نوع مقاله : مقاله پژوهشی
نویسندگان
1 دانشجوی دکترای دانشگاه تبریز
2 استاد دانشگاه تبریز
3 استادیار دانشگاه بوعلیسینا، همدان
4 دانشیار دانشگاه تبریز
چکیده
عنوان مقاله [English]
نویسندگان [English]
The range of the soil volumetric water content at which plant growth is least limited in relation to water potential, aeration, and mechanical resistance is the least limiting water range (LLWR).Experimentally, measurement of LLWR is expensive and time consuming. Using pedotransfer functions (PTFs) can facilitate its prediction. There are, however, contradictory information about the accuracy and reliability of the developed PTFs for soil hydraulic properties using various methods including artificial neural networks (ANNs), multi-objective group method of data handling (MGMDH) and multivariate linear regression (MLR). Evaluating the performance of these methods in direct prediction of LLWR was the main purpose of the present study. To this end, 188 undisturbed soil samples were used to determine water retention and soil resistance curves and finally four moisture coefficients (θpwp,θ fc,θ sr,θ afp) and disturbed samples for measurement of eleven various soil physical and chemical attributes. After calculation of LLWR from upper and lower limits (LLWRe), another time it was directly predicted from soil attributes (LLWRd) using the three mentioned methods. Accuracy and reliability of the developed PTFs was evaluated using root mean square error (RMSE), Akaike information criterion (AIC), and relative improvement (R.i). ANNs appeared as the most accurate and reliable one for LLWRd prediction (lower RMSE and more negative AIC); MGMDH and MLR ranked in descending order. Significant differences between accuracy and reliability of the developed PTFs by the three methods was evaluated using AIC. Differences between developed PTFs by ANNs and MGMDH versus MLR method were statistically significant, but differences between ANNs and MGMDH were only significant for the training step. Among the three methods studied, ANNs had the highest performance in LLWRd prediction.
)