Calibration of an on-line sensor for measurement of topsoil bulk density in all soil textures

Abstract

An on-line sensor for the prediction of bulk density (BD) has been previously developed, and calibrated for light textured soils. This study aims to expand the implementation of this sensor for all soil types, including clay soils. The on-line sensor consists of a single ended shear beam load cell to measure draught (D), a visible and near infrared (vis–NIR) sensor to measure moisture content (MC), and a wheel gauge equipped with a draw wire linear sensor to measure depth (d). This sensor was used to collect data from 14 fields across four European countries (Denmark, Czech Republic, the Netherlands and the UK) with different soil textures. The three measured parameters were substituted into a previously developed model to calculate BD. The calculated BD for 333 samples was compared with BD measured with a core sampling kit to validate the system and to calculate correction factors (CF) valid for each additional soil texture. A multiple linear regression (MLR), nonlinear multiple regression (NMR) and artificial neural network (ANN) analyses were carried out to establish models to predict CF, as a function of average field MC and soil texture fractions.For clay soils, the correlation between the measured and on-line predicted BD was satisfactory (R2=0.54–0.69), and suggested a large negative CF>−0.248Mgm−3. For other soil types with lighter soil textures, accuracy of measurements varied (R2=0.51–0.96), with lower CF values than those calculated for clay soils and ranged between negative and positive values (−0.007Mgm−3<CF<0.079Mgm−3). The ANN analysis resulted in the best model to predict CF as function of MC and soil texture fractions (R2=0.96), which allowed the utilisation of the on-line measurement system for any field having any texture and average MC. The on-line BD sensor was shown to be capable of predicting field dry BD rapidly for a large number of samples, as compared with traditional methods.