349

1 INTRODUCTION

The utilisation of PNT services is increasingly taking

place in current and planned applications where high-

precision and integrity of services are required. It

reflects especially on critical applications in maritime

navigation and transportation branches in general.

Developing applications are requiring demanding

tasks of sub-meter accuracy such as maritime

operations, traffic management, search and rescue

operations, marine engineering, as well as offshore

and port systems operations. Furthermore, satellite

positioning represents the primary navigational data

source onboard vessels. The timing data accuracy is

essential for the management of sailing passages,

ports and approaches and navigational lights

synchronisation (Thomas et al., 2011; Brčić, 2012).

Performance degradation of GNSS services can affect

certain system/application to a greater or lesser extent,

depending on the level of required availability,

integrity and accuracy (Thomas et al., 2011).

Satellite positioning performance is susceptible to

errors caused by a number of individual causes,

Reconstruction of Geomagnetic Event as Observed in Northern

Adriatic Region and Its Correlation with GPS Single

-frequency

Positioning Deviations

D. Br

čić, J. Ćelić & S. Valčić

University of Rijeka, Rijeka, Croatia

ABSTRACT: Space weather effects are generally recognized as causes of degradation of satellite positioning,

navigation and timing (PNT) services. We analyze GPS position estimation error during a geomagnetic storm,

focusing on manifestations of geomagnetic processes. The position estimation error was analyzed in terms of

GPS coordinates’ deviations (latitude, longitude and height) from their reference values. The storm’s impact

was studied in the Northern Adriatic region where GPS observables from two Global Navigation Satellite

System (GNSS) reference stations were analysed. Geomagnetic indices were elaborated, comprising readings

from interplanetary, magnetospheric and geomagnetic observatories. Total Electron Content (TEC) on both

stations was computed using dual frequency GPS pseudorange observables. The experiment was to reconstruct

the movement of geomagnetic disturbances entering the geospace, reaching the earth’s surface. The aim was to

correlate possible space weather manifestation on satellite positioning performance in terms of positioning

error. Regularities in changes in positioning deviations were identified with relation to influential indices. The

research offered a possibility of experimental positioning deviations assessment as well as forecasting.

Evaluation of generated rudimentary Classification and Regression Trees (CART) models showed that the risk

of satellite positioning errors could be assessed and predicted considering absolutes, as well as changes in

values of geomagnetic indices. During the research process, several activities emerged as preferable

continuation of the work, with the aim of further development of predictive models and the complement of

space weather scenarios and their consequences on navigational systems. Along with summarized results, they

are outlined in the conclusion section.

http://www.transnav.eu

the

International Journal

on Marine Navigation

and Safety of Sea Transportation

Volume 14

Number 2

une 2020

DOI:

10.12716/1001.14.02.11

350

varying from spatial distribution of satellites to user’s

micro-environment. Signal propagation error

represents the main cause of single-frequency

positioning degradation (Kintner and Ledvina, 2005;

Parkinson and Spilker, Jr., 1996) with ionospheric

delay as a prevalent positioning error cause

(Klobuchar, 1983, 1987). Ionospheric delay is

proportional to the amount of TEC on the path

between the satellite and receiver antenna (Klobuchar,

1988). TEC is expressed in TEC Unit (TECU), one

being 10

e / m

, equivalent to 1.624 meters of

measured pseudorange error at GPS frequency L1

(Parkinson and Spilker, Jr., 1996).

TEC behavior varies depending on different

occurrences and at different timescales (Mendillo,

2006); diurnal and seasonal variations, solar cycle and

storm time behavior. The motivation of the conducted

research was an attempt to create an image of near-

earth space environment during disturbed space

weather, before geomagnetic disturbances reach

GNSS orbit heights, with eventual positioning

deviations as final manifestation. A geomagnetic

event was reconstructed to determine the relation

between the geomagnetic storm effects and GPS

positioning error. Data describing geomagnetic

indices were analysed together with positioning

observables on two International GNSS Service (IGS)

stations in the Northern Adriatic area. Regularities of

positioning patterns were detected as a response to

the storm and variations in the geomagnetic field.

Subsequently, the ionospheric analysis was made

with local TEC behavior as a linkage between

magnetic variations and the final positioning

deviation. Event reconstruction was presented with

the objective to confirm geomagnetic impact on

satellite positioning performance. After analysed data

interpretation, modeling of latitude, longitude and

height deviations was introduced and evaluated. In

the concluding chapter, research results and

inferences were outlined along with planned activities

which emerged from the research.

2 BACKGROUND

Satellite navigation technology presents an inevitable

path towards improvements of transportation as a

master link between sustainability pillars (Brčić,

2012). PNT services are used in all transportation

branches, whether in traffic control management,

monitoring or navigation. The European Union Single

European Transport Area sub-projects are designed to

ease the citizens and cargo movements and to

enhance the European transport sustainability;

however, the vision is global. Current and planned

strategies imply an optimal infrastructure and reliable

sources of dynamic information regarding PNT data

provision (Thomas, 2011).

User Equivalent Range Error (UERE) encompasses

influential quantities which are affecting time

measurement of the satellite signal propagation,

leading to erroneous pseudorange computation

(Parkinson and Spilker, Jr., 1996; Subirana et al., 2013,

Kaplan and Hegarty, 2006):

( )

rct t t

ρ δδ

=+⋅ − +

(1)

where:

... pseudorange between satellite and user

antennae,

... geometric (true) range between

satellite and user antennae,

... user clock error,

... satellite clock offset,

δt

... the total ranging

error generated by other influential factors.

The range measurement timing relations are

shown in Figure 1, as interpreted from (Kaplan and

Hegarty, 2006).

The ionospheric delay can be expressed as:

40.3/ ^ 2f TEC

∆=

(2)

where

∆

… equivalent ionospheric delay of

determined pseudorange,

… system’s operating

frequency,

TEC

… Total Electron Content along an

equivalent column between satellite and user’s

antenna with a cross section of 1 m

Figure 1. Range measurement timing relations

Influences on GPS positioning error during

disturbed space weather have already been

investigated in the area (Filjar et al., 2013), together

with the development of GNSS positioning

performance forecasting (Filić and Filjar 2018).

Ionospheric disturbances are the consequence of

changes, variations and occurrences resulting from

conditions in the solar-terrestrial environment due to

the manifestation of the solar activity (Parkinson and

Spilker, Jr., 1996; Goodman, 2005). According to

National Oceanic and Atmospheric Association’s

Space Weather Prediction Center (NOAA SWPC),

geomagnetic storms, solar radiation storms and radio

blackouts are described by numbered levels and with

the corresponding indicators according to severity.

Geomagnetic activity is described by indices

presented in Table 1 (Perrone and De Franceschi,

1998; Zolesi and Cander, 2014). Solar eruptive events

create magnetic disturbances – Coronal Mass

Ejections (CME) which travel through interplanetary

space and interact with thermosphere, ionosphere and

magnetosphere (Lockwood et al., 1999). The planetary

Kp and Ap indices express the horizontal component

of the geomagnetic stability. There appears a negative

correlation between the Kp index and the storm-time

TEC during summer months (Mendillo, 2006; Ross,

1960). The southward orientation of earth-directed

interplanetary magnetic field (IMF) expressed with

the B

Z index causes the entering of power inputs in

351

the magnetosphere, producing geomagnetic

disturbances (McMorrow, 2011).

Table 1. Geomagnetic indices

_______________________________________________

Index Description

_______________________________________________

K/Kp Three-hour pseudo-logarithmic index representing

disturbances in horizontal component of the

geomagnetic field in relation to quite space weather

conditions.

Dst One-hour indicator of magnetospheric activity, as a

measure of the ring current in the magnetosphere.

It can be interpreted as longitudinal average of

horizontal geomagnetic disturbance.

AE One-hour indicator of auroral (electrojet) activity,

further divided in amplitude upper (AU) and

amplitude lower (AL) indices with the relation

AE=AU-AL A/Ap Linear equivalent of the K/Kp

index, representing average of geomagnetic field

variations

aa Global geomagnetic activity index derived from

two mutually antipodal magnetic observatories.

BT Geomagnetic field intensity indicator, which can be

further decomposed into northing, easting and

vertical component.

_______________________________________________

The Bz index is one of the first qualitative storm

indicators, as its southward pointing triggers

processes in the geospace, including an opening of the

magnetosphere (Lockwood et al., 1999). The

magnetospheric opening causes increased ring

current (Dst), allowing the energy and particles to

enter through auroral ovals (AE/AU/AL) and spread

over the globe (Mendillo, 2006; Booker, 1954). These

processes lead to disturbances in the ionospheric F2

region and on distribution and behavior of the total

electron content (Davies, 1965).

3 METHODOLOGY

The first group of analysed parameters refers to

geomagnetic indices, while the second represents

three-dimensional satellite positioning deviations

obtained with GPS. Readings of magnetic field

changes in the magnetosphere were retrieved from

the Geostationary Operational Environmental

Satellite (GOES) 15 measurements archive (Singer et

al., 1996). Apart from geostationary observations,

readings of magnetospheric field components were

retrieved from Advanced Composition Explorer

(ACE) spacecraft observables, measured at Libration 1

point.

Earth observations were analysed near to the

Northern Adriatic region, using nearest of

International Real-Time Magnetic Observatory

Network (INTERMAGNET) observatories (Chambon

la Foret, France). Global geomagnetic indices Dst, AE,

AU, AL and Kp, were retrieved from the World Data

Center for Geomagnetism (WDC) and Space Physics

Interactive Data Resource (SPIDR) databases.

Collection, processing, determination and creation of

GPS positioning solutions were enabled using Reader

Independent Exchange Format (RINEX) databases,

available at International GNSS Service and US NGS

CORS servers (Gurtner and Estey, 2009). Data from

two stations were analysed: Bolzano, Italy and Graz,

Austria (Table 2). It is assumed that the consistency

and quality of IGS data is ensured through (Kouba,

2009), focusing on the removal of non-dispersive

pseudorange error components, such as multipath (<

0.3 m), cycle slips (< 1 per 1000 observations),

northing, easting and height eccentricities to the

antenna reference point (≤1 mm), and other

requirements needed for the provision of high-quality

and high-integrity GNSS products.

Table 2. General information of Bolzano and Graz IGS

stations

_______________________________________________

ID City Location Longitude Latitude Height

(E) (N) (m)

_______________________________________________

bzrg Bolzano Italy 11.3368 46.4990 328.8

graz Graz Austria 15.4935 47.0671 538.3

_______________________________________________

GPS positioning files (RINEX.pos) were calculated

and archived using RTKLIB open source program

package, employing observation and navigation files.

The ionospheric correction was settled by employing

standard ionospheric model, coefficients of which

were taken from diurnal navigational messages.

Other standard methods and algorithms were settled

as follows: single positioning mode, L1 frequency

(single) positioning solution, 15° elevation mask,

broadcast ephemeris, continuous ambiguity

resolution and tropospheric model (Saastamonien).

Ionospheric TEC was calculated by using dual-

frequency (L1 and L2) GPS measurements from IGS

stations. Measured phase differences (

TEC

) were

smoothed with code differences (

TEC

), after which

they were leveled again using code differences to

correct TEC values (Dyrud et al. 2008):

0.104

TEC

mTECU

−

(3)

0.104

TEC

mTECU

φφ

−

(4)

After Differential Code Biases (DCB) correction

(Noll, 2010), it is assumed that all frequency-

dependent quantities except ionospheric delay were

eliminated (Subirana et al., 2013). Standard deviations

of computed values were presented as well. The total

electron content was derived using GPS-TEC software

from IGS stations’ RINEX files, by using described

mathematical relations.

Time series were analyzed from all parameters,

along with statistical summaries and distribution of

collected data. The event phases were further divided,

isolating specific influences and different behavior of

indices (e.g. depletion and enhancement of TEC).

Correlation between variables was made using

Pearson’s correlation coefficients for the two most

pronounced observed events.

Recursive binary splitting (partitioning) algorithm

was used to develop a decision tree model, which the

main objective is to minimise the Residual Sum of

Squares (RSS) given as (Hastie et al., 2009):

352

( )

m iR

RSS y yR

= ∈

= −

∑∑

(5)

where:

... partitions of partitioned feature space,

... response of a particular testing observation, and

... mean response of the training observation

within partition

Development of such models is based on the

concept of splitting the observed data set into subsets

and on the values of predictors in a series of iterations

(Filić and Filjar, 2018).

For the sake of simplicity, the procedure was

limited to the determined depth of a tree. The general

model is given as follows (Hastie et al., 2009, James et

al., 2013):

(

) (

)

( )

f yw

= =

∑

xx x

v

(6)

where:

…

the mean response for the

particular region

, and

represents how each

variable is split at a particular threshold value.

During splitting, Gini index as the main criteria

was used to assess the function (Hastie et al., 2009,

James et al., 2013):

( )

1 (|)

= −

∑

it p jt

(7)

where:

…

the number of possible output

categories, while the category

has a probability of

occurrence

( |)pjt

After several algorithm repetitions, the final

selection of indices was made, also considering the

correlation results. Regression trees as predictive

models were generated for each positioning

component as target variables. The final output of

models was the estimation of probable values of

latitude, longitude and height deviations,

respectively. Proposed models were evaluated

employing observed and predicted values.

Partitioning of used data-sets was divided as 70% for

training, 15% for validation and 15% for evaluation,

respectively.

4 THE EVENT

In this section, the geomagnetic event was described,

as occurred through June 21 – 27, 2015 following

official event reports, using available data and

followed with own interpretation.

4.1 Geomagnetic storm development

According to the United States Geological Survey

(USGS) National Geomagnetism report, the storm

occurred because of solar wind fast stream and CMEs,

arriving on June 21

at 16:45 UT (day of year (DOY)

172), June 22

at 05:45 UT (DOY 173), and June 22

18:30 UT, compressing the magnetosphere and

generating electric currents and geomagnetic field

perturbations.

Figure 2. IMF Bz component magnitude 4-minue level data

(top) and Kp index (bottom) 3-hours data through the storm

period, DOY 172-177

In a specific moment, IMF Bz component turned

southward (Figure 2). Geomagnetic activity arose

firstly at high latitudes, as shown in Figure 3. The

greatest Dst depression occurred on June 23

at 04:30

UT and recovering until another CME reached the

earth (Figure 3). The sub-events formed geomagnetic

storm period of seven days. Statistical parameters

were calculated based on the time series for the

specific indices, as presented in Table 3.

Figure 3. Auroral activity and Dst values, DOY 171 – 181

According to (Singer et al., 1996), nominal

(undisturbed) total geomagnetic field intensity (BT)

measured on satellite vary from -200 nT to +200 nT.

The BT maximum value and range exceeds the limits

several times.

353

Table 3. Statistical description of geomagnetic indices

during DOY 172-177

_______________________________________________

Index Min Mean SD Max Range

_______________________________________________

GOES BT 33.93 139.5135 72.2918 1,289.21 1,255.28

(nT)

Dst (nT ) -204.00 -46.34 41.20028 36.00 -240

AE (nT) 31.00 280.30 296.5678 1,636.00 1,605

AU (nT) 6.0 115.7 123.0969 841.0 835

AL (nT) -1,101.0 -164.6 190.7277 -13.0 1,088

Kp 0.300 3.273 8.300 1.9665 8.0

CLF BT 47,741.3 47,824.5 25.6576 47,976.3 234.96

(nT)

_______________________________________________

Pronounced values indicate geomagnetic storm

and travelling disturbances and high variations. The

Kp increased two times; the maximum of 8.3 took

place just before the commencement of day 174 (June

), associated with June 21

and June 22

CMEs.

The second increase occurred on day 176 (June 25

)

after another solar eruption and CME arrival,

respectively.

4.2 Storm-time GPS positioning deviations

Time series of deviations of positioning components

were calculated for the period June 22

(DOY 173) –

June 26

UTC (DOY 177). Horizontal deviation plots

are shown in Figure 4.

Figure 4. Horizontal positioning error as calculated for

stations Bolzano (top) and Graz (bottom).

Statistical description of GPS positioning error

through the observed period is shown in Table 4, with

arc units converted to meters.

Table 4. Bolzano (B) and Graz (G) horizontal positioning

error during the storm period

_______________________________________________

Station Lat error (m) Long error (m) Height error

(m)

Mean SD Mean SD Mean SD

_______________________________________________

G 4.191 1.017 -1.350 0.546 1.572 1.849

B 3.524 1.028 -0.126 0.6 0.098 1.875

_______________________________________________

The distribution of observed height values for both

stations is presented in Figure 5.

Although there is a constant offset relative to

reference values, latitude and longitude are

experiencing smaller deviations when compared to

height, tending northward (latitude) and eastward

(longitude), respectively. Standard deviations are

approximately the same at both locations and in all

axes; however, vertical component shows the most

prominent scattering.

Figure 5. Height histograms for Graz (top) and Bolzano

(bottom)

Statistical description of daily positioning

deviations is graphically presented in Figure 6,

showing minimal and maximal values and range,

mean values and standard deviation, respectively.

354

Figure 6. Bolzano (top) and Graz (bottom) daily horizontal

positioning error statistics

4.3 Interpretation of results

The TEC patterns during the storm period with

computed mean values and standard deviations are

shown in Figure 7.

Figure 7. TEC daily patterns mean values (top) and

standard deviations (bottom) as computed for locations

Bolzano and Graz during the storm period

In Table 5, statistical parameters of the computed

storm-time period TEC at locations Bolzano and Graz

are presented.

Table 5. Statistical description of the total electron content

_______________________________________________

Station Minimum Mean Standard Maximum Range

value value deviation value

_______________________________________________

Bolzano 2.09 10.36 5.59 28.55 26.46

Graz 2.03 9.78 5.51 26.57 24.54

_______________________________________________

On Figure 8, readings from GOES magnetometers

1 and 2 are presented, together with geomagnetic

field intensity as measured on Chambon-La-Foret

observatory. On Figure 9, combined plots of

positioning error deviations for Bolzano and Graz

locations are shown, together with associated TEC

computed on both locations. The illustrations are

followed with interpreted results.

Figure 8. Geomagnetic field intensity as measured with

Chambon-La-Foret (top), GOES 1 (middle) and GOES 2

(bottom) magnetometer

Abrupt change (depression) of TEC (DOY 173)

preceded the increase of geomagnetic field intensity.

After geomagnetic peak values and during TEC

decline (end of DOY 173 / beginning of DOY 174),

positioning deviations were most pronounced. At the

same time, the greatest Dst drop and AE increase

occurred. During first hours of day 174, a measurable

TEC peak occurred.

TEC commotions accompanied the second

positioning deviations increase on DOY 174. During

DOY 175, all the observed values were returning in

regular patterns. The increase of Dst indicated

355

possible commencement of new geomagnetic

disturbance.

The DOY 176 is characterized by Dst decline, an

increase of auroral activity and stirrings of all

(magnetospheric and ground) magnetic field

components. Total geomagnetic field flux measured at

GOES magnetometer 1 reached the greatest value

during this period. Second GOES magnetometer

shows oscillations of magnetic components around

the mean values, like ground readings. The drop of Bz

component as measured on the first GOES

magnetometer is here pronounced. Positional

deviations started to increase, reaching maximum

values before midday. The following hours were

characterised with TEC increase, reaching its greatest

observed values. The magnetospheric total field

strength was the highest observed.

The disturbance ceased the following day, which

can be interpreted from regular patterns of

geomagnetic components, as well as TEC lowering. In

line with described patterns, positioning deviations

are also decreasing, returning to normal values.

Figure 9. Graz (top) and Bolzano (bottom) latitude (red),

longitude (blue) and height (green) positioning deviation

patterns, with total electron content (black), computed at

each location

5 DISCUSSION

Two distinctive incidents (events) can be extracted:

abrupt depression and subsequent rise on DOY 173

and pronounced increase on DOY 176.

For these two periods geomagnetic, positioning

and TEC parameters were used to find mutual

correlations. The results are shown in Figure 10 and in

Figure 11. During the first event, the total magnetic

field intensity measured on all magnetometers

(BTCLF, BTG1 and BTG2) correlated positively with

latitude and height positioning errors (bzrg_fi, bzrg_h,

graz_fi, graz_h), while negative correlation were found

with longitude errors (bzrg_l, graz_l). The TEC

(bzrgtec, graztec) showed negative correlations with

components of GOES-measured geomagnetic activity,

and positive correlation with ground-based magnetic

readings. During the second event, BTG1 showed a

positive correlation with positioning components,

opposite to Chambon-La-Foret (ground) readings,

same being applicable for TEC.

Positioning deviations were most expressed

during pronounced changes in geomagnetic field

intensity components, measured with GOES

magnetometer 1 (BxG1, ByG1, BzG1, BTG1), GOES

magnetometer 2 (BxG2, ByG2, BzG2, BTG2) and

Chambon-La-Foret (BxCLF, ByCLF, BzCLF, BTCLF).

The TEC behavior showed correlation with

geomagnetic activity and the affectation of the

geomagnetic impact on both stations similarly.

Figure 10. Correlation matrix between variables during the

first event. The dots size indicates positive (white) or

negative (black) correlation.

Figure 11. Correlation matrix between variables during the

second event. The dots size indicates positive (white) or

negative (black) correlation.

The time of commencement and duration of the

main phase of the storm (DOY 173.5 – 174.5) was

extracted for modeling of positioning deviations of

356

components on station Graz: latitude (devFIG),

longitude (devLG) and height (devHG). During the

height model-building process, main components

were defined as AU, BTCLF, BxCLF, BzCLF, BzG1,

TEC, dTEC, and Dst. A similar process was made for

latitude and longitude models, respectively.

Governing indices were found as follows:

− Latitude model: AE, AL, ByCLF, ByG1, ByG2,

dBzCLF, TEC, Dst,

− Longitude model: AE, AU, BTG1, BTG2, ByCLF,

ByG1, BzCLF, TEC, Dst.

For the latitude target variable, the root node was

a northing magnetic component as measured on

second GOES magnetometer (ByG2), while the total

magnetic field intensity (BTG1) was the root node for

the longitude model.

Figure 12. Latitude (left), longitude (centre) and height

(right) deviations model evaluation

Decision tree models can develop certain

drawbacks in the form of extreme sensitivity and ease

of over-fitting. However, they represent a simple

decision-making tool. The success of models will

depend on the appropriate selection of possibly

influential quantities. As for the proposed models,

obtained results were compared with observed values

of positioning deviations. Evaluation of models is

shown in Figure 12. The plots are showing observed

versus predicted values, local regression scatterplot

smoothing, regression lines and confidence intervals.

Considering adjusted R-squared, the average

overall score of the efficiency of models is 71.4%

(Table 6).

The height model showed the best performance

(78%). In general, vertical satellite positioning

component is most susceptible to external influences.

The latitude model evaluation results showed 69%,

and longitude model showed 67% of performance

efficiency, respectively.

Table 6. Model evaluation output: observed against

predicted values regression analysis

_______________________________________________

Height deviation model

_______________________________________________

Coefficients Estimated Standard Pr(>|t|)

coefficient Error

_______________________________________________

Intercept 0.19438 0.06719 0.00421

Predicted deviation 0.96744 0.03458 < 2e-16

R-squared Multiple: 0.7845 Adjusted: 0.7835

F-statistic: 782.5 on 1 and 215 DF p-value: < 2.2e-16

_______________________________________________

Latitude deviation model

_______________________________________________

Coefficients Estimated Standard Pr(>|t|)

coefficient Error

Intercept -0.1192 0.3568 0.739

Predicted deviation 0.9921 0.04552 < 2e-16

R-squared Multiple: 0.6884 Adjusted:0.687

F-statistic: 475 on 1 and 215 DF p-value: < 2.2e-16

_______________________________________________

Longitude deviation model

_______________________________________________

Coefficients Estimated Standard Pr(>|t|)

coefficient Error

Intercept -0.1277 0.28384 0.653

Predicted deviation 0.9274 0.04413 < 2e-16

R-squared Multiple:0.6726 Adjusted:0.6711

F-statistic: 441.7 on 1 and 215 DF p-value: < 2.2e-16

_______________________________________________

6 CONCLUSIONS

The amount and intensity of space weather effects on

PNT services depend on an individual event. A

geomagnetic event which took place in June 2015 was

studied in this paper. The initial aim was the

reconstruction of the flow of geomagnetic occurrences

from magnetospheric heights up to the surface of the

earth, and to correlate geomagnetic indices with the

GPS positioning deviations. The storm period was

fragmented in sub-events, each characterised with

specific behavior of indices. Data interpretation and

subsequent analysis showed that positioning

deviations followed changes in geomagnetic field

intensity. Ionospheric TEC was introduced as an

entity which complemented the storm features and its

manifestation on positioning. Characteristic changes

in TEC behavior during times of geomagnetic

variations were observed. After the correlation

between all employed variables, regression tree

models were introduced to determine probable values

of positioning deviations as target variables. The

evaluation of models showed an acceptable level of

success, especially reflecting on height deviation

prediction as the most sensitive positioning

component. The proposed approach offers a simple

yet effective mean of assessment of GPS positioning

performance during space weather events directly or

indirectly affecting positioning deviations.

The storm was not classified as extreme; however,

it measurably influenced the positioning

performance. With ever growing GNSS utilisation,

such events could affect the simultaneous

performance of GNSS-based services that are

expected to be independent of each other. Further

research of similar events and employment of

additional indices is necessary to contribute to a more

comprehensive image of the solar-terrestrial

environment. Besides geomagnetic disturbances,

models should also employ solar, solar wind and

ionospheric indices, opening possibilities of predictive

357

analytics by monitoring upcoming disturbances from

their place of origin.

ACKNOWLEDGMENTS

This work has been fully supported by the University of

Rijeka under the project – uniri-tehnic 18-66. The authors

would like to thank N. Ness, Principal Investigator of Bartol

Research Institute and to Dr. Howard Singer, NOAA Chief

Scientist. Authors appreciate and support the access to open

software tools: R, R Studio and Rattle, RTKLIB and Rinex-

GPS-TEC.

REFERENCES

Booker, H.G., Morphology of Ionospheric Storms,

Proceedings of the National Academy of Science of the United

States of America, PNAS, Washington DC, 1954, vol. 40,

no. 10, pp. 931–943.

Brčić, D., Ensuring sustainability through utilisation of

satellite navigation technology, Proceedings of the 2012

International Conference on Transport Sciences (ICTS);

Portoroz, Slovenia, Fakulteta za pomorstvo in promet,

2012, p 14.

Davies, K., Ionospheric Radio Propagation, National Bureau of

Standards, Washington DC, 1965.

Dyrud, L. et al., Ionospheric measurement with GPS:

Receiver techniques and methods. Radio Science, 2008,

vol. 43, no. 6, pp. 1-11.

Filić, M. and Filjar, R., Forecasting model of space weather-

driven GNSS positioning performance. Lambert Academic

Publishing, Saarbrucken, 2018.

Filjar, R., Brčić, D. and Kos, S., Single-frequency Horizontal

GPS Positioning Error response to a moderate

Ionospheric storm over Northern Adriatic, In: Weintrit,

A. (ed) Advances in Marine Navigation, Marine Navigation

and Safety of Sea Transportation, CRC Press, Boca Raton,

2013, pp. 49-56.

Goodman, J.M., Space Weather and Telecommunications,

Springer, New York, 2005.

Gurtner, W. and Estey, L., RINEX: The Receiver Independent

Exchange Format, V3.01, IGS Central Bureau, Pasadena,

2009, p. 44.

Hastie. T., Tibshirani, R. and Friedman, J., The Elements of

Statistical Learning, Springer, New York, 2009.

James, G., Witten, D., Hastie, T. and Tibshirani, R., An

Introduction to Statistical Learning, Springer, New York,

2009.

Kaplan, E.D. and Hegarty, C.J., Understanding GPS:

Principles and Application, 2nd ed., Artech House, Boston,

2006.

Kintner, P.M. and Ledvina, B., The ionosphere, radio

navigation and global navigation satellite systems,

Advances in Space Research, 2005, vol 35, no. 5, pp. 788-

811.

Klobuchar, J.A., Ionospheric corrections for timing

applications, Proceedings of the 20th Annual Precise Time

and Time Interval (PTTI) Application and Planning Meeting,

Naval Observatory, Washington DC, 1988, pp. 193-201.

Klobuchar, J.A., Ionospheric Effects on Earth-Space Propagation,

Environmental research paper No. 866, Air Force

Geophysics Laboratory, Hanscom, 1983, p 33.

Klobuchar, J.A., Ionospheric Time-Delay Algorithm for

Single-Frequency GPS Users, IEEE Transactions on

Aerospace and Electronic Systems, 1987, vol. 23, no.3, pp.

325–331.

Kouba, J.A., Guide to using International GNSS Service (IGS)

products, IGS Central Bureau, Pasadena, 2009, p. 34.

Lockwood, M., Wild, M.N., Stamper, R., Davis, C.J. and

Grande, M., Predicting Solar Disturbance Effects on

Navigation Systems, Journal of Navigation, 1999, vol. 52,

no. 2, pp. 203-216.

McMorrow, D., Impacts of Severe Space Weather on the Electric

Grid, Report JSR-11-320, The MITRE Corporation,

McLean, 2011, p. 107.

Mendillo, M., Storms in the ionosphere: Patterns and

processes for total electron content, Reviews of Geophysics,

2006, vol. 44, no. 4, p. 47.

Noll, C., The Crustal Dynamics Data Information System: A

resource to support scientific analysis using space

geodesy, Advances in Space Research, 2010, vol. 45, no. 12,

pp. 1421-1440.

Parkinson, B.W. and Spilker, Jr. J.J., Global Positioning

System: Theory and Applications, Vol. I, AIAA,

Washington DC, 1996.

Perrone, L. and De Franceschi, G., Solar, ionospheric and

geomagnetic indices, Annals of Geophysics, 1998, vol. 41,

no. 5-6, pp. 843-855.

Ross, W.J., The determination of ionospheric electron

content from satellite Doppler measurements: 2,

Experimental results, Journal of Geophysical Research,

1960, vol. 65, pp. 2607–2615.

Singer, H.J., Matheson, L., Grubb, R., Newman, A. and

Bouwer, S.D., Monitoring space weather with the GOES

magnetometers, Proc. SPIE 2812, GOES-8 and Beyond,

1996, p. 29.

Subirana, J. S., Zornoza, J. M. J. and Hernandez-Pajares, M.,

GNSS Data processing, Volume I: Fundamentals and

algorithms, ESA Communications, Noordwijk, 2013.

Thomas, M. et al., Global Navigation Space Systems: reliance

and vulnerabilities, RAENG, London, 2011.

Zolesi, B. and Cander, R. L. J, Ionospheric Prediction and

Forecasting, Springer, New York, 2014.