ESWW13_Session6_P

Regional forecast of ionospheric
scintillation dedicated to
offshore operators
Ph. Yaya, L. Hecker
CLS (Collecte Localisation Satellites)
Toulouse, FRANCE
ESWW13, Session 6, 15 Nov 2016
Summary
Page 2
1. Introduction
2. Ionospheric scintillation phenomenon
3. Scintillation characteristics in West Africa
o An ionospheric dedicated network
o Time distribution of scintillations
4. An improved model based on measurements
o Main principles
o Performances of local forecast
5. Pilote sites of the service
6. Future steps
ESWW13, Session 6, 15 Nov 2016
Summary
Page 3
1. Introduction
2. Ionospheric scintillation phenomenon
3. Scintillation characteristics in West Africa
o An ionospheric dedicated network
o Time distribution of scintillations
4. An improved model based on measurements
o Main principles
o Performances of local forecast
5. Pilote sites of the service
6. Future steps
ESWW13, Session 6, 15 Nov 2016
Introduction
Page 4

Observation: some positioning processing based
on GNSS may be degraded, sometimes very
severely, on recurrent moment of the day

Consequences: very important financial issues for
off-shore applications.

A consortium (CLS, Fugro, Telecom B.) worked
under a CITEPH funding from 2011 to 2013 :




More than 10 m of error
Observation of the scintillations in West Africa
Correlation with positioning anomalies
Investigate means to forecast the scintillation
Work in 2014-2015: development of a model
based on the long-term observations (few years)
and the very last observations (few hours)  an
operational service is proposed.
Fugro-Topnav
ESWW13, Session 6, 15 Nov 2016
Introduction
S4 modelization near Libreville (Gabon)
PPP processing with CNES/GINS software on NKLG
IGS station (Libreville), from 13 to 21 March 2014:
•
•
•
•
•
Page 5
10 m
Interval: 30s
Constellation: GPS only
Mode: undifferenced ambiguity resolution
Cutoff : 15 deg
Corrections: none (ocean or atmospheric tides,…)
10 m
10 m
ESWW13, Session 6, 15 Nov 2016
Schematic impact
Page 6
OK
OK
OK
NOK
OK
NOK
GPS receiver
ESWW13, Session 6, 15 Nov 2016
Summary
Page 7
1. Introduction
2. Ionospheric scintillation phenomenon
3. Scintillation characteristics in West Africa
o An ionospheric dedicated network
o Time distribution of scintillations
4. An improved model based on measurements
o Main principles
o Performances of local forecast
5. Pilote sites of the service
6. Future steps
ESWW13, Session 6, 15 Nov 2016
Vizualisation of scintillation bubbles
Page 8
The origin of the scintillation phenomenon is an instability of Rayleigh-Taylor in the E
layer (100 km) at the sunset. These instabilities are sub-ionized structures which can
elevate up to 1400 km.
Numerical simulation of
bubbles
The color code corresponds
to an increasing ionization
density with the altitude
(Huton, 2008)
Observations of bubbles : Structures measured at
Jicamarca (9°N) : the altitudes are between 400 and
1400 km before 21:00 LT. On GPS signal there are
decreases of TEC and scintillations
(Valladares,2005)
ESWW13, Session 6, 15 Nov 2016
Ionospheric scintillations
Page 9
Yellowknife, 29 Oct. 2003 (Lassudrie, Fleury)
When the environment crossed by a
radioelectric signal is homogeneous, its
phase and amplitude are regular and
predictable.
But when the signal crosses a medium where
the electronic density is not constant, both
amplitude and phase are affected by
rapid fluctuations that may last up to a
few hours
ESWW13, Session 6, 15 Nov 2016
Quantification of degradation
Ascension, 27 March 2000
(Groves,2004)
 Define standardized indices
Page 10
Amplitude
Phase
Normalized standarddeviation of the signal
intensity (amplitude)
S4 index gives the level of perturbation on
the signal amplitude
Standard deviation of
the signal phase
sf index gives the level of perturbation on
the signal phase
ESWW13, Session 6, 15 Nov 2016
Climatological models
Climatological models (WBMOD and GISM) give global occurrence probabilities. They are not
adapted to local phenomenon (in time and space)
The accumulation of scintillation observations points out zones of high electronic density, the
equatorial « crests » at +/- 15-20° on each side of the equator.
Page 11
Map of scintillation intensity
(WBMOD climatological model). IPS.
Map of scintillation intensity
(GISM climatological model). IEEA
North
equatorial crest
South
equatorial crest
ESWW13, Session 6, 15 Nov 2016
Summary
Page 12
1. Introduction
2. Ionospheric scintillation phenomenon
3. Scintillation characteristics in West Africa
o An ionospheric dedicated network
o Time distribution of scintillations
4. An improved model based on measurements
o Main principles
o Performances of local forecast
5. Pilote sites of the service
6. Future steps
ESWW13, Session 6, 15 Nov 2016
Ionospheric Monitor Network
Page 13
ISM = Iono. Scintillation Monitor =
GNSS receiver robust to
scintillations, and rated at 50 Hz
(= frequency adapted to observe
the scintillations, which have
characteristic times of a few ms)
These monitors directly give S4 and
sf.
Monitor installed in 2011 on Fugro
bases on the Guinean Gulf.
Position of the ISMs w.r.t. the South
equatorial crest :
•
•
•
Port-Gentil station (POG1) on the maximum
Pointe-Noire station (PNR1) is a bit lower
Lagos station (LOS1) is almost on the
magnetic equator
ESWW13, Session 6, 15 Nov 2016
Time distribution of scintillations
Page 14
Data span : about 2 years and ½
Scintillation occurrence:
-
Depends on local time:
o
o
beginning at around19h LT,
ending at around 02h LT.
-
Depends on the season:
o
o
maximum at the equinoxes
(Spring & Automn), with an earlier
beginning (19h instead of 20h LT)
minimum at the solstices (even
though the occurrences are a bit
higher during Summer than during
Winter)
Télécom Bretagne
ESWW13, Session 6, 15 Nov 2016
Time distribution of S4
Page 15
Important : scintillation intensities are not lower at the solstices: they are less
frequent
Example for Pointe-Noire :
- In October 2011 S4 is > 0.8 almost each day
- In August 2012, there are also peaks of S4 > 0.8
This behavior cannot
be reproduced with
a climatological
model
 Need for a model
based on real time
observation
ESWW13, Session 6, 15 Nov 2016
Summary
Page 16
1. Introduction
2. Ionospheric scintillation phenomenon
3. Scintillation characteristics in West Africa
o An ionospheric dedicated network
o Time distribution of scintillations
4. An improved model based on measurements
o Main principles
o Performances of local forecast
5. Pilote sites of the service
6. Future steps
ESWW13, Session 6, 15 Nov 2016
Construction of the crest model
Page 17
Data span: January 2013 to
December 2014
Type of data:
• S4 from the three ISM
• S4 derived from GNSS
receivers, computed from a Std
Dev of SNR on L1.
Construction of the model:
• Each grid point is computed as
the 95th percentile of all data
values
• FFT low pass filter on the
cumulated data leads to a
model, which is then adjusted on
the last hours in order to
forecast the level of scintillation.
ESWW13, Session 6, 15 Nov 2016
Forecast method
Page 18
• Every hour, the model is adjusted to the observations of the last 2 hours
 Computation of an observed scale factor
•
The observed scale factor of the afternoon is correlated to the scale factor of the evening
Scale factor at 17 UT (blue) and 20 UT (red) over 24 months
ESWW13, Session 6, 15 Nov 2016
Scale factor statistics
Page 19
• Based on 24 months statistics, a forecasted scale factor is computed
for the next hours
• With observations until 17UT, the scale factor for 20UT is predicted
with an RMS of 0.24
ESWW13, Session 6, 15 Nov 2016
Regional Forecast examples
Page 20
•
•
Top : observations at the IPPs
(Ionospheric Pierce Points)
Bottom : Model at same locations
NB: only the South
crest is considered
ESWW13, Session 6, 15 Nov 2016
Local validation (ex. at Libreville)
Page 21
•
A forecast can be
given for any
location covered by
the model
ESWW13, Session 6, 15 Nov 2016
Local validation: 3-h forecast results
Page 22
•
The daily maximum is predicted 3 hours in advance with an RMS of 0.1
ESWW13, Session 6, 15 Nov 2016
1-6 hours forecast results
Page 23
•
For all forecast ranges between 1 and 6 hours, the performances for a +/- 0.15 interval
of S4 vary from 79% to 85% for daily maxima predictions.
ESWW13, Session 6, 15 Nov 2016
Combined model
Page 24
•
Sinus fitting on the
scale factor to
derive a
climatological
model
 improvement of performances
for few hours forecast
ESWW13, Session 6, 15 Nov 2016
Summary
Page 25
1. Introduction
2. Ionospheric scintillation phenomenon
3. Scintillation characteristics in West Africa
o An ionospheric dedicated network
o Time distribution of scintillations
4. An improved model based on measurements
o Main principles
o Performances of local forecast
5. Pilote sites of the service
6. Future steps
ESWW13, Session 6, 15 Nov 2016
Pilote web sites
Page 26
Put in place as a
demo for some
oil&gas
companies
NB. Many feedbacks
from TOTAL that
help us
improving our
model
Customer’s sites of interest
Hourly value of modelised index
(Observed on last 18h /
Prediction next 6 hours)
History of daily max.
ESWW13, Session 6, 15 Nov 2016
Summary
Page 27
1. Introduction
2. Ionospheric scintillation phenomenon
3. Scintillation characteristics in West Africa
o An ionospheric dedicated network
o Time distribution of scintillations
4. An improved model based on measurements
o Main principles
o Performances of local forecast
5. Pilote sites of the service
6. Future steps
ESWW13, Session 6, 15 Nov 2016
Perspectives
Page 28
Tuning the model in order to:
Improve the detection rate,
Decrease the false alarm rate.
15 March, NKLG
Color : forecasted S4<0.6
Grey : forecasted S4 >0.6
Time forecast per satellite  towards a
mitigation solution for positioning
during scintillation events
Analyse the possibility to construct and
apply a model for the South-East
Brasilian area, as it is also
affected by scintillations.
ESWW13, Session 6, 15 Nov 2016
Conclusions
Page 29
Ionospheric scintillations can severly affect the precision of GNSS-based position
techniques (up to several meters during high scintillation events)
Scintillations have a global well known behavior but they are very difficult to forecast at a
local level.
Thanks to ionosphere-dedicated instruments and classical GNSS receivers we propose a
model to predict the level of amplitude scintillation (S4) in the Guinean Gulf.
The model is adjusted on a few hours of near real time observations and is able to give
the maximum level of S4 3 hours before the beginning of the eventual event with a
confidence level of 0.15 on 83% of the cases (4% of false alarms, 13% of missed
events)
The performance can still be improved, and a prediction per satellite will be considered in
the future. The same model will be tested for the South-East Brasilian coast.
ESWW13, Session 6, 15 Nov 2016