Prediction of FC and PWP Using Neural Network and Statistical Regression in Bardsir-Kerman Area

Field capacity and permasent wilting point are the most important parameters in designing and programming irrigation, whose measurements are troublesome and time-consuming. But these parameters could be estimated by easy data characteristics such as soil texture, organic matter and gypsum, using Pedotransfer Functions (PTFs) with high precision. In order to estimate soil moisture at FC and PWP by easy data characteristic, using neural network (ROSETA) and regression models, 20 soil samples with 6 replications were collected from around Bardsir area in Kerman province and the charactersifics including, bulk density, clay, sand, silt, FC, PWP,T.N.V and organic matter were determined for each sample. The results showed that progress in neural network from a low level up to higher level needs new inputs (charactersifics), but without any considerable increase in the precision of prediction. Also, regression analysis for estimation of linear models to predict FC and PWP showed that PWP has a significant positive correlation with clay, and FC significantly correlated with sand, silt and clay. Therefore, two prediction models were constructed for FC and PWP with (R2= 69.2) and (R2= 76.6), respectively.