Development of a Multi-Sensor Platform for Proximal Soil Sensing.
James A. Taylor1, Alex. McBratney2, Raphael Viscarra Rossel2, Budiman Minansy3, Henry Taylor3, Brett Whelan2, and Michael Short3. (1) Australian Centre for Precision Agriculture, Univ of Sydney, McMillan Bld, A05, Sydney, Australia, (2) Australian Centre for Precision Agriculture, Univ of Sydney, McMillan Bld, A05, Sydney, Australia, (3) Australian Centre for Precision Agriculture, McMillan Bld, A05, The University of Sydney, Australia
The ability to map soil and soil properties using information collected from on-the-go proximal-mounted sensors has begun to be widely exploited in the past decade particularly in fine-scale digital aoil mapping and precision agriculture. Many different types of proximal sensors are available that utilize a range of techniques to measure soil and/or vegetative response. These techniques include, but are not inclusive of, electro-magnetic induction, magnetics, ã-radiometrics, near infrared spectrometry, ion-selective field effect transistors, ground-penetrating radar and multi- or hyper-spectral imagery. The sensors may be invasive, semi-invasive, or non-invasive but all are coupled to global navigation satellite systems (GNSS), such as the NAVSTAR Global Positioning System (GPS), to provide spatial coordinates for measurements. High-resolution GNSS receivers are capable of providing high accuracy digital terrain models and secondary and tertiary landform attributes. In general these proximal sensors have been run either individually (with a GNSS receiver) or in pairs. The different physical principals underlying the different sensors often produce complementary information however research into an integrated approach to using multiple sensors is lacking. Simultaneous analysis of data from coupled sensors should improve the accuracy of site-specific predications of soil properties and permit a greater range of soil properties, both chemical and physical, to be predicted. While this approach will still require some soil sampling for sensor calibration, M-SPs should reduce (or at least maintain) the current cost of soil analysis while greatly improving the density of information gather. This potential has prompted some recent commercial development in this area, for example, the Mobile Sensor Platform from Veris Technologies and Soil Information System from EarthIT. Researchers within the Australian Centre for Precision Agriculture and the soil science group at The University of Sydney are developing a multi-sensor platform (M-SP) to perform field and sub-catchment scale soil surveys. Currently the M-SP consists of two electro-magnetic induction instruments (Geonics Em38 and EM31), Veris 3100 electrical conductivity cart, the GPR 320 Gamma Radiometer and an Omnistar HP dual frequency GPS. Data is collected from the sensors at 1Hz frequency. The platform is driven at ~10-15 kmh-1 on 20-25m swaths. All the sensor data is logged into a ruggedised laptop.This paper will present experiences and results to date on the development of the M-SP and how soil property predictions from the M-SP compare to those using data from individual sensors. It will also discuss further aspects of M-SP development including incorporating a real-time pH/lime sensor from Computronics Holdings Ltd, the potential for incorporating a penetrometer and soil moisture senor and investigating improve methods for data fusion and soil property prediction.