Estimation of the values of soil absorption ratio using integrated geostatistical and artificial neural network methods

ABSTRACT : In recent Years several Methods based on mathematics and computer science have replaced experimental and laboratory methods. Because of increase in computational accuracy and decrease in laboratory costs, these methods have been gradually developed in most of the engineering sciences. From long ago the application of artificial neural networks in agricultural sciences has been accompanied by an effective role nowadays, with a population rise and increasing need of man to nutrition, the attention of experts has been focused on the sustainable agricultural sector. In sustainable agriculture, the discussion of soil fertility plays the most important role so any management practices on soil fertility grounds poses the greatest impacts on the agricultural cycle accurate studies of the physical, chemical and biological soil characteristics of each region before agricultural practices, will be of great importance in the more appropriate productivity of the land. in geological studies on physical, chemical and biological properties of soil, the soil profiles are prepared in the form of point sampling. In order to analysis and study these characteristics, all of the point samples should extended to the surface and for this purpose geostatistical methods are used. In this research by preparing zoning maps of those parameters that their calculations need lesser time and cost, we want to determine the values of those parameters that their calculations need experiments which are costly and time consuming the physical and chemical properties used in this research include the clay, sand, silt percentages, saturation percentage of soil(SP), acidity(PH), salinity(EC), bulk density(Bd), and sodium absorption ratio(SAR). The results indicate high level of accuracy of the incorporated Geostatistical model and the Artificial Neural Networks in the matter of prediction so that it has predicted the SAR values with correlation coefficient of R=0.94 and MSE=7.51423. As a result this method will be an assured replacement for costly experiments to estimate the SAR Values.