Use of a triple-sensor fusion system for on-the-go measurement of soil compaction

A general approach to mapping the within-field spatial variation in soil compaction is to measure a soil strength index such as penetrometer resistance (PR) using an on-the-go mechanical sensor. In general, these types of measurements depend on a number of soil properties, such as dry bulk density (rho(d)), organic matter (OM) content, water content (theta) and clay content (CC), and thus only provide a composite soil parameter that is somewhat inadequate for characterising the soil compaction status. In this study, a novel sensor fusion is proposed that combines a single-probe horizontal penetrometer, a dielectric-type soil water content sensor and a gamma-ray sensor (the Mole) for simultaneous measurement of PR, volumetric water content (theta(v)) and CC, respectively. In order to examine the applicability of the system for on-the-go measurement of soil compaction in terms of rho(d), experiments were conducted in two fields with CC of 148-313 g kg(-1). At 48 selected locations along transects where on-the-go measurements with the triple-sensor system were performed, core samples for determination of CC, OM, theta(v) and rho(d) were taken within the working depth of the horizontal penetrometer. A four-parameter statistical model was developed for rho(d) as a function of PR, theta(v) and CC based on the soil core (for rho(d), theta(v) and CC) and sensor measurements (only for PR) with R-2 = 0.71 and RMSE = 0.06 Mg m(-3). The model was employed to predict the spatial variations in rho(d) over one of the fields (for which a yield map was available) using the continuous sensor system data. Interpolated maps of crop yield, PR, theta(v), CC and rho(d) showed some similar local patterns at field scale. The sensor fusion system developed here can be a useful instrument for future studies on soil compaction, especially within the context of precision agriculture. Further evaluation of the sensor system over a wider range of soil texture and water content could provide an extensive data base for developing a general pedotransfer function for rho(d) as a function of sensor-based readily-measurable soil attributes. (c) 2012 Elsevier B.V. All rights reserved.