2.1.1. Inland Waters

2.1.2. Sea Ice

</section

The Sea Ice FRM products are derived from two sources: the primary activities carried out at the identified sea ice
opportunity sites through dedicated campaigns, and contributions from third-party activities.

2.1.2.1. Primary activities

The following table provides an overview of the data stemming from the primary activities.

2.1.2.2. Third-party activities

Once specific third-party activities and data have been identified, an overview will be provided following the same
structure as the primary activities.

This section provides an overview of the Land Ice FRM products. The Land Ice approach involves providing FRMs from
two opportunity sites:

• Antarctica – Adélie Land: Two fixed stations will be deployed (targeted for end of 2025, early 2026) to record time series of FRM measurands in the inland-coastal transition zone of the ice sheet, using custom-designed measurement stations near Sentinel-3
  ascending-descending crossover tracks.
• Svalbard – Austfonna: FRM measurements will be acquired along selected Sentinel-3 ground tracks transects
  during three dedicated campaigns with snowmobile-based surface GNSS profiling.

2.1.3.1. Antarctica – Adélie Land

The site in East Antarctica will provide daily time series of surface elevation measurements at the footprint scale of two
Sentinel-3 cross-over points through the deployment of fixed stations.
The custom-designed stations will monitor the surface elevation with several snow rangers which measure the distance
between the station frame and the surface, along with a GNSS receiver that measures the absolute position of the
frame.
To ensure the FRM measurements are directly comparable with the satellite data, in terms of scale and accuracy – we
plan to map the topography surrounding the stations using a UAV. This will be accomplished through two dedicated
campaigns.

2.1.3.2. Svalbard

The campaign-wise product from Svalbard contains precise snow surface elevations from Austfonna ice cap and its
vicinity based on kinematic GNSS surveys. Since Sentinel-3 tracks converge in this area (~80 degrees north), there will
always be numerous crossovers with in-situ GNSS transects, even if the latter is sometimes collected for other purposes
(i.e., glacier mass balance) or while commuting between the field camp and the ice cap. All collected kinematic GNSS
data (Figure 2) are therefore included in the data product, but with flags that indicate surface type (glacier, land, lake,
sea ice) and whether or not collected under a selected Sentinel-3 ground track (yes/no and S3 orbit/cycle number).

The GNSS raw data are provided as open-format RINEX ascii-files (level 0), including antenna height over the snow
surface, and processed results (Precise Point Positioning, PPP) are provided as a level 2 product including the mentioned
flags and other parameters for traceable parameters for data quality and uncertainty. The reference level for the
product is always the snow surface, consistent with the measurement target of Sentinel-3.

This section outlines the hierarchical structure and organization of data products used in generating the FRMs.
Data is categorized into levels – L0, L1, L2, and beyond – each representing progressively processed stages, from raw
measurements to higher-level, derived products. This structure, known as the Product Tree, provides a clear framework
for managing and understanding the progression of data from its initial acquisition through to FRM. By detailing each
level and its associated data attributes, we establish a view of how raw data evolves into valuable, standardized
measurements for each thematic surface.

2.2.1.1. Inland Waters

The St3TART-FO Inland Waters team collects water surface height measurements through both the micro-stations and
drone flights to produce FRMs. The level of the products is defined in the following table.

Uncertainties will be provided from L2a products. Those uncertainties will be computed from uncertainties of lower level products.

2.2.1.2. Sea Ice

The St3TART-FO Sea Ice team collects and processes data from several different sensors/instruments either as
permanent, semi-permanent or campaign-based data collections, organized in different ways reflecting the different
spatial, temporal, and parameter groupings. Below, these groupings are detailed.

In the L3/FRM processing baseline of the sea ice reference observations, we do not differentiate between L1 and L2 but
include both taking into account the information such as quality flags. We note that the reference data used to generate
L3/FRM is provided (or collected as third-party) data at L1 and L2., and we will only provide L3 FRM processed data on
the FRM Hub. Discussions are still ongoing on how to include data collected during the St3TART/St3TART-FO project, in
their initial format, for users to utilize them even when not processed to a FRM format.

2.2.1.3. Land Ice

2.2.1.3.1 Overview description of the product levels

2.2.1.3.2 Antarctica – Adélie Land

Different levels of products are defined to organize the data collected.
L0 products provide the raw data recorded by fixed station (Figure 1).
The L0 products are divided into:

• The list of parameters characterizing the station configuration at the beginning of each field campaign
  (L0_parameters);
• The sensor offsets from the GNSS antenna describing the position of each sensor at the beginning of each
  field campaign (L0_position);
• Data needed to define the position of the structures and the meteorological conditions (i.e., inclinometers
  and meteorological sensors) and Dataset from the sonic rangers (L0_snowvue_meteo);
• Data from the micro-laserscan (L0_SDMS40_range);
• Pitch and roll data of the GNSS antenna (L0_GNSS_inclinometer);
• The raw data from the GNSS i.e., the raw Trimble T02 files, every other day (L0_GNSS_station).

The L0 products also contain the information acquired to retrieve the Digital Surface Model (DSM) within a ~2 km radius
around the fixed station. These concern:

• The images obtained with UAV camera and the location of the calibrating Ground Control Points (GCP) used to
  perform the Structure-from-Motion (SfM) photogrammetry (L0_UAV);
• If UAV flights appear to be impossible to perform in the field due to bad weather conditions or implementation
  conditions that are too complex1, or a poor image contrast quality (pristine snow), the backup DSM would be
  obtained using a topographic wheel and differential GNSS positioning. The L0_GNSS_wheel product will
  provide the Trimble T02 files recorded during the surveys.

The L1 products contain:

• The calculation of the mean and median variations of the snow range between each snowvue and the surface
  at the fixed station using the data from the sonic rangers and from the SDMS40 (L1_snow_range).
• All the 36 estimations of the distance between the SDMS40 and the surface (L1_SDMS40_range).
• The processed position of the central mast computed with the PPP protocols (L1a_GNSS_gLAB_raw) (expected:
  2 daily surveys of 3-h continuous positioning of the GNSS). L1b_GNSS_gLAB_converged product will provide
  the filtered data. L1c_GNSS_resampled product will provide resampled
  data over 30 days, i.e. also when data transmission is not possible; in that case a computed position will be
  given assuming a constant surface displacement rate of the station.
• The DSM records using UAV in the area of the station (L1b_UAV).
• If UAV flights appear to be impossible to perform in the field due to bad weather conditions or implementation
  conditions that are too complex2, or a poor image contrast quality (pristine snow), the backup DSM would be
  obtained using a topographic wheel and differential GNSS positioning. The L1b_GNSS_wheel product will
  provide this record.

The Land Ice L2 product L2a_ snow_surface_elevation contains the snow surface positions at each surveyed location of
the station, whereas other L2 products (L2a_DSM, L2b_reanalysed_Shifted_DSM) have been designed to allow
comparing FRM data with S3 elevation at POCA locations at every satellite overpass. Based on the L2a_DSM, the
L2b_surface_roughness will provide useful information for researcher interested in the surface state within 2 km around
the fixed stations.

The Land Ice L3 products (L3_gridded_gap_filled_DSM and L3_reanalysed_gridded_gap_filled_DSM) will merge
St3TART-FO DSM information with existing high-resolution DSM of Antarctica to provide data to advanced altimetric
models requiring the DSM within the whole footprint.

Detailed product levels:

L0 products related to the fixed station provide the raw recorded data. This concerns L0_parameters, L0_position,
L0_snowvue_meteo. The later product will only be provided if our tests on the SDMS40 indicate that the sensor is
relevant for use in Antarctica. These data will be used to directly compute the hourly changes in the range between
each sensor and the surface at the fixed station (L1_snow_range) assuming changes in mast tilt angles and potential
data corrections.
The L0_SDMS40_range provides the raw data and is used to compute the L1_SDMS40_range product providing the
daily range of the sensor computed with beam corrected from the sensor angle.
Assuming an efficient satellite data transmission system, the Rinex files (L0_GNSS_station product) will be issued every
other day with a timeliness of one month in order to update the central station position. L0_GNSS_inclinometer will
provide the tilt and roll of the GNSS mast every day. Data will then be post-processed using the PPP positioning
protocols. The entire time dependent positions set of the central mast will be provided every other day in the
L1a_GNSS_gLAB_raw product (expected: 2 daily surveys of 3-h continuous positioning of the GNSS). From these time
series we restrain to the last 5 minutes
(L1b_GNSS_gLAB_converged). If data transmission is not possible, these data will not be available, instead, a computed
position will be interpolated assuming a constant trend obtained from a running average over the past 30 days, and
provided in the L1c_GNSS_resampled product. This gap filled product will be provided with a flag informing on the
method used to retrieve the position. In case of long or total communication failure, the GNSS data final product will
be a posteriori reanalysed from the entire GNSS data set downloaded every summer during field work so as to produce
accurate updated positions matching every S3 overpass. All the products will be updated a posteriori as well as the
higher-level products.
Using the daily changes in surface snow elevation of each snowvue (and SDMS40) sensor at the fixed station
(L1_snow_range) the position of the GNSS (L1c_GNSS_resampled) and the processed position of each snowvue sensor
(L0_position), we will compute the geolocation of each snowvue sensor, including surface elevation of each sensor.
These values will be provided in the L2a_snow_surface_elevation product.

During the first 2 summer field campaigns, we will collect data to calculate a new accurate digital surface model (DSM).
The raw data for calculating the DSM will be, in the best case, a series of images and the differential GNSS positioning
of Ground Control Points (GCP) in relation to a reference (here: the position of the station’s central mast). The L0_UAV
product will provide this information, acquired yearly (for the 2 first years), and will be released with a timeliness of
three months, because data will not be easily transferred directly from Antarctica. The position of the surface (DSM)
computed using the SfM photogrammetry performed from the L0_UAV will be provided in the L1b_UAV product with
a timeliness of 6 months to allow data treatment.
In case of UAV flights would not be possible or poor image contrast quality (due to pristine snow), we propose to retrieve
the DSM using surveys with a topographic wheel with a differential GNSS positioning. The L0 product (L0_GNSS_wheel)
will thus consist of the synchronous Rinex files from the topographic wheel and from our fixed GNSS station. Indeed, an
accurate differential GNSS positioning with a Post Processing Kinematic (PPK) analysis, requires both RINEX registrations
of the fixed base station and the wheel during the survey. The PPK analysis will provide precise positioning of the points
on the surface covered by the wheel (provided in the L1b_GNSS_Wheel product), relative to the position of the base
(fully georeferenced by Precise Point Positioning (PPP) analysis). The L1b_GNSS_wheel product will also be provided
yearly, but the annual raw data will be provided with a timeliness of six months.
Processed DSM of the 2km x 2km area around the fixed station will be obtained from field measurements and computed
yearly. This DSM will be by default computed using the SfM photogrammetry performed from the L0_UAV, but if UAV
flights are not possible, a lower resolution and smaller DSM will be computed by kriging of the positions obtained with
the topographic wheel (L0_GNSS_wheel and PPK analysis) and previous DSM or the REMA DSM as a backup. The
product, L2a_DSM, will be delivered with a timeliness of 6 months, because SfM involves manual operation. This
product will provide a flag informing on the method used to retrieve the DSM (use of data retrieved with the UAV or
with the topographic wheel).
Using the L2a_DSM, we will produce the L2b_surface_roughness, which will provide useful information for researchers
interested in the surface state within 2 km around the fixed stations. Data will be provided with a timeliness of six
months after the field surveys.
Using the L2a_DSM and the daily changes in surface snow elevation at the fixed station (L2a_snow_surface_elevation),
we will process into a shifted full-resolution DSM for the dates corresponding with S3A/B satellites overpasses. The
resulting full resolution DSM will be computed for every S3 overpasses, and provided yearly in the
L2b_reanalysed_shifted_DSM product, with a timeliness of 6 months after data acquisition in the field.
We will regrid the DSM provided in the L2b_reanalysed_shifted_DSM product, to a reduced resolution DSM, compatible
with the grid form of the Reference Elevation Model of Antarctica (REMA). This product will allow quick comparisons
with the REMA, assuming changes due to snow surface elevation changes at different dates and the displacement of
the fixed station due to the ice motion. The reduced resolution shifted DSM will be provided for every S3 overpasses, in
the product L3_reanalysed_gridded_gap_filled_DSM, with a timeliness of 6 months after field data acquisition.

Since the computed L3_reanalysed_gridded_gap_filled_DSM product will be provided only once per year, a monthly
product will be delivered with a one-month timeliness (product called: L3_gridded_gap_filled_DSM), providing a less
accurate DSM. This DSM will be based on the L2a_DSM product from the previous year, and the
L2a_snow_surface_elevation product. Here, we will mention whether the product was produced using processed GNSS
data using PPP protocols (if data transmission is possible), or using the approximated positioning of the fixed station
from the one measured during previous summer and assuming a constant ice movement (using the flag_GNSS variable).
This DSM corresponds to a less accurate version provided a few months later in the
L3_renealysed_gridded_gap_filled_DSM product after a new field campaign. Indeed, during this latter field campaign,
we will be able to reanalyze the 2km x 2km surface elevation changes in view of new processed data from the GNSS and
information from the UAV or the topographic wheel. The difference between the
L3_reanalysed_gridded_gap_filled_DSM and the L3_gridded_gap_filled_DSM products will be very similar to the one
found between operational analysis and reanalysis data in climate science.

2.2.1.3.1 Svalbard

The L0 product contains campaign-wise daily raw data recorded by the GNSS receiver, in the open RINEX ascii-format.
The manually measured antenna height over the snow surface will be included in the header as an antenna offset,
so that these files can be used directly in any processing without further need for other parameters or metadata from the survey.
The reference position is hence consistently at the snow surface even for the lowest level product.

Since all processing is done in one step with GNSS Precise-Point-Positioning (PPP) techniques, there is no need for an
intermediate L1 product. The results of the daily PPP processing form a campaign-wise merged L2 product with surface
height at given locations (longitude-latitude WGS84 and projected easting-northing in UTM 33) with a native 1 second
frequency. Data from breaks between surveys are excluded (i.e., receiver is kept continuously on), but otherwise all
collected kinematic data are included, regardless if they were specifically collected for Sentinel-3 comparison or for
other purposes such as mass balance measurements. The L2 product also contains various quality flags and traceable
uncertainty estimates. A surface flag will indicate if data were collected over glacier, land, lake or sea ice surface.
Auxiliary elevation data will be included as extracted at each location from ArcticDEM and the national Svalbard
reference DEM from the Norwegian Polar Institute.

No L3 product is planned, but if evaluations of the first campaign indicate a need for dense survey in a small area rather
than longer transects along POCA tracks, then an interpolated digital elevation model (DEM) from the GNSS data will be
considered as for the Antarctic site where it is needed to link fixed station data with variable Sentinel-3 ground tracks.

2.2.2.1. Inland Waters

On the St3TART-FO FRM Data Hub, Level 2 (L2) products derived from the collection and processing of inland waters
data are available. These products, generated by the St3TART-FO Inland Waters team, provide ready-to-use water
surface height measurements. Intermediate products (L0 to L1), which are internal steps in the computation process, are
not distributed. The Inland Waters product tree of the FRMs is outlined below:

2.2.2.2. Sea Ice

In the sea ice theme, we both collect and process data within the consortium and collect relevant datasets from third-party
and AO activities, all of which are expected to be provided at a L1/l2 product format for further L3/FRM processing.
Thus, the product tree will show in detail the L0 to L2 processing done at providers within the St3TART-FO consortium,
as well as the L2 product provided (or collected from third-party/AO activities) for further processing. L3 products will
be accessible via the FRM Data Hub.

2.2.2.3. Land ice

The product tree of the Land Ice surface products from the measurement stations Antarctica is presented in Figure 5.
The simpler GNSS product from survey campaigns on Svalbard follow a similar structure as for the station data,
with the relevant product specifications being: L0_GNSS_kinamatic (daily RINEX files) -> L2_snow_surface_elevation
(campaign-wise NetCDF files with time, positions, flags, uncertainties and standardized global attributes). For the fixed
station in Antarctica, L3 products will be available on the FRM Data Hub; and for Svalbard campaigns, L2 products
are available.

This section focuses on the contents of the FRM data at higher levels—primarily L2 or L3, depending on the thematic
surface. These processed levels of data are accessible through the FRM Data Hub. While the main emphasis is
on FRM data available to users, lower levels (such as L0 and L1) are also described in “L0 and L1 data content” for those
interested in the foundational raw data and intermediate processing steps. Selected lower-level data will also be made
available for users through a link on the Data Hub, though the specific data to be included is still under discussion.

The FRMs produced by the St3TART-FO Inland Waters team have a consistent content across all complexity levels.
The following table outlines the specific content of each FRM:

The L2 variables used in the project are presented below, from which the L3 FRM products are generated. L0 and
L1/L1b products are also presented in tables and can be found in “L0 and L1 data content” The output L3 products (FRM
data) are also presented below once the final structure and measurands have been defined. Currently, only the data
content of primary activities (not third-party or AO, as no third-party data or AO activities have yet been identified) is
listed below.

2.3.2.1. L2 products

Below, the expected L2 data content of the AWI IceBird, LOCEAN Ice-T buoy are provided, and the NORCE SnowDrone
(snow depth from snow-radar). We are still awaiting the final data content files (L2) of the LOCEAN mooring and the
Vortex/LEGOS L2 processing of total freeboard.

AWI IceBird

There are six potential quality flags, which provides information about the following: (1) whether acquiring over level
or deformed ice; (2) whether the total thickness is less than the total thickness + uncertainty + snow depth + snow
freeboard; (3) whether the maximum surface temperature (measured with another instrument) was above –5°C; (4)
whether the true sea ice bulk density (computed from a combination of the three main measurands) are above
100 kg.m-3; (5) whether the sea ice freeboard is less than 0 m; and (6) a flag related to the snow estimates, although
documentation does not state what this concerns. It is not certain whether all these flags will be used as inputs in the
final FRM processing or not.

This example is based on the 2019 campaign (only including variables of relevance), and we expect similar data content
for campaigns to be conducted. We await any updates regarding the processing chain until the AWI PhD student starts
in 2025.

LOCEAN Ice-T buoy

NORCE SnowDrone snow-radar

2.3.2.2. L3 FRMs

To be completed once agreed upon within the sea ice thematic team depending on measurands and differs depending
on the data product to be provided. Beyond the standard naming and global attributes to be included, the following
inputs are expected regardless of product (with the exception of xc and yc, which only apply to gridded products):

The measurands to be included can be one (or more) of the following with their associated uncertainties estimated
following the metrological approach and FRM protocols and procedures:

For products where there are 2 measurands or more available, all these variables will be included. If the products only
have one measurand available (e.g., moorings only have ice_draft), then only that measurand will be implemented. The
measurands along with uncertainties derived from conversion (using the conversion library in Module II) shall be named with“{variable_name}_derived”,
to indicate that those variables are not measured quantities, but rather computed from the conversion library.

2.3.3.1. Antarctica – Adélie Land

The L2 variables used in the project are presented below. L0 and L1 products are also presented in tables according to
the sensor used to measure them. These tables can be found in “L0 and L1 data content”.

Processed surface elevation at each sensor

Digital Surface Models

Variables describing surface roughness

2.3.3.2. Svalbard

NPI kinematic GNSS campaigns, Svalbard 

Instrumental validation
For each in-situ sensor used by the Inland Waters team, multiple validation of the data has been performed during the
development of the sensor in vorteX-io.
For the vorteX-io micro-stations, this work has been conducted during the development of the V1 micro-stations
regarding to the LiDAR performances and is still ongoing regarding the altimetric reference computation. LiDAR
measurement has been validated during multiple tests with third parties’ sensors from Vigicrues and other national
networks. We also performed a validation of the LiDAR during the basin tests in Marseille in the framework of the first
St3TART Project. As the LiDAR is the same in V1 and V2.1 micro-stations, there is no need for further tests on it.

A GNSS chip has been embedded in the V2.1 micro-stations to improve the accuracy of the GNSS positioning of the
micro-stations. The tests on the new system and its automatic processing chain to assess the accuracy of the new
processing based on the data of this GNSS chip are presented in FRM Super Sites Technical Description Document
(TD04) [RD14]. These tests are conducted by comparing the results of the positioning of the new system with the results
of PPP processing from the base we used for the V1 positioning and for drone flights. The V2.1 micro-stations are
installed following a new protocol that uses a generic boom arm. With this new boom arm, the distance between the
reference of the LiDAR and the GNSS antenna is standardized. For the V1 installation, it used to be the rope technicians
who measured manually the distance between the micro station and the antenna. With the new system, the accuracy
of the measurement of this distance is increased and human errors are limited.

As for the drone altimeter, we are developing a new version to improve the operationality of the altimeter and improve
the quality of the measurements with a new gyrostabilized pod. During this development phase, we have planned 3
tests days to validate all the measurements of the sensors. The first test has been conducted during summer 2024 and
the first results are excellent with a better stabilization (around 4° STD for pitch and roll compared to around 10° before.)
The two remaining days took place end of January 2025. During these tests, we compared the drone
altimeter measurements with a reference measured by a GNSS base and the Tide graph of Socoa in South of France
maintained by the SHOM (Service Hydrographique et Oceanographique de la Marine (French Navy Hydrographic and
Oceanographic Service). Cross comparison with micro-station will also be performed to ensure the data coherence
between our 2 sensors measurements.

Multi-sensor validation
A multi-sensor validation between our two sensors deployed during the framework of the project is planned for every
super site where we will perform a drone campaign. During these campaigns, the drone will perform a short static flight
near each micro-station. These flights will give us data to compare the water surface height measured by the altimeter
and the micro-station.

Multi-temporal
Finally, we perform regular validation of the micro-station measurement. This validation is performed via an automatic
editing of the water surface height time series to ensure that there are no aberrant measures. These aberrant measures
can come from multi detection due to fog or from masks in front of the LiDAR (as spider webs for example). This process
runs for every measurement performed by a micro-station.
We also will perform a regular validation of the altimetric reference of the station by a regular PPP positioning from
GNSS measurement by the station. With this processing we will verify the accuracy of the altimetric reference during
the whole life of the micro-station. Finally, we are currently developing new indicators to follow the quality of the
measurements. These indicators will complete the monitoring of the micro-station network and improve the predictive
maintenance.

In situ validation – including multi-temporal, -instrumental, and/or -campaign intercomparisons
Due to the remoteness and difficulties with the obtaining observations underneath the satellite orbit over sea ice,
considering the harsh environment, but also the highly dynamic nature of sea ice, and the atmosphere conditions, as
well as the drift of the ice, it is hard to validate reference observations themselves with under the same conditions.
Nonetheless, ideally the reference observations need to be validated independently. This can either be achieved by
using dedicated field campaigns completed during the collection of other observations (e.g., during drone campaigns)
or by identifying collocated, independent observations (e.g., drifting buoys) along the same path. Both of these
scenarios are usually the exception rather than expected.

When dedicated field campaigns have been conducted, these will be leveraged for validation purposes. For example, as
the drone with the SnowRadar as well as the lidar is currently in testing phases, ideally an in situ field campaign should
be conducted simultaneously. If not possible, the first test of the SnowDrone (only carrying the SnowRadar) conducted
ground-based field measurements (using magnaprobes), which can be leveraged providing they are processed to a
format comparable to the FRM. Important for any in situ field campaign is the sampling of something comparable to
the sensor to be compared with. Currently, there are no dedicated protocols or procedures defined to ideally sample
the sea ice for comparison with Sentinel-3. It is also not clear how the sampling strategy should be for comparison to
airborne campaigns, and what the impact of upscaling to satellite scales is.

Multi-temporal validation is hardly relevant for sea ice purposes, in the sense that sea ice constantly moves and
experiences significant thermodynamic and dynamic changes, which means that even a few hours can have significant
impact. Efforts have been made to fly airborne campaigns along the same flight tracks (or satellite orbits) every year,
but these efforts are primarily to evaluate seasonal differences rather than represent a full validation. Multi-campaign
intercomparisons follow similar considerations in the case of identical campaigns conducted at different time periods.
Former campaigns have been conducted with several airborne campaigns (and/or ground campaigns) simultaneously.
For example, in 2019, CryoVEx, Operation IceBridge, and AWI IceBird flew over similar satellite ground tracks/orbits
with days apart, and with a ground-team measuring the sea ice and snow conditions before, during, and after the flights.
However, such efforts are exhaustive and logistically difficult, hence rarely occur. Future multi-campaigns might be
conducted for some of the synergy or future altimetry missions. Finally, multi-instrumental is essential for being an FRM
over sea ice, meaning that we need at least two of the eight measurands relevant over sea ice
to properly convert between the measurands to validate against the Sentinel-3 Ku-band radar altimetry observations.
Hence, there would need to be additional, independent instruments to compare against.

During the operational phase, we will identify potential in situ observations (in whatever available format) that could be used to validate
the observations, whether it be dedicated campaigns or third-party provisions. Once identified, the protocols and
procedures to validate the validation observations with other reference observations will be discussed following the
metrological approach.

Data coherence and overall quality assessment
Overall, the data quality assessment of the FRM provision will be made according to the performance indicators set
forth for the ORR Successful Criteria.

2.4.3.1. Antarctica – Adélie Land

Due to the remote and challenging conditions of Adélie Land in Antarctica, there is no potential first-order validation in
this study area. The only secured solution at present would be to carry out a posteriori topographic surveys along
previous POCA locations using a topographic wheel and accurately positioning surface locations using PPP protocols.

2.4.3.2. Svalbard

Since surface elevations are determined from a single system, GNSS, there are limited potential for errors beyond the
processing itself, which is quantified in the processing itself, and the tape measurement of antenna height. For the
latter, antenna height over the surface is measured multiple times at different locations/conditions from which an
average value is calculated and used for all surveys. This is to eliminate a potential bias in antenna height over surface.

Internal validation is carried out by repeating parts of survey transects from day to day at random times. This
decorrelates many of the error sources related to satellite geometry, ionosphere and troposphere conditions, as well
as tidal impacts such as ocean loading. From this, we are able to calculate an empirical observational uncertainty which
is independent of the formal uncertainty from the GNSS processing and related assumptions. This internal validation
measure is included as a second uncertainty parameter in the data product, with one average root-mean-square (RMS)
for each campaign. Finally, the extraction of independent surface elevations from two external DEM surfaces (ArcticDEM
and NPI photogrammetric DEM) gives an additional control on the GNSS data quality, which can be used to flag potential
gross errors, if any. These external high-resolution DEMs (10 m and 5 m, respectively) can further be used for slope
correction between the point locations of Sentinel-3 ground tracks and surface GNSS transects.

The aim of the FRM Data Hub is to offer a centralized access to FRM measurements. This is why the filename convention
will cover the main characteristics of FRM measurements.
The filename will contain the following elements:

• Surface Type (SS)
• Geographic Area (GGG_ggg)
• Site/Campaign (site-or-campaign-name)
• Temporal period (YYYYMMDDThhmmss_YYYYMMDDThhmmss)
• Version

Therefore, the filename will look like this:

SS_GGG_ggg_site-or-campaign-name _ YYYYMMDDThhmmss_YYYYMMDDThhmmss _VX.Y.nc

The paragraphs below list the different component values.

The surface type designs which type of surface the measures address. It can take the following values:

The geographic area indicates the area where the measures have been taken. Each type of surface will have a specific
description; however, it is always composed of 2 trigrams.

2.5.2.1. Inland Waters

For Inland waters products: the ISO3 country code and the 3 first letters of the lake/river name are used to this end.
Examples:

2.5.2.2. Sea Ice

For sea ice products: prefix ARC/ANT, followed by 3 letters for the local sea name.
Example:

2.5.2.3. Land Ice

For land ice products: ISO3 country code + 3 letters of the name of the region.
Example:

The “Site/Campaign” element in the filename convention represents the specific site or campaign associated
with the Fiducial Reference Measurements (FRM) dataset. This component is key for identifying the particular
site or campaign where the data was collected, as FRM measurements are often gathered during dedicated field
campaigns or at longterm monitoring sites. For example, for the inland water surface, we will use the
virtual_station_name as the site name.
Example: marmande-2

The temporal period describes the time period covered by the file, from first measurement date to last measurement
date.
The temporal period will be given in the following format: YYYYMMDDThhmmssZ_YYYYMMDDThhmmssZ, in UTC.

The version describes the dataset version. Some datasets can be reprocessed, and in that case, it is important to
distinguish the first dataset version from the following ones.

2.6. Data Format and Associated Metadata

2.6.1. Conventions

2.6.2. Dimensions

2.6.3. Global Attributes

2.6.4. Variables

2.6.5. Maturity Matrix