Modeling soil enzyme activity using easily measured variables: Heuristic alternatives
Mitra Ebrahimia, Mohammad Reza Sarikhania, Jalal Shirib,c, Farzin Shahbazia
Department of Soil Science, Faculty of Agriculture, University of Tabriz, Iran.
In the present research, gene expression programming (GEP) and artificial neural network (ANN) techniques were used to estimate soil enzyme activity (SEA), including urease, alkaline phosphatase and dehydrogenase. Data from 65 soil samples located in Mirabad region, Suldoz plain (West Azerbaijan, Iran) were used to test the adopted methodology. The soil samples were selected from areas with different land usages (apple orchard, crop production, and rich pasture). Various combinations of the input parameters including soil texture, pH, organic carbon (OC), electrical conductivity (EC), microbial biomass carbon (MBC), and microbial soil respiration (SIR) were utilized to feed the applied models. The root mean square error (RMSE) and the coefficient of determination (R2) were employed for assessing the models' performance accuracy. The highest R2 and lowest RMSE were obtained for the models that used all available input parameters. The results showed that among targets (urease activity (UA), alkaline phosphatase (ALP) and/or dehydrogenase activity (DHA)), the highest performance accuracy was obtained for urease activity models. The obtained results revealed that the most effective parameters in estimating urease activity were soil texture, pH, EC and OC; where about 69% and 68% of its variability was predictable by the ANN and GEP, respectively.
Keywords: Artificial neural network, Dehydrogenase, Gene expression programming, Phosphatase, Urease.