Abstract
The knowledge on the relationships of protein and micronutrient concentration in wheat grain with edaphic characteristics could provide valuable information for site-specific fertilization of crops for producing grains denser in micronutrients such as iron (Fe) and zinc (Zn) in rain-fed agriculture. In this study, we used soil properties and topographic parameters in the artificial neural network (ANN) methodology as a power tool for improving models for predicting wheat grain micronutrient and protein contents in the hilly regions of western Iran. Soil and grain samples were collected from 1 m2 plots using the stratified random method, whereas the slope positions were considered as the basis of soil sampling, at 100 selected points. The mean grain Zn, Fe, Cu (copper) and Mn (manganese) concentrations were 37.02, 65.86, 14.79 and 44.93 mg–1 kg–1, respectively, and mean grain protein was 13.76%. Application of the ANN models for predicting Zn, Fe, Cu, Mn and protein contents in grains improved prediction by 96.77%, 95.45%, 124.13%, 125% and 109.75%, respectively, over the multiple linear regression (MLR) models.The topographic parameters wetness index, plan curvature and shaded relief, selected soil properties total nitrogen (TN), soil organic matter, available phosphorus and DTPA-extractable micronutrients were identified as the most important parameters for explaining the variability in wheat grain quality at the study area.