Tropospheric Products from High-Level GNSS Processing in Latin America

The present geodetic reference frame in Latin America and the Caribbean is given by a network of about 400 continuously operating GNSS stations. These stations are routinely processed by ten Analysis Centres following the guidelines and standards set up by the International Earth Rotation and Reference Systems Service (IERS) and International GNSS Service (IGS). The Analysis Centres estimate daily and weekly station positions and station zenith tropospheric path delays (ZTD) with an hourly sampling rate. This contribution presents some attempts aiming at combining the individual ZTD estimations to generate consistent troposphere solutions over the entire region and to provide reliable time series of troposphere parameters, to be used as a reference. The study covers ZTD and IWV series for a time-span of 5 years (2014–2018). In addition to the combination of the individual solutions, some advances based on the precise point positioning technique using BNC software (BKG NTRIP Client) and Bernese GNSS Software V.5.2 are presented. Results are validated using the IGS ZTD products and radiosonde IWV data. The agreement was evaluated in terms of mean bias and rms of the ZTD differences w.r.t IGS products (mean bias (cid:2) 1.5 mm and mean rms 6.8 mm) and w.r.t ZTD from radiosonde data (mean bias (cid:2) 2 mm and mean rms 7.5 mm). IWV differences w.r.t radiosonde IWV data (mean bias 0.41 kg/m 2 and mean rms 3.5 kg/m 2 ).


Introduction
Integrated Water Vapour (IWV) plays a fundamental role in several weather processes that deeply influence human activities. Retrieving IWV content in the atmosphere can be performed in different ways using independent techniques: from the traditional ones like radiosondes and ground-based microwave radiometers, up to the recent ones based on satellite techniques. In particular, the GNSS-based tropospheric Zenith Total Delay (ZTD) estimates allow inferring IWV values with high accuracy equivalent to that expected from direct observational techniques, such as radiosondes and microwave radiometers (Bonafoni et al. 2013;Van Baelen et al. 2005;Calori et al. 2016). Several studies have been devoted to the use of GNSS stations for the estimation of IWV over South America. Bianchi et al. (2016) estimated mean IWV based on GNSS data (IWV GNSS ) and its trends during 2007-2013 over more than a hundred GNSS tracking sites from SIRGAS-CON. Calori et al. (2016) analysed a period of 45 days where deep convective processes with hail precipitation took place over Mendoza province, in the Central-Western Argentina (CWA). For this assessment, the authors used IWV GNSS maps to draw insight into the accumulation and influence of humidity over the region. Even fewer studies have performed a validation of the IWV GNSS ; for this, Fernández et al. (2010) used radiosonde data from four locations over Central-North Argentina in order to validate IWV estimates from Global Positioning System (GPS) stations during a 1-year period (2006)(2007). The authors found an agreement between IWV GNSS and IWV estimated through radiosonde data (IWV RS ), with differences as large as 3 kg/m 2 . Llamedo et al. (2017) used GPSderived IWV to analyse moisture anomalies over South America during El Niño-Southern Oscillation phases, finding positive anomalies over northern Argentina during El Niño events. Camisay et al. (2020) estimated IWV GNSS time series for a 4-year period (2015)(2016)(2017)(2018), to assess the accuracy through a comparison in two GNSS Argentinean stations with radiosonde observations and explore the role of IWV in the development of regional precipitation events over the CWA. The obtained agreement between IWV GPS and IWV RS was close to 2 kg/m 2 in terms of mean absolute error. In Latin-American region, in situ meteorological observations are scarce; therefore, GNSS atmospheric monitoring has significant relevance for the understanding of regional meteorological processes. This kind of information is extremely valuable, and it can be used to achieve a better knowledge of IWV variable in the study region.
The GNSS allows monitoring the IWV from a network that surpasses traditional techniques due to its significant temporal and spacial density. This is of interest to study the regional trends of the climatic variable for which it is necessary to have a long time series by site and region. On the other hand, the ZTD can be estimated in real-time and near real-time mode, in order to be assimilated in regional forecast models. SIRGAS (Sistema de Referencia Geocéntrico para las Américas) is the geocentric reference frame in Latin America and the Caribbean. It is at present given by a network of about 420 continuously operating GNSS stations (Cioce et al. 2018) (Fig. 1). These stations are routinely processed by the SIRGAS Analysis Centres (AC), following the guidelines and standards set up by the International Earth Rotation and Reference Systems Service (IERS) and International GNSS Service (IGS). Since 2014, the routine GNSS data processing includes the estimation of hourly ZTD values based on GPS and GLONASS observations (Camisay et al. 2020;Sánchez et al. 2015;Brunini et al. 2012). Pacione et al. (2017) shows the great potential that a continental GNSS network offers in atmospheric studies. EUREF Permanent Network (EPN) (Bruyninx et al. 2019) had been used as a valuable database for the development of a climate data record of GNSS tropospheric products over Europe. It had been used as a reference in the regional numerical weather prediction reanalyses and climate model simulations and had been used for monitoring IWV trends and variability. Guerova et al. (2016) showed and discussed the advantages of the application of GNSS tropospheric products in operational weather prediction and in the climate monitoring.
In this contribution, we report on the estimation and validation of the ZTD and IWV values in Latin America GNSS stations, using as input data the ZTD values obtained in: (1) the operational processing of the SIRGAS regional reference frame and (2) applying the Precise Point Positioning (PPP) approach, with two softwares, BKG NTRIP Client (BNC) and Bernese v5.2. (BSW52). To assess the reliability of our results (ZTD and IWV values), they are compared with the operational IGS products (ZTD IGS ), IWV values extracted from radiosonde profiles (IWV RS ) and ZTD estimations inferred from integrate the correspondent radiosonde profile data (ZTD RS ).
In Sect. 2, the methodology used in operational SIRGAS processing to estimate ZTD product is reviewed. ZTD SIR internal consistency is presented. ZTD products estimated by PPP in SIRGAS stations are reviewed. Section 3 summarises the ZTD SIR and IWV SIR products validation with respect to ZTD IGS products and IWV radiosonde data. Conclusions, outlook and future work are given in Sect. 4.
The SIRGAS operational ZTD products (ZTD SIR ) are calculated with the final IGS products (orbits and earth rotation parameters, ERP). Table 2 summarizes the methodology implemented for the operational SIRGAS products and the testing PPP products.
Each SIRGAS-AC processes a different sub-network of SIRGAS GNSS stations. The distribution of the stations considers that each station parameter (ZTD i ) is available in three different solutions, so it is possible to evaluate the  internal consistency and generate the final combined ZTD products (ZTD SIR ). The ZTD i variance is used as a filter (¢ ZTD > 0.02 m), prior to the combination. The 5% of the ZTD i values are rejected in the analysed period. Table 3 shows the number of rejected estimates (in %) for each AC.

ZTD SIR Internal Consistency
Aweighted least-squares combination scheme using the inverse of the input data variances (¢ ZTD ) as a weighting factor is implemented to estimate ZTD SIR products. Figure 2 shows a detail per year of the number of stations in which the ZTD i data (3 or more solutions available, with ¢ ZTD < 0.02 m) are combined (Nc) compared to the number of stations that had only one solution. For the years 2015-2018 it was possible to have a data redundancy in more than 75% of the stations. The internal consistency of the ZTD SIR values is evaluated considering the residuals of each contributing ZTD solutions with respect to the combined ZTD value (ZTD i -ZTD SIR ). After a weighted least squares combination process, rms of each ZTD SIR parameters are determined. A mean rms is calculated per station and per year ( Table 4). The mean rms  is less than 1 mm in more than the 84% of the estimated values in the period 2014-2018 (Fig. 3).

ZTD SIR Validation with Radiosonde Data
ZTD SIR also, are compared with ZTD values calculated from data of 10 radiosonde stations (ZTD RS ).  Askne and Nordius (1987) with the physical constants for atmospheric refractivity from Rüeger (2002) (Eq. 1). The mean temperature of the atmosphere (Tm) used in (1) is calculated integrating the radiosonde profiles data (temperature and dew-point) in each level profiles up to GNSS station height (Eq. 2). The zenith hydrostatic delay values at the RS sites (ZHD RS ) are obtained according to Davis et al. (1985) (Eq. 3), where pressure is calculated to the GNSS height (P GNSS ) from pressure radiosonde data. An adaptation to the standard pressure model of Berg (1948) to correct for the height differences is applied (Eq. 4).

ZTD Estimation Applying PPP
In order to have a product in near real time to be used in numerical weather prediction model, we tested the Precise This estimation approached with two softwares, BNC (Weber et al. 2016) and BSW52, in the first case study. PPP with BSW52 showed better results (not shown). In the second period (year 2019) we decided to estimate ZTD PPP only by BSW52. In both cases of study, with BSW52 PPP, rapid IGS products (orbits, ERP and satellite clock corrections) were used so the ZTD PPP were estimated with 24 h delay. Table 2 summarizes the input data, models and main configuration used for each software.

Determination of IWV Values from GNSS-Based ZTD Estimates
The GNSS-based ZTD values are used to calculate the IWV applying the ratio of Askne and Nordius (1987) to the wet component of the delay (ZWD), (Eq. 1). In this work, the ZTD SIR and the one from applying PPP (from BSW52) were used. ZWD values were obtained by removing the ZHD, which was calculated according to Davis et al. (1985) (5) In this case, the weighted mean temperature of the atmosphere (Tm) was calculated in accordance with Mendes (1999) using the surface temperature (Ts) also provided by ERA-Interim. The values for the refractivity constants were taken from Rüeger (2002). Following this strategy, IWV SIR series from a 5 years (2014-2018) period were estimated in each SIRGAS station. We generated four daily IWV maps by Hunter (2007) (at 00:00, 06:00, 12:00 and 18:00 UTC) for the entire SIRGAS region, see some examples in Fig. 5 (24-6-2018).
The IWV SIR values were tested in the 10 radiosonde stations selected (Table 5). The Figs. 6 and 7 show the comparison of IWV SIR (inferred from ZTD SIR ) values with values obtained from radiosonde profiles (IWV RS ) at two SIRGAS stations: MZAC (located in an arid region) and IGM1 (located in a humid region), respectively.

ZTD SIR Validation
Our results presented a quite good agreement with the IGS products (see Fig. 4). Discrepancies between ZTD SIR and ZTD IGS values are compared at 15 IGS (SIRGAS) stations (Fig. 8). The results present a mean root mean square (rms) value of 6.8 mm (0.29% of the mean ZTD) with a negative mean bias of 1.5 mm (0.07% of the mean ZTD). The comparison of ZTD SIR w.r.t. ZTD RS is also very promising: discrepancies computed at 10 radiosonde stations (see Fig. 1 and Table 5) have a mean rms of 7.5 mm (0.32% of the mean ZTD) and a negative mean bias of 2 mm (0.09%

ZTD PPP Products Validation
Analysing the ZTD PPP products, the BSW52-based ZTD PPP estimates showed a better agreement than the BNC-based ZTD PPP estimates with respect to the corresponding ZTD SIR values. The rms and bias are the two indexes for the evaluation of the two estimations. Results of these two-test data set are shown in Table 6. BNC-based ZTD PPP estimates were less accurate as expected because real time IGS product were used. It may also be a consequence of the fact that ZTD SIR and the BSW52-based ZTD PPP use the same models to determine the tropospheric parameters. In the case 2 a bias-reduction scheme was implemented on a monthly basis as applied in Douša and Vaclavovic (2014). The comparison of the BSW52-based ZTD PPP estimates and ZTD SIR values at two selected SIRGAS station, EBYP (in a subtropical region) and MGUE (in an arid region), with the data in the case 1, are shown in Fig. 9.
The discrepancies between the ZTD PPP values estimated in the second case of study (Year 2019, 30 stations) with the respectively ZTD SIR values were also very promising (Fig. 10). The mean rms and mean bias per station is shown in the Fig. 10. The 84% of the stations had a mean rms < 28 mm and the rest 16% had a mean rms < 31 mm. In five GNSS stations, the BSW52-based ZTD PPP estimates were validated with respect to ZTD RS (detailed in bold in Table 5). Figure 11 shows this comparison in the IGS (SIRGAS) station CORD in the centre of Argentina, as an example.

IWV SIR Validation
The IWV SIR validation with IWV RS also showed agreement.
The results for a period of 5 years, in 10 RS -GNSS locations yielded a mean bias 0.41 kg/m 2 and a mean rms 3.5 kg/m 2 . The correlation coefficient of the two series (IWV SIR and IWV RS ) presented in Fig. 12 is 0.94, which indicates a very good agreement between both estimations. In the other hand, the comparison of IWV PPP (calculated from the BWS52-based ZTD PPP values) with IWV RS , produces discrepancies with a mean rms of 1 kg/m 2 , a standard deviation of 0.73 kg/m 2 and a bias of 2.37 kg/m 2 (not shown).

Conclusions
Latin America has SIRGAS network, an infrastructure of GNSS stations that generates ZTD (per hour), offering regional and continental coverage that can be used in atmospheric studies. The internal consistency of the ZTD SIR values, calculated by SIRGAS ACs, have been evaluated for a period of 5 years (2014)(2015)(2016)(2017)(2018)). An average rms less than 1 mm, in more than the 84% of the values, indicate the rigorous weighted least squares combination process implemented to get the SIRGAS reference products.
The ZTD SIR series for a 5-year period have been validated with two different time series. They agree with the corresponding values of the ZTD series obtained by the IGS (mean rms 6.8 mm; mean bias 1.5 mm) as well as those from the radiosonde technique (mean rms 7.5 mm; mean bias 2 mm). The ZTD obtained by PPP with BSW52, using the RAPID CODE products (ephemeris and clock corrections) are validated with respect to the post-processing products ZTD SIR . The mean rms of the differences is 22 mm (84% of the stations had a mean rms < 28 mm) for an annual case of study (2019, 30 stations). It remains to continue improving the methodology to increase accuracy and decrease the positive bias that on average resulted in 2 mm (0.07% of the ZTD mean value in the stations evaluated). Anyway, these accuracy of ZTD PPP complies with the threshold requirements for the operational NWP nowcasting -the relative accuracy of 5 kg/m 2 in integrated water vapor (IWV) and 30 mm in ZTD when approximating the conversion factor defined by Bevis  (1994) and Douša and Vaclavovic (2014). However, we must work to obtain a product in near real time (with 90 min of latency), applying ultra-rapid orbits and clocks, or even better using real-time corrections (Guerova et al. 2016).
The publication of this new product from SIRGAS opens the opportunity for new research topics that can be carried out both continentally and regionally in Latin America. As an example, it has been shown that SIRGAS ZTD products can be used to calculate the IWV over SIRGAS stations, thus providing IWV with a spatial and temporal density not existing in Latin America by conventional techniques. This variable has also been validated with radiosonde data (mean correlation coefficient 0.89, in 10 compared sites). SIRGAS ZTD products can be used as a reference for different scientific applications (e.g. validation of regional numerical weather prediction reanalyses) and they could be used for monitoring trends and variability in atmospheric water vapour in Latin America region, similar than EUREF Permanent network (Pacione et al. 2017).
Open Access This chapter is licensed under the terms of the Creative Commons Attribution 4.0 International Licence (http://creativecommons. org/licenses/by/4.0/), which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence and indicate if changes were made.
The images or other third party material in this chapter are included in the chapter's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the chapter's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.