Atmospheric water vapour plays a key role in the Arctic radiation budget,
hydrological cycle and hence climate, but its measurement with high accuracy
remains an important challenge. Total column water vapour (TCWV) datasets
derived from ground-based GNSS measurements are used to assess the quality of
different existing satellite TCWV datasets, namely from the Moderate
Resolution Imaging Spectroradiometer (MODIS), the Atmospheric Infrared
Sounder (AIRS) and the SCanning Imaging Absorption spectroMeter for
Atmospheric CHartographY (SCIAMACHY). The comparisons between GNSS and
satellite data are carried out for three reference Arctic observation sites
(Sodankylä, Ny-Ålesund and Thule) where long homogeneous GNSS time series of
more than a decade (2001–2014) are available. We select hourly GNSS data that
are coincident with overpasses of the different satellites over the three sites
and then average them into monthly means that are compared with monthly mean
satellite products for different seasons. The agreement between GNSS and
satellite time series is generally within 5
Water vapour has an important role in the Earth radiative balance (e.g. Kiehl and Trenberth, 1997; Trenberth and Stepaniak, 2003; Ruckstuhl et al., 2007; Trenberth et al., 2007), hydrologic cycle (e.g. Chahine, 1992; Serreze et al., 2006; Jones et al., 2007; Hanesiak et al., 2010) and climate change (e.g. Schneider et al., 1999, 2010; Held and Soden, 2000; Ramanathan and Inamdar, 2006; Rangwala et al., 2009). The rate of the Arctic climate change is two times larger than the global one due to greenhouse gas increase. The water vapour feedback loop is highlighted, as part of other feedback, as being responsible for the Arctic amplification (e.g. Winton, 2006; Francis and Hunter, 2007; Miller et al., 2007; Screen and Simmonds, 2010; Chen et al., 2011; Ghatak and Miller, 2013).
Water vapour measurements (total column and vertical profile information) using radiosondes have been available since the early 1940s and satellites since the 1980s primarily for meteorological purposes, while Global Positioning System (GPS) and more generally Global Navigation Satellite System (GNSS) measurements have been diverted from positioning to remote sensing of atmospheric water vapour since the 1990s (Bevis et al., 1992).
The total column of water vapour (TCWV), also called integrated water vapour
(IWV), is defined as the density of water vapour in an atmospheric column
over a unit area (
TCWV is characterised by large spatial and temporal variability. It affects the water cycle intensity and the atmospheric dynamics (Sherwood et al., 2010; Trenberth et al., 2005). Since 2010, the Global Climate Observing System (GCOS) declared the TCWV to be an essential climate variable and highlighted the importance of high-resolution long time series that could enable the detection of both local and global TCWV trends.
The available satellite remote sensing techniques to observe TCWV in microwave (MW), infrared (IR), near-infrared (NIR) and visible (VIS) spectral domains are promising, with a global coverage that enables climate studies, but with limited retrieval capability (e.g. only daytime, only clear skies, or over oceans only). Satellite observations are validated by ground-based techniques, traditionally radiosondes. However, radiosonde data sometimes suffer from systematic observational errors, as well as spatial and temporal inhomogeneity and instability (Gaffen, 1994; Wang, 2003), that could induce potentially regional biases if radiosondes alone are used to validate satellite data (Wang and Zhang, 2008, 2009; Bock and Nuret, 2009).
GNSS measurements complete the global radiosonde observations as another reliable reference to validate satellite water vapour retrievals and atmospherical models (e.g. Bock et al., 2007, and references therein). GNSS TCWV measurements are independent of the weather, performed with high temporal resolutions (a few minutes) and have continuously improved spatial resolution (from global down to a few kilometres for local networks). While GNSS is based on a delay measurement, it can be applied similarly to different sensors and is an ideal tool for long-term measurements, despite the presence of a possible bias in certain specific configurations (Ning et al., 2016).
Many studies comparing global satellite TCWV products with radiosonde, GPS and other reference data have pointed a dependence of bias and root mean square error (RMSE) on various observational factors like TCWV content (larger biases and RMSE are generally observed in regions with higher TCWV), reduced extreme values (e.g. wet bias at low TCWV values and dry bias at large values), solar zenith angle dependence (increased radiative transfer model error with larger zenith angles), day–night difference (increased background noise at daytime for VIS and NIR techniques), seasonal dependence (related to the two previous factors), latitude–geographical dependence (also partly connected with the former) and cloudiness dependence (usually increased biases and scatter with increasing cloudiness). Many of these aspects are discussed by (Vaquero-Martínez et al., 2017) for VIS, NIR and IR techniques over the Iberian Peninsula. Few studies investigated the polar and snow-covered regions. For example, Thomas et al. (2011) compared GPS to Moderate Resolution Imaging Spectroradiometer (MODIS) and Atmospheric Infrared Sounder (AIRS) over 13 Antarctic stations for 2004, and they found that GPS TCWV data are drier than MODIS, while wetter than AIRS. Palm et al. (2010) compared GPS with SCIAMACHY and GOME-2A data over Ny-Ålesund/Arctic and found GPS to underestimate both satellite sensors.
The current study provides intercomparisons of various measurements and methods allowing to quantify uncertainties, accuracies and limitations of several global satellite sensors/techniques available.
As common reference, we use a recently reprocessed version of GPS TCWV data
with hourly temporal sampling covering the period from 1996 to 2014. It
enables the largest number of coincident overpasses of three independent
selected satellites AIRS IR (from 2003 to 2014), MODIS NIR (from 2001 to
2014) and SCIAMACHY VIS (from 2003 to 2011) for intercomparisons. Three
Arctic ground-based GNSS observation sites were chosen: Ny-Ålesund
(78
Generally, satellites measurements are more accurate during clear sky conditions. In this work we use only cloud-cleared products in order to assess their uncertainties in optimal conditions in the Arctic region. However, cloud clearing is a challenging task. For this reason, we investigate the possible relation between satellite TCWV biases and the cloud cover at various timescales (seasonal and interannual, using time series with monthly, seasonal and annual sampling). In order to strengthen the conclusions, two different cloud fraction (CF) products are used (from MODIS and AIRS measurements). Though cloudiness dependence is not the only error source in satellite TCWV retrievals, it is one of the least well known, especially for the Arctic region. The impact of clouds on TCWV retrievals is to shield partly or totally, depending on the cloud opacity, the underlying atmosphere, so that the observed radiance is only a measure of the water vapour content above the cloud. The mixing of cloudy pixels with clear pixels tends to lower the TCWV estimate and lead to a dry bias. In contrast, depending on wavelengths, multiple scattering inside the clouds may increase the observed radiance and lead to overestimation of the water vapour content above the cloud. These effects are usually corrected in the retrieval algorithms using different methods depending on the instrument. However, in the end, both under- and overestimation of the retrieved TCWV can be observed.
Section 2 describes the datasets used and discussed the error sources specific to each technique. Section 3 presents results of TCWV comparisons (satellite retrievals compared to GNSS). Section 4 investigates the link between observed biases in the satellite data and cloudiness. Section 5 presents conclusions.
Originally designed for real-time navigation and positioning, GNSS was
rapidly seen as a cheap and accurate technique for measuring TCWV from the
ground (Bevis et al., 1992). The principle consists in
estimating the propagation delay induced by the atmosphere of the microwave
signals emitted by the GNSS satellites and received by ground-based
receivers. The zenith tropospheric delay (ZTD) is usually parsed into its
wet and hydrostatic components (ZWD and ZHD, respectively, for zenith wet
delay and zenith hydrostatic delay). Accurate estimations of surface
pressure and a weighted mean temperature are required to convert GNSS ZTD
into TCWV using the following formulas (Bevis et al.,
1992):
TCWV is converted from the ZWD as
In this study, we used GNSS ZTD data from the Geodetic Observatory Pecny
(Czech Republic) named “repro2 solution” and referred to as GO4
(Dousa et al., 2017). This ZTD dataset was produced with a homogeneous and optimised processing
of GPS observations. Outliers in the ZTD time series were detected and
removed using the range check and outlier check method described in
Bock et al. (2014). ZHD and
In order to overcome the satellite/GNSS timing error due to limited hours of
MODIS, AIRS and SCIAMACHY measurements during a month over a fixed point at the
surface, the satellites passing hours over the three Arctic GNSS stations
were defined through the IXION software (
Seasonal variations of the TCWV over all three sites for a common period of
11 years (2004–2014) exhibit a pronounced seasonal cycle (Fig. 1) with mean
values ranging from a maximum in July of 20, 14 and 13
Extreme hourly values could reach 40
Over passing hours of each sensor in universal time (UT) at three GNSS sites.
Annual cycle of TCWV from GNSS for the period 2004 to 2014 (in
Monthly time series of TCWV from GNSS over the full period of
observation at each site (in
MODIS is installed on
both platforms (Terra and Aqua) of the Earth Observing System (EOS). Both
satellites are launched on polar orbits since 1999 (Terra) and 2002 (Aqua).
They overpass the equator at 10:30 and 13:30, respectively. The
global coverage is provided within 1–2 days, through a nadir-looking
geometry at a solar zenith angle of 45
MODIS observes the NIR solar radiation reflected by sufficiently bright
surfaces and clouds and IR thermal emission in 36 channels covering the
spectral region 0.4–14.4
Five NIR channels are used for retrieving daytime water vapour. They are
centred on 0.865, 0.905, 0.936, 0.94 and 1.240
The TCWV data used in this study are from version 6 of the MODIS
instrument on board Terra platform, referenced as “Water vapour near
infrared – clear column (bright land and ocean sunglint only): mean of daily
mean” (Gao and Kaufman, 2003; Hubanks et al., 2008). The Aqua platform was not used because of many gaps in the
measurement. We retrieved global monthly mean files, gridded at 1 Dataset DOI:
We extracted TCWV from 2001 to 2014 for Sodankylä and Ny-Ålesund and from
2004 to 2014 for Thule for the comparison with GPS. The pixel selection
method is the following. MODIS data coordinates refer to the centre of each
gridded pixel, so a single pixel is considered per station (to avoid
interpolation and select the nearest pixel to GNSS/IGS stations) and defined
as follows:
For example, the Sodankylä MODIS pixel was selected as follows:
Launched on board the satellite ENVISAT-1 in March 2002, the Scanning Imaging
Absorption spectrometer for Atmospheric CHartographY (SCIAMACHY) was
designed to observe the earthshine radiance and the solar irradiance within
limb and nadir alternating viewing geometry. SCIAMACHY nadir and limb
observations cover the spectra from ultraviolet (UV) to NIR (214–2380
SCIAMACHY can measure water vapour at various wavelengths from the VIS to
the SWIR (shortwave infrared). This paper uses TCWV retrieved by the air-mass-corrected differential optical absorption spectroscopy method, shortly
AMC-DOAS (Noël et al., 2004), where water vapour is measured in nadir mode in the visible part
of the spectrum between 688 and 700
Though SCIAMACHY TCWV measurements are independent of the initial humidity
profile, they are affected by other factors. A dominant error
source in SCIAMACHY TCWV retrieval is caused by uncertainties of the
atmospheric radiative transfer, mainly due to effects of varying cloud cover
and surface albedo for different surfaces (Wagner et al., 2011) This error source is estimated to
be about 15
The three stations used in this study were part of the ground-based stations contributing to the SCIAMACHY validation effort (Piters et al., 2006) during which water vapour profiles alone were validated over Thule and Sodankylä, while TCWV was additionally validated over Ny-Ålesund.
TCWV data used in this paper are from
Noël et al. (2004), where all observations with AMF
This collocation is made by choosing data that meet the conditions
Then, SCIAMACHY TCWV monthly means are calculated from all the matched data to the given station. Note that SCIAMACHY data solar dependency results in missing data for winter months. Our study takes place from 2003 to 2011 over Sodankylä and Ny-Ålesund and from 2004 to 2011 for Thule.
The AIRS is carried on Aqua (EOS) since May 2002.
This platform has an equatorial over passing at 13:30 with a
sun-synchronous orbit. AIRS was dedicated to water cycle, energy and traces
gases observations. It provides twice daily global coverage with higher
vertical resolutions than all previous sensors and comparable accuracy to
radiosondes (Tobin et al., 2006). AIRS is a hyperspectral scanning infrared sounder. It
measures upwelling thermal radiation emitted from the atmosphere and the
surface. However, almost 30
Humidity profiles (level 2 products) are retrieved from cloud-cleared radiances (level 1). A set of different water vapour sensitive channels are used in addition to temperature sensitive channels. Water vapour mixing ratios at certain pressure levels are retrieved using the radiative transfer algorithm AIRS-RTA described by Strow et al. (2003). TCWV is obtained by integrating the vertical profile of water vapour mixing ratio.
The RMSE of the AIRS water vapour profiles is estimated to 10–15
Previous versions of AIRS TCWV were validated against radiosondes over
oceanic areas (Fetzer et al., 2006) and against reanalysis (ECMWF) (Susskind et al., 2006).
Gettelman et al. (2006) showed that AIRS retrievals in polar regions are unbiased relative to
in situ radiosondes. Most results indicate a small mean bias that does not
exceed 10
AIRS TCWV data
During this study, we additionally use the AIRS effective CF
(Kahn et al., 2014) monthly 1
Bias, RMSE and linear correlation coefficient between MODIS NIR,
SCIAMACHY VIS, AIRS IR clear column TCWV retrievals and GNSS TCWV estimates,
at Ny-Ålesund (78
Box plot of the TCWV differences (GNSS – MODIS) for 2001–2014 at
Sodankylä (67
MODIS time series of monthly means TCWV are compared to monthly means of
coincident overpassing (mentioned in Table 1) GNSS data over Sodankylä and
Ny-Ålesund for the period 2001–2014 and over Thule for 2004–2014. This
difference in the data range is linked to the GNSS data availability, as
GNSS dataset has some missing values at Thule during 2001–2003. The results
show an excellent overall agreement with a high coefficient of correlation
The mean biases and interannual variability of the individual months are analysed with box plots in Fig. 3. A seasonal variation can be seen at all three sites in the bias and in the dispersion (see the interquartile range in the box plots). The largest variations are observed at Sodankylä with large positive biases between September and February and slightly negative biases between July and August.
Dividing the year into four seasons, the statistics were also calculated and
given in Table 2. At Ny-Ålesund and Thule the relative bias does not exceed
13
At Sodankylä, the results are more complex to interpret. Multiple factors are
involved with the observed biases including clouds. During the snow season,
which lasts from October to April at Sodankylä, the solar angle has a strong
influence on the effective albedo, since Sodankylä is totally covered with
canopy, unlike both other stations, and its forests intercept the majority
of incoming solar radiation, as pointed out by Gryning et al. (2002). Additionally, Sodankylä
snow samples contain higher impurity concentrations (black carbon) than
measured elsewhere in Arctic Scandinavia or Greenland
(Doherty et al., 2010), as well as a bigger snow grain size. These two factors contribute to a
decrease in surface albedo
(Meinander et al., 2013). The chemical exchange between polluted atmospheric layers due to winter
biomass burning and snow surface opaque the lower part of the atmosphere at
the instrument's wavelengths. Since the MODIS retrieval capacities are
sensitive to surface albedo and atmospheric transmittance (Sect. 2.2), the
seasonal variation in these parameters and could explain the variation in
the MODIS TCWV bias, especially the dry bias during the snow season at
Sodankylä. During summer at Sodankylä, MODIS TCWV estimates were found
higher than GNSS TCWV measurements. This opposite bias can be explained by
the fact that the snow coverage nearly disappeared, in addition to the
tendency of increasing MODIS TCWV with increasing water vapour at sites
below 3000
Box plot of the difference (GNSS – SCIAMACHY) at Sodankylä
(67
Box plot of the difference (GNSS – AIRS) for 2003–2014 at
Sodankylä (67
Calculated monthly means of SCIAMACHY TCWV over Sodankylä and Ny-Ålesund for
2003–2011 and over Thule for 2004–2011 were compared to means of coincident
GNSS measurements. This comparison does not include winter pairs over Thule
and Ny-Ålesund because of missing SCIAMACHY measurements during polar
winter. Similarly to MODIS, SCIAMACHY underestimates TCWVs at all three
sites with mean absolute biases between 0.6 and 2.4
Consideration of surface albedo of complex surfaces could be also a challenge for the SCIAMACHY TCWV retrieval. The presence of snow with a nearby canopy (e.g. in Sodankylä) might result in a surface albedo significantly different from the prescribed surface albedo used in the AMC-DOAS method (e.g. 0.05 compared to 0.5), which would explain the winter biases (Noël, 2007b). Nevertheless, the DJF and SON absolute TCWV biases found here with SCIAMACHY are smaller than those found with MODIS. They are also smaller than those expected for SCIAMACHY in such conditions (Noël, 2007b). However, the JJA bias at Sodankylä is the most challenging and yet unexplained issue.
The AIRS TCWV monthly product shows excellent agreement with coincident GNSS
measurements at all stations. The overall correlation with GNSS is larger
than 98
Annual cycle of AIRS cloud fraction for 2004–2014; the error bars
show the standard deviation (
Correlation coefficients (%) between TCWV biases and coincident
cloud cover (AIRS) at Sodankylä (SODA) (67
MODIS and SCIAMACHY TCWV measurements are known to be sensitive to the presence of clouds, whereas the AIRS TCWV product is less impacted by clouds as it includes microwave water vapour measurements and a robust cloud clearing technique also based on microwave measurements (Susskind et al., 2003). This section uses the AIRS and MODIS CF products to examine the correlations between the TCWV biases found in Sect. 3 and cloud cover. The use of both products helps to minimise the influence of different overpasses between clouds fraction and satellites measurements. In this study, AIRS and MODIS cloud fractions show similar annual cycles only at Thule. This is not surprising, as previous comparisons between both cloud fractions showed the largest disagreement over the high latitudes (e.g. Wu et al., 2009). The observed inconsistencies in both cloud fractions are expected to be dominated by retrieval algorithm differences instead of differences in the observed radiances (Kahn et al., 2007). More significant differences between AIRS and MODIS retrievals can be found in areas of low clouds in the Arctic in summer (Weisz et al., 2007) as AIRS is less capable to detect the multiple layers summer clouds. However, AIRS is better suited to retrieve thin cirrus than MODIS (e.g. Kahn et al., 2007; Weisz et al., 2007) especially during the polar winter and at night (Kahn et al., 2005). Additionally, AIRS retrieval of cloud top pressure performs better than MODIS retrievals over polar regions, especially in presence of low-level temperature inversions (Weisz et al., 2007).
Figure 6 describes the annual cycle of CF at the three sites
based on monthly mean AIRS CF product for a common period of 11 years (2004–2014).
At Sodankylä, the 8-month period from May to December
shows a cloud cover above 50
Correlation coefficients (
Summer GNSS – MODIS TCWV differences (
Although this study uses only clear column water vapour observations, the
monthly time series of TCWV differences (GNSS–MODIS) show significant
correlations with the coincident AIRS (MODIS) CF at Thule and
Ny-Ålesund, with
The annual cycle of TCWV biases shows significant correlation with
coincident cloud fraction (both MODIS and AIRS) at Thule with
The interannual variability is generally more dominant at Ny-Ålesund out of
winter (
At Thule, the interannual variability is significant with both cloud
fractions on summer and additionally on winter and autumn when AIRS CF is
used (Table 3). At monthly scale, only 2 months are significant with AIRS
CF (November and December with
The high correlations between TCWV biases and cloud cover in JJA at both
sites could explain the poor agreement found in Sect. 3.1 (large biases
0.6 and 1.1
Regarding Sodankylä, TCWV differences show significant correlation with both
cloud fractions on November with
Summer GNSS – SCIAMACHY TCWV differences (
Winter GNSS – AIRS TCWV differences (
The monthly time series of SCIAMACHY TCWV biases are significantly
correlated at Thule with
At Ny-Ålesund and Sodankylä, monthly TCWV biases show similar sensitivity to
AIRS CF only, with
The correlations at annual scale at Thule and Ny-Ålesund behave again like
in Sect. 4.1. They increase at Thule (from
SCIAMACHY's TCWV retrieval is more sensitive to cloud cover than MODIS when AIRS CF is used, but MODIS retrieval shows more sensitivity to MODIS CF. Different sensitivity is observed to each of the used cloud fraction products, which is probably linked to closer SCIAMACHY overpasses with AIRS CF than with MODIS CF. The results at Sodankylä are thought to be more influenced by the diurnal variability and thereby the matched passing hours (CF and satellites). Similar sensitivities to both cloud fractions are marked in red in Table 4.
Generally, our results agree with the findings of Palm et al. (2010) who concluded that cloudy conditions introduce a severe bias at Ny-Ålesund, even if the SCIAMACHY measurement passes the cloud screening filter.
As found with MODIS (Sect. 4.1), TCWV biases and both cloud fractions are
strongly correlated at the interannual scale in JJA at Thule with
At Ny-Ålesund, TCWV biases are correlated with both cloud fractions in
August, while only with AIRS CF for the whole summer with
At Sodankylä, the interannual variability in TCWV biases and both cloud
fractions is significantly correlated in May with
The results with AIRS are quite different compared to SCIAMACHY and MODIS.
Whereas monthly time series of TCWV biases show significant positive
correlation with both cloud fractions at Thule (
Overall, Ny-Ålesund TCWV AIRS biases seasonality is almost linear with
negative slope with AIRS CF. Moreover, the interannual
variability of TCWV biases and AIRS cloud cover is significant at Ny-Ålesund
only for DJF (
As for MODIS (Sect. 4.1) and SCIAMACHY (Sect. 4.2), AIRS summer TCWV biases are also sensitive to MODIS CF at Thule (Table 4),
At Sodankylä, no significant correlations are found for the monthly means
and the annual cycle, but at interannual scale in March with AIRS CF (
Most correlations found are sparse temporally and do not show clear features. This might be due to the fact that AIRS TCWV biases are smaller in magnitude (Table 2) and show a different seasonality compared to MODIS and SCIAMACHY.
This paper found a general good agreement between satellite TCWV retrievals
and coincident measurements from three GPS instruments in the Arctic region.
MODIS and SCIAMACHY show overall mean dry biases compared to GPS with some
seasonal and latitudinal variation. We generally see better agreement
(higher correlation, smaller bias and RMSE) between GNSS and AIRS TCWV time
series than between GNSS and MODIS or SCIAMACHY. The seasonal
(3-monthly) biases do not exceed 1
For MODIS, the interannual agreement is getting better with latitude over all seasons except summer. During summer, the interannual variability is actually getting worse at higher-latitude sites. These increased summer biases are found to be sensitive to clouds cover. Additionally, MODIS dry biases during some snowy months at Sodankylä are also correlated with cloud fraction. However, the inaccurate estimation of the surface albedo over a complex mixed surface (snow and nearby canopies) also limits the MODIS retrieval capabilities at Sodankylä.
Summer SCIAMACHY–GNSS TCWV biases are found to be correlated with cloud cover at the higher-latitudes sites (Thule and Ny-Ålesund), in similar way as MODIS ones, but unlike AIRS. However, both MODIS and SCIAMACHY seem to be more sensitive to cloud fraction than AIRS as the annual cycle of TCWV bias for both satellites is well correlated with the annual variations of cloud fraction at Thule and Sodankylä. AIRS time series of TCWV differences to GNSS show a limited link with cloud fraction compared to MODIS and SCIAMACHY (except at Thule) with no clear features. Results reveal anticorrelated monthly differences with AIRS CF at Ny-Ålesund, probably due to opposite correlation with clouds in winter. Cloud presence is reported to affect satellite TCWV measurements more clearly at Thule compared to both other sites.
Overall, our results suggest a probable link between satellites TCWV biases to GNSS and cloud cover fraction, with contrasted regional and seasonal features. This sensitivity is strong to both AIRS and MODIS cloud fractions at Thule as both cloud fractions are more correlated at this station, and at all stations during summer. GNSS–AIRS biases are stronger correlated to the AIRS CF than to MODIS CF, whereas GNSS–MODIS biases are stronger correlated to MODIS CF. The use of two cloud fractions clears out a possible influence of the diurnal differences on studying the cloud impact. This effect is decreasing with latitude, as different sensitivity to both cloud fractions is mostly noticed at Sodankylä, which is thought to be linked to the diurnal variability.
We suggest that more robust information on clouds is included in the satellite data processing procedures in order to reduce the TCWV biases in the Arctic and then improve space-borne instrumental uncertainties. We suggest also using GNSS TCWV data in the calibration of satellite TCWV measurements.
All TCWV data are publicly available with the exception of the SCIAMACHY data set, which is obtainable by contacting Stephan Noel (data director, cited in the SCIAMACHY data section). All other data (which are cited with DOI and references in the article) are also available via Giovanni (NASA) online application. TCWV data and cloud cover at the three stations for the studied period are included in the Supplement.
The supplement related to this article is available online at:
The authors declare that they have no conflict of interest.
This article is part of the special issue “Advanced Global Navigation Satellite Systems tropospheric products for monitoring severe weather events and climate (GNSS4SWEC) (AMT/ACP/ANGEO inter-journal SI)”. It is not associated with a conference.
This work was developed in the framework
of the VEGA project and supported by the French program LEFE/INSU. This work
is a contribution to the European COST Action ES1206 GNSS4SWEC (GNSS for
Severe Weather and Climate monitoring;