Millimeter-wave deployment over short distances is a cornerstone of next-generation wireless communication systems. However, the applicability of existing models remains limited in tropical and heterogeneous urban environments. This underscores the need for localized model adaptation to ensure accurate signal estimation and efficient network design in emerging 5G markets. Millimeter-wave deployment over short distances is a key to enabling technology for next generation wireless communication systems. However, the accuracy and applicability of existing models are often limited when applied to environments with differing weather conditions and geographical features from those for which they were originally designed. Asma and Rafiqul [2] analyzed path loss for accurate signal estimation in Malaysia, focusing on outdoor microcellular environments at 38 GHz over a 300 meter path length. Their study examined the impact of rain attenuation on path loss, path loss exponent (PLE), and shadow fading (SF). Two channel models were used for simulation: the NYUSIM statistical spatial channel model (version 2), developed by New York University in 2019, and the 3GPP TR 38.900 Release 14 model. Conducted in a temperate climate, the measurement campaign highlighted the need for model adaptation in tropical regions. The underestimation observed in NYUSIM simulations was attributed to discrepancies in attenuation factors (AF) including pressure, humidity, temperature, and rain rate between simulated and measured data. The results indicated that the CI model within NYU Sub-Terahertz and Millimeter-Wave Channel Simulator provided more accurate path loss estimations compared to the 3GPP model, affirming its suitability for outdoor environments.

Agbotiname and Tolulope [3] conducted a comparative analysis of predicted and measured path loss over a lagoon environment. Propagation measurements were performed at 1800 MHz between May and August 2017 using Huawei Technologies drive test equipment. The measured path loss values were compared against predictions from several models, including Free Space, Log Distance, Two-Ray, COST 231 Hata, and Stanford University Interim (SUI). Among these, the COST-231 Hata model demonstrated the highest accuracy, yielding root mean square error (RMSE) values of 10.03 dB, 12.38 dB, 17.59 dB, and 7.67 dB for the initial measurement and subsequent monthly intervals. Further optimization of the COST-231 Hata model using the least squares algorithm resulted in improved prediction accuracy, with RMSE values reduced to 7.90 dB, 9.28 dB, 14.82 dB, and 5.28 dB, respectively. The optimized model achieved an average RMSE of 9.32 dB, representing a 21.81 % improvement over the original model’s average RMSE of 11.92 dB. These findings suggest that the optimized COST231-Hata model is suitable for characterizing radio channels in lagoon environments.

Accurate path loss prediction is a fundamental aspect of wireless communication network planning. Existing models such as Log-Normal, Okumura-Hata, and COST-231 often require adaptation when applied to different frequency bands. Adeyemo et al., [4] addressed this by modifying the Log-Normal model for High Speed Packet Access (HSPA) using an Adaptive NeuroFuzzy Inference System (ANFIS). Measurements of Received Signal Strength (RSS) were conducted via drive tests in the Ayetoro area of Lagos, Nigeria (Longitude 3.19647 °E, Latitude 6.59167 °N), using a system integrated with Testing and Evaluations Mobile System software (TEMS), Ericsson TEMS phones, and GPS. The performance of the developed models was evaluated using path loss values and RMSE metrics. Results indicated that the ANFIS, COST 231, and modified LogNormal models produced the lowest RMSE values and closely matched the measured data, making them suitable for HSPA signal prediction and future wireless network planning. However, the study was limited to RSS metrics and 4G LTE networks.

Methodology

Data Collection

Drive tests were conducted in each city using Testing and Evaluations Mobile System (TEMS) Investigation software, spectrum analyzers, Geographical Positioning System enabled receivers, and Digital elevation models (DEMs), that build footprint/land use layers, were used to characterize clutter and confirm site geometry. The study focused on the 3.5 GHz and 3.6 GHz frequency bands, commonly used in 5G deployments. Parameters recorded include: Reference Signal Strength Indicator (RSSI), Reference Signal Received Power (RSRP), Reference Signal Received Quality (RSRQ), Signal to Noise Ratio (SINR), distance, Building density and Terrain elevation.

Modeling Tools

OriginPro was employed for curve fitting, statistical analysis, and visualization. The following models were considered: Developed Lagos, Ibadan, Abuja and Kano urban environment models, okumura-Hata model, COST-235, ITU-R P-355 and Egli models

Model Formulation

– General path loss equation:
Pl(d) =Pl d_o+ 10n log(d) + Xσ 2.1

The city-specific model on two different service providers

The developed environmental model for Kano metropolis is
Kano (dB)=19.18+10nlogD+Ɣσ 2.2
The developed environmental model for Abuja Airtel metropolis is
Abuja (Airtel)(dB)=19.51+10nlogD+σ 2.3
The developed environmental model for Abuja MTN metropolis is
Abuja (MTN)(dB)=19.77+10nlogD+σ 2.4
The developed environmental model for Ibadan Airtel metropolis is
Ibadan (MTN)(dB)=30.37+10nlogD+σ 2.5
The environmentally developed model for Lagos Airtel metropolis is
Lagos(Airtel)(dB)=30.06+10nlogD+σ 2.6
The developed environmental model for Lagos MTN metropolis is
Lagos(MTN)amt (dB)=28.55+10nlogD+σ 2.7
Results and Discussion

1.1 Kano Metropolis Drive-Test Measurement Results
Collectively, these figures demonstrate that MTN’s 5G deployment in Kano achieves predominantly moderate-to-strong RSRP, high RSRQ and SINR across most surveyed areas, with localized pockets of degraded signal quality and throughput that warrant targeted capacity or coverage interventions as agreed by [5].

1.2 Comparison of Developed Model and Existing Models against Distance

Figure 1.5 compares measured path loss from drive tests in Kano with predictions from five propagation models: the Kano 5G MTN model, Okumura Hata, COST 235, Egli, and ITU R P.8339. The Kano 5G MTN model achieved by far the lowest RMSE (22.81 dB), indicating the best agreement with measurements. In contrast, Okumura Hata (235.09 dB), COST 235 (187.87 dB), Egli (214.87 dB), and ITU R P.8339 (435.93 dB) produced substantially larger errors, with ITU R P.8339 deviating most from empirical data. The superior performance of the Kano 5G MTN model reflects its calibration for high frequency 5G conditions and local urban characteristics; its path loss predictions closely track the measured values across the surveyed distances. All models reproduce the expected monotonic increase of path loss with transmitter–receiver separation, consistent with free space loss behaviour, but the locally tuned model is the only one that maintains close alignment over the measurement range. The remaining models’ higher RMSEs are attributed to their original design assumptions (lower frequency bands or different urban morphologies) and the combined effects of Kano’s low density foliage, low rise building stock, and high ambient temperatures on high frequency propagation.

1.3 Abuja Airtel Metropolis Drive-Test Measurement Results

1.4 Comparison of Developed Model and Existing Models against Distance

Model comparison across distance Figure 1.10 contrasts measured path loss with predictions from the Abuja 5G (Airtel) model, Okumura Hata, COST 235, Egli, and ITU R P.8339. The locally developed Abuja 5G model attains the lowest RMSE (21.69 dB), demonstrating superior fit and consistent alignment with empirical data across the measurement range. COST 235 shows the next best agreement, while Okumura Hata, Egli, and ITU R P.8339 produce substantially larger errors (RMSEs 158–299 dB), reflecting their limited suitability for high frequency urban 5G channels. All models reproduce the expected increase in path loss with distance, but only the calibrated Abuja model maintains close correspondence with measured values, underscoring the value of local calibration for accurate 5G propagation prediction.

1.5 Abuja MTN Metropolis Drive-Test Measurement Results

Figure 1.11 depicts RSRQ values occupy five intervals accounting for 0.00%, 64.05%, 33.40%, 2.15%, and 0.40% of points. No locations achieved the top-quality interval; most measurements (64.05%) indicate low negative RSRQ consistent with low interference and good link quality, while 0.40% suffer substantial degradation.

Figure 1.12 presents SINR is distributed across six bins: 56.34%, 30.22%, 9.20%, 3.28%, 0.68%, and 0.28%. Over half the area (56.34%) experiences high quality radio channels with low noise and interference; smaller fractions occupy progressively lower SINR bands indicating moderate to poor channel conditions as agreed by [7].

Figure 1.13 summarizes downlink throughput (Mbps) from drive test calls. Throughput samples fall into five categories comprising 7.80%, 16.03%, 41.81%, 23.12%, and 11.25%. Only 7.80% of samples reach the highest throughput tier, the largest share (41.81%) demonstrates intermediate performance, and 11.25% reflects the lowest data rates.

Collectively, these distributions indicate that MTN’s 5G deployment in Abuja provides predominantly moderate-to-high signal strength and quality across most surveyed locations, with localized pockets of degraded RSRQ, SINR, and throughput that merit targeted capacity or coverage interventions as shown in Figure 1.14.

1.6 Comparison of Developed Model and Existing Models against Distance

Figure 1.15 compares measured path loss against predictions from five models: the locally developed Abuja (MTN) 5G model, Okumura Hata, COST 235, Egli, and ITU R P.8339. The Abuja (MTN) model attains the lowest RMSE (20.19 dB), indicating the best agreement with drive test measurements; the other models exhibit substantially larger errors (RMSEs: Okumura Hata 219.46 dB, COST 235 158.56 dB, Egli 218.23 dB, ITU R P.8339 299.41 dB). All models reproduce the expected monotonic increase of path loss with transmitter–receiver separation, but only the locally calibrated Abuja model consistently aligns closely with observations across the measured distance range. The markedly higher errors of the legacy models reflect their design assumptions for lower frequencies or different urban morphologies, whereas the developed Abuja (MTN) model optimized for high frequency 5G channels and local urban characteristics provides superior predictive accuracy for coverage planning in Abuja as in line with [8].

1.7 Radio performance trends Ibadan drive tests

Figure 1.16 shows RSRP measurements across MTN’s Ibadan 5G coverage were binned into six intervals representing 0.29%, 47.09%, 28.69%, 13.22%, 9.39%, and 1.32% of samples. Only 0.29% of locations show very strong RSRP, while 1.32% record the weakest reception; the majority of points exhibit moderate to good signal strength.

Collectively, the Ibadan measurements indicate generally favorable signal quality and throughput in much of the surveyed area, with localized pockets of degraded RSRQ, SINR, and throughput that warrant targeted coverage or capacity interventions.

1.8 Comparison of Developed Model and Existing Models against Distance

The study compared five propagation models against measured distance data: a site-specific Ibadan MTN 5G model, Okumura Hata, COST 235, Egli, and ITU R P.8339. The mean root means square error (RMSE) values were Ibadan MTN 5G 34.69 dB; ITU R P.8339 127.90 dB; COST 235 167.10 dB; Okumura Hata 228.30 dB; Egli 325.94 dB. The Ibadan MTN 5G model exhibits substantially lower RMSE and closer agreement with measurements, demonstrating superior fidelity for high frequency 5G propagation in the local metropolitan environment. The relative performance of ITU R P.8339 suggests that attenuation terms addressing vegetation are relevant for Ibadan. Large RMSE differences among models underscore the necessity of localized calibration to capture site specific effects such as dense foliage, high rise structures, varied terrain, and high population density. Model optimization further reduced RMSE, confirming improved predictive accuracy after local tuning as agreed with [9] Figure 1.20.

1.9 Airtel 5G drive test distributions Lagos

Measured RSRP, RSRQ, and SINR decrease with increasing distance from base stations, consistent with free space loss combined with urban morphology and terrain effects. Threshold compliance was poor: only 0.29% of drive test locations met the RSRP threshold, 0.00% met the RSRQ threshold, and 58.74% met the SINR threshold. These outcomes indicate significant degradation from buildings, vegetation, topographic irregularities, and insufficient tower density. Recommended mitigations include network densification (additional sites or small cells), installation of signal boosters/repeaters, and adjustment of antenna count, height and orientation.

RSRP across six percentile groups was distributed as 0.10%, 0.35%, 12.00%, 30.81%, 52.85%, and 3.89%, indicating that most surveyed locations experienced moderate signal strength while a very small portion exhibited very weak (0.10%) or very strong (3.89%) RSRP. RSRQ percentages were 3.06%, 14.00%, 66.39%, 4.20%, and 2.05%, with the majority (66.39%) in a moderate RSRQ range and 2.05% in the poorest quality range likely attributable to high load or interference. SINR distribution (43.61%, 30.45%, 12.52%, 7.02%, 3.53%, 2.87%) shows that 43.61% of the area had high SINR while 2.87% suffered low SINR and high interference. Measured throughput classes (Mbit/s) fell into 2.29%, 17.81%, 22.11%, 25.40%, 19.11%, and 13.28% bands; only 2.29% of locations experienced peak data rates while 13.28% recorded the lowest rates. The results demonstrate that a site-specific propagation model materially improves path loss prediction in dense urban settings relative to generic models. Vegetation and built environment parameters significantly affect high frequency 5G propagation and should be included in model parameterization. For operational improvement, planners should pursue localized model calibration supported by iterative drive testing, targeted network densification (small cells and repeaters), and radio configuration optimization (antenna height, tilt, and load management) to enhance coverage, link quality, and data throughput as stated by [10] in wireless communication published book [Figure 1.21-1.24].

1.10 Comparison of Developed Model and Existing Models against Distance

Five models were evaluated by average RMSE: Lagos Airtel model 32.48 dB, COST 235 170.21 dB, Okumura Hata 230.48 dB, Egli 323.60 dB, ITU R P.8339 438.20 dB. The Lagos Airtel model exhibits the closest agreement with measurements and the lowest RMSE, reflecting its tailored parameterization for high frequency urban propagation. COST 235 and Okumura Hata produce moderate performance; Egli and ITU R P.8339 show large positive bias and poor fit, consistent with limited adaptability of those formulations to dense, high frequency urban environments. The aberrant behavior of ITU R P.8339 is attributed to its original design for lower frequencies, which yields exaggerated loss estimates at 5G frequencies. Path loss increases predictably with transmitter–receiver separation, corroborating free space loss theory under urban modification. This is inconsistent with free-space path-loss theory [11]. The Lagos 5G Airtel model’s predictions closely align with measured values, validating its effectiveness for local channel estimation. The dense urban landscape, characterized by high-rise buildings and Lagos’s humid climate, contributes to an average RMSE of 32.48 dB for Airtel, reinforcing the importance of environment-specific modeling [Figure 1.25].

1.11 Lagos 5G MTN Drive –Test Measurement Results

RSRP distribution across percentile groups indicates that only 3.89% of sampled locations met the RSRP threshold; the bulk of measurements fall into moderate signal ranges. RSRQ compliance was similarly low at 3.06%, indicating limited areas with low interference and high call quality. SINR was acceptable in 43.61% of locations, but 2.87% experienced low SINR and high interference. Throughput distribution shows only 2.29% of locations achieved peak data rates while 13.28% recorded the lowest rates, indicating constrained zones of high 5G capacity. The measured degradation of RSRP, RSRQ, and SINR with distance highlights the impact of urban obstructions, load, and network configuration on perceived performance.

RSRP distribution shows 3.61% of the area with very strong signal and 0.01% with very weak RSRP; the majority of samples are in intermediate ranges. RSRQ measurements indicate 24.58% of locations with favorable RSRQ and 13.99% in the poorest quality band; some RSRQ classes recorded no samples. SINR distribution reveals only 10.78% of locations with high SINR while 14.06% suffer low SINR and significant interference. Throughput bands indicate 2.56% of locations achieved the highest data rates while 6.90% observed the lowest rates. Coverage analysis for MTN reports that 3.61% met RSRP thresholds, 0.00% met RSRQ thresholds, and 10.78% achieved acceptable SINR, underscoring persistence of coverage and quality gaps despite isolated strong receptions (RSSI down to −67.18 dBm) that is line with [3] [Figure 1.26-1.28].

1.12 Comparison of Developed Model and Existing Models against Distance

Site specific models (Lagos Airtel and Lagos MTN) substantially outperform generic models, reducing RMSE and providing more reliable local channel estimation for 5G network planning. Environmental factors dense built form, vegetation, and humid climatic conditions drive elevated RMSE values even for tailored models, demonstrating the residual complexity of urban propagation at 5G frequencies. The disparity between model predictions and legacy model performance underscores the need to avoid direct application of low frequency models to high frequency deployments without recalibration. Measured radio metric shortfalls (low RSRP/RSRQ compliance and limited high throughput areas) point to operational issues: insufficient site density, suboptimal antenna configuration, interference and load imbalance. Adopt iterative, site-specific model calibration supported by regular drive testing to refine parameterization for high frequency urban channels. Increase network densification through small cells, repeaters, or boosters in identified low coverage corridors to reduce path loss and cell overshoot. Optimize antenna deployment (count, height, tilt) and apply interference mitigation and load balancing strategies to elevate RSRQ and SINR across the coverage area. Prioritize targeted interventions in zones showing the lowest throughput and SINR to maximize user perceived performance and spectral efficiency. Such measures are essential for managing cell overshoot and reducing network congestion, in alignment with recommendations by [12] [Figure 1.29].

Conclusion

This study shows that 5G path loss behavior in Nigerian urban environments is highly site dependent, with variations driven by local architecture, terrain and environmental conditions. Site specific models developed for Kano, Abuja, Ibadan and Lagos consistently outperform existing propagation models after local calibration, demonstrating the value of environment tailored parameterization for high frequency urban deployments. Measured coverage and quality indicators vary markedly by city: Kano: RSRP/RSRQ/SINR compliance of 0.97%, 81.37%, 72.40%. Abuja: Airtel 0.25% 0.00% 78.22%; MTN 0.47% 0.00% 56.34%. Ibadan: 0.29% 0.00% 58.74%. Lagos: Airtel 3.89% 3.06% 43.61%; MTN 3.61% 0.00% 10.78%. Optimized model accuracy (average RMSE) highlights the benefits of local tuning: Environmental models: MTN 22.04 dB, Airtel 21.33 dB (overall averages reported per city). City level RMSEs cluster lower for Kano and Abuja (~19.98–22.04 dB) and are higher for Ibadan and Lagos (~31.96–39.46 dB), reflecting greater propagation complexity in denser, more high-rise building areas. Developed models (Lagos Airtel, Lagos MTN, Ibadan MTN) show substantially lower RMSE than existing models, confirming that parameters accounting for vegetation, building density, and urban morphology materially improve prediction fidelity.

Differences in model performance and measured radio metrics are primarily attributable to urban morphology (building height/density), vegetation cover, terrain irregularities and operator site planning (site density and antenna configuration). Higher RMSE in larger metropolitan areas indicates residual modeling challenges at 5G frequencies, even after optimization. Prioritize site specific model calibration and iterative drive testing as standard practice for urban 5G planning. Increase antenna height where feasible and optimize antenna count, tilt and orientation to reduce shadowing and cell overshoot. Densify networks with small cells, repeaters or boosters in identified poor coverage corridors to raise RSRP and SINR and expand areas of high throughput. Incorporate vegetation and detailed urban morphology parameters into propagation models to improve high frequency predictions. Overall, the findings support a planning approach that combines localized propagation modeling, targeted densification and radio parameter optimization to achieve reliable, high capacity 5G coverage in Nigerian cities.

References
1. Tolulope TO, Pradeep K, Nicholas OO. Propagation path loss prediction modelling in enclosed environments for 5G networks: A review. American Journal of Electrical and Computer Engineering. 2022; (2): 112. Doi: 10.1016/j.heliyon.2022.e11581
2. Asma AB, Rafiqul IMD. Path loss models for outdoor environment with a focus on rain attenuation impact on shortrange millimeter-wave links. Advances in Electrical Engineering, Electronics and Energy. 2023; (100): 1-11.
3. Agbotiname LI, Tolulope EO. Propagation Measurements of a 4G LTE Network in Lagoon Environment. Journal of Engineering Research. 2018; 3: 19.
4. Adeyemo ZK, Olawuyi TO, Oseni OF, Semire FO. Modification of lognormal prediction model for HSPA network using adaptive neuro fuzzy inference system lautech. Journal of Engineering and Technology. 2018; 12(1).
5. Alison N, Shintaro R. Atmospheric Processes and Phenomena, Atmospheric Forces and Wind, Chapter Ten, Press Books. 2019; 8798.
6. Andre P. Innovative Technologies for Market Leadership Uniquitous Computing; From 5G To the Edge and Beyoung, Springer, Revised Edition. 2020; 2: 133151.
7. Akhideno IO, Eguasa O. Investigating the Path Loss of Cellular Mobile Network in Suburban Areas of Benin Metropolis of Nigeria at 910 MHz. FUDMA Journal of Sciences (FJS). 2023; 7(2): 110113.
8. Akpaida VOA, Anyasi FI, Uzairue SI, Idim AI. Determination of an Outdoor Path Loss Model and Signal Penetration Level in Some Selected Modern Residential and Office Apartments in Ogbomosho, Oyo State, Nigeria. Journal of Engineering Research and Reports. 2018; 1(2): 125.
9. Alexander SG, Kate G. Guide to Planning, Architecture and Benefits of 5G Enterprises. 2020.
10. Andrea G. Wireless Communications Cambridge, New York Cambridge University Press, First Edition. 2005; 1: 230239.
11. Aaron DMA, Lourdes RB, Emmanuel DR, Matthew ZM, Isabel N. Radio Wave Propagation: Simulation of Free Space Propagation Path Loss. International Journal of Emerging Trends in Engineering Research. 2020; 8(2): 281287.
12. Ayidu NJ, Elaigwu VO. Path-Loss Prediction Model in Wlan Propagation. FUDMA Journal of Sciences (FJS). 2023; 7(3): 1-5.