Return to Session D3

Session D3: GNSS Augmentation and Robustness for Autonomous Navigation

Ionospheric Effects on GNSS RTK
Uttama Dutta, RISE Research Institutes of Sweden & Chalmers University of Technology; Per Jarlemark, Carsten Rieck, RISE; and Jan Johansson, RISE & Chalmers University of Technology

Peer Reviewed

GNSS signals are influenced by free electrons as they propagate through the ionosphere. Studies have shown how the spatial variations of electron density in the ionosphere affects measurements with network-RTK [reference]. This paper aims to predict what can be expected from measurements during the next solar maximum that is expected to occur around 2025. The ionospheric activity and its impact on positioning in the coming solar cycle maximum is discussed. In order to perform a spatial characterization of the ionosphere, archived GPS data from the Swedish Network of permanent GNSS reference stations, SWEPOS, over an extensive period of time has been used. GNSS data from Kiruna, mainly captured in January 2014 – in the middle of the most active time during the last cycle – has been analyzed to predict the coming solar cycle. Based on the data, it was concluded that there is a risk of occasions with simultaneous signal slips on several satellites caused by the ionosphere which could cause temporary (minutes) loss of positioning ability for the RTK equipment. It is expected to occur a couple of times per month during the most active months of the solar cycle.
Radiation from the sun creates the ions and free electrons in the ionosphere, and a peak in the concentration of free electrons follow the sun around the earth. Most free electrons recombine with the ions at sunset. However, at the polar regions some free electrons can sweep from the daytime side to nighttime side of the earth guided by the magnetic field. The free electrons interact with GNSS signals and delay the reception. The frequency dependence of this delay enables us to form an “ionosphere-free” combination of satellite signals at two frequencies.
These ionosphere-free combinations are used in static precise positioning with GNSS, but seldom in RTK due to their larger measurement errors. In RTK the total signal delay at the rover equipment is compared to the total signal delay at a reference. The rover coordinates are calculated from the total delay differences for the visible satellites. The differences in the ionospheric part of the delay at reference and rover are therefore measurement errors that contribute to coordinate errors. The shorter the baseline between the reference and the rover, the smaller errors.
To summarize the major ionospheric effects on RTK measurements it can be said that large ionospheric delay differences contribute to large measurement errors that results in larger coordinate errors. Such large differences can also complicate the “phase ambiguity” resolution necessary for a RTK fix. Finally, ionospheric effects can also result in “signal slips” which are temporary loss of the signal causing sudden jumps in the phase data of a receiver. Large variations in electron concentration can be the reason for such reduction of the signal strength. Solar activity with increased UV, X-ray, and particle emissions has an 11-year cycle. The cycle has been followed as variations in the number of visible spots on the solar surface for nearly 300 years. The increased activity at a solar cycle maximum also leads to increased ionospheric activity. A way to predict the coming high ionospheric activity at the pits is therefore to study the activity in a previous solar cycle.
In order to predict the ionospheric situation in northern Europe in the coming years, GPS data from two receivers in Kiruna recorded during the last solar cycle maximum were studied. Polar region effects for GNSS receivers can be expected for Kiruna sites as they are located at high latitudes (68°N). Data set during 1-31 January 2014 was selected for this study. Figure(1) shows that January 2014 is in the middle of the most active period (November 2013 – April 2014) during the last solar cycle maximum. The calm period during 1-14 March 2019, and a recent period during 1-14 March 2021, were also included for comparison. Figure 1: The daily number of sunspots registered for the two last solar cycles. The periods investigated with GNSS data are marked as red “+” in the graph. The Kiruna receivers are separated by a baseline of 6 km in length which is a good choice for studying ionospheric effects for differential positioning techniques. In order to mimic the relatively high elevation angle cut-off in cities and forest environment, only satellite data with an elevation cutoff of 25° in Kiruna were used. In Figure (2), the GPS L1 ionospheric delay from one of the stations in Kiruna has been presented. Studies show that the delay in 2014 is significantly larger than that for the other sets. The meter-size of the uncompensated ionospheric delay can be easily noted from Figure (2). In the case of RTK, after differencing rover data with reference data from a station some kilometers away, only cm-level ionospheric delay differences will remain (as shown in Figure 3 below).
Figure 2: The ionospheric delay for the GPS L1 at one of the Kiruna receivers during the three studied periods in 2014, 2019, and 2021.
The ionospheric delay differences between the two Kiruna sites for GPS L1 using the “geometry-free” signal combination [3], were calculated. The standard deviations in 5-minute
intervals are presented in Figure (3). In the statistics, possible jumps in the signals, caused by signal slips, were removed before the standard deviation was calculated. (The slips are treated separately below.) Figure 3: The standard deviation of the ionospheric delay differences for the GPS L1 at one of the Kiruna receivers during the three studied periods in 2014, 2019, and 2021. In calculations below these delay differences are treated as the main source of measurement errors. As expected, the period from 2014 has the largest deviations. They are, smaller than 30 mm most of the time, and very seldom larger than 50 mm in the Kiruna data. The Kiruna baseline between the receivers is 6 km, significantly larger than in many normal RTK situations in surveying. Hence, the variations in the ionospheric delay differences between reference receivers and rovers is expected to be significantly smaller than for the Kiruna network. To get a rough idea of the expected consequences on coordinate estimates in a setup with, say, typically 1-2 km baselines one can scale the standard deviations of Kiruna by dividing by three. This means that the ionospheric measurement error standard deviation, ???? , would be smaller than 10 mm most of the time. With a PDOP value of, for example, 4 a contribution to the 3D coordinate error standard deviation of less than 40 mm is achieved most of the time, even in the worst periods of the solar cycle. It should be noted that emerging applications, such as automotive or maritime positioning with RTK techniques, will use fixed grid RTK networks with rather large mesh sizes, possibly in the order of 5-10 km.
Ionosphere induced slips are often caused by large, and maybe rapid, variations in electron concentration that reduce the strength of the satellite signal. A slip makes the specific satellite temporarily useless in the coordinate determination of the rover. In the studied Kiruna data sets, ionosphere induced signal slips (sudden changes in the recorded phase data from a satellite) were searched. Search for slips were conducted in either of the two main signals from the satellites (e.g., both L1 and L2 for GPS) and in the delay differences between reference and rover equipment. 1 second sampled data was used to find rapid changes that are characteristic for slips. Data were divided into 5-minute intervals, in order to count the number of almost simultaneously slipping satellite signals. The signals on the two main satellite frequencies were combined to an “ionosphere free” signal. In each interval, the one satellite with the smallest signal variations was chosen as reference, assuming this satellite had no slips in the time interval. By computing double differences (difference of the delay between the reference satellite and another satellite, differenced between the two receivers) ionospheric slips appear as distinct signal deviations, as shown in Figure (4). A slip is considered to occur when a group of at least three points deviate significantly (±75 mm) from the median value of the interval. Figure 4: The double difference between GPS satellites 21 and 18 received at the two Kiruna receiver sites on March 13, 2021. The “ionosphere-free” L3 combination of L1 and conventional L2 was used. Slips appear as distinct deviations in the data. Results for January 2014 Analysis of the data from this month gives an indication of what to expect during the most intensive times of the coming maximum. In the 2014 data only GPS and GLONASS satellites were tracked. The number of slipping satellite signals in the 5-minute intervals are shown in Figure (5). In total 89 slips were detected for this month, on the average 1.4 slips per day and per system. Most of the time the slips appear isolated, on either a GPS or a GLONASS satellite. There are however exceptions with even 16 slips within an hour (during 22-23 local time on 1st Jan 2014).
Figure 5: The number of satellites with detected signal slips in each 5-minute interval in January 2014. The counts are for the two systems, GPS and GLONASS separately, and the bottom graph also presents the total number of slipping satellites in black. At instances of isolated slips, it is likely, that the rover receiver can maintain accurate RTK positioning if the observation conditions are good (such as 5-6 non-slipping satellites, PDOP significantly better than 5). Isolated slips at worse observation conditions will lead to loss of RTK-fix and the positioning accuracy will deteriorate. There could be minutes with less accurate positioning. The time to reestablish an RTK-fix positioning solution will depend on the sizes of the ionospheric delay differences. During this time (usually in minutes) before the ambiguities have been resolved, the positioning mode is “RTK float” with decimeter-level accuracy or “DGNSS” with 0.5 – 1-meter accuracy.
At instances with many (almost) simultaneous slips, (such as during 22-23 local time on 1st Jan 2014), there might be a total loss of positioning for some minutes. However, these occasions are very rare. The slip distribution as function of time of day is shown in Figure (6). They often occur at night, as expected. There are, however, a significant part also at daytime. It is likely that the not all of these are induced by the ionosphere; there could also be, e.g., accumulation of wet snow on the antennas and man-made disturbances.
Figure 6: The detected signal slips in January 2014 as function of the local solar time.
Results for March 2021 In the 2021 data GPS, GLONASS, Galileo, and BeiDou are all present. It was expected to see similar pattern for the number of slips for all systems. However, this was not the case with GPS. The present weakness of conventional GPS L2 led to a domination of GPS slips, 101 out of 105 slips as shown in Figure (7).
Figure 7: The number of signal slips in the 2021 data. Left part: the slips when conventional L2 is used for GPS. 101 out of 105 slips originates from GPS. Right part: when L2C is used for GPS. In total 6 slips, 1 or 2 for each system When the use of the L2C version of GPS L2 was made instead, almost all GPS slips disappear; and only two slips remain. L2C is a modernized version, which contains a coding accessible for civilian users, facilitating signal tracking. It is vital that this difference in signal quality between L2 versions is considered when selecting signals for use in the RTK application. However, the settings in state-of-the-art RTK receivers, i.e., base and rover, is unclear since all GPS satellites still not transmit L2C. The optimal setting has to be checked with the RTK receiver manufacturer. In the March 2021 dataset, 23 out of 31 satellites had the L2C signal version, and the number of satellites will increase. Slip detection on other carriers and modulations will be presented as well.

The peak in the ionospheric activity is predicted to occur in 2025. In the years 2024-2026, a couple of months with higher frequency of slipping satellite observations can be expected,
corresponding to the situation in January 2014. It can be expected that in most slip instances in these months, only 1 or 2 satellites will slip, and not cause loss of RTK-fix accuracy, provided that the observation conditions are good. For worse observation conditions, there may be temporary fall back to the less accurate “RTK float” or “DGNSS” for considerable time not acceptable by the user application. It is also likely that there will be rare occasions with many simultaneous slips, a few times per month in the worst months, where there could be a total loss of positioning for some minutes. It should be remembered that predictions of the situation in the coming solar cycle based on old data is uncertain. The coming solar cycle maximum could, e.g., be worse than the previous. On the other hand, the receiver’s satellite tracking abilities could be improved since the data of January 2014 were captured, thereby reducing the probability of signal slips. Finally, it is worth noting the availability of several existing services for ionospheric prediction. NOAA has a space weather prediction service that can provide warnings [4]. Lantmäteriet’s SWEPOS ionosphere monitor can provide information on the situation in Sweden at the moment [5]. Some care is required using these services. Regarding the NOAA service it should be remembered that these are predictions. When looking at the Swepos ionosphere monitor the graphs are based on real data and the ionospheric influence on the achievable position estimates are calculated in near real time. The SWEPOS monitor is developed for Network RTK (not RTK), and the quality measures presented are not directly applicable for the situation in conventional RTK. Actually, the situation may be better in a local RTK system since the
baseline length between the base station and the rover receivers are generally much shorter.

1. Odolinski, R., Teunissen, P. J. G., and Odijk D., Combined BDS, Galileo, QZSS and GPS single-frequency RTK. GPS Solutions (2015) 19:151–163. DOI 10.1007/s10291-
2. Emardson, R., Jarlemark, P., Bergstrand S., Nilsson T., and Johansson J., Measurement Accuracy in Network-RTK. SP report 2009:23, ISBN 978-91-86319-10-6, 2009.
Available for download at RISE.
3. Hofmann-Wellenhof, B., Lichtenegger, H.,K. and Wasle, E., GNSS - Global Navigation Satellite Systems. Springer-Verlag, Wien, Austria, 2008

Return to Session D3