4. ASCAT Products Overview

4. ASCAT Products Overview

4.1 Inheritance from ERS scatterometers and key features of the ASCAT mission

The design of ASCAT is based on the robust and well-understood concept of the ERS scatterometers. Among the common features are:

• fan-beam geometry with antennas oriented at 45°, 90° and 135° with respect to the satellite track
• VV polarisation
• Radar frequency: C-band
• Swath gridded into nodes, one triplet of averaged backscatter measurements per node
• Generation of an operational product at 50 km resolution on a nodal grid of 25 km

The experience acquired with nearly one decade of operations of ERS scatterometers and the use of their data have also identified areas where improvement was possible, in order to allow operational use of the data, as well as for emerging scatterometer applications to develop. This experience has been taken into consideration by introducing a number of new key features, which are listed below and further discussed in detail in the following paragraphs:

• Continuous data acquisition and product generation without sharing operation time in orbit with other instruments, which would increase the data coverage
• Increased spatial coverage via the use of a double swath
• Higher incidence angle range towards an area which offers a higher wind direction sensitivity in the geophysical model and therefore better performance of the wind retrieval algorithm
• Additional generation of a research product at a higher resolution of approximately 30 km on a 12.5 km nodal grid
• Improved instrument design, resulting in higher stability and reliability
• Improved on-board processing concept, allowing for a lower raw data rate

4.2 The ASCAT instrument

4.2.1 Technical description

ASCAT is a real aperture radar operating at 5.255 GHz (C-band) and using vertically polarised antennas. It transmits a long pulse with Linear Frequency Modulation ('chirp'). Ground echoes are received by the instrument and, after de-chirping, the backscattered signal is spectrally analysed and detected. In the power spectrum, frequency can be mapped into slant range provided the chirp rate and the Doppler frequency are known. The above processing is in effect a pulse compression, which provides range resolution.

The contribution of the thermal noise to the radar measurement is observed within each pulse-repetition interval by monitoring the output of the receiver during a period of time when all the pulse echoes have decayed. Noise measurements are sent to the ground with echo measurements, to enable measurement noise subtraction to be performed during ground processing.

Considerable data rate reduction is achieved by pre-processing the noise and echo measurements on board. The on-board processing consists of a power spectrum estimation followed by a spatial low-pass filtering. Different averaging takes place for echo and noise measurements, reducing the raw data rate by a factor of approximately 25. On-board processing carries as a result a degree of spatial correlation between different and within received echoes, which is taken into account later on by the Level 1b processing.

4.2.2 Instrument modes

The ASCAT instrument may operate in three different modes: Measurement, Calibration and Test. Additionally, in Measurement mode the instrument occasionally generates special source packets for Gain Compression Monitoring.

The nominal instrument mode, and the only one that generates science data for the users, is Measurement mode.

During external calibration campaigns, the instrument is operated in Calibration mode when it passes over the ground transponders. Additionally, the instrument is operated in Calibration mode over the ground transponders for one ascending pass and one descending pass every orbit cycle (29 days). Operating the instrument in Calibration mode implies an interruption of the Measurement mode for that pass of approximately 10 minutes (4400 km of swath length).

In addition, the instrument Gain Compression needs to be monitored periodically. This is a special variant of measurement mode, called Gain Compression Monitoring mode, and monitors the relationship between the transmitted power and the RFU drive level setting. These measurements are not used to retrieve science data. The acquisition of Gain Compression Monitoring data interrupts the nominal Measurement mode for approximately 5 minutes. The instrument needs to monitor the Gain Compression once per orbit cycle over land.

4.2.3 Instrument calibration

The Internal Calibration loop is part of the nominal operation. Within every pulse repetition interval, a calibration phase and a noise estimation phase are included. The calibration signal and the noise data are used to correct for the instrument internal effects during the ground processing. This mechanism allows the combined variation of the transmitter power and the receiver gain to be monitored and accounts for internal instrument drifts during the mission lifetime. At the beginning of the mission, the internal calibration level is established, and its monitoring later on is done with respect to this reference value.

The objective of the External Calibration is to ensure that the normalised radar backscatter value measured from a target is correct (absolute calibration) for all incidence angles (relative calibration). In order to achieve this, the absolute gain of each antenna and the pointing need to be established in flight, and this is done with the help of ground transponders.

Transponders are active devices that, after receiving a pulse from ASCAT, send a delayed pulse back. Each transponder has a known and stable radar cross section as well as a known and stable response delay, and its position on the Earth’s surface is known accurately. ASCAT receives and detects the transponder pulses from within its swath. By comparing the measured transponder echo level and its associated localisation data with the expected data for each transponder, three antenna depointing angles and antenna gain correction values are estimated. This allows a reference calibrated system to be established, against which the system performance can be evaluated and monitored.

The three transponders used for the ASCAT absolute calibration are situated in Turkey. Their relative positions have been carefully chosen to allow a continuous and sufficiently dense sampling of the antenna gains for all elevation angles. For the planned Metop orbit, over a period of one month, this will allow transponder measurements to be obtained for each antenna beam over the full range of incidence angles with cuts at approximately 10 km intervals in ground range.

The external calibration with three transponders has been carried out during the Commissioning Phase (see [RD23]).

4.3 Measurement principle, geometry and coverage

Like the ERS scatterometers, the ASCAT system geometry is based on the use of fan-beam antennas. It covers two 550-km swaths which are separated from the satellite ground track by about 336 km for the minimum orbit height. The ASCAT incidence angle ranges from 25° to 65° (Figure 4.1). For each swath, three antennas illuminate the sea surface at three different azimuth angles, measuring the backscattered signal. At such incidence angles, the main backscattering mechanism is considered to be Bragg resonance, which describes the interaction of the radar signal with short sea surface waves having a wavelength of a few centimetres.

The wind speed and direction near the ocean surface with respect to the antenna viewing angles are determined by using an empirical Geophysical Model Function (GMF), which relates these parameters to the observed backscatter normalised radar cross section (σ₀) [RD43].

For the Metop operational orbit (see Appendix B), a coverage map can be seen in Figure 4.2 for a full day.

Figure 4.1: ASCAT swath geometry for a Metop minimum orbit height (822 km)

 

Figure 4.2: ASCAT swath coverage for 1 day

4.4 Overview of the ground processing and calibration

4.4.1 The Level 1b ground processing

The Level 1 ground processing chain is illustrated in Figure 4.3. The goal of Level 1 ground processing for ASCAT is the generation of the radiometrically calibrated Level 1b backscatter product, namely σ₀ values.

The functionality of the ASCAT main processing chain is highlighted in blue in Figure 4.3. This chain is data driven and is applied to every science source packet to generate Level 1b products. The operational processor also includes an external calibration processing chain, highlighted in red, which is driven by the acquisition of calibration source packets and the need for generating in-flight instrument characterisation functions after an external calibration measurement acquisition campaign. In-flight instrument characterisation parameters are validated outside the ASCAT operational processor; instead they are validated in the EPS Calibration and Validation Facility, highlighted in green. After they are satisfactorily assessed, the instrument in-flight characterisation parameters are uploaded to the main processing chain.

4.4.1.1 The Level 1b main processing chain

The ASCAT power echoes corresponding to both Measurement mode and Calibration mode are subject to front-end processing, namely internal calibration (Power-Gain product correction) and noise corrections (receive filter shape correction and noise power subtraction). Corrected measurement power echoes are then normalised into 256 σ₀ values along every antenna beam projection on the ground and localised on the surface of the Earth.

Spatial averaging (smoothing) in the along- and across-track directions is further applied per beam over all available σ₀ values, with the objective of obtaining a set of three σ₀ values (one from each beam) for each grid node of each swath at the desired radiometric resolution (also called Kp). The weighting function used to carry out the spatial filtering is a two-dimensional Hamming window centred at every node position, which is based on a cosine function and designed to peak in the centre of the samples and taper to zero at the ends. Its width determines in effect the spatial resolution of the σ₀ averaged values, which is defined in the principal cut in each of the along- and across-track directions, as the distance around the centre of the spatial filter when its intensity reaches 50% of the peak value. The averaged σ₀ values at two different horizontal scales and spatial grids are generated this way, leading to two smoothed σ₀ products: with 50 km and approximately 30 km horizontal resolution on grids of 25 km and 12.5 km respectively. Associated with the spatial filtering, the Kp of the averaged measurement is estimated as the standard noise within the averaging area defined by the Hamming window. In that step, the measurement correlation due to the previous on-board along-track and across-track average is additionally taken into account.

Figure 4.4 and Figure 4.5 give an idea of the ground geometry of the spatial filtering. Note that for the higher resolution product, a trade-off was necessary between spatial resolution and radiometric accuracy, by applying a Hamming window width across the swath for the different beams. The resulting spatial resolution ranges from 25 km in the near swath to 34 km in the far swath..

(a) Before 9 September 2009

(b) After 9 September 2009

Figure 4.3: Functional overview of the ASCAT Level 1 ground processing chain before 9 September 2009 (a) and thereafter (b). (In Section 4.4.1.3 this change is explained.)

 

Figure 4.4: Ground geometry of the spatial smoothing for σ₀ values corresponding to the right mid-beam for a given node N, for the 50-km resolution product.

Figure 4.5: Ground geometry of the across-track component of the spatial smoothing for σ₀ values corresponding to two across-track adjacent nodes

4.4.1.2 The external calibration processing chain

Before launch, the instrument is delivered with a set of characterisation parameters in the form of several LUTs, estimated on-ground and characterising the expected instrument performance. Among them are the antenna absolute gain patterns and their electrical boresight pointing, which determine the values of the normalisation factors used later for the generation of σ₀ values during the Level 1b processing. Antenna absolute gain patterns and their electrical boresight pointing are expected to depend on in-flight conditions, and they have to be estimated subsequently in flight. Once these parameters have been characterised in flight, a corresponding in-flight set of normalisation factors can be generated (i.e., key data). The normalisation factors are the conversion coefficients needed within Level 1b processing to convert instrument raw data into σ₀ and are estimated based on the in-flight knowledge of the instrument.

As illustrated in Figure 4 3, antenna gain measurements in the antenna coordinate system are obtained from every transponder calibration measurement and a set of antenna gain patterns and pointings are derived, which need to be examined off-line and their quality evaluated, before they can be uploaded to the operational processor.

A two-dimensional reconstruction of the antenna gain patterns and the estimation of three depointing angles is the method applied in the external calibration processing, based on fitting absolute gain values estimated at different angular positions (antenna azimuth and elevation) by using transponder measurements.

4.4.1.3 The power-to-sigma0 normalisation factors and deblooming kernels

The normalisation factors are the conversion coefficients needed within Level 1b processing to convert instrument raw data into σ₀. Their estimation is based on calculating the frequency power profile corresponding to a distributed target with constant σ₀ = 1.

The deblooming kernels are coefficients intended for (optional) σ₀ deblooming subtraction. Their estimation is based on calculating the inverse impulse response function.

In both cases, these contributions are estimated, by applying the corresponding two-way radar equation while accounting for on-board echo processing corresponding to the ASCAT measurement mode.

The deblooming kernels are currently not calculated, nor applied in the Level 1 operational processor.

With respect to the normalisation factors, an important change in the processing chain was introduced on 9 September 2009. Previously, the normalisation factors were generated after an external calibration campaign by assuming a reference Metop orbit, thus leading to a static file which, after off-line evaluation, was activated in the Level 1 operational processor. On that date, a processor generating normalisation factors for every actual Metop orbit was introduced and those dynamic normalisation factors are used now by the operational Level 1 processing chain. The auxiliary files containing the normalisation factors are therefore no longer static.

4.4.1.4 Level 1b nominal and degraded operation modes

The different instrument modes generate data intended for different types of processing. These are all considered nominal and can be of three types;

• The main nominal processing mode is driven by continuous availability to the processor of measurement data and corresponds to the generation of ASCAT Level 1b products within the main processor chain. These are the products intended for users.
• Another nominal processing mode is driven by the availability to the processor of a sequence of Gain Compression Monitoring data, to the end of generating instrument performance assessment data, which are not sent to users but are used internally in the EPS CGS.
• Finally, the third processing mode considered nominal is the generation of calibration products within the external calibration processing chain, driven by the availability to the processor of a sequence or a number of sequences of calibration data. Calibration products are not available to users either.

Non-nominal operation modes are not defined as such for the ASCAT processor, but both the main Level 1b and the external calibration processing chains can generate products of non-nominal quality, referred to as degraded products. The ASCAT processor has the capability of flagging any degradation of the products in near real-time, by introducing the necessary flags and qualifiers within the products themselves. Those flags and qualifiers are addressed in detail in the sections of this document describing the ASCAT Level 1b product contents, but here the main product degradation scenarios are outlined:

• degradation due to non-nominal antenna pointing, due to spacecraft orbit manoeuvres or orbit/attitude anomalies
• degradation due to instrument performance anomalies, assessed by examining instrument telemetry data
• degradation of the measurements of the Left Fore antenna due to reflections from the solar array
• degradation at the start or end of a measurement mode sequence, due to non-availability of enough echo or noise data to compute the necessary internal calibration and noise correction functions
• degradation due to unavailability of good quality calibration data

As part of the product validation activities, an assessment of the impact on the product accuracy of the above scenarios has been carried out during the commissioning phase (see the 2009 commissioning quality report [RD23]).

Guidelines on the usage of degraded products are provided in the product validation section in this document.

4.4.2 The Level 2 ground processing

4.4.2.1 Ocean winds processing

Level 2 ocean winds processing takes the radiometrically calibrated backscatter product to derive winds near the ocean surface (at a nominal height of 10 m). ASCAT Level 2 ocean wind global and regional products are respectively produced and distributed by the EUMETSAT Ocean and Sea Ice Satellite Application Facility (OSI SAF) and the EARS Project. In both cases the core activities are carried out by KNMI (Royal Netherlands Meteorological Institute).

Near-surface ocean wind vectors are derived from ASCAT Level 1b data, by using a Geophysical Model Function (GMF), which relates σ₀ values to wind speed and direction. The wind retrieval is an inversion problem at each node where, given a set of three σ₀ values and the GMF, the wind vector is computed that has the highest probability of representing the true wind. An inversion problem presented in these probabilistic terms is usually equivalent to the minimisation of a cost function. Usually, two wind vectors are obtained as the most likely solutions, with directions separated by 180°. The measurement noise and the harmonic nature of the GMF result in this wind direction ambiguity as two equal-probability minima occurring in the cost function. The GMF currently used to retrieve ASCAT winds is CMOD5.n [RD33].

A wind direction ambiguity removal step is further applied, based on variational meteorological analysis and relying on prior numerical weather prediction (NWP) model information. In 2D-VAR a cost function is minimised. The cost function is formulated in terms of wind increments and penalises deviations from both a background wind field and the ambiguous scatterometer wind solutions obtained from scatterometer wind retrieval.

Finally, sea ice screening is applied, based on the scatterometer data itself and on sea ice history information.

4.4.2.2 Soil moisture processing

Level 2 soil moisture processing takes the radiometrically calibrated backscatter product to derive soil moisture over land surfaces. ASCAT Level 2 Soil Moisture Index products will be produced and distributed by EUMETSAT following the method developed and prototyped for EUMETSAT by the Institute for Photogrammetry and Remote Sensing of the Vienna University of Technology.

The method for retrieving soil moisture from ERS scatterometer data is a change detection method. Instantaneous backscatter measurements are extrapolated to a reference incidence angle (taken at 40°) and compared to dry and wet backscatter references. By knowing the typical yearly vegetation cycle and how it influences the backscatter-incidence angle relationship for each location on the Earth, the vegetation effects can be removed, revealing the soil moisture variations. Soil moisture is considered to have a linear relationship to backscatter. The historically lowest and highest values of observed soil moisture are then assigned to the 0% (dry) and 100% (wet) references respectively and a value for the relative (surface) soil moisture estimated. The backscatter references are derived a-priori from analysis of multi-annual scatterometer backscatter time series, starting with the first ERS scatterometer measurements in 1991. Details can be found in [RD48].

4.5 ASCAT product characteristics and use

Table 4.1 summarises the main characteristics of ASCAT products available to users. All products contain quality control and other information about the retrieval and their use, which are important to know when you choose the product needed for your application.

The ASCAT Level 0 product is not considered in this section, because no direct research or operational application at the user side has been identified for it. The Level 0 data stream is however available to users interested in the instrument raw measurements and their further processing, thus its contents and format is discussed in the section dedicated to ASCAT product formats and dissemination.

Table 4.1: Summary of the main characteristics of ASCAT products

(*)Note that the figures on accuracy are, for this issue of the ASCAT Product Guide, pre-launch estimations. During Commissioning, the actual accuracy of all products will be assessed and the figures updated.

The set of reprocessed Level 1b and Level 2 Soil Moisture products corresponding to the ASCAT reprocessing Phase 1 are not listed above as a separate product, because even if they have been generated off-line, the same configuration as the operational processor has been used. The purpose of these products is the establishment of a consistent data record, including all the improvements incorporated into the operational processor after the validation activities since the beginning of the mission.

In that sense, the Phase 1 of the Level 1b products covering the 2007-2008 period complements directly the near-real-time products generated in the operational processor as of 2 Dec 2008 and throughout 2009, until 9 Sep 2009. On this date, the dynamic generation and utilisation of the normalisation factors starts a slightly different data record from the processing configuration point of view.

Equally, the Phase 1 of the Level 2 Soil Moisture products covering the 2007-2008 period complements directly the near-real-time products generated as of 4 May 2009 until today. Note that there is a period of data between 1 Jan 2009 and 4 May 2009 which is currently not consistent with this data record.

4.5.1 The ASCAT Level 1b spatially averaged products

These are distributed in near real-time and the main geophysical parameter contained is the spatially-averaged normalised backscatter radar cross section values, known as σ₀ triplets. Each triplet contains three σ₀ values corresponding to each instrument beam for the corresponding swaths, and localised on the surface of the Earth to a set of nodes on a horizontal surface grid. This grid is generated with a node distance of 25 km for the 50-km resolution product and 12.5 km for the higher resolution product. Other important information provided for each triplet is incidence angle, azimuth angle, Kp value and geographical position of the node.

Important to understand is that, due to the characteristics of the spatial averaging, the data from adjacent nodes are correlated. As explained in Section 4.4, the size of the spatial resolution window is about 100 km, and thus the σ₀ value corresponding to a node contains information from individual σ₀ values corresponding to several adjacent nodes along and across track.

Note that due to the antenna geometry, σ₀ triplets may be partially filled at the start or end of a Measurement mode sequence. Those empty σ₀ values are replaced by a default value.

The accuracy of the σ₀ triplets is summarised by the following parameters.

• radiometric accuracy and inter-beam radiometric stability, also called calibration accuracy, expected to be less than 0.5 dB peak-to-peak and between beams for both resolution products.
• radiometric resolution, also called Kp, expected to be from 2 to 5% for the 50-km resolution product and from 4 to 11% for the higher resolution product.
• localisation accuracy, expected to be about 4 km for both resolution products.

Level 1b data contain a number of flags that indicate the quality of the σ₀ and Kp values. The summary flags that take an integer value are:

• F_USABLE – summary flag for σ₀ (0, 1, 2 correspond to good, usable, bad)
• F_KP – Kp quality flag (good = 0, bad = 1)

The remaining flags take a value from 0 to 1 (where 0 indicates the highest quality) and are:

• F_F – synthetic data have been used to fill data gaps (not validated yet)
• F_V – if amount of synthetic data used is above a configurable threshold (not validated yet)
• F_OA – quality of satellite orbit and attitude
• F_SA – data may be corrupted by solar array reflections (affecting the FORE MID beam on descending passes only)
• F_TEL – instrument and satellite housekeeping telemetry data are present and within limits (currently not activated)
• F_EXT_FIL – extrapolated filter shape has been used to correct data
• F_LAND – indicates the fraction of land contributing to a σ₀ node value.

Additionally, the processor can be configured to consider the instrument calibration and/or the status of the products as commissioning and to reflect this in the summary F_USABLE flag by setting its value directly to 2. This flagging was activated until processor version 5.3.1. From 5.6.0, the products are considered formally operational, hence F_USABLE is not affected by configuration but based solely on a summary of the other flags.

The implementation of F_SA is based solely on geometry considerations (relative angle between solar array panel and the ASCAT antennas). It still remains to assess the actual impact (if any) of the solar array reflections on the product quality (σ₀ and Kp).

The intention of the F_OA flag is to detect orbit or attitude anomalies that make the normalisation from echo power to σ₀ inaccurate or inappropriate, i.e., if the actual orbit or attitude deviates from the assumed ones in the generation of the normalisation factors.

F_OA also takes into account the effect of manoeuvres. During Out Of Plane (OOP) manoeuvres, the ASCAT instrument is switched off and product dissemination only resumes once the spacecraft is stable and back to its reference orbit. During In Plane (IP) manoeuvres, the ASCAT instrument is left on and the processor running. In that situation, a manoeuvre flag is triggered, which contributes to F_OA.

F_USABLE is an advisory flag on the overall quality of the product. Its value is nominally set to 0 (=GOOD) if:

• the value of all the specific flags described above is 0 and
• the products are considered operational (set via configuration) and
• the instrument calibration is considered good (set via configuration).

F_USABLE can also have a value of 1 (=USABLE), triggered by:

• non-zero values of F_F (if F_V less than a certain threshold), F_TEL, F_EXT_FIL, F_KP, or
• cases of inaccurate instrument calibration (set by us via configuration),
• while F_OA and F_SA still hold a value of 0.

F_USABLE is set to a value of 2 (=BAD) if it is neither GOOD nor USABLE, i.e., typically

• if F_V is greater than a certain threshold,
• if either F_OA or F_SA are non-zero (i.e., the normalisation of echo power to σ₀ is not correct or the σ₀ value is potentially affected by solar array reflections),
• or if the products are still considered under commissioning (set by us via configuration).

With respect to F_SA, note that up to processor version 5.6.0, this flag sets the value of F_USABLE to 2 (=BAD). Given the experience with the use of the data and no observed quality degradations of nodes potentially affected by this flag, this constraint has been relaxed to allow the F_SA flag to set the value of F_USABLE to 1 (=USABLE). This change has been implemented from processor version 6.2.0.

It is important to understand that, due to the antenna geometry, the triplets generated at the beginning and end of a measurement mode sequence are incomplete, when the corresponding node locations are only viewed by one or two antennas. There is no flag for incomplete triplets, the missing values are set to the default field value.

For the flags listed above referring to a 'fraction of data' affected by some anomaly, it is important to understand that the flag is calculated by applying the same Hamming window used for the spatial averaging to the corresponding individual flags associated with the individual σ₀ values averaged for that node. The flags are therefore not a 'fraction' in the strict sense, but rather a weighted average of flags associated with individual σ₀ values.

4.5.2 The ASCAT Level 1b full resolution product

These are not distributed in near real-time, but are available from the EUMETSAT Data Centre. The main geophysical parameter contained in the product is the normalised backscatter radar cross section - 256 values along each antenna beam, localised on the surface of the Earth. This product does not include measurement collocation from the three different beams that observe a point of the Earth. It corresponds to an intermediate product generated just before the swath node generation and spatial averaging steps in the main processing chain, hence called a 'full resolution' product. The data are then organised per position along each beam, and not per node in a swath. Other important information provided for each σ₀ value is incidence angle, azimuth angle and geographical position of the sample.

The resolution of these σ₀ values is defined across-beam by the diffraction effect due to the finite size of the antenna. The antenna pattern averages about 20 km across-beam on ground, and the sampling rate corresponds to 1.4 km on ground. Along-beam the resolution of the individual σ₀ values is given by the range discrimination provided by the azimuth pulse compression. This resolution is variable along the beam but is of the order of 10 km. The sampling of individual σ₀ values along-beam is of approximately 2 km for mid beams and 3 km for side beams. As a result, the full resolution σ₀ values along every antenna beam represent footprints of about 10 x 20 km of various shapes and orientations, depending on the Doppler pattern over the surface. Due to the on-board processing, the individual σ₀ values carry a certain amount of measurement correlation. For a deeper understanding of this correlation, see more details about the on-board processing in [RD16].

The accuracy of the σ₀ individual values is summarised by the following parameters.

• radiometric accuracy and inter-beam radiometric stability, also called calibration accuracy, expected to be less than 0.5 dB peak-to-peak and between beams for both resolution products.
• localisation accuracy, expected to be of about 4 km for both resolution products.

A number of quality flags are generated during the Level 1b processing, associated with every individual σ₀ sample along each antenna beam, and included in the Level 1b full resolution products. These contain equivalent information to those found in the averaged Level 1b product, but assigned to individual values of σ₀. A full list and detailed explanation of all flags is later given in the Level 1b full resolution product content and format description.

4.5.3 The ASCAT Level 2 ocean surface wind product

Here, only general information is given. For more details, please refer to the OSI SAF scatterometer webpage.

The ASCAT Level 2 ocean surface wind product is generated and distributed in near real-time. The main geophysical parameter is near surface (10 m height) ocean (neutral) wind vectors. The operational Level 2 product is nominally generated by the OSI SAF based on the swath-based 50-km resolution Level 1b product received from EUMETSAT on a 25 km spacing grid.

The product contains the two ambiguous wind vectors located at each node, as well as a flag indicating which one has been selected as corresponding to the correct wind direction by the ambiguity removal step. As an integral part of the product, the necessary geophysical parameters used for the retrieval from the Level 1b product are also included in the Level 2, namely:

• σ₀ values over non-land surfaces, as well as their corresponding Kp values
• measurement geometry information, such as incidence angle, azimuth angle, latitude and longitude position.

Additionally, the Level 2 product contains other auxiliary information used in the retrieval, such as collocated NWP ocean wind values used in the ambiguity removal step.

The expected accuracy of the ASCAT 50-km resolution ocean winds is 2 m/s RMS difference in wind vector components and 0.5 m/s wind speed bias with respect to a reference 'true wind'.

With each ASCAT Level 2 wind vector, a value reflecting the likelihood of that retrieved wind solution is also given in the product, which accounts for the residual of the wind retrieval and a probability associated with that wind direction, as given by the wind direction ambiguity removal step.

4.5.4 The ASCAT Level 2 Soil Moisture product

Here, only general information is given. For more details, please refer to [RD48].

4.5.4.1 General characteristics

The ASCAT Level 2 Soil Moisture product is generated and distributed in near real-time. The main geophysical parameter is relative land surface soil moisture, based on the swath-based grid.

4.5.4.2 Accuracy

The expected average RMS error of the ASCAT soil moisture index is about 25%, which corresponds to about 0.03-0.07 m³ water per m³ soil, depending on soil type.

4.5.4.3 Quality information in the products

In order to help users in judging the quality of the soil moisture product, several quality and processing flags are delivered with the data. The quality flags comprise information about the intrinsic product quality, internal quality checks and specific processing details. The flags are derived directly from the incoming scatterometer data.

Additionally to these quality flags, advisory flags are defined. The advisory flags will support the user in judging the reliability of the soil moisture product for a particular location on a particular day of the year and will allow the user to consider rejecting unreliable measurements. These indicators are based on average land cover characteristics, soil type and topography, which cannot be derived from the scatterometer data set, but need to rely on external data sets.

For more information on the ASCAT soil moisture flags, see [RD48].

4.6 Summary of ASCAT product current and potential applications

A number of operational applications of scatterometer data have been developed in meteorology and climate monitoring, after the success of the recent scatterometer missions. The main operational product from scatterometers remains the wind over the ocean surface, and the main operational application is the use of scatterometer winds in weather analysis and forecasting. Scatterometer winds are used routinely to track tropical and extra-tropical cyclones.

A second group of operational applications, based on the use of scatterometer σ₀ estimates, have been developed recently, such as sea ice edge detection and monitoring, where maps of sea ice coverage based on scatterometer data are routinely produced. Besides the operational use of scatterometer data, there are a number of scatterometer applications at various levels of maturity, in the fields of oceanography as well as for monitoring sea ice, snow cover and land surface.

4.6.1 Scatterometer ocean wind applications

4.6.1.1 Ocean winds assimilation in weather forecasting

The main application of scatterometer ocean winds is the assimilation in NWP models. Since the main limitation for this application is the problem of the wind direction ambiguity, the impact and the benefits of the assimilation of scatterometer winds in NWP models were clearly improved when variational data assimilation was developed. These schemes can ingest the two wind direction ambiguities and perform the ambiguity removal within the assimilation process, by weighting one wind direction or another according to the information provided by the background field and the other observations used in the assimilation. ERS and SeaWinds winds have been and are operationally assimilated in global models at ECMWF and the Met Office (UK). Also regional NWP models use scatterometer data following similar principles. ASCAT winds are expected to be used in the same way.

4.6.1.2 Nowcasting and forecasting of extreme events

Scatterometer ocean winds can be used also in Nowcasting and forecasting of extreme events. The added value of the scatterometer data in this particular application is two-fold. First, the scatterometer measurements are available in data-sparse oceanic regions where extreme weather events such as tropical cyclones are generated, and they are available in cloudy and rainy conditions. Secondly, scatterometer winds contain much sub-synoptic scale information that, though difficult to assimilate into NWP, is vital for Nowcasting applications. Moreover, the increased coverage and timeliness of ASCAT with respect to ERS winds facilitates the routine use of scatterometer winds by operational weather forecasters.

4.6.1.3 Winds forcing ocean and wave models

An important limitation in the interpretation of the results obtained by ocean models is considered to be the uncertainties in the surface wind fields, especially in tropical areas, where the ocean reacts strongly and quickly to the atmosphere's variability. Again in this case, conventional instrumentation data is generally sparse in these areas. The main parameter of interest in this case is ocean surface wind stress, related to the near-surface ocean winds.

The scatterometer winds play also an indirect role in operational ocean wave forecast models, as most of these models are driven by an atmospheric model wind analysis that has assimilated scatterometer winds (e.g. ECMWF model).

4.6.2 Other scatterometer applications

Sea ice detection based on scatterometer data is mainly based on two properties of the backscatter response from sea ice. First, sea ice surfaces have isotropic backscatter response, in contrast to the strong anisotropic response from the open water. Secondly, the dependency of the backscatter response with respect to incidence angle is generally smaller for sea ice than for the ocean. Based on these properties, and on available in situ validation sea ice data, models describing the scatterometer sea ice backscatter signatures have been developed, which point at one key sea ice characteristic driving the backscatter response: sea ice age/thickness.

Based on these results, multi-sensor sea ice detection and classification algorithms relying on multivariate analysis with microwave (active and passive), infrared and visible data as input, have been developed, in which the sea ice information contained in the scatterometer measurements plays a key role.

4.6.2.2 Snow accumulation over great ice sheets

In order to understand the relationship between global warming and changes in the ice sheets, it is necessary to quantify mass balance and the time and space scales of variability in snow accumulation and ablation. Microwave backscattering depends on both the roughness and physical characteristics of the snow and ice. This provides an efficient means of monitoring backscatter changes, which accompany diagenetic changes in the ice sheets forced by seasonal accumulation, densification, temperature cycling, ablation and/or metamorphic processes.

Radar backscatter from land corresponds to a mixture of surface and volume scattering associated with the radar penetration depth. The surface roughness, vegetation cover and terrain dielectric properties play an important role in determining what the dominant backscatter regime is. The response of the land surface to the radar with respect to incidence angle (slope) is different for both regimes. Hence examining not only the backscatter intensity but also the backscatter slope allows assessing surface type, vegetation cover and soil moisture content.