نوع مقاله : مقاله پژوهشی
نویسندگان
1 فیزیک خاک، گروه علوم و مهندسی خاک، دانشکده کشاورزی و منابع طبیعی، دانشگاه محقق اردبیلی، اردبیل، ایران
2 گروه مهندسی آب، دانشکده کشاورزی و منابع طبیعی، دانشگاه محقق اردبیلی، اردبیل، ایران
چکیده
کلیدواژهها
موضوعات
عنوان مقاله [English]
نویسندگان [English]
The saturated hydraulic conductivity (Ks) of soil plays a key role in the transport of water, nutrients, and pollutants in soil. Direct measurement of Ks in soil is time-consuming and expensive, and due to soil heterogeneity and error, the results obtained may not be very reliable. Therefore, indirect methods such as regression pedotransfer functions and intelligent models have been used to estimate this soil hardness variable. The aim of this study was to compare the performance of three intelligent models including neurofuzzy (NF), gene expression programming (GEP), and random forest (RF) in estimating Ks from readily available soil variables. For this purpose, information on 102 soil samples selected from the agricultural lands of the Ardabil plain, including sand, silt, clay, geometric mean and geometric standard deviation of particle diameter, bulk and particle density, organic carbon, CaCO3, and Ks, was used. Eighty two data were used for training and 20 data were used for testing the intelligent models. Five different combinations of readily available soil variables were selected as model inputs to estimate Ks by NF, GEP, and RF models. The results of all three intelligent methods used in the study indicated that the combination of two input variables including clay and soil bulk density, had the highest accuracy and was the best model in estimating Ks. The values of the coefficient of determination (R2), normalized root mean square error (NRMSE), mean error (ME), and Nash-Sutcliffe coefficient (NS) were calculated based on the test data as 0.83, 0.118, 0.002 cm min-1, 0.81, and 0.83, 0.145, 0.020 cm min-1, 0.71, and 0.74, 0.151, 0.020 cm min-1, 0.69 for the best NF, GEP, and RF models in estimating Ks, respectively. The results showed that the NF model, compared to the other two intelligent models, was able to estimate Ks with high accuracy in the study area due to its low NRMSE and high NS.
کلیدواژهها [English]