LoS/NLoS modeling in 5G planning requires accurate 3D building vectors and vegetation data, which allow the propagation model to trace the geometric path between a base station antenna and a receiver location and determine whether any obstacle interrupts that path. Without 3D data, LoS/NLoS classification defaults to statistical probabilities per land-use class, an approximation that introduces 10 to 30 dB discrete prediction errors at 3.5 GHz and above.
In 4G planning at 700 to 2100 MHz, LoS and NLoS links experience different path loss, but the difference is manageable within the link budget. Empirical propagation models like COST-231 Hata account for urban NLoS statistically.
In 5G at 3.5 GHz and especially 26 GHz mmWave, the difference between LoS and NLoS is not a statistical correction. It is a fundamental change in propagation regime:
This means that whether a specific receiver location is in LoS or NLoS is the single most important factor in its link budget. Getting the LoS/NLoS classification wrong by even a few meters of terrain or building height introduces a discrete planning error that cannot be averaged away.
Statistical models assign a LoS probability to each location based on environment type. The 3GPP specifications define LoS probability formulas for different scenarios:
These models are used when 3D geometric data is unavailable or when modeling a statistically large area without per-location precision. They are appropriate for initial feasibility studies, sub-1 GHz macro planning, and national coverage estimates.
Limitation: Statistical models cannot determine whether a specific receiver location is in LoS or NLoS, only what the probability is for a location at that distance in that environment type. This is insufficient for FWA premise qualification or precise small cell coverage prediction.
Deterministic classification computes LoS status for every specific transmitter-receiver link using 3D geometric data:
This approach requires accurate 3D building vectors (LOD1 or LOD2) with building heights, 3D vegetation data with individual tree positions and heights, a DTM for terrain-related intersections, and a propagation model capable of deterministic ray calculation (ray-tracing or Knife-Edge Diffraction).
A clear LoS link in RF planning is not simply the absence of complete blockage. For a link to be considered in full LoS, the First Fresnel Zone must be at least 60% clear of obstacles.
Fresnel Zone radius at the midpoint of a path of length d (meters) at frequency f (GHz):
r₁ = 8.66 × √(d / 4f) (simplified formula for the midpoint case)
This means that at 3.5 GHz, an obstacle that does not physically intersect the direct path but enters the Fresnel zone still causes diffraction loss. At 26 GHz, the Fresnel zone is smaller, but so is the margin before obstruction becomes total blockage.
Propagation models account for Fresnel zone clearance differently:
Atoll uses 3D building vectors or DSM data for automatic LoS/NLoS computation per prediction pixel. With 3D buildings loaded and an appropriate propagation model selected (standard propagation model, ITU models, or partner models like Volcano), Atoll can perform Knife-Edge Diffraction calculations over each building. The building height attribute in the input shapefile directly drives the diffraction loss computation.
The AIM (Advanced Intelligent Model) in InfoVista Planet explicitly models building diffraction, wall reflections, and street canyon effects using 3D building data. AIM was designed with urban 5G planning in mind and uses 3D geometry as a first-class input, not an optional add-on.
Full ray-tracing propagation models trace multiple ray paths between transmitter and receiver, including direct LoS ray, reflected rays (off building facades), and diffracted rays (around building edges). These models provide the highest LoS/NLoS accuracy but require the highest-quality 3D input data and the greatest computational resources.
| Input Layer | Required Accuracy | Why It Matters |
|---|---|---|
| 3D Building Vectors | Height RMSE ≤2 m; capture rate ≥93% | Buildings are the primary blockers |
| DTM (bare earth) | Vertical RMSE ≤2 m at 5 m resolution | Terrain below building heights must be accurate |
| 3D Vegetation | Individual tree heights ≤2 m accuracy | Trees at 26 GHz cause 35.3 dB path loss |
| Antenna height | Precise (within 0.5 m) | Small antenna height errors shift LoS boundary |
Building capture rate matters as much as height accuracy. A building absent from the dataset is treated as free space, creating a false LoS classification. At a 93%+ capture rate, the remaining approximately 7% of missing buildings occasionally cause this error. For mmWave FWA applications where per-premise qualification is required, capture rates ≥95% are preferable.
Vegetation cannot be treated as binary block or pass at 5G frequencies. Instead:
Modeling vegetation correctly requires individual tree polygon data with canopy heights, not just a binary forest classification from a 2D clutter map.
LuxCarta provides the 3D building and vegetation datasets required for deterministic LoS/NLoS classification in 5G planning. Building data achieves 93%+ capture rate with 87.67% precision and 88.44% recall, benchmarked by customers including Telefónica Deutschland for 26 GHz FWA planning. Vegetation data includes individual tree polygons with total height, trunk height, and canopy height, enabling accurate computation of vegetation attenuation at both 3.5 GHz and 26 GHz.
LuxCarta's research on vegetation attenuation (published with measured data from Barcelona) established the 12.8% accuracy improvement at 3.6 GHz and the 35.3 dB single-tree loss figure at 26 GHz, statistics now used by RF planners to quantify the value of adding 3D vegetation to their propagation models.
All LuxCarta building and vegetation products are delivered in formats directly compatible with the LoS/NLoS computation engines of Forsk Atoll, InfoVista Planet (AIM), and TEOCO Asset, without requiring data transformation.
LoS probability (from 3GPP TR 38.901 or similar) gives the statistical likelihood that a link at a given distance and environment type is in LoS. Deterministic LoS classification computes the actual LoS/NLoS status of each specific link using 3D geometric data. Deterministic classification is always more accurate where 3D data is available, particularly important for FWA qualification and small cell planning.
Massive MIMO beamforming uses channel state information (CSI) to steer signal energy toward the receiver. In NLoS conditions, the dominant path is typically a diffracted or reflected ray, not the geometric direct path. The beam must be steered toward the strongest arriving path direction, not the direct geometric direction. 3D building data helps identify candidate reflected and diffracted path directions during the planning phase.
Indoor 5G planning (for private networks or indoor coverage extension) models wall penetration, floor-to-floor penetration, and internal LoS/NLoS separately from outdoor planning. Tools like iBwave use indoor 3D geometry (floor plans, wall materials) for indoor ray-tracing. Outdoor-to-indoor modeling requires both outdoor 3D building data and indoor building construction data.
Outdoor-to-indoor propagation adds building entry loss to the outdoor path loss. Entry loss depends on building wall construction (glass, concrete, brick) and estimated wall thickness, which can be approximated from LOD2 building geometry. For precise outdoor-to-indoor modeling, combining outdoor 3D building data with indoor floor plan data and material properties gives the most accurate results.
For fixed users (FWA subscribers at a defined premises location), LoS/NLoS is computed once per premises location, a static binary determination. For mobile users, LoS/NLoS status changes as the user moves. The propagation model must handle the continuous transition between LoS and NLoS as the user's position changes relative to buildings and vegetation. Statistical LoS probability models are typically used for mobile user simulation, while deterministic classification applies to fixed subscriber qualification.
LuxCarta provides AI-powered 3D geospatial data solutions for telecom, simulation, and smart city applications worldwide. Learn more at luxcarta.com or explore on-demand extraction at BrightEarth.