**1. Introduction**

In the last few decades, Wireless Sensor Networks have become indubitably one of the most important technologies for both development and investigation in various fields as health care [1], agriculture [2], domestic monitoring [3], and means of land transport [4], among others. However, although in most of applications, numerous sensor nodes are distributed around a wide geographical area while the reliability of the wireless communication in terms of network coverage is a crucial requirement, some environments can be tough and make it hard to achieve the desired communication between nodes; thus resulting in a non-responsive application. Inhomogeneous vegetation environment is one of the most common cases that cause wireless communication complexities. Although WSN are expected to automatically monitor the ecological evolution and wildfire habits in forests [5,6], as over 400,000 wildfires have occurred in Spain over the last 30 years [7] also many rescue interventions are carried out to retrieve people lost in the woods using mobile wireless communication, inhomogeneous vegetation environments have the special feature of acting as scatterers of electromagnetic waves. The signal scattering is translated into an excess of attenuation which can limit the performance of the Device-to-Device (D2D) communications envisaged at low power, high data rates and low latency expected for the upcoming 5G mobile wireless communications. D2D is expected to be used in almost every environment and scenario, and very especially in smart cells. The reduced coverage area and need for high data rate performance of this type of cell deployment can be limited by the attenuation due to vegetation.

In fact, the appearance of the trees in the path of the communication link between the transmitter and the receiver has significant effects on the quality of the received signal. This is because a forest is characterized by vegetation of different canopies and components with different physical natures in terms of trees height, leaves pattern and thickness, trunk sizes and number of branches, which determine the rate of attenuation of radio waves that propagate through it via scattering, absorption, refraction and diffraction of the waves. The signal scattering is translated into an excess of attenuation which can limit the performance of the Internet of Things (IoT) envisaged at high data rates and low latency expected for the upcoming 5G mobile wireless communications. In this context, radio planning tasks become necessary in order to assess the validity of future D2D communications operating in vegetation environments.

For that purpose, path loss models for scenarios with vegetation play a key role since they provide RF power estimations that allow an optimized design and performance of the wireless network [8]. Moreover, a path loss model may contribute to the evaluation of the maximum effective distance between adjacent terminals and hence to the estimation of the number of sensors needed to cover a certain area. Finally, the signal strength loss is related to the quality of service (QoS), causing unreliable communication between nodes that will increase both the number of data packet retransmissions and the power consumption of the nodes, causing radio link failure in last term. Therefore, there is a need for reliable through-vegetation radio channel modelling for vegetation environments, which will assess the propagation behavior in terms of both path loss and multipath propagation.

Since the 1960s, a significant amount of work has been done to investigate the characterization of the radio channel blocked by vegetation elements [9,10], which proposed different analytical and empirical models to estimate the power attenuation or level excess loss introduced by a signal blockage due to vegetation obstacles, mainly trees [9–23]. Among the conventional forest models, we can mention the COST 235 [11] and the Weissberger [12]. These models were developed to model the excess attenuation encountered in a forest beyond that predicted by either free-space or two-ray propagation. Moreover, the path loss was empirically modeled in [11,13–15]. In [16–18], path loss models for a specific 1.9 GHz radio frequency are presented. More characteristics such as the shadowing loss [19] and small-scale fading [19,20] for trees as well as tall food grass fields [21] were empirically modeled. In addition, a model for ultra-wideband (UWB) channels is presented in [22].

However, although a significant amount of research has been performed on the empirical propagation loss modeling, it is still a challenge to describe the radio wave propagation within the forest environment accurately. Especially in inhomogeneous vegetation environments, since a forest can be made up of mixed or homogeneous tree types resulting in different effects on radio waves even at the same frequency by the same group of trees, depending on the geometry of the link. A model combining path loss and multipath is rarely considered.

In this contribution, which is an extension of a conference paper [24], we present a simple model to characterize the attenuation due to the isolated trees in an air-to-air communication channel occurring between a static transmitter and a mobile user which moves linearly toward the tree. The developed channel model also considers parameters due to the multipath presence obtaining the value of the delay spread parameter.

Since an accurate modeling of the propagation of radio waves through tree foliage, generally requires accurate electromagnetic description of the tree geometry, including its branches and leaves, the radio characterization was performed by means of simulations based on 3D Ray Launching software, where the specific material parameters of the vegetation elements are considered, such as

dielectric constant and conductivity. A geometrical and dielectric model of the trees were created and introduced in the simulation software for two species: pine and oak. The scenario simulated and measured at 2.4 GHz to determine the effects of signal level blockage consisted of a medium-size single oak tree immersed in a forest. The path loss was estimated as dependent of the radio link range. Simulation and measurement results corroborate the existence of a double zone within the path loss curve due to different dominant propagation mechanisms: scattering (near the tree) and line of sight obstruction (far from the tree). Furthermore, 2D planes of delay spread values obtained by simulation similarly reflect this two-zone propagation model.

The measurements were carried out using devices operating at the band of 2.4 GHz. For the isolated tree case, a specific forest (called Chandebrito, situated in Galicia, Spain) composed of different type of trees was the chosen scenario, since it suffered a fire and some isolated trees can be found there nowadays. On the other hand, the Orgi Forest, located in Navarra, Spain, was the selected scenario for the analysis of non-isolated tree case.

The paper is organized as follows. In Section 2, the radio channel characterization for a D2D communications blocked by an isolated single tree is presented. Both simulations by the in-house developed 3D Ray Launching algorithm and measurements in a real forest environment are shown. In Section 3, the same approach is used for the radio channel characterization when the blocking tree is a non-isolated single tree, i.e., when dense vegetation is present in the surroundings. Section 4, Discussion, closes the paper.

### **2. Isolated Single Tree Radio Channel Characterization**

In this section, the radio channel characterization at 2.4 GHz when an isolated single tree is blocking the communication path is presented. First, wireless channel estimation methodology based on deterministic 3D Ray Launching algorithm is described and employed. For that purpose, a simulation scenario is implemented, including dimensions and electromagnetic frequency dispersive properties of the employed materials. Then, measurements in a real scenario have been performed in order to complete the obtained simulation-based radio channel model.

### *2.1. Simulation Software*

In order to perform wireless channel estimation, in which multiple elements such as vegetation are considered, different approaches can be employed. These can be based on analytical models, usually employing first degree approximations to simplify them, or empirical/semi-empirical models, which rely on measurement-based regressions.

Theoretical or analytical approaches are based on statistical theory and usually provide accurate propagation phenomena predictions [25–28]. Two different types of theoretical models can be distinguished in the literature, namely the radiative energy transfer (RET) model [29] and the analytical theory approach by Foldy [30]. These two models have successfully been used to simulate radio wave propagation in vegetation environments but, due to the complexity of the mathematical equations on which they are based, and the difficulty to extract the input parameters of the models, such as density, area or thickness of leaves and branches, they tend to be unaffordable for real-sized scenarios, such as forests [31].

On the other hand, empirical techniques are based on extensive measurement campaigns in the considered environment [32–36]. They have successfully characterized path loss attenuation impact in diverse scenarios with inhomogeneous vegetation within them, and at different frequency bands. However, their main drawback is their lack of accuracy when different site-specific environments are analyzed. As an example, the work in [37] verifies that the well-known propagation empirical models can lead to an error percentage of 30% in propagation predicted values for a classic tomato greenhouse.

To overcome these limitations, ray-tracing based methods have been proposed in the literature. In [38], a propagation model based on uniform theory of diffraction (UTD) has been presented for urban environments, considering the impact of propagation over buildings and the vegetation attenuation

and scattering. In [39], another propagation model based on geometrical optics (GO) is presented to assess the scattered field caused by vegetation elements in the radio path. Methods based on GO such as ray tracing or ray launching achieve a good trade-off between simulations accuracy and computational cost [40]. Due to this fact, in this work, an in-house developed 3D Ray Launching (3D-RL) algorithm has been used to characterize inhomogeneous vegetation environments. The 3D-RL model is a deterministic technique based on GO and UTD and it is divided in three main steps:


The ad-hoc 3D ray launching algorithm has been described in detail in [40] and validation in complex environments including vegetation has been described in [41]. In the algorithm, the consideration that the vegetation medium is homogeneous has been assumed, thus it has been treated as an isotropic dielectric material with constant permittivity and conductivity. The 3D RL algorithm considers ray/object interaction in terms of reflection, refraction and diffraction. In particular, diffraction phenomena can be optionally activated in the simulation parameters, which is the case of the simulation results presented in this work.
