Toprak Profil Duvarında Farklı Regresyon Ağacı Algoritmaları Kullanılarak Sayısal Renk Parametreleri ile Organik Maddenin Tahmin Edilmesi

Abstract

Soil organic matter has a critical role for the physical, chemical and biological properties of the soil and for sustainable soil and agriculture. Quick and cost-effective prediction of soil organic matter can provide basic data support for precision agriculture. The study area is located in the Muttalip pasture of Tepebaşı, Eskişehir. The soil profile wall (1x1 m) was dug and divided into 10x10 cm raster cell. A total of 100 soil samples were taken from center of each raster cell of the soil profile wall. The field-based and lab-based digital color parameters (CIE Lab) were measured depending on the grid sampling model. The ordinary Kriging interpolation method was used in geostatistical distribution maps of the amount of organic matter (OM) and field-based and lab-based CIE Lab values. CHAID, Ex-CHAID, and CART regression tree algorithms were used to predict the OM with field-based and lab-based CIE Lab values. The OM in the soil profile wall varies between 4.65-10.54% in the topsoils, while it varies between 0.01-0.41% in the subsoils. According to the results, lab-based CIE Lab values obtained high predicting performance and more effective than field-based CIE Lab values. It concluded that the CART algorithm can be used rapidly and economically in prediction OM with high prediction performance (R2=0.89) with lab-based digital color parameters.