Keywords

1 Introduction

The European Plate Observing System (EPOS) is a European research infrastructure that brings European nations together with the goal to enhance the integration, accessibility, and utilization of solid Earth science data, products, services, and facilities. EPOS maintains a sustainable and cross-disciplinary central data portal (https://www.ics-c.epos-eu.org/) that offers open access to standardized metadata and quality-controlled data from various Earth science disciplines, accompanied by essential tools for analysis and modeling. (EPOS-ERIC 2017)

Fig. 1
figure 1

The EPOS-GNSS data dissemination workflow

Full size image

EPOS is organised in Thematic Core Services (TCS), such as TCS seismology, TCS geomagnetism, or TCS GNSS data and products (also called EPOS-GNSS). Each TCS organises the contribution of their specific community to EPOS. EPOS-GNSS is dedicated to delivering GNSS data, metadata, products, and software to EPOS (Fernandes et al. 2022). Within the scope of its participation in EPOS-GNSS, the Royal Observatory of Belgium is developing the EPOS-GNSS Data Quality Monitoring Service (DQMS) with as frontend the EPOS data quality monitoring web portal (https://gnssquality-epos.oma.be/). The primary goal of the DQMS is to oversee and evaluate the availability and quality of daily GNSS data that are discoverable through the EPOS-GNSS architecture. Similar GNSS data quality assessment procedures are also provided by EUREF Permanent Network (Bruyninx et al. 2019, 2022), Nevada Geodetic Laboratory (Blewitt 2021), and International GNSS Service (Johnston et al. 2017).

This paper describes the current status of the EPOS-GNSS data quality monitoring web portal and illustrates with some practical examples how the information provided by the portal can help interpreting GNSS data products targeting geodetic or geophysical applications.

2 EPOS-GNSS Data Dissemination Concepts

GNSS data become discoverable and accessible within EPOS after going through several procedures which aim at applying the same standard to GNSS (station and file) metadata and also ensuring the quality of GNSS data.

Figure 1 illustrates the EPOS-GNSS data dissemination workflow to deliver GNSS station and file metadata to the EPOS data portal. Currently, there are ten EPOS-GNSS data nodes that operate on top of GNSS data repositories. These data nodes use the open-source Geodetic Linkage Advance Software System (GLASS) (Fernandes et al. 2022), a set of softwares developed within EPOS-GNSS, to establish a virtualization layer on top of the data repository. GLASS performs several tasks, including indexing GNSS data files (in RINEX format), data validation with respect to station metadata from M3G (https://gnss-metadata.eu/) (Fabian et al. 2021), data quality checks using G-Nut/Anubis software (Vaclavovic and Dousa 2015), and it stores all relevant information in the local node database. Furthermore, GLASS ensures the synchronization of file metadata to the EPOS-GNSS Data Gateway (DGW), where all EPOS-GNSS data are accessible through an Application Program Interface (API) and a web portal.

3 EPOS-GNSS Data Quality Monitoring Portal

3.1 Introduction

The backend of the EPOS-GNSS DQMS retrieves from each of the EPOS-GNSS data nodes every day the list and metadata of available daily GNSS data as well as their data quality metrics. This information is then stored in the DQMS database and used to compute the GNSS data quality indicators (DQIs) from which a selection is presented and made available on the data quality monitoring portal (Bamahry et al. 2022).

This portal is structured into two main sections. The first section focuses on RINEX data availability. It allows to check the distribution and the availability of the GNSS data that are discoverable through EPOS and filter them by EPOS-GNSS data node, GNSS network, or M3G metadata maintainer. For each station, the resulting number of available daily GNSS data files are then graphically represented on a map (see Fig. 2).

Fig. 2
figure 2

Web page of the GNSS data quality monitoring web portal showing the map of the RINEX data availability

Full size image

The second section focuses on RINEX data quality and provides plots of several GNSS DQIs, such as:

  1. (a)

    The percentage of observed vs. expected observations computed as the ratio of the number of actual observations with respect to the number of expected observations for each constellation. This DQI plot provides the ratio on at least one frequency (1freq) or at least two frequencies (2+freq). 2+freq is more relevant for geodetic positioning applications as it is needed to reduce the effect of the ionospheric delay. Additionally, the lowest elevation cut-off observed is also provided on this plot to help identifying data quality degradations at low elevations.

  2. (b)

    The number of epochs without observations.

  3. (c)

    The number of observed satellites for each constellation tracked by the station.

  4. (d)

    The maximum number of observations counted for each frequency and each constellation tracked by the station.

  5. (e)

    The number of cycle slips represents the ratio of the number of identified phase cycle slips × 1000 with respect to the number of phase observations for each constellation tracked by the station.

  6. (f)

    The Standard Point Positioning (SPP) coordinates, representing the daily mean XYZ coordinates estimated from the code observations obtained from a specific satellite constellation. The plot shows the deviation of the estimated SPP coordinates with respect to their median value over the station history.

  7. (g)

    The multipath values on code observations represent the daily mean of code multipath per frequency band and satellite constellation. These values are calculated for all pseudo-range codes of all GNSS satellites providing dual-frequency carrier-phase observations. They can be used to characterize the quality of the observed signal and the site environment.

    Fig. 3
    figure 3

    Number of daily RINEX files discoverable from the EPOS-GNSS Data Gateway as function of time

    Full size image
    Fig. 4
    figure 4

    Map of the 1631 EPOS-GNSS stations with daily RINEX data; the color code shows the number of days with data quality metrics (converted in years). The 315 EPOS-GNSS stations without data are shown in grey

    Full size image

3.2 The Availability of EPOS-GNSS Data

Based on the information gathered within the DQMS database, the number of daily GNSS data discoverable through the EPOS-GNSS Data Gateway increases clearly over the time (see Fig. 3).

Currently there are 1631 stations with data distributed throughout ten active EPOS-GNSS data nodes (see Fig. 4). Almost 40% of GNSS data that are discoverable from the EPOS-GNSS Data Gateway are coming from the ROB-EUREF data node (see Fig. 5), which makes the daily RINEX data of the EUREF stations that agreed to share their data within EPOS available.

A significant number (35%) of the daily GNSS data files were indexed by the data nodes in 2023. This value can be attributed to the additional number of GNSS stations with long tracking history that agreed to share data with the EPOS this year (327 GNSS stations) as well as the official launch of the EPOS data portal at the end of April 2023. Figure 4 shows that there are 315 EPOS-GNSS stations without data, primarily because some EPOS-GNSS data nodes are still in the process of populating their databases.

Fig. 5
figure 5

Number of daily RINEX observation files discoverable from ten active data nodes

Full size image

3.3 The Quality of EPOS-GNSS Data

The evolution of GNSS data quality over time was calculated in this study to evaluate the data quality performance of all EPOS-GNSS stations. We determine the daily median value for constellation tracking performance: 2+freq above 0°, 1freq above 0°, and 1freq above 10° over all EPOS-GNSS stations. These values were smoothed over three-month periods to visualise long-term trends in tracking performance for each constellation.

Fig. 6
figure 6

Historical trends of observed vs. expected observations for each constellation. Upper panel: 3 months moving averages of the daily median overall station of 2+freq above 0°, 1freq above 0°, 1freq above 10° for GPS (blue), GLONASS (GLO, red), Galileo (GAL, green), and BeiDou (BDS, yellow). Lower panel: number of EPOS stations used to calculate these trends

Full size image
Fig. 7
figure 7

Degraded tracking at low elevation in 2003 caused spurious subsidence at the station ACOR00ESP. Upper panel: the percentage of the observed with respect to expected observations, minimum elevation is shown in purple. Lower panel: detrended position time series

Full size image

In Fig. 6, we observe that the values for GPS 2+freq above 0° over all EPOS-GNSS stations have been increasing 20% over the past 28 years. Conversely, GLONASS 2+freq above 0° exhibited an upward trend only until 2016. From 2016 on, the performance of GLONASS 2+freq above 0° slowly degraded until mid of 2021. This phenomenon was explained by Bruyninx et al. (2019), attributing the degradation of GLONASS 2+freq above 0° observed in EPN stations to several GLONASS satellites with impaired L2 observations. In contrast, both Galileo and BeiDou demonstrated an improvement in their performance at 2+freq above 0° after mid 2019. In addition, we also can see that the number of stations that track GLONASS observations has dramatically decreased in 2010. Based on our investigation, this feature is due to a bug in the software used by EPOS-GNSS and this period is currently being reprocessed by the nodes in order to solve the issue.

The degraded quality of daily GNSS data can impact the accuracy of GNSS products. We used the detrended GNSS position time series from Nevada Geodetic Laboratory (Blewitt et al. 2018) of the three stations (ACOR00ESP, BUCK00GBR, TRO100NOR) to illustrate how GNSS DQIs plots can help to correctly interpret GNSS position time series. Figure 7 depicts the degraded tracking observed at low elevation at ACOR00ESP station in 2003, resulting in spurious subsidence in its position time series. Similarly, undocumented elevation cut off changes in 2014 at the BUCK00GBR station have a modest effect on its position time series (see Fig. 8). As response to these issues, we are presently developing an algorithm for automatically detecting changes in elevation cut-off angle using Online Bayesian method (Adams and Mackay, 2007) and provide the information to station owners with the final goal to correct station metadata.

Fig. 8
figure 8

Example of an undocumented change affecting the cut off angle in 2014 for the station BUCK00GBR. Upper panel: the percentage of observed with respect to expected observations, minimum elevation is shown in purple. Lower panel: detrended position time series, a purple vertical line in 2014-05-16 shows an un-documented change of the cut off angle observed

Full size image

In Fig. 9, a strong correlation is evident between anomalies in the cycle slips, multipath values, and the position time series at TRO100NOR station in 2000–2004. There is no specific information available concerning the cause of this quality degradation. Nevertheless, given the seasonal signals in the DQI, the data quality degradation could be attributed to environmental conditions in the vicinity of the antenna.

4 Summary and Future Works

The new DQMS web portal monitors the availability and quality of the daily GNSS data that are discoverable through the EPOS-GNSS Data Gateway and EPOS Data Portal. In this paper, we outlined several use cases and analyzed GNSS data quality indicator plots available from this web portal. Drawing from our experience at ROB, these data quality indicators are employed to gain insights into outliers or irregularities within position time series. However, it is important to note that the correlation between these indicators and position time series is very complex. Consequently, we are also exploring the utilization of artificial intelligence to better comprehend the complex correlation between DQIs and position time series, enabling us to automatically identify degraded GNSS data quality that might be contributing to outliers or anomalies in position time series (Bamahry et al. 2023).

Fig. 9
figure 9

Degraded tracking performance between 2000 and 2004 affecting the number of cycle slips, the multipath values and the position time series for the station TRO100NOR. Upper panel: number of cycle slips per number of observations. Middle panel: daily mean of code multipath per frequency band and satellite constellation. Lower panel: detrended position time series

Full size image