Evaluation of Aerodynamic Performance of a Multi-Rotor eVTOL During Landing Using the Lattice Boltzmann Method

Abstract

Electric vertical take-off and landing (eVTOL) aircraft are transforming urban air mobility (UAM) by providing efficient, low-emission, and rapid transit in congested cities. However, ensuring safe and stable landings remains a critical challenge, particularly in constrained urban environments with variable wind conditions. This study investigates the landing aerodynamics of a multi-rotor eVTOL using the lattice Boltzmann method (LBM), a computational approach well-suited to complex boundary conditions and parallel processing. This analysis examines the ground effect, descent speed, and crosswind influence on lift distribution and stability. A rooftop landing scenario is also explored, where half of the rotors operate over a rooftop while the rest remain suspended in open air. Results indicate that rooftop landings introduce asymmetric lift distribution due to crosswind and roof-induced flow circulation, significantly increasing rolling moment compared to ground landings. These findings underscore the role of descent speed, crosswinds, and landing surface geometry in eVTOL aerodynamics, particularly the heightened risk of rollover in rooftop scenarios.

1. Introduction

Electric vertical take-off and landing (eVTOL) aircraft have emerged as a transformative technology in urban air mobility (UAM), offering the potential to revolutionize transportation by providing efficient, low-emission, and rapid-transit solutions in congested urban environments [1]. These multi-rotor aircraft are designed to operate in complex aerodynamic conditions, particularly during critical phases such as take-off and landing, where ground effects, crosswinds, and rotor–rotor interactions significantly influence their performance and safety [2]. As eVTOLs transition from conceptual designs to operational reality, understanding their aerodynamic behavior during landing is crucial for ensuring safe and stable operations, especially in urban settings with limited landing zones and unpredictable environmental conditions [3].

The aerodynamic performance of eVTOLs during landing is inherently complex due to the interplay of multiple factors, including ground effect, descent speed, and crosswind interactions. Ground effect, a phenomenon where the proximity of the ground alters the airflow around the aircraft, can significantly enhance lift force but may also lead to instability if not properly managed [4]. Crosswinds, common in urban environments, further complicate the aerodynamic landscape by inducing asymmetric lift distribution across the rotors, potentially leading to rolling moments and an increased risk of rollover [5]. These challenges are exacerbated in scenarios where eVTOLs must land on elevated surfaces, such as building rooftops, where partial rotor coverage can create additional aerodynamic disturbances [6].

Traditional computational fluid dynamics (CFD) methods, based on the Navier–Stokes equations, have been widely used to study aerodynamic phenomena. However, these methods often require extensive computational resources and complex meshing procedures, particularly for multi-rotor configurations and complex geometries [7]. In recent years, the lattice Boltzmann method (LBM) has gained prominence as an alternative approach for simulating fluid dynamics [8]. LBM, rooted in kinetic theory and statistical mechanics, offers several advantages, including inherent parallelism, the ease of handling complex boundary conditions, and the ability to simulate multi-physics phenomena without the need for intricate mesh generation [9]. These features make LBM particularly well-suited for studying the aerodynamic behavior of eVTOLs, especially in scenarios involving ground effect, crosswinds, and rotor interactions [10].

The LBM exhibits several notable advantages, making it a robust and versatile tool for computational fluid dynamics [11,12]. The algorithm to handle a moving solid boundary condition for free surface fluid simulations with the LBM is also discussed [13] and validated [14]. Several studies have demonstrated the effectiveness of the LBM in simulating complex aerodynamic flows. For instance, Zhang utilized the LBM to investigate the ground effect on rotorcraft performance, highlighting its ability to capture the intricate flow patterns near the ground [15]. Chang applied the LBM to numerically simulate the rotor flow field of the quadrotor UAV, revealing the change law of the rotor flow field [16]. These studies underscore the potential of the LBM as a powerful tool for analyzing eVTOL aerodynamics, particularly in challenging landing scenarios.

For the landing scenario, the ground effect is a key factor. Research on this topic is primarily experimental and is often compared to the theoretical values by Cheeseman et al. [17]. For example, Sanchez-Cuevas et al. [18] conducted experimental studies on a single rotor and reported excellent agreement with Cheeseman’s theory. Powers et al. [19] experimentally investigated the ground effect on a quadcopter, indicating that the ground effect persists up to a normalized height of five. Similarly, Sharf et al. [20] examined the ground effect in an X8 multi-rotor configuration. A significant ground effect was observed, particularly for the bottom rotor, up to a normalized height of five, which aligns well with Powers’ results. Li et al. [21] elucidates the aerodynamic coupling mechanisms between the sidewall and the ground. When the ground height and sidewall proximity are minimal, propeller wake compression in the wall gap reduces flow velocity and amplifies duct thrust loss, while further ground height reduction generates ground vortices near the duct that decelerate surrounding flow, causing rapid declines in both thrust and side force.

Aerodynamic interactions in multi-propeller configurations also serve as an important benchmark for validating numerical CFD tools [22]. In recent years, the rotorcraft industry and academic research have focused extensively on developing mid-fidelity aerodynamic solvers for the preliminary design of novel VTOL aircraft configurations [23,24], and for investigating complex flow phenomena in rotorcraft designs like tiltrotors and compound helicopters [25,26].

Despite these studies, there remains a gap in the literature regarding the comprehensive evaluation of eVTOL landing performance under realistic urban conditions, including the combined effects of ground effect, descent speed, and crosswinds. Furthermore, the unique challenges posed by rooftop landings, where partial rotor coverage can lead to asymmetric lift distribution and increased rolling moments, have not been thoroughly investigated. Addressing these gaps is critical for the safe integration of eVTOLs into urban airspace and for the development of robust control strategies to mitigate potential risks during landing.

This study aims to fill these gaps by employing the lattice Boltzmann method to evaluate the aerodynamic performance of a multi-rotor eVTOL during landing. Specifically, the influence of the ground effect, descent speed, and crosswinds on lift force distribution and stability are investigated. Additionally, a unique scenario where the eVTOL lands on a building rooftop, with half of its rotors positioned on the roof and the other half suspended in the air is explored. By analyzing these scenarios, this study aims to provide insights into the aerodynamic challenges associated with eVTOL landing and to identify potential risks, such as rollover, that may arise in urban environments.

2. Numerical Methodology

2.1. Lattice Boltzmann Method

The lattice Boltzmann method (LBM) is a mesoscopic numerical simulation technique rooted in kinetic theory and statistical mechanics. It describes fluid dynamics by discretizing the Boltzmann equation, avoiding the complexity of directly solving the Navier–Stokes equations. The core idea of the LBM is to recover macroscopic fluid behavior by simulating particle collision and propagation processes on a regular lattice.

The foundation of the LBM lies in the Boltzmann equation, which governs the evolution of the particle distribution function:

$${\displaystyle \frac{\partial f}{\partial t}}+v\cdot \nabla f=\Omega (f)$$

(1)

where $f(x,v,t)$ represents the distribution of particles at position x, velocity v, and time t, and $\Omega (f)$ is the collision operator.

In the LBM, the Boltzmann equation is discretized into finite velocity directions and spatial grids. Common discrete velocity models include D2Q9 (2D, 9 velocity directions) and D3Q19 (3D, 19 velocity directions). For example, in D2Q9, the discrete velocities $e_i$ are defined as follows:

$$e_i=\left\{\begin{array}{ll}(0,0)\hfill & i=0\hfill \\ (\cos {\theta }_i,\sin {\theta }_i)c\hfill & i=1,\dots ,4\hfill \\ \sqrt{2}(\cos {\theta }_i,\sin {\theta }_i)c\hfill & i=5,\dots ,8\hfill \end{array}\right.$$

(2)

where c = Δx/Δt is the lattice speed, and θ_i is the direction angle.

The LBM solution process consists of two steps: the collision process and the propagation process.

• Collision Process:

The distribution function $f_i(x,t)$ is updated using the collision operator:

$$f_i^{\mathrm{post}}(x,t)=f_i(x,t)+{\Omega }_i(f)$$

(3)

The BGK (Bhatnagar–Gross–Krook) collision model is commonly used:

$${\Omega }_i(f)=-{\displaystyle \frac{1}{\tau }}\left(f_i-f_i^{\mathrm{eq}}\right)$$

(4)

where τ is the relaxation time, and $f_i^{\mathrm{eq}}$ is the equilibrium distribution function:

$$f_i^{\mathrm{eq}}=w_i\rho \left(1+{\displaystyle \frac{e_i\cdot u}{c_s^2}}+{\displaystyle \frac{{(e_i\cdot u)}^2}{2c_s^4}}-{\displaystyle \frac{u^2}{2c_s^2}}\right)$$

(5)

where $w_i$ is the weighting factor, ρ is the density, u is the macroscopic velocity, and c_s is the speed of sound.

b.

Propagation Process:

The post-collision distribution function propagates to neighboring lattice points:

$$f_i(x+e_i\Delta t,t+\Delta t)=f_i^{\mathrm{post}}(x,t)$$

(6)

Macroscopic quantities are obtained by statistical averaging of the distribution functions:

$$\rho ={\displaystyle \sum _i{f}_i},\rho u={\displaystyle \sum _i{e}_i}{f}_i$$

(7)

2.2. Turbulent Model

The wall-adapting local eddy-viscosity (WALE) model was selected as the turbulent model. It performs well both near and far from walls and is applicable to both laminar and turbulent flows. Also, it accurately reproduces the asymptotic behavior of turbulent boundary layers when they can be directly resolved and avoids introducing artificial turbulent viscosity in shear regions outside the wake. The WALE model is formulated as follows:

$${\nu }_{\mathrm{turbulent}}={\Delta }^2{\displaystyle \frac{{(G_{\alpha \beta }^d{G}_{\alpha \beta }^d)}^{3/2}}{{(S_{\alpha \beta }{S}_{\alpha \beta })}^{5/2}+{(G_{\alpha \beta }^d{G}_{\alpha \beta }^d)}^{5/4}}}$$

(8)

$$S_{\alpha \beta }={\displaystyle \frac{1}{2}}\left({\displaystyle \frac{\partial v_{\alpha }}{\partial r_{\beta }}}+{\displaystyle \frac{\partial v_{\beta }}{\partial r_{\alpha }}}\right)$$

(9)

$$G_{\alpha \beta }^d={\displaystyle \frac{1}{2}}\left(g_{\alpha \beta }^2+g_{\beta \alpha }^2\right)-{\displaystyle \frac{1}{3}}{\delta }_{\alpha \beta }{g}_{\gamma \gamma }^2$$

(10)

$$g_{\alpha \beta }={\displaystyle \frac{\partial v_{\alpha }}{\partial r_{\beta }}}$$

(11)

$$\Delta =C_w{\mathrm{Vol}}^{1/3}$$

(12)

where Δ is the filter scale, S_σβ is the strain rate tensor of the resolved scale and the Smagorinsky constant (Cs) usually has a value between 0.1 and 0.2; by default, Cs = 0.12. The WALE constant (Cw) is typically 0.2.

2.3. Wall Treatment

To reduce computational cost near walls, XFlow employs a wall function to estimate velocity and temperature at the node closest to the wall. This wall function incorporates a generalized law of the wall, accounting for both adverse and favorable pressure gradients to model the boundary layer. It is valid across all y+ ranges and accurately captures turbulent boundary layer behavior, including the effects of curvature and pressure gradients. The velocity estimation formulation is based on the work of Shih et al. [27] and is given as follows:

$$\begin{array}{l}{{\displaystyle \frac{U}{u}}}_c={\displaystyle \frac{U_1+U_2}{u_c}}={\displaystyle \frac{u_{\tau }}{u_c}}{\displaystyle \frac{U_1}{u_{\tau }}}+{\displaystyle \frac{u_p}{u_c}}{\displaystyle \frac{U_2}{u_p}}\\ ={\displaystyle \frac{{\tau }_w}{\rho u_{\tau }^2}}{\displaystyle \frac{u_{\tau }}{u_c}}{f}_1\left(y^{+}{\displaystyle \frac{u_{\tau }}{u_c}}\right)+{\displaystyle \frac{\mathrm{d}{p}_w/\mathrm{d}x}{|\mathrm{d}{p}_w/\mathrm{d}x|}}{\displaystyle \frac{u_p}{u_c}}{f}_2\left(y^{+}{\displaystyle \frac{u_p}{u_c}}\right)\end{array}$$

(13)

$$y^{+}={\displaystyle \frac{u_{c}y}{\nu }}$$

(14)

$$u_c=u_{\tau }+u_p$$

(15)

$$u_{\tau }=\sqrt{{\displaystyle \frac{\left|{\tau }_w\right|}{\rho }}}$$

(16)

$$u_p={\left({\displaystyle \frac{\nu }{\rho }}\left|{\displaystyle \frac{\mathrm{d}{p}_w}{\mathrm{d}x}}\right|\right)}^{1/3}$$

(17)

where $y^{+}$ is the normal distance from the wall, $u_{\tau }$ is the skin friction velocity, ${\tau }_w$ is the turbulent wall shear stress, $\mathrm{d}{p}_w/\mathrm{d}x$ is the wall pressure gradient, $u_p$ is a characteristic velocity of the adverse wall pressure gradient, and U is the mean velocity at a given distance from the wall.

2.4. Setting

The numerical simulations were conducted using Xflow 2020x [28], a computational fluid dynamics (CFD) software based on the LBM, which has demonstrated its robustness and accuracy in different benchmarks [29]. Unlike traditional CFD approaches, which often involve laborious and time-consuming meshing operations, XFlow eliminates the need for extensive mesh generation by requiring only a closed geometry as input. For this study, a closed-surface geometry of the eVTOL (electric vertical take-off and landing) aircraft was specifically prepared for CFD analysis, as shown in Figure 1. Notably, each propeller in the model was treated as an independent component, capable of rotating around its axis at a predefined rotational speed, allowing for an accurate representation of the aerodynamic interactions during operation. The planar positions of the rotor axes are tabulated in

Table 1. Planar positions of rotor axes.

Axis ID	X (m)	Z (m)
1	−0.648	1.564
2	−1.564	0.648
3	−1.564	−0.648
4	−0.648	−1.564
5	0.648	−1.564
6	1.564	−0.648
7	1.564	0.648
8	0.648	1.564

. The T-motor U15L KV43 motors are equipped with 47-inch carbon fiber composite blades [30].

The computational domain for the flow field is defined as a cuboid with dimensions of 20 m × 20 m × 20 m. The surfaces of the eVTOL (including the fuselage and propellers) and the ground (for cases involving ground effect) are modeled as non-slip walls. Boundary conditions are applied to the sides of the domain: a pressure outlet is used for typical cases, while a velocity inlet is employed for scenarios involving crosswind.

To investigate the influence of the ground effect in the discussion of the landing process, a simulation of hovering at a height of 0.7 m above the ground was first conducted (Case 1). Using Case 1 as the baseline, simulations were performed for landing speeds of 1.5 m/s (Case 2) and 3 m/s (Case 4), respectively. Also, the impact of crosswinds on the eVTOL at different landing speeds was investigated (Case 3 and 5).

Inspired by the V-22 “Osprey” tiltrotor aircraft, which experienced a rollover crash during shipboard landing under similar conditions due to a sudden and significant imbalance in lift, cases of landing on a roof were conducted to assess whether the eVTOL also faces the risk of rollover accidents under such conditions. In these scenarios, half of the eVTOL’s rotors were positioned on the roof, while the other half were suspended in the air, as shown in Figure 2. The roof is a square with a side length of 30 m, and the building height is 21.1 m. The computational domain is 90 m (length) × 50 m (width) × 80 m (height). Four cases (Case 6–9) were performed to investigate the effects of descent speed and crosswind speed. Detailed information for each case is provided in Table 5.

The Courant number was set as a constant of 0.1. The time step was dynamically adjusted based on the Courant number to ensure numerical stability. Default air properties, as listed in

Table 2. Air properties.

Gas Properties	Value
Molecular weight	29
Density	1.2 kg/m3
Temperature	288 K
Viscosity coefficient	1.8 × 10−5 Pa·s
Thermal conductivity	0.024 W/(m·K)
Specific heat capacity	1006.4 J/(kg·K)

, were utilized throughout the simulations.

2.5. Mesh

To determine the appropriate mesh size for the blades, a mesh independence study was conducted using a single rotor of T-motor U15L at an rpm of 2400 (14,400 deg/s) [30]. Figure 3 illustrates the relationship between lift force and mesh density. Initially, the target resolved scale of the blade was set to 0.1. After refining the mesh and doubling its density, the target resolved scale was reduced to 0.05. Further refinement to a mesh density of 4 (corresponding to a target resolved scale of 0.025) resulted in a lift force of approximately 300 N, which is slightly less than the test data. Additional refinements yielded no significant changes in lift force, confirming mesh convergence. The results indicate that the acceptable range for the target resolved scale is between 0.003125 and 0.025.

The lattice structure within the domain was automatically generated by XFlow, with the maximum lattice size set to 1 m in regions far from the eVTOL. Near the eVTOL, the lattice sizes are refined to 15.6 mm for the fuselage and 3.9 mm for the blades. The vortex resolution was set to 15.6 mm. Initially, the domain comprised 9 levels and approximately 4 million elements, as illustrated in Figure 4a. Through the application of adaptive grid refinement technology, the number of grid elements increased to 82 million after 1 s of simulation, as shown in Figure 4b.

2.6. Validation

To assess the accuracy of the LBM in predicting aerodynamic performance, the lift forces obtained from CFD simulations were compared with flight test data. The hover test was conducted on a flat concrete surface, with the eVTOL’s takeoff mass of 374 kg and a hover height of approximately 1 m (Figure 5a). The flight test was carried out at Nanhu Paddy Park (Figure 5b). The aircraft’s unloaded takeoff mass is 374 kg and its moment of inertia along the z-axis is 184 kg·m².

The flight altitude was 20 m, with a flight speed of 6 m/s. The flight performance was stable, and the aircraft landed after completing 2.5 laps, with a total flight duration of 16 min and 30 s, covering a distance of 4.2 km (Figure 6). During the flight, the aircraft attitude (speed and pitch angle) and the throttle input of each rotor were recorded from the flight control system. The values during two stable level-flight phases were averaged, as shown in

Table 3. Configurations of validation cases.

Validation Case	Velocity (m/s)	Pitch Angle (deg)	Rotational Speed (deg/s)
			Axis 1 and 8	Axis 2 and 7	Axis 3 and 6	Axis 4 and 5
1	0	0	14,364	14,364	14,364	14,364
2	4	−1	13,068	13,716	14,364	14,724
3	6	−2.5	11,664	13,068	14,401	15,300

. The rotational speeds were converted from the throttle input.

The results are summarized in

Table 4. Errors of lift for verification cases.

Validation Case	Lift Force (N)	Weight (N)	Error (%)
1	3689.1	3668.9	0.6
2	3465.8		5.5
3	3316.7		9.6

, which presents the percentage errors in the simulated lift forces relative to the experimental values. The analysis reveals that the error in the simulation-predicted lift force increases with flight speed. Specifically, in the hovering condition, the error is minimal, demonstrating the method’s capability to accurately capture the aerodynamic behavior in a static state. For forward flight at 4 m/s, the error remains relatively low, indicating good agreement with experimental data. At 6 m/s, the error increases but stays below 10%, which is within an acceptable range for engineering applications. These findings confirm that the LBM-based simulations provide sufficient accuracy for studying the aerodynamic characteristics of the eVTOL aircraft, particularly in the context of landing performance and multi-rotor interactions. The consistent and relatively small errors across the tested speed range validate the robustness of the LBM approach for this application.

3. Results and Discussion

The data presented in

Table 5. Configurations and lift of landing cases.

Case	Landing Area	Descent Speed (m/s)	Crosswind Speed (m/s)	Lift Force—Right Side (N)	Lift Force—Left Side (N)	Lift Force (N)	Variation (%)	Rolling Moment—Right Side (N·m)	Rolling Moment—Left Side (N·m)	Rolling Moment (N·m)
1	Ground	0	0	-	-	4073.7	-	-	-	-
2		1.5	0	-	-	4112.1	0.9	-	-	-
3		1.5	10	-	-	4224.7	3.7	-	-	-
4		3	0	-	-	4297.3	5.5	-	-	-
5		3	10	2138.1	1834.7	4348.0	6.7	−2420.8	2040.1	−380.6
6	Roof	0	0	1972.1	1987.8	3779.6	−7.2	−2179.6	2201.0	21.4
7		0	10	2233.2	1498.9	3754.0	−7.8	−2492.3	1745.1	−747.2
8		3	0	2037.5	2047.6	4089.2	0.4	−2254.3	2263.7	9.4
9		3	10	2729.1	1695.1	4284.6	5.2	−3083.4	1949.4	−1134.0

reveal a clear relationship between lift force, descent speed, and crosswind speed for the eVTOL aircraft during landing. The velocity contours for Case 5, 7, 8, and 9 are presented in Figure 7 and Figure 8, while the vorticity contours are presented in Figure 9. From Case 1, it is evident that the lift force increases by approximately 13% compared to that in free air due to the ground effect. As the descent speed increases from 0 m/s (hovering) to 1.5 m/s and further to 3 m/s, the lift force exhibits a consistent upward trend. Specifically, at a 1.5 m/s descent speed (Case 2), the lift force increases by 0.9% compared to the hovering condition (Case 1). At a 3 m/s descent speed (Case 4), the lift force increases by 5.5% compared to Case 1. This suggests that higher descent speeds contribute to a moderate increase in lift force, likely due to enhanced aerodynamic interactions and dynamic pressure effects. The presence of a crosswind further amplifies the lift force. For example, at a 1.5 m/s descent speed, the introduction of a 10 m/s crosswind (Case 3) results in a 3.7% increase in lift force compared to the no-crosswind condition (Case 2). At a 3 m/s descent speed, the addition of a 10 m/s crosswind (Case 5) leads to a 6.7% increase in lift force compared to the no-crosswind condition and a rolling moment of 380.6 N·m.

When the eVTOL hovers close to the ground, it can be observed that the rotor downwash strikes the surface and spreads outward, generating an upward recirculating annular airflow around the drone’s perimeter—commonly referred to as the fountain effect. The data demonstrate that both descent speed and crosswind speed positively correlate with lift force. However, the combined effect of higher descent speed and crosswind results in the most significant increase in lift force. This indicates that crosswinds significantly influence the aerodynamic performance of the eVTOL, likely due to changes in flow dynamics and increased effective airspeed over the rotors.

Unique scenarios involve landing on a building’s roof, where half of the rotors are positioned on the roof and the other half are suspended in the air. During hover at 3 m above the roof (Case 6), the results indicate that the eVTOL aircraft experiences a 7.2% lift loss while generating negligible roll moment. This observation suggests that the asymmetry of ground effect between the rotors has a limited influence on the aircraft’s rolling tendency. With the crosswind condition (Case 7), the lift of the windward rotor increases significantly, while the lift of the rotor positioned above the rooftop (on the leeward side) decreases sharply. Consequently, although the total lift is only slightly reduced, a substantial lateral rolling moment of 747.2 N·m is generated. The combined effect of lift reduction and lateral tilting renders the hovering eVTOL highly susceptible to crash conditions. During landing with a 3 m/s descent speed scenario (Case 8), the lift force increases by approximately 7% compared to Case 7, whereas the roll moment decreases to become negligible. This confirms the conclusion from Case 6, namely that the presence or absence of the ground effect on the two sides of the eVTOL’s rotors does not result in a significant rolling moment. The wake on one side descends along the building wall, while on the other side it extends outward across the rooftop. This flow pattern—where the wake is re-ingested by the rotors as part of the ground effect—leads to high airflow velocity and low pressure above the rooftop, thereby diminishing the lift enhancement typically provided by the ground effect.

Finally, in Case 9, the roll moment substantially escalates to 1134 N·m and the rotors on the left side of the aircraft do not exhibit an increased lift due to the ground effect from the roof below. Instead, the combined effects of the crosswind and the roof-induced flow circulation reduce the lift on this side, resulting in a significant increase in the rolling moment from 380.6 N·m (Case 5) to 1134.0 N·m (Case 9). The windward-side wake vortices are more pronounced, indicating stronger aerodynamic forces, primarily due to increased flow velocity across the rotor blade sections. The interaction between the crosswind and the building sidewall raises the pressure beneath the outer rotor, inhibiting the downward development of its wake and deflecting it toward the leeward rotor. As a result, the leeward rotor experiences increased inflow and higher downward flow velocity, which leads to a significant reduction in its aerodynamic force. This unique phenomenon highlights the potential risk of rollover when operating under crosswind conditions with partial rotor coverage over a rooftop, emphasizing the need for careful aerodynamic evaluation in such scenarios.

4. Conclusions

This study evaluated the aerodynamic performance of a multi-rotor eVTOL aircraft during landing using the lattice Boltzmann method (LBM). Simulations were conducted for multiple hovering scenarios, including cases above the ground and with partial rotor coverage over a building roof. The key conclusions are as follows:

• The ground effect significantly increases lift force by approximately 13% compared to free-air conditions.
• Higher descent speeds and the presence of crosswinds further amplifies lift forces, with the combined effect of a 3 m/s descent speed and 10 m/s crosswind resulting in a 6.7% increase in lift.
• Asymmetric lift distribution, due to the interaction between crosswinds and roof-induced flow circulation, leads to a substantial decrease in the lift on the rotors above the roof and an increase in rolling moment.

These findings highlight the critical influence of descent speed, crosswinds, and landing surface geometry on eVTOL aerodynamic behavior, particularly the risk of rollover during rooftop landings. The results underscore the importance of advanced aerodynamic modeling for ensuring the safety and stability of eVTOL operations in complex environments. Future work should explore additional factors, such as turbulence intensity and rotor–rotor interactions, to further optimize eVTOL landing performance.