Evaluation of tropospheric estimates from CentipedeRTK, a collaborative network of low-cost GNSS stations

Abstract

AbstractThe CentipedeRTK network is a collaborative Global Navigation Satellite System (GNSS) network launched in 2019, consisting mainly of low-cost GNSS receivers and antennas. This network enables free Real-Time Kinematic (RTK) positioning with centimeter accuracy for all users. The raw GNSS measurements from the CentipedeRTK network are routinely archived by the French scientific network RÉseau NAtional GNSS permanent, with the aim of exploiting raw GNSS measurements for geoscience applications. This paper presents a first assessment of the use of this dataset for tropospheric monitoring. We considered all the data provided in 2023 by more than 400 low-cost GNSS stations in mainland France. After selecting the stations with dual-frequency observations over the period, the data of 331 stations were analyzed using precise point positioning , resulting in a set of 265 stations satisfying our screening procedure and providing data covering more than 50% of the year 2023. A first indication of the quality of the analysis is given by the repeatability of the stations, of the order of $$2.2\pm 1.1$$ 2.2 ± 1.1 , $$2.1\pm 0.8$$ 2.1 ± 0.8 and $$6.9\pm 2.6$$ 6.9 ± 2.6  mm respectively on the East, North and Up components. These values are slightly higher than those obtained for nearby conventional stations, especially for the vertical component ($$5.4\pm 0.8$$ 5.4 ± 0.8  mm). The tropospheric delays were compared with those retrieved from nearby GNSS reference stations (less than 30 km away) belonging to conventional networks (186 stations considered). The comparison shows a good agreement between low-cost and conventional stations, with a root mean square of differences of $$7.4\pm 3.0$$ 7.4 ± 3.0  mm; a mean bias of 2.7 mm is highlighted and shown to be stable over time; its origin has not yet been determined but its magnitude seems related to the antenna type of the CentipedeRTK stations. In a second step, the integrated water vapor content were derived from the tropospheric delays and compared with those of the European Centre for Medium-range Weather Forecasts fifth reanalysis (ERA5). Only stations located at an altitude less than 100 m around the ERA5 orography were considered (240 stations). The differences between the two techniques are similar to those reported in the literature for traditional networks, with a mean bias of $$0.06\pm 0.82$$ 0.06 ± 0.82  kg m$$^{-2}$$ - 2 and a mean standard deviation of $$1.48\pm 0.18$$ 1.48 ± 0.18  kg m$$^{-2}$$ - 2 . This again confirms the quality of the dataset. Finally, the value of such low-cost stations for monitoring and describing meteorological phenomena is illustrated by the study of an atmospheric river affecting the central–western part of France in December 2023. All these results underline the considerable potential of low-cost GNSS networks in geoscience applications, especially in regions with limited instrumentation. Their role could be particularly important in meteorological or climatological contexts, where GNSS-based water vapor monitoring is widely used.