The CO2MVS Research on Supplementary Observations (CORSO) project aims at reducing the uncertainties in the predicted biogenic carbon fluxes by leveraging new types of satellite observations, machine learning and data assimilation. The objective of this work is to implement the assimilation of active microwave and solar-induced chlorophyll fluorescence (SIF) satellite observations in the Integrated Forecast System (IFS) developed at the European Centre for Medium-Range Weather Forecasts (ECMWF) to jointly analyze soil moisture and leaf area index (LAI). The assimilation of observation requires the development of an observation operator to predict the model-simulated counterpart of the observation from the model fields. In this paper, we present the development of machine learning-based observation operators for the normalized backscatter at 40° from the Advanced Scatterometer (ASCAT) instrument and SIF derived from the TROPOspheric Monitoring Instrument (TROPOMI). Results show that deep learning methods provide accurate predictions of the satellite signals from the IFS model fields. The next step will be the implementation of these operators in the IFS and the evaluation of the impacts of analyzing both soil moisture and vegetation variables on the estimation of carbon and water fluxes along with near-surface meteorological variables.

Recent applications of passive L-band observations from space are summarized for ocean, land surface and cryosphere applications. The main limitation of the measurements performed by the current generation of sensors is the spatial resolution. The need of a mission ensuring the continuation of L-band measurements from space with high spatial resolution (10-15 km) is discussed.

Radio Frequency Interference (RFI) is a growing threat to all Earth Observation (EO) passive microwave missions. Many RFI detection algorithms are used on-board and in the ground segment data processing, but there is no single algorithm that can detect all RFI instances. The best strategy is always to combine several detection methods. This paper presents the results of the new Ground RFI Detection System (GRDS). The GRDS uses a combination of a wide variety of RFI detection algorithms to clean the Earth Observation measurements from RFI. The system is built to be able to scan for RFI for any EO mission. The initial results show the important reduction on RFI in the EO data as measured by the European Center for Mid-Range Weather Forecast (ECMWF) first guess departure statistics.

In this paper we present L-band data usage for Numerical Weather Prediction applications and Emergency Services at the European Centre for Medium-Range Weather Forecasts (ECMWF).

Radio Frequency Interference (RFI) is a serious constraint affecting the Earth Observation (EO) passive microwave missions whose effect has been increasing. There is no single algorithm that can detect all RFI instances. The best strategy always is to combine several detection methods. This paper presents a new system concept: Ground RFI Detection System (GRDS). The GRDS uses a wide variety of RFI detection algorithm to clean the Earth Observation measurements from RFI. The system architecture is modular and it is built to be able to scan for RFI for any EO mission. The initial assessments reported show the potential for this GRDS system.

A set of Soil Moisture and Ocean Salinity (SMOS) soil moisture (SM) data assimilation (DA) experiments are presented. The SMOS soil moisture dataset used in this study was produced training a neural network (NN) using SMOS brightness temperatures as input and ECMWF H-TESSEL SM fields as reference for the training. The DA experiments are computed using a surface-only Land Data Assimilation System (so-LDAS) based on the HTESSEL land surface model. SMOS NN SM DA experiments were compared to Advanced Scat-terometer (ASCAT) SM DA. In both cases, experiments with and without 2 metre air temperature and relative humidity DA are discussed. The different SM analysed fields are evaluated against a large number of in situ measurements of SM. On average, the SM analysis gives similar results to the model open loop with no assimilation. The effect of the soil moisture analysis on the Numerical Weather Prediction (NWP) was evaluated using the analysed surface fields to perform atmospheric forecast experiments. In the Northern Hemisphere both with ASCAT and SMOS, the experiments using 2m air temperature and relative humidity improve the forecast in April-September. SMOS alone has a significant positive effect in July-September. Maps of the forecast skill with respect to the open loop experiment show that SMOS improves the forecast in North America and to a lesser extent in Northern Asia for up to 72 hours.

This paper presents the Soil Moisture and Ocean Salinity (SMOS) mission data assimilation activities conducted at the European Centre for Medium-Range Weather Forecasts (ECMWF) to analyse soil moisture for Numerical Weather Prediction (NWP) applications. Two different approaches are presented based on SMOS brightness temperature and SMOS neural network soil moisture data assimilation, respectively. For the first approach, SMOS brightness temperature data assimilation relies on forward modelling. Long term results, spanning the SMOS period, of SMOS forward modelling, monitoring and data assimilation are presented. They emphasize the relevance of SMOS data for monitoring and to support NWP model developments. For the second approach, a SMOS soil moisture product has been produced based on a Neural Network (NN) trained on ECMWF soil moisture. So, the SMOS-ECMWF NN soil moisture product captures the SMOS signal variability in time and space, while by design its climatology is consistent with that of the ECMWF soil moisture, which makes it suitable for data assimilation purpose. This approach, initially tested for 2012 in a global scale stand alone approach, shows that SMOS NN data assimilation slightly improves the two-metre air temperature forecast in the short range at regional scale. For NWP applications this approach has been further developed with a near real time production of the SMOS-ECMWF NN soil moisture product, with the implementation of the SMOS NN data assimilation in the ECMWF Integrated Forecasting System (IFS), and with high resolution (9km) global scale testing compatible with the current ECMWF NWP system.

This study investigated the use of physically based statistical regressions to retrieve a global and long term (e.g. 2003–2014) surface soil moisture (SSM) record based on a combination of passive microwave remote sensing observations from the Advanced Microwave Scanning Radiometer (AMSR-E; 2003-Sept. 2011) and the Soil Moisture and Ocean Salinity (SMOS; 2010–2014) sensors. Statistical regression methods based on bi-polarization (horizontal and vertical) brightness temperatures (Tb) observations obtained from AMSR-E. The coefficients of these regression equations were calibrated using SMOS level 3 SSM maps (SMOSL3) as a reference. This calibration process was carried out over the June 2010-Sept. 2011 period, over which both SMOS and AMSR-E observations coincide. Based on these calibrated coefficients global SSM maps could be computed from the AMSR-E Tb observations over the whole 2003–2011 period. In this study, the SSM maps were successfully evaluated against the SMOSL3 SSM products over the period of calibration (Jun. 2010-Sept. 2011). Correlations (R) and Root Mean Square Error (RMSE) were computed between the AMSR-E retrievals and the reference (SMOSL3) SSM products. The R (mostly > 0.75) and RMSE (mostly3/m3) maps showed a good agreement between the retrieved and SMOSL3 SSM products particularly over Australia, central USA, central Asia, and the Sahel. In conclusion, the statistical regression method is capable of retrieving a coherent "SMOS-AMSR-E" SSM time series for the period 2003–2014.

The 2-D interferometric radiometer on board the Soil Moisture and Ocean Salinity (SMOS) satellite has been providing a continuous data set of brightness temperatures, at different viewing geometries, containing information of the Earth's surface microwave emission. This data set is affected by several sources of noise, which are a combination of the noise associated with the radiometer itself and the different views under which a heterogeneous target, such as continental surfaces, is observed. As a result, the SMOS data set is affected by a significant amount of noise. For many applications, such as soil moisture retrieval, reducing noise from the observations while keeping the signal is necessary, and the accuracy of the retrievals depends on the quality of the observed data set. This paper investigates the averaging of SMOS brightness temperatures in angular bins of different sizes as a simple method to reduce noise. All the observations belonging to a single pixel and satellite overpass were fitted to a polynomial regression model, with the objective of characterizing and evaluating the associated noise. Then, the observations were averaged in angular bins of different sizes, and the potential benefit of this process to reduce noise from the data was quantified. It was found that, if a 2° angular bin is used to average the data, the noise is reduced by up to 3 K. Furthermore, this method complements necessary data thinning approaches when a large volume of data is used in data assimilation systems.

In the above paper (ibid., vol. 49, no. 4, pp. 1177-1189, Apr. 2011), there is an error in equation (7). The explanation and corrected equation are presented here.

The launch of the Soil Moisture and Ocean Salinity (SMOS) satellite of the European Space Agency opens the way to using a new type of satellite data that are very sensitive to soil moisture for numerical weather prediction. The European Centre for Medium-Range Weather Forecasts (ECMWF) has developed an operational chain which makes it possible to process SMOS data in near real time (NRT) and compare it with a model equivalent. This process has been very challenging. The main reasons are the particular characteristics of the SMOS observation system and the large volume of data. Despite these obstacles, SMOS data are being processed successfully in NRT within the ECMWF Integrated Forecasting System (IFS). The ultimate objective is to assimilate these data in the IFS. It is expected to have an impact on the weather forecast at short and medium ranges. Prior to assimilation experiments, the quality of the data has to be assessed. This can be done through monitoring activities. Monitoring is a routine task performed with all satellite data, and among other things, it makes it possible to localize temporal (or spatial) bias or drifts in the data, thus providing NRT reports to the calibration and validation teams, which can act accordingly. In this letter, the implementation of SMOS data in the ECMWF IFS for monitoring purposes is discussed. The system was developed using a simulated file for the NRT processor, and it was tested using real data from the first year since the launch date.

The present paper analyzes the effects of roughness on the surface emission at L-band based on observations acquired during a long-term experiment. At the Surface Monitoring of the Soil Reservoir Experiment site near Toulouse, France, a bare soil was plowed and monitored over more than a year by means of an L-band radiometer, profile soil moisture and temperature sensors, and a local weather station, accompanied by 12 roughness campaigns. The aims of this paper are the following: 1) to present this unique database and 2) to use this data set to investigate the semiempirical parameters for the roughness in L-band Microwave Emission of the Biosphere, which is the forward model used in the Soil Moisture and Ocean Salinity soil moisture retrieval algorithm. In particular, we studied the link between these semiempirical parameters and the soil roughness characteristics expressed in terms of standard deviation of surface height (σ) and the correlation length (LC). The data set verifies that roughness effects decrease the sensitivity of surface emission to soil moisture, an effect which is most pronounced at high incidence angles and soil moisture and at horizontal polarization. Contradictory to previous studies, the semiempirical parameter Qr was not found to be equal to 0 for rough conditions. A linear relationship between the semiempirical parametersNand σ was established, while NHand NVappeared to be lower for a rough (NH~ 0.59 and NV~ -0.3) than for a quasi-smooth surface. This paper reveals the complexity of roughness effects and demonstrates the great value of a sound long-term data set of rough L-band surface emissions to improve our understanding on the matter.

In the forward model [L-band microwave emission of the biosphere (L-MEB)] used in the Soil Moisture and Ocean Salinity level-2 retrieval algorithm, modeling of the roughness effects is based on a simple semiempirical approach using three main “roughness” model parameters:$H_{R}$,$Q_{R}$, and$N_{R}$. In many studies, the two parameters$Q_{R}$and$N_{R}$are set to zero. However, recent results in the literature showed that this is too approximate to accurately simulate the microwave emission of the rough soil surfaces at L-band. To investigate this, a reanalysis of the PORTOS-93 data set was carried out in this paper, considering a large range of roughness conditions. First, the results confirmed that$Q_{R}$could be set to zero. Second, a refinement of the L-MEB soil model, considering values of$N_{R}$for both polarizations (namely,$N_{\rm RV}$and$N_{\rm RH}$), improved the model accuracy. Furthermore, simple calibrations relating the retrieved values of the roughness model parameters$H_{R}$and$(N_{\rm RH} - N_{\rm RV})$to the standard deviation of the surface height were developed. This new calibration of L-MEB provided a good accuracy (better than 5 K) over a large range of soil roughness and moisture conditions of the PORTOS-93 data set. Conversely, the calibrations of the roughness effects based on the Choudhury approach, which is still widely used, provided unrealistic values of surface emissivities for medium or large roughness conditions.

The emission of bare soils at microwave L-band (1-2 GHz) frequencies is known to be correlated with surface soil moisture. Roughness plays an important role in determining soil emissivity although it is not clear which roughness length scales are most relevant. Small-scale (i.e., smaller than the resolution limit) inhomogeneities across the soil surface and with soil depth caused by both spatially varying soil properties and topographic features may affect soil emissivity. In this paper, roughness effects were investigated by comparing measured brightness temperatures of well-characterized bare soil surfaces with the results from two reflectivity models. The selected models are the air-to-soil transition model and Shi's parameterization of the integral equation model (IEM). The experimental data taken from the Surface Monitoring of the Soil Reservoir Experiment (SMOSREX) consist of surface profiles, soil permittivities and temperatures, and brightness temperatures at 1.4 GHz with horizontal and vertical polarizations. The types of correlation functions of the rough surfaces were investigated as required to evaluate Shi's parameterization of the IEM. The correlation functions were found to be clearly more exponential than Gaussian. Over the experimental period, the diurnal mean root mean square (rms) height decreased, while the correlation length and the type of correlation function did not change. Comparing the reflectivity models with respect to their sensitivities to the surface rms height and correlation length revealed distinct differences. Modeled reflectivities were tested against reflectivities derived from measured brightness, which showed that the two models perform differently depending on the polarization and the observation angle.

The future Soil Moisture and Ocean Salinity satellite time of overpass is 6 A.M. and 6 P.M. at the equator. In many regions, morning dew is expected at the time of the satellite overpass and might play a role in soil moisture retrievals. The aim of this study was to assess the effects of dew on L-band measurements. Radiometric, biomass, and dew measurements were performed over a natural grass field. Our results show that at the diurnal scale, vegetation internal water content changes play a major role in emission. A direct impact of dew on measurements was not identified. However, a brightness temperature increase of 0.5 and 1 K was observed at vertical and horizontal polarizations, respectively. This increase was detected later after dew, suggesting that it was not directly caused by the presence of dew on the grass blades. Our hypothesis is that the observed increase in brightness temperatures is due to the absorption of dew water by the litter layer.

The community microwave emission model (CMEM) is the low frequency forward observation operator developed at ECMWF. It is used in this paper to simulate brightness temperatures at local and regional scales over SMOSREX (France) and AMMA (West Africa), respectively. Background errors in simulated brightness temperatures are quantified at different frequencies and incidence angles for these two sites.

To retrieve soil moisture from L-band microwave radiometry, it is necessary to account for the effects of temperature within both vegetation and soil media. To compute the effective soil temperatureTG, several simple formulations accounting for soil temperatures at the surface and at depth and surface soil moisture have been developed. However, the effects of the soil physical properties in terms of texture, density, or structure, which all may be important variables in the modeling ofTG, have never been investigated. In this paper, several simple formulations ofTGat L-band, accounting for or ignoring the effects of soil texture and density, were developed and compared based on a very large simulated data set. The best configurations and parameterizations of these simple formulations were computed and could be directly used for operational applications in future soil moisture retrieval studies. For instance, we showed that the use of the surface temperature in the estimation ofTGcan be significantly improved by using additional information on the soil temperature at depth (the average error in the estimation ofTGdecreased from ~ 4 to ~ 1.8 K). On the contrary, almost no improvement was obtained if air temperature was used instead of surface temperature. Also, it is shown that the use of additional information on the soil properties, mainly the soil clay content and density, led to improved results by about 0.2 K in the estimation ofTG. The improvement was found to be larger for sandy and dry soils: simplified formulations accounting for soil properties are able to represent the fact thatTGis closer to the soil temperature at depth for these soil conditions.

The CoSMOS (Campaign for validating the Operation of the Soil Moisture and Ocean Salinity mission) campaign was conducted during November of 2005 in the Goulburn River Catchment, in SE Australia. The main objective of CoSMOS was to obtain a series of L-band measurements from the air in order to validate the L-band emission model that will be used by the SMOS (Soil Moisture and Ocean Salinity) ground segment processor. In addition, the campaign was designed to investigate open questions including the sun-glint effect over land, the application of polarimetric measurements over land, and to clarify the importance of dew and interception for soil moisture retrievals. This paper summarises the campaign activities, and presents progress on the analysis of the CoSMOS data set.

This paper compares L-band measurements from three different experiments in areas covered by grass. The main objective is to assess soil moisture retrievals based on the L-band Microwave Emission of the Biosphere model (L-MEB) used by the Soil Moisture and Ocean Salinity mission (SMOS). Results indicate that over grass the vegetation is isotropic to the microwave propagation at horizontal polarisation, while at vertical polarisation non-zero scattering is observed for all the grass data sets. Surface soil moisture is retrieved with enough accuracy for all data sets as long as the soil roughness and litter emission are calibrated beforehand. The study also highlights the importance of detecting strong attenuation by wet vegetation and litter due to rainfall interception. We show that strong rainfall interception can be flagged using a microwave polarisation index.

This paper presents an analysis of ENSIVAT advanced synthetic aperture radar data acquired over a Sahelian region located in Mali, West Africa. The considered period is 2004-2005 and includes two rainy seasons. Emphasis is put on two ScanSAR modes, namely, the global monitoring (GM) and the wide swath (WS) modes characterized by spatial resolutions of about 1 km and 150 m, respectively. Results show that the WS mode offers better performance in terms of radiometric resolution, radiometric stability, and speckle reduction than the GM mode. The latter is more appropriate for studies at large scale (> 10 times 10 km). In both modes, pronounced angular and temporal signatures are observed for most soil surfaces, and azimuthal effects are observed on markedly orientated rocky surfaces. In contrast, polarization differences (VV/HH) are small during the dry season except on flat loamy soil surfaces. Finally, a relationship is observed between the normalized WS backscattering signal at HH polarization and the surface soil moisture of sandy soils.

A simple reflectivity model of a bare soil at L-band is developed to account for the effects of soil roughness at different angles and polarizations. This model was developed using a long-term dataset acquired over the bare soil in the framework of the Surface Monitoring Of the Soil Reservoir EXperiment (SMOSREX). It is shown that the roughness effects are different depending on the measurement configuration, in terms of incidence angle and polarization. However, in this paper, a simple parameterization that is based on a single roughness parameter was calibrated in order to account for this angular and polarization dependencies. This parameter was found to be dependent on soil moisture: drier conditions were associated to higher ldquoroughnessrdquo conditions. The root-mean-square error between the measured and modeled reflectivities on days when no precipitation events were detected at vertical polarization (V-pol) is 0.0275, and at horizontal polarization (H-pol), the rmse is 0.0237; all incidence angles were considered. When all data are considered, the rmsd for V-pol is 0.0350, and for H-pol, the rmse is 0.0373. This new simple model is suitable for soil moisture retrieval from Soil Moisture and Ocean Salinity data. By means of this simple parameterization, almost two years of soil moisture data were retrieved with a good accuracy. The SMOSREX dataset allowed to ensure a long-term suitability of the proposed parameterization.

In this paper, microwave remote sensing, together within situmoisture probes, is used to investigate temperature effects on the soil dielectric constant. Field and specific laboratory measurements were performed for different soil water content over a wide range of temperatures. The experimental results lead to the following evidences: (1) temperature effect is different for bound and free waters in soil; (2) bound-water relaxation frequency falls within the range of frequencies that are used by impedance soil moisture probes for field measurements; and (3) the increase of bound-water relaxation frequency with soil temperature interferes in a significant way with moisture measurements when bound-water fraction is important. These results have implications in field experimentation since most moisture sensors operate under 500 MHz and are affected by this phenomena of relaxation.

Soil Vegetation Atmosphere Transfer(SVAT) models have been used for a long time to provide realistic descriptions of surface processes in order to simulate energy and water fluxes at the interface soil-vegetation-atmosphere. The initialization of biophysical variables of slow temporal variation, mainly root zone soil water content (w/sub 2/) and biomass, has an operational interest in atmospheric and hydrological models since they condition the surface fluxes. Therefore their correct initialization constitutes an important issue for short to medium term meteorological modelling. In this paper a first approach to assimilate remote sensing data in the Interaction Soil Biosphere Atmosphere SVAT model of Meteo-France, modified to account for the carbon dioxide concentration (ISBA-A-gs) [J.-C Calvet, 1998], is addressed in order to describe the slow temporal evolution of w/sub 2/ and the biomass. The experimental set-up at the south of Toulouse, SMOSREX (Surface Monitoring of the Soil Reservoir Experiment), is providing continuous data of meteorological forcing, energy fluxes, soil water content, soil temperature profiles and vegetation biomass. Moreover the addition in 2003 of a new very precise radiometer (LEWIS) is supplying radiometric measurements in L-band, reinforced by measurements in the infrared and solar domain by respectively an infrared radiometer and a reflectance-meter. All these continuous measurements will permit to test the assimilation techniques of multi-spectral remote sensing data in the surface models over fallow and bare soil with the goal of analyzing the soil water content in the root zone and the biomass.

This paper is part of an ongoing study aimed at exploring the potential of biangular measurements at L band to estimate near surface soil moisture. A statistical approach based on a physical radiative transfer model is tested over a natural grassland. Soil moisture shows an encouraging correlation with a combination of brightness temperatures measurements performed at two different angles. Nevertheless, the assumptions involved into the definition of a valid statistical index need to be studied. For that purpose, an insight into the modeling of grassland L band emission is also presented

A simple approach for correcting the effect of vegetation in the estimation of soil moisture (w/sub S/) from L-band passive microwave observations is presented in this study. The approach is based on statistical relationships, calibrated from simulated datasets, which requires only two observations made at distinct incidence angles (/spl theta//sub 1/,/spl theta//sub 2/). A sensitivity study was carried out, and best retrieval remote sensing configurations, in terms of polarization and couple of incidence angles (/spl theta//sub 1/,/spl theta//sub 2/), were investigated. Best estimations of w/sub S/ could be made at H polarization, for /spl theta//sub 1/ varying between 15/spl deg/ and 30/spl deg/, and with a difference (/spl theta//sub 2/-/spl theta//sub 1/) larger than 30/spl deg/. The method was tested against two experimental datasets acquired over crop fields (soybean and wheat). The average accuracy in the soil moisture retrievals during the whole crop cycle was found to be about 0.05 m/sup 3//m/sup 3/ for both crops.

In the framework of the preparation of the Soil Moisture and Ocean Salinity (SMOS) mission, several field experiments are required so as to address specific modeling issues. The goal is to improve current models and to test retrieval algorithms. However, adequate ground instrumentation is scarce and not readily available "off the shelf". In this context, a high-accuracy L-band radiometer was required for a specific long-term campaign for the preparation of the SMOS mission. For this purpose, a dual-polarized radiometer was designed and built to check algorithms for surface soil moisture retrieval from multiangular dual-polarized brightness temperatures. This radiometer has been tested in the field for 20 months and is operational since end of January 2003. The aim of this paper is to give details of the system architecture, calibration procedures, together with the performances obtained and some preliminary results.