Exploring Ground Reflection Effects on Received Signal Strength Indicator and Path Loss in Far-Field Air-to-Air for Unmanned Aerial Vehicle-Enabled Wireless Communication

Abstract

Unmanned aerial vehicle (UAV)-enabled wireless communications are becoming increasingly important in applications such as maritime and forest rescue operations. UAV systems often depend on wireless networking and mobile edge computing (MEC) devices for effective deployment, particularly in swarm UAV-enabled MEC configurations focusing on channel modeling and path loss characteristics for air-to-air (A2A) communications. This paper examines path loss characteristics in far-field (FF) ground reflection scenarios, specifically comparing two environments: FF1 (forest floor) and FF2 (seawater floor). LoRa modules operating at 868 MHz were deployed for communication between a transmitting UAV (Tx-UAV) and a receiving UAV (Rx-UAV) to conduct this study. We investigated the received signal strength indicator (RSSI) and path loss characteristics across channel bandwidths of 125 kHz and 250 kHz and spread factors (SF) of 7, 9, and 12. Experimental results show that ground reflection has minimal impact in the FF1 scenario, whereas, in the FF2 scenario, ground reflection significantly influences communication. Therefore, in the seawater environment, a UAV-enabled LoRa MEC configuration using a 250 kHz bandwidth and an SF of 7 is recommended to minimize the effects of ground reflection.

1. Introduction

Unmanned aerial vehicles (UAVs), commonly known as drones, have gained significant attention in recent years due to their versatility and potential in a wide range of applications [1]. Among the most impactful areas of UAV deployment are rescue maritime operations and forest rescue operations [2,3], where UAV-enabled wireless communication plays a critical role [4]. In both fields, the ability to provide real-time data transmission, communication, and monitoring is essential for operational efficiency, safety, and decision-making.

In maritime rescue operations [2], UAVs provide real-time monitoring over vast oceanic areas, enabling faster response times during search and rescue (SAR) missions. Equipped with cameras, sensors, and communication modules [5], UAVs can quickly locate distressed vessels or individuals stranded at sea, even in poor visibility conditions. They can also serve as communication relays, transmitting live data and video feeds to rescue teams onshore or aboard ships, facilitating more informed decision-making. However, reliable wireless communication is essential in these operations, as the open sea environment often lacks the infrastructure required for traditional communication systems. UAV-enabled wireless communication networks, including mobile edge computing (MEC) systems [6], can address these challenges by providing robust, low-latency communication channels between UAVs and ground control stations.

Similarly, forest rescue operations pose unique challenges due to dense vegetation, uneven terrain, and limited line-of-sight communication [7]. UAVs are invaluable in such scenarios, providing a bird-eye view that enables search teams to cover large forested areas in a fraction of the time required by traditional methods. They can be deployed rapidly to survey remote and inaccessible regions, relay communications to ground rescue teams, and even deliver supplies to isolated individuals. However, wireless communication in forest environments is complex, as signal propagation can be affected by ground reflections, vegetation, and other environmental factors [8]. Ensuring reliable communication in such scenarios requires careful consideration of channel modeling and path loss characteristics for air-to-air (A2A) and air-to-ground (A2G) communications [9,10].

This paper focuses on UAV-enabled MEC in rescue operations for maritime and forest environments as shown in Figure 1, particularly examining the path loss characteristics and propagation challenges in these distinct terrains. Through an analysis of theoretical and empirical models, this study provides insights into optimizing UAV communication systems for efficient and reliable operation in both maritime and forest rescue missions.

2. Related Works

The widespread deployment of Internet of Things (IoT) devices, such as IoT-based WiFi (IEEE 802.11) [11], low power wide area networks (LPWAN) [12], Narrowband IoT (NB-IoT) [13], and Sigfox [14], has become integral to modern applications including healthcare [15], transportation [16], smart farming [17], and smart cities [18], among others [19]. One of the key challenges in IoT communication lies in channel modeling and ensuring energy efficiency, coverage, latency, reliability, and connectivity. Notably, the propagation limitations of mobile edge computing (MEC) devices significantly impact the coverage range and the reliability of communication systems. Therefore, there is a critical need for accurate analytical channel models to address MEC requirements for wide coverage, reliability, and energy efficiency.

Similarly, air-to-ground (A2G) channel modeling presents a significant challenge in UAV-enabled wireless communication systems [20]. In contrast, air-to-air (A2A) channel modeling introduces new complexities, especially for vehicle-to-vehicle (V2V) networking [21]. Accurate channel modeling for UAVs is critically needed, particularly in IoT and 5G applications [22]. For IoT device networks, UAVs can act as mobile base stations (BS), effectively supporting IoT services in uplink and downlink communication across dispersed geographical areas. In uplink scenarios, UAVs can dynamically adjust their position based on IoT device locations to optimize energy efficiency and reliability in data collection. This concept of UAV-enabled BS for uplink communication was explored in [23]. In downlink scenarios, UAVs can be strategically positioned closer to IoT devices, enhancing coverage for IoT systems [24]. Integrating UAVs and IoT necessitates reliable wireless communication technologies, such as WiFi or LoRa, to ensure connectivity.

LoRa communication is part of the LPWAN technology family, alongside Sigfox and NB-IoT, and is widely recognized for enabling energy-efficient, long-distance communication, particularly in UAV applications [25,26]. Communication performance requirements between multiple UAVs and ground stations are defined by standards such as 3GPP TS 22.261, TR 22.261, and TR 36.777, which are crucial for swarm UAV operations [27]. When deploying UAV swarms, it is essential to evaluate air-to-air channel modeling (A2A-CM) across different environments. Due to large-scale propagation characteristics, A2A-CM is influenced by factors such as free-space path loss, ground reflections, Doppler shifts, and shadowing. Compared to mobile wireless channels, UAV-based air-to-ground (A2G) and air-to-air (A2A) channels generally experience higher dispersion, more substantial terrestrial shadowing attenuation, and faster channel variations.

The application of LoRa modules in UAV communication channels was examined in [28], where it was shown that LoRa provides stable signal strength and high accuracy over long distances. The Chirp Spread Spectrum (CSS) modulation used by LoRa enables more efficient communication links in aerial applications [29]. Additionally, channel models for LoRa at 433 MHz and 868 MHz frequencies, based on ground-to-ground (G2G) communication models, have been studied in urban environments using path loss-based empirical measurements [30]. LoRa’s capabilities make it particularly advantageous for smart city applications, such as UAV-based air pollution monitoring [31], as well as in smart agriculture for tasks like soil quality analysis [32].

In the field of path loss modeling, several studies have aimed to optimize path loss and channel characteristic predictions. In [33], a machine learning-based approach was introduced for predicting air-to-air (A2A) path loss. The authors used a ray-tracing simulation model to generate data for the A2A channel, with the transmitting UAV (Tx-UAV) equipped with a directional antenna and the receiving UAV (Rx-UAV) moving at 2 m intervals across different altitudes. Path loss predictions were then conducted using k-nearest neighbors (kNN) and random forest algorithms. While the random forest algorithm yielded more accurate predictions, the study did not consider realistic environmental factors that impact UAV communication channels.

Subsequently, a path loss prediction model under more realistic conditions was developed for air-to-ground channel modeling (A2G-CM) using both kNN and random forest algorithms [34]. The results showed that the random forest algorithm outperformed kNN in channel prediction accuracy. Additional studies have also applied machine learning techniques to path loss prediction [35,36], demonstrating that these algorithms can effectively model and predict the uncertainties associated with path loss characteristics in both A2A-CM and A2G-CM scenarios.

For air-to-air channel modeling (A2A-CM), a comprehensive survey on UAV communication channels was presented in [37], highlighting that the propagation characteristics of high-altitude UAV A2A links are largely influenced by free space models and Doppler shift effects. In contrast, low-altitude UAVs must consider additional factors, such as ground reflections and shadowing. The primary objective of A2A communication is typically to utilize UAVs as relay networks for multi-hop communication. In [38], the authors proposed an A2A data link designed to support addressed, multicast, and broadcast communication for multi-hop networks. They introduced key optimizations, including a medium access control (MAC) scheme, forward error correction codes, and a modulation scheme to enhance the performance of the A2A data link.

Further, the propagation characteristics of A2A channels in urban environments were explored in [39], where the authors utilized a ray-tracing technique to model path loss in large-scale fading environments. They demonstrated that excess fading loss (EL) and close-in free space (CI) models could be employed as effective tools for A2A channel modeling. Additionally, the A2A channel path loss in flying ad hoc networks (FANETs) was examined using the Stanford University Interim (SUI) model [40]. However, it was observed that the SUI model may be unsuitable for A2A communication as the results were uncorrelated with propagation environments, especially in rugged terrains.

Moreover, a wideband A2A channel model was simulated in [41] to evaluate its suitability for drone-to-drone communications. The results aligned with realistic propagation mechanisms, revealing a rich multipath environment caused by ground reflections. Ground reflection effects were further studied in terms of angle-of-arrival (AOA) and angle-of-departure (AOD) for ground scattering in A2A channel modeling [42], where the authors proposed methodologies to analyze multi-scattering cluster scenarios effectively.

Three-dimensional (3D) wideband non-stationary models for A2A-CM have been discussed in [43,44], but these studies lacked consideration of ground reflection conditions in realistic environments, particularly for drone-to-drone communications and small UAVs.

This paper presents an empirical study on path loss channel characterization for air-to-air communication modeling (A2A-CM) under a ground reflection model, specifically in far-field (FF) communication scenarios. The study employs LoRa-based communication modules operating at 868 MHz for UAV-enabled MEC applications. Experimental results are analyzed in terms of received signal strength indicator (RSSI) and path loss, with measured data compared against predictions from the analytical A2A Two-Ray (A2AT-R) model and the modified Log-distance model [45].

3. Methodology

The A2AT-R model is an analytical model designed to investigate air-to-air (A2A) communication based on ground reflection. Figure 2 illustrates the model, where the Tx-UAV represents the transmitter in a UAV-enabled wireless communication system, and the Rx-UAV represents the receiver. The proposed A2AT-R model, along with the modified Log-distance model, was introduced in previous work [45] to characterize the propagation influenced by ground reflections for near-field UAV communications. However, these models did not account for far-field conditions. In this work, we investigate the far-field conditions of two ground surfaces: FF1 (forest floor) and FF2 (seawater floor). On the Rx-UAV side, we evaluate the RSSI and path loss characteristics. By comparing the actual data with the analytical A2AT-R and modified Log-distance models, we aim to better understand the propagation behavior in these environments.

3.1. Path Loss Analysis Based on A2AT-R Model

As mentioned in Figure 2, the total energy at the Rx-UAV receiver site is written as

$$E_U\left(d_c\right)=E\left(d^{\prime }\right)+E_r\left(d^{″}\right)$$

(1)

where $E(d^{\prime })$ is the line-of-sight (LOS) of $E$-field, and $E_r\left(d^{″}\right)$ is the ground reflected ray of $E$-field.

Note that two propagation waves arrive at the Rx-UAV such as the direct wave that is a distance $d^{\prime }$; and the reflected wave that is a distance $d^{″}$. Thus, $d_c$ is the separation distance between Tx-UAV and Rx-UAV. The $E$-field in LOS at the Rx-UAV is given as

$$E\left(d^{\prime }\right)={\displaystyle \frac{E_0{d}_0}{d^{\prime }}}\cos \left(2\pi f_c\left(t-{\displaystyle \frac{d^{\prime }}{c}}\right)\right)$$

(2)

and the ground reflected from the floor, which has a distance of $d^{″}$, is given as

$$E_r\left(d^{″}\right)={\Gamma }_{Floor}{\displaystyle \frac{E_0{d}_0}{d^{″}}}\cos \left(2\pi f_c\left(t-{\displaystyle \frac{d^{″}}{c}}\right)\right)$$

(3)

where $t$ is the duration time, $E_0$ is the $E$-field at a reference distance $d_0$ from the Rx-UAV when $d>d_0$, where $d$ denotes the separation distance of Tx-UAV and Rx-UAV, ${\Gamma }_{Floor}$ denotes the reflection coefficient from the floors, $c$ is the velocity of the light, and $f_c$ is the carrier frequency.

Thus, the $E_U\left(d_c\right)$ can be expressed as

$$E_U\left(d_c\right)={\displaystyle \frac{E_0{d}_0}{d^{\prime }}}\cos \left(2\pi f_c\left(t-{\displaystyle \frac{d^{\prime }}{c}}\right)\right)+\hspace{0.17em}\left(-1\right){\displaystyle \frac{E_0{d}_0}{d^{″}}}\cos \left(2\pi f_c\left(t-{\displaystyle \frac{d^{″}}{c}}\right)\right)$$

(4)

where ${\Gamma }_{Floor}=-1$ is the perfect ground reflection component from the floor.

In terms of RSSI, we obtain the received signal for the analytical A2AT-R model, that $r_{T-R}\left(t\right)$ can be expressed as

$$r_{T-R}\left(t\right)=\left({\displaystyle \frac{\lambda }{4\pi }}\left|r_{d^{\prime }}\left(t\right)+r_{d^{″}}\left(t\right)\right|\cdot e^{j2\pi f_{c}t}\right)$$

(5)

$$r_{d^{\prime }}\left(t\right)={\displaystyle \frac{\sqrt{G_{T,d^{\prime }}{G}_{R,d^{\prime }}}}{d_0}}{e}^{-j2\pi {\displaystyle \frac{d_0}{\lambda }}}\cdot {\left|E\left(d^{\prime }\right)\right|}^2$$

(6)

$$r_{d^{″}}\left(t\right)={\displaystyle \frac{\sqrt{G_{T,d^{″}}{G}_{R,d^{″}}}}{d_0}}{e}^{-j2\pi {\displaystyle \frac{d_0}{\lambda }}}\cdot {\left|E_r\left(d^{″}\right)\right|}^2$$

(7)

where $G_{T,d^{\prime }}$ and $G_{R,d^{\prime }}$ are the antenna gain of the Tx-UAV and Rx-UAV in LOS direction, respectively, $G_{T,d^{″}}$ and $G_{R,d^{″}}$ are the antenna gain of the Tx-UAV and Rx-UAV in ground reflection ray, respectively.

The $h_{\mathrm{Tx}-\mathrm{UAV}}$ and $h_{\mathrm{Rx}-\mathrm{UAV}}$ are the Tx-UAV and Rx-UAV altitudes. The different distance and the phase between the LOS ray and the ground reflection ray are given by

$$\Delta d={\displaystyle \frac{2h_{\mathrm{Tx}-\mathrm{UAV}}{h}_{\mathrm{Rx}-\mathrm{UAV}}}{d_c}}$$

(8)

$$\Delta \theta ={\displaystyle \frac{4\pi h_{\mathrm{Tx}-\mathrm{UAV}}{h}_{\mathrm{Rx}-\mathrm{UAV}}}{\lambda d_c}}$$

(9)

The RSSI can be approximately calculated as follows

$$\begin{array}{l}{P}_{\mathrm{Rx}-\mathrm{UAV}}(d_c)\approx P_{\mathrm{Tx}-\mathrm{UAV}}{\left({\displaystyle \frac{\lambda }{4\pi }}\right)}^2{\displaystyle \frac{G_T{G}_R}{d_c^2}}\hspace{0.17em}{\left({\displaystyle \frac{4\pi h_{\mathrm{Tx}-\mathrm{UAV}}{h}_{\mathrm{Rx}-\mathrm{UAV}}}{\lambda d_c}}\right)}^2\\ \hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}=P_{\mathrm{Tx}-\mathrm{UAV}}{G}_T{G}_R{h}_{\mathrm{Tx}-\mathrm{UAV}}^2{h}_{\mathrm{Rx}-\mathrm{UAV}}^2{d}_c^{-4}\end{array}$$

(10)

where $G_T$ is the represented value for $G_{T,d^{\prime }}$ and $G_{T,d^{″}}$. The $G_R$ is the represented value for $G_{R,d^{\prime }}$ and $G_{R,d^{″}}$, respectively.

Thus, the path loss analytical A2AT-R model can be expressed by

$$P{L}_{\mathrm{A}2\mathrm{AT}-\mathrm{R}}\left(\mathrm{dB}\right)\approx 10{\log }_{10}\left[{\displaystyle \frac{1}{4G_T{G}_R}}{\left({\displaystyle \frac{4\pi d_c}{\lambda }}\right)}^2\cdot \left({\displaystyle \frac{2\pi h_{\mathrm{Tx}-\mathrm{UAV}}{h}_{\mathrm{Rx}-\mathrm{UAV}}}{\lambda d_c}}\right)\right]$$

(11)

3.2. Path Loss Analysis Based on Modified Log-Distance Model

The modified Log-distance model for the channel modeling of UAV communications can be extended from the conventional Log-distance model as

$$P{L}_f=K{\left({\displaystyle \frac{d_c}{d_0}}\right)}^{\alpha }$$

(12)

where $K$ denotes the unit-less scaling factor, and $\alpha$ is the path loss exponent. Then, the path loss in dB scale is given by

$$P{L}_f\left(\mathrm{dB}\right)=K\left(\mathrm{dB}\right)+10\alpha {\log }_{10}\left({\displaystyle \frac{d_c}{d_0}}\right)$$

(13)

For the modified Log-distance model, one form closely similar to the general form of Log-distance model is the floating intercept model given by

$$P{L}_f\left(\mathrm{dB}\right)=10\alpha \left(h_{\mathrm{Rx}-\mathrm{UAV}}\right){\log }_{10}{d}_c+\beta \left(h_{\mathrm{Rx}-\mathrm{UAV}}\right)$$

(14)

where $\alpha$ and $\beta$ are usually jointly determined by minimizing the mean square error between the model and the empirical measurements. The $h_{\mathrm{Rx}-\mathrm{UAV}}$ denotes the height of Rx-UAV, $\alpha \left(h_{\mathrm{Rx}-\mathrm{UAV}}\right)$ and $\beta \left(h_{\mathrm{Rx}-\mathrm{UAV}}\right)$ determined based on measurement set up are given by

$$\alpha \left(h_{\mathrm{Rx}-\mathrm{UAV}}\right)=\max \left(3.9-0.9{\log }_{10}\left(h_{\mathrm{Rx}-\mathrm{UAV}}\right),2\right)$$

(15)

$$\beta \left(h_{\mathrm{Rx}-\mathrm{UAV}}\right)=-8.5+20.5{\log }_{10}\left(\min \left(h_{\mathrm{Rx}-\mathrm{UAV}},h_{\mathrm{FSPL}}\right)\right)$$

(16)

where $h_{\mathrm{FSPL}}$ is the height where free-space propagation loss (FSPL).

Finally, the modified Log-distance model which the antenna characteristics, certain propagation environment, and frequency term are considered. The expression is given by

$$\begin{array}{l}PL\left(\mathrm{dB}\right)=P{L}_f\left(\mathrm{dB}\right)+10\alpha \left(h_{\mathrm{Rx}-\mathrm{UAV}}\right){\log }_{10}\left({\displaystyle \frac{d_c}{d_0}}\right)\\ \hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}-10{\log }_{10}\left({\displaystyle \frac{h_{\min }}{\Delta h}}\right)+C_l+10{\log }_{10}\left(1+f_c\right)\end{array}$$

(17)

where $\Delta h=\left|h_{\mathrm{Rx}-\mathrm{UAV}}-h_{\min }\right|$, $h_{\min }$ is the minimum height of the Rx-UAV that gives the lowest path loss for a given environment, $C_l$ is a constant polarization loss from UAV antenna orientations.

4. Experimental Setup

4.1. LoRa Communication Network

LoRa [46] is a wireless communication technology under a low-power wide area network (LPWAN), LPWAN communication between connected devices. It is primarily used in Internet of Things (IoT) applications where devices require low data rates and operate in environments with limited power resources. LoRa module uses a proprietary modulation scheme called chirp spread spectrum (CSS), LoRa offers different SF typically ranging from SF = 7, SF = 9, and SF = 12. Thus, the relationship between the data rate of CSS modulation is determined by the chirp rate, SF, and the symbol rate of LoRa is expressed

$$R=SF\times {\displaystyle \frac{B}{2^{SF}}}$$

(18)

where $B$ is the channel bandwidth.

However, when LoRa uses forward error correction (FEC) for data validation, the receiver can determine the appropriate code rate (CR), where the CR is defined as 4/(4 + n), with n ranging from 1 to 4. Considering Equation (18), it can be rewritten as follows

$$R=SF\times {\displaystyle \frac{B}{2^{SF}}}\times CR$$

(19)

Figure 3 illustrates the LoRa communication network, comprising three key sections: edge computing or the LoRa module (between the transmitter and receiver), the Internet or cloud network, and the recorder or web applications. At the edge computing level, the LoRa transmitter sends downlink data to the LoRa receiver. The uplink channel then facilitates communication from sensors or actuators to the Internet when the network server processes the device user interface (DeviceUI) is processed by the network server. Within the Internet, the Message Queue Telemetry Transport (MQTT) protocol subscribed DeviceUI to the network server and published data to the application server, enabling real-time monitoring of data from the edge computing layer.

Although LoRa is capable of long-range communication, the actual range over water can vary due to environmental factors like signal reflection from the ground, and signal absorption in dense forests might worsen in the environment.

4.2. Measurement Setup

Figure 4 shows the UAV under test where Tx-UAV as shown in Figure 4a and Rx-UAV as shown in Figure 4b. Figure 5 shows the 2D graphical scheme and 3D satellite map of the FF1 and FF2 scenario.

The field measurement test was conducted at KMITL Prince of Chumphon Campus, Thailand. In the FF1 and FF2 scenarios, the Tx-UAV was positioned at a fixed altitude of 5 m at coordinates 10.726136, 99.375188, and for the FF2 scenario, it was located at 10.721133, 99.385508. The Rx-UAV was tested at varying altitudes ranging from 1 m to 20 m, with coordinates of 10.730588, 99.381898 for the FF1 scenario, and 10.719542, 99.393762 for the FF2 scenario. The parameters for the measurement setup are outlined in

Table 1. Parameters of the measurement setup.

Parameters	Values
LoRa frequency	868 MHz
LoRa channel bandwidth, B	125 kHz, 250 kHz
Spread factor (SF)	7, 9, 12
Code rate (CR)	4/8
Tx-UAV and Rx-UAV flight time	20 min.
Transmitted power	23 dBm
LoRa module antenna gain	3.2 dBi
Tx-UAV height	5 m
Rx-UAV height	1–20 m
Separation distance	1 km
LoRa received signal sensitivity	−137 dBm

, and Figure 6 illustrates the field test setups for both FF1 and FF2 scenarios.

5. Result and Discussion

The results of RSSI and path loss characterizations are presented in Figure 7, Figure 8, Figure 9 and Figure 10. The measurement data for RSSI and path loss were analyzed by comparing two channel bandwidths, 125 kHz and 250 kHz, of the LoRa module, with varying spread factors (SF) of 7, 9, and 12.

The RSSI in the FF1 (forest floor) scenario is shown in Figure 7a,b. Figure 7a illustrates the RSSI (dBm) versus Rx-UAV altitudes, ranging from 1 m to 20 m. The blue line represents measurement data for SF = 7, with RSSI values ranging from −104.63 dBm to −100.50 dBm. The red line shows SF = 9, where RSSI ranged from −106.21 dBm to −101.50 dBm, and the black line represents SF = 12, with values from −106.41 dBm to −103.50 dBm. A notable drop in RSSI was observed as the SF increased from 9 to 12, with a reduction of approximately 4.2 dBm. The phenomenon observed in Figure 7a,b, where the RSSI increases between 1 m and 15 m, and then decreases beyond 15 m, can be explained by the effects of forest shadowing and separation distances. At altitudes between 1 m and 15 m, the Rx-UAV is initially within the forest shadowing zone, where signal attenuation is higher due to obstructions from trees and vegetation. As the Rx-UAV rises above this shadowing zone (between 1 m and 5 m), the signal experiences less attenuation, leading to an increase in RSSI. Between 5 m and 15 m, the signal encounters fewer obstacles, resulting in higher RSSI values.

However, beyond 15 m, the distance-dependent path loss becomes the dominant factor. As the altitude increases further, the signal travels over a greater distance, causing the RSSI to degrade due to free-space path loss.

Figure 8 reports the RSSI characteristics in the FF2 (seawater floor) scenario. In Figure 8a, we examine a 125 kHz channel bandwidth. The blue line represents measurement data for SF = 7, with RSSI values ranging from −103.31 dBm to −101.60 dBm. The red line shows SF = 9, where RSSI ranges from −101.31 dBm to −99.70 dBm, and the black line indicates SF = 12, with values from −98.51 dBm to −97.70 dBm. Unlike the FF1 scenario, RSSI levels in FF2 fluctuate at Rx-UAV altitudes between 1 m and 8 m over the seawater, influenced by ground reflection when the Rx-UAV hovers near the water surface. This fluctuation is due to multipath fading, a characteristic propagation effect over seawater, impacting UAV-enabled wireless communication. However, this ground reflection effect diminishes when the Rx-UAV altitude exceeds 10 m. Additionally, increasing SF impacts RSSI levels.

For the 250 kHz channel bandwidth, shown in Figure 8b, the RSSI level increases. The blue line represents SF = 7, with RSSI ranging from −106.41 dBm to −103.50 dBm. The red line shows SF = 9, with RSSI values from −106.21 dBm to −101.50 dBm, and the black line for SF = 12, where RSSI ranges from −104.63 dBm to −100.50 dBm.

From the RSSI results, we found that ground reflection in the FF2 (seawater floor) scenario significantly impacted Rx-UAV altitudes between 1 m and 8 m over the seawater.

The path loss characteristics for FF1, shown in Figure 9, and FF2, depicted in Figure 10, the measurement data curved are compared with the A2AT-R model and the modified Log-distance model. The simulated path loss for the A2AT-R model ranges from 84.70 dB to 95.27 dB, while the modified Log-distance model ranges from 73.31 dB to 85.85 dB.

Figure 9a illustrates the path loss characteristics at a channel bandwidth of 125 kHz. For SF = 7, the path loss ranges from 84.11 dB to 80.77 dB; for SF = 9, it ranges from 86.11 dB to 81.77 dB; and for SF = 12, it ranges from 86.11 dB to 83.77 dB. Figure 9b shows a channel bandwidth of 250 kHz. For SF =7, the path loss ranges from 79.34 dB to 78.23 dB; for SF = 9, it increases from 81.23 dB to 79.87 dB; and for SF = 12, it varies from 84.12 dB to 81.54 dB. Both channel bandwidth parameters indicate that path loss increases with higher SF values.

At Rx-UAV altitudes between 1 m and 15 m, path loss decreases as the RSSI increases because the Rx-UAV rises above the forest shadowing zone. However, above 15 m, path loss increases due to the greater separation distance between the Tx-UAV and Rx-UAV.

Additionally, the measurement data reveal that the path loss curves differ from those predicted by the A2AT-R and modified Log-distance models. This confirms that, in the FF1 scenario, ground reflection has minimal impact on wave propagation, while shadowing, attenuation, and absorption effects from the forest are the dominant factors.

In contrast, in the seawater floor scenario (FF2), the actual path loss data align more closely with the modified Log-distance model, as shown in Figure 10a,b. Ground reflection effects are strongly evident in this scenario, particularly at Rx-UAV altitudes between 1 m and 8 m. At these lower altitudes, signal fluctuations are observed at the Rx-UAV due to the impact of ground reflections, which introduces multipath fading characteristics and variation in RSSI. Above 8 m, the path loss fluctuations decrease compared to lower altitudes. In the FF2 scenario, the actual path loss data trend aligns increasingly with the modified Log-distance model, demonstrating that this model is more accurate in predicting path loss than the analytical A2AT-R model.

Figure 10a presents the path loss characteristic curves for the measurement data at a 125 kHz channel bandwidth in the FF2 scenario. For SF = 7, the path loss ranges from 70.21 dB to 78.37 dB; for SF = 9, it ranges from 78.04 dB to 81.47 dB; and for SF = 12, it ranges from 79.43 dB to 84.52 dB. Here, the path loss for SF = 7 is the lowest, while SF = 9 and SF = 12 closely align with the modified Log-distance model.

At a 250 kHz channel bandwidth, shown in Figure 10b, the path loss curves are as follows: SF = 7 ranges from 70.01 dB to 77.12 dB, SF = 9 from 73.55 dB to 80.12 dB, and SF = 12 from 75.23 dB to 82.56 dB. The ground reflection effect is observed at Rx-UAV altitudes between 1 m and 8 m over the seawater.

To discuss this study, the results highlighted the influence of ground reflection floors and signal characteristics such as channel bandwidth and SF on RSSI and path loss in UAV-enabled wireless communication.

In the forest floor scenario (FF1), RSSI measurements exhibited distinct behaviors across varying altitudes and spreading factors (SF). At higher SFs, the RSSI values decreased, indicating that signal strength weakened as the SF increased. Additionally, at Rx-UAV altitudes between 1 m and 15 m, RSSI values increased noticeably, while actual path loss decreased due to the Rx-UAV flying above the shadowing effects from the forest canopy. Beyond 15 m, RSSI values began to degrade due to the increased distance. These findings confirm that ground reflection has a minimal effect on the Rx-UAV, and both the analytical A2AT-R model and modified Log-distance model show limited accuracy in matching the actual data.

In contrast, the seawater floor scenario (FF2) displayed unique characteristics due to the reflective nature of the seawater surface. RSSI measurements fluctuated considerably between 1 m and 8 m due to ground reflection, highlighting the impact of the seawater surface on signal propagation. The ground reflection effect diminished at altitudes above 10 m, and increased SF values led to higher RSSI levels, contrasting with the forest floor scenario. At a higher channel bandwidth (250 kHz), RSSI levels generally increased across SFs, suggesting that the wider bandwidth may help mitigate signal attenuation over reflective surfaces like water.

Path loss measurements at 125 kHz and 250 kHz channel bandwidths reveal further insights into signal behavior in both forest and seawater environments. In the forest floor (FF1) scenario, the path loss increased with higher SFs, indicating that the dense forest attenuated the signal more significantly at increased SF values. The fit curve model from both the A2AT-R and modified Log-distance models was used to predict the ground reflection effect for exploring actual path loss data.

In the seawater floor (FF2) scenario, the actual path loss data align more closely with the modified Log-distance model, indicating a more predictable ground reflection effect on signal propagation. Ground reflection effects were particularly evident at Rx-UAV altitudes between 1 m and 8 m, where signals likely fluctuated or faded due to multipath interference from reflected waves. Additionally, the findings suggest that the increased channel bandwidth of the LoRa module contributes to RSSI degradation, resulting in higher path loss. Although this study observed effects from reflected wave components, it did not analyze the power delay profile. Future work could analyze delay spread analysis to better capture multipath scattering effects from the ground surfaces.

6. Conclusions

This study investigated RSSI and path loss characteristics of UAV-enabled wireless communication in two distinct far-field ground reflection scenarios: FF1 (forest floor) and FF2 (seawater floor). Using LoRa-based communication at 868 MHz, with channel bandwidths of 125 kHz and 250 kHz and spreading factors (SF) of 7, 9, and 12, we analyzed the impact of ground reflection on signal behavior, comparing the results against the A2AT-R model and the modified Log-distance model as predictive baselines. The findings indicate that the forest floor environment (FF1) has minimal ground reflection impact. Instead, shadowing and absorption effects from the forest canopy weaken signal strength, resulting in increased path loss at higher SFs and lower bandwidths. This outcome suggests that specialized path loss models may be required for forested environments to account for shadowing and absorption effects. In contrast, the seawater floor scenario (FF2) exhibits a pronounced ground reflection effect, particularly at lower Rx-UAV altitudes (1 m to 8 m), where multipath interference from reflected waves causes fluctuating RSSI levels and increased path loss. Here, the modified Log-distance model closely aligns with the measured data, highlighting its suitability for environments with significant reflective surfaces like water. Based on these results, we recommend configuring a UAV-enabled LoRa MEC system with a 250 kHz bandwidth and an SF of 7 in maritime environments to mitigate ground reflection effects and support stable communication. These findings contribute to refined channel modeling and improved system design for UAV-enabled MEC across various environmental settings, enhancing performance in rescue and monitoring applications in forested and maritime environments.

Future work will examine dynamic propagation factors, such as antenna radiation patterns and UAV mobility, to further clarify their effects on path loss and RSSI, along with multipath fading and shadowing, which are vital for UAV communication networks in disaster response, environmental monitoring, and similar applications.