Optimization of nitrogen concentration of plant tissue for increased quantity and quality of tobacco leaf using an artificial neural network

Nitrogen (N) affects adversely the tobacco yield quantity and quality as it increases yield, Chlorine and nicotine contents, but decrease potassium content. This experiment was aimed at optimization of (the balance between) N concentration in leaf, stem and root to increase both yield quantity and quality (high potassium, low Chlorine and medium nicotine contents) using artificial neural network. Two field experiments based on complete block design with three replications were conducted in Tirtash and Urmia tobacco research centers. Treatments were factorial arrangement of two N sources (urea and nitrate ammonium) and four application patterns (basal, 2/3 basal and 1/3 after initiation of rapid growth (AIRG), 1/2 basal and 1/2 at AIRG, 1/3 basal and 2/3 at AIRG). The N concentration of leaf, stem and root (model inputs) was measured in 30, 50, 70, 85 and 100 days after transplanting. After harvesting, the quantity of cured leaf and its Cl, K and nicotine content (model outputs) were also determined. The results indicated that a model with one hidden layer and configuration of 15-15-4 is appropriate and there were no significant different between two N sources. The best pattern was use of nitrate ammonium in 2/3 basal and urea 1/3 basal. The average value of optimized N concentration was 3.06, 2.42 and 1.5 percent for leaf, stem and root, respectively. These optimized concentrations can lead to potential increase in quality and quantity of tobacco which should be taken into consideration by breeders and agronomists.