Determination of Measurement Points in Urban Environments for Assessment of Maximum Exposure to EMF Associated with a Base Station

A base station (BS) antenna operates in accordancewith the established exposure limits if the values of electromagnetic fields (EMF) measured in points of maximum exposure are below these limits. In the case of BS in open areas, the maximum exposure to EMF probably occurs in the antenna’s boresight direction, from a few tens to a few hundred meters away. This is not a typical scenery for urban environments. However, in the line of sight (LOS) situation, the region of maximum exposure can still be analytically estimated with good results. This paper presents a methodology for the choice of measurement points in urban areas in order to assess compliance with the limits for exposure to EMF.


Introduction
The mobile communications technology has revolutionized the telecommunications industry worldwide over the last two decades.In order to attend the demand of users of cellular services, a substantial growth has been observed in both the amount of BS and the number of available frequency bands.International bodies have established EMF exposure limits for personal protection [1,2] that are endorsed by World Health Organization (WHO) [3] as well as measurement standards to assess the compliance of radio communication stations with the exposure limits [4].
Three basic components have to be taken into account when evaluating human exposure to EMF: radio frequency (RF) source, wireless channel, and exposed person.The first considers technical characteristics of the radiating system, like radiated power, antenna gain, height, tilt, and half-power angle.The second is related to the path loss, considering that the wave propagates in an uncontrolled and lossy medium and is subject to variations not only due to distance between EMF source and exposed person but also due to shadowing and multipath [5].The third component is related to characteristics of the exposed person, including, but not limited to, height, weight, and position (standing up, sitting down, lying, etc.).
It is worth noting that EMF exposure and coverage prediction deal with the propagation phenomena.Nevertheless, the human exposure to EMF is a quite different issue compared with the coverage problem, given that the relevant region of the former is in the vicinity of the station, where higher power density levels can be measured.Therefore, the selection of measurement points located in the region of maximum exposure due to the BS being evaluated is very important to guarantee compliance with EMF exposure limit.
For practical reference level assessment in far field region, measurement of derived quantities like electric field intensity, magnetic field intensity, or power density are sufficient to ensure that the basic restrictions are satisfied [2].
This paper presents a methodology to estimate the probable location of maximum exposure to EMF associated with a BS antenna in urban areas as well as suburban or rural areas, filling a gap not covered by current international standards [6][7][8], which address selection of points of investigation, without covering how to identify locations of maximum exposure.In [9], an approach was developed based on open area modeling, including simulations and measurements in 2 International Journal of Antennas and Propagation such type of environment.The present work improves and generalizes the previous methodology, so that it could be suitable for any environment, including paths with slopes, for LOS condition.
The remainder of this work is structured as follows.In Section 2, we present an overview of urban environments and traditional propagation models are revisited.In Section 3, we present the methodology of how to select points of investigation based on chances of maximum exposure.In Section 4, the proposed methodology is tested using simulation and measurement results.Finally, in Section 5, some conclusion remarks are presented.

Propagation Models
For proper evaluation of the path loss of radio waves, the propagation models should adequately consider the main characteristics that impact the wireless system, such as frequency, antenna height, and the environment (terrain, building, vegetation, LOS, NLOS, etc).
Urban areas are more complex than open areas scenarios in relation to the RF, presenting multiple objects in the environment that produce reflected, diffracted, or scattered replicas of the original signal.In these cases, the free-space propagation model overestimates the real exposure level; besides, complex environments enhance nonuniform field distribution along an exposed person, once those replicas reach the human body with different amplitudes, phases, and time delays.
Figure 1 shows a typical scenario of the set RF source, wireless channel, and person exposed.Point  1 (building rear) is reached with the maximum radiation intensity direction, part of the energy is reflected, and part will penetrate and pass through the building, being both refracted and absorbed in a multilayer environment.Part of the signal will pass through the building and reach  1 and  2 .The absorbed wave may be more relevant at  1 than  2 and  3 diffracted waves.However, at certain distance the diffracted signal becomes predominant, for instance, at  2 .The  3 person will be reached by a direct ray combined with reflected ( 3 and  4 ) and scattered components ( 5 ).Basically, shadowing is the effect of diffraction while multipath is the effect of reflection and scattering.
Multipath (or small-scale) fading creates nonuniform field distribution along human body.Therefore, performing a spatial averaging process is highly recommended for a wholebody assessment [5].It must be noted that spatial averaging, for the purpose of human exposure to EMF, is performed by averaging the electric field intensity squared and then taking square root of the result, or by averaging the power density.Spatial averaging of the electric field intensity will result in higher values compared with simple averaging.
The combination of path loss and shadowing can be written as where  and  can be function of one or more of the following parameters: frequency, antenna height, and environment;  is the distance from the antenna;  0 is a reference distance in the same unit of ; the shadow (or large-scale) fading parameter  is a zero-mean Gaussian random variable, with standard deviation  [10].Nevertheless, it can be assumed that the spatial variations at a local scale are only due to small-scale fading [11].For instance, for the free-space model  = 32.4+20⋅log(),where  is frequency, in GHz,  0 = 1 m, and  = 2.
If the path loss follows (1), then the average power density, in W/m 2 , can be estimated using the following generalized formula: where  is the power supplied to the antenna, in ;  max is the maximum gain of the antenna; (, ) is the relative numerical gain, varying with elevation and azimuth angles;  is a fix value for the specific considered model;  is the path loss exponent.The influence of (, ) for NLOS is studied in [12].

The Two-Ray Propagation Model.
The two-ray model considers that between the EMF source and the receiver there are just the direct ray and a single specular reflected ray that dominates the multipath effect.The direct ray propagates through free space and the reflected ray is proportional to the direct electric field intensity, by a complex factor Γ (reflection coefficient).With this model, the nonuniform distribution along the whole body can be shown, mainly for ultrahigh frequencies (UHF) or above.
For a fast and conservative estimation of the power density, it is common to consider that the reflected ray is inphase with the direct ray, so that the power density can be calculated as where |Γ| is the modulus of the ground reflection coefficient (a typical value is 0.6).Considering the relationship between the equivalent isotropically radiated power (EIRP) and the power received by an isotropic antenna, it is possible to estimate the power density at the receiving point in scenarios that follow LOS COST-WI model, as (see Appendix)

Walfisch-Ikegami
It should be noted that the exponent in ( 4) is the same path loss exponent , while the value 2.08 is associated with the parameter .

Other Models.
Reference [14] verified that for their LOS measurements, the results were very close to free space, with average path loss decaying as  −2 .Therefore, for these cases,  = 2 and  ̸ = 4.Reference [15] also identified  close to 2 in rural areas (2.1 for receiving antenna at 10 m and 2.7 for receiving antenna at 6 m).
Other path loss propagation models like Hata, ECC-33, and SUI can also be written in the form PL = +10⋅⋅log(), with some correction or gain factors being added.
All these complex models present better results for NLOS and typically overestimate path loss for LOS environments.In general, LOS presents  ≈ 2 (2.6 for LOS COST-WI), while NLOS presents  between 3 and 5.

Methodology for Determination of Measurement Points
In [9] an approach was developed based on open area modeling, including simulations and measurements in such type of environment.The present work improves and generalizes the previous methodology, so that it could be suitable for any LOS condition, including paths with slopes.
It is quite clear that the maximum exposure region is likely to occur parallel to the antenna boresight for open area cases.Nevertheless, depending on the antenna installation and technical characteristics, the main lobe might not be responsible for maximum exposure point, but the sidelobes [16].On the other hand, as the maximum exposure may be located some hundreds of meters away the BS, chances are  that buildings constructions might obstruct the LOS in urban environment.Therefore, not only positions in the direction of the antenna's boresight are relevant, but also other positions with LOS to the antenna may be a point of interest.
Figure 2 illustrates a simulation in downtown of São Paulo, Brazil, where just one sector of the BS is presented in order to evaluate the single source influence.It must be noted that the antenna azimuth is 0 ∘ north, the crossing street orientation is approximately 35 ∘ , and there are huge buildings in 3 of 4 corners, as shown in Figure 3.In this case, relevant maximum exposure points will occur not in the direction of the antenna's boresight but according to the orientation of crossing streets that have LOS with the BS.The simulation tool used is ICS Telecom V12 Radio Planning and Technical Spectrum Management software [17], configured to run propagation model based on ITU-R Rec.525, calculation of free-space attenuation, and Rec.P.526, propagation by diffraction.
The tridimensional radiation pattern (, ) can be approximated by () ⋅ (), where () and () are vertical and horizontal radiation patterns, respectively.This is a good approximation for the forward radiation and a sufficient approximation for the backward radiation, which is satisfactory from the point of view of exposure assessment International Journal of Antennas and Propagation is not exactly equivalent to (a) because the -axis should also be rotated by  degrees.However, the error introduced by this assumption is small compared to the estimated maximum exposure location.[18].The main lobe of () can be approximated by cos  ( − ), where  is the antenna tilt [9].A BS antenna operates in accordance with the established exposure limits if the values of EMF measured in points of maximum exposure are below these limits.Additional measurement points may be required, but the proposed methodology provides guidelines on how to judiciously select the necessary measurement location, that is, the region where the antenna main lobe reaches the exposed person, in a LOS case.Reference [16] showed that, in some cases, side lobes can be responsible for maximum exposure, even transporting less energy.Nevertheless, it is very unlikely that any EMF exposure problem will happen in these cases.
The estimated maximum power density location can be , where  is the horizontal distance from the BS to a reference point,  is the antenna height, and ℎ is the approximate head level height.Taking / = 0 and isolating , we obtain where  = log(1/2)/ log[cos( bw /2)] and  bw is the halfpower angle in the vertical plane.Consequently,  exp max is a reference measurement point, since it is likely that the maximum exposure location is near this point.
The impact of street inclination () in  exp max can be mitigated by adequately adding/subtracting  to the antenna tilt , turning the sloped path a plane region, as presented in Figure 4.The error introduced by this assumption in the axis is ℎ ⋅ sin , which is much less than  exp max .Therefore, (5) can be applied for these cases with adequate correction in , generating a   .

Testing the Proposed Methodology
Three urban areas cases are presented to test the methodology.The first case considers the site of Figures 2 and 3.The technical characteristics of the BS are presented in Table 1, as well as the results for different  values and simulation.Figure 5 shows that the simulated and calculated maximum exposure locations are compatible.The result matched the  = 2 due to the propagation model selected in the simulation, that is, free space combined with diffraction.
The second case considers path loss propagation measurements at 910 MHz made in downtown core of Ottawa, Canada, using transmissions from an antenna at 8.5 m height to a receiving antenna mounted on a van at 3.65 m height [19].The author conducted LOS and NLOS measurements in areas with buildings taller than the antennas in both street sides, but just LOS measurement in the westward of Slater Street is considered for testing the proposed approach.
The transmitter and receiver antennas are omnidirectional and the elevation discrimination was discounted, so  the presented results are just path loss, which were recovered from the graphical curves presented in [19].The received power was sampled approximately once per meter with a calibrated receiver.The resulting path loss is shown in Figure 6.Based on a simple linear regression analysis, a path loss model is given by PL = 23.4+ 26.6 log () , where  is the distance, in meter, and  0 = 1 m.It can be seen that the proposed model is very close to LOS COST-WI model, given by PL = 23.8+ 26 log(), at 910 MHz,  in meters, and  0 = 1 m.
Using the formulation presented in the Appendix, the power density can be estimated by  (, , 0) =  ⋅  max ⋅  (, 0) 1.89 ⋅  2.66 .Simulations with the proposed path loss model and freespace model were executed following procedures used in [20], as presented in Figure 7.The BS has the technical characteristics presented in Table 2.As expected, the simulated and calculated maximum exposure locations are equal or very close.
The third case uses [12] data collected in a measurement campaign performed in the centre of Aarhus, an urban medium city in Denmark, with average building height about 15-18 m and street width about 20 m.The technical characteristics of the BS are presented in Table 3, as well as the results for different  values and simulation.Based on GPS collected data and Google Earth path profile, the average street slope is approximately 3.8 ∘ within a 100 meters radius from the BS (Sector 1 of [12]).The receiver was a van with a 5 dBi omnidirectional antenna with ground plane at 2.5 m height and a network scanner.The sampling rate was 50 samples/s for an average driving speed of 15 km/h.Of course, the maximum received power point is the same as the maximum power density point.In the present work, no correction for receiving vertical antenna pattern was used; therefore, only measurement farther than 20 m was considered in order to minimize its influence.Figures 8 and  9 show that the expected maximum exposure locations are coherent with the proposed methodology.
As it was shown in the cases of study, the estimated location  exp max indicates the probable region of local maximum exposure.Nevertheless, it must be taken into account that scattered and diffracted fields may shift the real local maximum exposure closer to or farther from the radio communication station.Besides, there are uncertainties related to the exact BS technical information as well as GPS altitudes to estimate street slopes and results are sensible for these parameters.Therefore, a procedure of walk around the estimated location measuring the electric field intensity must be performed to identify the point of maximum exposure.Spatial averaging might be used when appropriate.

Conclusion
This paper presented a methodology to estimate the probable location of maximum exposure to EMF associated with a BS antenna in any environment for LOS cases, filling a gap not covered by current international standards.
The results of the proposed methodology are consistent with the presented case studies and previous work.For LOS scenarios the estimated maximum exposure location is dependent of path loss exponent, , but its influence does not play major role for short distances, as maximum exposure point is usually located in vicinity of the base station, below few hundreds of meters.
Results show that higher  brings maximum exposure location closer to the base station.Results suggest that in case of urban environment with canyon streets characteristics and LOS, technical staff should consider  = 2.6 for a reference point, and then walking around the estimated location measuring the electric field intensity to identify the real maximum exposure point.
Results also suggest that in other LOS cases, technical staff should consider  = 2 for a reference point, knowing that there are more chances to have maximum exposure location closer to the base station, without disregarding the walking around measurement procedure.
NLOS cases are likely to present low power density levels for human exposure concerns, although it would provide satisfactory signal level for mobile communications.Therefore, when selecting measurement points for EMF exposure assessment, LOS places should be preferable compared with NLOS locations.

Figure 1 :
Figure 1: Representation of the main propagation mechanisms in a complex environment: LOS, reflection, diffraction, absorption, and scattering.
Model: Line of Sight Scenario.COST-Wal-fisch-Ikegami model (COST-WI model) takes into account more information to describe the urban environment, like height and separation of buildings, widths of streets, and street orientation with respect to the direct radio path.This model distinguishes between LOS and NLOS cases, where the EMF source and the receiver are within a street canyon in the former case.This formulation is based on measurements performed in Stockholm, Sweden[13].For the LOS COST-WI model,  = 42.6 + 20 ⋅ log(), and  = 2.6.This model is suitable for use over the ranges of 800-2,000 MHz, height of base station antenna () between 4 and 50 m, height of mobile antenna (ℎ) in range of 1-3 m, and 20 m ≤ d ≤ 5,000 m.

Figure 2 :
Figure 2: Simulation of a BS located in downtown of São Paulo.Buildings in both sides of the streets block the EMF propagation.However, directions with fewer blockages present stronger EMF signals, even for back radiation of the antenna.

Figure 3 :
Figure 3: A tridimensional view of the BS among densely tall buildings.Antenna height is 30 m, azimuth is 0 ∘ , tilt is 10 ∘ , and half-power angle is 88 ∘ and 26 ∘ , for horizontal and vertical planes, respectively, EIRP = 1000 W and  = 900 MHz.

Figure 4 :
Figure4: Approximation of a sloped path by a plane path with tilted antenna.(b) is not exactly equivalent to (a) because the -axis should also be rotated by  degrees.However, the error introduced by this assumption is small compared to the estimated maximum exposure location.

2 )Figure 5 :
Figure 5: Power density evaluated based on ICS Telecom simulations.The simulated maximum exposure point is located at 75 m, same as estimated for  = 2 and higher than estimated for  = 2.6.

Figure 6 :
Figure6: Path loss westward of the transmitter located at 300 Slater Street.In this case, the PL = 23.4+ 26.6 log(), with  = 3.84, was obtained using least square (LS) regression analysis and then compared with free space (FS).The intercept point between both models is 18 m.

2 )Figure 7 :Figure 8 :
Figure 7: Power density depending on the distance  from the BS, as computed with the free-space model and the case specific PL model.It is being considered that PL below the intercept point at 18 m is equivalent to free-space propagation.