Data fusion methods are usually classified into three levels: pixel level (ikonic), feature level (symbolic) and knowledge or decision level. Here, we will focus on the development of ikonic techniques for image fusion. Image transforms such as the Intensity-Hue-Saturation (IHS) or Principal Component (PC) transform are widely used to fuse panchromatic images of high spatial resolution with multispectral images of lower resolution. These techniques create multispectral images of higher spatial resolution but usually at the cost that these transforms do not preserve the original color or spectral characteristics of the input image data. In this study, a new method for image fusion will be presented that is based on filtering in the Fourier domain. This method preserves the spectral characteristics of the lower resolution mul-tispectral images. Examples are presented for SPOT and Ikonos panchromatic images fused with Landsat TM and Iko-nos multispectral data. Comparison with existing fusion techniques such as IHS, PC or Brovey transform prove the su-periority of the new method. While in principle based on the IHS transform (which usually only works for three bands), the method is extended to any arbitrary number of spectral bands. Using this approach, this method can be applied to sharpen hyperspectral images without changing their spectral behavior.

Multi-spectral and panchromatic remote sensing images fusion is discussed. High spatial resolution panchromatic image can provide detail geometric features ,while multi-spectral image can provide very good spectral information. A high spatial resolution multi-spectral image can be obtained by combining these two images. Multi-wavelets can be simultaneously orthogonal, symmetric, short supported and of high vanishing moments, all these properties can be used to maintain spatial information of panchromatic images and spectral information of multi-spectral images. A multi-wavelet transform based fusion scheme is presented in this paper. Experiment results show that the multi-wavelet transform based fusion method is better than the ordinary wavelet transform based fusion method and is superior to IHS fusion method and PCA fusion method. The fused image can provides more spatial information.

Automation of image classification is a challenge to the image interpretation community. One of the most time consuming task is certainly the collection of training data. The introduction especially of low cost Global Positioning System (GPS) receivers and the higher accuracy of the GPS signal after turning-off the Selective Availability has enhanced the ease and versatility of spatial data acquisition. Has also made the approaches by which it is integrated with GIS and remote sensing data more flexible. The emphasis of this paper is to present a method for improving the training data collection for classification purposes using remotely sensed imagery as well as various data sources combined in a GIS. Firstly, a methodology for defining and describing training areas is demonstrated. The training data were stored in a vector data base (shape files) by using the geometry of the land parcels of the test site. Secondly, in addition to a conventional field mapping approach, an advanced GPS based field mapping methodology was used to collect new training data. Within this new approach single point information of the target ground truth class were collected along the roads in this test area. In this step, the following attributes were recorded: ID, left or right of the street and biotope class. The goal for this approach is that one single person should handle the field mapping while driving in a car. The implementation of this approach is performed in ArcPad 6.03 and Application Builder from ESRI. The standard version of ArcPad was modified so that a one hand collection of training data is possible. After the field survey, the results were used within Erdas Imagine (version 8.7). In our approach all "left" points were moved to the adjacent left field - "orthogonal" to the street. All "right" points were shifted to the adjacent right field. Now those moved points were used as a seed pixel in a Euclidian distance algorithm to automatically derive new training sets. In conclusion it can be stated that the proposed training collecting method and its implementation have been proved to be a very valuable and reliable method for image classification purposes.

The motivation for data fusion is to reduce the limitations and uncertainties associated with data coming from a single sensor only. In the context of remotely sensed data the fusion is often performed by combining high spatial with high spectral resolution imagery at different levels. In contrast to pixel-based approaches like the IHS-transformation, in this paper we will focus on a fusion of data at the feature level. In high spatial resolution data the geometry of urban objects can be determined very accurately. But high spatial resolution data often contains low spectral information such as, for example, a three band RGB image. Thus, similar feature values for thematic classes like water, dark pavements or dark rooftops lead to classification errors. If hyperspectral data is used to classify urban materials, the definition of endmembers representing those materials is needed. The problem is that endmembers representing urban surface types are often the result of a mixture of spectral pure materials which leads to flat spectra. Consequently, those thematic endmembers can hardly be detected by standard algorithms like the Pixel Purity Index (PPI) so that standard classification procedures fail. In order to improve the classification process, our approach fuses hyperspectral data recorded by the HyMap sensor with high spatial resolution imagery (digital orthophotos) for a combined endmember selection, classification, and structural analysis. The endmember for the thematic classes will be determined in a semi-automatic process. After a segmentation of the high spatial resolution dataset the resulting segments will be used to detect those pixels in the hyperspectral data sets, which represent candidates for the definition of thematic endmembers. The endmembers are stored in a spectral library and are used for the classification of hyperspectral data. The segments in the high spatial resolution data will be processed based upon the classification of the hyperspectral dataset and the application of overlay rules.

In this paper we present the use of ASTER data for the creation of a Digital Terrain Model (DMT) of high accuracy. The area of study is Milos Island in the Aegean Sea. For the creation of the DTM we used an ASTER stereo-pair dating August 07, 2002. The images cover an area of 60X60 km. The images have been received in the near infrared (0,78-0,86μm) part of the spectrum with a spatial resolution of 15m. We created a DMT a 30m-pixel size. Then we compared the DTM with the following DTM's: a) A GTOPO 30 DTM b) A DTM created from digitized contours of 1/50.000 scale topographic maps. We first made an optical comparison of the DTM's. Then, we proceed to statistical control of the histogram values of the three DTM's. From the first two controls we conclude that the GTOPO30 DTM is less accurate than the other two DTM's and we proceed in more detailed control only for the DTM from the digitized contours and for the DTM from the ASTER stereo-pair. We verified the DTM's accuracy using different points of well-known elevation. All the results demonstrated that DTM's derived from ASTER data have better accuracy than the other two DTM's.

A land use/cover classification system is important for land resource management and it is one of the key research issues in Land Use /Cover Change (LUCC) and land change science research. Lots of work has been done in this area, however, a universally-accepted classification system has not been available yet. This paper proposed several guidelines for building a land use/ cover classification system, which encompasses the basic concept of land use and land cover. Then a preliminary framework for Chinese land use/cover classification system at different scale based on remote sensing data was detailed. The framework is made up of four scales, including national scale, regional scale, county scale and country scale. The classes of first level at national scale in the system are agriculture land, woodland, natural grassland, built-up land, water, wetland and barren land. The regional scale includes 27 classes of land use/cover and the county scale includes 43 types. The general diagnostic criterion of the first level of classification system is the situation of existing vegetation, soil and water, artificial and natural surface. Monitoring on land resource in Beijing-around area as an example, this paper introduce the respects need to be paid attention of this classification system. Because of the complexity and difficulty of this question itself, this system was based on synthesis of relevant research achievement; its actual feasibility still remains to be verified.

The ECMWF Re Analysis (ERA-40) refers to the rerun of the European Centre of Medium Range Weather Forecast (ECMWF) Numerical Weather Prediction (NWP) model for the period September 1957- August 2002 employing all state-of-the-art information and satellite data input presently available. A selection of the reanalysis atmospheric output data can potentially be used to run a hydrological model to simulate historic river flows for the whole of Europe. Once evaluated against observed time series of rainfall and river flow, the output would constitute an extensive and coherent 40+ year database of pan-European calibrated river flow time series, providing a wealth of information and allowing a range of evaluation possibilities. Here, in order to separate the meteorological model performance from the hydrological model performance, the ERA-40 near-surface rainfall aggregates, which come as a by-product of the ECMWF NWP system, are evaluated against interpolated fields of observed surface precipitation. The ERA-40 rainfall fields consist of forecast data with a 36 hr lead time at midday and midnight and a 6 hr lead time at 6:00 and 18:00 UTC, allowing different combinations of lead and base times to compute daily rainfall aggregates. The evaluation of these aggregated precipitation fields against observed totals is relevant to the spin-up time of the ECMWF NWP system and the forecast reliability with increasing lead time. Interpolated fields of observed daily rainfall totals are provided by the Monitoring Agriculture with Remote Sensing (MARS) database (1990-2001) based at the European Commission Joint Research Centre (JRC).

Weather Satellite data has great potential for Precipitation forecast which plays an important role in flood disaster monitoring. In this paper, the GMS-5 infrared cloud imagery combined with surface temperature data for two years in Binjiang reaches of Guangdong province in China is used to study the relationship between infrared cloud imagery and surface rainfall rates. First, parameterization estimate of infrared cloud imagery is made one the base of atmospheric probing principle, then some parameterization estimate result have been obtained under different analysis field from 3×3 to 15×15 pixels. The result shows:1 there exist obvious correlation between the probability of rain and parameterization estimate such as average brightness temperature(Tb), brightness temperature variance(fc), equivalent cloudage(CN),brightness temperature area index(A1--the first A5--the fifth grade, A6-the sixth grade );2 The rainfall intensity increase with Tb and f and CN, and that it decrease with Tb and A1.Finally,the prediction empirical formula of rainfall intensity has been established by means of optimized subclass regression under different analysis field. The following formula is made under analysis field of 11×11 pixels. The statistical result shows that the average precision of rainfall intensity is about 80% using infrared cloud imagery parameters and the size of analysis field has slight effect on it. If the rainfall intensity reached the storm standard, the flood alarm would be sent out.

The National Assembly for Wales (NAW) is responsible for monitoring the effects of dredging for fine aggregate from sandbanks off the coast of South Wales. A key monitoring objective is the analysis of changes to the sandbank bathymetry and the adjacent coastline. This paper reviews the monitoring strategy, with a particular emphasis on the use of laserscanning with LiDAR over the last six years for large-scale topographic beach mapping and analysis. The focus is on the methodologies that were implemented in order to make the data compatible, consistent and usable within a geographical information system (GIS). The issues that are addressed include data handling strategies; automatic error/blunder detection of spurious data; identifying sources of errors; projection and datum transformations; LiDAR artefacts; quality control; choice of digital terrain model and spatial resolution; choice of interpolation algorithm; the calibration of LiDAR surveys to ensure consistency; and LiDAR accuracy compared with land surveys. Some of these issues have proved problematic, which if not correctly resolved, can produce significant application errors, thus reducing confidence in this technology. The paper concludes with some examples of the analyses undertaken to date.

In many European countries air- and spaceborne remote sensing data is operationally used for oil spill monitoring. For the Baltic Sea the yearly results of the aerial surveillance are collected by the Helsinki Commission (HELCOM) and for the North Sea by the Bonn Agreement Secretariat. To improve knowledge of the oil spill situation in the North Sea and the Baltic Sea these data sets were analyzed and visualized. If a geographical phenomenon may be reasonably modeled as point data is largely a question of scale. Oil spills on a sea basin level (North Sea, Baltic Sea) can be considered as point data. During the analysis we are essentially looking for patterns in the data. However, to combine oil spill surveillance results from different countries the data must be standardized. For standardization purposes it is important to have information about the surveillance effort, which means for instance the number of pollution control flights per year and the area covered or the number of acquired and analyzed satellite. Special incidents (e.g. accidents) may lead to a multitude of oil spills in a certain period of time which has to be considered during the analysis. The data was visualized taking into account additional information and available information about the surveillance activities. Kernel estimation was used to calculate oil spill density estimation. First results are promising. The strongest impediment is the unavailability of information for data standardization.

In the paper a methodology of RS-based thematic mapping is introduced which uses an original RS imagery interpretation approach. The implementation of the methodology is based on application of GIS MapInfo Professional and original imagery processing and interpretation system "LandMapper" developed in Tomsk Polytechnic University (TPU). The paper considers the basic principles of imagery interpretation approach adopted in the "LandMapper" system as well as gives the results of its application for Tomsk region oil-fields pollution mapping with use of high resolution images acquired by QuickBird satellite.

Airborne imaging spectrometers have a history of about 20 years starting with the operation of AIS in 1982. During the following years, many other instruments were built and successfully operated, e.g., AVIRIS, CASI, DAIS-7915, and HyMap. Since imaging spectrometers cover a spectral region with a large number of narrow contiguous bands they are able to retrieve the spectral reflectance signature of the earth allowing tasks such as mineral identification and abundance mapping, monitoring of vegetation properties, and assessment of water constituents. An essential prerequisite for the evaluation of imaging spectrometer data is a stable spectral and radiometric calibration. Although a considerable progress has been achieved in this respect over the last two decades, this issue is still technically challenging today, especially for low-to-medium cost instruments. This paper introduces a new airborne imaging spectrometer, the ARES (Airborne Reflective Emissive Spectrometer) to be built by Integrated Spectronics, Sydney, Australia, and co-financed by DLR German Aerospace Center and the GFZ GeoResearch Center Potsdam, Germany. The instrument shall feature a high performance over the entire optical wavelength range and will be available to the scientific community from 2006 on. The ARES sensor will provide 150 channels in the solar reflective region (0.47-2.42 μm) and the thermal region (8.1-12.1 μm). It will consist of two co-registered optical systems for the reflective and thermal part of the spectrum. The spectral resolution is intended to be between 12 and 16 nm in the solar wavelength range and should reach 150 nm in the thermal range. ARES will be used mainly for environmental applications in terrestrial ecosystems. The thematic focus is thought to be on soil sciences, geology, agriculture and forestry. Limnologic applications should be possible but will not play a key role in the thematic applications. For all above mentioned key application scenarios, the spectral response of soils, rocks, and vegetation as well as their mixtures contain the valuable information to be extracted and quantified. The radiometric requirements for the instrument have been modeled based on realistic application scenarios and account for the most demanding requirements of the three application fields: a spectral bandwidth of 16 nm in the 0.47-1.8 μm region, and 12 nm in the 2.02 - 2.42 μm region. The required noise equivalent radiance is 0.05, 0.03, and 0.02 Wm-2sr-1μm-1 for the spectral regions 0.47- 0.89 μm, 0.89 - 1.8 μm, and 2.02 - 2.42 μm, respectively. In the thermal region similar simulations have been carried out. Results suggest a required noise equivalent temperature of 0.05 K for the retrieval of emissivity spectra in the desired accuracy. Nevertheless, due to system restrictions these requirements might have to be reduced to 0.1 K in the wavelength range between 8.1 and 10 μm and 0.1-0.2 K from 10 to 12.1 μm.

This paper describes an Airborne Multi-Spectral Imaging System (AMIS) and the development of its system software. This system has been developed so as to be rapidly deployed in response to episodic events such as hurricanes and tropical storms which may occur year round in coastal zones. The system uses digital video cameras to provide high resolution images at a very high collection rate. The system is software controlled so as to provide a minimum distraction for the aircraft pilot by providing for the remote manipulation of the camera and the GPS receiver. The system is viable for many applications that require good resolution at low cost. Such applications include vegetation detection, oceanography, marine biology, and environmental coastal science analysis.

The infrared sensor system FASA (Fire Airborne Spectral Analyser) operated in a DLR research aircraft is a unique combination of an imager and a Fourier Transform Spectrometer used to detect, analyse and classify fires and volcanoes. In such cases, the scenes in the FTS field of view are generally inhomogeneous and not known a priori, and the data fusion approach of image and spectral data is supposed to overcome the difficulty hereby involved. The analysis of an artificial coal fire has provided us fundamental insights in how to model such fires and potential strategies for data evaluation. First and promising results will be presented. Moreover, the importance of aerosol contributions is examined and the feasibility to retrieve smoke particle parameters in the infrared is demonstrated.

This study was aimed at developing a new sensitive approach to account for small sediment dust particles using spectral reflectance across the shortwave spectral region (1250-2400 micron). The NIRA (Near Infrared Analyses) approach was adopted in order to examine its capability to predict gravimetric weight of sediment dust particles solely from the reflectance data. In order to quantitatively characterize the dust loading process, two model composition mixtures representing homogeneity (talc powder) and heterogeneity (Environmental Protected Agency (EPA) dust) of chemical compounds were examined. A wind tunnel was constructed and used to simulate the different amounts of dust loadings over an indoor environment. Different spectral manipulations most commonly used to analyze spectral data were tested. On these manipulated spectra, a multivariate data analysis based on Partial Least Squares (PLS) regression was run and prediction modeling between NIR spectroscopy and the dust loadings was generated. For this purpose, the relationship between spectroscopic measurements and the total gravimetric weight was used. Using reflectance values in the PLS analysis was found to demonstrate the best performance in EPA dust relative to other manipulations employed (with RMSEP of 4.8%). For the talc dust, the first derivative of absorbance manipulation was found to demonstrate the best performance relative to other manipulations with RMSEP of 5.4%. Although the RMSEP might seem somewhat high, one should note that this concerns a relatively small amount of dust with a narrow gravimetric weight of ±0.0001 g. Moreover, validation and examination tests applied to the population studied have presented very significant results. This method can be further used to assess very small amounts of dust in indoor environments and accordingly to identify shade on the environmental air quality on regular non dusty-days.

As part of a wider study of carbon cycling in boreal peatlands, radar remote sensing was used with the objective of obtaining diverse environmental information related to these environments. An analysis of multi-temporal Fine beam mode RADARSAT-1 images was carried out, with the support of collected field data, to verify if water table height and volumetric water content influenced radar backscatter coefficient. A maximum likelihood classification (MLC) on speckle filtered and textural images was also carried out, evaluated and compared with a similar classification procedure on Standard beam mode images. Significant changes in water table position and soil moisture have been observed but these were not reflected in radar backscatter coefficient. C-band wavelength, shallow incidence angle and high volumetric water content of peat are some factors that would limit hydrological conditions monitoring with Fine beam mode images. Further analyses have to be done in order to confirm these conclusions. MLC classification using textural images generated from multi-temporal Fine beam mode images brought poorer results than a similar classification using multi-temporal Standard beam mode textural images. This can be explained by the lower radiometric resolution of Fine beam mode images. If only radar imageries are available for boreal peatland mapping, Standard beam mode images should be used, even if they have a lower geometric resolution.

The island Tenerife is characterized by an increasing tourism, which causes an enormous change of the socio-economic situation and a rural exodus. This development leads - beside for example sociocultural issues - to fallow land, decreasing settlements, land wasting etc., as well as to an economic and ecological problem. This causes to a growing interest in geoecological aspects and to an increasing demand for an adequate monitoring database. In order to study the change of land use and land cover, the technology of remote sensing (LANDSAT 3 MSS and 7 ETM+, orthophotos) and geographical information systems were used to analyze the spatial pattern and its spatial temporal changes of land use from end of the 70s to the present in different scales. Because of the heterogeneous landscape and the unsatisfactory experience with pixel-based classification of the same area, object-oriented image analysis techniques have been applied to classify the remote sensed data. A post-classification application was implemented to detect spatial and categorical land use and land cover changes, which have been clipped with the socio-economic data within GIS to derive the driving forces of the changes and their variability in time and space.

Digital change detection using multi-temporal remotely sensed imagery is a key topic in the studies of the global environmental changes. Significant efforts have been made in the development of methods for digital change detection. Among the methods, the multivariate alteration detection (MAD) shows great promising. However, the use of mean and covariance matrix of feature vectors in the method makes the detection non-robust because the mean and covariance matrix are influenced by the presence of outliers. In this article two schemes are proposed to improve the robustness of the MAD method. The two schemes, based on different strategies of outlier handling, consist of a two-pass and a one-pass processing, respectively. Finally a preliminary study was carried out to evaluate the feasibility and effectiveness of the proposed schemes.

Detection of changes in remotely sensed geographic images is required for a variety of applications including natural disasters. Change detection is an important process utilized for updating the geographic information system (GIS) data, monitoring natural resources and urban developments. It provides quantitative analysis of the spatial distribution of the area of interest. Different types of change detection techniques include: Multi-date visual composite (multitemporal composite), image differencing, post-classification, etc. In this work, we propose a change detection algorithm based on multiresolution analysis and motion estimation. We use multispectral satellite imagery and apply the dual tree complex wavelet transform to get images ensembles. We have also compared our proposed algorithm with some of the efficient methods reported in the literature. Experimental results are given using the IRS images of the Bam city before and after the earthquake.

The Kam Kotia mine tailings generated acidic mine drainage waters which killed large areas of adjacent forest and badly damaged surrounding ecosystems over a 30-year period. With the start of the site's rehabilitation in 2001, a remote sensing-based monitoring program was initiated. In this first phase, the baseline study was carried out to come up with a method to delineate the Kam Kotia mine tailings areas into distinct zones, which enable the monitoring of the rehabilitation status. This study was based on airborne hyperspectral TRWIS III imagery. Data pre-processing included the retrieval of surface reflectance and corrections of radiometric and spectral errors. After application of a destriping procedure, surface reflectances were retrieved indicating a varying across-track wavelength shift, known as the spectral smile. The detection and correction of this phenomenon used an algorithm based on the comparison of measured and modeled at-sensor radiance data within certain wavelength regions. Finally, sensor calibration problems required a scene-based radiometric calibration performed on the destriped and spectrally corrected reflectance data. The subsequent spectral unmixing analysis included an iterative error analysis (IEA) technique to automatically extract endmembers from the data. The resulting fraction images were first grouped into three major surface materials (vegetation, vegetation residues and oxidized tailings). Three boundaries were determined, delineating the three surface materials and a less vegetated transition by subdividing the entire tailings area into four distinct zones. The area change of each zone and the course of the boundaries in future data sets will provide information of the site's rehabilitation status.

Field and airborne hyperspectral data was used to map residual contamination after a mining accident, by applying spectral mixture modelling. Test case was the Aznalcollar Mine (Southern Spain) accident, where heavy metal bearing sludge from a tailings pond was distributed over large areas of the Guadiamar flood plain. Although the sludge and the contaminated topsoils have been removed mechanically in the whole affected area, still high abundance of pyritic material remained on the ground. During dedicated field campaigns in two subsequent years soil samples were collected for geochemical and spectral laboratory analysis and spectral field measurements were carried out in parallel to data acquisition with the HyMap sensor. A Variable Multiple Endmember Spectral Mixture Analysis (VMESMA) tool was used providing possibilities of multiple endmember unmixing, aiming to estimate the quantities and distribution of the remaining tailings material. A spectrally based zonal partition of the area was introduced to allow the application of different submodels to the selected areas. Based on an iterative feedback process, the unmixing performance could be improved in each stage until an optimum level was reached. The sludge abundances obtained by unmixing the hyperspectral spectral data were confirmed by the field observations and chemical measurements of samples taken in the area. The semi-quantitative sludge abundances of residual pyritic material could be transformed into quantitative information for an assessment of acidification risk and distribution of residual heavy metal contamination based on an artificial mixture experiment. The unmixing of the second year images allowed identification of secondary minerals of pyrite as indicators of pyrite oxidation and associated acidification.

The goal of this study is to develop remote sensing desertification indicators for drylands, in particular using the capabilities of imaging spectroscopy (hyperspectral imagery) to derive soil and vegetation specific properties linked to land degradation status. The Cabo de Gata-Nijar Natural Park in SE Spain presents a still-preserved semiarid Mediterranean ecosystem that has undergone several changes in landscape patterns and vegetation cover due to human activity. Previous studies have revealed that traditional land uses, particularly grazing, favoured in the Park the transition from tall arid brush to tall grass steppe. In the past ~40 years, tall grass steppes and arid garrigues increased while crop field decreased, and tall arid brushes decreased but then recovered after the area was declared a Natural Park in 1987. Presently, major risk is observed from a potential effect of exponential tourism and agricultural growth. A monitoring program has been recently established in the Park. Several land degradation parcels presenting variable levels of soil development and biological activity were defined in summer 2003 in agricultural lands, calcareous and volcanic areas, covering the park spatial dynamics. Intensive field spectral campaigns took place in Summer 2003 and May 2004 to monitor inter-annual changes, and assess the landscape spectral variability in spatial and temporal dimension, from the dry to the green season. Up to total 1200 field spectra were acquired over ~120 targets each year in the land degradation parcels. The targets were chosen to encompass the whole range of rocks, soils, lichens, and vegetation that can be observed in the park. Simultaneously, acquisition of hyperspectral images was performed with the HyMap sensor. This paper presents preliminary results from mainly the field spectral campaigns. Identifying sources of variability in the spectra, in relation with the ecosystem dynamics, will allow the definition of spectral indicators of change that can be used directly to derive the desertification status of a land.

The HRSC-A (High Resolution Stereo Camera - Airborne) systems are multiple line "pushbroom" instruments developed by the German Aerospace Center (DLR). In the image data of the blue and green band the effects of atmospheric scattering are most prominent. The correction of atmospheric absorption is not addressed here. When strong atmospheric effects are observed, a simple atmospheric correction can be applied. This is done for each individual flightline and each channel separately, based on view-angle dependent histogram statistics. For the correction it is assumed that due to the small pixel size and the large area a significant number of pure shade pixels with a theoretical signal of 0 DN are covered across the whole swath. An across-track Dark-Pixel-Profile (DPP) is used to determine the view angle dependent correction offset. The shape of these profiles varies with the phase angle. This simple atmospheric correction has successfully been tested in several projects, clearly increasing the image homogeneity of the individual flightlines and also improving the quality of the resulting image mosaics. Local atmospheric effects within the flightlines (e.g. due to topography, adjacency effects) are not corrected.

Air qualities studies in urban areas embrace several directions that are strongly associated with urban complexity. In the last centuries cities evolution implied changes in urbanization trends: urban sprawl has modified the relationship between cities and surroundings settlements. The existence and protection of urban green and open areas is promoted as a mean to improve the quality of life of their citizens and increase the satisfactory level of the inhabitants against pollution and noise adverse effects. This paper outlines the methods and approaches used in the EU research project Benefits of Urban Green Space (BUGS). The main target of BUGS is to assess the role of urban green spaces in alleviating the adverse effects of urbanization trends by developing an integrative methodology, ranging from participatory planning tools to numerical simulation models. The influence of urban structures on atmospheric pollutants distribution is investigated as a multi-scale problem ranging from micro to macro/regional scale. Traditionally, air quality models are applied on a single scale, seldom considering the joint effects of traffic network and urban development together. In BUGS, several numerical models are applied to cope with urban complexity and to provide quantitative and qualitative results. The differing input data requirements for the various models demanded a methodology which ensures a coherent data extraction and application procedure. In this paper, the stepwise procedure used for BUGS is presented after a general presentation of the research project and the models implied. A discussion part will highlight the statements induced by the choices made and a conclusive part bring to the stage some insights for future investigations.

Presently no reliable synoptic picture of number, extent, distribution and emissions from mining waste sites exists, neither for EU member states, nor for the Accession and Candidate Countries. At EU level, this information is needed to assess the large range of environmental impacts caused by mining wastes and their emissions in a coherent way across the different policies addressing the protection and sustainable use of environmental resources. The core task lies in the harmonised collection and standardised compilation and evaluation of existing data and in connecting them to a geographical reference system compatible with other European data sets. In the proposed approach information from national registers of mining wastes is linked to related standardized spatial data layers such as CORINE Land Cover (the classes of mineral extraction sites, dump sites) or other data sets available in the EUROSTAT GISCO data base, thus adding the spatial dimension at regional scale. Higher level of spatial detail and distinction between mineral extraction site and waste sites with or without accumulation of potentially hazardous material is added by remote sensing, applying a semi-automated principal component analysis (PCA) to selected spectral channels of geo-referenced Landsat-TM full scenes. The method was demonstrated on large areas covering approximately 120000 km2 of Slovakia and Romania and was validated against mining-related features from Pan-European and/or national databases, detailed geological maps, mineral resource maps, as well as by a GIS analysis showing the distribution of anomalous pixels in the above-mentioned features compared to the main land cover classes.

The aim of this work is to use information from various sources, including remote sensing images from which land use change may be identified, in order to produce landslide hazard maps. We designed a fuzzy neural network which allows us to incorporate all the levels of uncertainty in the informations used in order to draw conclusions about the severity of the landslide hazard. The scale of operation of such a system is at the regional level rather than the local microlevel where ground local measurements may be performed and detailed geotechnical mathematical models may be applied to calculate soil stresses. It is not possible to apply such accurate detailed models for large scale hazard assessment. The proposed system is expected to be less accurate but more widely applicable than what is currently used in geotechnics.

The prediction of subsidence rates and magnitudes is a challenging problem due to the range of complex variables that combine to determine the displacement of the surface. Many subsidence prediction models utilise an approach that involves detailed modelling of mechanical behaviour of strata transferring strain from the underground void to the surface. Such approaches are typically calibrated using subsidence records. Even after this calibration they generally struggle to predict accurately and reliably actual subsidence in virgin terrain. In this paper a model is presented based on an alternative data-driven approach using statistical techniques. This approach utilises past patterns of monitored subsidence to predict future movements at any point in space and time as a consequence of mining activities. Testing of the model proved that 89% of the estimations are between -1.65 mm/year and +1.40 mm/year of the actual subsidence value and 51% of the estimations are between -0.6 mm/year and 0.4 mm/year of the actual subsidence value.

Laser interferometry is one of the most sensitive methods for small displacement measurement. This technique is successfully applied in several fields of physics with very good performances. Its large diffusion is mainly due to the availability of high-quality low-cost optical components and laser systems, but not less important is the general versatility of laser interferometric techniques. In this paper, we present a laser interferometric system for one-dimension local seismic noise measurement. After a description of the theoretical model, the experimental results are presented and discussed, also in comparison with a standard mechanic accelerometers, like the Episensor model FBA ES-T from Kinemetrics. The obtained result are encouraging, so that further studies are scheduled to check the performance of this system at the very low frequencies and to measure its intrinsic noise. Finally, although some problems, arosen in connection with the sensitivity and the stability of the interferometric system, have to be more deeply studied, it is already possible to think to design new versions of this system aimed to the measurement of more degrees of freedom.

Urban environmental quality is an important part of efficient urban environment planning and management. A scientific management system for protection, conservation and restoration must be based on reliable information on bio-geophysical and geomorphologic, dynamics processes, and climatic change effects. Synergetic use of quasi-simultaneously acquired multi-sensor data may therefore allow for a better approach of change detection and environmental impact classification and assessment in urban area. As is difficult to quantify the environmental impacts of human and industrial activities in urban areas , often many different indicators can conflict with each other. The spatial and temporal distribution of land cover is a fundamental dataset for urban ecological research. Based on Landsat TM, ETM, SPOT and SAR data for Bucharest metropolitan area in Romania, it was performed a land cover classification based on spectral signatures of different terrain features used to separate surface units of urban and sub-urban area . A complete set of criteria to evaluate and examine the urban environmental quality, including the air pollution condition indicators, water pollution indicators, solid waste treated indicators, noise pollution indicators, urban green space have been widely used .

During the last decades, land subsidence has occurred in three main river deltas of China, especially in some metropolitans. With the assistance of some geodetic technologies, such as digital leveling, GPS, total station, and so on, repeat-pass radar interferometry has the capability to monitor land subsidence in 3 dimensions with high accuracy. The potential and limitation of repeat-pass radar interferometry are still being evaluated till now. Considering the operational application of this technique, this paper describes applications of D-InSAR the urban subsidence of China. The case study with a small data stack in Hong Kong presents about 1cm height change in two test sites. The two test sites have mainly been investigated, and some of the key steps of this technique have been discussed in detail. The evolution of the D-InSAR and the future research involve more accurate and robust measurement with large data stacks in a long time series.

Natural hazards monitoring and analysis have come to be very important sectors in environmental management. In the last decades, natural slope dynamics of high mountain relief in Aosta Valley has been analyzed by means of detailed geological and geomorphological field surveys. For a long time, remote sensing techniques have been an effective tool within inaccessible locations thanks to their wide analysis. This article intends to present some results coming from investigations on high elevation alpine environments, based on the high resolution hyperspectral airborne sensor MIVIS images. Considering the great problem due to the high geometric distortion of such data, related both to the sudden relief displacement and to the intrinsic whiskbroom recording system, we are willing to show an experimental neural network approach for geometric correction which is a very important item for measurement instances. To enhance geomorphological characterization and hazards studies on large slope instabilities we have successively applied a neural network LVQ 2 (Modified Learning Vector Quantization) self-developed classifier exploiting spectral information coming from images. The application of the method to the Val Veny- Val Ferret area (Mont Blanc zone) allowed a better definition of the active deformational features connected to flexural toppling, deep creep and superficial fracturing along double ridges and steep slopes, eventually preparing for collapse along high rock walls. Hazard scenarios for deep seated gravitational slope deformations of the Upper Aosta Valleys can be more precisely outlined exploiting MIVIS images hyperspectral contents, but first the geometric correction problem has to be preventively solved.

Digital Terrain Models (DTM) represent an effective tool for many applications and in particular for terrain morphology investigation and orthoimages generation. The availability of satellite stereo images allows to generate updated DTMs through digital photogrammetric algorithms especially in those areas where old and poor maps exist. In this work we show some quality tests results about DEMs obtained from ASTER data (15m geometric resolution), elaborated through commercial software. We consider this information very important to understand which kind of applications can reasonably use these data. Shown results refer to a mountain area located in the NW Italian Alps, characterized by different height type regions (flat, hilly, mountain) whose evaluation can drive to consistent results for a better understanding of limits and forces of these data. Such tests, making use of the commercial software AsterDTM, take into consideration both qualitative and quantitative aspects. Height profiles comparisons, statistical analysis on differences and data mining have been carried out in order to evaluate accuracies and to define the nature of possible systematic errors. Reference data is the available Regional DTM (50m x 50m grid, accuracy of 2.5m) and the single height points extracted from technical maps at 1:10000 scale for more precise local investigation.

Prestige fuel oil tanker was damaged during a storm in November 13rd, 2002, close to the coast of Galicia (Spain). After some days the Prestige broke in half and sank, leaking about 40.000 tons of oil which affected more than 1.000 km of the coast in Spain, Portugal and France. Some months later, layers of fuel contamination still appear at different depths in the sand of the beaches. The tidal process is that the first tide brings fuel over the sand but, if it is not removed, following tides place clean layer of sand on the top of fuel, and the beaches appear to be clean. Layers of fuel appears at different depths in the sand, from some cm to 1-2 meters. The lateral extent of the contamination also varies from some cm to more than 1 m. Radar sensors could be used in-situ to detect and imaging fuel layers below sand in some inland areas, which are under the influence of high winter tides but remain out of the influence of salt water from the sea during spring and summer time. This study show some tests carried out on the beaches with a ground-penetrating radar system operating with 500 & 800 MHz nominal frequency antennas, and a study case made in the beach of Carnota (Galicia) where it was possible to detect an imaging a buried fuel layer 6 months later.

Monitoring dust storms with recently available medium and moderate resolution satellites (Meris, Modis and SeaWiFS) is providing new global information regarding the sources, transportation tracks and affected areas. Saharan dust plumes reach the SE region of the United States and the Caribbean region in summer and the Amazon basin in winter. Generally these Saharan plumes branch off in dust tracks along the North Atlantic reaching Western Europe as far north as the Scandinavian countries. Furthermore, dust storms originating in the Eastern Sahara and Northern African deserts form dust plumes propagated by the Sirocco winds that, after crossing the Mediterranean Sea, affect Southern and Central Europe particularly during spring and summer. Dust storms originating in the Gobi and Taklamakan deserts blow in an easterly direction propagating dust plumes affecting Korea, Japan and reach the United States after crossing the Pacific Ocean. The large amount of cyclic deposition generated by dust storms produces an environmental impact that causes the decay of coral reefs in the Caribbean, the origin and distribution of red tides and the disappearance of sea grasses. The relationship of dust plumes with the increasing number of asthma and allergy cases in the Caribbean correlates well with the appearance of similar cases in Europe and elsewhere during the mid 1980s. The recurrence presence of insecticides in regions where these products were banned long ago, or where they were never used, may be partly due to Saharan dust plumes. The loss of agricultural soil, literally blown away by dust storms in the source areas, creates hardship, hunger and forced-migration. Dust storms should be considered as an important security issue.

In Spain there are more than 500,000 ha of Eucalyptus plantations. These represent 3,5% of the national forest and the 25% of the timber harvested. Galicia monocultures of Eucalyptus globulus Labill. plantations cover 177,679 ha, and mixed stands of eucalyptus cover 200,000 ha more. This high productivity has been powered by the absence of pests and pathogens. However, since 1991 the health and productivity of these stands has been threatened by the Eucalyptus snout beetle (Gonipterus scutellatus Gyll.), which causes a severe defoliation to eucalyptus stands in Galicia. The aim of this paper is to establish a workflow to locate the areas affected by the defoliator, and determinate the basics patterns of spatial distribution, in order to predict future hot spots and develop more integrated pests management. This information will be part of a wider Information System, develop to improve the forest management and monitoring of these stands. The damaged area and the level of defoliation will be mapped using satellite imagery. The additional information of stand conditions, such as site index, climate and microclimatic conditions, digital terrain model, dendrometric and dasometric variables, will be integrated also in a Geographical Information System.

A method is developed to integrate topographical (elevation and slope) and climatic (precipitation and temperature) information with multi-temporal VGT images into a coarse scale land cover classification. The Normalized Difference Vegetation Index (NDVI), cumulated NDVI (SNDVI), Normalized Difference Water Index (NDWI) and the cumulated NDWI (SNDWI) were used in a two-step classification approach. The two steps encompass an unsupervised classification based on the ISODATA (Iterative Self-Organizing Data Analysis Technique) method, and a supervised classification based on a dichotomous hierarchical tree classification at the landscape patch scale. Results demonstrate the potential of the integrated method to estimate forest and grassland areas with VGT imagery. The method reduces confusion between different land cover classes with same spectral characteristics and slightly improves classification accuracy. 58% forest and 57% grassland were obtained for the Gansu Province. We suggest two main reasons for the high percentage of land cover misclassification: Confusion of the different land cover classes with same spectral characteristics and the spatial scale of observation unsuited for classifications of a highly fragmented land cover. The integrated data source approach is therefore limited to applications in regional land cover classification. The classification method could be improved in the critical value initialization of the classification tree.

During the last decades, human activities endanger the biological and economic productivity of drylands, observable by processes like soil erosion and long-term loss of vegetation. To identify these changes and underlying driving processes, it is essential to monitor the current state of the environment and to include this information in land degradation models. A frequently used input parameter is the degree of vegetation surface cover, thus there is a demand for quantitative cover estimation of large areas. Multispectral remote sensing has a limited ability to discriminate between dry vegetation components and bare soils. Therefore hyperspectral remote sensing is thought to be a possible source of information when applying adequate preprocessing and specific spectroscopic methodologies. The proposed approach is based on multiple endmember spectral unmixing, where the mixture model is iteratively improved using residual analysis and knowledge-based feature identification. It is believed that this automated methodology can provide quantitative fractional cover estimates for major ground cover classes as well as qualitative estimates of scene components. This apporach is currently tested using HyMap imaging spectrometer data of Cabo de Gata, Southern Spain, and will be adapted to larger areas based on hyperspectral data of future satellite instruments.

In this paper, an object based classification approach for land cover and land use classes is presented, and first test results are shown. Recently, there is an increasing demand for information on actual land cover resp. land use from planning, administration and science institutions. Remote sensing provides timely information products in different geometric and thematic scales. The effort to manually classify land use data is still very high. Therefore a new approach is required to incorperate automated image classification to human image understanding. The proposed approach couples object-based clasification technique -a rather new trend in image classification - with machine learning capacities (Support Vector Classifier) depending on information levels. To ensure spatial and spectral transferability of the classification scheme, the data has to be passed through several generalisation levels. The segmentation generates homogeneous and contiguous image objects. The hierarchical rule type uses direct and derived spectral attributes combined with spatial features and information extracted from the metadata. The identified land cover objects can be converted into the current CORINE classes after classification.

The seismic hazard of Romania is relatively high, mainly due to the subcrustal earthquakes located at the sharp bend of the Southeast Carpathians, in Vrancea region, one of the most seismically active area in Europe. It is crossed by a series of principal and secondary faults. Vrancea area is assumed to be a conjunction of 4 tectonic blocks which lie on the edge of the Eurasian plate. Several GPS monitoring data revealed the motion of the blocks both in horizontal direction (relative motion of 5- 6 millimeters/year), as well as in vertical direction(of a few millimeters/ year).All data information available on the study area have been integrated in a unique database of geologic maps, thematic maps from cartography, land use maps provided by satellite images acquired in different spectral wavelengths by Landsat MSS, TM and ETM, SAR ERS and ASTER during a long term period (1975-2002). Satellite data are excellent for recognizing the continuity and regional relationships of faults . Synergy use of satellite data and image analysis techniques is essential for neotectonic applications, improving greatly the interpretability of the images and subsequent more accurate terrain features and lineament analysis of geologic structures in active seismic areas.

The coastal zone contains that unique environmental triple point where the water, land and atmospheric components of the terrestrial surface converge and interact. This paper is an application of remotely sensed images in marine coastal land cover classification for change detection assessment. The nature of the gradients in coastal region land cover composition among the map classes can therefore be identified.A supervised approach uses the prior knowledge about the area and thus it is very useful in getting better results than an unsupervised classification. The study test area was North-Western Black Sea coastal region, characterized by no so fast drastic changes,as it is a slow and continuous process. Satellite images (Landsat MSS, TM, ETM, SAR ERS, ASTER, MODIS) over a period of time between 1975 and 2003 were chosen for change detection analysis.In the fuzzy approach, it is possible to describe change as a degree, this being the main reason for fuzzy approach using for classification and change detection of major land cover classes in a marine coastal area.The results can be utilized as a temporal land-use change model for a region to quantify the extent and nature of change, and aid in future prediction studies, which helps in planning environmental agencies to develop sustainable land-use practices .

The aim of this work is to apply hyperspectral remote sensing technique to land quality monitoring to explore its application potential in this field. According to the characteristics of hyperspectral remote sensing technique combining with spectral features of land quality indicators, we use multivariate statistics methods and approaches based on spectral position variables, explore the spectral indicators sensitive to land quality, set up the retrieval models of land quality indicator, study the potential of applying hyperspectral remote sensing technique in land quality monitoring. Attempt to ameliorate land quality monitoring techniques, expand monitoring extents, decrease the self cost of survey, shorten the survey period, and make the results more scientific, objective and stable by this technique.

This paper analyzed major characteristics of land use changes in the Beijing-around region, based on TM(ETM) in 1991 and 2002. On the basis of that, we studied the differences in districts of land use change on county area scale, using intensity, state and trend parameters of land use change. In addition, we investigated the effects of land use change on eco-environment in this region. We found that the area of arable land decreased greatly, with a gradually increasing trend from southeast to northwest from 1991 to 2002. On the other hand, the area of forested land and grassland increased, especially in the northwestern area. The total area of sandy land increased, with a gradually decreasing trend from east to west. Land use change was characterized by low intensity, and the area of net change in each kind type of land use was much more for the most counties in the studied region. From south and north to middle and east to west, the intensity of land use change increased gradually. The degree of single-direction interchange between different land use types decreased gradually from west to east. In recent 10 yr, quality and productivity of land was decreased increasingly in this region. But with the construction of various forest zones by reversing arable land to grassland and forest land, the descent dust amount per yr in Beijing suburban decreased with increase of the area of grassland and forestry land in the Beijing-around region.

A multiprocessor board for SAR raw data processing is presented. It has been designed especially for application on small platforms, e.g. UAVs. The challenges were to provide high processing power for real-time processing of SAR raw data on a small printed circuit board with low power consumption. These constraints can be met by a digital signal processor (DSP) like the HiPAR-DSP 16, which has been developed at the Laboratorium fur Informationstechnologie. The multiprocessor board of the size of 230x160x20mm3 is equipped with six HiPAR-DSP 16 and has a peak power consumption of 15 W. For generation of SAR images the wk algorithm has been selected. It has been implemented for the HiPAR-DSP 16 and mapped to the multiprocessor board. Each HiPAR-DSP 16 receives one set of rangelines and returns the image data. A demonstration system emulates exactly the timing of the SAR sensor. In this way the multiprocessor board works in a real environment. The SAR images generated by the multiprocessor board can be stored and displayed. The resulting images have been compared to images generated by the reference code.

This work presents applications of an unsupervised method capable of providing estimates of temporal coherence starting from a pair of multilook detected Synthetic Aperture Radar (SAR) images of the same scene. The method relies on robust measurements of the temporal correlation of speckle patterns between the two pass dates. To this end, a nonlinear transformation aimed at decorrelating the data across time while retaining the multiplicative noise model was defined as the pixel geometric mean and ratio of the two overlapped images. The temporal correlation coefficient (TCC) of speckle is analytically derived from the noise variances, measured in the transformed pair of images as regression coefficients of local standard deviation to local mean, calculated on homogeneous, i.e., non-textured, pixels. Experiments carried out on two pairs of multitemporal SAR observations, from the ERS-1/2 Tandem mission show that coherence quicklooks having comparable accuracy, even though considerably lower spatial resolution than those calculated from single-look complex (SLC) data, are easily obtained for coherence browsing.

An extensive data set, made up of different remote sensing experiments, six carried with a ground-based instrument, a scatterometer, and two with the AIRSAR sensor, Washita '92 and SMEX '02, has been investigated. The aim was to study the feasibility of soil parameters extractions in different environmental conditions and with different sensors. The extraction algorithm is a combination of Bayesian methodology with theoretical models. The chosen theoretical model is the Integral Equation Model because its range of applicability covers most of experiment surface conditions. Bayesian methodology allows meaningful and rigorous incorporations of all information sources into the inverse problem solution. The key point is the evaluation of a joint posterior probability density function based on the contemporary knowledge of data sets consisting of soil parameters measurements and the corresponding remotely sensed data. In this study, it is obtained by applying the maximum likelihood principle (MLP). The inversion procedure has been applied to bare and vegetated fields. The correlation coefficient between measured and estimated dielectric constant values are R = 0.41 and R = 0.81 for bare fields and for C and L band respectively. In the case of the vegetated soils, the correlation coefficients are variable between 0.34 and 0.94, according to the different level of vegetation. It can be noted that the drying phase changes considerably from one part to another of the same field. The in-homogeneity of the fields introduces further errors in the inversion procedure.

A study has been carried out to test which one of two different approaches: use of L-band active and passive data or use of L-C-X-bands passive data, is more effective to retrieve soil moisture of bare soils. Simulated and measured data were used. Simulated data were generated implementing IEM model for active and L-C passive data, and GO model for X band passive data. Measured data derive from the Soil Moisture Experiment "SMEX-02". As a preliminary investigation, retrieval was solved by the application of artificial feed forward backpropagation neural networks. Three different input configurations were considered: 1a) L-band: emissivity H and V polarizations-backscattering coefficient HH polarization ; 1b) L-band: emissivity H and V polarizations-backscattering coefficient VV polarization ; 2) L-band--C-band--X-band emissivity H polarization. For all three input configurations the requested outputs were root mean square of heights s, correlation length l and dielectric constant er. To test the methodology, the best performing nets were chosen to simulate first a retrieval with an artificial dataset with noise added. All chosen configurations permit an excellent retrieval of the real part of the dielectric constant on every soil type (smooth, medium and rough), while roughness parameters, especially autocorrelation length, are not well retrieved. Active-passive approach proved to be more efficient, as a consequence only active-passive configurations were used with real data. The algorithm confirmed to be efficient when neural networks have been trained with "noisy data". However, there is always an underestimation, probably due to vegetation. Further investigations need to be carried out in order to understand the cause of this underestimation.

At Faculty of Physics, Moscow State University, the new image processing methods for different physical measuring systems are created. The main feature of the proposed super-, ultra-resolution methods consists in the diminishing of the dimensions of problems under consideration. In super- resolution method every actual (or virtual) ray has its own local vision domain. The local-linear super- resolution problem was solved on the special arranged set of actual (or virtual) rays. The introduced resolving function R [1] was not used. Point Spread Function (PSF) O and resolved O: R*O were needed for the illustration of results of the local-linear super-resolution method [1]. In ultra-resolution (point) method, the resolving function R is directly used on small size vision domains ,and so is PSF O. The ultra-resolution method gives point results. In the super-resolution method each pixel was divided onto 2x2 and 4x4. The method of ultra-resolution gives us practically unlimited capability for "interpolation of pixels". "The pixel interpolation" certainly increases the dimensions of problem, but it enables us to perform a better presentation of the PSF O of the image measuring system. From the point of view of super- resolution method, the number of virtual rays in ultra-resolution method corresponds to the number of the small "interpolated pixels". The new ultra-resolution method is more effective and stable in comparing with the super-resolution method [1]. Numerous applications are considered, too.

An experiment whose aim is the retrieval of surface wind fields from SAR imagery coupled to a high resolution mesoscale numerical atmospheric model in semi-enclosed sea basin, is presented. A sea region belonging to North-Western Mediterranean Sea, which spans in W-E direction from Corsica (8.8 E) to Italian coast (10.5 E) and in N-S direction from Lygurian Sea (44.0 N) to North Tyrrhenian Sea (42.2 N), was selected as test area. Two consecutive ERS-2 SAR frames from the pass of March 30, 2000, along with a set of NOAA/AVHRR and MODIS images acquired on the same day were used for the analysis. SAR wind speeds and directions at 10 m above the sea surface were retrieved from predictions of the semi-empirical backscatter models CMOD4 and CMOD-IFREMER, which describe the dependency of the normalized radar cross section (NRCS) on wind vector and ERS-2 SAR image geometry. Surface wind vectors predicted by the meteorological ETA model were exploited as guess input to SAR wind inversion procedure to describe atmospheric conditions in the area, according a Bayesian approach recently proposed in literature. ETA is a three-dimensional, primitive equation, grid-point operational model running at the National Centers for Environmental Prediction of the U.S. National Weather Service. The model was adapted to run on selected regions of the Mediterranean basin with a nested very high, up to about 4.0 Km, resolution. The latter feature makes ETA model particularly suitble for its use in combination with SAR images. Besides, to simulate and predict several specific atmospheric weather phenomena, ETA outputs also include the vertical distributions of physical parameters such as air pressure, temperature, moisture up to about 25 Km. Apart some discrepancies in sparse and small areas, an overall agreement between SAR inversion results and ETA predictions was found. More importantly, it was found that the inversion methodology was not able to resolve wind speed modulations due to the manifestation of an atmospheric gravity wave, which occurred in the analyzed area as a result of the terrain disturbance to the air flow imposed by the peninsula located North of Corsica. Temporal evolution of the wave propagation phenomenon was allowed by inspection of NOAA/AVHRR and MODIS images through the detection of a cloud band associated to the atmospheric wave. A wave propagation model describing waves in the atmosphere owing to the disturbing action on the primary air flux by terrain features was thus used to account for the observed surface wind speed modulation on SAR image. Synergy with ETA model outputs was further exploited as atmospheric parameters up-wind the atmospheric wave were considered as input to the wave propagation model.

Various phenomena in the radar imaging mechanism, such as velocity bunching, azimuthal cutoff, tilt modulation, range dependence etc affect the SAR image. Among them, velocity bunching and azimuthal cutoff, straightly related to SAR azimuth direction, have been most investigated. However, in this study, the range factor, mainly due to lower radar image intensity for far range, is proved to be more pronounced. In the study framework, 2nd order texture analysis was performed in order to investigate a) the potential of texture to detect wave direction, b) wave range dependence, and c) velocity bunching effects. For this purpose, the Haralick Cooccurrence matrix was calculated on a despeckled ERS2 image for four directions. Eight texture images were generated and compared to the wave direction resulting from the TOPEX/POSEIDON model. The comparison showed that a) the texture image produced in the range direction detected sea wave direction efficiently, and b) accuracy was affected by the range factor. Thus, a range correction was proposed and implemented on the despeckled texture images. Following this correction, an accuracy of 88.6% was achieved for wave direction detection, if texture is calculated along the range direction and 74.9% if texture is calculated along the azimuth direction.