You are on page 1of 13

ain-Cell Identification and Modeling for

Propagation Studies from


Weather Radar Images
V. Pastoriza
1
, A. Nuiiez
2
, P. Mariiio
1
, F. P. Fontim
2
, and U.-C. Fiebig
3
1University of Vigo, Electronic Technology Department
Campus As Lagoas, 36310, Vigo, Spain
Tel: +34986818717; E-mail: vpastoriza@uvigo.es; pmarino@uvigo.es
2University of Vigo, Signal Theory and Communications Department
Campus As Lagoas, 36310, Vigo ,Spain
E-mail: fito@uvigo.es;fpfontan@tsc.uvigo.es
3German Aerospace Center (OLR), Institute of Communications and Navigation
0-82230, Wessling, Germany
E-mail: uwe.fiebig@dlr.de
Abstract
_. a'ms to characterize and model the rain fields for propagation studies from weather radar images of the northwest
S::ain. The study includes the modeling of these images according to two well-known rain-cell models, EXCELL and
~ r this analysis, a procedure for detecting rain fields and an algorithm for identifying rain cells within the images is
This work is part of an ongoing activity, trying to develop dynamic space-time models of rain fields for simulating
........~ - < : . ~ . J effects on millimeter-wave communication systems.
=Iectromagnetic propagation; radar data processing; rain cell modeling; rain; millimeter wave radio propagation
_ ......,-;,1 \" "'1 factors
1. Introduction
-nowledge of the structure of rain cells is important in
- ":eling of rain-induced propagation effects, for radio
_.::ion applications at frequencies above about 10 GHz
-5jJRce and terrestrial applications [1-7). Rain-induced
- ;::an be measured using satellite beacons, or, to a point,
" .,eters. This kind of experimental work can be per
:. at a few geographical locations, and for a limited
. -- :'requencies and link geometries. Unfortunately, the
:.:!l ed can not easily be extrapolated to other locations.
_ . e this problem, several attenuation models based on
5".::al data have been developed, which provide adequate
that allows the estimation of link margins for locations
_ "': the world [8] . Moreover, new satellite networks will
- ':nic-reconfiguration techniques, which must be ade
.:-.::gned. This requires the use of space-time models for
= :.'":e mentioned design.
_ .::escription of the horizontal structure of rain fields is
: ex, due to its high space-time variability. This explains
-;::al-statistical nature of most studies perfoTI11ed on this
:everal research groups have developed i"ain-cell models
_ -- a -radar data. These are of a statistical nature, but do
;-'e construction of two-dimensional rain fields [10-14].
" as and Propagation Magazine, Vol. 52, No.5, October 2010
Their application in the study and characterization of atmospheric
propagation enables evaluating the following at any location:
diversity gain, terrestrial-path attenuation, or slant-path attenuation
for different azimuth and elevation angle directions. Moreover, the
short-term forecasting of rain rate to derive estimates of rain-field
attenuation, if integrated into a fade-mitigation and resource-man
agement scheme, could lead to an increase in system uptime.
Simple rain-cell characterization can be done using only a
limited number of parameters. Simplicity is an inherent advantage
over sophisticated techniques, and reduces the computation time
and the storage space necessary for extracting model parameters
from the radar observations [9]. In most studies, the mathematical
representation of rain cells follows a stochastic approach, describ
ing rain cells as randomly generated entities that behave similarly
in a statistical sense - to the observed rain cells. The work reported
here divides the observed rain fields into sets of rain cells that are
described via simple mathematical formulations.
Two fairly simple models found in the literature allow a
representation of their essential hydrological characteristics: geo
graphical extent and small-scale variability. These two models
characterize rain cells by means of an ellipse. The EXCELL model
assumes that intensity within the cell decays exponentially from a
maximum at its center [15] . The HYCELL model describes the
intensity decay as following a combination of an exponential func-
ISSN 1045-9243/20101$25 2010 IEEE 117
tion and a Gaussian function (quadratic exponential) [9]. Two par
ticular cases have been identified for the HYCELL model: a purely
exponential case and a purely Gaussian case, for which the cells
can be characterized by means of three parameters each, instead of
the required seven parameters for the general HYCELL model.
This paper is organized as follows. First, an introduction on
rain-cell modeling is presented. The experimental database of
weather-radar images (WRIs) used is then described. How rain
cells are identified within weather-radar images for model param
eterization is next discussed. Finally, the results of this study are
presented.
2. Rain-Cell Modeling
The term "rain cell" is frequently found in the meteorology
and hydrology literature. However, there is not a widely accepted
definition for this concept [16]. The terms "rain cel!" and "rain
area" are generally found in studies on the structure of rain fields
based on weather-radar data. The term "rain cell" is often the pre
ferred term when there is an implicit or explicit reference to the
dynamic structure in the rain phenomenon. However, the term
"rain area" is more appropriate when the number of peaks within
these entities is not taken into consideration as a classification cri
terion [9] . In this paper, where the interest is in the effects of rain
on radio-communication systems, the term "cell" is used, as many
others have done in the past when referring to this issue [9, 12, 15,
17-20].
Rain cells are basically defined in two different ways in the
literature. The first definition is as the region inside a given con
tour with a local reflectivity peak, where the reflectivity values
inside the region are greater than a given threshold that is a func
tion of the peak value. Examples include -10 dB in [10], -6 dB in
[21], and -3 dB in [22]. Second, the cell consists of a region within
which the intensity, R (or the reflectivity, Z), is greater than or
equal to a fixed threshold, r [11, 16,23-25]. In this paper, we use
the second definition, and neglect the rain intensity below the
threshold, r .
The modeling of rain cells must take into consideration their
geometric shape and their horizontal intensity distribution. Circular
cell models, where the diameter is equal to that of an equivalent
circle of the same area, and a uniform-rate horizontal distribution,
were proposed in [26-29]. Circular-cell models with horizontal
distributions including uniform, exponential, and Gaussian were
assumed in [30-31] . A circular-cell model with Gaussian horizon
tal rain-rate distribution was proposed in [32-35].
In [9, 15, 36-42], the shape of the cell was described by
means of an equivalent ellipse, the parameters of which are the
semi-major and semi-minor axes, and the ellipse's orientation. The
horizontal distribution of the intensity was characterized by means
of a uniform function [36], an exponential function [15, 37, 40,
42], a Gaussian function [41], or a combination of the last two [9].
In the literature, rain cells are also classified into two catego
ries as a function of the horizontal distribution of the intensity
within them: "stratiform cells," with a gentle decay from the center
with a maximum rate below 10 mm h -I , and "convective cells,"
which show areas of intense rain, greater than 10 mm h -I [43-45].
Moreover, convective cells are generally surrounded by a strati
form area of lower intensity.
2.1 Modeling the
Horizontal Structure of Rain Cells
The rain fields observed in each weather-radar image
into rain cells that can be identified by applying the algoritm
cussed in Section 4. Each identified cell is then modele.::
parameterized using a mathematical description. In this stue;
rain-cell models have been tested: EXCELL and HYCELL.
characteristics are summarized below.
Capsoni et al. [15] proposed the EXCELL model to c,"
the horizontal structure of rain cells by means of an ellipse
rate, R, that exponentially decays around a single
analytic expression for the horizontal distribution of R
cell in the horizontal plane (Oxy) follows the expression
2 J1/2]
R(X,y)=REexPr-
[

2
,R?:.R2 ,
L
where RE is the peak intensity, and aE and bE are the
along the Ox and Oy axes for which the peak decays by a :"1:,
1/e. R2 is the minimum value considered (Figure 1). The ::,;
mulation of the model was valid for a value of R2
5 mm h-
I
[15]. Later, this value was extended by Awak3.
0.4 mm h-
I
, and by Goldhirsh [40] to 0.5 mm h-
I
. In
sis, we have set the minimum level to 1.01 mm h-
I
(23. 8
example of a rain cell using the EXCELL model is ilh.;,::--,
Figure 2.
As discussed above, in the EXCELL model a cell is :
characterized by means of the three parameters in
R
E
, aE, and bE' The parameter R2 remains cons!c:
parameters must be estimated on a cell-by-cell basis, ane :
rion used for such estimation [15] involves minimizing :::e ..
rain-intensity error between the actual and the modeled
by Equation (2):
(REI -II,
S= I(Rr I
- - - - . ----r------, ---,--------
sol R
45
40
35
E 30
"'

25
a:
c
"m
a:
15
10'
ot

-8 -2 0
Oistance X tk..rn,
Figure 1. The EXCELL model: t he --ertical pl aoe
Ox axis. The vertical axis is r ain r:ne in mmb: t
axis is distance in km.
JEEE .. E..-.:
118
The analytical expression for the distribution of rain rate
within the cell in the horizontal plane (Oxy) of the HYCELL model
is described by
X DislmCe!km)
of an EXCELL rain cell with the
aE = 1.67 km, bE = 3.34 km, and
~ - - - - - - ~ - - - - - - - - ~ ~ J
I
-4 -2 0
Dislance X (km )
__ The HYCELL model: the vertical plane along the Ox
is the EXCELL-modeled rain-intensity mean value,
- the mean measured intensity in the cell.
EX CELL model is extensively used. However, it pre
.e shortcomings. In the original model, the cell intensity
zero at infinity. To overcome this problem, a modified ver
e model, called "lowered EXCELL," was developed. This
--:e rain-intensity profile to reach a zero intensity value at a
-.ance from the center [6, 46].
- ' :her shortcoming of this model is that the horizontal
.)n of intensity in the vicinity of the peak does not in real
_ abruptly as an exponential function. It seems more real
_se a Gaussian function for the higher intensity values in
-"'f of the cell. Feral et al. [9] defined a model similar to
-=.. to overcome this issue. In the HYCELL model, the hori
m cture is also elliptical, while the vertical profile follows
.: Gaussian-exponential law (Figure 3). From a conceptual
" \ iew, the Gaussian component describes the convective
_': the cell (area of elevated rates), while the exponential
ent characterizes the stratifonn region that surrounds it.
="'ents the peak from taking on values that are too high,
......' ., .. <0' " with those recorded by the radar.
-ennas and Propagation Magazine, Vol. 52, No.5, October 2010
(3)
where the parameters R
G
, aG, and b
G
define the Gaussian
component. Respectively, they are the peak of the rain intensity,
and the distances along the Ox and Oy axes for which the rain rate
decays by a factor of l/e with respect to RG _ Similarly, the
parameters R
E
, aE, and bE define the exponential part as in the
EXCELL model. R
J
is the crossover point between the Gaussian
and the exponential regions_ The validly domain for this hybrid
model is R ;:o: R
2
, with R2 = I mm h-
1
. The value of R2 is in line
with the threshold of 1.0 I mm h - 1 used in the analysis presented
here.
In the EXCELL model, one cell is completely characterized
by means of three parameters, R
E
, aE' and bE' However, the
HYCELL model requires seven parameters, i.e., R
G
, aG' b
G
,
R
E
, aE, bE' and R
I
. These parameters have to be estimated on a
cell-by-cell basis. The criterion used here in this analysis is the
same as that used in [9], where the error, S, is minimized. This
error is defined as follows:
that is, the sum of the absolute values of the nonnalized rain-inten
sity mean value, \ R), the root-mean-square value, ( R2 ) , the
mean, \ C), and the root-mean-square value, (c
2
), of the horizon
tal rain-intensity gradient. Subscripts r and H refer to the radar
image and the HYCELL model, respectively.
Three special cases can be distinguished in the HYCELL
model, as a function of the relative value of R
J
: a hybrid model ,
when R2 < R1 < RG ; a purely exponential model, when
R
J
= RG = RE (in this case, the HYCELL model is the same as the
EXCELL model) ; and a purely Gaussian model , when R1 = R
2
. In
the last two cases, one cell is fully defined by only three parame
ters: R
E
, aE , and bE' or R
G
, aG, and b
G
, respectively .
In the analysis performed, HYCELL model parameterization
of rain cells was only carried out using the purely exponential and
purely Gaussian special cases. This decision was made because the
cells in this study will be used at a later stage in a space-time
dynamic model, where rain fields will move according to gi ven
advection laws. In this extension, related to the one in Feral et al.
[9], the idea is to work on rain cells the parameters of which
change in time as they move through the simulation scenario. To
do this, it is very convenient to reduce the complexity of the vari
ous elements used in the overall modeling.
119
3. Experimental Data
The data used in this research were weather-radar images
provided by the meteorological radar that the Spanish Meteoro
logical Agency (AEMET) owns at Cerceda (Figure 4), on the
northwest coast of Spain (43.1690 N, 8.5261 W). The climatic
characteristics in this area correspond to a temperate maritime cli
mate, subjected to frequent frontal systems arriving from the
Atlantic Ocean. These weather-radar images corresponded to plan
position indicator (PP!) images from a C-band Doppler weather
radar. The images contained reflectivity (Z) values in d8Z, where
6
Z = 10 loglo (z), with z in mm mm-3. The space resolution of the
data was 1 km x I km, while the time resolution was 10 min [47).
Table I summarizes the radar parameters. The radar reflectivity, z
3
(mm
6
mm- ), was converted into rain rate, R (mm h-
I
) . Several
empirical Z-R conversion relationships are available in the litera
ture [48). The expression Z = 238R
I5
is considered by the opera
tors, the AEMET, as the most appropriate for the type of rain seen
by this radar. .
The data were corrected (attenuation, clutter, beam-blockage
compensations), and a 121 km x 12 I km region inside the weather
radar images was selected. Finally, the data were transformed for
convenience into pseudo-reflectivity units (pZ), to work with a
limited number of discrete integer values of reflectivity (0 to 127)
[47]. The conversion rule was as follows:
0, Zs12.8d8Z
pZ= 10(Z-12.8)/4, 12.8 < Zs63.6dBZ (5)
127, Z;?: 63.6 dBZ
1
The initial database consisted of corrected 121 km x 121 km
weather-radar images, recorded every 10 minutes from November
2003 to April 2008. From this data set, weather-radar images with
a significant rain geographical extent and a duration of several
hours were selected. This selection was performed according to
[29], based on the "rain coverage" in each weather-radar image.
The rain coverage is defined as the percentage of pixels in the
weather-radar image with rain intensity above or equal to a given
threshold with respect to the total number of pixels in the weather
radar image. For this work, a threshold of 18 pZ (19.8 dBZ,
0.54 mm h-
I
) was established to separate dry pixels from wet pix
els. The whole procedure of data selection is described next.
Table 1. The general characteristics of
Cerceda's weather radar.
General Characteristics
Latitude 43 10' I" N
Longitude 8 31' I" W
Height 608 m asl
Wavelength 5.33 cm (C-band)
D0I!E!ler-Mode Characteristics
Pulse width 0.5
Scan elevation 0.5
0
Pulse repetition rate (PRR) 900 Hz-1200 Hz
Main lobe width 0.92
Observation range 120 km
Space resolution I x I km
2
10 min
I
Figure 4a. The location of weather radars throughout [
The location ofCerceda's weather radar is marked by
01 Apr 2005 1704 UTe
Figure 4b. An example of a weather-radar .
Cerceda's weather radar.
-' =J- '
03 fl' \
02 . . . . . j' \ __
0.1 .. ... .. . . .\
" ! .. -- ,
o /- - -- =====:-
'2- -
- ... ...
Figure 5 . .-\n exa mple of the co\erage and
age 11["
:: 6-: _ _-:
120
- : ..- ents" were identified as a function of their coverage:
.-: started when, in the image, the raincoverage time
.up a given threshold, and ended when it crossed
=- eshold. The "event duration" or "lifetime" was the
.' between these two crossing points. To avoid small
3! the edges of the coverage threshold - which cause
ncations of very-small-duration rain events - a
peration was performed on the coverage time series.
',n consisted in a running mean of the last seven points
i.e. , an average of the coverage in the last hour.
____""-""on was only done when at least five of the seven
images were available. Otherwise, coverage
not available. Here, a threshold of 10% for the
_, '. erage was set for identifying events. Figure 5 shows
a coverage time series before and after smoothing
\Iarch 25, 2008).
.ing the selection method described above, a set of 267
_ initially identified. After the smoothing step, only 196
_ _ for further analysis. From these events, 101
.' a lifetime of over three hours - were selected. For
concentrated on those events that did not exceed a
_: ncoverage value of 22%. Finally, 25 events were thus
- slUdy.
4. Identification of Rain Cells
step in modeling rain cells consisted of the
: :1 of all cells within each weather-radar image. The
.. algorithm presented in this paper only used the
\ alues to define rain cells. The algorithm developed
- 10 a group of algorithms [21, 23, 25, 49-54] used for
::-precipitation-forecast (QPF) methods. These algo
a rain cell as a contiguous region that exceeds a cer
.... ity threshold and outranges a certain minimum size.
.;'.<: specific algorithmic method differs from algorithm
-:. A brief description of these procedures was given in
_:OUT [50, 51], a rain cell is defined as such if pixels
a certain reflectivity threshold directly touch.
_:E [21] determines rain cells by drawing a contour con
: 6 dB below each enclosed reflectivity maximum.
::1 employs a mass-centroid approach, with a single
:. t is chosen by the user. SC1T [52] uses a mass-cen
with seven thresholds (30, 35, 40, 45, 50, 55,
. define rain cells. TRACE3D [53] applies a first thresh
.:: the weather-radar image into "clusters." Within each
_ second threshold, which depends on the maximum
. in the cluster, is then used to define rain cells belonging
- er. In [25] and [54], the identification of the rain cells
---;ned by searching for the maximum reflectivity, adding
pixels (if their values were above a certain threshold),
' ::, the number of cells (by merging them if the distance
. ;-nate, or removing them if their area did not exceed a
- -,mum size).
- , work, a hierarchical three-level rain-intensity structure
: <:red . The lowest level corresponded to rain cells that
d into clusters (second level), which, in tum, were
--::0 rain events (third level). Rain-rate-threshold criteria
10 define the cells and clusters. In tum, events were
on the coverage method described in Section 3.
as and Propagation Magazine, Vol. 52, No.5, October 2010
A cluster into a given weather-radar image of a rain event
was defined as the region made up of pixels with a reflectivity, Z,
that was higher than or equal to a fixed threshold of 18 pZ
(19.8 dBZ, 0.54 mm h-
I
). This threshold was the same used previ
ously to separate dry from wet pixels in a given weather-radar
image to calculate its coverage. A rain cell within a cluster was
defined as the pixel region with pixel reflectivity, Z, that was
higher than or equal to a fixed threshold of 28 pZ (23.8 dBZ,
1.0 I mm h - J). However, rain-cell identification within a weather
radar image is not always an easy task, as it might seem from the
above criteria, and several additional operations were required.
The following is a description of the algorithm used for
identifying the rain cells for each weather-radar image. This
description will be illustrated by an example for better under
standing (Figure 6). The example corresponds to a 40 km x 16 km
region, part of a weather-radar image from the database (Fig
ure 6a). The rain-cell identification algorithm ran as follows:
I. First, "groups of pixels" with reflectivity values higher
than or equal to 28 pZ (23.8 dBZ, 1.01 mm h-
I
) were
identified. A group was understood as a contiguous
area. In Figure 6b, five groups were identified, and they
are shown in different colors.
2. Next, " pixel subgroups" with reflectivity values higher
than or equal to 55 pZ (34.6 dBZ, 5.28 mm h -I) inside
the previous group were determined. If only one sub
group within a group existed, then the whole group was
considered as an "initially isolated cell." In other cases,
pixels that did not belong to any subgroup (pixels with
values between 28 pZ and 54 pZ) were associated with
the closest subgroup. The closest subgroup was the one
with a center of mass that was the closest to the pixel
under consideration. In the example, only the largest
group contained several - in this case, three - sub
groups (Figure 6c) .
3. After the above operation, it could happen that a sub
group was formed by more than one pixel aggregate
(the black subgroup in Figure 6c was made up of three
pixel aggregates). To avoid this situation, it was
checked that each subgroup contained only one aggre
gate. If not, the following steps were performed:
a. First , the "main aggregate" (the one with the
maximum number of pixels fulfilling p Z 55)
was identified, and this aggregate was considered
as the subgroup.
b. Then, each pixel from the other subgroup aggre
gates was assigned to a different subgroup next to
the aggregate to which it belonged. If there were
several subgroups next to the aggregate, then the
pixel was assigned to the subgroup with the clos
est pixel.
c. The process was repeated until there was only one
aggregate in each subgroup.
In the example, the black initial subgroup obtained for
the group with the larger area (Figure 6c) was corrected
through this process. Resulting subgroups are shown in
Figure 6d.
121
(a) (b) (c) (d) (e)
(f) (g) (h) (i)
Figure 6. The sequence of operations for the rain-cell identification algorithm.
These subgroups from each group were now considered 7. Initial grouped cells with an area smaller than 8 L
"initial grouped cells." [n a previous step, groups with a became part of one of the grouped cells with whi.:
unique subgroup were already considered initial iso was in contact. None of the three grouped cells ir.
lated cells. Figure 6e shows the "initial cells" identified example satisfied this condition (Figure 6g).
in the example.
8. Rain cells with an area smaller than 8 km
2
4. Next, initial isolated cells in contact with the image bor
removed (Figure 6h).
ders and with an area smaller than 8 km
2
(eight pixels)
were removed. [n the example, the lowest left cell of 9. Finally, all the cells in contact with the image
Figure 6e was removed (Figure 6f).
and with an area smaller than 12 km
2
were remc'
None of the three grouped cells in the example sa!:,
5. Then, each remaining initial isolated cell with an area
this condition (Figure 6h).
smaller than 8 km
2
was merged with its closest cell in
a three-pixel distance. Otherwise, the cell was kept
The resulting weather-radar image, after applying
unaltered. [n the example, the lowest centered cell of
cell identification algorithm, is shown in Figure 6i.
Figure 6f became part of the one just above it (Fig
ure 6g).
This algorithm was applied to the 25 selected e\e-..:
ther rain-cell parameterization. It must be borne in mine ::
6. Each isolated cell after this operation was merged, if it
the definitions and criteria used - i.e., rain cells with pl _
existed, to the neighboring cell with a center of mass
less than 5 km from the center of mass of the isolated
clusters with pZ 2 18 - once cells were identified,
cell. Otherwise, the cell remained unaltered. None of resided within clusters; however, clusters could exist '.' :
the two isolated cells in the example satisfied this con cells. In this case, they were considered clusters that w
dition (Figure 6g). or driving away, and they were not used in this study.
IEEE Antennas and Propagation Magazine, Vol. 52, NO.5. 122
Table 2. A summary of all 25 events selected for analysis.
I
I
I
I
1'1
[ vent
Date Time
WRIs
d
Total Total
I 01 DEC 2003 [07:14,13:241 38 443

18 DEC 2003 [08:24, 13 :34J 32 500
-
,
20 JUN 2004 [ 18:04, 21:341 22 252
.!
13 JAN 2005 [03: 14, 06:44] 22 135
5 15 JAN 2005 [04:14,12: 14] 49 324
6 05 FEB 2005 [10:24,13:541 22 163
-
26 MAR 2005 [19:24,23:24] 25 414
8 01 APR 2005 [16:54,21 :54] 31 215
9 15 APR 2005 [08:44, 14:341 36 563
:0 23 APR 2005 [01 :04, 05:14] 26 280
il 14 MAY 2005 [08:04, II :541 24 315
:1 17 OCT 2005 [03:44,07:04] 21 135
!3 27 NOV 2005 [01 :34, 11 :24J 60 670
02 DEC 2005 [07:54,14:341 41 413
15 02 DEC 2005 [15:34,20:44] 32 466
16 31 DEC 2005 [15:24,23:141 48 589
! 7 15 JAN 2006 [16:24,20:041 23 388
18 16 FEB 2006 [10:44,14:54] 26 223
19 15 APR 2006 [07:54, 12:041 26 322
20 16 APR 2006 [10:04,13:441 23 252
21 21 OCT 2006 [07:24,13:14] 36 257
01 DEC 2006 [07:34, II: 141 23 124

_.J 30 DEC 2006 [10:24, 13:34] 20 136
14 30 MAR 2007 [19:04,22:24] 21 235
25 27 FEB 2008 [04:44,08:041 21 135
Clusters Rain Cells Cells per Cluster
Mean STD Total Mean STD Mean STD
11.66 3.06 785 20.66 5.63 1.77 1.61
15.63 5.09 934 29.19 6.03 1.87 2.06
11.45 4.94 386 17.55 5.07 1.53 1.16
6. 14 2.10 279 12.68 2.85 2.07 1.61
6.61 3.09 801 16.3 5 6.03 2.47 2.75
7.41 1.97 336 15.27 3.38 2.06 2.07
16.56 4.09 865 34.60 7.80 2.09 3.24
6.94 3.77 689 22.23 9.77 3.2 4.74
15.64 3.31 1090 30.28 6.92 1.94 2.17
10.77 3.22 608 23.38 6.66 2.17 2.26
13.13 2.66 664 27.67 5.79 2.11 2.01
6.43 2.50 243 11.57 3.72 1.8 1.39
11.17 3.72 1350 22.50 5.34 2.01 1.95
10.07 2.97 1138 27.76 8.46 2.76 3.29
14.56 3A9 855 26.72 6.41 1.83 2.17
12.27 2.56 1202 25.04 4.62 2.04 1.72
16.87 4.31 625 27.17 5.81 1.61 1.22
8.58 3.53 513 19.73 5.13 2.3 2.63
12.38 2.37 739 28.42 5.71 2.3 2.83
10.96 2.58 541 23.52 5.21 2.15 2.46
7. 14 2.52 675 18.75 6.01 2.63 3.65
5.39 2.27 266 11 .57 3.27 2. 15 1.99
6.80 1.99 247 12.35 2.74 1.82 IAI
11.19 3.14 506 24.10 6.97 2.15 3.04
6A3 1.69 282 13.43 4.93 2.09 1.70
: "rovides a summary of the number of clusters and
-=" in each one of the 25 events in our database, fol
considerations and criteria. In this table, for
;..,-:e total number of identified clusters, the average
:. ;.he standard deviation per weather-rad\lr image, as
_ e:age number and standard deviation of the cells per
:- ":5 listed. Additionally, in the same table, we indicate
- date, and time, and the total number of weather
_= of which it consists. From Table 2, it could be
the average number of clusters per weather-radar
:-::ween SA and 16.9, the average number of cells per
......= image was between 11.6 and 34.6, and the average
per cluster was between 1.53 and 3.20.
ain-Cell Model Parametrization
-:-sent the results of thi s study in this section. The ratio
_ ;:1 [9], was used for quantitatively comparing the
d ing results. The ratio .lml-m2 between cell parame
for two given models, ml and m2, is defined as
(6)
_-,J! ratio between the sums of the errors of the statistics
:. . (G), and (G
2
) - of each model with respect to
-easurements, r. These integral parameters of the cell
are the mean, (R), and root-mean-square, (R2) , values of the rain
rate, R, and the mean, (G), and root-mean-square, ( G
2
), values of
the horizontal gradient of R. If .l
m
l-m2 is greater than one, then
Model 2 is better than Modell, and vice versa.
First, following the steps given in [9] for two populations of
rain cells - one in Bordeaux, France, and another in Karlsruhe,
Germany - the EXCELL (ex), pure exponential HYCELL (he),
and pure Gaussian HYCELL (hg) models were parameterized for a
population of rain cells on the northwest coast of Spain. For greater
reliability, modeled rain cells with a minor axis for the ellipse
smaller than 2 km, a major axis larger than 50 km, or those cells in
contact with the borders of the image were excluded. Then, the
ratios .lex-he' .lex-hg' and .lhe-hg were computed. Finally, for
each ratio, the mean value for each event was calculated (Table 3).
Moreover, from the he and hg models, a different analysis
was done. This new analysis (heg) consisted in selecting for each
cell the best model (he or hg). Table 3 also shows the mean value
of the ratio .lex- heg for each event, and the percentage of cases for
each event where the he and hg models were obtained as the best
models in heg modeling.
For each ratio, the last row of Table 3 shows the mean value
for the total set of 25 events selected. Results indicated that the he
and hg models behaved generally better than the ex model. The heg
results were much better than the he or hg with respect to the ex
model. This better parameterization achieved with the heg model
ing will then improve route diversity and site diversity for terres
trial systems, for example. From this table, it could also be appre
ciated in the heg modeling how cells were better described (higher
percentage) using the hg instead of the he (heg model). Some
"'_r-ao; and Propagation Magazine, Vol. 52, No. 5, October 2010 123
table 3. A ratio comparison for the four models in the Results indicated that in general, on the northwest coast 0,0
25 selected events. the HYCELL models performed better than the EXCELL m
Event
Xer-he Xer-hg Xhe- hg Xex-heg
heK Modeling
# %he %hK
I 1.87 2.77 1.24 2.94 46.21 53.79
2 2.20 3.32 1.29 3.51 39.43 60.57
3 1.96 2.69 1.22 2.85 41.97 58.03
4 1.15 1.22 0.98 1.39 70.89 29.11
5 1.43 1.80 1.10 1.99 54.55 45.45
6 1.36 1.52 1.00 1.72 63.16 36.84
7 2.08 2.97 1.26 3.14 35.97 64.03
8 2.15 3.08 1.22 3.30 45.40 54.60
9 2.24 3.35 1.31 3.53 38.54 61.47
10 1.86 2.56 1.18 2.74 48.03 51.97
II 2.31 3.56 1.34 3.74 38.48 61.52
12 1.47 1.66 1.04 1.86 53.30 46.70
13 2.07 3.03 1.27 3.21 40.45 59.55
14 2.36 3.61 1.37 3.78 32.87 67.1 3
15 2.20 3.25 1.27 3.45 41.32 58.68
16 2.14 3.04 1.26 3.23 37.37 62.63
17 2.09 3.05 1.23 3.24 40.92 59.08
18 1.77 2.37 1.16 2.56 46.01 53.99
19 2.14 3.15 1.27 3.34 40.90 59.10
20 1.80 2.52 1.19 2.72 48.2 1 51.79
21 1.73 2.26 1.13 2.46 51.80 48.20
22 1.53 1. 88 1.06 2.05 54.33 45.67
23 1.24 1.31 0.96 1.49 65 .98 34.02
24 1.71 2.28 1.13 2.46 47.93 52.07
25 1.1 8 1.20 0.94 1.36 68.18 31.82
Total 1.97 2.81 1.22 2.99 44.03 55.97
examples of rain cells fitted with the he and hg model s are shown
in the appendix.
6. Summary
We have presented results from a study that aimed to charac
teri ze and model the rain fields from weather-radar images in the
northwest coast of Spain. However, the approaches used here can
be easily configured to be applicable to weather-radar observations
of other climatic regions. Thi s work is part of an ongoing activity
trying to develop dynamic space-time models of rain fields for
simulating rain-induced effects on millimeter-wave communica
tion systems [55, 56] . Rain identification and modeling are two of
the building blocks in this research. The next steps will be focused
on the development of a space-time model capable of reproducing
the evolution of rain fields. Both simulation and nowcasting (short
term prediction) are the envi saged applications. Moreover, further
work could include growth and decay trends of rain cells to
enhance their dynamics .
Detailed modeling of these images according to two well
known rain-cell models, EXCELL and HYCELL, has been per
formed. For this modeling, rain events were first detected through
a procedure based on coverage and rain-intensity criteria. An algo
rithm for identifying rain cells within the images was then applied.
7. Ackno,,!,ledgment
The authors would like to acknowledge the regional
the Spanish State Meteorological Agency (AEMET) at
providing the radar images. This research was partly carrie-
the framework of the European Network of Excellence
8. Appendix
Here, three examples of the he and hg model-fitted ,_
(Figures 7, 8, and 9) are shown. These three
selected under the following cons iderations:
I. Rain cells belonged to different events fo r each e:c
pIe.
2. At least one example where the hg model was be:-:
than the he model.
3. At least one example where the he model was
than the hg model.
In each of these examples, we show three two-dirr.
images (the radar observation, the he-modeled cell, ani!
modeled cell), as well as three three-dimensional view'so
include a table for each example, summarizing the
each model: R
E
, 0E , and bE, or R
G
, 0G, and b
G
. In
these three parameters in both models are denoted as RH 0
b
H
, respectively.
In these tables (Tables 4,5, and 6) we also compare'
properties of the radar observation and the modeled
These properties are the cell area for R2 =0 1.0 I mm h
ellipticity factor, El; the cell-orientation angle, B; the
rate, (R); its root-mean-square value, (R2); the mean
gradient of R, (G); its root-Olean-square value, (G
2
) :
error, ,; , according to Equation (4). Note that parameters
and B always take on the same values for both the he an':
els.
Each rain-cell shape from radar data or modele.:
approached by an equivalent ellipse, the three shape
which are the major-axi s length, the minor-axis length, aT.
entation angle, B. These geometrical features are corr ...........
minimization of the second moments of inertia. The e
factor is defined as the ratio of the minor-to-major-axis Ie!'
orientation, B, is taken between the major axis and the
system. It thus takes on a value equal to 0 along the d
the x axis, and increases in the counterclockwise sense.
124
"
Q)
'"
Q.
1J
a
"t)
Q)
<Q
Ol
g:
'"
s:
Q)
'g
5-
5
<

Ol
-'"
Z
0
'_Ol
0
0
s-
a

'"

a

'"
Ol
> - I ...- .
-'
... . ..

X (l<m)
Y(km)
X (km)
I -.a__ U''''''
oj .. .
..". J: > 34. e {6.2II) oj " . . . ...... .
.. 2S
Y(km)
2S
X (l<m) x(km)
I
>
2S
Y (km)
X (km)
X (km)
Figure 7. An example of he or hg modeling for one cell corresponding to event 12.
The top row is the radar observation. The middle row is the he modeled cell. The
bottom row is the hg modeled cell.
"
>
r1
X (km) Y (km)
X(km)
!
>
X(km)
Y (km)
X (km)
!
>
X(km)
Y(km)
X(km)
Figure 8. An example of he or hg modeling for one cell corresponding to
event 14. The top row is the radar observation. The middle row is the he
modeled cell. The bottom row is the hg modeled cell.
I
>
Y(km)
X(km) x (km)
I
>
2!i
Y(km)
X(J<m) X(J<m)
~
>
20
Y(km)
X(km) X(J<m)
Figure 9. An example of he or hg modeling for one cell corresponding to event 19. The top
row is the radar observation. The middle row is the he modeled cell. The bottom row is the
hg modeled cell.
Table 4. A comparison of modeling results for the cell in Figure 7.
[mm' h'lj [kml [km
2
1 lmm,h" 'km"\ Immh"\ [1
2
() QH (R) RH or Rmox El ( R2) Ar (G) (G ) ~
b
H
2,25 1.24 he 2.13 10.4 2.73 1.38 1.52 0.1 2
46 0.79 -56.7
1.65 2.83 2.45 2.84 1.37 1.43 0.30
Radar
6.3 hR
- -
7.0 0.75 -44.9 2.27 2.65 1.38 1.65 35
Table 5. A comparison of modeling results for the cell in Figure 8.
..J
[kmj fmmh" \ [km
2
]
[0]
QH b
H
R,., or Rmax Ar El
()
(R) ( R2) (G) (G
2
) ~
he 0.46 0.98 46.3
47 0.38 -12.3
4.57 7.74 4.34 6.08 0.27
hfZ 1.46 3.11 25.9 5.46 7.74 4.45 5.27 0.08
Radar - -
22.3 31 0.31 -11.0 5.45 7.80 4.35 5.54
IEEE Antennas and Propagation Magazine, Vol. 52, No. 5, O c t c : ~ 126
Table 6. A comparison of modeling results for the cell in Figure 9.
[km] [mmh-'] [km
2
]
[0] [mmh-
1
1 (mmh-
I
km-
1
1
aH b
H
RH or Rmax Ar El e (R) (R2) (G) ( G
2
)
.;
he

0.59
0.80
1.29
1.74
102.5
91.5
48 0.76 28_8
7.08 13.33 7.43 10.84 1.80
13.8 23.40 13 .34 17.21 0.19
Radar
-
97.5 32 0.74 31.7 13.6 23.67 12.96 19.78
9. References
I Radio Consultative Committee (CClR), " Recom
-: and Reports of the CClR, Volume V, Propagation in
-pd Media," CCfR 15th Plenary Assembly, Genoa, Italy,
-= iliani, A. Paraboni, J. Lemorton, L. Castanet, L. Feral
.. -usquet, "Short- and Mid-Term Prediction Techniques of
:m Conditions from Weather Radar Observations. Appli
Management in SATCOM Systems," Proceed
. t: 2006 International Workshop on Satellite and Space
:3tions, IWSSC2006, Leganes, Spain, September 2006,

::met, J. Lemorton, S. Bertorelli, C. Capsoni, F. Lacoste
lartellucci, "Advanced Modelling of the Space-Time
- of the Propagation Channel for Future Satcom Sys
'eedings of the First European Conference on Antennas
EuCAP 2006, Nice, France, November 2006, pp.
in, L. Feral, H. Sauvageot, L. Castanet, J. Lemorton and
"Stochastic Spatio-Temporal Modelling of Rain
_.. for Propagation Studies," Proceedings of the Second
Conference on Antennas and Propagation, EuCAP 2007,
UK, November 2007, pp. 1-6.
ni, C. Capsoni, L. Luini and C. Riva, "An Enhanced
. , he SC EXCELL Model Considering Cells Movement, "
of the Second European Conference on Antennas and
EuCAP 2007, Edinburgh, UK, November 2007, pp.
- ni, L. Luini, A. Paraboni, C. Riva and A. Martellucci,
Prediction Model of Rain Attenuation that Separately
:-,)r Stratiform and Convective Rain, " IEEE Transactions
Propagation, AP-57, I, January 2009, pp. 196-204.
_ _-::. Multiexcell: A Large-Scale Rain Field Model for
on Applications and its Experimentql Identification
- . PhD Thesis, Politecnico di Milano, Milan, Italy, 2009,
_ - : lita, M. D'Amico, 8 . Gremont, M. Hajny, F. S.
T. Palade and C. Riva, " Rain Attenuation," in O. Fiser
_-.ST Action 255 Final Report: Radiowave Propagation
;: for New Satcom Services at Ku-Band and Above,
.:.' The Netherlands, ESA Publications Division, March
-: 2: Fixed Propagation Modelling, Chapter 2.2, pp. 2.2.1
: Iwww.cost255.rl.ac.uklpart2.pdf.
_.. H. Sauvageot, L. Castanet and J. Lemorton, "HYCELL
:-Iybrid Model of the Rain Horizontal Distribution for
Propagation Studies: I. Modeling ofthe Rain cal," Radio Science,
38,3, 1056, June 2003.
10. T. G. Konrad, "Statistical Models of Summer Rainshowers
Derived from Fine-Scale Radar Observati ons," Journal ofApplied
Meteorology, 17,2, February 1978, pp. 171-188.
II. J. Goldhirsh and 8 . Musiani, " Rain Cell Size Statistics Derived
from Radar Observations at Wallops Island, Virginia," IEEE
Transactions on Geoscience and Remote Sensing, GE-24, 6,
November 1986, pp. 947-954.
12. R. K. Crane, Electromagnetic Wave Propagation Through
Rain, New York, John Wiley & Sons, Inc., 1996, pp. 1-273.
13. F. Mesnard and H. Sauvageot, "Structural Characteristic of
Rain Fields," Journal of Geophysical Research, 108, D 13, 4385,
July 2003.
14. J. Goldhirsh and B. Musiani, "Dimension Statistics of Rain
Cell Cores and Associated Rain Rate Isopleths Derived from Radar
Measurements in the Mid-Atlantic Coast of the United States,"
IEEE Transactions on Geoscience and Remote Sensing, GE-30, I,
January 1992, pp. 28-37.
15. C. Capsoni, F. Fedi, C. Magistroni, A. Paraboni and A.
Pawl ina, "Data and Theory for a New Model of the Horizontal
Structure of Rain Cells for Propagation Applications," Radio Sci
ence, 22, 3, 1987, pp. 395-404.
16. H. Sauvageot, F. Mesnard and R. S. Tenorio, "The Relation
between the Area-Average Rain Rate and the Rain Cell Distribu
tion Parame'ters," Journal of the Atmospheric Sciences, 56, I,
January 1999, pp. 57-70.
17. P. Misme and P. Waldteufel, "A Model for Attenuation by Pre
cipitation on a Microwave Earth-Space Link," Radio Science, 15,
3,1980, pp. 655-665.
18. R. K. Crane, "A Two-Component Rain Model for the Predic
tion of Attenuation Statistics," Radio Science, 17, 6, 1982, pp.
137 1-1387.
19. C. Capsoni, F. Fedi and A. Paraboni, "A Comprehensive Mete
orologically Oriented Methodology for the Prediction of Wave
Propagation Parameters in Telecommunication Applications
beyond 10 GHz," Radio Science, 22, 3, 1987, pp. 387-393.
20. L. Feral, H. Sauvageot, L. Castanet and J. Lemorton,
"HYCELL - A New Hybrid Model of the Rain Horizontal Distri
bution for Propagation Studies: 2. Statistical Modeling of the Rain
Rate Field," Radio Science, 38, 3, 1057, June 2003.
_ 'laS and Propagation Magazine, Vol. 52, No.5, October 2010 127
21. P. 1. Bremaud and Y. B. Pointin, "Forecasting Heavy Rain
from Cell Motion Using Radar," Journal of Hydrology, 142, 1-4,
February 1993, pp. 373-389.
22. R. K. Crane and H. C. Shieh, "A Two-Component Rain Model
for the Prediction of Site Diversity Performance," Radio Science,
24, 5, 1989, pp. 641-665.
23. M. Dixon and G. Wiener, "TITAN: Thunderstorm Identifica
tion, Tracking Analysis and Nowcasting - A Radar-based Method
ology," Journal of Atmospheric and Oceanic Technology, 10, 6,
December 1993, pp. 785-797.
24. G. Drufuca, "Radar-Derived Statistics on the Structure of Pre
cipitation Patterns," Journal of Applied Meteorology, 16, 10, Octo
ber 1977, pp. 1029-1035.
25. E. Morin, D. C. Goodrich, R. A. Maddox, X. Gao, H. V. Gupta
and S. Sorooshian, "Spatial Patterns in Thunderstorm Rainfall
Events and their Coupling with Watershed Hydrological
Response," Advances in Water Resources, 29, 6, June 2006, pp.
843-860.
26. P. M. Eagleson, "The Distribution of Catchment Coverage by
Stationary Rainstorms," Water Resources Research, 20, 5, 1984,
pp. 581-590.
27. D. R. Cox and V. S. Isham, "A Simple Spatial-Temporal
Model of Rainfall," Proceedings of the Royal Society: Mathemati
cal, Physical and Engineering Sciences, 415, 1849, February 1988,
pp.317-328.
28.1. B. Marco and 1. B. Valdes, "Partial Area Coverage Distribu
tion for Flood Frequency Analysis in Arid Regions," Water
Resources Research, 34, 9, 1998, pp. 2309-2317.
29. H. S. Wheater, V. S. Isham, D. R. Cox, R. E. Chandler, A.
Kakou, P. 1. Northrop, L. Oh, C. Onof and I. Rodrfguez-lturbe,
"Spatial-Temporal Rainfall Fields: Modelling and Statistical
Aspects," Hydrology and Earth System Science, 4, 4, 2000, pp.
581-601.
30. I. Rodriguez-Iturbe, D. R. Cox and P. S. Eagleson, "Spatial
Modelling of Total Storm Rainfall," Proceedings of the Royal
Society: Mathematical, Physical and Engineering Sciences, 403,
1824, January 1986, pp. 27-50.
31. P. Eagleson, N. Fennessey, W. Qinliang and I. Rodriguez
Iturbe, "Application of Spatial Poisson Models to Air Mass Thun
derstorm Rainfall," Journal of Geophysical Research, 92, 08,
1987, pp. 9661-9678.
32. 1. Goldhirsh and E. 1. Walsh, "Rain Measurements from Space
Using a Modified Seasat-Type Radar Altimeter, " IEEE Transac
tions on Antennas Propagation, AP-30, 4, July 1982, pp. 726-733.
33. D. E. Barrick and B. 1. Lipa, "Analysis and Interpretation of
Altimeter Sea Echo," Advances in Geophysics, 27, 1985, pp. 61
100.
34. J. Testud, P. Amayenc, X. Dou and T. Tani , "Tests of Rain
Profiling Algorithms for a Spaceborne Radar Using Raincell Mod
els and Real Data Precipitation Fields," Journal of Atmospheric
and Oceanic Technology, 13, 2, April 1996, pp. 426-453.
35.1. Tournadre, "Determination of Rain Cell Characteristics from
the Analysis of TOPEX Altimeter Echo Waveforms," Journal of
Atmospheric and Oceanic Technology, 15, 2, April 1998, ;-:
406.
36. P. Northrop, "A Clustered Spatial-Temporal Model l':
faJI," Proceedings of the Royal Society: Mathematical.
and Engineering Sciences, 454, 1975, July 1998, pp. 1875- :
37. 1. Awaka, "A Three-Dimensional Rain Cell Model
Study of Interference due to Hydrometer Scattering," Jo.
the Communications Research Laboratory, 36, 147, Marc"
pp. 13-44.
38. A. Pawl ina and M. Binaghi, "Radar Rain Intensity t .
Ground Level: New Parameters for Propagation Impairm<O"
diction in Temperature Regions, " Proceedings of the
Commission F Open Symposium on Wave Propaga .
Remote Sensing, Ahmedabad, India, November 1995, pp. :
39. L. Feral, F. Mesnard, H. Sauvageot, L. Castane:
Lemorton, "Rain Cell Shape and Orientation Distribution ir
West of France," Physics and Chemistry of the Earth . .=
Hydrology, Oceans and Atmosphere, 25, 10- I 2, October ::
1073-1078.
40. J. Goldhirsh, "Two-Dimension Visualization of IU
Structures," Radio Science, 35, 3,2000, pp. 713-729.
41. P. Willems, "A Spatial Rainfall Generator for Sma::
Scales," Journal of Hydrology, 252, 1-4, October 200 I. ;
144.
42. J. von Hardenberg, L. Ferraris and A. Provenzale, "n _
of Convective Rain Cells," Geophysical Research Lellen,
2280, December 2003.
43. A. Pawlina, "Essential Knowledge of Rain Structure [
Applications based on Available Data and Models,"
of 3rd Regional Workshop on Radio Communication j
RADIO AFRICA'99, Gaborone, Botswana, October 1999
44. D. Rosenfeld, E. Amitai and D. B. Wolff, "Classi fi.
Rain Regimes by the Three Dimensional Properties of Re
Fields," Journal of Applied Meteorology, 34, I, January
198-211.
45. E. Matricciani and A. Pawlina, "Statistical Characte:
Rainfall Structure and Occurrence for Convective and
Rain Inferred from Long Term Point Rain Rate Data," P
of Millennium Conference on Antennas and Propagation. -
Davos, Switzerland, April 2000.
46. C. Capsoni , L. Luini, A. Paraboni and C. Riva, "Stra
Convective Rain Discrimination Deduced from Local P
Transactions on Antennas Propagation, AP-54, II , ,
2006,pp.3566-3569.
47. A. Nunez, V. Pastoriza, P. Marino, F. P. Fontan
Fiebig, "A Space-Time Model for Reproducing R;!.
Dynamics," Proceedings of the Second European
Antennas and Propagation, EuCAP 2007, Edinburgh, L i\...
ber 2007, pp. 1-6.
48.1. A. Smith and W. F. Krajewski, "A Modeling Stue:
fall Rate-Reflectivity Relationships," Water Resources .
29,8, 1993,pp. 2505-2514.
IEEE Antennas and Propagation Magazine, Vol. 52, No.5, Ocr.:
128
: :;; Ienburg, A. Jurczyk, J. Szturc and K. Osrodka, "Quan
--c:-:: ipitation Forecasts (QPF) based on Radar Data for
'" : al Models," COST 717: Use of Radar Observations in
. and NWP Models, Working group: 1-8, Working
- WDF_01_200203_2, March 2002; http://www.smhi.se/
_ 01_200203 _2.pdf.
:. -:31t, T. Denoeux and G. Jacquet, "A Radar Rainfall
- :vtethod Designed for Hydrological Purposes," Journal
<r" 114,3-4, May 1990, pp. 229-244.
- alt, T. Denoeux, and G. Jacquet. "The Development of
_ > ILO Rainfall Forecasting Method," in L D. Cluckie and
.r (eds.), Hydrological Applications of Weather Radar,
Ellis Horwood Ltd., 1991, pp. 359-367.
. nson, P. L. MacKeen, A. Witt, E. DeWayne Mitchell,
-.pf, M. D. Eilts and K. W. Thomas, "The Storm Cell
and Tracking Algorithm: an Enhanced WSR-88D
.. Weather and Forecasting, 13, 2, June 1998, pp. 263
. werker, "Cell Tracking with TRACE3D - A New
_.. Atmospheric Research, 61, I, January 2002, pp. 15
_.:eer, "A Short Term Radar Rainfall Forecasting Proce
"h Resolution Radar Data," Physics and Chemistry of
Part B: Hydrology, Oceans and Atmosphere, 24, 8,
9, pp. 879-882 .
..;.oriza, Space-Time Modeling for the Simulation and
:: of Rain Fields through Image Processing Techniques
ata, PhD Thesis, University of Vigo, Vigo, Spain,
pp. 1-134 (in Spanish).
_-.ariza, A. Nunez, F. Machado, P. Marino, F. P. Fontan
_ Fi ebig, "Combining Meteorological Radar and Network
. uges Data for Space-Time Model Development," Inter
';mrnal of Satellite Communications Systems and Net
::: blished online (early view), September 2009.
ucing the Feature Article Authors
ote Pastoriza was born in Bueu, Pontevedra, Spain, in
received the MS degree in Industrial Engineering and the
_ from the University of Vigo, Vigo, Spain, in 1998 and
:>cctively. In 2000 he joined the Electronic Technology
. University of Vigo, as an Associate Professor; since
as been an Assistant Professor there. He is an IEEE
ric is currently a researcher with the Digital Communi
- 0 Instrumentation Group (TE3), University of Vigo. His
ch activities are in the fields of atmospheric propaga
': D communication systems, signal processing, and com
_ mtelligence, and their application in decision making
modeling.
as and Propagation Magazine, Vol. 52, No.5, October 2010
Adolfo Nunez Fernandez was born in Vigo, Spain, in 1975.
He is studying mathematics at the University of Santiago de Com
postela (USC), Spain. Since 2003, he has worked in channel mod
eling for satellite communications in the Signal Theory and Com
munications Department of the Telecommunications Engineering
School of the University of Vigo, Spain. In 2004, he was a guest
researcher at the Institute of Communications and Navigation at
DLR (German Aerospace Research), Oberpfaffenhofen, Germany.
His current research interests are channel modeling for satellite and
mobile radio links .
-
Perfecto Marino received the PhD degree in
Telecommunications Engineering from the Polytechnic University
of Madrid, Spain, in 1984. From 1978 to 1993, he was an Associ
ate Professor in the Electronic Technology Department, University
of Vigo, Spain. In 1988, he was a Visiting Scientist in the Com
puter Science Department, Carnegie Mellon University, Pittsburgh,
PA, USA. Since 1993, he has been a full Professor in the Elec
tronic Technology Department, University of Vigo.
His main research interests include distributed electronic
instrumentation, wireless sensor networks, fieldbus technology for
industrial control, and artificial vision. He has authored more than
20 articles in national and international technical journals, coau
thored 12 books, and published more than 100 papers in interna
tional conferences. He was technical co-chair for such international
conferences as IECON, ISlE, SIECI, CAIP, CONTROLO, SAAEI,
and evaluator of several journals, such as Mechatronics, IEEE
Transactions on Industrial Electronics, IEEE Transactions on
Industrial Informatics, CIT, ESIME, Elsevier, and others. He has
been involved in more than 40 research and applied projects for
several national enterprises, such as Citroen, Televes, Ence, Union
Fenosa, Anfaco, Castrosua, and others. He holds a European patent
on distributed instrumentation for power transformers (2005). He
is also the Director of the Digital Communications Division of the
Electronic Technology Department, University of Vigo. He has
been distinguished as an Expert on Information Technology from
the Commission of the European Union for the SPRINT (Luxem
bourg, 1991); COPERNICUS (Brussels, 1994); and ACTS (Brus
sels, 2002) programs. He has served as General Manager of the
Information Society (2002-2004) and Communications Technol
ogy (2002-2006) programs from the State R&D Office (Autono
mous Government, Galicia, Spain). He is also an expert from the
R&D National Program (Ministry of Education and Science,
Madrid) in the evaluation of research projects in the field of infor
mation and communication technologies. He is an IEEE Senior
Member.
129

You might also like