evaporation, water capacity, and permeability coefficient have been used. 
Perceptrons are arranged in layers, with the first layer taking in inputs and the last layer producing outputs. The middle layers have no connection with the external world and hence are called hidden layers. Each perceptron in one layer is connected to every perceptron on the next layer. Hence information is constantly "fed forward" from one layer to the next. There is no connection among perceptrons in the same layer. 
Radial basis function (RBF) networks have three layers: an input layer, a hidden layer with a non-linear RBF activation function, and a linear output layer. The input can be modeled as a vector of real numbers. The output of the network is then a scalar function of the input vector, and is given by where is the number of neurons in the hidden layer, is the center vector for neuron, and is the weight of neuron functions in the linear output neuron. Functions that depend only on the distance from a center vector are radially symmetric about that vector. 
The best results were calculated using the average savings in the treatment section compared to the observed section, 31.7%. It also shows water consumption in the treatment section and the control is calculated as 5793 and 6566.9 m3/ha, respectively, which indicates an 11.8% decrease in water consumption reduction (733.9 m3/ ha) of the treatment compared to the control. According to the obtained results and after comparing the results of RBF and GFF networks, RBF networks (function with radial base) with parameters of different irrigation levels as input were recognized as the best network. The R2 is equal to 0.92 and the square root of the RMS is equal to 0.035.
It can be stated that the method of two-sided furrow irrigation, in addition to reducing water consumption, increased crop yield. Also, there was the highest water loss in the first irrigation. The average efficiency of water application efficiency in the treatment and control sections was calculated to be 2.24 and 1.52 kg/m3, respectively, with the majority of losses being deep penetration. The RBF model had better results in predicting than the GFF neural network model. RBF neural networks with the parameter of different irrigation levels as input were recognized as the best network.