Workshop on Surface Properties>
Session I: Assimilation of surface sensitive observations in Numerical Weather Prediction systems (Chairs: Godelieve Deblonde and Niels Bormann)
Benjamin Ruston, NRL, Monterey, CA, USA
Sid-Ahmed Boukabara, IMSG Inc., NOAA/NESDIS/STAR, Camp Springs, USA
Abstract: A survey was taken from the different operational global weather modeling centers of the world. We asked a series of questions designed to ascertain the use of surface sensitive microwave and infrared radiances in operational assimilation. It was found that microwave radiances over ocean were used unanimously, while use of microwave over land was being tested at many centers. Infrared was not as popular for assimilation, and the efforts seem to lag slightly behind the microwave efforts. Finally though recognized as an important factor, no center as yet is using spectrally variant skin temperature.
Fuzhong Weng1, Banghua Yan2, Paul van Delst3, Jim Jung4, John LeMarshall5, Jim Yoe1
1NOAA/NESDIS, Camp Springs, MD, USA
2QSS Inc., NOAA/NESDIS, Camp Springs, MD, USA
3EMC, Madison, WI, USA
4CIMMS, Madison, WI, USA
5JCSDA, Camp Springs, MD, USA
Abstract:Note that both radiance and Jacobian computations require accurate knowledge of surface emissivity and reflectivity. Without an emissivity model, the measurements from those channels of current and future advanced sounders that are sensitive to the lowest atmospheric layers may not be assimilated into NWP models. Shortly after the launch of the first AMSU in 1998, an ocean microwave emissivity model was developed at NOAA/NESDIS for AMSU radiance assimilation. The model developed by the Metoffice, United Kingdom (English and Takashima, 1998) was also tested in the operational environment. It is found that the Metoffice fast emissivity model (FASTEM) produced better results in assimilating the AMSU data at high frequencies, whereas the NEDSIS emissivity model performs better at frequencies less than 37 GHz.
At microwave frequencies, a land emissivity model was also developed for improving radiance assimilation (Weng et al., 2001). Prior to this model, constant emissivity values were used for land, snow and sea ice, respectively in the NCEP global data assimilation system. In the new emissivity model, the volumetric scattering from surface scatters is computed from a strong fluctuation theory which computes perturbation in wave-number and is utilized to calculate scattering and emission coefficients (Weng et al, 2001). In the case of vegetation covered land, geometrical optics is used to calculate the leave reflectivity and transmissivity since the leaf size is typically larger than the wavelength. The land surface roughness is approximated by the small perturbation theory. However, the convincing evidence of the model deficiencies comes from comparisons of the simulated global emissivity distribution with satellite retrievals from the AMSU (Weng et al., 2001) and from the high degree of spectral variability in aircraft brightness data observed during the NASA Cold Land Processes (CLP) experiments (Stankov, et al., 2004). In cold climate regimes (e.g., Greenland) where snow structure is very complex, exhibiting both stratification and metamorphosis, simulated biases are significantly high. The snow emissivity simulation is further improved using a hybrid algorithm in which satellite window channels are used to detect snow type and to retrieve emissivity at a few discrete frequencies. These retrievals are then used to adjust the theoretical snow emissivity model.
From visible to infrared wavelengths, the emissivity over land is derived from a look-up table, according to surface type and wavelength. In the table, the emissivity spectra are specified as function of surface types including water, old snow, fresh snow, compacted soil, tilled soil, sand, rock, irrigated low vegetation, meadow grass, scrub, broadleaf forest, pine forest, tundra, grass soil, broadleaf pine forest, grass scrub, oil grass, urban concrete, pine brush, broadleaf brush, wet soil, scrub soil, broadleaf 70-pine 30, and new ice. Currently, the theory and models are also being developed for infrared emissivity.
Jesse Meng, Kenneth Mitchell, Helin Wei
NOAA/NCEP/EMC, Camp Springs, MD, USA
Niels Bormann, Peter Bauer, Marco Matricardi, and Jean-Noel Thepaut
European Centre for Medium-range Weather Forecasts (ECMWF), Reading, UK
Abstract: The presentation gives an overview of the current status of the assimilation of surface-affected microwave or infrared radiances over land at ECMWF, and outlines plans for the refinement of the estimation of surface emissivity.
The current approach of assimilating the lower-troposphere sensing AMSU-A channels 5 and 6 over land, snow, and ice is using a parametric model to estimate the surface emissivity. It uses the AMSU-A window channels 1, 2, 3, and 15, a land/sea mask, and an estimate of the surface skin temperature to perform a simple scene classification into surface types, based on the frequency-dependent emission or scattering signature. The surface emissivity is then estimated from a type-specific parametric emissivity model applicable to nadir viewing. In case of fractional sea-ice cover, surface emissivity is calculated as a weighted average from the emissivities for water and ice. For snow-free land surfaces, a direct emissivity regression model is applied. Shortcomings of the approach include the assumption of nadir viewing, errors arising from misclassification, and lack of variability in surface emissivity for a given surface type.
Plans for refinement of the estimation of surface emissivity in the microwave are based on a direct retrieval of surface emissivity from AMSU-A window channels. Earlier work has demonstrated the feasibility of an online estimation of surface emissivity using the window channels and the atmospheric profiles from the First Guess available in the assimilation (Prigent et al. 2005). We also intend to adapt this methodology to SSMIS.
For the infrared, a land surface emisivity model is planned to be developed for use in radiative transfer computations. The model will be based on a static classification-based look-up table. It will use laboratory measurements of reflectance of various types of materials and it will allow for corrections due to the viewing geometry and the condition of the surface (i.e. moisture of the soil, snow cover, etc).
Stephen English, Fiona Hilton and TR Sreerekha
Met Office, UK
Abstract: Extensive use is made of satellite data in data assimilation systems but the use of channels with sensitivity to both surface and atmosphere has proven difficult and the impact of this data is still very limited. There are several reasons why this is the case. Firstly there is a data assimilation issue. That is, given an a new measurement which is sensitive both to the surface and the atmosphere, how do we partition the information between surface and atmospheric analysis changes. Secondly there is a state space issue. What are we trying to analyse and how many degrees of freedom does the problem have and is it underconstrained (the answer usually being yes). Thirdly there is a an issue of prior information. What prior information do we have about surfaces which is relevant to the data assimilation problem we are trying to solve and is it sufficient to make the problem well posed. Lastly there is an issue of radiative transfer, that is to say how do we model the relationship between the state space, the space occupied by the prior information and the observation space. Each of these four aspects of the problem is challenging and solving each one individually is a necessary but not sufficient criterion for success. This presentation will touch on all aspects but will focus on the data assimilation problem and how we might solve it.
Stephen English, Met Office, UK
Abstract: Ocean emissivity models have been available since the 1970s. There are three main components to such models: the permittivity of saline water; the treatment of roughness; foam. Recent laboratory measurements from Ellison et al. have provided reliable permittivity as a function of sea surface temperature. There are many models available for treatment of surface roughness. However for AMSU and SSM/I frequencies it has proven adequate to solve as geometric optics. The challenge for operational NWP was to make this fast enough and a fast model for the surface roughtness, fastem, has become quite widely used. A definitive model for foam remains elusive. There has been recent work but the modelling of the foam contribution remains the largest source of uncertainty above 15 m/s. Recently the models have been challenged to reproduce the behaviour of the full Stokes vector and its dependence on the relative azimuth angle between the wind direction and the look angle. This lends itself to fast modelling because the form of the azimuthal variation is well known and only the windspeed and sea surface temperature dependence needs to be established.
In terms of data assimilation we benefit (compared to the land and sea ice problem) from the fact that the NWP model has a good sea surface temperature analysis (error usually better than 1 K) and near surface wind speed short range forecast (error usually lower than 2 m/s). Therefore assimilation of sounding channels over the ocean is rather straightforward and the main purpose to the emissivity model is to allow wind speed or wind vector information to be retrieved as well as total column water vapour, cloud and rain (and sea surface temperature at low frequency).
In this presenation the status of the use of near surface window channels at NWP centres will be presented alongside their choice of emissivity model and the main characteristics of the models.
Louis Garand, MSC, Dorval, Canada
Abstract: MSC is planning to assimilate AIRS radiances by the end of 2006 from 100 channels. Surface-sensitive channels will be assimilated over oceans only. However, all is in place for research activities on the assimilation over land as well. Currently, surface emissivity over land is fixed for a given frequency and land type. The presentation will describe the system and examine the nature of skin temperature increments obtained. Various aspects of the problem will be briefly covered: horizontal length scale for skin temperature Ts, correlation between Ts and low-level temperature, and inter-channel error correlation.
Fatima Karbou, Elisabeth Gérard, and Florance Rabier
Météo-France, Toulouse, France
Abstract: AMSU-A & -B measurements are still not extensively used over land surfaces for atmospheric applications. Recent studies have shown that it is now possible to take advantage of the information content of these instruments if land emissivity and skin temperature estimates are improved. This paper reports on three land surface scheme comparisons using the Météo-France four-dimensional variational (4D-var) assimilation system. Firstly, a monthly mean estimated land emissivity atlas using AMSU-A & -B data is used to feed the model with updated emissivities. Two mean atlases have been tested: from the year 2000 and from the year 2005 . A second land surface scheme based on direct emissivity calculations is developed to obtain dynamically emissivity values. The third approach is based on the first one with the addition of a dynamic skin temperature estimation based on one AMSU-A or AMSU-B surface channel. The land surface schemes described above have been implemented within the 4D-var system and their results have been compared with those obtained while using the operational system. With improved emissivities and/or skin temperatures, the observation operator simulations are clearly improved compared with the operational model. We also noticed that many more data could be assimilated when the surface scheme was updated.
Fatima Karbou, Elisabeth Gérard, and Florance Rabier
Météo-France, Toulouse, France
Abstract: Assimilation experiments with the improved microwave land surface emissivities (Karbou et al., 2006) within the 4DVar assimilation system at Météo-France will be presented. This new scheme allows for a better description of land surface properties and the assimilation of a greater number of surface sensitive AMSUA and AMSUB channels. We will focus on the impact of such a modification on the analysis quality and the forecast skills.
Harald Schyberg, Vibeke W. Thyness, Frank T. Tveter
Norwegian Meteorological Institute, Oslo, Norway
Abstract: The Norwegian Meteorological Institute (met.no) operationally runs a 20km resolution version of the HIRLAM (High Resolution Limites Area Model) limited-area model for short range weather forecasting. The HIRLAM model includes a 3D-Var assimilation scheme. Here we present the work undertaken to extend the assimilation of AMSU-A brightness temperatures to include observations over Arctic areas covered with sea ice.
Usually the properties of the ice surface change slowly, so information from recent past passages of microwave radiometers can help determine the surface emissivity in the AMSU-A channels. As a first approach, typical emissivity values for First-Year and Multi- Year sea ice are used in combination with daily maps of SSM/I based sea ice type and concentration retrievals running operationally under the EUMETSAT Ocean and Sea Ice Satellite Application facility (OSI SAF).
We will present quality control issues and impact studies on the effect of including AMSU-A observations over the Arctic sea ice. We will discuss both the impact of assimilating the channels not sensitive to the surface and the effect of adding surface channels using the simple emissivity formulation described above. With the current implementation, the AMSU-A observations over sea ice contribute on average slightly positively to the forecast quality of the HIRLAM 20km model at met.no. Some plans for future developments on the topic will also be presented.
Session II: Emissivity Modeling (Chairs: Norman Grody and Stephen English)
Dan Tarpley1, Le Jiang2, Kenneth Mitchell3, Helin Wei3, and Vince Wong3
1Center for Satellite Applications and Research, NESDIS/NOAA, Camp Spring, MD, USA
2I. M. Systems Group, Inc. at NOAA/NESDIS, Camp Spring, MD, USA
3Environmental Modeling Center, NCEP/NWS/NOAA, Camp Spring, MD, USA
Abstract: The NCEP operational surface physics model, Noah, uses several prescribed fields of vegetation properties, including vegetation class, leaf area index (LAI) and green vegetation fraction (GVF), to partition the surface evaporative flux between bare soil and vegetated surfaces. In this paper we briefly review the physics of the Noah model as it is affected by vegetation characteristics and show some results that demonstrate the sensitivity of the model to vegetation properties.
NESDIS delivers an operational weekly green vegetation fraction product as one of the prescribed fields for operational NCEP NWP models. The GVF is derived by regression against a scaled NDVI, called Vegetation Condition Index, which, in turn, is dependent on the current NDVI and a weekly NDVI climatology. The underlying assumption of the process is that the observed NDVI from each AVHRR (or in the future, VIIRS) is well-calibrated and is derived from bands that have the same spectral band shape as those that went into the NDVI climatology. This is obviously not the case among the different AVHRRs or between AVHRR and VIIRS. To adjust the GVF from any imager to one comparable to the standard AVHRR, correction methods are being evaluated. The most promising is adjusting the current sensor NDVI to the standard NDVI (climatology) by matching empirical distribution functions between the operational instrument and the standard. If successful, the correction should allow quick delivery of operational GVF products upon the launch of a new AVHRR or from VIIRS. Results of experiments with the procedure will be shown.
Albert Olioso1, Frédéric Baret1, Stéphane Jacquemoud2
1INRA/CSE, Domaine Saint Paul, Agroparc, Avignon, France
2Laboratoire Environnement et Développement, Université Paris 7, France
Abstract: The PROSPECT model was developed by Jacquemoud and Baret (1990) to simulate reflectance and transmittance spectra of plant leaves between 0.4 and 2.5 µm. In the present study, this model is adapted in order to simulate leaf spectra between 3 and 14 µm. Required information consists in: water and dry mater specific absorption spectra, a structural indices describing the level of leaf internal reflection, water and dry matter leaf content, leaf refractive index and an incidence angle determining the first interface transmittivity. Water absorption coefficients were set from the literature, while other leaf characteristics were obtained by combining a model sensitivity analysis and an inversion of the model against measured spectra. Results show that absorption by water, and dry matter at a lesser level, is very high as compared to what happens between 0.4 and 2.5 µm. In consequence reflectance is usually low (few percent) and transmittance very close to zero. However reflectance may increases up to 30% between 4 and 5 µm and up to 20 % between 11 and 13 µm when the leaf is very dry (transmittance also increases). This is in agreement with measurements available in ASTER and MODIS spectral library. For fresh leaves, variation in reflectance level between species may be interpreted by considering leaf surface effect. The model may be used to study spectral response of plant leaves in the thermal infrared, as for example to analyze the variation of leaf spectral emissivity as a function of leaf water content.
Albert Olioso1, José Sobrino2, Guilhem Soria2, Michaël Chelle3, Benoit Duchemin4, Frédéric Jacob5
1INRA/CSE, Domaine Saint Paul, Avignon, France
2Department of Thermodynamics, Faculty of Physics, University of Valencia, Burjassot, Spain
3INRA, UMR Environnement et Grandes cultures, Thiverval-grignon, France
4CESBIO - Centre d'Etudes Spatiales de la Biosphère, Toulouse, France
5Laboratoire de Télédétection et de Gestion des Territoires, Ecole Supérieure d'Agriculture de Purpan, Toulouse, France
Abstract: The SAIL model was developed by Verhoef (1984, 1985) for simulating land surface directional spectral reflectances in the solar domain. Olioso (1992, 1995) adapted the model for simulating radiative transfers in the thermal infrared, and particularly for simulating land surface emissivity. This version of the model was called SAIL-Thermique. Simulations of land surface emissivity using another version of the SAIL model were also recently presented by Verhoef et al. (2005) and by Sobrino et al. (2005). Due to the difficulty to measure land surface emissivity, almost no emissivity model has ever been validated against ground measurements. In this study, several datasets extracted from the literature and from various databases were used in order to evaluate emissivity simulations by the SAIL-Thermique model.
We used data acquired in Barrax (Spain, EFEDA and DAISEX experiments), in Marrakech (Morocco, WATERMED European project), in the Alpilles test site (South East of France, ReSeDA European project), in Ardèche and Hérault (South East of France, Valor and Caselles, 1996), in Tsukuba (Japan, Sugita et al. 1996), and in Bostwana (Van de Griend et al. 1991). These datasets were acquired on several types of land surfaces including natural and agricultural vegetations at different levels of growth and of water status. Compiled land surface emissivities for the 8-14µm spectral band ranged between 0.92 and 0.99.
Model simulations were performed from the knowledge of leaf area index, leaf inclination distribution, direction of observation, and leaf and soil emissivities. As data on leaf inclination and leaf emissivity were usually not available for a specific experiment, stochastic simulations were performed from an a priori knowledge on their distribution which was extracted from an intensive survey of the literature. In general, simulated 8-14 µm emissivities were favorably compared to measurements with a root mean square difference around 0.006. When considering only herbaceous species, the root mean square difference was 0.004. The model is also used for simulating emissivity spectra in order to provide information for the interpretation of multispectral emissivity estimation from the ASTER sensor.
Jean-Pierre Lagouarde, Mark Irvine, Sylvia Dayau, Britta Kurz Jérôme Ogée, Patrick Moreau, Dominique Guyon, Isabelle Champion
INRA Unité EPHYSE, Villenave d'Ornon, France
Abstract: Measurements of surface temperature performed in the thermal infrared (TIR) domain display important directional anisotropy and significant 'hot spot' effects during daytime. These depend (1) on the surface structure which governs the temperature profiles inside the canopy via the coupled energy-radiative transfers simultaneously to the spatial distribution of the facets seen by the sensor, and (2) on the solar position. Characterizing the TIR directional anisotropy is important for several purposes : (1) access to the surface temperature of the different canopy layers for improving sensible heat flux estimates, (2) assimilation of multi-angular remotely sensed data in the surface models, (3) correction and normalization of large swath satellite sensors with the scope of analyzing temporal or spatial variability, and (4) definition of optimal viewing configurations and recommendations for future TIR spatial systems.
Several experiments based on airborne measurements have been performed to characterize the TIR directional anisotropy over different types of surfaces: pine forests, vineyards, urban areas. The protocol is based on the use of a TIR camera equipped with wide-angle lenses and installed aboard a small aircraft flying different directions. It allows retrieving directional anisotropy (differences between oblique and nadir viewing temperatures) in a range of zenithal viewing angles up to 60° and in all azimutal directions. The possible sources of errors related to the instruments and the atmosphere are analysed and corrections proposed.
Experimental results reveal significant directional anisotropy whatever the surface type. Measurements performed at different times of day and at different periods of the year show systematic hot spots in relation with the sun position. The impact of the surface structure is also illustrated. Over maritime pine stands it is shown that the size of the hot spot directly depends on the stand structure (i.e. size, geometry and spacing of trees on the stand), with values reaching ±2 K in the principal plane. Results obtained over vineyards reveal that anisotropy results from the combination of a 'macro structure' effect (orientation of rows) with a 'micro-structure' effect (hot spot of bare stony soil). The strong contrasts between dry bare soil and vegetation walls explain huge directional variations from -12 to 4 K in the principal plane. The impact of the size of buildings, the built-up ratio, and the vegetation ratio on the TIR anisotropy over urban areas is also illustrated from results obtained in the framework of the ESCOMPTE experiment performed at Marseille in 2001 which show ranges of variation between -5 K up to 7 K. An example of application for the correction of two temporal series of NOAA 14 and NOAA 16 data over the city is given.
Finally we discuss the difficulty of assessing the contribution of angular variations in surface emissivity to the overall thermal anisotropy, and we briefly present a directional anisotropy modelling approach of surface temperature based on combining 3D canopy models with surface models.
Benjamin Ruston, NRL, Monterey, CA, USA
Abstract: Comparisons between infrared and microwave emissivities retrieved in a 1DVAR assimilation scheme, with the initial estimates from a combination of model and databases. The initial estimates of microwave emissivity over land, snow, and sea-ice are taken from the JCSDA Microwave Emissivity Model (MEM). Infrared emissivity estimates are given by the ASTER spectral reflectance database indexed to a one-half degree resolution global land database of soil and vegetation. Results show the JCSDA model does well over most vegetated surfaces, and suffers in arid and desert regions where over estimates may approach 5%. The infrared emissivities have similar regions of error, with again desert regions overestimated by the indexed first guess but the discrepancy is less severe (~2%) around the 11 micron window but grows to around (~5%) in the near-infrared window around 4 microns.
Christian Mätzler1 and Philip W. Rosenkranz2
1Institute of Applied Physics, University of Bern, Switzerland
2Massachusetts Institute of Technology, Cambridge, MA, USA
Abstract: For significant surface reflection, the brightness temperature above planetary surfaces not only depends on surface temperature, emissivity and on atmospheric emission, but also on the type of bistatic scattering. This dependence can be specified by an effective incidence angle from zenith of downwelling radiation. We obtained analytic expressions for the reflectivity and for typical scattering functions, such as Lambert, Lommel-Seeliger, multiple-isotropic (Chandrasekhar). A plane-parallel atmosphere was assumed. In all cases the effective incidence angle decreases with increasing zenith opacity (considered range: 0 to 1), and in most cases the dependence on observation direction is small. These results are in contrast to specular reflection where the effective incidence angle is given by the observation angle, which of course, is independent of opacity.
The dependence of terrestrial brightness temperatures on the type of bistatic scattering was studied with multi-frequency and multi-angular data provided by the Advanced Microwave Sounding Unit-A (AMSU-A). A potential exists to infer a parameter, AL, describing the relative contributions of Lambertian (AL=1) and specular (AL=0) scattering, as demonstrated for the case of Antarctica. Whereas the open ocean has AL values in the range from 0 to 0.3, sea ice and land ice are more Lambertian with AL values in the 0.5 to 1 range. The existence of some super-Lambertian regions with AL>1 was revealed.
Rasmus Tonboe, Center for Ocean and Ice, Copenhagen, Denmark
Abstract: The strategic importance of sea ice emission models is to relate physical properties of the target to brightness temperatures at different microwave frequencies and polarisations and further to interpolate between observations at different places and frequencies. Because the radiances received at the satellite generally contains contributions from both atmosphere and surface, the surface emission model relationships are needed for the retrieval of both surface and atmospheric parameters. About 4-5% of Earths surface is covered by sea ice and meteorological measurements in arctic regions are sparse. Atmospheric conditions in the Arctic never the less impact the weather in Northern Europe and other populated regions at these relatively high latitudes. Open ocean microwave brightness temperatures measured by satellite radiometers at sounding frequencies e.g. 150 and 183GHz from the AMSU instruments are assimilated into numerical weather prediction (NWP) models. The measurements are prominent for information on atmospheric temperature and humidity and have a significant positive impact on NWP model performance. The use of these data for atmospheric sounding is possible because of models describing the open ocean surface emissivity at these frequencies. Similar models do not exist for snow and ice emissivity.
A sea ice module is added to the microwave emission model of layered snowpacks (MEMLS; Wiesmann & Mätzler, 1999) suitable for snow on land in the 5-100GHz rage. Measured properties of snow and ice, collected in the Arctic Ocean, are used as input to the model and the seasonal variability is simulated using the output from a thermodynamic and mass model for sea ice. The scattering formulation is the most significant limitation for extending the MEMLS range of validity into the 100-200GHz range. The use of different scattering formulations is discussed.
The aim with the presentation is to give a short overview over the quantitative influence of different microphysical parameters in sea ice and the snow on top of it and how these link to emissivity, brightness temperature, dielectric properties and scattering.
J-P Wigneron1, Y. Kerr2, P. Waldteufel3, P. Richaume2, K. Saleh1,4, J. P. Grant1,5, A. Van de Griend5, F. Demontoux6, G. Ruffié6, P. Ferrazzoli7, P. De Rosnay2, M-J Escorihuela2, J-C Calvet8, J. Fenollar4, E. Lopez-Baeza4
1INRA, EPHYSE, Villenave d'Ornon, France
2CESBIO, Toulouse, France
3IPSL/Service d'Aéronomie, Paris, France
4Universitat de Valencia, Spain
5Vrije Universiteit Amsterdam, Netherlands
6PIOM, UMR Bordeaux, France
7Universita di Roma Tor Vergata, Roma
Abstract: In the framework of the ground segment activities for the SMOS mission, geophysical products (e.g. soil moisture (SM) and vegetation characteristics) will be produced by an operational SMOS Level 2 Soil Moisture retrieval algorithm. The principle of the algorithm is based on an iterative approach, minimizing a cost function computed from the sum of squared weighted differences between measured and modelled brightness temperature (TB) data, for a range of incidence angles. The retrievals provide the best suited parameters driving the direct TB model that minimize the cost function. The selected direct model is the so-called L-MEB (L-band Microwave Emission of the Biosphere) which was used in the first ESA studies aiming at assessing SMOS capabilities from synthetic data set. In the algorithm process, for each incidence angle, the different cover types (vegetated area, open water, urban area, land use, etc.) present within the SMOS footprint are estimated from high resolution land use maps. For low vegetation and forest categories, these maps allow to distinguish between a large number of sub-categories (N > 200) corresponding to grasslands, crops, matorral, tropical & boreal forests, etc. for a variety of climatic and geographic conditions. Based on these refined vegetation classes, corresponding to specific vegetation structure, and on maps of soil properties (for soil texture, roughness and bulk density) parameters driving the L-MEB can be selected and tabulated.
Of course L-MEB is not frozen but prone to evolve so as to gradually incorporate significant improvements made in the field. Recently, in the framework of SMOS research activities, advances in understanding and modelling the L-band microwave emissivity of vegetation have been achieved. These results were mainly obtained over natural land covers, over which very few informations were available to date. They could be derived from a substantial number of experimental campaigns carried out throughout Europe, for a variety of vegetation/soil characteristics and climatic conditions; To name but a few, SMOSREX over fallow and bare soil near Toulouse, MELBEX over matorral in the Mediterranean environment of Valencia, Bray and Jülich over, respectively, coniferous (Les Landes) and deciduous forests, etc.
The combination of these experimental data with modelling activities over the last years helped to improve significantly our knowledge of the key processes that drive the emission of natural vegetation canopies: effect of water interception by the standing vegetation, attenuation of soil emission by mulch and litter, dependence of vegetation attenuation on configuration parameters (incidence angle, polarization), etc. The most significant recent results obtained over prairies, forests and matorrals are presented and discussed in this communication.
Session III: Emissivity retrieved from observations (Chairs: Dan Tarpley and Christian Mätzler)
Eva E. Borbas, Suzanne Wetzel-Seemann, Robert O. Knuteson, Paolo Antonelli,Jun Li and Hung-Lung Huang
Space Science and Engineering Center, University of Wisconsin-Madison, USA
Abstract: A global infrared land surface emissivity database with high spectral and high spatial resolution will be introduced. The database is derived from a combination of high spectral resolution laboratory measurements of selected materials, and MODIS (MOD11) observed land surface emissivities at 3.7, 3.9, 4.0, 8.5, 11.0 and 12.0 micron wavelengths. For a given month, a spectrum of emissivity from 3.7 to 14.3 micron is available from this database for every latitude/longitude point globally at 0.05-degree resolution.
Two methodologies will be discussed. In the first approach, a baseline emissivity spectra (BES) is derived from laboratory measurements of high spectral emissivity. The BES are derived at moderate spectral resolution (8 values in the 3.7-14.3 micron range), with wavelengths chosen as the inflection points that best characterize the shape of each relevant laboratory emissivity spectra. Then BES are applied by adjusting the magnitude of the emissivity at each of the inflection point wavelengths based on the observed MODIS MOD11 emissivity values. The result of the BES adjustment is a spectrum of emissivity at the inflection points for each month at each MOD11 latitude and longitude point (0.05 degree resolution) over land.
In the second approach, principal component analysis of 332 laboratory spectra and MODIS emissivity observations were used to create high spectra resolution emissivity dataset.
Both method will be implemented and run for one year of global data, and the results will be compared to each other. For validation the datasets will be applied to MOD07 and AIRS retrieval algorithms and results will be compared.
E. Péquignot1, A. Chédin1, C. Gautier2, N. A. Scott1
1Laboratoire de Météorologie Dynamique (LMD), Institut Pierre-Simon Laplace, Ecole Polytechnique, Palaiseau, France
2Institute for Computational Earth Systems Science (ICESS), University of California, Santa Barbara, California, USA
Abstract: Surface temperature and emissivity are variables essential for greatly improving the estimation of the longwave surface energy budget and, consequently, improving the performance of surface-atmosphere interaction models. Continental surface infrared emissivity strongly depends on the wavelength and on the type of the surface. For example, in the reststrahlen spectral band of quartz around 8-10 µm, emissivity values as low as 0.6 may be observed, particularly over Sahara desert. In this paper, Atmospheric InfraRed Sounder (AIRS) satellite observations (aboard the NASA/AQUA platform) are interpreted in terms of monthly mean surface "skin" temperature and emissivity. Clear sky radiances are first determined by using Kohonen maps as a classification tool for clear and cloudy AIRS pixels. Then, the difficulty of the interpretation of window channel radiances in terms of surface emissivity and temperature is accounted for by choosing, instead of simple linear regressions, the non linear way offered by neural network techniques. In particular, such inference techniques allow potential non linearities to be properly taken into account and do not necessitate to explicitly formulate the functional linking the knowns and the unknowns. Here, use is made of the Multi-Layer Perceptron (MLP). Training the MLPs is performed using the Thermodynamic Initial Guess Retrieval (TIGR) climatological library of about 2300 representative atmospheric situations selected by statistical methods from 80,000 radiosonde reports. This technique is applied to the simultaneous determination of surface temperature and emissivity at 9 wavelengths. Computing more wavelengths would require reducing significantly the computational burden. The second algorithm relies on an estimation of the surface temperature provided by a dedicated MLP and an evaluation of the atmospheric thermodynamic state through a proximity recognition scheme within the TIGR dataset. With this prior information all terms of the radiative equation are evaluated independently by using the 4A line-by-line radiative transfer model. Therefore emissivity spectra may be calculated for all atmospheric windows (transmittance greater than 0.3). In order to estimate the quality of these two methods, corresponding retrievals are compared with the Moderate Resolution Imaging Spectroradiometer (MODIS) monthly mean L3 products also flying aboard the NASA/AQUA platform.
J. A. Sobrino, J. C. Jiménez-Muñoz, J. Cuenca, M. Gómez, G. Sòria, M. Romaguera, M. M. Zaragoza-Ivorra, Y. Julien and M. Atitar
Global Change Unit. Dpt. of Earth Physics and Thermodynamics, University of Valencia, Spain
Abstract: This paper shows the application of different methodologies to retrieve land surface emissivity (LSE) from VNIR and TIR data. Methods for LSE retrieval from VNIR data include the NDVI Threshold Method (NDVITHM) which has been used and adapted in order to obtain LSE from low, medium and high spatial resolution sensors such as Advanced Very High Resolution Radiometer (AVHRR), Moderate Resolution Imaging Spectrometer (MODIS), Advanced Along-Track Scanning Radiometer (AATSR), Spinning Enhanced Visible and Infrared Imager (SEVIRI), Thematic Mapper (TM) and Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER). The method has been also applied to airborne sensors such as Digital Airborne Imaging Spectrometer (DAIS) and Airborne Hyperspectral Scanner (AHS). LSE has been retrieved from TIR data using the Temperature and Emissivity Separation (TES) algorithm, which can be only applied to sensors with more than four TIR bands, such as ASTER, DAIS or AHS. Besides, an analysis of methods for retrieving emissivity spectra from ground-based measurements is also presented, since they are of fundamental importance to test the values obtained from satellite or airborne data. For this purpose the box method and the TES algorithm were used. Finally, a study of the angular variations on surface emissivities was carried out using a field goniometric system and results predicted by theoretical models.
Thomas Schmugge1, Kenta Ogawa2
1New Mexico State University, Las Cruces, New Mexico, USA
2School of Engineering, University of Tokyo, Tokyo, Japan
Abstract: Knowledge of the land surface emissivity is important for estimating the longwave radiation budget, a decrease of soil emissivity by 0.1 will decrease net and upward longwave radiation by about 6.6 and 8.1 Wm^2, respectively. In most atmospheric models the spatial variability of the emissivity is not accounted for because of a lack of data. The multi-spectral thermal infrared data from the Advanced Spaceborne Thermal Emission and Reflection (ASTER) radiometer and the Moderate Resolution Imaging Spectrometer (MODIS) provide new tools for observing the land surface emissivity. ASTER has 5 channels in the 8 to 12 micrometer wave band with 90meter resolution. MODIS has 3 channels in this wave band with 1 km resolution and it provides almost daily coverage. Both sensors are onboard the NASA Earth Observing System (EOS) Terra satellite, which was launched in 1999. MODIS is also on the Aqua satellite. Data from ASTER can be used to assess the spectral and spatial variations of surface emissivity when used with the Temperature Emissivity Separation (TES) algorithm. TES makes use of an empirical relation between the range of observed emissivities and their minimum value to extract the temperature and 5 emissivities from the 5 channels of ASTER data. The approach was validated with ASTER data acquired over the Jornada Experimental Range and the White Sands National Monument in New Mexico between 2001 and 2005 yielding good agreement with ground measures of emissivity. The approach was extended to produce maps of emissivities over a 400 x 1200 km area for a desert region of North Africa, including the sand dunes of the Grand Erg Oriental. The spectra for the sand dunes showed good agreement with that expected for quartz sand based on laboratory and field measurements. A multiple regression approach was used to relate the emissivities of the 5 ASTER channels to the window channel emissivity. The spatial variation of the window channel emissivity observed by ASTER is from 0.8 to 1, which corresponds to a range of 15 w/m^2 in the net surface longwave radiation under a dry atmosphere. These results show that ASTER data can be used to map the spatial and spectral variations of surface emissivity over large areas in particular the deserts of the world for which there is much exposed soil and sand. A broadband emissivity map for North Africa was generated using the coarser resolution MODIS data The range of the broadband emissivity was found to be between 0.80 and 0.96 for the desert area. The expected RMS error of the map is about 0.02. Such an emissivity map has been used as an input to a climate model and improves the prediction of surface and air temperatures by up to 1 degree C .
Françoise Nerry, Geng-Ming Jiang and Zhao-Liang Li
Laboratoire des Sciences de l'Image, de l'Informatique et de la Télédétection,Illkirch, France
Abstract: This work addresses the retrievals of Land Surface Emissivity (LSE) from combined mid-infrared and thermal infrared data of Spinning Enhanced Visible and Infra-Red Imager (SEVIRI) onboard the geostationary satellite - Meteosat Second Generation (MSG). To tackle the low temporal resolution problem of the ECMWF atmospheric data (only four profiles per day), a new atmospheric correction scheme was developed. For the middle infrared (MIR) channel, because it is less sensitive to water vapor, the clear-sky and time-nearest atmospheric data are used for the images where no atmospheric data are available. For the thermal infrared (TIR) channels, a new physics-based temperature diurnal cycle model was developed to remove the atmospheric attenuation for the images when no atmospheric data are available. The separation of surface temperature and LSE is based on the concept of the Temperature Independent Spectral Indices TISI constructed with one channel in MIR and one channel in TIR: Assuming that a special ratio of day and night emissivities do not change between day and night, the bi-directional reflectivity in the MIR channel is determined by extracting the contribution of solar reflection occurring in this channel. Then, taking into account the angular variations of the BRDF in this channel by using a whole day of measurements (6:00 am to 6:00 pm), the directional hemispherical reflectivity and the directional emissivity in this channel are deduced from the bi-directional reflectivity. Using the spectral properties of TISI again, emissivities in the TIR channels are extracted.
The results show that the SEVIRI data are properly corrected form the atmospheric perturbations using new atmospheric correction method. The results confirm too that the land surfaces do not behave as a Lambertian reflector and that the retrieved emissivities in channels 4, 9 and 10 over extended areas are consistent with those obtained from former researches.
The methodology is then applied to sets of data covering the 0-60N and 20W-40E region and for the 4 main periods winter, spring, summer and autumn to attain the seasonal variation of the emissivity parameters.
Leonardo Peres1, Carlos DaCamara1, Isabel Trigo1,2, and Cristina Madeira2
1Centro Geofísico da Universidade de Lisboa, Lisboa, Portugal
2Instituto de Meteorologia, Lisboa, Portugal
Abstract: Land Surface Temperature (LST) and Land Surface Emissivity (LSE) are two key parameters for the characterization of land-surface processes. Good estimates of LSE are essential for correct assessment of surface energy budget in Numerical Weather Prediction and Climate Models, as well as for the retrieval of surface and low-troposphere variables from remote-sensed information over land.
The Satellite Application Facility on Land Surface Analysis (LSA SAF) currently generates, on a operational basis, LSE maps for Meteosat Second Generation/Spinning Enhanced Visible and Infrared Imager (MSG/SEVIRI) channels IR10.8 and IR 12.0, used for the retrieval of LST. The LSE algorithm is based on the so-called vegetation cover method (VCM), and uses another LSA SAF product - the Fraction of Vegetation Cover (FVC). This methodology has been developed for the currently retrieved LSE maps (IR10.8 and IR 12.0), as well as for the remaining IR channels (IR3.9 and IR8.7), and for a broadband LSE (3-14 µm), necessary for the estimation of longwave surface fluxes. The generation of LSE maps - broadband plus all IR MSG/SEVIRI channels - by the LSA SAF operational system is expected to be implemented during the current year.
Laboratory measurements of LSE were obtained from the Johns Hopkins University (JHU) and Jet Propulsion Laboratory (JPL) spectral libraries included in the Advanced Thermal Emission Reflectance Radiometer (ASTER) library, as well as from the Moderate Resolution Imaging Radiometer-University of California, Santa Barbara (MODIS-UCBS) spectral library.
The LSE methodology presented here provides an expected LSE value, on pixel-by-pixel basis, as well as an estimate of the uncertainty and variation of LSE in SEVIRI channels IR 3.9, IR8.7, IR10.8 and IR 12.0 for each surface type. Those ranges may be used in sensitivity studies of algorithms where an a priori knowledge of LSE is required, such as the generalized split-window scheme used by the LSA for the retrieval of LST. In the case of broadband LSE (3-14 µm), the estimated uncertainties are useful in the impact assessment of spatial and spectral LSE variations on energy balance and climate modelling studies (where constant and uniform LSE values are generally assumed).
In addition, the LSE maps supply first-guess estimates to be used in algorithms that allow a simultaneous retrieval of LST and LSE. This is the case of the Two Temperature Method (TTM), which assumes that LSE does not change between observations and that atmospheric effects may be adequately estimated by means of a radiative transfer model. Previous studies performed within the framework of the LSA SAF strongly suggest that TTM may be used as a complementary method for LST and LSE retrievals, particularly in areas where LSE is not well known a priori, and where the traditional Split-Window (SW) algorithm for LST estimation has a higher sensitivity to LSE (e.g., over dry and semi-arid areas, dominated by bare soil or sparse vegetation). The first results of a methodology, which merges the SW and TTM schemes to produce realistic LSE maps over (semi-)desert regions are also shown here.
Catherine Prigent1, Filipe Aires2, Bill Rossow3
1 LERMA/CNRS, Observatiore de Paris, France
2 LMD/CNRS, Université Paris VI, France
3 NASA/GISS, New York, USA
Abstract: Microwave land surface emissivities have been calculated over the globe for 10 years between 19 and 85 GHz for both orthogonal polarizations, using satellite observations from SSM/I. Ancillary data (IR satellite observations and meteorological reanalysis) help remove the contribution from the atmosphere, clouds, and rain from the measured satellite signal. The method to calculate the emissivity is very general and can be applied to other sensors.
The objectives of these microwave emissivity calculations are manifold. First, these emissivities can help characterize the land surface properties. They are sensitive to the vegetation density, to soil moisture, to the presence of standing water at the surface, to the snow behavior. Being available over a decade, it is now possible to analyze the inter-annual variability of land properties such as the global inundation extent. Second, with these estimates, one can calculate the surface contribution to the radiance measured from satellite and as a consequence deduce the atmospheric characteristics from passive microwave even over land. Lastly, this data set can help evaluate emissivity models at global scales.
The monthly mean emissivities are available for the community, with a 0.25°x0.25° spatial resolution over the 07/1992-10/2001 period.
Jean-Luc Moncet, Pan Liang, John F. Galantowicz, and Alan E. Lipton
Atmospheric and Environmental Research, Inc., Lexington, MA, USA
Abstract: Land surface emissivities in the AMSR-E and SSMI channels have been retrieved over several months from combined AMSR-E, AIRS, and MODIS instruments and physical land-surface models and analyses. We have examined the temporal variability of retrieved local surface emissivities over selected regions of the globe and provide a preliminary assessment of error sources and the potential for minimizing them. Accurate a priori knowledge of local surface emissivity is required for lower tropospheric microwave remote sensing over land and for land surface parameter retrievals like land surface temperature (LST). Ideally, for a stand-alone microwave system (i.e., without an external source of surface temperature), a priori emissivity accuracies of 0.01 or less are needed to minimize the impact of cloud liquid water on temperature and water vapor retrievals and to improve surface temperature retrievals to 2 K or better. Because surface properties may change rapidly and unpredictably, an emissivity database must be frequently updated and be able to self-report on the quality of each emissivity estimate. Surface emissivity may be best characterized, for example, using collocated microwave and infrared observations in many cloud-free scenes where microwave penetration of the land surface media is minimal. In other scenes, terrain, surface type inhomogeneities, residual (undetected) clouds, and other factors will reduce emissivity retrieval quality.
Globally, emissivity retrieval is especially sensitive both to the way the surface effective emitting temperature is modeled and to its source in our 1-D VAR retrieval algorithm. Retrieval quality depends on time-of-day, effective emission depth at each microwave frequency, LST source quality (especially cloudiness and timeliness), radiative transfer model, and other factors. We have attempted to characterize the differences first between various LST sources (measurements and models) and second between the LST estimate and the land surface effective emitting temperature. We have also used land surface thermal and emission models to gauge the influence of emission depth and time-of-day factors. Our processing stream now incorporates both SSMI F13 (0600/1800 equator crossing time) and AMSR-E (1330/0130) data and additional SSMI and the TRMM TMI may be added. We are building a collection of test grid cells covering both problem areas and areas where retrieval confidence is high, and we are using the test cases to develop solutions for separating the emissivities from secondary signals. Our final product will be a full year database of clear- and cloudy-sky emissivity estimates, quality control flags, and ancillary data. The database will both support a prototype system for all-weather AMSR-E retrievals and serve as a pathfinder for future systems such as NPOESS CMIS with more comprehensive microwave spectral coverage and enhanced atmospheric sounding capabilities.
Sid-Ahmed Boukabara1 and Fuzhong Weng2
1IMSG Inc., NOAA/NESDIS/STAR, Camp Springs, USA
2INOAA/NESDIS/STAR, Camp Springs, MD, USA
Christian Mätzler, Institute of Applied Physics, University of Bern, Switzerland
Abstract: For background information satellite retrievals need first-guess emissivity spectra and their cross correlations. Such information was obtained from a long-term experiment of surface-based measurements made with the 5-frequency microwave radiometer system PAMIR from 1978 to 1995. The dataset covers measurements made over snow-covered ground, vegetation-covered ground (grass, barley, oat), bare ground, sea ice and open ocean, at vertical and horizontal linear polarisation with concentration on an observation nadir angle of 50°. A review of the experimental data made over land was presented by Mätzler (1994). Here we we will add cross correlations between emissivities at different frequencies to infer the accuracy of spectral interpolations. The uncertainty of interpolations is required as error estimate in the ill-posed retrieval problems.
Catherine Prigent1, Elodie Jaumouillé1, Frédéric Chevallier2, Filipe Aires3, Fuzhong Weng4
1Observatoire de Paris, France
2LSCE, Gif-sur-Yvette, France
3Université Paris VI, France
4NOAA/NESIDS, Camp Silver, USA
Abstract: Land surface emissivities have been calculated for TMI, SSM/I, AMSU-A conditions, for two months (July 2002 and January 2003) over the globe at ECMWF 1) directly from the satellite observations, 2) using the Weng et al. 2001 emissivity model. From this data set, a parameterization of the microwave emissivities is proposed that accounts for frequency, angle, and polarization dependences. It is anchored to a climatological monthly-mean maps of the emissivities at 37 GHz, calculated from SSM/I. For each location and time of the year, it can provide realistic first guess estimates of the microwave emissivities from 10 to 100 GHz, for all scanning conditions, as well as the emissivity statistics required in an assimilation environment.
Session IV: Satellite retrievals of atmospheric and surface parameters over continental surfaces (Chairs: Thomas Schmugge and José Sobrino)
Jesse Meng, Kenneth Mitchell, Helin Wei, Dan Tarpley, and Istvan Laszlo
NOAA/NCEP/EMC, Camp Springs, MD, USA
Abstract: Land surface skin temperature (LST) is a critical boundary condition for meteorological and hydrological processes and is an indicator of soil moisture, vegetation health and energy exchange between the land and the atmosphere. Accurate spatial and temporal knowledge of LST has the potential to significantly improve estimates of soil moisture, latent, sensible, and ground heat fluxes and to help environmental monitoring and prediction. Operational Numerical Weather Prediction (NWP) modeling and assimilation systems are found to have non-trivial bias in LST, leading to errors in predicting the boundary layer structure of air temperature, humidity, and wind speed. LST retrieved from remote sensing platforms also are of poor accuracy due to uncertainties in surface emissivity and reflectivity which provides crucial information in surface emission and cloud screening. Therefore, it is yet suitable for operational assimilation systems. The offline uncoupled Land Data Assimilation System (LDAS) is motivated to utilize state-of-the-art land surface models, observation-based forcing, and advanced satellite data assimilation techniques to generate enhanced land surface states and fluxes for further weather, climate, and hydrology assessments. Improved LST simulation is one of the focuses among the LDAS objectives.
In this study, LST derived from various sources are compared. The data include:
- - NCEP operational NWP modeling and assimilation systems of GFS/GDAS and NAM/NDAS;
- - Global and North-American LDAS;
- - GOES retrieval;
- - NOAA SURFRAD and ATDD observations.
The analysis will be made on various spatial and temporal scales, from point to regional, and from diurnal cycle to monthly averages. This work has been performed on regular basis and has been used by developers from NOAA NCEP and NESDIS to monitor the performance of their operational products.
Lihang Zhou1, Chris Barnet2, and Mitchell Goldberg2
1QSS Group at NOAA/NESDIS, Camp Springs, MD, USA
2NOAA/NESDIS, Camp Springs, MD, USA
Abstract: The operational AIRS emissivity retrieval uses NOAA regression emissivity product (Goldberg et al., 2003) as a first guess over land. The NOAA approach is based on clear radiances simulated from the ECMWF forecast and a surface emissivity training dataset. The training dataset used for the AIRS v4.0 algorithm had a limited number of soil, ice, and snow types and very little emissivity variability in the training ensemble. An updated regression coefficient set has been generated using a number of published emissivity spectra (12 spectra for ice/snow, 14 for land) that were blended randomly together for land and ice respectively. Statistics as well as example emissivity spectra and maps will be shown from v4.0 and the new system.
Robert O. Knuteson, University of Wisconsin-Madison, USA
Abstract: Techniques for obtaining accurate land surface skin temperature and infrared emissivity will be demonstated using AIRS observations and implications for the use of IASI and CrIS observations will be identified. Validation of measurements using ground based and aircraft observations will also be presented. Results from a site in the U.S. Southern Great Plains and from sub-Saharan Africa will be highlighted.
F. Karcher, V. Ferreira
Météo-France CNRM, Toulouse, France
Abstract: The ozone column products that we are developing for the " Ozone Monitoring Satellite Application Facility " of Eumetsat use observations of the nadir radiances of the Earth and atmosphere system collected from the Meteosat Second Generation and Metop platforms. Over the subtropical desert regions, the products experience unrealistic total ozone diurnal variations that we assign to surface spectral emissivity differences between the 9.7 and the 11 micron spectral regions.
The retrieval method assumes, as a first approximation, that the upwelling radiance at the tropopause level can be considered in the 9.7 micron band as the blackbody radiance corresponding to the atmospheric window radiance near 11 micron.
This assumption suffers from differences in tropospheric absorption in the 2 above mentioned channels, but also from surface emissivity differences that may be enhanced by hot surface skin temperatures. While the former difference can be corrected by information extracted from other IR low resolution channels, the latter correction is related to the geographical variation of surface properties.
The impact of surface properties on the radiance upwelling at the tropopause level has been modelled using the RTTOV fast radiative code and we are currently investigating a simple formulation for a correction that could be used for the determination of ozone using the observation of the 9.7 micron band as a whole. The use of MSG observations should help the determination of a high resolution atlas over the Sahara for surface properties pertinent for the ozone column determination.
Filipe Aires1, Catherine Prigent2, Bill Rossow3
1CNRS, Université Paris VI, France
2CNRS, Observatoire de Paris, France
3NASA/GISS, New York, USA
Christian Melsheimer, Georg Heygster, and Nizy Mathew
Institute of Environmental Physics, University of Bremen, Germany
Abstract: The polar regions belong to the regions of which the least information is available about the current and predicted states of surface and atmosphere. We present advances in a method to retrieve the total (column) water vapour (TWV) of the polar atmosphere from spacebourne microwave radiometer data, in particular data from the sensor AMSU-B (Advanced Microwave Sounding Unit B) on the new generation polar orbiting satellites of NOAA (National Oceanic and Atmospheric Administration), NOAA-15, NOAA-16, and NOAA-17.
The starting point of our TWV retrieval is the algorithm first proposed by Miao (2001): Using the three AMSU-B channels centred around the 183 GHz water vapour line and the window channel at 150 GHz, TWV can be retrieved independent of the surface emissivity, provided the surface emissivity in these four channels is the same. The method works up to TWV values of about 6 to 7 kg/m2; higher water vapour contents cannot be retrieved. While in Antarctica, TWV is below that limit most of the year, in the Arctic, this is true only for autumn, winter and spring.
In order to extend the retrievable range to higher water vapour values, the window channel at 89 GHz might be used -- it has some sensitivity to water vapour. However, the emissivity of sea ice and ocean at 89 GHz is significantly different from the emissivity at 150 and 183 GHz, and thus the key assumption of the original algorithm is not valid any more. It is nevertheless possible to develop an "extended" algorithm which needs additional information on surface emissivity at 89 and 150 GHz. We have extracted this information from emissivity measurements over sea ice and open water during the SEPOR/POLEX (Surface Emissivities in Polar Regions-Polar Experiment) campaign.
The resulting algorithm can retrieve TWV up to about 15 kg/m2, butwith considerably reduced accuracy with respect to the original algorithm. It now allows to monitor the TWV fields over the centralArctic sea ice during most of the year. They can show details which might be missed out by standard analysis data. There is still room for improvement since the emissivity measurements during the SEPOR/POLEX campaign were restricted to winter, thus the method could benefit form, e.g., measurements or models of sea ice emissivity during summer melt.
Nizy Mathew, Christian Melsheimer, Georg Heygster
Institute of Environmental Physics, University of Bremen, Germany
Abstract: Polar regions play an important role in global variability and change. The regions are composed of different surface types, e.g., open water, different types of sea ice and land ice. However the retrieval of near surface profiles in polar regions is restricted due to the high and highly variable surface emissivity except for open water. Fore example the emissivity of first year ice surfaces is often greater than 0.9 although it can be much lower during melting.
The surface emissivities are retrieved using the brightness temperatures at window channels (23.8, 31.4, 50.3, 89 and 150 GHz) of the AMSU (Advanced Microwave Sounding Unit) instrument on new generation satellites of the National Oceanic and Atmosphere Administration (NOAA) since these measurements are least affected by the atmospheric absorption and emission. The calculations are done under the assumption of flat and specular surface and a clear atmosphere. In the surface emissivity retrieval method, brightness temperatures are simulated at the top of the atmosphere assuming surface emissivities 0 and 1 using atmospheric profiles, e.g., from ECMWF as input for a radiative transfer model. The simulated brightness temperatures are compared with the satellite measured ones in order to retrieve the emissivity. The retrieval is done over both the Arctic and the Antarctic and for different seasons. The angular and frequency dependence of surface emissivities are studied.
Polar temperature profiles are retrieved from AMSU brightness temperatures using an iterated minimum-variance algorithm including a surface emissivity model. An a-priori emissivity spectrum is assigned to the surface emissivity model to each of the eight different surface types classified and used an analytic function of frequency with a small number of free parameters to represent the frequency dependence of surface emissivity. The results are compared with the collocated radiosonde observations. The retrieval errors are quantified. The sensitivity to errors in prescribed emissivity in the algorithm for the temperature retrieval in the near surface part is analyzed.
P. de Rosnay1, J-P Wigneron2, Escorihuela, M.J.1, T. Holmes2,3, Y. Kerr1, J.C. Calvet4
1CESBIO (CNRS, CNES, IRD, UPS), Toulouse, France
2INRA, EPHYSE, Villenave d'Ornon, France
3NASA/Goddard Space Flight Center, Greenbelt, USA
4Météo-France/CNRM, Toulouse, France
Abstract: An accurate value of the effective temperature is critical for soil emissivity retrieval, and hence soil moisture content retrieval, from passive microwave observations. Computation of the effective temperature needs fine profile measurements of soil temperature and soil moisture.
The availability of a three-year long data set of these surface variables from SMOSREX (Surface Monitoring Of the Soil Reservoir Experiment) makes it possible to study the features of the effective temperature at L-band and the IR temperature of the surface at the seasonal to the inter annual scale. This paper reviews the main parameterizations of the effective temperature which have been proposed in the literature by Choudhury et al. (1982), Wigneron et al. (2001) and Holmes et al. (2006). An inter-comparison and validation of the effective temperature models at the inter-annual time scale is performed. Based on the SMOSREX data set, the paper investigates the effective temperature dependency to soil moisture and the effects on the retrieved soil emissivity. The 3-year long SMOSREX data set is used to investigate the features of the L-band emissivity at both vertical and horizontal polarizations, under various conditions of soil moisture soil temperature, and soil roughness.
Alain Royer1, Arnaud Mialon1,2, Michel Fily2, Ghislain Picard2
1Centre d'applications et de recherches en télédétection, Université de Sherbrooke, Sherbrooke Québec, Canada
2Laboratoire de glaciologie et géophysique de l'environnement, Grenoble, France
Abstract: Observing boreal and sub-polar ecosystems is important as they are suspected to evolve significantly in response to the expected increase in temperature for the next decades. This project deals with the development of a new method to derive the surface temperature over the northern latitude areas (>45°N) from satellite microwave measurements. The processed data are derived from the daily SSM/I (Special Sensor Microwave Imager) 25 km by 25 km resolution EASE-Grid Dataset, provided by the NSIDC, Boulder, CO.
Decoupling temperature and emissivity in the measured brightness temperature at 19 and 37 GHz is based on a linear relationship between the emissivities in the vertical and horizontal polarizations over snow- and ice-free land surfaces (Fily et al., 2003, RSE, vol. 85). As the seasonal behavior of the snow cover extent dynamics is strongly variable, the snow- and -ice contaminated pixels have been carefully determined on a pixel-by-pixel basis, using a normalized difference brightness temperature Index [DTb = (37V-19V)/19V]. The proposed approach allows deriving two surface geophysical parameters: the fraction of water surface (FWS) from emissivities and the surface temperature. As passive microwaves are very sensitive to liquid water, the emisivity is related to inundated surfaces (open water, small lakes and reservoirs), wetlands (bogs, swamps associated with low vegetation) and to a lesser extent, soil and vegetation humidity. The second parameter is a near-surface temperature characterizing the surface and the air above the ground over vegetated and forested areas. A method to normalize the temperature is presented to overcome the variation of the time acquisition of data (variable overpass time, differences between satellites and satellite drift), as well as to interpolate missing data. The normalization method uses a diurnal cycle model derived from the ECMWF ERA40 re-analysis (2.5° resolution dataset), which is fitted to the two satellite measurements available per day. The derived parameter is thus a consistent hourly series of temperature during the summer (without snow), The mean accuracy is of the order of 2.5 - 3 K compared with synchronous in situ air temperature, and other gridded datasets.
The results are discussed according to the methodology used (assumptions, sensitivity to the atmosphere, limitations). The derived seasonal dynamics and interannual variations of both FWS (wetland and inundated surface) and temperature are briefly shown for the 1992-2002 period over Canada and Alaska (Mialon et al., 2005, JGR; Mialon et al., 2006, JAM, submitted).