Premium
Evaluation and analysis of different regression models for estimation of ECe from EC 1:5 —With a case study from Buin‐Zahra, Iran *
Author(s) -
Hassannia Mehrdad,
Nazari Bijan,
Kaviani Abbas,
Sotoodehnia Abbas
Publication year - 2020
Publication title -
irrigation and drainage
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.421
H-Index - 38
eISSN - 1531-0361
pISSN - 1531-0353
DOI - 10.1002/ird.2488
Subject(s) - linear regression , mean absolute percentage error , salinity , mathematics , regression analysis , soil science , exponential function , statistics , soil salinity , mean squared error , environmental science , soil water , geology , mathematical analysis , oceanography
Abstract Soil salinity is an important parameter in irrigation, drainage, and environmental studies. Determining the electrical conductivity of soil saturated paste extract (ECe) is a well‐known method, but its use is limited because its production process is time‐consuming and difficult. The 1:5 solution electrical conductivity (EC 1:5 ) is an alternative simplified method. The aim of this study was to evaluate ECe‐from‐EC 1:5 conversion models. A total of 123 samples from 3 soil layers were analysed. The research was planned in two phases: 1) model evaluation in soil layers, 2) model evaluation in ECe categorized data. Results show that the linear model was the best model in the second layer ( R 2 = .66, mean absolute percentage error [MAPE] = 0.058), and the exponential model was the best in the other layers ( R 2 = .65 to 0.67, MAPE = 0.085 to 0.060). Also, exponential and linear regressions were the best ECe estimation models in the ECe < 4 and ECe > 4 categories, respectively. In addition, the conversion factor ( f ) for converting ECe and EC 1:5 was obtained as 3.7–13.9, 2.8–5.0, and 3.4–5.8 for three soil layers, respectively. Results show that the use of general recommendation tables for f will lead to considerable error, and the factor must be determined specifically for each region.