A frequency response-based dielectric sensor for real-time sensing and long-term monitoring of soil water content, nitrogen, and salinity

Abstract

Practical sensors capable of in-field, simultaneous, real-time sensing and long-term monitoring of several soil properties that are important to production agriculture and environment, including water content, fertility, and salinity, are not available. Worldwide research has concentrated on optical (spectral), electrochemical, and dielectric types of sensors that measure only one or two properties. In the Instrumentation and Control Laboratory of the Department of Biological and Agricultural Engineering at Kansas State University, a frequency response-based dielectric sensor (“FR sensor”) was developed in 2004 and it has been used to measure multiple properties of various types of dielectric materials, including soil, water, air, and fuel. The early work showed that the sensor is multifunctional, inexpensive, and rugged for field applications. However, the potential of the sensor for use in the field for simultaneous measurements of soil water and nutrients is unknown. This dissertation reports work since 2017, when a new sensor probe was designed specifically for “buried-in-soil” applications. Tests showed that the sensor could conduct simultaneous measurements of soil water content, density, and salinity. When salinity was replaced with a nitrate-N fertilizer, an ammonium-N fertilizer, or an organic N fertilizer, the sensor reported good results. The sensor, in general, had difficulties in separately measuring two types of the nitrogen fertilizer when both existed in soil. To simulate the in-field applications, corn was planted in two large pots. Two sensor probes were placed in each of them, and the pots were placed outdoors for more than three months. Dielectric spectral data from the sensors were measured every other day, and, at the same time, soil samples were taken for analysis of water content, nitrate-N, and ammonium-N using standard instrument and at the KSU Soil Testing Laboratory to provide reliable references. Computer programs were written to analyze the data and to develop prediction models using the partial least squares regression method. Prediction results were compared with the reliable references. For water content and nitrate-N, good results were obtained using data from single sensors as well as data from across multiple sensors. For ammonium-N, good results were obtained using single-sensor data, but the results from across-sensor data were not satisfactory. From the partial least squares regression analyses, a limited number of frequencies - the “signature frequencies” – were identified for each soil properties. With the reduced number of frequencies, the cost and measurement time can be reduced, while the measurement accuracy can be improved. In the outdoor experiment, one fifth of the data were lost due to broken wires. Thus, improvements in the design of the probe, control box, and wire connections are needed to insure longevity and reliability of the sensor.