Rain Attenuation Prediction Model for Satellite Communications

Based on The Météo-France Ensemble Prediction System PEARP

DAHMAN Isabelle123 , ARBOGAST Philippe2 , JEANNIN Nicolas1 , and BENAMMAR Bouchra3
1
ONERA - DEMR, 2 Avenue Edouard Belin, 31055 Toulouse - France
2
CNRM – GMAP, 42 avenue Gaspard Coriolis 31057 Toulouse - France
3
CNES - DCT-RF-ITP, 18 Avenue Edouard Belin, 31400 Toulouse - France
Correspondence: Philippe ARBOGAST (philippe.arbogast@meteo.fr)

Abstract. This paper presents an example of usage of Ensemble Weather Forecast for the control of Satellite-based Com-
munication Systems. Satellite communication systems become increasingly sensitive to weather conditions as their operating
frequency is increasing to avoid electromagnetic spectrum congestion and enhance their capacity. In the microwave domain,
electromagnetic waves that are conveying information are attenuated between the satellite and Earth terminals in presence of
5 hydrometeors (mostly rain drops and more marginally cloud droplets). To maintain a reasonable level of service availability,
even with adverse weather conditions considering the scarcity of amplification power in spacecraft, fade mitigation techniques
have been developed. The general idea behind those fade mitigation techniques is to re-route, change the characteristics, or
re-schedule the transmission in case of too significant propagation impairments. For some systems, a scheduling on how to
use those mechanisms some hours in advance is required, making assumptions on the future weather conditions affecting the
10 link. To this aim the use of weather forecast data to control the attenuation compensation mechanisms seems of particular
interest to maximize the performances of the communication links and hence of the associated economic value. A model to
forecast the attenuation on the link based on forecasted rainfall amounts from deterministic or ensemble weather forecasting is
presented and validated. In a second phase, the model’s application to a simplified telecommunication system allows demon-
strating the valuable contribution of weather forecasting in the system’s availability optimization or in the system’s throughput
15 optimization. The benefit of using ensemble forecasts rather than deterministic ones is demonstrated as well.

1 Introduction and Background

Since a few decades, satellites have become absolutely essential in modern society. Their field of application is expanding
constantly. Nowadays, they are widely used in various areas such as navigation, weather forecasting, disaster management,
or telecommunications. In fact, geostationary telecommunication satellites can offer a global coverage which make them par-
20 ticularly attractive for bringing broadband Internet in isolated areas, where the access to terrestrial networks remains very
limited.
The data transmitted from satellites to Earth are conveyed by radiowaves whose frequency is comprised between 1 and
40 GHz. Frequencies within this range are classified into frequency bands, dedicated to specific applications for satellite

1
 communications but also shared with other systems as fixed terrestrial wireless systems, microwave remote sensing instruments,
radar or positioning systems. These frequency band labels (L,S,C,X, Ku,Ka and Q/V) are detailed in Table 1.
To increase the overall capacity of communication satellites (and hence the number of users of the system and/or the offered
data rate), the use of the Ka and Q/V bands for which large modulation bandwidths are available is becoming widespread
5 among operational systems. However, the possibility to transmit data at a given data transmission rate is also dependent on the
power level of the Electromagnetic wave received by the terminal. An insufficient power level will result in data losses. Power
losses between the satellite and the Earth terminals are mostly caused by the dilution of the wave in space during its propagation
and by some atmospheric phenomenons. In particular, atmospheric gases and more importantly the presence of hydrometeors
attenuate electromagnetic waves. The scattering theories (Rayleigh or Mie) tell us that the level of the attenuation depends
10 on the ratio between the hydrometeor diameter and the wavelength (Gunn and East, 1954), (Oguchi, 1983)). This attenuation
can reach several tens of decibels in case of liquid precipitations. Furthermore, for a given atmospheric state the attenuation
tends to increase significantly with the frequency. The occurrence of propagation losses higher than a given threshold linked
to the way the information is modulated results in data loss. To quantify this probability of data loss, the ITU (International
Telecommunication Union) has established a climatology of the probability to exceed a given level of attenuation. The model
15 strives to quantify the probability of exceedance of tropospheric attenuation function of the radio-electrical characteristics of
the link (frequency, elevation) and of the geographical position of the Earth terminal (ITU-R P.618-13, 2017). To this aim,
climatological databases taking into account the rain regime (ITU-R P.837-7, 2017) are used. The height of the 0o C isotherm
(ITU-R P.839-04, 2013) is also useful since solid hydrometeors have a negligible impact on electromagnetic wave propagation
in the RF domain due to different electrical properties than liquid ones.
20 Figure 1 shows the attenuation due to rain exceeded 0.1 % of the time as computed by the ITU, for two different frequencies
of transmission : 12 GHz and 50 GHz. The difference between the two scenarios is significant. At 12 GHz, the rain attenuation
exceeded 0.1% of the time is a few dB for Europe whereas it is several tens of dB at 50 GHz. As the systems are usually
designed to ensure a link availability larger than 99.9%, the attenuation values presented in Fig. 1 represent the power margins
required to cope with the atmospheric impairments and to comply with the availability requirement function of the geographical
25 location of the receiving station.

2
 Figure 1. Attenuation exceeded 0.1% of the time computed from (ITU-R P.618-13, 2017) for a link with a geostationary satellite at 35
degrees of elevation at 12 GHz(a) and 50 GHz(b).

The margins required in some area to maintain the communication 99.9% of the time can reach 30 dB at Q/V band in mid-
latitudes areas and even more in tropical areas. In other terms, it means that to ensure the availability of the link with a 99.9%
probability a power 1000 times higher than the one required to maintain the link without rain is needed. As the power on-
board of satellites is a scarce resource, techniques to mitigate adaptively the impairments have been developed (Panagopoulos
5 et al., 2004) to allow the efficient use of Ka and Q/V bands for satellite communications. These techniques rely either on
signal re-routing (Jeannin et al., 2014) (i.e using another station ), delaying the transmission (Arapoglou et al., 2008), data rate
decrease (Cioni et al., 2008), or in some extent in payload reconfiguration (Paraboni et al., 2009), (Resteghini et al., 2012).
Some of them can be operated in closed loop. In this case the state of the channel is analyzed in real time and the decision
to activate one or another mechanism is taken in near real-time. This is for instance the case of techniques using adaptive
10 coding and modulation that aim at adjusting the data rate and as a direct consequence the robustness of the communication
link to the attenuation experienced by the communication link. One of the consequence is that a return communication channel
between the Earth terminals and the satellite is required to convey the feedback. It is not necessarily available (for instance in
broadcasting systems) and is also a cost and a security issue.
Other techniques require a forecast of the attenuation some hours in advance in order to prepare and to optimize the link
15 configuration through the telecommand of the satellite. In this respect the use of meteorological forecasts constitutes a promis-
ing approach to control the decision process associated with those fade compensation mechanisms. This has for instance been

3
 studied in (Paraboni et al., 2009), (Biscarini et al., 2016) or (Jeannin et al., 2017) for different applications. One of the major
difficulty highlighted by those studies is that the attenuation induced by rain on Earth space links is not strongly correlated to
the model outputs due to the model accuracy and its limited space and time resolutions. Hence, it has been found that the use
of the microwave attenuation computed directly from model outputs in the decision process shows poor skill.
5 To increase the attenuation prediction skill, probabilistic precipitation forecasts based on ensemble predictions could be used
as long as cost-loss models are known and available.
The objective of this paper is to propose and describe a fully probabilistic approach to forecast rain attenuation by forecasting
a probability of exceeding a given rain attenuation level rather than a deterministic value. To this aim Ensemble Prediction
Systems (Descamps et al., 2015) will be shown to be particularly suited when using the probabilistic precipitation forecasts in
10 the control loop of a simplified communication satellite.
The organization of the paper is the following. The first section is devoted to the description of the model, where the different
steps to obtain rain attenuation probability distributions conditioned to ensemble forecasts from Météo-France are described.
In the second section, various scores are analyzed to assess the relevance of the proposed attenuation model. In a last section,
the performances of the forecasts to maximize either the link capacity, the link availability or both are analyzed considering
15 concurrently measured attenuation data and the simulation of a simplified communication system.

2 DESCRIPTION OF THE ATTENUATION FORECAST MODEL

To develop the forecast model, it is needed to relate actual rain attenuation data to precipitation forecasts. The data used to this
aim are detailed in a first part of this section. The elaboration and the development of the model is detailed in a second stage.

2.1 DATA

20 2.1.1 Beacon Data

The attenuation due to rain on an Earth space-link can be characterized by measurement on Earth of power fluctuations of
beacon signals (unmodulated signals) emitted by satellites. As the signal transmitted by the satellite has a constant power, the
fluctuations of the received power are linked to the fluctuations of the tropospheric fade undergone by the signal during its
propagation. Furthermore, the temporal scales of variation of water vapor, oxygen, clouds and rain attenuations differ signifi-
25 cantly which allows to discriminate the various contributors to the tropospheric attenuation. In particular, the rain attenuation
dominates by large the total attenuation and can easily be deduced from the fluctuation of the beacon. Another possibility is to
isolate the attenuation due to rain from the other components using concurrent radiometric measurements, that can be used to
quantify clouds and gaseous attenuation.
ONERA conducts its own measurements campaigns analyzing the 20 GHz (Ka band) beacon signal of the Astra 3B geo-
30 stationary satellite in various experimental facilities. Throughout the years 2014 et 2015, Ka band attenuation measurements
have been collected for two receiving sites located in Toulouse, France (latitude 43.5◦ N and longitude 1.5◦ E) (Boulanger

4
 et al., 2015) and in Salon de Provence, France (latitude 43.6◦ N and longitude 5.1◦ E). These attenuation measurements are
sampled at 1Hz but have been averaged on a 5-minute basis to save computational resources. An illustration of the Ka band
rain attenuation time series used in the study is given in Fig. 2.

Attenuation due to rain - KA Band

40

Attenuation (dB)
35
30
25
20
15
10
5
0
0 50 100 150 200 250 300 350 400
Time (Days)
40
Attenuation (dB)

35
30
25
20
15
10
5
0
0 5 10 15 20 25
Time (Hours)

Figure 2. Ka band attenuation measurements operated by ONERA in 2014 ( upper Figure ) for a link between the geostationary satellite
Astra 3B and a receiving station located in Toulouse, France. The lower figure focus on the 19 July 2014.

Those rain attenuation have been used as predictands in the statistical model discussed in part 2 as well as the probabilistic
5 precipitation forecasts described in the next subsection.

2.1.2 Weather Forecast Model

The probabilistic precipitation predictions are built upon the sampling using the French global ensemble PEARP (Descamps
et al., 2015). The horizontal resolution is variable and is about 10 km over France (60 km on the opposite side of the globe).
Initial condition uncertainties are sampled using singular vectors and an Ensemble of 4DVar (Desroziers et al., 2014). The
10 model error component is based on the multiphysic approach. For instance we run two different deep convection schemes such
as those described in (Bougeault, 1985) and in (Piriou et al., 2007). 35 members are calculated twice a day (at 0600 UTC and
1800 UTC). Available archived data consists in three hours cumulative rain rate forecasts. The daily PEARP forecasts archived
in 2014 and 2015 have been used on a 0.5°×0.5° grid in this study to build predictive attenuation distributions. Specifically,
we build a super ensemble by using together the 35 members of the 36h forecast run at 1800UTC day D with the 35 members

5
 of the 24h forecasts run at 0600 UTC D+1. In the following, the ensemble forecast is denoted as F such as F =[member 1,
member 2 ... member 70] where each member constitutes an estimation of the rain amount expressed in mm/3 hours.

2.2 The attenuation statistical prediction model

The temporal variability of attenuation due to rain is high (5 minutes) compared to the time resolution of the model forecasts
5 (3 hours). It would not make sense to average rain attenuation data on a 3-hour basis. Furthermore, this would dramatically
reduce the size of the training dataset used to compute the complementary distribution function of attenuation conditioned
to the PEARP forecasts. The strategy adopted in order to provide a statistical link between predicted rain amount and rain
attenuation is the following. First, we select all members of the PEARP forecasts archived in 2014 and 2015 around the
receiving station of interest. In some cases the weather forecast is realistic but even a slight phase error may lead to the
10 double-penalty problem (Nurmi, 2003). Therefore, we consider here a 100x100 km box around the attenuation observation site
such as in (Theis et al., 2005). Second, theses PEARP members are duplicated 36 times. Because there is 36 times 5 minutes
in three hours, this allow to obtain an equal number of observations and forecasts data. Theses forecasts are then classified
into the 5 categories, denoted c0 to c4 with c0 = [0., 0.01] mm/3h, c1 = [0.01, 0.1] mm/3h, c2 = [0.1, 1] mm/3h, c3 = [1, 6]
mm/3h, c4 = [6, 50] mm/3h. Finally the members of the forecast thus classified and the attenuation observations are combined
15 to compute the probabilities P (A > AT h |F ∈ ci). Theses probabilities are memorised to predict, in the operative context,
the overunn of an attenuation threshold AT h . The Fig. 3 gives an example of the complementary cumulative rain attenuation
distribution conditioned to PEARP forecasts obtained from the previously described methodology.

6
 Cumulative distribution functions of rain fall measurments in Toulouse
conditionned to PEARP rain forecast
PEARP in :
0.0 to 0.01 mm/3H
0.01 to 0.1 mm/3H
101
0.1 to 1.0 mm/3H
1.0 to 6.0 mm/3H

given on abcissa (%)

Probabilite to exceed the attenuation A*
6.0 to 50.0 mm/3H

100

10−1

10−2
0 5 10 15 20
Attenuation not exceeded A* (dB)

Figure 3. Probability to exceed the attenuation threshold given in abscissa based on data recorded in 2014 and 2015 in Toulouse, France.

In the operative context, the goal will be to represent the future state of the channel by a unique predictive attenuation
distribution. Yet the PEARP forecasts are constituted of 70 members. Each of theses members leads to the selection of a
specific complementary distribution of rain attenuation. A methodology must be defined in order to obtain a single probabilistic
estimation of the rain attenuation occurring in future. The formula of total probability is written as follows:

n−1
X
5 P (A > AT h |F ) = P (F ∈ ci )P (A > AT h |F ∈ ci ) (1)
i=0

where AT h is the attenuation threshold, expressed in dB. F is the PEARP forecast with F = [member 1, member 2 . . . . member
70] . c0 to cn are the PEARP classes previously described with n being equal to 5. P (A > AT h |F ∈ ci ) is the complementary
cumulative distribution of attenuation, conditioned to the PEARP classes presented in Fig. 3. P (F ∈ ci ) is the probability
that, for a draw at a given time, the forecast belong to a given class. This probability can be directly sampled by counting the
10 members of the ensemble. For example, let us assume that, for a given time horizon, half of the PEARP members belongs to
the first PEARP class c0, half belongs to the second PEARP class c1 . In such case, P (F ∈ c0 ) = 0.5, P (F ∈ c1 ) = 0.5 and
P (F ∈ c2 , c3 , c4 ) = 0.
This methodology is equivalent to averaging the 70 rain attenuation distributions. An illustration is presented in Fig. 4. For
the sake of simplicity, we consider only two realizations of the forecast. Both members lead to a complementary cumulative
15 distribution of attenuation (black and green line) which, after averaging, leads to the probabilistic forecast of rain attenuation

7
 noted P (A > AT h |F ) (gray line). Where AT h is the attenuation threshold, in dB, exceeded and F the PEARP forecast system.
In this specific example F= [1.48,0.78] mm/3H.

Cumulative distribution functions of rain fall measurments

conditionned to PEARP rain forecast
Membre 1 =1.48 mm/3H
Membre 2 =0.78 mm/3H
Average distribution
101

given on abcissa (%)

Probabilite to exceed the attenuation A*
100

10−1

10−2
0 5 10 15 20
Attenuation not exceeded A* (dB)

Figure 4. Probability to exceed the attenuation threshold given in abscissa. 2 distributions are selected according to the value of 2 ensemble
members (black and green lines). The gray line stands for the average of those 2 distributions.

The use made of this predictive attenuation distribution will depend on the application of interest. One possible use is to
get the attenuation threshold exceeded only 0.1% of the time, which is the tolerated unavailability threshold. This attenuation
5 threshold is equivalent to the power margin required to prevent inappropriate communication interruptions.

3 MODEL EVALUATION

The low spatial and temporal resolutions of the PEARP archives used for the learning process may cast some doubt on the
utility of the attenuation forecasts. An appropriate use of the model requires an evaluation of its potential and weaknesses. A
probabilistic forecast model is expected to present reliability and resolution. The reliability assesses the ability of a model to
10 provide a probabilistic prediction of a given event closed to the observed frequency of the same variable. The resolution is the
ability to discriminate between events and non-events.
Only scores based on binary events are considered here. In a first stage, the reliability of the attenuation forecasts is addressed
using the reliability diagram and the rank diagram also known as the Talagrand diagram. Resolution is evaluated using Receiver
Operating Characteristic curves,referred to as ROC hereafter, and sharpness diagrams. The scores proposed here are computed

8
 using the available observations and model outputs over the 2-year period defined above. Nevertheless, in order to evaluate the
possible overfitting of the statistical model a bootstrapping approach is used: 6/10 of the sample is taken for the learning step
and the remaining 4/10 for the score calculation. The procedure is repeated 100 times.

3.1 Model reliability assessment

5 The reliability of an ensemble model characterizes its ability to provide forecasts frequencies consistent with the observed
ones. For example, let’s assume that the forecast system provides, for a particular event, a probability of occurring of x%.
Ideally, this event should be observed x% of the occasions on which such forecast is made.
The reliability diagram consists in plotting the observed frequencies against the forecast probabilities, previously classified
into a few bins. For perfect reliability, the curve must merge with the diagonal line. A reliability curve located to the right
10 of the diagonal line is typical of a model overestimating the probability of the event. Similarly, a model underestimating
systematically the probability event presents a curve located to the left of the diagonal line. It is also conventional to represent
the climatological probability of the event in the forecast by a vertical line and the climatological probability of the event in
the observation by a horizontal line. The last one brings a complementary information on the model resolution. A forecast
which provides the effective climatological probability has no ability for discriminating between cases of event and cases of
15 non-events, this means that it has no resolution.
In the following, a positive event will be defined as the overrun of an attenuation threshold AT h alternatively set to 1 dB, 3
dB and 6 dB. The model developed provides the probability P (A > AT h |F ) which is the predictive probability that a positive
event occurs as a function of the PEARP forecasts. Once fixed AT h it comes that P (A > AT h |F ), for a given F, is a number.
From the time series of the PEARP forecasts archived in 2014 and 2015, the time series of P (A > AT h |F ) have been computed
20 for the three value of AT h . Times series of attenuations observations and predictive probabilities P (A > AT h |F ) allowed to
build the Fig. 5. The climatological probability is indicated by the dotted lines. The small windows show sharpness diagrams
which are discussed later.

9
 Reliability Diagram
3.0
2.5 Attenuation Threshold =1 dB

Repartition(%)
climatology
2.0 50

Sample
40
1.5 30
20
10
1.0 0
0.0 0.5 1.0 1.5 2.0 2.5 3.0

Observed relative probability (%)

0.5 Forecast probability (%)
0.0
0.0 0.5 1.0 1.5 2.0 2.5 3.0

0.7 Attenuation Threshold =3 dB

Repartition(%)
0.6 climatology
50

Sample
0.5 40
0.4 30
20
0.3 10
0
0.2 0.00.10.20.30.40.50.60.70.8
0.1 Forecast probability (%)
0.0
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
0.30
Attenuation Threshold =6 dB
0.25

Repartition(%)
climatology
0.20 50

Sample
40
30
0.15 20
10
0.10 0
0.000.050.100.150.200.250.30
0.05 Forecast probability (%)
0.00
0.00 0.05 0.10 0.15 0.20 0.25 0.30
Forecast probability (%)

Figure 5. Reliability diagrams showing observed relative frequency as a function of forecast probabilities for attenuation thresholds 1dB
(blue), 3dB (green) and 6 dB (red). Inset boxes indicate the frequencies of use of the forecasts. The dotted lines represent the climatological
probability. By convention, the horizontal dashed line represents the climatological probability of the event in the observations. The vertical
dashed line represents the climatological probability in the forecast.

First of all, high probabilities are rarely met for those precipitation thresholds. That is why only low probabilities are shown.
The shape of the reliability curves meets the expectations: the observed frequencies grow with the forecast probabilities and
the curves deviate little from the diagonal line. This reflects the reliability of the attenuation forecast model and confirms its
value for the forecast of the exceed of attenuation thresholds. However, the reliability of our statistical model is not perfect:
5 while low probabilities tend to be underestimated, high probabilities tend to be overestimated.
An other useful tool for determining the model reliability without considering thresholds is the rank diagram, also knows as
the Talagrand diagram (Hamill, 2001). In the case of an ensemble prediction system, the rank histogram of the position of the
verifying observation with respect to the predicted ensemble values provides a measure of reliability. When the ensemble is
reliable, say the verification and the evaluated system are from the same distribution the diagram is flat. If the diagram follows
10 a U-like shape then the observation is often outside the ensemble suggesting the ensemble is under dispersive. Since we have
access to the cumulative distribution function discretized in m quantiles rather than members we attribute the observation to
its quantile. With regard to this, the abscissa of the rank diagram shown on the Fig. 6 is as well discretized in m bins. Non-
zero values of precipitation (or attenuation) are often met for the few highest ranks only. Then, considering our sampling this
diagram could be considered as flat.

10
 Rank Histogram
6

Frequency (%)
3

0
0 5 10 15 20
Quantiles

Figure 6. Rank histogram, also called Talagrand diagram,of probabilistic attenuation forecasts.

3.2 Resolution and sharpness

Resolution is another desired quality we expect from a probabilistic prediction. The Brier score decomposition tells us that res-
olution is the difference between the curves of the Fig. 5 and the observation climatology represented by the horizontal dashed
line. It turns out that the slope of our curves are steeper that the perfect-reliability curve suggesting pretty high resolution.
5 Sharpness is a property of the forecast alone. Nevertheless, it is also an indication of resolution since only sharp predictive
distributions show high resolution. Here, the shape of the diagrams is consistent with a good sharpness although very low
probabilities dominate.
Another approach to address the resolution as well as the value of a probabilistic prediction is to draw Receiver Operating
Characteristic (ROC) curves. ROC graphs are widely used, in particular in the area of medicine (Zweig and Campbell, 1993),
10 to evaluate the power of discrimination of a classification model as a function of a decision threshold. These are of particular
interest to evaluate and compare the performances of predictive models and are as well useful to determine the statistical
decision-making thresholds (in terms of probability level) that will limit the impact of poor forecasting. As detailed in (Fawcett,
2006), the ROC analysis is based on the computation of the true positive and true negative rate for different classification
thresholds. This method is applicable to binary classification problems only. The observation is either considered as belonging
15 to the positive or to the negative class. As a function of the decision-making threshold set, the classifier also assigns the
prediction to one of these two classes. The confusion matrix between observations and forecasts presented in Table 2 lists the
four outcomes: True Positives (TP) and False Positives (FP), True Negatives (TN) and False Negatives (FN).

11
 The true positive rate T Pr represents the rate of true positive (T P ) among the total numbers of positive. It characterizes the
sensitivity of the model, meaning its ability to predict an event when the event occurs.

TP
T Pr = (2)
TP +FN
The false positive rate, also called False Alarm rate F Ar , represents the rate of False Positives (F P ) among the total number
5 of non occurring cases.

FP
F Ar = (3)
FP +TN
The ROC curves is a plot of T Pr against F Ar for a range of decision thresholds. The point whose coordinates are (0,0) is
obtained for a classifier which systematically assigns the forecast to the negative class. Similarly, the point whose coordinates
are (1,1) refers to a classifier which systematically assigns the forecast to the positive class. The best algorithm for classification
10 would yield a point in the upper left corner of the ROC space (coordinates (0,1)). In fact, this point is associated to a zero-rate
of False Alarm and a zero-rate of False Positives. On the contrary, a ROC curve approaching the diagonal from (0,0) to (1,1)
indicates that the forecast skill is not better than chance level.
As described in part 2.2 the model evaluates P (A > AT h |F ), the probability to exceed AT h as a function of weather
forecasts. The problem is as follows: from which probability threshold P (A > AT h |F ) the forecast should be considered as
15 positive?
The Fig. 7 illustrates the decision process involved in identifying positive forecasts as well as the possible scenarios leading
to True Negative, False Negative, False Alarm and True Positive events.

12
 Setting
of AT h

Current forecast Setting

P (A > AT h |F ) of DT h

Is
P (A > AT h |F ) > DT h
?

Forecast Forecast
yes
considered considered
no
as negative as positive

This is
Is the Real Is the Real This is
a True no
attenuation attenuation a False
Negative no
>AT h ? >AT h ? Alarm (FA)
(TN)

yes yes

This is This is
a False a True
Negative Positive
(FN) (TP)

Figure 7. Block diagram of the discrimination process between positive and negatives forecasts.

Let DT h be the decision threshold from which the predictive probability of exceeding a given attenuation threshold is
considered significant enough for establishing protective measures (for example to reduce the link capacity). In other words,
DT h is the probability threshold from which the forecast of exceedance of a given attenuation threshold is considered as
positive.
5 Based on the process described in Fig. 7, false alarm and true detection rates have been computed for a range of DT h , from
0.01 to 3 %, and for three attenuation thresholds AT h respectively set to 1, 3 and 6 dB. A 100-folds cross validation have been
performed (the original sample was randomly partitioned into ten equal sized subsamples, six used as training data, 4 used as
testing data). The average of the values computed in the loop leads to the solid lines in Fig. 8. The standard deviation of this
data is as well given by the 2D-box plot.

13
 ROC curves
1.0
0.34 0.17 0.08
0.04
0.02
0.01
0.0
0.69 SA =1 dB
0.8
0.6
1.39
0.4
0.2

True Positive Rate (Sensibility)

0.0
0.0 2.81 0.2 0.4 0.6 0.8 1.0
1.0
0.08 0.04 0.0
0.02
0.01
0.17 SA =3 dB
0.8
0.6
0.4
0.34
0.2
0.0
0.69
0.0 1.39
2.81 0.2 0.4 0.6 0.8 1.0
1.0
0.04 0.02
0.01
0.0
0.08 SA =6 dB
0.8
0.6
0.4
0.2 0.17
0.0
0.0 0.34
2.81
1.39
0.69 0.2 0.4 0.6 0.8 1.0
False Positive Rate (1-Specificity)

Figure 8. ROC curve obtained averaging 100-fold cross validation for attenuation thresholds set to 1dB(blue), 3dB(green) and 6dB(red).
The box plots indicate the standard devision of each point. The diagonal line corresponds to random forecasts. A curves away from the
diagonal represents a high levels of performance. Are indicated in the boxes, in percent, the forecast probability DT h associated to the points
designated by an arrow.

The curves profile highlights the model ability to exploit the information provided by the weather forecasts. The first obser-
vation that can be made is that the ROC curves are located above the diagonal line (in black). This means that the model
allows improvements over a strategy of randomly guessing the state of the forecast. The model is thus able to detect, in a rather
accurate way, the over run of an attenuation threshold AT h ,whether is set to 1, 3, 6 dB.
5 It should be stressed that the departure from the diagonal of the ROC curve are equivalent for the three used values of AT h .
However, the standard deviation of the data appears to be highly dependent of AT h . It turns out that the highest attenuation
rate considered here is a rather rare event, the strong variability being related to the poor sampling of the ROC calculation.
When the costs of False Alarms and the ones of non-detection are in the same order of magnitude, DT h can be deduced
from the ROC curves. The optimal DT h is then the one given by the left and uppermost point of the curve. Actually, the
10 cost-loss ratio is here non trivial. A methodology to choose optimally the decision threshold in the present context, based on
the maximization of a predetermined economic value, will be discussed in Part 4.1.4.

14
 4 MODEL APPLICATION

In the previous part a methodology to develop a statistical forecast model of rain attenuation based on numerical weather
forecasting has been detailed and evaluated. This final part is devoted to the description of methodologies for the optimization
of the offered capacity or of the economic value for a predetermined user oriented service offer. Both proposed methods use
5 the attenuation forecast model outlined in section 2.

4.1 Optimization of radio resource management algorithms using weather forecasts

4.1.1 Link adaptation algorithms to propagation impairments

As detailed in 1, satellite transmissions become particularly sensitive to weather propagation conditions with growing frequen-
cies. Especially in Ka and Q/V bands, attenuation due to rain may sometimes reach critical levels and cause the interruption of
10 the communication. In such context, adaptive power control systems are particularly costly and can compensate only a limited
portion of the tropospheric fades. It is however essential for satellite operators to ensure a minimal availability to the final user,
typically set to 99.9% of the transmission time.
In order to maintain the link even in adverse propagation conditions without the need of radiating more power, an alternative
is to adapt the link data rate to the weather conditions. The idea is to modify the Modulation and Coding used to carry the
15 information as the tropospheric attenuation is varying. Here, the purpose is not to detail the Modulation and Forward Error
Correction Coding techniques. More informations can be found in (Ziemer and Peterson, 2001) and (Watson et al., 2002). The
whole point of this is that the nature of the modulation and coding combination used to transmit the information determines the
achievable capacity of the link. Furthermore, the use of a specific modulation and coding combination is subject to propagation
conditions. In fact, the most efficient combinations are also the less robust to signal degradation. The challenge is therefore
20 to select the most efficient coding and modulation suitable to maintain the link considering the experienced propagation con-
ditions. Hereafter, the different combinations of modulation and coding available are referred to as MCS ( Modulation and
Coding Scheme).
The characteristics of some MCS that are used in the following of the study are listed in the Table 3.The MCS are labeled
from 1 to 6. The achievable capacity C is given as a function of the MCS selected for a bandwidth of 540 MHz.
25 The capacity characterizes the data amount, here in Gigabit, transmitted each second on the RF channel. An attenuation
threshold AT h indicating the maximal level of attenuation below which the MCS can be used to transmit the data stream
without errors is also given. Those thresholds are of course depending on the characteristics of the link (bandwidth, wavelength)
and of the communication equipments (radiated power, antenna gains and receiver figure of merits).
To result in a valid transmission the attenuation threshold noted AT h associated to the MCS used must be higher than the
30 experienced attenuation on the link. For example, the MCS n°6 should not be used unless the tropospheric attenuation is lower
than 1 dB whereas the MCS n°1 can be used up to 10.8 dB of attenuation but the capacity of the link shrinks from 1.13 to 0.4
Gb/s. More generally, the most robust MCS may be used even while the tropospheric attenuation is strong and thus guarantees
the transmission in most of the experienced weather conditions. However, in that case the price of this resilience to adverse

15
 atmospheric conditions is a reduced capacity of the link. It is then preferable to adjust the MCS following weather conditions.
For instance, an MCS offering high link capacities should be used under clear sky conditions when the tropospheric attenuation
is weak.
It is understood that an inaccuracy in the estimation of the tropospheric attenuation leads to an inadequate selection of the
5 MCS which can have significant consequences on the performances of the link. In case of overestimation of the tropospheric
fading, a less efficient MCS that the one allowed by the real propagation conditions is selected. The difference between the
capacity offered by the achievable MCS and the one used is lost. In case of the underestimation, the MCS selected does not
allows to face the propagation impairments. Such a scenario inevitably leads to the interruption of the communication, namely
to the unavailability of the link and a null capacity. Therefore, it is clear that the underestimation of the attenuation is far more
10 prejudicial than an overestimation.
To illustrate this it can be assumed that the attenuation of the link reaches 2 dB. According to the Table 3, the MCS number
5 should be used to carry the information. In this scenario, the capacity of the link can reach 1 Gbit per second. Now, assuming
that the attenuation has been overestimated using the weather forecast, the MCS n°4 would have been selected instead. The
provided capacity would only reach 0.87 Gbit per second. This mistake would result in a capacity loss of 0.23 Gbit per second.
15 On the contrary, assuming for the same experienced attenuation conditions an underestimation of the attenuation leading to the
use of the MCS n°6 would lead the interruption of the communication and would results in a loss of 1 Gbit per second.
It is now clear that the efficiency and the availability of the satellite transmission is highly dependent of the MCS selection,
based on the analysis of the propagation attenuation. Nonetheless, the propagation losses affecting the link are not always
known as it requires a feedback. As a function of the available information on the state of the channel, different Modulation
20 and Coding strategies can be applied. Theses are detailed in the next subsection.

4.1.2 Transmission strategies

The selection method of the MCS depends on the nature of the available information on the propagation channel. Four different
scenarios have been considered with various assumptions on the type of information available to control the modulation and
coding of the link:

25 – The propagation channel is perfectly known,

– The channel is unknown,

– Probabilistic weather forecasts are available,

– Deterministic weather forecasts are available.

The characteristic of those scenarios are summarized in the table 4.

30 In the first scenario, the attenuation experienced by the link is supposed to be known. The optimal strategy can be adopted,
namely to dynamically select the most suitable MCS considering the current propagation conditions.

16
 This strategy is referred to as ACM for Adaptive Coding and Modulation ((Zhu et al., 2006)). ACM technique is extremely
efficient but requires a quasi-instantaneous feedback of the tropospheric attenuation affecting the channel to be sent to the
modulator through a return channel. The existence of this return channel raises various issues in terms of payload complexity,
security of the telecommand (TC), and required infrastructure. In particular in the context data downlink of Low Earth Orbit
5 satellite, it would need significant developments before entering in operation.
The second scenario assumes the total absence of information about the propagation conditions. When it is impossible to
implement ACM, a constant coding and modulation (CCM) scheme is applied to the transmission. The objective is to use
the MCS that will be compatible with the targeted availability. From (ITU-R P.618-13, 2017) the tropospheric attenuation not
exceeded 99.9% of the time, noted A99.9% can be computed. The CCM strategy consists in selecting the most efficient MCS
10 satisfying the following condition : Ath < A99.9% . For example, let us consider a satellite system configuration operating in Ka
band for which the tropospheric fading exceeds 8 dB, 0.1 % of the time. The CCM strategy would consists in applying as unique
MCS, the numbered 2 in the Table 3. Only robust but relatively inefficient MCS are able in the context of Constant Coding and
Modulation transmission to face the high tropospheric attenuations occurring in Ka and Q/V bands. An intermediate solution
would be to schedule a plan of MCS depending on the probabilistic attenuation forecasts.
15 Third scenario makes use of the weather forecasts. The model developed and described in 2 gives the attenuation distributions
conditioned to the PEARP forecasts. From those distributions, the corresponding attenuation thresholds exceeded 0.1% of the
time can be deduced. It is proposed to choose the MCS in advance as a function of these probabilistic attenuation forecasts.
In the following this strategy will be called PCM, for Programmed Coding and Modulation. Such strategy would provide a
flexibility close to the one offered by an ACM link without requirement of a constant return link between the receiving stations
20 and the satellite.Let us consider the specific case of Low Earth Orbit (LEO) Satellites. Once or several times a day, the LEO
satellite comes in visibility during a few minutes of the receiving stations. None are able to transmit information to the satellite.
However at pre-set times the satellite contact as well a control station equipped with a telecommand link which could be used
to program a plan of MCS for the next orbit.
In order to assess the interest of using probabilistic forecasts over deterministic ones, a last scenario consisting in using
25 deterministic forecasts instead of PEARP forecasts is investigated. This strategy is referred to as PCM-D strategy (PCM-
Deterministic). In the following, the first member of the PEARP ensemble, called the control member is arbitrarily chosen as
the deterministic forecast.
The Adaptive Coding and Modulation based scenario relying on currently experienced attenuation is obviously the most
favorable. In theory, ACM strategy allows a perfect optimization of the capacity and highly limits the unavailability. Thus, the
30 Programmed Coding and Modulation strategy based on the PEARP forecasts, is not expected to be as efficient as the ACM
one. Here, the aim is rather to enhance the performances offered by a Constant Coding and Modulation strategy that relies only
on local climatology without requirement of real attenuation measurements and of a return link. This point is discussed in the
following subsection.

17
 4.1.3 Results

As explained, the PCM strategy would be particularly suitable for the management of Low Earth Orbit Satellite transmissions
for which return links are only implemented at the level of the control station, not at the level of the receiving stations. However,
the lack of Ka band measurements for Low Earth Orbit Satellite prevents the model from being tested in this context. As a first
5 step, an evaluation of the performances of the PCM strategy is thus proposed in the context of a geostationary satellite. A Ka
band transmission line between the geostationary satellite Astra 3B and the receiving station located in Toulouse, France, have
been simulated. The four scenarios listed in the table 4 have been tested.
The mission parameters used for the simulation are the following : a bandwidth of 540 MHz, an elevation of the satellite of
35° and a EIRP+G/T of 80 dBW/K. The EIRP, for Effective Isotropic Radiated Power, measures the ability of the transmission
10 antenna to direct the power emitted in a given area. The term G/T is a figure of merit in the characterization of ground
station antenna performance, where G is the antenna gain in decibels at the received frequency, and T is the equivalent noise
temperature of the receiving system in kelvins.
The PEARP forecasts of 2014 and 2015 have been used as input of the PCM and the PCM-D decision algorithm. The ACM
strategy have been based on the Ka band measurements of the same years assuming an idealized adaptation of the MCS to the
15 channel state.
The target availability has been set to 99.9% of the communication time. Here again a bootstrapping approach has been
used. The available samples were partitioning in 10 subsets, 6 constituting the training dataset, 4 constituting the testing set.
The procedure has been repeated 100 times and the capacities obtained have been averaged. The mean capacities obtained for
ACM, CCM, PCM and PCM-D strategies are presented in Fig. 9.

18
 Geostationnary satellite - KA Band
2500

2000

Mean Capacity Mbps

1500

1000

500

0
ACM CCM PCM PCM-D

Figure 9. Mean capacities obtained for a target availability of 99.9% and considering a GEO satellite. Comparison of ACM, CCM, PCM and
PCDM deterministic strategies. The box plots indicate the standard deviation of the data.

As expected, the best performance have been obtained for the ACM strategy which is an upper bound. Due to limited weather
forecast predictability, the proposed PCM strategy is not able to offer the same level of capacity. Nevertheless, it is clear that,
with an increase of the capacity close to 17 %, this technique significantly improves the throughput ensured by a constant
coding and modulation without any prior on the instantaneous state of the propagation channel.
5 It also appears that the use of ensemble forecast outperforms the use of deterministic forecast. The results obtained are highly
dependent on the systems parameters such as the satellite elevation, the ground station location or the transmission frequency.
Furthermore, the use of higher frequencies (Q/V bands) or of receiving station located in tropical areas would inevitably result
in even more noticeable differences in the throughputs achievable.
So far, the methodology proposed, consisting in programming in advance the modulation and coding using weather forecasts,
10 requires to set a level of target availability. In fact, the MCS are choose in order to ensure this availability. Without this constraint
the number of perspectives would certainly be increased. In the next part, a more general approach is adopted. A decision
process based on the optimization of an economic value taking into account both capacity and availability offered is proposed.

4.1.4 Economic value of the forecast

When it comes to use a probabilistic forecast, the decision to change or not the link MCS amounts to search the forecasted
15 probability above which the forecast is considered as positive. In this context the issue is to find the forecast probability

19
 P (A > AT h |F ) above which it is considered that the attenuation threshold AT h is exceeded, and consequently above which
an action must be taken.
So far, the required availability of the link has been set to 99.9% in order to respond to the operator’s needs. In such context,
the probabilistic attenuation forecast has to be considered as positive when the chance to exceed the attenuation threshold of
5 interest AT h are equal or higher than 0.1 %, which is the value of the tolerated unavailability. In fact, the choice of a decision
threshold higher than 0.1% would not allows to bring the required availability. On the contrary, a lower decision threshold
would result in a capacity loss.
This requirement of 99.9 % of availability is typical for communication satellites, since the tolerance of the final users to
communication outages is highly limited. However, it is easy to imagine further applications for which both availability and
10 mean capacity have to be optimized without prerequisites on any of these parameters. The challenge in this case is to determine
the optimal decision threshold, noted DT h in the following.
As an example, still considering the particular case of an Earth observation satellite. The images acquired by the satellite on
its path must be transmitted on Earth as soon as a receiving station is in visibility. The visibility periods of the Earth stations
are limited to a few minutes. Especially under rainy conditions, it could happen that the data sent by the satellite do not reach
15 the receiving station. In this case data are definitively lost. It might then be sometimes more careful to temporary store the
data in the on-board memory, waiting for the next contact with the ground. This strategy has been evaluated in (Jeannin and
Dahman, 2016). In this context the probabilistic attenuation forecasts could be of great value to decide in advance either the
data transmission or the data storage, provided that the decision threshold is chosen carefully.
A methodology to determine DT h , defined here as the forecast probability from which the data storage is preferred to the
20 data transmission, is now proposed. The optimal value of DT h is the one that allows both, the maximization of the amount of
transmitted data and the minimization of the amount of lost data. For the sake of simplicity, it is considered that the data sent
can only be received if the attenuation do not exceed a given attenuation threshold AT h . A perfect decision algorithm would
result in the data transmission when the real attenuation is lower than AT h and otherwise in the data storage.
The ROC curves introduced in part 3 could be used to determine the probability forecasts threshold that minimize the false
25 alarms rate while maximizing the true detection rate (Greiner et al., 2000). Such method shows its limitations once it is required
to penalize either False Negative or False positive events. We rather propose here a decision process based on the optimization
of an economic value that one could adapt as a function of the needs of the systems targeted.
To account for the high cost of lost data and for the successfully transmitted ones, the economic value to be maximized,
noted EV is defined in Eq. (4).

30 EV = C · 10−L (4)

Where C is the total amount of transmitted data and L is the ratio of total lost data over total transmitted data.
Decreasing the value of DT h limits the amount of transmitted data as well as the amount of lost data and thus impacts the
Economic Value.

20
 In the Fig. 10 is illustrated the evolution of the mean EV on a two years period simulation (2014 and 2015) as a function of
DT h for several attenuation thresholds AT h .

Maximisation of the Economic Value

Att=8.7 dB
70 Att=6.4 dB
Att=3.9 dB
60 Att=2.4 dB
Att=1.0 dB

Mean Economic Value

50

40

30

20

10

0
0.0 0.5 1.0 1.5 2.0 2.5 3.0
Decision Threshold (%)

Figure 10. Evolution of the mean economic value as a function of the decision threshold used to discriminate between positive and negative
forecasts. The mean economic have been obtained averaging the economic values computed following equation 4 for a simulation period of
two years (2014-2015)

It appears in the Fig. 10 that, for all attenuation threshold, an optimal threshold D_T h can be found. It should be used
for deciding either the data transmission or the data storage. As example, for an attenuation threshold of 1.0 dB, the optimal
5 decision threshold is 0.24 %. This means that when the tropospheric attenuation is tolerated up to 1.0 dB, the appropriate
strategy would consist in transmitting data only when the probability P (A > 1dB|F ) is below 0.24% . Otherwise the data
should be stored.It was noted that, for a same attenuation threshold the ROC curves in Fig.8 indicates a decision threshold
comprised between 0.7 and 1.4 %. The difference between the results is because the cost of misdetection is larger than the one
of false alarm. The ROC curves do not integrate this information.

10 5 CONCLUSION AND PERSPECTIVES

This study has presented a methodology for predicting the rain attenuation which affects the satellite transmissions. The sen-
sitivity of satellite transmission to rain becomes particularly sensitive with the on-going trend to use high frequency bands,
from 20 to 50 GHz. The proposed model exploits the probabilistic rain forecasts of Météo-France short-range ensemble pre-
diction system PEARP and delivers probabilistic attenuation forecasts at 20 GHz. In particular due to the inhomogeneity, in

21
 terms of temporal resolution, of the predictand used for the model’s learning process, a bias into the model was expected. It
turns out that reliability diagrams show forecasts frequencies close to the observed ones. The figures obtained suggest that the
statistical model shows only a small remaining bias. For a more complete assessment, ranks diagrams and ROC curves demon-
strate the model ability to discriminate between event and non event cases and to give forecasts frequencies different from the
5 climatology one. Consequently, it can be concluded that the model shows satisfactory reliability resolution and sharpness.
In satellite communication, the main concerns are the link availability and capacity. The primary hypothesis tested in the
study was that the probabilistic weather forecasts could be very helpful to maintain the high availability required by the satellite
operators while optimizing as far as possible the capacity. It has been shown that to condition the type of waveform (modulation
and coding scheme) used to transmit the information, to probabilistic weather forecasts allows increasing the mean capacity
10 of the link while ensuring the availability of 99.9 % usually required. It has also be proven that the benefit is higher using
probabilistic weather forecasts over deterministic one.
The link availability and capacity are highly interdependent. Within a certain limit, increasing one of these parameters is
detrimental to the other one. The request of a high availability inevitably results in a limitation of the capacity which may be in
some context particularly unfortunate. It would then be sometime profitable to be able to find the least expensive combination
15 of these two parameters.
In a last stage, a strategy to maximize an economic value accounting for the transmitted data volume and for the fraction
of successfully transmitted data has been proposed. This economic value could be adapted to the targeted application. For this
initial attempt to optimize high frequency band satellite transmissions from ensemble weather forecast systems, encouraging
results have been obtained. It should be stressed that the application to the forecast of rain attenuation around 50 GHz, or to
20 more sensitive ground station locations such as in tropical regions could show more value. Unfortunately, we do not have any
attenuation observations for those contexts.
It has also to be mentioned that the horizontal resolution and the temporal resolution of the PEARP forecast as well are a
non-negligible drawbacks. It would make sense to replace the global ensemble PEARP by the regional mesoscale ensemble
prediction system PEARP (Bouttier et al., 2011) based on AROME in the attenuation prediction process. The authors would
25 like to stress the fact that the time resolution of the model output used for the prediction process should be higher than 3 hours
and probably even higher than 1 hour according to the large time-variability of attenuation observations.
As a conclusion, though perfectible, the model developed allows to demonstrate the benefit of using ensemble weather
forecasts in the field of satellite communications. The wide range of applications of the model developed includes the following,
particularly relevant weather dependent applications which could be address in further publication:

30 – site diversity (Jeannin et al., 2014) for which an anticipation of the attenuation would allows to schedule in advance the
switching from the ground station under unfavorable propagation conditions to redundant ones.

– deep space links (Biscarini et al., 2016) for which an anticipation of the attenuation would allow scheduling the data
transmission when the propagation conditions are the most favorable.

22
 References

Second Generation Framing Structure, Channel Coding and Modulation Systems for Broadcasting,Interactive Services, News Gathering and
Other Broadband Satellite Applications; Part 2: S2Extensions (DVBS2X), vol. ETSIEN302307-2, Digital Video Broadcasting (DVB)
Standard, 2014.
5 Arapoglou, P.-D. M., Panagopoulos, A. D., and Cottis, P. G.: An analytical prediction model of time diversity performance for earth-space
fade mitigation, International Journal of Antennas and Propagation, 2008, 2008.
Biscarini, M., Marzano, F., Montopoli, M., De Sanctis, K., Iess, L., Montagna, M., Mercolino, M., and Lanucara, M.: Optimizing Data Vol-
ume Return for Ka-Band Deep Space Links Exploiting Short-Term Radiometeorological Model Forecast, IEEE Transactions on Antennas
and Propagation, 64, 235–250, https://doi.org/10.1109/TAP.2015.2500910, 2016.
10 Bougeault, P.: A simple parameterization of the large-scale effects of cumulus convection, Monthly Weather Review, 113, 2108–2121, 1985.
Boulanger, X., Gabard, B., and Casadebaig, L. C.: Four Years of Total Attenuation Statistics of Earth-Space Propagation Experiments at Ka-
Band in Toulouse, IEEE Transactions on Antennas and Propagation, 63, 2203–2214, https://doi.org/10.1109/TAP.2015.2407376, 2015.
Bouttier, F., Nuissier, O., Vié, B., and Raynaud, L.: Convective-scale ensemble prediction experiments with the AROME model, in: EMS
Annual Meeting Abstracts, vol. 8, p. 804, https://doi.org/11th EMS / 10th ECAM, 2011.
15 Cioni, S., De Gaudenzi, R., and Rinaldo, R.: Channel estimation and physical layer adaptation techniques for satellite networks
exploiting adaptive coding and modulation, International Journal of Satellite Communications and Networking, 26, 157–188,
https://doi.org/10.1002/sat.901, http://dx.doi.org/10.1002/sat.901, 2008.
Descamps, L., Labadie, C., Joly, A., Bazile, E., Arbogast, P., and Cébron, P.: PEARP, the Météo-France short-range ensemble prediction
system, Quarterly Journal of the Royal Meteorological Society, 141, 1671–1685, https://doi.org/10.1002/qj.2469, http://dx.doi.org/10.
20 1002/qj.2469, 2015.
Desroziers, G., Camino, J.-T., and Berre, L.: 4DEnVar: link with 4D state formulation of variational assimilation and different possible
implementations, Quarterly Journal of the Royal Meteorological Society, 140, 2097–2110, https://doi.org/10.1002/qj.2325, http://dx.doi.
org/10.1002/qj.2325, 2014.
Fawcett, T.: Introduction to ROC analysis, 27, 861–874, 2006.
25 Greiner, M., Pfeiffer, D., and Smith, R.: Principles and practical application of the receiver-operating characteristic analysis for diagnostic
tests, Preventive veterinary medicine, 45, 23–41, 2000.
Gunn, K. L. S. and East, T. W. R.: The microwave properties of precipitation particles, Quarterly Journal of the Royal Meteorological Society,
80, 522–545, https://doi.org/10.1002/qj.49708034603, http://onlinelibrary.wiley.com/doi/10.1002/qj.49708034603/abstract, 1954.
Hamill, T. M.: Interpretation of rank histograms for verifying ensemble forecasts, Monthly Weather Review, 129, 550–560, 2001.
IEEE: IEEE Standard Letter Designations for Radar-Frequency Bands, IEEE Std 521-2002 (Revision of IEEE Std 521-1984), pp. 01 −
−3, https : //doi.org/10.1109/IEEEST D.2003.94224, 2003.
30ITU-R P.618-13: Propagation data and prediction methods required for the design of Earth-space telecommunication systems, 2017.
ITU-R P.837-7: Characteristics of precipitation for propagation modelling, 2017.
ITU-R P.839-04: Rain height model for prediction methods, 2013.
Jeannin, N. and Dahman, I.: Sizing and optimization of high throughput radio-frequency data down link of earth observation satellites, Int. J.
Satellite Communications Networking, 34, 231–250, https://doi.org/10.1002/sat.1115, https://doi.org/10.1002/sat.1115, 2016.

23
 Jeannin, N., Castanet, L., Radzik, J., Bousquet, M., Evans, B., and Thompson, P.: Smart gateways for terabit/s satellite, International Journal of
Satellite Communications and Networking, 32, 93–106, https://doi.org/10.1002/sat.1065, http://oatao.univ-toulouse.fr/11239/, 2014.
Jeannin, N., Castanet, L., Dahman, I., Pourret, V., and Pouponneau, B.: Smart gateways switching control algorithms based on tropospheric
propagation forecasts, International Journal of Satellite Communications and Networking, https://doi.org/10.1002/sat.1233, http://dx.doi.org/
5 10.1002/sat.1233, sat.1233, 2017.
Nurmi, P.: Recommendations on the verification of local weather forecasts., Shinfield Park, Reading, https://www.ecmwf.int/en/elibrary/
11401-recommendationsverification-local-weather-forecasts, 2003.
Oguchi, T.: Electromagnetic wave propagation and scattering in rain and other hydrometeors, Proceedings of the IEEE, 71, 1029–1078,
https://doi.org/10.1109/PROC.1983.12724, 1983.
10Panagopoulos, A. D., Arapoglou, P. D., and Cottis, P. G.: Satellite Communications at KU, KA, and V Bands: Propagation Impairments and
Mitigation Techniques, Commun. Surveys Tuts., 6, 2–14, https://doi.org/10.1109/COMST.2004.5342290, https://doi.org/10.1109/COMST.
2004.5342290, 2004.
Paraboni, A., Buti, M., Capsoni, C., Ferraro, D., Riva, C., Martellucci, A., and Gabellini, P.: Meteorology-Driven Optimum Con-
trol of a Multibeam Antenna in Satellite Telecommunications, IEEE Transactions on Antennas and Propagation, 57, 508–519,
15 https://doi.org/10.1109/TAP.2008.2011238, 2009.
Piriou, J.-M., Redelsperger, J.-L., Geleyn, J.-F., Lafore, J.-P., and Guichard, F.: An Approach for Convective Parameterization with
Memory: Separating Microphysics and Transport in Grid-Scale Equations, Journal of the Atmospheric Sciences, 64, 4127–4139,
https://doi.org/10.1175/2007JAS2144.1, https://doi.org/10.1175/2007JAS2144.1, 2007.
Resteghini, L., Capsoni, C., Nebuloni, R., Gabellini, P., Maggio, F., Martellucci, A., and Rinous, P.: Effect of non-ideal components on the
20 performance of a reconfigurable on-board antenna for broadcasting applications, in: 2012 6th European Conference on Antennas and Propa-
gation (EUCAP), pp. 1487–1490, https://doi.org/10.1109/EuCAP.2012.6206205, 2012.
Theis, S. E., Hense, A., and Damrath, U.: Probabilistic precipitation forecasts from a deterministic model: a pragmatic approach, Meteorological
Applications, 12, 257–268, https://doi.org/10.1017/S1350482705001763, 2005.
Watson, M., Luby, M., and Vicisano, L.: Forward Error Correction (FEC) Building Block, RFC, 3452, 1–16, 2002.
25Zhu, P., Tong, W., and Royer, C. C.: Adaptive coding and modulation, uS Patent 7,043,210, 2006.
Ziemer, R. E. and Peterson, R. L.: Introduction to digital communication, vol. 2, Prentice Hall Upper Saddle River, NJ, 2001.
Zweig, M. H. and Campbell, G.: Receiver-operating characteristic (ROC) plots: a fundamental evaluation tool in clinical medicine., Clinical
Chemistry, 39, 561–577, http://clinchem.aaccjnls.org/content/39/4/561, 1993.

24
 Band Frequency range Applications for Bandwidth available
satellite communica- for communication
tion systems satellites
L 1-2 GHz Mobile Satellite ser- Some tens of MHz
vices
S 2-3 GHz Mobile Satellite ser- Some tens of MHz
vices
C 4-6 GHz TV Broadcasting and ≈ 200 MHz
point to point com-
munication link
X 7-9 GHZ Military communica- 1 GHz
tion systems
Ku 11-16 GHz TV Broadcasing, 1.5 GHz
point to point link,
internet access
Ka 20-30 GHz Internet access 2.5 GHz
Q/V 40-50 GHz Future systems for 5 GHz
Internet access
Table 1. Frequency bands used for satellite communications.More details in (IEEE, 2003).

25
 Observation

Positive Negative

Positive True Positives (T P ) False Positives (F P )

Forecast
Negative False Negatives (F N ) True Negatives (T N )
Table 2. Confusion Matrix for ROC curves construction

26
 Number of the MCS Modulation Code Rate Attenuation thresh- Capacity achievable
old AT h (dB) C (Gb/s)
1 Q-PSK 0.36 10.8 0.4
2 Q-PSK 0.51 8.7 0.55
3 8-PSK 0.47 6.4 0.71
4 8-PSK 0.62 3.9 0.87
5 8-PSK 0.70 2.4 1
6 16-APSK 0.79 1 1.13
Table 3. Extract of possible modulation and coding schemes as defined in the Digital Video Broadcasting - Satellite - Second Generation
(DVB-S2X) standard (DVB, 2014) and of their operating characteristics.

27
 Scenario n°1 : Scenario n°2 : Scenario n°3 : Scenario n°4 :

The real State of the channel Only PEARP Only Deterministic

attenuation is unknown forecasts available forecasts available
known (1 data every 3 (1 data every 3
(1 data every 5mins) hours) hours)

Coding and Modula- ACM CCM PCM PCM-D

tion Strategy Adaptive Coding and Constant Coding and Programed Coding Programed Coding
Modulation Modulation and Modulation and Modulation
based on PEARP based on
forecasts deterministic
forecasts
Model required None Climatology of Ka Ka band attenuation Ka band attenuation
band attenuation forecast model based forecast based on
(ITU-R P.618-13, on PEARP forecasts deterministic
2017) forecasts

MCS Decision Fac- Real Attenuation Statistical Attenua- Statistical Attenua- Statistical Attenua-
tor tion threshold tion threshold con- tion threshold con-
ditioned to PEARP ditioned to the deter-
forecast ministic forecast
Decision time Continuously Once Programming of Programming of the
In near real time During the design the MCS one or two MCS one or two
phase of the satellite days in advance de- days in advance de-
pending on the fore- pending on the fore-
cast availability cast availability
Availability ensured Close to 100.0% Close to 99.9% Close to 99.9% Close to 99.9%

Table 4. Proposition of Coding and Modulation strategies as a function of the available information on the state of the channel.

Competing interests.

28