ASSESSING PLANT AVAILABLE POTASSIUM OF ILLITIC LOESS SOILS POSSESSING HIGH SPECIFIC SURFACE AREA AND WEAK AGGREGATION
|S. Amin Shafiei*, S. Alireza Movahedi Naeini
Dep. of Soil Sciences, Gorgan University of Agricultural Sciences and Natural Resources, Pegah St., Basij Sq., Gorgan City, Golestan province,Iran
|Corresponding Author: Seyed Amin Shafiei; E-mail: [email protected] Tell: +989359794580, +982122308040|
|Received: 30 March 2014 Accepted: 04 May 2014|
|Related article at Pubmed, Scholar Google|
Although Quantities of potassium extractable by NH4OAc in some illitic Loess slowly Swelling Soils possessing high specific surface area is high, potassium has been identified as the limiting plant growth factor in this areas. Due to potassium deficiency Golestan provincesoils, taking advantages of theNH4OAc method was not feasible and what has been proposed was sodium tetra phenyl boron (KNabph4) method with 1minute extraction. The aim of this study was soil potassium assessment (KNabph4), specifying whether status of aggregation (SOA) and soil specific surface area (SSA) can effect onKNabph4, and feasibility studies for prediction of the KNabph4. 183 soil samples from 0- 30 centimeter depth with varied physicochemical properties were obtained. Linear regression stepwise (LRS) models were established between the measured 17 parameters. Figures were tested via genetic algorithm (GA) solver and artificial neural network (ANN). MLP, RBF and ELMAN were used in ANN. By investigation of the interactions, it was specified that SOA (0.52**), surface charge excess of potassium (0.69**) and SSA (0.72**) were highly correlated with KNabph4. Therefore, with ascending of the SSA amount, further surfaces of soil particles will be weathered and this process certainly ascends the K release. If the weak aggregation does not reduce the ascending trend of K release, the quantities of potassium availability will be increased. About predicting soil potassium, results indicated the produced models via ANN were much accurate than the LRS and GA. The best model was obtained in MLP network by selecting SOA and SSA as the inputs parameters and KNabph4 as the output of model while the RMSE= 59.36, MEF= 84.06, AEP= 15.46, RTest= 0.99, R2= 0.84 was.