Variations in the distribution of atmospheric water vapor are the most significant source of noise for large-scale interferometry analysis using C-band Sentinel-1 data. The propagated noise depending on partial pressure of dry air, temperature and partial pressure of water vapour, decreases the accuracy of interferometric synthetic aperture radar (InSAR) measurements in identifying and interpreting slow deformation signals. In this study, we evaluate the use of zenith total delay (ZTD) information derived from around 200 permanent global navigation satellite system (GNSS) stations with the separation between adjacent stations from around 13 km to 70 km, for correcting country-scale Sentinel-1 interferograms in Norway. The GNSS data was provided by the Norwegian mapping authorities.

Several stochastic and deterministic techniques including linear interpolation, machine learning techniques like linear regression, support vector machines (SVM), and Gaussian processes (GPs), canonical correlation analysis (CCA), as well as the Archimedean Copula-based analysis are investigated for downscaling of GNSS and InSAR data in both temporal and spatial resolution. First, we derive the dependence structure between GNSS and InSAR atmospheric components. Then, we simulate the high temporal and spatial atmospheric components based on the captured relationship between GNSS and InSAR time series. Next, the generated solutions are compared with the correction derived from ERA-Interim to assess the performance of each method in characterizing the atmospheric component in large-scale Sentinel-1 interferograms.

We utilize 14 SLC images of Sentinel-1 acquired in interferometric wide (IW) swath mode covering May–November 2016 from a descending track. Mainland Norway is fully covered by 27 frames of Sentinel-1 images acquired in a descending orbit along five tracks. We used Gamma software to consecutive frames of the SAR images along each acquisition track and for the interferometric proceeding. A network of small baselines interferograms have been designed and the created interferograms are then corrected based on the downscaled ZTD data. The results are presented and the derived ground surface displacement time-series are compared with the GNSS displacement time-series for validation.

GNOS (GNSS Occultation Sounder) is a new radio occultation (RO) payload onboard the Chinese FY-3 series satellites, which probes the Earth’s neutral atmosphere and the ionosphere. GNOS is capable of tracking the signals of both the Beidou (the Chinese navigation satellite system) and the GPS navigation satellite systems. The first FY-3C satellite with GNOS launch on 23 September 2013, and has more than 500 RO events daily, including approximately 400 GPS and 100 Beidou RO events. The second FY-3 polar meteorological satellite, FY-3D, also launch on 15 November 2017 successfully. Due to the improvement of payload capability, FY-3D satellite can receive more RO data than FY-3C. The FY-3C and FY-3D are separately belonged to the morning and afternoon orbit satellite, therefore the RO data can cover 85% the earth in the six-hour assimilation window. In this paper, the data quality of GNOS, including GPS and Beidou RO, will be assessed. The impact experiments of assimilating GNOS radio accultation in GRAPES prediction system of China Meteorological Administration, will also be presented. The combined impact of morning and afternoon orbit RO will be studied. The impact of Beidou RO will be evaluated firstly. GNOS of FY-3C and FY-3D receive the signals of the second generation Beidou satellite, which has 14 operational satellites in orbit now, including five geostationary satellites, five inclined geosynchronous satellites and four middle earth orbit satellites. The orbits of Beidou navigation satellite differ from those of the GPS, GLONASS and Galileo navigation satellite systems, resulting in different geographical distribution of RO events.

The distinct capability of satellite altimeters to observe the open ocean repeatedly and globally with centimetre-level accuracy has enhanced our knowledge both of the absolute sea level and of the ocean state and space-time variability, therefore revolutionizing various sciences, among which are oceanography and geodesy. Extending over a period of more than two and a half decades, the satellite altimetry dataset holds potential for understanding the influence of the ocean on the global climate and its changes, providing that stable and homogeneous products, in terms of the applied corrections, are available. Measurements with such accuracy are currently available since precise orbits are accessible and the accurate modelling of all instrument, range and geophysical effects can be achieved over open-ocean with the same accuracy.
However, altimeters are designed to observe the open ocean and major problems arise in coastal regions, due to several factors, namely the large size of the altimeter footprint and even larger microwave radiometer’s one, that sense land, simultaneously with water, that possesses a completely different surface emissivity and backscatter. Additional problems arise from the inaccurate modelling of the tidal and atmospheric forcing and of the tropospheric content of water vapour. Even considering that improved tracking techniques of the radar echoes are used, the inexistence or invalidity of the corrections in the coastal regions lead to the rejection of the measurements in these regions that remain poorly sampled, thus preventing a successful management of the coastal zone with satellite altimetry.
Amid the mentioned corrections, the wet tropospheric correction (WTC) that accounts for the wet path delay in the altimeter observations is still considered one of the major sources of uncertainty due to high spatial and temporal variability of the water vapour content in coastal regions. Over open-ocean, the measurements acquired by the microwave radiometer (MWR), collocated in time and space with those of the altimeter, lead to the most precise WTC correction; over the ~30-km narrow coastal strip, these measurements generally become invalid, leading to the rejection of the altimeter data or obliging the use of a WTC computed from numerical weather models, which does not correctly account for the variability of the wet path delay.
To overcome this obstacle in the coastal regions and compute an improved global WTC, generating simultaneously a long-term stable dataset for use in climate change studies, the University of Porto developed the GPD+ (GNSS-derived Path Delay Plus) algorithm. For each altimeter mission, the GPD+ combines, using objective analysis, valid wet path delays (WPD) observations i) from the microwave radiometer (MWR) on board that mission, whenever they exist, ii) derived from GNSS data acquired at coastal and island stations, and iii) from scanning imaging radiometers (SI-MWR) on board various remote sensing satellites, to estimate a new WTC for all invalid MWR measurements. To ensure the long-term stability of the WTC datasets, the measurements from the various radiometers have been inter-calibrated prior to their use in the algorithm adopting the Special Sensor Microwave Imager (SSM/I) and Special Sensor Microwave Imager/Sounder (SSM/IS) on-board the Defense Meteorological Satellite Program (DMSP) satellite series as reference.
GPD+ products have been computed for all altimetry missions and validated against other available WTC. Results have shown that the GPD-derived WTC is the most effective in reducing the sea level anomaly variance in the coastal regions, particularly for the ESA missions. Since GPD+ deals with not only land contamination but also rain and ice contamination, and instrument malfunctioning, the GPD+ WTC is also an added-value for open-ocean studies. Moreover, the correction becomes of great importance for the missions that do not possess an on-board MWR, such as CryoSat-2, being computed for this mission operationally and made available in CryoSAt-2 level 2 products. The GPD+ WTC has been selected for use in the generation of products from the ESA Climate Change Initiative Sea Level (SL-cci) project, being currently made publicly available by AVISO (Archiving, Validation and Interpretation of Satellite Oceanographic data), through the Center for Topographic studies of the Ocean and Hydrosphere (CTOH) webpage, in addition to the University of Porto webpage.
This study shall review the main features of the GPD+ algorithm and its impact, particularly in coastal studies and those at high latitudes. The GPD+ methodology highlights the advantages of combining different data sets to serve a given application.
The potential contribution of wet path delay measurements computed using data synergy from different sources (e.g., infrared radiometers, INSAR, etc.), allowing the characterisation of the field at small spatial scales, as well as from future missions (e.g. Sentinel-5) shall be discussed.

Atmospheric Phase Screens (APS) are one of the components in SAR interferograms, together with topography (hence, DEM), and deformations. In the case of small-baseline missions like Sentinel-1, they are those mostly contributing to the phase. However, up to now, they have been mostly regarded as a noise field to get rid of.

Recently, it has been shown that the fine space resolution and the sensitivity of the APS to water vapor leads to products with great potential when assimilated into weather models. Their value could be much increased if APS could be generated at fine temporal resolution, like from geostationary SARs.
Up to now, APS are generated as by-products by Advanced DInSAR approaches, such as PSInsar, SqueeSAR, SBAS and many others. These methods are designed to estimate deformation, therefore APS are suppressed as much as possible, as noise, and recovered back at a later stage, after residual topography and subsidence have been estimated. We can argue that the procedure is not optimal in terms of quality nor efficiency.

In the proposed approach, a novel algorithm is designed, to retrieve APS from both distributed and Persistent Scatterers. The method exploits Phase Linking to provide a high-quality estimation of the phases by exploiting ground patches whose size compares with the desired spatial accuracy in APS. The method is quite suited to short revisit, C-band systems like Sentinel-1, where sufficient coherence is achieved when estimating phases over large areas, compared with the size of the APS correlation. The dense ground coverage of the achieved estimate allows for direct phase unwrapping, to provide phase screens that are processed, at a later step, to remove residual topography and deformations. In particular, as the method does not apply any space or time derivative, it does not suppress the low-frequency components of the phase field, that are fundamental for meteorological applications.

Notice that few images need to be processed to take advantage of distributed targets and, at the same time, minimize the impact of deformations. This, together with an efficient implementation of phase linking, makes the method quite fast and robust. The method has been developed in the frame of the H2020 project TWIGA, aimed to transform weather and water data into value-added information services, with a focus on sub-Saharan Africa.

Results achieved in different test cases are shown and compared with other methods.

An example of a full 250 x 250 km three swath APS is in the figure. A single stack of 10 images can be processed in less than one hour, starting from the coregistered, DEM compensated RAW data.

The processing of SAR images to generate maps of Precipitable Water Vapor (PWV) is becoming a useful tool in the community of atmosphere scientist to study atmospheric physical processes. Of particular interest is the assimilation of PWV maps in high-resolution Numerical Weather Prediction (NWP) models. It has been demonstrated that information content on water vapor distribution of interferometric SAR images, can help to better model phenomena as deep convections with an impact on the correct forecast of extreme weather events as intense rainfalls. This has a crucial importance also on Civil Protection practices as it is tightly related to the flood and other hydrogeological risks. The most used data are those of the ESA Sentinel-1 SAR mission as those images cover large areas and have a revisiting time of up to six days. The assimilation of PWV maps generated by Sentinel-1 data in NWP models (e.g. the Weather Research and Forecasting model) showed an enhancement of numerical forecasts also when compared to the assimilation of PWV measurements obtained by dense GNSS networks. The enhancements were assessed in different ways, e.g. by comparing the PWV estimated numerically by the NWP model with GNSS measurements or the cumulated rainfall provided by the model with in-situ gauge measurements. The assessment was conducted both at the time of assimilation, and in a time window after the assimilation, to study the persistence of information on water vapor introduced by the assimilation of SAR maps. However, it was observed that both the impact of SAR information on the output of NWP models changed depending on the specific atmospheric phenomenon and the NWP initial state. Also, the “memory effect” of the assimilated PWV maps was changing from a few hours up to more than six hours.
The aim of this work is to present a method to know in advance the impact that the assimilation of PWV maps obtained by a time series of interferometric SAR images would have on the output of an NWP.

ACKNOWLEDGMENTS
Fundação para a Ciência e Tecnologia (FCT) – project UID/GEO/50019/2013 and SFRH/BPD/96069/2013.

High resolution Numerical Weather Prediction (NWP) models are presently able to produce forecasts with a spatial resolution in the order of 1 km (the so-called cloud resolving grid spacing). However, poor knowledge of the initial state of the atmosphere at small scales may imply an inaccurate simulation of the weather phenomena in terms of timing, location and intensity. Hence, to fully exploit the advances in numerical weather modelling, it is necessary to feed them with high spatio-temporal resolution information over the atmospheric column.
In this context, Sentinel-1 data can be very useful to characterize the atmosphere with unprecedented spatial resolution by applying the SAR interferometry (InSAR) technique to derive the columnar water vapor from Atmospheric Phase Screen (APS), i.e., maps of the atmospheric delay at the time of the SAR observation relative to the atmospheric delay at the acquisition time of a master image. The possible availability of a spatially dense network of Global Navigation Satellite Systems (GNSS) stations could also be exploited to estimate the columnar water vapour. In fact, GNSS data can be used to estimate time series of Zenith Total Delays (ZTDs), representing the delay induced by the troposphere on GNSS signals in the zenith direction above a GNSS receiver. This paper presents the first experimental results of a study on the assimilation, in the Weather Research and Forecasting (WRF) model, of atmospheric water vapour data obtained from GNSS measurements and Sentinel-1 derived APS maps.
In agreement with the previous considerations, the STEAM (SaTellite Earth observation for Atmospheric Modelling) project, funded by the European Space Agency, aims at investigating new areas of synergy between numerical weather models and data from spaceborne remote sensing sensors. An example of synergy is just the assimilation of EO derived data on the precipitable water vapour. This work presents some experimental results of a study, carried out in the framework of the STEAM project, on the assimilation, in the Weather Research and Forecasting (WRF) model, of atmospheric water vapour data obtained from GNSS measurements and APS derived maps. The STEAM project focuses on high impact weather events (HIWEs), which are the most interesting from an applicative point of view. Note that it is still a challenge to accurately forecast the onset and/or the spatio-temporal evolution of a HIWE, especially in complex topography areas (e.g. Mediterranean Region). The experimental results concern two flood events. The first one occurred in Tuscany (Central Italy) in September 2017 and caused nine casualties. Actually for this event only GNSS data were used because, at the time of the Sentinel-1 acquisition closest in time to the onset of the event, few water vapour was present in the atmosphere. The second event considered for the assimilation of EO derived columnar water vapour maps occurred in Abruzzo (Central Italy) in November 2017. In this case, both GNSS maps and APS derived maps were assimilated in two separate WRF model runs.
The results of the experiments demonstrate that the assimilation of EO derived water vapour data, has a positive impact of the forecast accuracy of the WRF model.