Network of RF Ground Sensors for Applications in Precision Agriculture

Abstract

The viability of a ground sensor based remote sensing network for applications in precision agriculture is investigated in this study. While systems based on aerial and orbital platforms have been developed to monitor crop physiology and to improve crop yield, such systems are not flexible enough for application-specific or crop-specific uses. In addition, for observation of an agricultural field with space-borne radars, even if the spatial resolution is adequate, the temporal resolution is often limited by the overpass frequency of the satellite and therefore may be too coarse to capture short term phenology as well as rapidly changing, localized meteorological conditions crucial to proper crop growth. For monitoring small-scale land features such as agricultural plots, an autonomous ground based sensor network could be more economical to implement compared to off-ground systems and can offer all-weather, 24-hour uninterrupted observation capabilities. In order to demonstrate the practicality of a ground based sensor network, the signal response and frequency response signature and sensitivity of a vegetation layer must be examined in an accurate manner. A web of RF sensors can be embedded inside and scattered throughout a crop field; in such a network, given that a suitable inverse or reconstructive algorithm exists, measured inter-nodal responses can be used as the means to infer an assortment of properties intrinsic to the propagation medium. In this preliminary analysis, the node-to-node propagation characteristics are calculated by modeling the crop and soil layers as an isotropic, two-layer half-space, homogenous medium with the assumption that the system is operating at a wavelength that is large in comparison to the physical features of the individual plants. The source of excitation at the transmitter node is assumed to be an electric dipole, and the observation node is to be located in the far field region of the source. Near-earth wave propagation in the presence of a vegetation layer has been considered in a number of studies in which asymptotic methods are applied to characterize a transmitter’s far field radiation pattern. The nature of the dominant wave component stimulating a receiver depends on the locations of the transmitter and receiver with respect to the vegetation layer and can be classified into lateral waves, geometrical-optics waves, and Norton-type correction waves; the relevancy of each of these types of waves are discussed in detail in [1], [2]. The successful deduction of useful quantities such as soil composition, vegetation transpiration rate, crop biomass, and crop health state and growth rate from electromagnetic observables (e.g., signal path loss, spectral response, etc.) requires a thorough understanding of the interaction between the propagating wave and the vegetation medium. The physical properties of the vegetation are partly reflected in the value of its effective dielectric constant. It is wellknown that healthy or robust plants can be differentiated from sick or stressed plants by examining their reflectance in the near-infrared wavelengths. Similar health monitoring techniques can be exploited using the propagation data collected by a ground sensor grid. For a uniformly irrigated crop field, for example, potentially diseased patches of plants can be distinguished from healthy ones by noting that healthy plants would intake and retain water more consistently and proficiently; the moisture content of the plants, in turn, presents a direct impact on the signal path loss among the nodes of a sensor network. In the same manner that multispectral information is employed in radar imagery to infer additional soil and plant properties, a frequency response characterization can also extend the capability of a ground sensor system.