Autonomous Intelligent Irrigation Systems in Hop Plantations (Republic of Chuvashia, Russia)

The possibility of implementing intelligent irrigation has a number of undeniable advantages, mainly including the fact that the time can be determined and the volume of irrigation water can be adapted to specific plant types on a specific soil. A neural network has been trained to describe the dynamics of soil moisture based on the basic soil water retention curve (SWRC). It is able to take into account a wide range of input data, such as the specific surface area of the solid phase of soils, porosity, humidity, etc., for a given initial soil moisture profile. Preference is given to a recurrent neural network, since this type works well with sequential data and is able to take into account time dependence and solve the problem of decaying gradients of soil hydrophysical properties. The neural network processes the vector of incoming signs—humidity, temperature, volume of incoming/outgoing water, etc.—and connects them with the dynamics of humidity from sensors located at different depths. When modeling mass–salt transfer with different boundary and initial conditions, the dependence of moisture retention on the moisture conductivity function is used, which allows us to calculate how moisture with dissolved nutrients moves through the soil under the influence of pressure and concentration gradients. Since the SWRC is constructed as a function of directly measured data, it is easy to set it for each point of interest in the field and at each depth. During modeling, the soil is divided into elementary volumes (from 2–3 mm to 1 cm), and an array with data sets is compiled at each point. The research was conducted in a real hop plantation (the village of Opytny, Tsivilsky district, Republic of Chuvashia). The values of the soil moisture sensors at different depths, together with the data from the portable weather station, are sent to the input of the neural network. According to the minimum allowable humidity for hops, the model predicts situations when humidity reaches critical values and initiates watering. Thus, the implemented approach makes it possible to automate irrigation management, increase water use efficiency and ensure optimal conditions for plants.