Determination of the best model for predicting soil available phosphorus based on soil organic carbon

Abstract

There are several methods for determining the determination of soil available phosphorous. These methods require many chemicals and different devices, and determining soil available phosphorous in the laboratory is difficult, time-consuming, and costly. Consequently, pedotransfer functions more suitable and economical were used to predict the soil available phosphorous(AP) from soil organic carbon (OC). To predict soil available phosphorous (AP) from soil organic carbon (OC), five models were utilized, including exponential, linear, logarithmic, polynomial, and power models. Soil AP was predicted as a function of soil OC. The soil AP determined by laboratory tests was compared with the predicted soil AP based on the AP-soil OC model using the Bland-Altman method. The 95% limits of agreement for comparison of soil AP were determined with laboratory tests, and the soil AP pedotransfer function was computed at -1.707 and 1.542 mg kg -1. The mean soil AP difference between the two methods was - 0.083 mg kg -1. The polynomial model (AP = 38.173 - 4.462 x OC+ 0.146 x OC2) is given the best fit to predict soil available phosphorus, due to its high R2 (0.868) and low RMSE (0.846).

Keywords

• Polynomial model, soil, available phosphorus, pedotransfer function, Bland- Altman method