1. Introduction
Concern about exposure to radio frequency (RF) electromagnetic fields (EMF) is an issue of growing interest in modern society. The evolution of the network towards 5G technology has increased this concern. In the words of the European Commission, “Some citizens perceive the fifth generation of wireless networks—5G—as a threat to public health, as they think that EMF exposure is higher than exposure from current 4G networks” [
1]. The spread of alarming news about the issue and the skepticism of the population can only be combated with dissemination tools that contribute to an adequate perception of the risk. The clear presentation of the results in the form of exposure maps is one way to contribute to an adequate perception of the exposure levels and their temporal evolution. Interpolation techniques such as kriging are used to produce these exposure maps [
2,
3]. In addition, this type of large-scale study allows for further analysis related to the appearance of or increase in different diseases.
According to commonly accepted procedures, there are two types of measurements for monitoring RF field exposure: mobile (personal) and fixed. Fixed measurements, determined with a spectrum analyzer or a frequency-selective meter, are the most accurate, although they require more effort in terms of cost and personnel [
4]. For rapid characterization at a fixed location, there is also the possibility of using a broadband probe. When the location to be assessed is very large, it is of crucial importance to optimize the number of measurements to be performed to be able to carry them out without increasing the cost of the equipment or greatly increasing the number of hours required. Mobile (personal) measurements use exposure meters that allow many measurements to be collected for a single individual or a population group. They allow for the analysis of spatial variation for a single user but present the problem of generalization of the obtained results. When measurements of multiple environments are required, the number of exposure meters needed or the time needed to carry out the study is also increased if the measurements are taken consecutively. The cost of the equipment (thousands of Euros per device) is a very important limiting factor for both fixed measurements and personal exposure meters. This has led to the search for low-cost solutions for the measurement of the new 5G mobile networks [
5].
Up to now, to ensure the adequacy of EMF exposure levels at a site regarding the current regulations, two methods are recommended: measurements with a selective meter or spectrum analyzer and broadband measurements averaged over a 6 min period. Both are based on the original 1998 ICNIRP guidelines [
6], EN50413 and IEC 62311 [
7], on which the European regulations are based [
8]. There has recently been an update of the ICNIRP recommendations on EMF protection [
9]. Among the most notable changes affecting this work are the change in the averaging interval to 30 min (as opposed to the previous 6 min) for the whole body and the limit of 2 GHz in the definition of the reference values. These changes have not yet been incorporated into the legislation. For this reason, and to facilitate comparison with previous work, the measurements made in this proposal were averaged over a 6 min interval.
The novelties of this study are as follows:
A smartphone-based method is proposed to obtain a large number of measurements of an area in which to characterize EMF exposure.
An area of 1.9 km2 was established by installing the terminal on a bicycle. Exposure maps were made using the equivalent interpolation techniques for fixed sites. The obtained results were compared with those of standard methodologies.
A tool is provided for obtaining values equivalent to EMF exposure that can, in a simple way, to contribute to an adequate risk perception.
The proposal is organized as follows.
Section 2 describes the location of the area under study and the measurement points for the cases of fixed sites and the smartphone. It also includes the characteristics of both devices and briefly mentions the programming tools used to design the application that allows for the cell phone to be used as a power meter. The relationship between both measured values is also theoretically justified.
Section 3 discusses the measurements taken with both methods, their statistical distribution, and the possibility of generating exposure maps with both methods.
Section 4 discusses the advantages and limitations of the presented proposal. Finally,
Section 5 highlights the conclusions of this study.
3. Results
First, the statistical distribution of both data sets was analyzed. On the one hand, the distribution of the values measured with the EMR 300 broadband meter was fitted to the distribution, and on the other hand, the values measured with the smartphone were fitted to the distribution. Since the values measured with the app are given in dBm, to compare the statistical distribution, the equivalent electric field was obtained. Since no realistic gain estimate is available, it was decided in this case to normalize the maximum value to unity. In both cases, in accordance with previous work, the values were fit to a lognormal distribution. Statistical analyses were carried out using Statgraphics 19 [
17] software. This package can determine the means and variances, compare sample independence, and fit points to different probability density functions. The results of both adjustments can be seen in
Figure 5 and
Figure 6.
As can be seen, both measurement sets follow the lognormal distribution fit. Clearly, the statistical properties of the mean and variance of both sets are not the same, but this is perfectly explainable by the exact conversion value between both measurement methods being unknown. Additionally, the measurement condition itself is different in both cases. In the case of the meter, it is a measurement averaged over a 6 min interval at the same location; in the case of the smartphone, it is a power measurement received by the terminal taken every 5 s along the indicated route. Measurements taken with the smartphone are subject to greater variability (fading or loss of signal) and may therefore lead to problems in the collected data. A further statistical analysis in
Section 3 discusses the statistical validity of measurements taken along a trajectory instead of in a fixed point.
Therefore, a rigorous calibration process would be necessary to establish a direct relationship between the two measurements. This is also useful in terms of the existence of proportionality between measurements. This result could lead to an immediate application for the establishment of comparisons, for example, between indoor and outdoor exposure, using a mobile terminal, or variations in exposure between different areas. Following this idea, EMF exposure maps were obtained using both techniques.
Figure 7 shows the interpolation obtained using the values measured with the EMR300 meter.
Figure 8 shows the interpolation obtained using the values taken with the smartphone. The maps were created using ArcGIS Pro 3.0 [
18]. The methodology chosen for the geographical interpolation of exposure values was ordinary kriging.
The maps show the BTS in the study’s area of influence with red dots. The fixed locations where measurements were taken with the EMR300 are indicated by red dots (LOS) and red squares (NLOS) on the map in
Figure 7. The points measured along the trajectory taken with the bicycle are indicated by lines on the map in
Figure 8.
The results shown in
Figure 7 and
Figure 8 prove that both methods can detect the areas of greatest exposure in an equivalent manner. The greatest differences are seen in the northern and southeastern zones. This is an artifact of the interpolation due to the lack of sampling points in the perimeter of the area. An island effect is also seen in some of the areas of the exposure map generated from the broadband meter values. This effect is possibly due to insufficient sampling in some areas. The average distance between measurement points was 300 m, which is at the limit obtained in previous work [
2]. A larger range of values also appeared in the case of measurements taken with the smartphone (approximately 40 dB vs. 20 dB). That is, in the case of the broadband meter, a range of values from the maximum to a 100 times lower value is obtained, while in the case of the smartphone, a range of values from the maximum to a 10,000 times lower value is obtained. This is due to the different sensitivities of both meters.
Table 1 shows the measured values for the fixed sites and the interpolated values at the same locations taken from the map in
Figure 8. The mean difference between the measured and interpolated EMF values is 72.3 dB for LOS cases and 67.8 dB for NLOS cases. This result is also consistent with the one obtained in [
16], where two conversion constants were also observed between exposure meters and broadband meters depending on the LOS/NLOS situation.
To investigate the relation between the results of the different methods, measurements were taken at three additional points, marked A, B, and C in
Figure 2. At each point, a measurement was taken with the EMR300 meter averaged over 6 min. Simultaneously, smartphone measurements were taken every second. One smartphone was fixed on the same measuring tripod as the EMR300 meter and the other measurements were taken while walking around the fixed point (see the yellow dots on the inner map in
Figure 2).
The density traces of the measurements made with the fixed smartphone and with the itinerant one are shown in
Figure 9.
Measurements taken while moving around the same point have lower values and greater variability than those taken at a fixed point. This may be due to multiple factors: other wireless services present, the reception of several frequencies simultaneously, different directions of arrival, different sensitivities of the meter and the smartphone, temporary signal fades, etc. But, as can be seen in
Figure 9, the statistical distribution of the signals received in the six-minute interval is similar in both cases. This leads to the conclusion that small movements around a site follow a similar trend to single-point measurements. In other words, a continuous measurement at a site can be replaced, without noticeable statistical difference, by a trajectory around the site. Therefore, when cycling a route, nearby points behave statistically as a single fixed point.
A different question is the possibility of establishing a direct relationship between the two measurements.
Table 1 compares the results of the interpolation with the kriging of both methods, reaching differences in the range of 60 to 87 dB between the values measured with the EMR300 and the values measured with the smartphone.
Table 2 shows the results corresponding to the three additional points that were analyzed. Points A and B present LOS conditions with respect to the nearest emitter, while point C corresponds to an NLOS condition.
Table 2 shows that there is a relationship between the measured values in the sense that a higher exposure value measured with the EMR300 corresponds to a higher value of power detected by the smartphone. Beyond that, with the available data, it seems difficult to expect that a smartphone can easily replace an exposure meter.
However, from the point of view of the characterization of large surfaces, its value has been proven as an element that can efficiently detect the areas of greatest exposure and, therefore, its greatest usefulness will be in the determination of sensitive areas where later measurements can be made with calibrated instrumentation.
4. Discussion
In this proposal, we analyzed the possibility of using a smartphone to carry out measurements equivalent to those of EMF exposure and thus contribute to an adequate risk perception. The proposed system presents, firstly, advantages in terms of the cost of the equipment needed to obtain the measured values. This could allow for collaborative studies or studies with a large number of volunteers, making it possible to generalize the results, which is the main limitation of the methods based on exposure meters. It also allows many measurements to be obtained over time, similar to what happens with personal exposure meters, although this point was not the subject of this work. The existence of an offset factor (in dB) between the exposure values measured with standard methods and the values collected with the smartphone makes this method suitable for comparative studies.
A methodology for exposure mapping was also proposed, in this case, by cycling through an area. The methodology, combined with interpolation techniques, proved to be equivalent to that based on traditional measurement methods. It also offers the advantage of less time being invested in data collection. For the 1.9 km2 area under study, the time taken to complete the route with the bicycle was 1 h and 10 min. In the case of the fixed measurement sites, each of them was averaged for 6 min. We must add the time corresponding to the travel between points, so the total time spent in this method was 4 h. This significant reduction in the required time provides an advantage.
The main drawback of the proposed methodology is that, lacking a precise conversion factor, the values do not faithfully represent the exposure, but a proportional value. Another limitation of the proposed smartphone measurement technique is that it is focused only on the measurement of BTS signals. Although the measurement possibilities can be extended to other wireless technologies, such as WiFi or similar technologies, contributions to exposure due to TV or FM are excluded from the measurement possibilities.
In terms of risk perception, the contribution can be significant since it provides the public with a comparative and transparent tool of the received values. The application allows the user to learn, in real time, the values of the set of signals (of the different BTS) to which he is exposed and to establish his own comparisons of the different exposure situations in which he may find himself. In addition, the agile creation of exposure maps using the measured values contributes to the effective and accessible dissemination of the exposure of the different areas. Furthermore, these values can be contrasted with those measured by the user himself, helping to increase confidence in the published data.
5. Conclusions
In this study, we compared the performance of EMF exposure measurements and maps following the measurement methodologies for fixed points with a broadband meter and sampling from the data collected by a smartphone. To calculate the sum of the powers received by the cell phone, an application was designed that uses the possibilities of Android terminals.
The data collected by the terminal are proportional to the exposure measurements and follow similar statistical and spatial distributions, so they are useful for comparative studies and provide the advantages of automation and less time being spent in gathering them.
The accessibility of the received signal data in any terminal in which the application is installed helps to generate confidence in the exposure values by making it possible to contrast the collected values with those that can be obtained by any user of the application.
In this proposal, we presented an alternative for the rapid characterization of an area with respect to cell phone emissions. It is not intended to replace the measurement by means of a personal exposure meter with a smartphone. For this, an extension of the research work to a much more exhaustive characterization of the terminal would be necessary to establish a relationship between the exposure values and the instantaneous values. The complete characterization of the gain of the cell phone is also indispensable to determine this relationship. Similarly, the influence on the exposure of services not covered in this work (FM, DTV, WiFi, etc.) must be assessed when comparing the two values.