Irrigation management in substrate culture with root zone sensors and estimation of crop evapotranspiration with neural networks

This research was conducted in 4 growing seasons consists of 2 spring and 2 fall seasons between 2011-2013. Tomato (cv. Duru) plants were studied in closed system substrate culture under 2 irrigation regimes (Full: %20, Deficit: %50) and 2 growing media (perlite and cocopeat). Irrigation thresholds were determined by the allowed percentage of depletion from readily available water and automated with root zone sensors. Supplied and circulated amounts of daily irrigation water, crop evapotranspiration, yield, etc. were measured in all growing seasons. It was concluded that, root zone sensors can be used in substrate culture and with optimum thresholds, deficit irrigation can minimize enviromental impact, irrigation (water and system) and nutrient costs with minimum yield loss. During all growing seasons, daily climate parameters, crop evapotranspirations (ETc) and crop coefficients (Kc) were also determined and standart equation's (FAO-PM and Hargreaves) prediction rate of tomato's ETc grown in substrate culture was evaluated. Generalized regression models and artificial neural network (ANN) models were also generated and compared with empirical equations. ANNs performed better in all input cases, especially, ANN model without crop coefficient and solar radiation inputs performed better than FAO-PM equation with all necessary inputs. ANN models were also generated to forecast daily crop evapotranspiration with the day before's data and performed very well. It was concluded that, ANN models can be used in site specific ETc predictions better than standart equations and can be implemented into irrigation automation with their forecasting ability.