Weather Effects Impact on the Optical Pulse Propagation in Free Space

M. S. Awan*, Marzuki*, E. Leitgeb*, F. Nadeem*, M. S. Khan*, C. Capsoni**

* Graz University of Technology, Institute for Broadband Communication, Graz, Austria

** Poltecnico Di Milano, Milan, Italy

Abstract- Optical wireless links offer gigabit per second data we require first, based on theoretical models as well case
rates and low system complexity. For ground-space and or studies, to study, analyse and evaluate in detail the impact of
terrestrial communication scenarios, these links suffer from such weather effects on the transmitted optical pulse in free
atmospheric loss mainly due to fog, scintillation and space that jeopardize the performance of FSO links in terms
precipitation. We investigate here the impact of fog, rain and of availability, reliability and quality of service.
snow effects and evaluate their performance on the basis of
In this work, we investigate the impact of fog, rain and
attenuation data collected for the optical pulse propagated
through the troposphere. snow effects on the transmitted optical signal by evaluating
their attenuation behaviour based on measurement data
Keywords: Free-Space Optics (FSO), specific attenuation, collected at Graz, Milan, Nice and Prague. The data used in
visibility, empirical model. the analysis for Graz, Milan, Nice and Prague consists of six
months (September 2005-February 2006), two months
I. FSO INTRODUCTION (January 2005 – February 2005), eight days (24 June 2004 –
Free-Space Optics (FSO) or optical wireless 01 July 2004) and two months (September 2007 & December
communication is the concept of transmitting very high 2007), respectively.
bandwidth digital data using laser beam directly through the
II. THEORETICAL BACKGROUND
atmosphere. Recently FSO –links are identified as an
attractive alternative to the existing radio links for Transmitted optical pulses in free space are mainly
applications involving ground-to-ground (short and long influenced by two main mechanisms of signal power loss –
distance terrestrial links), satellite uplink/downlink, inter- absorption and scattering. Absorption is mainly due to water
satellite, satellite or deep space probes to ground, ground-to- vapours and carbon dioxide, and depends on the water
air/air-to-ground terminal (UAV, HAP etc.). Moreover, vapour content that is dependent on the altitude and
growing demands for higher data rates and wider bandwidths humidity. By appropriate selection of optical wavelengths for
from the end user to manipulate multimedia information in transmission the losses due to absorption can be minimised.
the recent years allegorises a challenge for the future Next It was found that scattering (especially Mie scattering) is the
Generation Networks (NGN). The prime advantages of FSO main mechanism of optical power loss as the optical beam
usage are: a wider bandwidth with data rates exceeding looses intensity and distance due to scattering. The beam loss
easily 100 Gbit/s using WDM techniques, low power due to scattering can be calculated from the following
consumption, better security against eavesdropping, better empirical, visibility range dependent, formula:
0.195V
protection against interference, and no frequency regulation 17 ⎛ 0.55 ⎞
α scat = ⎜ ⎟ (1)
issues [1]. V ⎝ λ ⎠
However, a well known disadvantage of FSO links in the where V is visibility range in km, λ is transmission
troposphere is their sensitivity to weather conditions – wavelength in nm. Visibility range is typically defined as the
primarily to fog and precipitation, causing substantial loss of distance at which the transmitted optical intensity is
the optical power over the channel path [2]. Even though, if attenuated to 5% of its original value [5]. The visibility range
fog, snow and rain concerns only a small fraction of the time V can be calculated from transmittance (τ) according to the
and or affects only a small portion of the overall propagation formula mentioned in the following equation [8]:
path yet it severely impairs the optical link performance and
limits the communication link availability primarily due to ln(τth ).dL −3.27m
the Mie scattering phenomena [3]. The main cause for such V (m) = ≅ (2)
ln(τM ) ln(τM )
significant losses is the sensitivity of the transmitted optical
signal to the drop size distribution of the atmospheric where dL is the link distance, τth and τM are the threshold of
particles of fog, snow and rain. Recent investigations showed transmission (5%) and the measured percentage of
that optical signal losses under dense maritime fog and transmission at a minute scale. To calculate the total received
moderate continental fog conditions reached up to 480 power and the link budget, following link equation:
dB/km and 120 dB/km respectively [4, 5]. Similarly, the d2
Prec = PT ( 2R 2 )τatm τT τR (3)
optical power loss under intense shower conditions can be as ΘT L
large as 30 dB/km [6], and for a heavy snow event the
attenuations can be more than 55 dB/km [7]. To counteract where dR is the transmitter aperture diameter, ΘT is the
these losses with enough excess power by FSO technology, transmitter beam divergent angle, L is the link distance,

978-1-4244-2517-4/09/$20.00 ©2009 IEEE 1

 τ atm is the atmospheric transmission factor, and τT & τ R are 3.91 λ − q
α Fog = ( ) (6)
the transmitter and receiver optics efficiency, respectively. V 550
To determine the probability of attenuation that it is equal where q is the parameter related to size distribution of the
to or exceeds a certain threshold value x dB during an droplets
arbitrary time interval t0 such that the optical link is
unavailable, is PAtot(x) given by the following relationship
[13]:
PAtot ( x) = P{ Atot (t 0 ) ≥ x} = EAtot ( x) (4)

where x is the normalized link margin such that x=x*/L In the case of rain, the optical power loss is relatively
(dB/km) and its value also represents the highest value of the wavelength insensitive. For rain, an accepted empirical
atmospheric attenuation, and x* is the link margin whose model based on the visibility range from [11] is,
upper bound is determined by the dynamical range of the 2.9
α Rain = (7)
receiver having a typical value of about 30-40 dB. PAtot ( x ) is V
the cumulative distribution of the total attenuation and The attenuation in case of snow can be more severe than
the attenuation in case of rain due to the much larger droplet
EAtot ( x) is the cumulative exceedance probability of the
size. Based on [12], the attenuation for snow based on
atmospheric attenuation. The percent availability of the FSO visibility range is,
links such that it fulfils the quality of service (QoS) 58
requirements can be expressed by the following equation: α Snow = (8)
V
The underlying electromagnetic theory behind the
Availability = (1 − PAtot ( x)) *100% (5) theoretical (microphysical) models is well known. For these
models, we assume that multiple scattering is negligible, and
In order to determine the reliability and the quality of the free
all the scatterers are acting independently and are distributed
space link, the attenuation coefficient α scat is considered a
uniformly along the atmospheric transmission path. For fog,
stochastic process whose distribution is empirically defined the attenuation of optical pulse is mainly due to Mie
about a lognormal distribution. It is important to mention scattering effect and the loss effects due to absorption can be
here that the availability and quality of service (QoS) of FSO ignored. Therefore the extinction coefficient is evaluated by,
links is strongly dependent upon and is influenced by various r2
weather effects especially fog, rain and snow besides the β Fog = ∫ πr 2Qext (m, r / λ)n(r )dr [dB/km] (9)
r1
wavelength used for transmission purpose [4, 5]. Where m and r are the refractive index and radius of the
fog droplets, respectively. Qext is the extinction efficiency and
III. EMPIRICAL AND THEORETICAL ATTENUATION MODELS
n(r) is the size distribution (i.e., modified gamma) of the fog
In FSO systems, modulating the instantaneous intensity
droplets. For rain conditions the usual equations are:
of a laser source data are transmitted and the received r2
intensity is detected with a photodiode. During propagation β Rain = 4.34∫ α scat (r )n(r )dr [dB/km] (10)
r1
of optical pulses in free space, photons are absorbed and
Where n(r) and r is the gamma distribution and radius of the
scattered by the atmospheric particles, such as fog, rain and
rain drops, respectively. The rain rate can be calculated as,
snow droplets. Generally, these droplets have radii in a range ∞
between 1 µm to 5000 µm, with drop-size concentration R = 4.8∫ r 3v(r )n(r )dr [mm/hr] (11)
0
decreasing sharply with increasing drop sizes [9] and cause
and the terminal velocity v (r ) of rain droplets in stagnant air
varying attenuation depending on their distribution.
However, the droplets is given by,
The Beers-Lambert law defines the optical power loss v(r ) = 9.65 − 10.3e −1.2 r [m/s] (12)
through the atmosphere that varies on the order of hours [10]. The model dealing with snow attenuations is,
The appropriate model for the attenuation coefficient (α) ρ R
varies depending on the presence of precipitation. Normally, β Snow = 0.3619 w [dB/km] (13)
C3vt
the optical links in free space measure visibility data with the
attenuation coefficient. In order to calculate the attenuations where terminal velocity vt of the snowflakes is,
caused by fog, rain and snow usually we rely on two gC3
approaches –empirical approach and the theoretical approach vt = 0.1155 [m/s] (14)
ρ a CD
(microphysical models). The empirical approach is very easy
and quick to implement whereas, the theoretical approach is Here, ρw is the water density in g/cm3, R is snowfall rate in
somewhat difficult and time consuming. The one very mm/hr, C3 is the thickness of snowflakes in g/cm2, ρa is the
commonly adopted empirical approach concerns computing
air density in g/cm3, CD is the drag coefficient for snow.
signal attenuations based on the visibility range estimate. In
the case of clear or foggy weather with no rain, hail or snow, Although snowflakes have a very complicated shape, yet
Kim’s model is employed to compute the attenuation due to they do not exhibit any preferred dimension, hence assuming
fog ( α Fog ), that is very accurate for the narrow wavelength the spherical shape of snow droplets seems reasonable.
Additionally, the sizes for rain droplets can be much larger
range between 785 – 1550 nm [10],

2
than 5 mm, but since they are hydro-dynamically unstable so
9
they tend to break into relatively smaller dimensions.
8
IV. IMPACT OF WEATHER EFFECTS AND THE ANALYSIS

Specific attenuation (dB/km)

Fall of the received optical power below the receiver 7

sensitivity threshold introduces a fade. This fade can be 6

described as short-term fade or the long-term fade depending
upon the time duration of the optical power interruption. The 5

type of fade that occur due to rain, fog and snow is the long- 4
term fade whose duration could be a few minutes or it can
reach up to several hours. In this contribution we would 3

focus our discussion only on the long-term fades introduced Prague - Rain attenuations at 850 nm
2
by fog, rain and snow conditions to the FSO link in free Milan - Rain attenuations at 785 nm (path)
space. We analyse here one month rain attenuation data from 1
Milan - Rain attenuations at 785 nm (V)

Prague, one snow event 0 5 10 15 20

Rain rate
25 30 35 40 45

Fig. 1 shows an empirical simulation of the rain, snow Fig. 2: Rain attenuations against a rain rate in mm/hr from rain events at
and fog attenuation effects against visibility range. From this Prague and Milan for FSO links operating at 850 nm and 785 nm
plot it is evident also that fog, rain and snow cause
significant propagation losses to the optical signals In order to assess the optical signal loss effects due to
transmission in free space. snow, we analyze three snow events that were recorded; one
800
long snow event at Graz during 25 – 28 November 2005 and
Kim Fog Model at 950 nm
the two in Milan on 18.01.2005 and 28.01.2005. The
Krus e Fog m odel at 950 nm
700
Rain attenuation at 950 nm maximum attenuation measured for snow at Graz reached to
600
Snow attenuation at 950 nm 55 dB/km with a mean attenuation of about 3 dB/km
averaged on a minute scale. While the changes in specific
Visibility in meters

500 attenuation occurred around ± 15 dB/km & ± 18 dB/km on a

400
minute and second scales, respectively. Whereas, for Milan
the maximum attenuations reached to 65 dB/km, a mean
300 attenuation of about 8 dB/km and with a change in specific
attenuation of about ± 16 dB/km. Figure 3 below, shows the
200
time series of these particular snow events on a second scale.
100 It is concluded that the optical signal loss effects due to snow
attenuations are much higher when compared with the rain
100 200 300 400 500 600 700 800 attenuations.
Attenuation in (dB/km )
Fig. 1: Simulation of empirical model for fog, rain and snow conditions for 70
Graz - Snow attenuations at 950 nm (25-28.11.2005)
FSO link at 950 nm
60 Milan - Snow attenuations at 785 nm (18.01.2005)
Milan - Snow attenuations at 785 nm (28.01.2005)
We will analyze here first the attenuations caused by rain
Spec ific attenuation (dB/k m )

50
to the propagation of optical pulses transmitted in free space.
We take a rain event recorded at Prague for the transmission 40

of 850nm pulses on 850 m link. From the analysis of this 30

measured rain event data, it was observed that rain
20
attenuation occurs for percentages of time smaller than
0.66% and the maximum measured value of 18 dB 10
corresponds to 0.0005% of time of year. The minimum
0
measured visibility during rain events is 1000 m. Fig. 2
shows two sample rain events occurred during the -10

attenuation measurement campaign of year 2006 – 2007 on

-20
18.09.2006 at Prague and at Milan during year 2005-2006 0 1000 2000 3000 4000 5000 6000 7000 8000 9000 10000

measurement campaign on January 17th 2005. The maximum Tim e in minutes

rain rate during this particular rain event was recorded to be Fig. 3: Snow events measured attenuations at Graz and Milan for a FSO
link during 25 -28 November 2005 and January 2005, respectively
about 44 mm/hr and the maximum attenuation reached
during this rain event was 8.5 dB/km at Prague. The
From the knowledge of the Mie scattering theory, we
maximum change in specific attenuations during the rain
know that fog is the biggest attenuating effect to the
events recorded were about ± 1.70 dB/km averaged on a
propagation of optical signal transmission especially in the
minute scale. While for Milan the maximum attenuation
troposphere. The time series analysis of the measured fog
reached was about 6 dB/km at a rain rate of about 3 mm/hr
(radiation and advection) attenuations for Graz (Austria),
with a maximum change in specific attenuation of about ±
Milan (Italy), La Turbie Nice (France) and Prague (Czech
0.7 dB/km. It is evident from these sample rain events that
Republic) have been already reported and discussed in details
rain events have minimum impact on the optical signal
[4, 6, 13, 14]. From these analyses, it was concluded that the
propagation.

3
fog attenuations are strongly dependent on the local climate It is evident from the above plot that the changes in
and the fog microphysics as the fog attenuations vary specific attenuation for dense maritime fog events are more
spatially and temporally with time. It was also observed that concentrated around 0 dB/km suggesting that most of the
on the average the changes in fog specific attenuation values time the fog was stable. We believe that the large changes
occur about ±6 – ±8 dB/km in case of moderate continental observed in the changes were mainly due to the wind
fog conditions like as in Graz, Milan and Prague. turbulence causing changes or the order of ±300 dB/km. The
Fog can be classified into many types on the basis of its Table 1 below summarises some basic statistics of the optical
formation mechanism, climatic conditions and the drop size links installed at the above mentioned four locations.
distribution. However, the two most common types of fog
are continental and maritime fog. The fog conditions at Graz,
Milan and Prague are the continental type while at La Turbie
Nice it is the dense maritime fog. The fog microphysics of
these two types of fog is very different, so is their
attenuations to the optical signal transmissions in these fog
environments. The attenuations in continental fogs reach up Table 1: Basic statistics of free space optical links at four locations
to 120 dB/km while in dense maritime fogs the average value * The range for atmospheric link losses attenuation for Milan is 21 dB
is about 480 dB/km.
The apparent changes in the average and the maximum
values of specific attenuations measured and mentioned
under Table 1 (for Graz, Milan and Prague) for continental
fog conditions are strongly dependent upon the fog
microphysics and also due to the local weather conditions.
However, despite Milan, Prague and Graz are located in the
same temperate area, we feel that there are two major
differences which are expected to affect to some extent the
fog attenuation measurements: a) the climate during winter is
colder in Graz than Prague and Milan, where daily
temperature minima are often below 0°C, while in Milan
temperature rarely falls below 0°C but average temperature
in Prague during winter remains -0.4°C , and b) Milan and
Prague are large cities the microclimate is that of a dense
Fig. 4: Fog attenuations comparison for Graz and La Turbie (Nice)
urban area. On the other side, Graz is a midsize town, with
the optical link being located in a suburban environment with
Fig. 4 above, shows the time series of these two fog types no tall buildings and many wide open areas around.
attenuations on a minute scale. The time series of these two Therefore it is reasonable that fog episodes are heavier in
fog types changes in specific attenuations shows that the Graz than in Milan and Prague. We present in Table 2, some
changes occur much more sharply in dense maritime fog statistics of all the fog events measured during the September
conditions than in moderate continental fog conditions. 2005 to February 2006 measurement campaign at Graz since
These changes in specific attenuations are normally between
the fog attenuations measured at Graz shows the worst case
± 6 - ± 8 dB/km under moderate continental fog conditions
fog events as compared with the Milan and Prague fog
with an average temperature change of about 10˚C during the attenuation events.
whole fog event, while these changes for dense maritime fog
environments can reach up to ± 300 dB/km as shown in Fig.
5 in the form of a histogram of changes in specific
attenuations.

Table 2: Fog episodes detected at Graz from September 2005 to February

2006. The last two columns show laser attenuation values measured during
fog: alongside with the 50% and 99% percentiles of the attenuation
distribution

Fig. 5: Histogram of changes in specific attenuations for La Turbie

(Nice) For Nice, the fog conditions are dense maritime type
that are not dependent on any particular season but are more

4
dependent on the marine environment. Fig. 6 as shown months winter starts to come and finish, respectively. So
below, presents the scatter of the fog attenuation data during there is a very high likelihood of the less severe or a kind of
a fog event measured on 28th July 2004 with 950 nm very light fog conditions. Additionally, the very high
wavelength as fog attenuations for Nice shows a worse fog attenuation individual events are mostly occurring during the
attenuation scenario. The measured data points are compared evening and especially very late night hours (i.e., during the
with the four commonly used empirical models to measure 18-24 time interval), while the attenuations are generally
the fog attenuations. It is apparent that the dependence of lesser in the 00-06 time interval. From this data set, however
specific attenuation on visibility for dense maritime fog does it is very difficult to clearly identify the regions and time
not converge to a particular model. events of very high attenuation levels.

V. CONCLUSIONS
This contribution explains the impact of rain, snow and fog
effects on the propagation of optical wireless links in free
space. It was found out that for optical wireless links in the
troposphere, fog is the most limiting factor as compared with
the losses incurred by rain and snow.
This paper provides some useful statistics of the fog
attenuations from the measured attenuations of terrestrial
optical wireless links. We also tried to evaluate the atmospheric
environments of Graz, Milan, Nice and Prague on the basis of
three weather conditions i.e., rain, snow and fog and studied
their effect on the optical signal propagation. Obviously, more
attenuation data is required to make an in depth analysis of
Fig. 6: Measured attenuation comparison with different empirical models
these links under fog, rain and snow conditions. Additionally,
we tried to figure out the time intervals of fog evolution, its
Since the fog attenuations at Graz shows a worse
dissipation and its occurrence based on the fog attenuation
scenario of fog attenuations under the moderate continental
trends. This information can prove very useful when to apply a
fog conditions, we draw a contour plot of the measured
certain fade margin, that can have a significant impact on the
attenuations of six months duration in order to visualize their
optical wireless links availability and improving the quality of
diurnal behaviour. This knowledge can be handful to identify
service (QoS) demands for applications requiring high data
the time intervals that can have the highest & lowest
rates communication.
attenuation values for such type free space optical links. This
information would consequently be useful in the link budget REFERENCES
designs for locations having such fog environments.
[1] G. Hyde and B. I. Edelson, “Laser satellite communications: Current
status and directions,” Space Policy, vol. 13, pp. 47–54, 1997.
[2] McCartney E. J, “Optics of the atmosphere. Scattering by Molecules and
Particles”, J. Willey & Sons, 1977.
[3] C. F. Bohren and D. R. Huffman, “Absorption and scattering of light by
small particles”, (John Wiley and sons, 1983).
[4] M. Gebhart, E. Leitgeb, S. Sheikh Muhammad, B. Flecker, C. Chlestil,
M. Al Naboulsi, H. Sizun, F. De Fornel, “Measurement of light
attenuation in dense fog conditions for optical wireless links”, SPIE
proceedings, Vol. 589, 2005.
[5] S. Sheikh Muhammad, B. Flecker, E. Leitgeb and M. Gebhart,
“Characterization of fog attenuation in terrestrial free space links”,
Journal of OE 46(4) 066001 June 2007.
[6] C. Capsoni, R. Nebuloni, “Effects of Rain on Free Space Optics”, 11th
Symposium on Radio Wave Propagation and Remote Sensing, Rio de
Janeiro, 30 Oct-02 Nov 20007, Brazil.
[7] R. Nebuloni, C. Capsoni, “Laser attenuation by Falling Snow”,
CSNDSP 2008, 23-25 July 2008, Graz, Austria.
[8] Art MacCarley: “Advanced Image Sensing for Traffic Surveillance and
Fig. 7: Contour plot of fog attenuations for Graz at 950 nm Detection”, California PATH Research Report, 1999.
[9] Van de Hulst: “Light scattering by small particles”, 1957.
We plotted the collected six months fog attenuations data [10] I. I. Kim, B. McArthur and E. Korevaar, “Comparison of laser beam
propagation at 785 nm and 1550nm in fog and haze for optical wireless
for Graz in Fig. 7. From this contour plot we can see, that communications,” Proc. SPIE, vol. 4214, pp. 26-37, Feb. 2001.
there are few individual high attenuation events scattered in [11] D. Atlas, “Shorter Contribution Optical Extinction by Rainfall,” J.
the plot. However, it is difficult to conclude from this plot Meterol., vol. 10, pp. 486-488, 1953.
any particular time slots when the fog attenuations are [12] H. W. O’Brien, “Visibility and Light Attenuation in Falling Snow,” J.
Appl. Meteorol., vol. 9, pp. 671-683, 1970.
particularly high. From this plot, it is clear that especially in [13] B. Flecker et. al: Results of measurements for optical wireless channels
October to end of January the overall attenuations are high under dense fog conditions regarding different wavelengths, SPIE
most of the time in the range of 120 dB/km - 180 dB/km on a proceedings, vol. 6303, 63030P, 2006.
second scale. Also in the start of September and the end of [14] V. Kvicera, M. Grabner, O. Fiser, “Visibility and Attenuation Due to
Hydrometeors at 850 nm Measured on an 850 m Path”, CSNDSP 2008,
February the attenuations are comparatively lesser as 23-25 July 2008, Graz, Austria.
compared the other months. The reason could be that in these