The technique of the orthogonal projection of ionosphere electronic content variations for mapping total electron content (TEC) allows us to visualize ionospheric irregularities. For the reconstruction of global ionospheric characteristics, numerous global navigation satellite system (GNSS) receivers located in different regions of the Earth are used as sensors. We used dense GNSS networks in central Europe to detect and investigate a special type of plasma inhomogeneities, called travelling ionospheric disturbances (TID). Such use of GNSS sensors allows us to reconstruct the main TID parameters, such as spatial dimensions, velocities, and directions of their movement. The paper gives examples of the restoration of dynamic characteristics of ionospheric irregularities for quiet and disturbed geophysical conditions. Special attention is paid to the dynamics of ionospheric disturbances stimulated by the magnetic storms of two St. Patrick’s Days (17 March 2013 and 2015). Additional opportunities for the remote sensing of the ionosphere with the use of dense regional networks of GNSS receiving sensors have been noted too.

Advanced processing of collected global navigation satellite systems (GNSS) observations allows for the estimation of zenith tropospheric delay (ZTD), which in turn can be converted to the integrated water vapour (IWV). The proper estimation of GNSS IWV can be affected by the adopted GNSS processing strategy. To verify which of its elements cause deterioration and which improve the estimated GNSS IWV, we conducted eight reprocessings of 20 years of GPS observations (01.1996–12.2015). In each of them, we applied a different mapping function, the zenith hydrostatic delay (ZHD) a priori value, the cut-off angle, software, and the positioning method. Obtained in such a way, the ZTD time series were converted to the IWV using the meteorological parameters sourced from the ERA-Interim. Then, based on them, the long-term parameters were estimated and compared to those obtained from the IWV derived from the radio sounding (RS) observations. In this paper, we analyzed long-term parameters such as IWV mean values, linear trends, and amplitudes of annual and semiannual oscillations. A comparative analysis showed, inter alia, that in terms of the investigation of the IWV linear trend the precise point positioning (PPP) method is characterized by higher accuracy than the differential one. It was also found that using the GPT2 model and the higher elevation mask brings benefits to the GNSS IWV linear trend estimation.

The spatial correlation between different stationsfor individual components in the regional GNSS networksseems to be significant. The mismodelling in satelliteorbits, the Earth orientation parameters (EOP), largescaleatmospheric effects or satellite antenna phase centrecorrections can all cause the regionally correlated errors.This kind of GPS time series errors are referred to ascommon mode errors (CMEs). They are usually estimatedwith the regional spatial filtering, such as the "stacking".In this paper, we show the stacking approach for the setof ASG-EUPOS permanent stations, assuming that spatialdistribution of the CME is uniform over the whole regionof Poland (more than 600 km extent). The ASG-EUPOSis a multifunctional precise positioning system based onthe reference network designed for Poland. We used a 5-year span time series (2008-2012) of daily solutions in theITRF2008 from Bernese 5.0 processed by the Military Universityof Technology EPN Local Analysis Centre (MUTLAC). At the beginning of our analyses concerning spatialdependencies, the correlation coefficients between eachpair of the stations in the GNSS network were calculated.This analysis shows that spatio-temporal behaviour of theGPS-derived time series is not purely random, but there isthe evident uniform spatial response. In order to quantifythe influence of filtering using CME, the norms L1 and L2were determined. The values of these norms were calculatedfor the North, East and Up components twice: beforeperforming the filtration and after stacking. The observedreduction of the L1 and L2 norms was up to 30% dependingon the dimension of the network. However, the questionhow to define an optimal size of CME-analysed subnetworkremains unanswered in this research, due to thefact that our network is not extended enough.