Elevation Dependent Shadowing Model for Mobile Communications via High Altitude Platforms in Built-Up Areas

Elevation Dependent Shadowing Model for Mobile Communications via High Altitude Platforms in Built-Up Areas Jaroslav Holis, Student Member, IEEE, and Pavel Pechac, Senior Member, IEEE Ab

Elevation Dependent Shadowing Model for Mobile Communications via High Altitude Platforms in

Built-Up Areas

Jaroslav Holis, Student Member, IEEE, and Pavel Pechac, Senior Member, IEEE

Abstract—An empirical propagation prediction model is

de-scribed for mobile communications from high altitude platforms

(HAPs) in different types of built-up areas The model introduced

here is defined as a function of the angle of elevation The target

frequencies are selected from the 2 to 6 GHz frequency band

prospective for 3G and 4G mobile systems, namely at 2.0, 3.5, and

5.5 GHz This new HAP model recognizes two cases—line of sight

(LOS) and non-line of sight (NLOS) between a HAP and a user at

street level The simulation of the urban environment is based on a

statistical approach Additional shadowing path loss is calculated

using the uniform theory of diffraction for NLOS conditions.

Normal distribution of the additional shadowing path loss was

distinguishable from the simulation results The shadowing path

loss is defined as a function of the elevation angle The results

of the empirical model developed for idealized conditions are

verified by measurements taken from a remote-controlled airship

in different types of urban environment Close correlation was

achieved between the theoretical model and the experimental

data The HAP elevation dependent shadowing model is easy to

implement and can be used for realistic planning and simulations

of mobile networks provided via HAPs in built-up areas.

Index Terms—Empirical model, high altitude platforms (HAPs),

propagation, wireless communications.

I INTRODUCTION

OVER THE past 10 years, mobile communications have

changed the world more than any other area of technical

achievement One weak point of mobile networks is the relative

ease with which they can be disabled by disasters or terrorist

attacks High altitude platforms (HAPs) could provide a

pos-sible alternative to the terrestrial provision of mobile services

A suitable HAP is usually considered to be a quasi-stationary

unmanned vehicle in the stratosphere at an altitude of between

17 and 22 km [1], [2] It should maintain its position within a

sphere having a radius of 0.5 km [3] HAPs can combine the

benefits of both terrestrial and satellite communications In

par-ticular, they have low free space loss (FSL) when compared to

satellites and limited shadowing at high elevation angles, i.e

they provide good coverage when compared to terrestrial

net-Manuscript received July 13, 2007; revised November 28, 2008 This work

was supported in part by the Czech Ministry of Education, Youth and Sports

within the framework of the OC092-COST Action 297 and MSM 6840770014

projects.

The authors are with the Department of Electromagnetic Field, Faculty of

Electrical Engineering, Czech Technical University in Prague, Praha 6, Czech

Republic (e-mail: holisj1@fel.cvut.cz; pechac@fel.cvut.cz).

Digital Object Identifier 10.1109/TAP.2008.919209

works in dense urban areas A major advantage of using HAPs

is the low cost of deployment and, especially in the event of a disaster, their rapid deployment

The principal mobile network prospective for HAPs involves 3G systems and, potentially, mobile WiMAX systems The fre-quency spectrum for 3G mobile networks (around 2 GHz) is also allocated to HAPs [3], [4] 3G networks could be replaced by a Mobile WiMAX operating in a frequency spectrum between 2 and 6 GHz [5]

A simple FSL propagation model is unsuitable for system simulations of mobile systems provided via HAPs in urban areas An empirical roadside shadowing model (ERS) [6] or an empirical model for satellites [7] is used in order to achieve a more realistic approach to HAP propagation prediction A sto-chastic modeling approach, which can be utilized for coverage planning of HAP communication systems, is introduced in [8], [9] A propagation model for building blockage for satellite mo-bile systems was presented in [10] The results of deterministic wideband channel modeling of a satellite propagation channel with building blockage are shown in [11], [12] The impact

of vegetation on the propagation environment for terrestrial and satellite mobile communications is demonstrated in [13] Complex information about propagation effects in megacells is provided in [14] An overview of channel modeling for HAPs

is given in [15]

An elevation dependent propagation model is required for HAP scenarios This type of model plays a key role in real-istic studies of mobile services from HAPs The propagation model presented in this paper is based on simulations using

a randomly generated urban environment The theoretical ap-proach is partly verified by measurements using a remote-con-trolled airship The aim of this paper is to introduce an HAP shadowing model defined as a function of the elevation angle

in different types of built-up areas and the prospective for HAP mobile applications in the 2–6 GHz frequency band, namely at 2.0, 3.5, and 5.5 GHz This model can be used for radio network planning of mobile systems provided from HAPs

A comprehensive modeling approach based on stochastically generated urban environment and ray tracing was introduced in [16] The fade distribution can be derived as an output according

to specific parameters that describe the scenario The statistical distribution of building heights, street widths, the azimuth, etc

as well as the position of the mobile antenna within the sce-nario can be defined The model presented in this paper is based

on the same principles, but for the urban environment the sta-tistical ITU-R Rec P.1410 model [17] was utilized in order to 0018-926X/$25.00 © 2008 IEEE

describe different types of built-up areas in general terms

More-over, using very extensive simulations for all possible locations

of a mobile terminal within the modeled area, the line of sight

(LOS) probability as well as the shadowing were derived as a

closed form expression By this method path loss can be

cal-culated as a function of the elevation angle and can be directly

used in system level simulations

Section II presents the propagation modeling approach The

simulation results and the shadowing model are described in

Section III The propagation model defines the probability of

LOS connections at street level and predicts an additional loss

due to the shadowing effects of buildings using a statistical

dis-tribution Finally, Section IV deals with the verification of the

model by measurements

II PROPAGATIONMODELINGAPPROACH

A Statistical Model of Building Deployment

As was mentioned above, a randomly generated urban

en-vironment was utilized to simulate different types of built-up

areas The statistical ITU-R Rec P.1410 model [17] was

adopted for building deployments The application of the

sta-tistical model was motivated by its universality and simplicity

The great advantage of this model is that the city can be modeled

without specific information concerning building shapes and

distribution The statistical model requires only three empirical

parameters describing the built-up area The ratio of land area

covered by buildings to total land area , the mean number

of buildings per unit area , and a parameter determining

building height distribution The parameter a ranges from

0.1 to 0.8 and ranges from 750 to 100, respectively The

parameter parameterizes the Rayleigh statistic distribution of

building heights where represents the most frequent building

height The Probability Density Function (PDF) for the building

heights based on the Rayleigh distribution is

(1)

where denotes a probability that the building height is equal

to in the given urban environment The ratio of land area

covered by buildings can easily be obtained from 2D city

plans The mean number of buildings per unit area is the

least important parameter since it has a minor impact on the final

simulation results presented below The buildings are generated

with random heights based on the Rayleigh distribution This

guarantees a realistic approach when analyzing the shadowing

conditions in terms of statistics, although the buildings are

de-ployed in an unrealistic regular grid

Four different types of environment were selected for the

sce-narios presented here:

1) Suburban area;

2) Urban area;

3) Dense urban area;

4) Urban high-rise area

Table I summarizes the parameters of the statistical ITU-R

Rec P 1410 model used to define the different types of

en-TABLE I

P ARAMETERS OF THE ITU-R P.1410 M ODEL FOR S ELECTED E NVIRONMENTS

Fig 1 Geometry of basic LOS and NLOS scenarios.

vironment ranging from suburban districts to city centers with skyscrapers

It must be noted that in some references the building height density is shown to be closer to log-normal distribution than

to Rayleigh [8], [14] Information on distributions of building heights can be found in [14], [17]–[19]

B Simulation Method

An urban area with dimensions of 2 by 2 km was considered for the analysis The resolution of building deployment for the simulations was equal to 1 m This fine resolution, a sufficiently large area, and the extent of the simulation ensure that the re-sults are not dependent on a specific case of randomly generated building heights

The simulations were divided into two cases First the LOS probability in the streets was analyzed as a function of an ele-vation angle for different types of built-up area Second, addi-tional path loss due to shadowing effect of buildings was ana-lyzed using the uniform theory of diffraction (UTD) [20], [21] Building rooftops were modeled as dielectric wedges with a rel-ative permittivity equal to 4 Rooftop diffraction loss was calcu-lated for both vertical and horizontal polarizations separately

A more detailed ray tracing could not be applied here since the statistical model of the environment does not include realistic street structures with street canyons etc Fig 1 illustrates the basic geometry covering both LOS and NLOS cases

The calculations were made for azimuth angles ranging from 0–360 with a variation of 9 degrees Initially, buildings were randomly generated using the statistical model Then the posi-tion of the HAP was determined for each point on the streets in

Fig 2 LOS probability in the streets as a function of the elevation angle for

selected environments.

a fine grid for a given elevation and azimuth angles in order to

simulate and analyze an extremely large number of scenarios

The LOS probability for a specific elevation angle was

calcu-lated as the median value of the data obtained for all azimuth

angles In this way the results are independent of the azimuth

angle, since in the real world buildings are not usually located

in a regular structure

The simulations were performed for the full range of

eleva-tion angles from 1 to 89 degrees

III NEWPROPAGATIONMODEL

The simulation results can be divided into two cases First,

the probability of LOS connections in the streets is shown In

this case the free space loss (FSL) model can be used for the

mean path loss calculation Second, the additional path loss due

to the shadowing effect of buildings must be considered for the

NLOS connections The realistic elevation angles for HAP

mo-bile applications in cities range from 60 to 90 (the HAP is

situated at the zenith above the mobile terminal), but lower

el-evation angles are not excluded as they might be important, for

example, in terms of interference or for other studies For

in-stance, the distance between the user and a sub-platform point

on the ground (the point vertically below the HAP) is about 211

km for an altitude of 22 km (including the impact of the earth

curvature) and about 168 km for an altitude of 17 km, if the

el-evation angle is equal to 5 degrees

A LOS Probability

The LOS probability in the streets as a function of

the elevation angle was obtained for the four environments, see

Fig 2

A simple function was then found to approximate the

simu-lated data in Fig 2

(2)

TABLE II

P ARAMETERS FOR LOS P ROBABILITY C ALCULATION (2)

where is the probability of LOS in percent, is an eleva-tion angle in degrees and , , , , are the empirical parameters given in Table II for the four typical environments Parameters for an environment described by the arbitrary values of , ,

of the ITU-R Rec P 1410 statistical model could also easily be derived from the simulation results

To a certain extent the results can be compared with the roadside building-shadowing model [6], which uses the same Rayleigh distribution of building heights It gives the proba-bility of the link blockage as a function of the elevation angle, the azimuth angle relative to street direction, and the distance

of the mobile terminal from the buildings, while the model (2) is based on statistics covering all possible positions of the mobile terminal within the area Appropriate parameters for both models were selected to define several test cases which would enable tentative comparisons As expected, both models give similar results

B Additional Shadowing Loss

This section presents the impact of the shadowing effects of buildings on NLOS connections The great benefit of HAP sta-tions as compared to satellites is that the relatively short path length also enables NLOS links between the mobile station and the HAP station Fig 3 illustrates an example of the simulation results—a normalized histogram for additional rooftop diffrac-tion loss at 2.0 GHz for vertical polarizadiffrac-tion at an elevadiffrac-tion angle

of 70 Normal distribution can be distinguished from the figure The PDF for the normal distribution fitted to the simulated data

is also shown in Fig 3 The PDF of the normal distribution can

be written as

(3)

where is normalized probability, the mean value in dB, and the standard deviation in dB For the specific case in Fig 3

The cumulative distribution functions (CDF) for the simu-lated data are depicted in Fig 4 for elevation angles equal to

85 , 80 , 70 , and 50 The dashed curve in Fig 4 illustrates the CDF of normal distribution fitted to the simulated data The close conformity between the simulated data and the normal dis-tribution is distinguishable from this figure The CDF for eleva-tions between 10 and 50 are very similar to the 50 curve so they have not been included in Fig 4 in order to maintain clarity Parameters of the normal distribution (mean value and stan-dard deviation ) for elevation angles ranging from 1 to 89 are depicted in Fig 5 All the parameters are given as an averaged

Fig 3 Normalized histogram and PDF of shadowing loss at 2.0 GHz for

ver-tical polarization in a dense urban area.

Fig 4 CDF for shadowing loss at 2.0 GHz for vertical polarization and for a

range of elevation angles (solid line—simulation results, dashed line—CDF of

fitted normal distribution).

value of vertical and horizontal polarizations in the following

stages

The results presented in Figs 3–6 were obtained for a dense

urban area Simulations for the other selected environments

were also performed The same results for additional rooftop

diffraction loss were achieved, because the incidence angle

(see Fig 1) depends much more on the elevation angle, which

was calculated for each point in the streets independently,

than on building heights This means that the same range of

angles was included in the UTD wedge diffraction geometry

regardless of the environment; the impact of street widths was

also insignificant The type of built-up area only determines

the number of NLOS points in the scenario, i.e., the LOS

prob-ability Table III presents the normal distribution parameters

for an elevation angle of 70 and a carrier frequency equal to

Fig 5 Normal distribution parameters for shadowing loss as a function of the elevation angle for vertical and horizontal polarization.

Fig 6 Normal distribution parameters as a function of an elevation angle for carrier frequencies equal to 2.0, 3.5, and 5.5 GHz.

TABLE III

P ARAMETERS OF N ORMAL D ISTRIBUTION FOR S ELECTED E NVIRONMENTS

( F = 2 GHZ, 2 = 70 )

2.0 GHz for the four environments Only the suburban area has slightly different parameters, but these are insignificant The simulations and analysis were also performed for the 3.5 and 5.5 GHz frequencies It is obvious that a higher shadowing loss was recorded at these higher frequencies Fig 6 presents the parameters of normal distribution for three different carrier frequencies and elevation angles ranging from 1 to 89

TABLE IV

P ARAMETERS OF (4) F OR F = 2:0 GHZ, A LL E NVIRONMENTS

TABLE V

P ARAMETERS OF (4) FOR F = 3:5 GHZ, A LL E NVIRONMENTS

TABLE VI

P ARAMETERS OF (4) FOR F = 5:5 GHZ, A LL E NVIRONMENTS

The data were divided into two intervals of the elevation

approximate the simulation results by a simple function A

simple fractional rational function was derived as a best fit to

simulation results to approximate the mean value as well as

standard deviation of the normal distribution

(4)

where is the elevation angle in degrees and , , are empirical

parameters

The approximations are shown in Fig 6 by continuous lines

The empirical parameters are summarized in Table IV for a

carrier frequency of 2.0 GHz, in Table V for 3.5 GHz, and in

Table VI for a carrier frequency equal to 5.5 GHz The

parame-ters for other frequencies in the 2 to 6 GHz range can be obtained

as an interpolation of the parameters presented in Table IV–VI

Parameters from Tables II–VI should be used with the given

dec-imal points precision in all calculations

C HAP Shadowing Model

Finally, the path loss in a built-up area can be expressed in dB

as

LOS

where is the free space loss which can be calculated as

(6) where is the distance between the transmitter and the

re-ceiver in km and frequency in GHz represents a random

Fig 7 CDF of a shadowing model for a dense urban environment at a fre-quency of 3.5 GHz.

shadowing in dB as a function of the elevation angle It is cal-culated using the normal distribution parameterized by (4) and appropriate empirical parameters from Table IV–VI Because of the need for realistic system-level modeling of mobile systems, random components and in dB are added as a loca-tion variability utilizing the log-normal distribuloca-tion with a zero mean Based on the measurement results presented in Section IV below and in relation to [14], the standard deviation of the loca-tion variability is from 3 to 5 dB for LOS connecloca-tions and from

8 to 12 dB for NLOS connections

As an example of the practical implementation of (5), (6) shows the generation of random shadowing loss at 2 GHz for

an elevation angle of

LOS

NLOS (7) where is the distance in km, is the elevation angle in de-grees and the generates random numbers using the normal distribution with the mean and the standard devi-ation in dB is a standard Matlab function Another option is to express the elevation dependent shad-owing model in the form of CDF as

(8) where is probability in percent, and represent the shad-owing loss in dB, is the mean value in dB and the standard deviation in dB of the normal distribution by (4), and is the line of sight probability at street level by (2)

As an example of the type of calculation Fig 7 shows a CDF calculated from (8) for 3.5 GHz, for a dense urban environment

as defined in Table I, and elevations of 80 , 70 , 50 , and 60

Fig 8 Path loss at 2.0 GHz measured in front of a flat roof building during an

airship flyover.

IV VERIFICATION BYMEASUREMENT

These methods of built-up area modeling and the

calcula-tion of addicalcula-tional shadowing loss using wedge diffraccalcula-tion may

appear over-optimistic Different roof profiles do occur in real

cities and the model only employs the shortest path between a

HAP station and a mobile terminal That is why measurements

were taken from a low altitude remote-controlled airship [22] (9

m long) in order to verify the feasibility of the simplifications

employed

The signal was transmitted from 26 dBm CW generators at

frequencies of 2.0, 3.5, and 5.5 GHz using patch antennas

sit-uated at the bottom part of the airship gondola A spiral

broad-band antenna was used to receive the transmitted signal at street

level Several trials were performed in the city of Prague The

re-ceiver station was located in real conditions in front of a building

on the street The airship flew across the rooftops at an altitude

of about 150 m The low altitude enabled a wide range of

el-evation angles (20 –90 ) to be obtained Measurements were

taken using different types of buildings and roofs Two samples

per second were recorded at each frequency and the measured

data were compared with the calculations The position of the

airship was logged using the Global Positioning System (GPS)

The 3D track of the airship was restored using the GPS data

Possible measurement errors caused by the roll and pitch of the

airship (changing the radiation patterns of the transmitting

an-tennas) were below about 5 dB

The measured path loss was compared with the theoretical

model described in Section II-B Additional shadowing loss was

calculated as rooftop diffraction loss for a given geometry using

the UTD wedge diffraction For the UTD calculations the

build-ings were modeled as idealized blocks, in the same way as in the

simulations described above

Fig 8 presents the measured path loss from a trial in front of a

flat roof building at 2.0 GHz The modeled mean path loss is

de-picted by a dashed curve The shadowing effect of the building

Fig 9 Path loss measured in front of saddle roof building during an airship flyover The carrier frequency was equal to 5.5 GHz.

is accompanied by a significant fading The close conformity between the measurements and the model can be distinguished

in Fig 8

Fig 9 illustrates a comparison between the model and the measurement at a frequency of 5.5 GHz in front of a saddle roof building A high level of conformity was observed here as well The measurements proved that the shadowing path loss esti-mation based on the UTD presented in Section II was adequate

V CONCLUSION The narrowband elevation dependent shadowing model was introduced for mobile systems provided from high altitude plat-forms in built-up areas in the 2–6 GHz frequency band prospec-tive for 3G and 4G mobile systems The probability of LOS con-nections between the HAP and a mobile station and the addi-tional shadowing path loss for NLOS connections is presented

as a function of the elevation angle The probability of LOS is defined for four different types of built-up area, from suburban

to high-rise urban Simple empirical formulas derived from ex-tensive simulation results enable easy implementation of this model as a radio network planning tool for system-level sim-ulations and availability estimations A measurement campaign demonstrated the applicability of the new propagation model

REFERENCES [1] G M Djuknic, J Freidenfelds, and Y Okunev, “Establishing wireless communications services via high-altitude aeronautical platforms: A

concept whose time has come?,” IEEE Commun Mag., pp 128–135,

Sep 1997.

[2] T C Tozer and D Grace, “High-altitude platforms for wireless

communications,” IEE Electron Commun Eng J., vol 13, no 3, pp.

127–137, Jun 2001.

[3] “Revised technical and operational parameters for typical IMT-2000

terrestrial systems using high altitude platforms stations and CDMA radio transmission technologies,” ITU-R Study Groups Document 8-1/

307-E, Mar 1999, ITU.

[4] “Minimum performance characteristics and operational conditions

for high altitude platform stations providing IMT-2000 in the bands

1885–1980 MHz, 2010–2025 MHz and 2110–2170 MHz and in

regions 1 and 3 and 1885–1980 MHz and 2110-–2160 MHz in region

2,” Recommendation ITU-R M.1456, 2000, ITU.

[5] Air Interface for Fixed and Mobile Broadband Wireless Access

Sys-tems-Amendment for Physical and Medium Access Control Layers for

Combined Fixed and Mobile Operation in Licensed Band, IEEE

Stan-dard 802.16e-2005, 2005.

[6] “Propagation data required for the design of earth-space land mobile

telecommunication systems,” Recommendation ITU-R P.681-6, 2003,

ITU.

[7] M A N Parks, B G Evans, and G Butt, “High elevation angle

prop-agation results applied to a statistical model and an enhanced empirical

model,” IEE Electron Lett., vol 29, no 19, pp 1723–1725, Sep 1993.

[8] C Oestges, “A stochastic geometrical vector model of macro- and

megacellular communication channels,” IEEE Trans Veh Technol.,

vol 51, no 6, pp 1352–1360, Nov 2002.

[9] C Oestges and D Vanhoenacker-Janvier, “Coverage modeling of

high-altitude platforms communication systems,” IEE Electron Lett., vol.

37, no 2, pp 119–121, Jan 2001.

[10] P A Tirkas, C M Wangsvick, and C A Balanis, “Propagation model

for building blockage in satellite mobile communication systems,”

IEEE Trans Antennas Propag., vol 46, no 7, pp 991–997, Jul 1998.

[11] Z Blazevic, I Zanchi, and I Marinovic, “Deterministic wideband

modeling of satellite propagation channel with buildings blockage,”

IEEE Trans Veh Technol., vol 54, no 4, pp 1225–1234, Jul 2006.

[12] C Oestges and D Vanhoenacker-Janvier, “Propagation modeling and

system strategies in mobile-satellite urban scenarios,” IEEE Trans Veh.

Technol., vol 50, no 2, pp 422–429, Mar 2001.

[13] T Sofos and P Constantinou, “Propagation model for vegetation

ef-fects in terrestrial and satellite mobile systems,” IEEE Trans Antennas

Propag., vol 52, no 7, pp 1917–1920, Jul 2004.

[14] S R Saunders and A Argo-Zavala, Antennas and Propagation for

Wireless Communication Systems, 2nd ed. New York: Wiley, 2007.

[15] M Vazquez-Castro, F Perez-Fontan, and B Arbesser-Rastburg,

“Channel modeling for satellite and HAPS system design,” Wireless

Commun Mobile Comput., vol 2, no 3, pp 285–300, 2002.

[16] C Oestges, S R Saunders, and D Vanhoenacker-Janvier, “Physical

statistical modeling of the land mobile satellite channel based on ray

tracing,” Proc Inst Elect.Eng Microw Antennas Propag., vol 146,

pp 45–49, Feb 1999.

[17] “Propagation data and prediction methods required for design of

ter-restrial broadband millimetric radio access system operating in a fre-quency range about 20–50 GHz,” Recommendation ITU-R P.1410-2,

2005, ITU.

[18] P Ledl and P Pechac, “Area coverage simulations for millimeter point-to-multipoint systems using statistical model of buildings

blockage,” Radioengineering, vol 11, no 4, pp 43–47, 2002.

[19] C Cheon, G Liang, and H L Bertoni, “Simulating radio channel

statistics for different building environments,” IEEE J Sel Areas

Commun., vol 19, no 11, pp 2191–2200, Nov 2001.

[20] R G Kouyoumjian and P H Pathak, “A uniform geometrical theory of

diffraction for an edge in a perfectly conducting surface,” Proc IEEE,

vol 62, pp 1448–1461, Nov 1974.

[21] R J Luebbers, “Finite conductivity uniform GTD versus knife edge

diffraction in prediction of propagation path loss,” IEEE Trans

An-tennas Propag., vol 32, no 1, Jan 1984.

[22] [Online] Available: www.airshipclub.com

Jaroslav Holis (S’05) received the M.Sc degree in

radio electronics from the Czech Technical Univer-sity in Prague, Czech Republic, in 2005, where he is currently working toward the Ph.D degree His research interests are focused on physical layer

of 3G and 4G mobile systems and on radiowave prop-agation.

Pavel Pechac (M’94–SM’03) received the M.Sc

de-gree and the Ph.D dede-gree in radio electronics from the Czech Technical University in Prague, Czech Re-public, in 1993 and 1999, respectively.

He is currently a Professor in the Department of Electromagnetic Field, Czech Technical University

in Prague His research interests are in the field of radiowave propagation and wireless systems.