The ASCAT Soil Moisture Product: A Review of its Specifications, Validation Results, and Emerging Applications

TL;DR: The Advanced Scatterometer (ASCAT) is a C-band active microwave remote sensing instrument flown on board of the Meteorological Operational (METOP) satellite series as discussed by the authors.

Abstract: Many physical, chemical and biological processes taking place at the land surface are strongly influenced by the amount of water stored within the upper soil layers. Therefore, many scientific disciplines require soil moisture observations for developing, evaluating and improving their models. One of these disciplines is meteorology where soil moisture is important due to its control on the exchange of heat and water between the soil and the lower atmosphere. Soil moisture observations may thus help to improve the forecasts of air temperature, air humidity and precipitation. However, until recently, soil moisture observations had only been available over a limited number of regional soil moisture networks. This has hampered scientific progress as regards the characterisation of land surface processes not just in meteorology but many other scientific disciplines as well. Fortunately, in recent years, satellite soil moisture data have increasingly become available. One of the freely available global soil moisture data sets is derived from the backscatter measurements acquired by the Advanced Scatterometer (ASCAT) that is a C-band active microwave remote sensing instrument flown on board of the Meteorological Operational (METOP) satellite series. ASCAT was designed to observe wind speed and direction over the oceans and was initially not foreseen for monitoring soil moisture over land. Yet, as argued in this review paper, the characteristics of the ASCAT instrument, most importantly its wavelength (5.7 cm), its high radiometric accuracy, and its multiple-viewing capabilities make it an attractive sensor for measuring soil moisture. Moreover, given the operational status of ASCAT, and its promising long-term prospects, many geoscientific applications might benefit from using ASCAT soil moisture data. Nonetheless, the ASCAT soil moisture product is relatively complex, requiring a good understanding of its properties before it can be successfully used in applications. To provide a comprehensive overview of themajor characteristics and caveats of the ASCATsoil moisture product, this paper describes the ASCAT instrument and the soil moisture processor and near-real-time distribution service implemented by the European Organisation for the Exploitation of Meteorological Satellites (EUMETSAT).A review of themost recent validation studies shows that the quality of ASCAT soil moisture product is – with the exception of arid environments –comparable to, and over some regions (e.g. Europe) even better than currently available soil moisture data derived from passive microwave sensors. Further, a review of applications studies shows that the use of the ASCAT soil moisture product is particularly advanced in the fields of numerical weather prediction and hydrologic modelling. But also in other application areas such as yield monitoring, epidemiologic modelling, or societal risks assessment some first progress can be noted. Considering the generally positive evaluation results, it is expected that the ASCAT soil moisture product will increasingly be used by a growing number of rather diverse land applications.

Summary

1 Introduction

• The Advanced Scatterometer is an active microwave remote sensing instrument that was designed for monitoring of winds over the oceans in support to operational applications such as numerical weather prediction (NWP), tropical cyclone analysis, and ocean waves forecasting (ISAKSEN and STOFFELEN, 2000; FIGA-SALDAÑA et al., 2002; LIU, 2002).
• Over land, no operational services were initially foreseen.
• And indeed, many of the initial ESCAT validation studies carried out by independent research teams unexpectedly found quite encouraging results (PELLARIN et al., 2006; BROCCA et al., 2009; RÜDIGER et al., 2009).
• Secondly, the spatial resolution of ESCAT and ASCAT, which is in the order of tens of kilometres (25-50 km), is commensurate with the requirements of NWP models, while e.g. hydrological models run on much finer spatial grids.

2 Mission specifications

• The ASCAT soil moisture service owes several of its attractive features to the long and successful heritage of space borne ocean wind vector monitoring programmes.
• In particular Europe can look back to a series of successful scatterometer missions, starting with the ERS satellite programme operated by the European Space Agency (ESA), and continuing to the on-going Meteorological Operational satellite programme operated by EUMETSAT.
• The high continuity provided by these European satellite programmes (Section 2.1), in combination with the strong heritage in the sensor design from one instrument generation to the next (Section 2.2), is the basis for the continuity, reliability and promising longterm prospects of the ASCAT soil moisture service.

2.1 Satellite programmes

• The first European scatterometer was the one flown on board of the two European Remote Sensing Satellites ERS-1 and ERS-2. Z., 22, 2013 eschweizerbart_xxx Operational satellites, whereas the first satellite (-A) was launched in October 2006 and the second (-B) in September 2012.
• Even for the successor instrument of ASCAT, which will be flown on board of one of the Second Generation (SG) satellites of the EUMETSAT Polar System, plans are already well advanced (LIN et al., 2012).
• As one can see in Fig. 1a, which shows the daily global coverage achieved by one satellite (e.g. METOP-A), the gaps in coverage are largest near the equator, while at higher latitudes full daily coverage is achieved over the two poles (>65 ) and in the latitudinal belt between about 35 and 55 .
• This is an important constraint in using the ASCAT soil moisture data, because applications need to be developed in such a way as to cope with the highly irregular coverage, or to settle for using interpolated (and thus more uncertain) measurements.

2.2 Instrument

• ASCAT is a fixed fan-beam scatterometer which uses six side-ways looking antennas to illuminate two 550 km wide swaths to each side of the satellite track (Fig. 2).
• It is operated at a frequency of 5.3 GHz (C-band) in VV polarisation, i.e. it both transmits and receives electromagnetic waves in vertical polarisation only (ver- tical polarisation means that the electric field vector, which defines the polarisation of the electromagnetic wave, has a vertical component relative to the earth’s surface).
• After reception, the backscatter echoes are amplified and further processed for echo power detection.
• This strong dependence of the backscattering intensity on the soil moisture content implies that ASCAT r0 measurements provide a relatively direct measure of the soil moisture content over bare soils.
• Other favourable technical specifications of ASCAT are: d ASCAT backscatter measurements are well calibrated and very stable over time (WILSON et al., 2010).

3.1 Physical basis

• The physical basis for the capability of ASCAT to measure soil moisture is the strong dependence of C-band backscatter on the soil moisture content in the top soil layer (usually held to be 1-2 cm thin).
• Vegetation moisture content and geometric structure are thus key factors for the backscatter, especially since most structural elements of forests, shrubs etc. are comparable in size with typical microwave wavelengths (1-25 cm).
• This was especially the case when the radar echoes were observed at lower incidence angles.
• The droplets of the clouds are randomly located and considered to be held in place by the vegetative matter.
• Also the ASCAT soil moisture product retrieval scheme uses a model that is very similar in functionality to the Cloud Model, depicting e.g. enhanced vegetation scattering at large incidence angles and a reduced sensitivity to soil moisture during the peak of the vegetation season (WAGNER, 1998; WAGNER et al., 1999a).

3.2 Algorithm

• The algorithm for the ASCAT soil moisture product was developed by the Vienna University of Technology (TU Wien) and is from its conception a change detection method.
• While static vegetation effects are implicitly accounted for by these assumptions, there is still a seasonal vegetation component that needs to be corrected for.
• In other words, there is an incidence angle where the backscattering coefficient r0 is stable despite seasonal changes in above ground vegetation biomass.
• The surface soil moisture content ms is estimated in one of the last processing steps using ms ¼ r0 r0dry r0wet r0dry ð1Þ where r0 is the backscatter measurement to be inverted and r0dry and r 0 wet are the backscattering measurements representing a dry and wet earth respectively.
• Overall, the results obtained in experimental validation studies for both ESCAT and ASCAT suggest that the assumptions of the TU Wien change detection model are in general quite reasonable (Section 4).

3.3 Product properties

• Both EUMETSAT and TU Wien generate and distribute a soil moisture product based on the same algorithm but with different product properties.
• The products can be classified according to the processor from which they are generated, their spatio-temporal representation and the production time.
• The time series are infrequently reprocessed and updated at TU Wien, taking always the most recent algorithmic updates into account.
• Due to the processing effort in deriving model parameters and the requirement for temporally representative data time series WARP NRT uses model parameters produced by the WARP system (Fig. 5).
• In the year 2012 both products became part of EUMETSAT’s Satellite Application Facility on Support to Operational Hydrology and Water Management (H-SAF) which is an important milestone in guaranteeing the long-term operations of these products.

3.4 Error propagation

• The goal of the error propagation is to provide along with each soil moisture estimate ms a measure of the uncertainty pertaining to it, expressed as standard deviation of its error distribution.
• One exception is the estimate of the noise of the slope and curvature parameters, which is obtained not by error propagation, but by employing a Monte Carlo approach.
• The ESD characterises the uncertainty due to noise sources that affect the backscatter measurements, from speckle to geo-location uncertainty and residual azimuthal effects (WAGNER et al., 1999c).
• Over tropical forests and other densely vegetated regions, backscatter variations and hence the sensitivity are very small (< 2dB), thus yielding a high soil moisture retrieval error (Fig. 6).
• 12 W. Wagner et al.: The ASCAT Soil Moisture Product Meteorol.

3.5 Advisory flags

• In certain situations, for example when open water, snow or frozen soils dominates the satellite footprint, the retrieval of soil moisture is heavily impacted or not possible at all.
• The impact of these effects is not explicitly part of the TU Wien change detection model, which nonetheless estimates a soil moisture value in these situations.
• Therefore, aside from the astute analysis of soil moisture values themselves, the subsequent advisory flags for snow, frozen soil, surface water fraction, and topographic are also provided with EUMETSAT’s NRT product.
• Even so, users of the ASCAT soil moisture product are advised to use the best auxiliary data available to them for improving the flagging of non-valid soil moisture retrievals.
• If users are only interested in historic time series they may use reanalysis data to improve the flagging of snow and frozen soil.

3.5.1 Snow

• Backscatter from snow is often considered to consist of three components: scattering from the top snow surface, the underlying ground surface and the volume scattering from within the snow pack (ULABY et al., 1986; FUNG, 1994).
• The exact scattering behaviour depends on several physical parameters of the snow layer, including the liquid water content, the roughness of the air-snow interface, the layering of the snow pack, and the grain size and shape.
• In terms of the backscattering characteristics a snow layer can be classified into dry or wet, depending on the liquid water content, which in turn has an influence on the penetration depth of the signal.
• A wet snow with a smooth surface might have a lower signal than a dry bare soil.
• A snow advisory flag based on a historic analysis of SSM/I snow cover data (NOLIN et al. 1998) gives the probability of the occurrence of snow for a particular day.

3.5.2 Frozen soil

• The soil dielectric constant strongly decreases at temperatures below 0 C due to the inability of the soil water molecules to align themselves to the external electromagnetic field.
• As a result, backscatter drops and frozen soil shows comparable backscatter characteristics as dry soil at microwave frequencies (HALLIKAINEN et al., 1984).
• In order to exclude soil moisture estimates governed by frozen soil conditions, a frozen land surface flag based on a historic analysis of modelled climate data (ERA-40) (UPPALA et al., 2005) is part of the advisory flags.
• It gives, similar to the snow advisory flag, the probability of frozen soil conditions for each day of the year.

3.5.3 Surface Water Fraction

• Due to the short penetration depth (< 1-2 mm) of C-band microwaves into water, backscatter characteristics are primarily controlled by the roughness of the water surface.
• In case of a smooth, calm surface, water acts like a mirror (so-called specular reflection) and almost the complete signal scatters into the forward direction.
• It is exactly this effect that is exploited for the retrieval of the wind direction of open water (STOFFELEN, 1998).
• In case of surface soil moisture retrieval the contribution of open water has a disturbing influence on the signal if the area covered by open water surface within the footprint is large.
• Therefore an inundation and wetland flag, derived from the Global Lakes and Wetlands Database (LEHNER and DÖLL, 2004), provides information on the fraction of water covered by the surface.

3.5.4 Topographic complexity

• In mountainous areas backscatter can show significant variations which are not necessarily coupled with soil moisture changes.
• The high variability of the surface topography directly influences the scattering behaviour.
• Calibration errors resulting from the differences between the real surface and the assumed ellipsoid can also have an impact on the backscatter.
• For this reason, a topographic complexity flag, derived from a global digital elevation model (GTOPO30) data is provided.
• The flag contains a standard deviation of the elevation normalized to the values between 0 and 100 and enables an initial understanding of the underlying local topographic conditions.

3.6 Higher Level Products

• As the requirements of different applications may vary significantly, there is a need to combine the original ASCAT satellite retrievals with auxiliary data to produce a range of value added soil moisture product.
• Many applications are not interested in the soil moisture content of the thin (1-2 cm) remotely sensed soil layer, but require estimates of the soil moisture content in the soil profile.
• This requirement is addressed by the Soil Water Index (SWI) product (Section 3.6.1) and by data assimilation schemes as the one of the European Centre for Medium-Range Weather Forecasts discussed in Section 3.6.2.
• Another important requirement of many applications is to have finer resolution soil moisture data.

3.6.1 Soil Water Index

• Estimating the profile soil moisture content from one single ASCAT surface soil moisture image is not possible; the deeper soil layers may either be wetter or drier than the soil surface depending on the weather conditions within the last few days to weeks.
• The resulting SWI time series has an exponential autocorrelation function with a characteristic time length T, agreeing with theoretical expectations (DELWORTH and MANABE, 1988) and empirical observations (VINNIKOV et al., 1996).
• The operational dissemination of the SWI product started in fall 2012.
• Z., 22, 2013 eschweizerbart_xxx may accordingly behave similar, particularly at short time scales.

3.6.2 Profile soil moisture through data assimilation

• Another approach to estimate root zone soil moisture from near surface soil moisture relies on satellite data assimilation in Land Surface Models.
• For all these approaches, the Land Surface Model used in the data assimilation scheme describes the physical processes that control land-atmosphere interactions, including vertical transfer of soil moisture between the surface and root zone reservoirs.
• The retrieved ASCAT root zone soil moisture is an optimal combination between the modelled first guess, the screen-level temperature and humidity analyses, and the ASCAT-derived surface soil moisture which is propagated forward in time through the root zone profile.
• It has been extensively evaluated against ground soil moisture measurements and showed to yield better estimates of soil moisture conditions when compared to model or satellite estimates alone (ALBERGEL et al., 2010; ALBERGEL et al., 2012).

3.6.3 1 km disaggregated soil moisture

• To disaggregate coarse scale microwave measurements they are usually combined with finer resolution satellite data acquired either by synthetic aperture radars (DAS et al., 2011) or visible/infrared imagers (PILES et al., 2011).
• This means that the relationship between local scale and regional scale measurements may be approximated by a linear model.
• In other words, when the regression parameters of Meteorol.
• The coefficients cASAR and dASAR are the two scaling parameters which are derived from long ASAR backscatter time series using the methods described in WAGNER et al. (2008).

4 Validation

• Given that the ASCAT soil moisture product had initially not been planned as part of the METOP operations, there have been no dedicated calibration and validation (Cal & Val) activities as usually being performed after the launch of new satellite missions.
• Even so, ASCAT has profited significantly from Cal & Val activities performed within the framework of other satellite missions used for global mapping of soil moisture.
• SMOS is the first spaceborne mission that was designed specifically for the purpose of soil moisture monitoring over land (KERR et al., 2010).
• Its launch has been an important impetus for setting up new in-situ soil moisture networks, carrying out intensive field and airborne campaigns, and pursuing novel validation and data assimilation approaches (DELWART et al., 2008).
• And finally, also the increasing availability of soil moisture data derived from multi-frequency microwave radiometers such as AMSR-E (Advanced Microwave Scanning Radiometer for EOS) or WindSat have invigorated research- and validation activities in the soil moisture domain (WAGNER et al., 2007b).

4.1 Validation issues

• The validation of spaceborne soil moisture retrievals is challenging for two main reasons:.
• Firstly, soil moisture is highly variable in space and time (WESTERN et al., 2002), making it very difficult to match the intermittent and spatially irregular satellite measurements with independent reference data.
• In fact, satellite and model data often compare better with each other than each of them with the in-situ measurements (PELLARIN et al., 2006).
• In light of these issues, it is probably more appropriate to interpret validation results in a relative context (e.g. assessing the relative performance of a number of different satellite data sets against the same in-situ and model data) rather than attributing ‘‘absolute’’ meaning to the results.
• The main goal of validation activities is to determine the bias and root mean square error (RMSE) through a direct 16 W. Wagner et al.: The ASCAT Soil Moisture Product Meteorol.

4.2 Validation over experimental sites

• The ASCAT soil moisture data have already been validated over several well instrumented test sites situated in different climatic regions with different land cover.
• Overall the results were quite positive, albeit at two stations (one located in a mountainous region) no significant correlations were obtained.
• BROCCA et al. (2010a) validated an improved version of the ASCAT product (produced off-line by TU Wien) over a site in Central Italy using both in-situ and simulated soil moisture data.
• This may cause volume scattering from deeper soil layers or scattering by subsurface discontinuities e.g. a rock surface beneath a shallow soil layer (MÄTZLER, 1998; ELSHERBINI and SARABANDI, 2010), potentially leading to enhanced backscatter and hence higher soil moisture retrievals.
• They found average correlations of 0.53 and 0.45 for ASCAT and SMOS respectively, suggesting that ASCAT retrieval capabilities are comparable to the ones of SMOS.

4.3 Triple collocation

• The validation of the ASCAT soil moisture data over experimental sites allows a quantitative assessment of the retrieval accuracy.
• Triple collocation (albeit called differently by some authors) has for long been applied for estimating the errors of different satellite products, such as evapotranspiration (ROSEMA, 1993) or ocean winds (STOFFELEN, 1998).
• The basic idea behind triple collocation is that the error structure of three independent data sets can be resolved if the errors are uncorrelated.
• One combination of three independent data sets is the triple of ASCAT soil moisture retrievals, AMSR-E retrievals obtained with the LPRM model, and a modelled soil moisture data set such as the one using the Noah model of the Global Land Data Assimilation System .
• One finds that the estimated errors of the anomalies are somewhat larger than the errors of the absolute values as obtained by error propagation (Fig. 6), but overall, the spatial patterns are comparable.

5 Emerging applications

• The use of a new data type in applications is usually very challenging, simply because models are built around input data that were available at the time when the models were developed.
• This process usually takes many years, and even though the first global soil moisture data set derived from the ASCAT predecessor ESCAT was already released in 2002 (SCIPAL et al., 2002), the development of applications for the ASCAT soil moisture products is only in its beginning.
• In the following, several of the emerging applications of the ASCAT soil moisture data will be discussed, reviewing published applications studies for ESCAT and ASCAT and presenting some results of the authors for a better illustration of the challenges and the potential of using this new data type.

5.1 Numerical Weather Prediction

• Reasons to use soil moisture data in Numerical Weather Prediction (NWP) are manifold.
• On a regional scale, MAHFOUF (2010) assimilated globally bias-corrected ASCAT data with a simplified Extended Kalman Filter (sEKF) and focused mainly on forecasts of 2m temperature and humidity, showing some improvement for bias over Central Europe.
• In the study briefly presented here, the impact of soil moisture assimilation on rainfall forecasts, especially convective precipitation in complex terrain was investigated.
• The model has a horizontal grid point spacing of 9.6 kilometres and 60 vertical levels, the global coupling model is Météo Frances ARPEGE.
• Both a global and a local CDF matching were applied to the data set.

5.2 Runoff forecasting

• Accurate flood forecasts rely on appropriately estimated current hydrological conditions at the time of the forecast.
• It is difficult to cover large areas by the sensors due to logistic constraints, and the spatial support or footprint of one measurement is usually only a few centimetres (GRAYSON and BLÖSCHL, 2000).
• The good predictive ability of SWI values for the prediction of runoff response for lead times in the order of 10 days up to several weeks at large catchments is also supported by studies in South Africa (VISCHEL et al., 2008) and the Zambezi (MEIER et al., 2011).
• Different remotely sensed soil moisture products are used along with various hydrologic models, ranging from physically based approaches to simple conceptual models.
• The positive impact of assimilating ASCAT surface soil moisture into hydrological models with an explicit description of the surface soil moisture seems to be smaller compared to the assimilation of SWI into the root zone layer (BROCCA et al., 2012a).

5.3 Vegetation and Crop Growth Monitoring

• The root zone moisture supply is one of the main factors limiting plant growth, particularly in arid, semi-arid and temperate climatic zones.
• To illustrate how these two parameters are related over larger domains, Fig. 12 shows the correlation of monthly NDVI and SWI time series over Africa for the years 2007 to 2009.
• But they can equally be applied in a distributed model at regional scale (DE WIT and VAN DIEPEN, 2007).
• Especially precipitation, soil input data and related soil water content variations need to be considered, because of their importance for soil water storage and water availability for crops (EITZINGER et al., 2008).
• The use of information on spatial variability of top soil moisture as crop model input could improve the spatial crop yield simulations as compared to the use of the point information of single weather stations.

5.4 Epidemic risk assessment

• Soil moisture data can be used for modelling infectious diseases forced by weather and environmental parameters, particularly mosquito-borne diseases (MONTOSI et al., 2012).
• Under recent global warming, however, mosquito-borne disease outbreaks are also observed in mid-latitudes more frequently.
• The Bluetongue virus (BTV) gained public attention due to economical losses of 150 M€ (HOOGENDAM, 2007) caused by the first outbreak of BTV serotype 8 in North-western Europe in 2006 (CONRATHS et al., 2009).
• In their study the authors used temperature and precipitation forecasts from the Austrian meso-scale NWP model ALADIN (WANG et al., 2011) and ASCAT soil moisture interpolated to a 10 km grid.
• Having estimates of the spatio-temporal distribution of the vectors and the hosts allows calculating risk maps.

5.5 Societal risk assessment

• In this context risk analysis includes the assessment of threats that a natural hazard poses to an exposed social system and of the potential impacts it could cause.
• Water shortage on the other hand can result in a rapid decrease of soil water storage.
• Wildfires and its spatial patterns have been set in relation with soil moisture conditions for various case studies in different regions of the world such as Canada’s Northwest Territories (LEBLON et al., 2002), Alaska (KASISCHKE et al., 2007), Siberia (BARTSCH et al., 2009a) and Africa (AUBRECHT et al., 2011).
• Through the application of freely available global datasets and different satellite-data derived flood masks (e.g. from MODIS, Landsat etc.) an impact assessment on population, land cover and infrastructure was carried out.
• In that case information on thresholds for the hazard (such as saturation or dryness of soil) has to be combined with vulnerability factors reflecting susceptibility and the lack of resilience of the society, in order to allow assessing the risk and associated potential impacts.

6 Conclusions and outlook

• The ASCAT soil moisture product can be regarded as an example that, often, science does not proceed along predetermined pathways.
• Of course, there are also situationswhere the quality of the ASCAT retrievals is problematic, e.g. over mountainous regions or over some desert areas where, for the time being, it might be better to use the SMOS or AMSR-E retrievals.
• A better understanding of sensor performances will also open the door for new innovative approaches for merging the different data sets in order to improve the overall product accuracy and the spatio-temporal coverage (LIU et al., 2011).
• Considering the initial challenges when starting to use ASCAT soil moisture data in a particular application, the progress made in the various application domains is very promising.

Figures

Figure 2: Imaging geometry of ASCAT.

Figure 4: Illustration of the r0(h) dependency on vegetation and soil moisture. Adapted from WAGNER et al. (1999a).

Figure 7: Comparison of ASCAT Soil Water Index (SWI) and Antecedent Precipitation Index (API) time series to precipitation (top) and in-situ (IN SITU) and modeled (MOD) soil moisture data over the Bibeschbach catchment in Luxemburg for the years 2007 and 2008.

Figure 9: Spatial error field of the ASCAT soil moisture product obtained by applying the triple collocation technique to ASCAT, AMSR-E and GLDAS soil moisture data. The figure is based upon methods presented in DORIGO et al. (2010).

Figure 14: SWI anomaly within the Indus river basin (a) and a map of the population affected by the 2010 Pakistan Floods (b). From FLEISS et al. (2011).

Figure 12: Correlation between monthly SWI and NDVI time series over Africa for all land cover classes (left) and for cropping areas only (right).

Figure 8: Comparison of ASCAT surface soil moisture time series with 5 cm in situ measurements acquired at one station (Y3) of the OZNET network located in south eastern Australia (RÜDIGER et al., 2007), and modeled 10 cm soil moisture data extracted from the Global Land Data Assimilation System (GLDAS) for the year 2010. In this example, both the ASCAT and GLDAS data were scaled to match the absolute value and dynamic range of the in situ measurements.

Table 1: Summary of the product properties for WARP and WARP NRT as produced and distributed by TU Wien and EUMETSAT. In the future these products will also distributed by the H-SAF (http://hsaf.meteoam.it/).

Figure 5: Overview of the data flow for WARP and WARP NRT.

Figure 1: Daily global coverage achieved by the ASCAT instrument over land with only METOP-A in orbit (a) and with METOP-A and METOP-B in orbit (b).

Figure 11: Top: Antecedent rainfall during the past 12 hours, Centre: remotely sensed ASCAT surface soil moisture, Bottom: Simulated skin soil moisture in the Kamp catchment (1550 km2). Left: 22 June 2009 immediately before a 5 year flood. Right: 23 June 2009 during the flood.

Figure 6: Spatial error field of the ASCAT soil moisture product obtained from error propagation. The figure is based upon the methods presented in NAEIMI et al. (2009b).

Figure 13: Mean daily basic reproduction number R0 for June to September 2010, depicting the spatio-temporal distribution of the potential risk for a bluetongue virus (BTV) outbreak in Austria. Green colours depict regions at minor risk, red at moderate risk and dark red at high risk. The location of the first BTV cases confirmed in Austria is marked by a white asterisk.

Figure 3: Illustration of the interaction of radar pulses with a vegetated surface: a) surface scattering from the ground; b) volume scattering in canopies; c) ‘‘multi-bounce’’ effects between vegetation and ground.

Figure 10: A (left) and S (right) score for ALADIN precipitation forecasts without (OPER) and with (EXP1-4) assimilation of soil moisture in Austria during July 2009. Verification is separated for a flatland (F) and an Alpine (A) region as well as for precipitation events with different mean areal precipitation, analyzed by INCA (abscissa). The ordinate is indicating values for A and S, where 0 indicates a perfect forecast.

Full text

eschweizerbart_xxx
The ASCAT Soil Moisture Product: A Review of its
Specifications, Validation Results, and Emerging Applications
Wolfgang Wagner
1,
*
, Sebastian Hahn
1
, Richard Kidd
1
, Thomas Melzer
1
, Zoltan Bartalis
2
,
Stefan Hasenauer
1
, Julia Figa-Saldan
˜
a
3
, Patricia de Rosnay
4
, Alexander Jann
5
,
Stefan Schneider
5
,Ju
¨
rgen Komma
6
, Gerhard Kubu
7
, Katharina Brugger
8
,
Christoph Aubrecht
9
, Johann Zu
¨
ger
9
, Ute Gangkofner
10
, Stefan Kienberger
11
,
Luca Brocca
12
, Yong Wang
5
,Gu
¨
nter Blo
¨
schl
6
, Josef Eitzinger
7
, Klaus Steinnocher
9
,
Peter Zeil
11
and Franz Rubel
8
1
Department of Geodesy and Geoinformation, Vienna University of Technology, Vienna, Austria
2
ESA, ESRIN, Frascati, Italy
3
EUMETSAT, Darmstadt, Germany
4
ECMWF, Reading, United Kingdom
5
ZAMG, Vienna, Austria
6
Institute of Hydraulic and Water Resources Engineering, Vienna University of Technology, Vienna, Austria
7
Institute of Meteorology, University of Natural Resources and Applied Life Sciences, Vienna, Austria
8
Institute for Veterinary Public Health, University of Veterinary Medicine, Vienna, Austria
9
AIT Austrian Institute of Technology GmbH, Vienna, Austria
10
GeoVille, Innsbruck, Austria
11
Centre for Geoinformatics, University of Salzburg, Salzburg, Austria
12
Research Institute for Geo-Hydrological Protection, National Research Council, Perugia, Italy
(Manuscript received May 5, 2012; in revised form November 21, 2012; accepted February, 2013)
Abstract
Many physical, chemical and biological processes taking place at the land surface are strongly influenced by the
amount of water stored within the upper soil layers. Therefore, many scientific disciplines require soil moisture
observations for developing, evaluating and improving their models. One of these disciplines is meteorology
where soil moisture is important due to its control on the exchange of heat and water between the soil and the
lower atmosphere. Soil moisture observations may thus help to improve the forecasts of air temperature, air
humidity and precipitation. However, until recently, soil moisture observations had only been available over a
limited number of regional soil moisture networks. This has hampered scientific progress as regards the
characterisation of land surface processes not just in meteorology but many other scientific disciplines as well.
Fortunately, in recent years, satellite soil moisture data have increasingly become available. One of the freely
available global soil moisture data sets is derived from the backscatter measurements acquired by the Advanced
Scatterometer (ASCAT) that is a C-band active microwave remote sensing instrument flown on board of the
Meteorological Operational (METOP) satellite series. ASCAT was designed to observe wind speed and
direction over the oceans and was initially not foreseen for monitoring soil moisture over land. Yet, as argued in
this review paper, the characteristics of the ASCAT instrument, most importantly its wavelength (5.7 cm), its
high radiometric accuracy, and its multiple-viewing capabilities make it an attractive sensor for measuring soil
moisture. Moreover, given the operational status of ASCAT, and its promising long-term prospects, many
geoscientific applications might benefit from using ASCAT soil moisture data. Nonetheless, the ASCAT soil
moisture product is relatively complex, requiring a good understanding of its properties before it can be
successfully used in applications. To provide a comprehensive overview of the major characteristics and caveats
of the ASCAT soil moisture product, this paper describes the ASCAT instrument and the soil moisture processor
and near-real-time distribution service implemented by the European Organisation for the Exploitation of
Meteorological Satellites (EUMETSAT). A review of the most recent validation studies shows that the quality of
ASCAT soil moisture product is – with the exception of arid environments –comparable to, and over some
regions (e.g. Europe) even better than currently available soil moisture data derived from passive microwave
sensors. Further, a review of applications studies shows that the use of the ASCAT soil moisture product is
particularly advanced in the fields of numerical weather prediction and hydrologic modelling. But also in other
application areas such as yield monitoring, epidemiologic modelling, or societal risks assessment some first
progress can be noted. Considering the generally positive evaluation results, it is expected that the ASCAT soil
moisture product will increasingly be used by a growing number of rather diverse land applications.
Keywords: Soil moisture, earth observation, scatterometer, hydrometeorological applications, accuracy
assessment.
*
Corresponding author: Wolfgang Wagner, Department of Geodesy and Geoinformation, Vienna University of Technology, Vienna, Austria, e-mail:
wolfgang.wagner@geo.tuwien.ac.at
Meteorologische Zeitschrift, Vol. 22, No. 1, 5–33 (February 2013)
Open Access Article
Ó by Gebru¨der Borntraeger 2013
DOI 10.1127/0941-2948/2013/0399
0941-2948/2013/0399 $ 13.05
Ó Gebru¨der Borntraeger, Stuttgart 2013

eschweizerbart_xxx
1 Introduction
The Advanced Scatterometer (ASCAT) is an active
microwave remote sensing instrument that was designed
for monitoring of winds over the oceans in support to
operational applications such as numerical weather pre-
diction (NWP), tropical cyclone analysis, and ocean
waves forecasting (I
SAKSEN and ST OFFELEN, 2000;
F
IGA-SALDAN
˜
A et al., 2002; LIU, 2002). Over land, no
operational services were initially foreseen. However ,
research carried out with its predecessor instrument, the
ERS-1/2 scatterometer (ESCAT), provided increasing
evidence that ASCAT might also be used for monitoring
of soil moisture (P
ULLIAINEN et al., 1998; WAGNER
et al., 1999c; WEN uND SU,2003; WAGNER et al.,
2007a) even though it was not clear at that time whether
soil moisture products derived from these instruments are
able to meet the accuracy requirements of potential appli-
cations. The main concern was, and to some extent still
is, that ESCAT and ASCAT are operated at a wavelength
of 5.7 cm (C-Band) which has often been stated to be
sub-optimal for the task of soil moisture retrieval due
to a reduced sensitivity to soil moisture in the presence
of vegetation compared to longer wavelengths such as
L-Band (K
ERR, 2007). It has however been overlooked
that ESCAT, and even more so ASCAT, are well-cali-
brated instruments with a high radiometric accuracy. In
other words, while they offer a somewhat reduced sensi-
tivity to soil moisture compared to L-Band instruments,
their signal-to-noise ratio may still suffice to achieve an
acceptable retrieval accuracy. And indeed, many of the
initial ESCAT validation studies carried out by indepen-
dent research teams unexpectedly found quite encourag-
ing results (P
ELLARIN et al., 2006; BROCCA et al., 2009;
R
U
¨
DIGER et al., 2009).
The first user community to take note of the opportu-
nities offered by ESCAT and ASCAT was the NWP com-
munity. There are probably two reasons for this: Firstly,
the importance of soil moisture for modelling land-atmo-
sphere interactions had increasingly been recognised
since the 1990s (Z
HENG and ELTAHIR, 1998; KOSTER
et al., 2004), prompting much research in the NWP com-
munity to improve the representation of soil moisture
processes over the last decade (J
EREZ et al., 2010;
B
ARTHLOTT et al., 2011). Secondly, the spatial resolution
of ESCAT and ASCAT, which is in the order of tens of
kilometres (25-50 km), is commensurate with the
requirements of NWP models, while e.g. hydrological
models run on much finer spatial grids. The interest of
several European NWP centres in a potential ASCAT soil
moisture product led to the decision of the European
Organisation for the Exploitation of Meteorological Sat-
ellites (EUMETSAT) to develop an operational global
ASCAT soil moisture processing and dissemination ser-
vice (B
ARTALIS et al., 2007; WAGNER et al., 2010). EU-
METSAT implemented this service in cooperation with
the Vienna University of Technology (TU Wien) and
put it into operations in 2008, roughly two years after
the launch of METOP-A, the first satellite to carry
ASCAT. To meet the requirements of the NWP commu-
nity, this service is being operated in near -real-time, i.e.
the ASCAT soil moisture data are continuously being
processed and distributed worldwide within 135 minutes
after data acquisition. This allows the NWP centres to
assimilate the ASCAT soil moisture data in their opera-
tional forecasts.
Up until now the work of the NWP centres with the
ASCAT soil moisture product has concentrated on valida-
tion activities, quality assessments and scientific studies.
But thanks to quite positive outcomes from several data
assimilation experiments some NWP centres have
already started using the ASCAT soil moisture product
in an operational fashion (D
HARSSI et al., 2011; DE
ROSNAY et al., in press a). This shows that even though
much more scientific work is still required to characterise
the spatio-temporal accuracy of the retrievals, the
ASCAT soil moisture product starts having a positive
impact in applications. Yet, being relatively new and
the first-of-its-kind, the ASCAT soil moisture service is
not yet widely known to a broader research community.
Given the central and unifying role of soil moisture in
understanding atmospheric, hydrologic, biologic, and
geomorphic processes and their interactions (L
EGATES
et al., 201 1) it has however also a significant potential
in many other applications. This review paper was thus
written with a view on new potential applications, dis-
cussing the strengths, limitations, and potential applica-
tions of the ASCAT soil moisture product from a user’ s
perspective.
2 Mission specifications
The ASCAT soil moisture service owes several of its
attractive features to the long and successful heritage of
space borne ocean wind vector monitoring programmes.
In particular Europe can look back to a series of success-
ful scatterometer missions, starting with the ERS satellite
programme operated by the European Space Agency
(ESA), and continuing to the on-going Meteorological
Operational (METOP) satellite programme operated by
EUMETSAT. The high continuity provided by these
European satellite programmes (Section 2. 1), in combi-
nation with the strong heritage in the sensor design from
one instrument generation to the next (Section 2.2), is the
basis for the continuity, reliability and promising long-
term prospects of the ASCAT soil moisture service.
2.1 Satellite programmes
The first European scatterometer (ESCAT) was the one
flown on board of the two European Remote Sensing
Satellites ERS-1 and ERS-2. Together, these two satel-
lites provided scatterometer measurements for a period
of 20 years (1991–2011). ASCAT itself is being flown
on a series of three polar orbiting Meteorological
6
W. Wagner et al.: The ASCAT Soil Moisture Product Meteorol. Z., 22, 2013

eschweizerbart_xxx
Operational (METOP) satellites, whereas the first satellite
(METOP-A) was launched in October 2006 and the sec-
ond (METOP-B) in September 2012. METOP-B will be
flown in parallel to METOP-A in a dual satellite constel-
lation to improve the spatio-temporal coverage of their
sensors. Together with METOP-C, which is scheduled
for launch in 2018, the three satellites can be expected
to provide an uninterrupted stream of ASCAT backscatter
observations well into the 2020s. Even for the successor
instrument of ASCAT, which will be flown on board of
one of the Second Generation (SG) satellites of the EU-
METSAT Polar System, plans are already well advanced
(L
IN et al., 2012).
Similar to the two ERS satellites, METOP flies in a
near-polar sun-synchronous orbit at an altitude of about
817 km with a repeat cycle of 29 days. In this orbit the
METOP circles the earth within about 100 min, which
means that the satellite completes 14 orbits per day.
The equator crossing times are at 9:30 for the descending
pass and 21:30 for the ascending pass, meaning that for
all equatorial and mid-latitude regions ASCAT data
acquisitions take place at around 9:30 respectively
21:30 local time (±1 hour). Considering the two
550 km wide swaths of ASCAT as described in the next
section, the daily global coverage with one METOP
satellite is about 82%. As one can see in Fig. 1a,which
shows the daily global coverage achieved by one satellite
(e.g. METOP-A), the gaps in coverage are largest near
the equator , while at higher latitudes full daily coverage
is achieved over the two poles (>65°) and in the latitudi-
nal belt between about 35° and 55°. In the latitudinal belt
between 55° to 65° there are some gaps in the daily cov-
erage maps, but nevertheless, on average, the number of
acquisitions per day is higher than one for all regions
north of 40°. With two satellites in orbit, the gaps
between 55° to 65° disappear , but near the equator they
are still present (Fig. 1b).
A consequence of the irregular spatial coverage
achieved by either the one or two satellite constellation
is that also the temporal coverage is highly irregular:
Sometimes two (or near the poles even more) acquisi-
tions are taken on a single day over a selected region
of interest, but on other days no data are being acquired
at all. This is an important constraint in using the ASCAT
soil moisture data, because applications need to be devel-
opedinsuchawayastocopewiththehighlyirregular
coverage, or to settle for using interpolated (and thus
more uncertain) measurements.
2.2 Instrument
ASCAT is a fixed fan-beam scatterometer which uses six
side-ways looking antennas to illuminate two 550 km
wide swaths to each side of the satellite track (Fig. 2).
It is operated at a frequency of 5.3 GHz (C-band) in
VV polarisation, i.e. it both transmits and receives
electromagnetic waves in vertical polarisation only (ver-
tical polarisation means that the electric field vector ,
which defines the polarisation of the electromagnetic
wave, has a vertical component relative to the earth’s sur-
face). After reception, the backscatter echoes are ampli-
fied and further processed for echo power detection.
The echo power measurements are then used as input into
the radar equation to calculate the so-called backscatter -
ing coefficient r
0
, given in units of m
2
m
2
or , more com-
monly, in decibels (dB). Simply put, r
0
is a measure of
the electromagnetic energy intercepted and reradiated at
the same wavelengths by an areal unit of the Earth’s land
surface. The nominal spatial resolution of the ASCAT
backscatter measurements is 50 km, but a higher resolu-
tion r
0
product with about 25 km is also available (the
resolution of the higher resolution product varies some-
what across the swath from 25 km to 34 km). Complying
with the Nyquist–Shannon sampling theorem, the grid
spacing of the 50 km product is 25 km, and 12.5 km
for the 25 km product.
The technical specifications of ASCAT make it a suit-
able sensor for soil moisture retrieval for several reasons.
First of all, its operation frequencies of 5.3 GHz is within
the range of microwave frequencies (< 10 GHz) where
the addition of liquid water to soil strongly increases
the soil dielectric constant (approximately tenfold from
dry to wet soils). Therefore, when the soil moisture
content increases, so does the dielectric constant at the
air-soil boundary and thus backscatter. This strong depen-
dence of the backscattering intensity on the soil moisture
content implies that ASCAT r
0
measurements provide a
relatively direct measure of the soil moisture content over
bare soils. In the presence of vegetation, the response of
r
0
to changes in the soil moisture content is dampened,
making it important to correctly model the effect of veg-
etation on the overall backscatter. Of course, also surface
roughness has an important effect on the r
0
measure-
ments and needs to be corrected for (Section 3.1). Other
favourable technical specifications of ASCAT are:
d ASCAT backscatter measurements are well calibrated
and very stable over time (W
ILSON et al., 2010).
According to A
NDERSON et al. (2012) the worst-case
calibration error is 0.15-0.25 dB and annual changes
are in the order of 0.02 dB. This means that ASCAT is
very well suited for tracking changes in soil moisture.
d With its three antennas for each swath, ASCAT takes
for each pixel three independent and quasi instanta-
neous r
0
measurements at three different azimuth
angles and two different incidence angles. Particularly
the last feature is important because the incidence
angle behaviour of r
0
is an important indicator for
the vegetation density, and can hence be exploited
for correcting vegetation effects in the soil moisture
retrieval.
d Its spatio-temporal sampling properties allow captur-
ing the large-scale soil moisture patterns driven by
atmospheric processes (precipitation, evapotranspira-
tion) quite well (V
INNIKOV et al., 1999).
Meteorol. Z., 22, 2013 W. Wagner et al.: The ASCAT Soil Moisture Product 7

eschweizerbart_xxx
3 Product specifications
3.1 Physical basis
The physical basis for the capability of ASCAT to mea-
sure soil moisture is the strong dependence of C-band
backscatter on the soil moisture content in the top soil
layer (usually held to be 1-2 cm thin). But besides soil
moisture, also surface roughness has a significant effect.
There are several semi-empirical and theoretical models
for describing backscatter from a rough soil surface,
but unfortunately, the correct characterisation of the
roughness of natural surfaces still poses significant chal-
lenges (V
ERHOEST et al., 2008). Therefore, for the
ASCAT soil moisture retrieval a change detection
approach has been adopted which circumvents the prob-
lems of surface roughness parameterisation by just inter-
preting changes in r
0
over time.
In addition to surface roughness, also vegetation has a
significant effect on r
0
. As illustrated in Fig. 3 total back-
scatter from vegetation is composed of several contribu-
tions including surface backscatter from underlying
ground (subject to attenuation in the canopy), canopy
volume scattering, multiple path interactions between
canopy and ground and double-bounce configurations
between tree trunks and ground (so-called corner reflec-
tors, a multiple-bounce over perpendicularly oriented sur-
faces returning the backscatter in its incident direction).
Vegetation moisture content and geometric structure are
thus key factors for the backscatter, especially since most
structural elements of forests, shrubs etc. are comparable
in size with typical microwave wavelengths (1-25 cm).
Dense forests and shrubs are usually opaque to C-band
radar , while s parse forest, grassland and agricultural
crops are partly transparent. This has e.g. been demon-
strated by experimental studies using range-resolving
Figure 1: Daily global coverage achieved by the ASCAT instrument over land with only METOP-A in orbit (a) and with METOP-A and
METOP-B in orbit (b).
8 W. Wagner et al.: The ASCAT Soil Moisture Product Meteorol. Z., 22, 2013

eschweizerbart_xxx
radar systems which recorded significant soil responses at
C-band frequencies over these latter vegetation types
even in cases when some of the theoretical backscatter
models would not have predicted this (P
ULLIAINEN
et al., 1996; BROWN et al., 2003). This was especially
the case when the radar echoes were observed at lower
incidence angles.
One of the most important models to describe back-
scatter from vegetation is the so-called Cloud Model,
where vegetation is modelled as one or several layers
(clouds) of water over a surface (A
TTEMA and ULABY,
1978). The droplets of the clouds are randomly located
and considered to be held in place by the vegetative mat-
ter. Because of the complexity it would add to the model,
multiple scattering is usually not considered. Due to the
aforementioned corner reflection mechanisms and possi-
ble resonant scattering, even the more complex versions
of the Cloud Model remain usually just coarse approxi-
mations of the observed phenomena. Most soil moisture
retrievals algorithms developed for ESCAT and ASCAT
make use of the Cloud Model formulation or variants
thereof (M
AGAGI and KERR, 1997; PULLIAINEN et al.,
Figure 2: Imaging geometry of ASCAT.
Figure 3: Illustration of the interaction of radar pulses with a vegetated surface: a) surface scattering from the ground; b) volume scattering
in canopies; c) ‘‘multi-bounce’’ effects between vegetation and ground.
Meteorol. Z., 22, 2013 W. Wagner et al.: The ASCAT Soil Moisture Product
9

Frequently asked questions

Q1. What have the authors contributed in "The ascat soil moisture product: a review of its specifications, validation results, and emerging applications" ?

Yet, as argued in this review paper, the characteristics of the ASCAT instrument, most importantly its wavelength ( 5. 7 cm ), its high radiometric accuracy, and its multiple-viewing capabilities make it an attractive sensor for measuring soil moisture. To provide a comprehensive overview of themajor characteristics and caveats of the ASCATsoil moisture product, this paper describes the ASCAT instrument and the soil moisture processor and near-real-time distribution service implemented by the European Organisation for the Exploitation of Meteorological Satellites ( EUMETSAT ). Moreover, given the operational status of ASCAT, and its promising long-term prospects, many geoscientific applications might benefit from using ASCAT soil moisture data. Further, a review of applications studies shows that the use of the ASCAT soil moisture product is particularly advanced in the fields of numerical weather prediction and hydrologic modelling. 

Q2. What are the future works mentioned in the paper "The ascat soil moisture product: a review of its specifications, validation results, and emerging applications" ?

This question needs to be addressed in future studies that analyse and compare the end-to-end error budgets of ASCAT, AMSR-E, SMOS, SMAP and other Meteorol. In other application domains such as vegetation and crop yield monitoring, epidemic risk modelling and societal risks assessments some first encouraging results have been obtained, but much further work is required to optimally use the information provided by ASCAT. A better understanding of sensor performances will also open the door for new innovative approaches for merging the different data sets in order to improve the overall product accuracy and the spatio-temporal coverage ( LIU et al., 2011 ). Mechanistically, their result can be explained by enhanced moist convection over dry soils and/or meso-scale variability in soil moisture, yet this negative soil moisture feedback was not correctly modelled by six state-of-the-art global weather and climate models. 

Q3. What is the role of soil moisture in the forecasting of floods?

As it plays an important role in partitioning rainfall into runoff and infiltration, soil moisture is one of the key variables in flood forecasting models. 

Q4. What are the components of the backscatter from snow?

Backscatter from snow is often considered to consist of three components: scattering from the top snow surface, the underlying ground surface and the volume scattering from within the snow pack (ULABY et al., 1986; FUNG, 1994). 

Q5. Why is the soil moisture variability difficult to characterise using in-situ measurements and?

because soil moisture may vary strongly within meters due to variable soil properties, vegetation, and fine-scale topography, spatial soil moisture patterns are difficult to characterise using in-situ measurements and soil maps. 

Q6. What is the promising method for estimating crop yield over regions?

The most promising method for estimating crop yield over regions more accurately is therefore to combine ecosystem models and remote sensing data (DE WIT and VAN DIEPEN, 2007; VERSTRAETEN et al., 2010). 

Q7. Why is the soil moisture data difficult to match with independent reference?

soil moisture is highly variable in space and time (WESTERN et al., 2002), making it very difficult to match the intermittent and spatially irregular satellite measurements with independent reference data. 

Q8. What is the main reason why the use of a new data type is often difficult?

The use of a new data type in applications is usually very challenging, simply because models are built around input data that were available at the time when the models were developed. 

Q9. What is the effect of the local bias correction on the forecasts of relative humidity at 2m?

forecasts of relative humidity at 2m can be improved due to the assimilation during the first six hours of the model run, and overall, forecasts tend to be cooler and moister when assimilating soil moisture in comparison to Austrian SYNOP stations which has a positive impact on model bias during night-time. 

Q10. What are the physical parameters that determine the exact scattering behaviour of snow?

The exact scattering behaviour depends on several physical parameters of the snow layer, including the liquid water content, the roughness of the air-snow interface, the layering of the snow pack, and the grain size and shape. 

Q11. What is the need for a range of value added soil moisture product?

As the requirements of different applications may vary significantly, there is a need to combine the original ASCAT satellite retrievals with auxiliary data to produce a range of value added soil moisture product. 

Q12. What is the effect of the sEKF on the soil moisture distribution in lowlands?

This leads to the conclusion that in mountainous regions, orographic features are playing an important role in the localisation of convective initiation, while in lowlands the more stochastic nature of initiation is benefitting from the improved soil moisture distribution in the ground. 

Q13. What is the physical basis for the ability of ASCAT to measure soil moisture?

The physical basis for the capability of ASCAT to measure soil moisture is the strong dependence of C-band backscatter on the soil moisture content in the top soil layer (usually held to be 1-2 cm thin). 

Q14. What is the way to disaggregate the coarse scale microwave measurements?

To disaggregate coarse scale microwave measurements they are usually combined with finer resolution satellite data acquired either by synthetic aperture radars (DAS et al., 2011) or visible/infrared imagers (PILES et al., 2011). 

Q15. What makes ASCAT a suitable sensor for soil moisture monitoring?

In particular, this review highlighted the important role of other sensor characteristics – most importantly radiometric accuracy, multiple-viewing capabilities and spatio-temporal coverage – that make ASCAT a suitable sensor for soil moisture monitoring.