Our work investigates the potential of high-resolution hyperspectral satellite data for detecting atmospheric methane concentrations. We employ the matched filter with Albedo correction and reweiGhted L1 sparsity Code (MAG1C) algorithm, which integrates a sparsity prior, a matched filter, and albedo correction techniques. For the analysis, we utilize hyperspectral data from the PRISMA mission, leveraging its high spatial resolution to potentially enable more accurate localization of point emission sources. Comparing the methane column estimation resulting from our work with corresponding products provided by both the Sentinel-5P and GHGsat missions, a good agreement was found. In particular, a bias of 5 ppb with respect to the methane abundance estimated from GHGsat was reached.

Land surface temperature (LST) is one of the most important variables for the terrestrial ecosystem. [1] It stands as a fundamental Essential Climate Variable (ECV) [2] and hold paramount significance across different environmental and agricultural domains [3].Temperature estimation from satellites is increasingly widespread, which allows to obtain large-scale and almost real-time informations.However, estimating temperature using satellite data has some limitations: presence of cloud cover has an impact on remotely sensed observations [4].This paper presents a data fusion approach for enhancing Sentinel-3 LST products by replacing cloudy pixels with data from MODIS, GCOM-C and ERA5-Land. The model has been tested on a trial study area, but its versatility allows it to be applied worldwide.The resulting fused LST product is subjected to an evaluation against LANDSAT 8 and 9 LST data. The performance of the data fusion demonstrates the efficacy of the proposed fusion method, with Pearson correlation values ranging from 0.60 to 0.93.The study not only contributes to advancements in LST data quality but also establishes a benchmark for future research in satellite data fusion.

Remote Sensing (RS) is applied for a variety of purposes, thanks to the large availability of heterogeneous data. Furthermore, the growing number of CubeSat missions is encouraging increasingly advanced, flexible, and configurable RS missions, that also include the use of Artificial Intelligence (AI) on-board. Indeed, specific hardware allows advanced processing on-board the satellites even if the computational capability is not the same as on the ground. In the context of on-board processing, the compression of acquired images is crucial because permits to save bandwidth for data transmission. We propose an AI-based lossy image compression algorithm for multispectral images that can be executed on-board a CubeSat. The algorithm is based on a Convolutional AutoEncoder (CAE) Neural Network (NN). In lossy compression part of the information stored in the original image is lost. Therefore, the results evaluation includes the assessment of the usability of the decompressed images for common applications.

Post-disaster analysis is an important challenge in disaster risk management during the recovery phase. This study focuses on the benefits of the use of multispectral images by Copernicus Sentinel-2 to assess long-term urban change and to monitor the progress of reconstruction after earthquakes. The innovative and straightforward methodology aims at monitoring changes in urban areas, and more specifically related to buildings, after a seismic event. It is based on the analysis of extended time series of the spectral reflectance values in the Red-Green-Blue (RGB) bands and the resulting Perceived Lightness (PL) values, as well as the Normalized Difference Vegetation Index (NDVI). This procedure is effective and reproducible, with the advantage that Sentinel-2 observations, unlike commercial Very High Resolution (VHR) images, are free and regularly available on a wide scale. Based on a rich set of multispectral data, this research provides valuable insights into urban dynamics, and aims to optimize post-disaster risk management.

This study explores the synergy of electromagnetic data modeling and Artificial Intelligence (AI) algorithms for soil moisture retrieval over agricultural fields using Polarimetric SAR (PolSAR) data. SAR acquisitions are considered as a valuable source for accurate estimation of soil moisture in agricultural areas. However, its retrievals are influenced by various factors, including vegetation cover. In this context, the polarimetric information of SAR acquisitions allows the interpretation of the occurring scattering processes. The proposed approach involves a physics-aware AI algorithm based on two Artificial Neural Networks (ANNs), trained on electromagnetic (EM) model simulations at L-band. Starting from a full polarimetric Covariance Matrix, the first AI model separates the different scattering contributions, estimating surface and double-bounce scattering mechanisms while minimizing the attenuation effects caused by the vegetation layer. Then, the estimated surface and double-bounce components are fed to an additional AI model to retrieve soil moisture. Field campaign data over actual corn fields were considered and ingested by the EM model to generate synthetic SAOCOM-like case studies which were used to validate the approach within the simulated domain.

The activities undertaken within the EcoNet project aim at the design and development of an integrated system for the monitoring of changes in surface waters natural status based on different sensoristic techniques. The proposed integration approach combines ground measurements and hyperspectral satellite images. The promising dialogue that occurs between these two multi-sensoristic technologies requires the implementation of appropriate tools for data handling and analysis which in this work are represented by Artificial Intelligence (AI), particularly suitable to retrieve very subtle relationships among the data. This integration can open enormous potential for overcoming the limits of traditional environmental monitoring and diagnostic techniques.

In this work a standard approach for the collection of multi-angular reflectance measurements by means of UAS (Unmanned Aerial System) is presented, as well as the modelling of the reflectance anisotropy through the Ross-Li-Maignan BRDF (Bidirectional Reflectance Distribution Function) model. The analysis is performed over three different types of surfaces, in particular a wheat field, an asphalted area and a corn field, and the measurements are acquired by means of MAIA multispectral camera on board UAS, which is characterized by the same spectral bands of Copernicus Sentinel-2 MSI (MultiSpectral Instrument). The results are promising being the relative RMSE between modelled and measured reflectances below 10% for all the test sites.

The presented methodology aims to create a daily Aerosol Optical Depth (AOD) fusion product by integrating observations and forecasts from various EO data sources. The data used for this purpose are from the Ocean and Land Colour Instrument (OLCI) and Sea and Land Surface Temperature Radiometer (SLSTR) sensors on Sentinel−3, the Second Generation Global Imager (SGLI) on the Global Change Observation Mission−Climate (GCOM−C), and the forecasts from the Copernicus Atmosphere Monitoring Service (CAMS). Leveraging the strengths of each dataset, the developed algorithm adapts to different formats and resolutions, providing a unified and higher−resolution AOD dataset.The Sentinel−3 AOD product ensures high resolution (4.5 km), the GCOM−C product ensures dataset accuracy, and CAMS forecasts offer predictive insights. The algorithm employs a mathematical averaging technique for coincident pixels, facilitating precise AOD estimates in overlapping regions, while a mosaic technique seamlessly integrates non−coincident areas.This fused dataset enhances spatiotemporal coverage, contributing to a more in−depth understanding of atmospheric composition variations. Emphasizing the importance of complete AOD data, the methodology is versatile and has been validated against Moderate Resolution Imaging Spectroradiometer (MODIS) aerosol products, achieving a Pearson coefficient of 0.87.The product was tested on the Italian Po River Basin area for the year 2021. This region represents an area of great interest for the study of air quality and atmospheric dynamics due to its geographic complexity and the significant impact of anthropogenic activities.The final product is an improvement over the reliable SYN−AOD product from Sentinel−3, providing daily data obtainable in less than an hour through an automated algorithm.

The PRIMARY (PRIsma for Monitoring AiR quality) project objective is to address air quality monitoring, especially in urban areas, exploiting the PRISMA (PRecursore IperSpettrale della Missione Applicativa) mission. Utilizing PRISMA's hyperspectral data, the project aims to gain insights into atmospheric aerosol content and composition, crucial for understanding environmental and health impacts, especially in urban areas. Overcoming spatial resolution limitations and the inverse problem's complexity in satellite-based characterization, PRISMA's decametric spatial resolution and artificial intelligence play crucial roles. A synthetic PRISMA-like dataset, relying on data provided by the Copernicus Atmosphere Monitoring service (CAMS), was generated for training neural networks for estimating aerosol characteristic exploiting PRISMA data. Preliminary results are encouraging. Properly field campaigns were performed in Rome (autumn 2022) and Milan (winter to summer 2023) to support the validation of the PRIMARY project's outcomes. In addition, drone-based campaigns are currently ongoing.

This paper introduces a hybrid approach to dimension reduction of PRISMA hyperspectral data, employing both linear and non-linear techniques: Principal Component Analysis (PCA) and autoencoders. The study aims to validate the efficacy of autoencoders by comparing results with the well-established PCA method. Our primary objective is to harness the complementary strengths of both methods in a hybrid framework, wherein certain bands may exhibit superior performance with autoencoders, while others fare better with PCA in dimension reduction. This strategic amalgamation not only accelerates data transfers and lowers computational costs for real-time applications but also leverages the specific advantages offered by each technique. The paper underscores the potential of this hybrid approach for optimizing hyper-spectral data for enhanced feature extraction in various neural network applications.

This research utilizes radiance data from the PRISMA hyperspectral satellite mission, employing the MAG1C algorithm that incorporates sparsity and albedo correction. The algorithm is applied to convert level 1b data into methane concentration-pathlength maps, providing essential insights for addressing methane emissions. Data from the TROPOMI imaging spectrometer on board the Sentinel-5P mission will be used to compare the experimental results obtained through the algorithm. To bridge the difference in resolutions, the data comparison will be performed by degrading the PRISMA data by means of a weighted average on the different data resolutions, resulting in an RMSE (Root Mean Square Error) of 73.7175 ppb. The use of new high resolution data is one of the factors on which the maximum effort must be focused with the aim of being able to localize the point emission sources.

This paper aims to introduce the concept and objectives of the recently funded AI-OBSERVER Horizon Europe Twinning project titled “Enhancing Earth Observation capabilities of the Eratosthenes Centre of Excellence on Disaster Risk Reduction through Artificial Intelligence”. The AIOBSERVER project aims to significantly strengthen and stimulate the scientific excellence and innovation capacity of the ERATOSTHENES Centre of Excellence on the use of Artificial Intelligence for Earth Observation in the Disaster Risk Reduction thematic area, as well as the research management and administrative skills, of the Centre. This will be achieved through a series of capacity building and targeted research activities, having the support of internationally leading institutions, i.e., the German Research Centre for Artificial Intelligence from Germany and the University of Rome Tor Vergata from Italy, assisting the ERATOSTHENES Centre of Excellence to reach its longterm objective of raised excellence on Artificial Intelligence for Earth Observation on environmental hazards.

The amount of raw data generated by instruments on board Earth Observation (EO) satellites is quite often more than what can be transmitted to the ground, so new advanced on-board processing procedures are required. Artificial Intelligence (AI) and Deep Learning (DL) can provide advanced information from EO data and thanks to specific hardware platforms these algorithms can be used also in space. We present here the Convolutional AutoEncoder (CAE)-based algorithm developed for on-board lossy image compression of the European Space Agency (ESA) Φ-Sat-2 mission. DL algorithms have already been successfully applied for image compression however performance degradation may occur in the context of a representative on-board environment. Therefore, besides analyzing the results for the local hardware environment, we investigate the performance variation for the on-board setting. Moreover, we introduced an applicative metric for the evaluation of the compression to assess the applicability of the reconstructed images for other tasks.

Super-resolution is aimed at enhancing image spatial resolution and it has been intensively explored for many years. The recent advancements, underpinned with deep learning, also include techniques developed specifically for hyper-spectral data. However, most of the emerging methods are validated in application-independent scenarios, which often rely on an unrealistic experimental setup—the reconstruction is performed from simulated low-resolution images (degraded from an original image) with the goal of inverting the degradation process and restoring the original image. This leads to over-optimistic assessment of super-resolution capabilities and limits their practical applications. In this paper, we demonstrate task-based validation for different types of hyperspectral PRISMA image super-resolution, including pan-sharpening, fusion of multispectral and hyperspectral data, as well as single-image super-resolution. The obtained results reported in the paper are encouraging and they help better understand the value of super-resolved PRISMA images.

There is an urgent need to assess the condition of transport network in many countries all over the world as in Italy and, in this context, the consideration of non-destructive techniques has particular importance. These techniques, such as space-born systems, laser scanners, ground-penetrating radar (GPR), and monitoring tests, provide valuable data but have limitations in assessing specific aspects of the infrastructure. This research aims to overcome these limitations by using a "data fusion" approach to achieve a comprehensive understanding of the asset’s condition. The ongoing project "EXTRA-TN" focuses on road infrastructures selected as case-studies, located in Salerno (Italy), utilizing the various aforementioned methods. This integrated approach aims to enhance infrastructure resilience, provide valuable information for maintenance scheduling, and monitor both external and internal factors affecting safety and functionality.

The InnEO’Space PhD program addresses the growing demand for skilled individuals in Earth Observation (EO) data management. It offers modern and transferable courses to enhance researchers' innovation-oriented skills. The program includes InnEO Startech, focusing on entrepreneurship and leadership, and the Machine Learning for Earth Observation InnEO Summer school, emphasizing open science and technical skills. Small Private Online Courses (SPOCs) provide blended learning through a user-friendly platform. The InnEO PRO section facilitates interaction among PhD students and professionals, offering various resources and self-assessment tests. The program's success has led to collaborations with UNIVERSEH and FabSpace, while AI4AGRI aims to utilize its resources for AI in EO research.

Soil moisture represents a significant guiding factor for agricultural activities, especially for smart irrigation and crop yield estimation. In this context, SAR data is one of the most valuable sources of information for accurate and continuative estimation of soil moisture in agricultural areas. However, SAR-derived soil moisture retrievals are affected by several factors, including the vegetation cover, which is responsible for additional signal attenuation and scattering mechanisms. Concerning the algorithms for soil moisture estimation, Machine Learning (ML) has proved to be a valuable instrument for finding relations between SAR data and the soil dielectric properties. For this purpose, this study aims to synergically adopt electromagnetic data modelling and a ML algorithm to estimate the scattering contributions associated with the ground and demonstrate that they are more sensitive to soil moisture than the total received signal. To apply such approach to real SAR data, airborne acquisitions at L-band will be considered.

EcoNet is a joint project between the Italian Space Agency (ASI) and the Institute of Nanostructured Materials of the National Research Council of Italy (CNR-ISMN) with the participation of University of Tor Vergata (UTOV) that aims to develop an integrated sensor-driven system managed by artificial intelligence (AI) for monitoring surface waters near human settlements. The present paper outlines the foundational concepts of the project, the methodology and the ongoing activities, with an outlook towards the innovation for downstream applications in the field of aquatic ecosystem monitoring and management.

This work addresses the Bidirectional Reflectance Distribution Function (BRDF) characterization by means of MAIA multispectral camera onboard an Unmanned Aerial System (UAS). The proposed procedure relies on the design and execution of UAS flight plan for multi-angular acquisitions, which can be automatically repeated over different land cover types. Then, the inversion of the RossThick-LiSparse (Ross-Li) BRDF model is pursued in order to retrieve the fundamental parameters, allowing the complete characterization of the considered surface in terms of reflectance distribution for each band. A key point of this work is the challenge we face related to the development of an optimum strategy for the collection of ground-based dataset for BRDF model inversion.

Air pollution and its effects on human health pose a significant challenge in modern society. The PRIMARY (PRIsma for Monitoring AiR quality) research project aims to utilize the capabilities of the Italian Space Agency's (ASI) PRISMA (PRecursor HyperSpectral Application Mission) to enhance air quality monitoring, particularly in urban areas. In particular, the project focuses on the exploitation of the hyperspectral PRISMA data to obtain detailed qualitative and quantitative data on atmospheric aerosol load and composition in urban environments. Current satellite-based characterization of particulate matter is limited due to spatial resolution constraints and to the complexities of the underlying inverse problem involving multiple variables. The PRIMARY project addresses the first issue through the decametric spatial resolution of PRISMA images, while the second issue is tackled by leveraging artificial intelligence approaches.

In this paper, we present our study focused on applying a vision transformer-based pansharpening technique to enhance PRISMA satellite hyperspectral data. The PRISMA mission, launched by the Italian Space Agency, captures hyperspectral images comprising visible and near infra-red, as well as short-wave infra-red channels. By integrating the panchromatic image of high spatial resolution with the hyperspectral data of high spectral resolution, the pansharpening process consists in producing spatially-enhanced hyperspectral imagery. Our research involves modifying and adapting the state-of-the-art HyperTransformer architecture to effectively process real-life PRISMA data. The evaluation of our model’s performance utilizes PRISMA L2D data, encompassing simulated low-resolution data and real-life data. We employ quantitative metrics and visual examination to assess the results. We also highlight the importance of choosing right PRISMA data processing level for the pansharpening process. The proposed pansharpening model successfully enhances PRISMA data for practical applications, contributing to the advancement of Earth observation techniques.

The growing amount of data currently collected by earth observation satellites requires new processing procedures able to manage huge quantity of information. Among these, data reduction techniques represent a viable solution. In particular, data reduction on-board is significant because allows to save on-board storage space and bandwidth for data transmission to the ground. However, the algorithm used for compression must be able to preserve the key information contained in the acquired data, so that the applicability of the collected information is still guaranteed in the different fields of work. Artificial intelligence, and in particular deep learning, are well suited for this purpose because of their ability to extract valuable information from complex data. This work proposes a lossy image compression procedure based on a Convolutional Autoencoder (CAE) that can be performed on-board the satellite. The images acquired by the sensor can be compressed through the algorithm, stored, and sent to the ground where they are reconstructed, saving space and bandwidth for data transmission. The performance of the compression algorithm will be evaluated in terms of original-reconstructed image similarity and also with regard to the applicability of the reconstructed images to common applicative cases. The algorithm here proposed has been idealized and is currently in development in the context of the European Space Agency's (ESA) PhiSat-2 mission, that aims at demonstrating the advantages of the Artificial Intelligence (AI) on-board for Earth observation applications.

We present a physics-based machine learning (ML) approach for estimating corn plant height from simulated synthetic aperture radar (SAR) data. The proposed study intends to demonstrate the physical awareness of datadriven approaches such as ML. In this regard, a multilayer perceptron (MLP) artificial neural network (ANN), designed for corn plant height estimation, was trained with simulated C- and L-band SAR data generated using a state of the art electromagnetic model for microwave backscattering from terrain covered with vegetation. Here we show how the most significant connections between the nodes composing the network and the most relevant input variables can be detected, demonstrating the physical meaning behind the mapping criteria of the network itself.

In this work the classification of Sentinel-3 Sea and Land Surface Temperature (S3-SLSTR) images with a focus on volcanic cloud detection through a Neural Networks (NNs) approach is presented. Since the hazardous nature of eruptions, a fast and reliable method to monitor the evolution of volcanic clouds in real time is of primary interest. NNs represent a suitable tool for this purpose given their short processing time once trained, and their ability to solve complex problems as those related to natural events. The present research starts from the generation of the training patterns by means of MODerate resolution Imaging Spectroradiometer (MODIS) data collected during the 2010 Eyjafjallajokull (Iceland) eruption, it goes through the training of the NN, and ends with the application of the NN-based model to SLSTR data collected during the 2019 Raikoke (Kuril Island, Russia) eruption.

In this study we present a novel approach dedicated to the retrieval of volcanic ash parameters by means of data acquired from the Sea and Land Surface Temperature Radiometer on board of the Copernicus Sentinel-3. In this framework, we developed a procedure combining Radiative Transfer Model simulations and Neural Network for estimating three volcanic ash parameters such as aerosol optical depth, effective radius and ash mass. The Radiative Transfer Model simulations have been considered for producing synthetic training sets, which have been used in the training phase of the Neural Networks development. In particular, nine latitude's belts have been identified for training several Neural Networks ensuring the global coverage of the method. The approach has been tested by comparing the results of the trained NN with the ones obtained by applied the state-of-art Look Up Table and the Volcanic Plume Retrieval procedures. The results of the methodologies applied on Raikoke, 2019 eruption demonstrated the feasibility of the proposed approach by registering values of the correlation coefficient between all the three methods ranging between the 65% and the 94%.

Different technologies currently applied for monitoring transport infrastructures show promising results in providing responses for their particular field of application, but they are still missing to be assimilated in an integrated solution that would exploit the advantages of each of them. Aiming at filling this gap, here we propose an approach based on the fusion of nondestructive measurements from satellite and insitu surveys, for a full-assessment of transport infrastructures. In particular, we describe the preliminary results obtained by the space segment, including the application of satellite SAR interferometry technique and satellite SAR and multispectral land cover classification over an area west of the city of Salerno, south Italy, that is characterized by severe instability and human pressure. In addiction, this work provides a first formalization of the proposed data-fusion scheme. Identified methods show promising results and will be further developed in the context of Italian research Project of Relevant National Interest (PRIN) already ongoing.

In this paper we describe the experimental set-up of a study aiming at testing the capability of UAV (Unmanned Aerial Vehicle) multispectral imagery for the calibration of the electromagnetic quantities measured by the medium resolution satellite Sentinel-2, launched by European Space Agency. This is made feasible by mounting on the UAV a camera characterized by acquisition bands which are designed in order to mimic those on the satellite. Preliminary analysis over a vegetated area shows encouraging results because the spectral signatures of the two instruments appear quite consistent. Theoretical modelling of BRDF (bidirectional reflectance distribution function) is also considered in order to guide the acquisition plan of the UAV measurements

This work focuses on the development of a fully-automated integration approach, which seeks to combine data from multiple sources with the aim of implementing a dependable soil moisture estimation framework based on Neural Networks (NNs). Several papers have dealt with inverse modeling of soil moisture using NNs, often trained by way of Synthetic Aperture Radar (SAR) data generated through the Integral Equation Model (IEM); our approach is designed to harness the newer IEM calibrated version modified by Baghdadi (IEM_B), integrating synthetic data with real data in order to monitor possible improvements in NNs estimation efficiency. The experiment involves two steps: a first NN is trained with SAR Sentinel-1 data (taken from the Google Earth Engine, GEE, Catalog and granted freely by the European Union, EU, under the Copernicus programme) and in situ soil moisture measurements, taken from the International Soil Moisture Network (ISMN); in the second part, a further NN is trained by enlarging the training dataset with IEM_B generated SAR data. Combination of the two large-scale data sources and IEM_B generated data may lead to an improvement in separating soil/vegetation contributions, possibly improving NN soil moisture estimation accuracy; the study presented may serve as a baseline demonstration for future exploitation of the large-scale data sources for both general and field-specific soil moisture estimation.

Machine learning can be considered as a very important area within artificial intelligence and it is characterized by algorithms and techniques that learn by examples. In the last decade, mainly due to the improvements obtained in the field of high performance computing, such as the enhanced exploitation of cloud technology and of graphics processing units (GPU), machine learning models have gained considerable progress as far as remote sensing and Earth Observation (EO) applications are concerned. However, the need of huge quantities of data necessary for the training phase, may be still a limiting factor especially in problems addressing the quantitative estimation of geo-physical parameters. In this paper, we report about the design and the development of a new platform capable of meeting the requirements of scientists and researchers who are attracted by the use of machine learning but meet difficulties in the generation of reliable data sets. The platforms relies on the implementation of radiative transfer models, plus a bunch of appropriate functionalities, in order that simulated data can be added to those available by ground-truth campaigns.

In recent years many studies concerning the monitoring of volcanic activity have been carried out to develop ever more accurate and refine methods which allow to face the emergencies related to an eruption event. In our work we present different approaches for the volcanic ash cloud detection and retrieval using Sentinel-3 Sea and Land Surface Temperature Radiometer (SLSTR) data. As test case the SLSTR image collected on Raikoke volcano the 22 June 2019 at 00:07 UTC has been considered. A neural network based algorithm able to detect and distinguish volcanic and meteorological clouds, and the underlying surfaces, has been implemented and compared with two consolidated approaches: the RGB (Red-Green-Blue) and the Brightness Temperature Difference procedures. For the ash retrieval parameters (aerosol optical depth, effective radius and ash mass), three different methods have been compared: the reliable and consolidated LUTp(Look Up Table) procedure, the very fast VPR (Volcanic Plume Retrieval) algorithm and a neural network based model.

During a volcanic eruption, large quantities of Sulphur dioxide (SO2) are sometimes emitted into the atmosphere. Rapid detection and tracking ofvolcanic SO2 clouds might be beneficial to air traffic security and to predict any correlated impact on the environment; for example, the possibility of acid rain events. Within the presented work, we exploited Sentinel-5p radiance data (Level 1 b) to detect and retrieve SO2 volcanic emissions through a neural network based algorithmthat produces rapid SO2 vertical column estimates. The dataset used for training the net was composed of 13 TROPOMI Level 2 “Offline” SO2 data collected during the Etna Volcano eruption that occurred in 2018 from 22 December to 1 January. Experimental results are very encouraging and open to the perspective ofmake available a new and stable product for monitoring atmospheric SO2 clouds on a global scale based on Sentinel-5p acquisitions.

Studies of oil slicks in the ocean environment with synthetic aperture radar (SAR) have found that one of the most complex challenges to oil spill detection is the separation of mineral oil spills from slicks that are biogenic in origin. The possible occurrence of multiple scattering mechanisms beyond Bragg scattering for the sea surface, with or without biogenic or mineral oil slicks, and even under low to moderate wind conditions, has also been a subject of debate because the measured signals from these radar-dark surfaces can be contaminated easily by noise. Therefore, the use of noise-uncontaminated data is required for oil spill study in order to avoid significant alteration in the measured radar backscatter, which can lead to misinterpretation and misclassification of the scattering mechanisms involved. To this end, this study uses uninhabited aerial vehicle SAR data, with a noise-equivalent sigma zero as low as −53 dB, to investigate slick classification within a deep learning framework in order to assess deep architectures’ capabilities for providing a reliable and accurate three-state classifier capable of separating mineral oil films from biogenic slicks and from the clean sea. The study exploits parameters with sensitivity to the dielectric constant and ocean wave damping properties, and convolutional neural networks’ (CNNs’) capability for learning nonlinear features, shapes, and textural and statistical patterns, in order to obtain significant classification accuracy. Very high accuracy results have been achieved, with values up to 0.91, 0.94, 0.98, and 0.99 under the most probable real-world spill acquisition conditions.

In this work we present a new coastline extraction approach, which seeks to enhance performances and to provide automation in shoreline generation with SAR (Synthetic-aperture radar) data. Our approach is designed to harness Multi-Pol SAR acquisitions, while single-pol acquisitions are used in most of the approaches in this area, employing an Autoassociative Neural Network (AANN) for data fusion purposes and a Pulse-Coupled Neural Network (PCNN) for the generation of a final coastline. Using RADARSAT-2 data, main findings are shown, exhibiting better and comparable results with respect to consolidated approaches and with a recent automated method which may be regarded as within the state-of-the art methods in the field of coastline extraction from SAR data.

Synthetic Aperture Radar (SAR) images taken over a certain area at different bands and also with a short time interval are now more widely available. This is due to the increase of SAR acquisitions following the last space missions, such as Sentinel 1 and COSMO-SkyMed (CSK). New paradigms capable of performing effective analysis and synthesis stemming from such a type of information are then required in order to exploit better and disseminate the information contained in the data. In this letter, a data fusion technique between CSK and Sentinel-1 data is described. To this purpose, an ad hoc Nonlinear Principal Component Analysis (NLPCA) with Auto-Associative Neural Networks (AANNs) algorithm is designed and developed. The network extracts the most relevant features from the combination of the different scattering mechanisms. The extracted features are then used as inputs for a land cover classification exercise. A comparison between the results obtained with the original images and those yielded by the new synthesized data, with lower dimensionality, demonstrates the ability of the algorithm to generate useful final products.

In this paper a new approach based on the use of Synthetic Aperture Radar COSMO-SkyMed products to verify urban change detection and to observe new constructions is presented. SLC products information has been exploited, since the proposed procedure combines backscattering coefficient and coherence information as extracted from two interferometric data, acquired in a short time interval. The algorithm exploits the information from backscatter intensity and interferometric coherence. Firstly, the interferometric SAR couple is processed by an unsupervised Neural-Networks, particularly PCNN (Pulse Coupled Neural Network) is applied to create a preliminary changes map based on the difference of backscatter intensity information. Then, accuracy is further improved by the fusion with the coherence information. The achieved results shown as the combination of backscattering and coherence information, extracted from Very High Resolution SAR data, allows to provide very accurate urban change detections with a fast and unsupervised procedure. The latter is particularly suitable to process quickly huge amount of SAR data since its lower computational requirements with respect to e.g. supervised algorithms.

This study aims at analysing local government efficiency using EO data to focus on the construction of new buildings in burned areas. We detected all forests and pastures of Sardinia (Sardinia) that witnessed at least a one-hectare wildfire, for biennium 2005-2006, using Landsat-TM data. We then monitored all the burned areas for a period of at least 10 years following the wildfire, up until 2016, by using Landsat products and VHR imagery provided by the Sardinian Region. The idea was to verify compliance with the Italian Framework Law on forest fires, which prohibits the construction of any structure or infrastructure aimed at civil settlements on woods and pastures affected by a fire for the next ten years. We retrieved as many as 748 burned areas, over which 148 violations were detected. These data have been used to carry out an econometric pilot analysis. Results suggest that, in general, the emergence of violations is inversely related to wildfire dimensions and that, on average, law-breaking occurs within the first five years.

In this paper a new approach based on the fusion of Sentinel-1 and Sentinel-2 products to map urban change detection and to observe suburb's development is presented. The algorithm developed can process data in a fast, automatic and accurate way. To reach this goal, the processing chain uses an iterative multitemporal approach based, for each iteration, on three procedures. The first and second ones are based on Pulse Coupled Neural Network (PCNN) applied to SAR and optical images, respectively, while the third processing is an optical multiband filter, implementing the spectral difference computation. The three outputs of each iteration are fused together by means of a weighted average formulation. The algorithm may deal with multitemporal acquisitions to improve the overall accuracy in the detection of urban changes by the integration of the outputs at different time intervals.

Livestock productivity is likely to be adversely affected by climate change mainly in terms of feed supply variations. Principal livestock food resources in Zambia and Malawi are grass areas, but very few data are available for supply management and amount estimation. The aim of this paper is to illustrate the procedure adopted for preliminary estimations of grassland biomass retrieval and grass growth cycle identification over a wide area between the Lukulu District and the Mongu District, West Zambia, for the period 1996–2016. The procedure takes advantage of remote sensing observations from multiple sensors and neural networks. The preliminary results obtained are in accordance with the expectations and the seasonal variation is clearly visible in the growth cycles.

A Neural Network approach to classify Sentinel-3 sea and land surface temperature radiometer (SLSTR) pixels over polar regions is presented. The proposed approach is based on a careful preliminary analysis aimed to simulate SLSTR observation by means of MODIS data. The latter have been considered because of the long available time series and the quality of cloud mask products. A large set of MODIS AQUA and TERRA products has been applied to develop the training set of the Neural Network classificator that has been tuned to discriminate clear ocean, clouds and sea-ice surfaces on the scene.

In this manuscript, we address the problem of change detection for Sentinel-2 data. The proposed method is based on deep features representation. First, multilevel convolutional neural network (CNN) features are extracted from input images acquired at different times. Then, euclidean distance is applied to generate dissimilarity map that indicate change probabilities of each pixel. Finally, bounding boxes corresponding the change areas can be obtained with clustering and an optimizing connected component labeling algorithm. Experiments on a manually annotated dataset demonstrate the feasibility and effectiveness of the proposed method.

The aim of this paper is to introduce new algorithms for the oil spill detection taking fully advantage of the polarimetric and textural features contained in new generation SAR data such as those provided by Radarsat-2 and COSMO-SkyMed missions. The SAR information is exploited using a new statistical decomposition method based on AANN. Thanks to the AANN the original image is represented in terms of Nonlinear principal components (NLPC). The oil spill detection procedure is then directly applied to the new generated components.

The FabSpace 2.0 project (the open-innovation network for geodata-based innovation — by leveraging Space data in particular, in universities 2.0), funded by European Union under the Horizon Programme, aims at making universities open innovation centres for their region and improving their contribution to the socio-economic and environmental performance of societies. In order to achieve these general objectives, the FabSpace 2.0 project focuses on Earth observation data, an area with high expected socio-economic impact. In this context the universities involved in the project must endorse a new role beyond knowledge providers: co-creators of innovations. This does not mean that they will replace businesses and give up basic research, but that they collaborate with businesses to tackle market challenges and capitalize on opportunities.

The finer spatial, spectral and radiometric resolutions of current and planned sensors are rendering increasingly-high data rates which, coupled with limited on-board storage, downlink bandwidth and receiving ground station availability, make high-throughput, high-performance data-reduction techniques essential in forthcoming missions. On this paper we describe an algorithm well suited to high-dimensional data as those produced by multispectral and hyperspectral sensors, both highly relevant in a broad range of Earth Observation activities with the latter becoming increasingly available and delivering the highest data rates. The performance of parallel implementations of the algorithm on multi-core and GPU architectures is also evaluated.

SAR images from Italian COSMO-SkyMed mission can have a significant impact on the production and updates of land cover maps. However, for the full exploitation of the data and their application to nationwide extensions, robust automatic procedures need to be designed. In this paper we present the preliminary results obtained by the implementation of a processing scheme using COSMO-SkyMed images to provide, and regularly update every six months, land cover maps for the whole Italian territory. Most of the automatic processing is based on Neural Networks (NN) algorithms. In particular PCNN (Pulse Coupled NN) have been considered for change detection purposes while Multi-Layer Perceptrons (MLP) have been used for classifying the pixels belonging to a detected changed area.

The recently launched H2020 project URBANFLUXES (URBan ANthrpogenic heat FLUX from Earth observation Satellites) investigates the potential of EO to retrieve anthropogenic heat flux, as a key component in the Urban Energy Budget (UEB). URBANFLUXES advances existing Earth Observation (EO) based methods for estimating spatial patterns of turbulent sensible and latent heat fluxes, as well as urban heat storage flux at city scale and local scale. Independent methods and models are engaged to evaluate the derived products and statistical analyses provide uncertainty measures. Optical, thermal and SAR data are exploited to improve the accuracy of the UEB components spatial distribution calculation. Synergistic use of different types and of various resolution EO data allows estimates in local and city scale. Ultimate goal of the URBANFLUXES is to develop a highly automated method for estimating UEB components to use with Copernicus Sentinel data, enabling its integration into applications and operational services.

Spectral unmixing provides information on a sub-pixel level, which is extremely useful for studying the urban areas. Nevertheless, the high spatial diversity of man-made structures, the spectral variability of urban materials and the three-dimensional structure of the cities makes the sub-pixel mapping of urban surfaces one of the most challenging tasks of remote sensing science. In this study, these issues are addressed using an artificial neural network trained with endmember and non-linearly mixed synthetic spectra to inverse the pixel spectral mixture in high resolution multispectral imagery. A spectral library is built, consisting of endmember spectra collected from the images and synthetic spectra, produced using a non-linear model specifically developed for urban scenes. The proposed method is easily transferable to any city and fast in terms of computations, which makes it ideal for implementation with operational services for cities.

Detailed, frequent, and accurate land surface temperature (LST) estimates from satellites may support various applications related to the urban climate. When satellite-retrieved LST is used in modeling, the level of uncertainty is important to account for. In this letter, an uncertainty estimation scheme based on Monte Carlo simulations is proposed for local-scale LST products derived from image fusion. The downscaling algorithm combines frequent low-resolution thermal measurements with surface cover information from high spatial resolution imagery. The uncertainty is estimated for all the intermediate products, allowing the analysis of individual uncertainties and their contribution to the final LST product. Uncertainties of less than 2 K was found for most part of the test area. The uncertainty estimation method, although demanding in terms of computations, can be useful for the uncertainty analysis of other satellite products.

High-resolution thermal infrared (TIR) remote sensing has a wide range of applications. In this paper, we describe the different applications and requirements identified in a literature review and during a consultation meeting with researcher experts in different fields. As a result, more than 30 applications were identified within three different fields: 1) land and solid Earth; 2) health and hazards; and 3) security and surveillance. A complete set of requirements (spatial, temporal, and radiometric resolution, algorithms used, and supporting data, among others) for each application is also provided. The results presented in this paper provide useful information to enhance the importance of high-resolution TIR data for civil applications and may serve as a reference document for future TIR mission concepts.

Urban surfaces are highly inhomogeneous because of the high spatial and spectral diversity of man-made structures. Spectral unmixing techniques although developed to be used with hyperspectral data, are useful for assessing sub-pixel information on multispectral data as well. The large spectral variability imposes the use of multiple endmember spectral mixture analysis techniques, in which many possible mixture models are considered to produce the best fit. The use of many endmembers and mixture models result in prohibitive computational time. In this study, an artificial neural network is used to inverse the pixel spectral mixture in Landsat imagery. Endmember spectra, collected from the image were used to train the network and capture the spectral variability of man-made structures.

Volcanic eruptions affect at different levels the population and economy of interested areas. Moreover, volcanic ash detection represents a key issue for aviation safety due to the harming effects on aircraft. For these reasons, an accurate and fast analysis of the data is needed to monitor the phenomena's evolution and to manage the risk mitigation phase. In this scenario, the introduction of an inversion approach based on Neural Networks (NNs) has significant interest to reduce the need of human interpretation of the ash detection maps as those generated by the application of brightness temperature difference approach. In this work we show that NNs algorithms are suitable for an accurate mapping of ash cloud on Moderate Resolution Imaging Spectroradiometer (MODIS) images in a very cloudy scenario as the ones of 2010 Eyjafjallajökull and 2011 Grimsvötn eruptions.

In this paper a new approach from the combination of band ratioing function and MLP Neural Networks technique is proposed to differentiate between clouds and background in Landsat ETM+ and MSG SEVIRI data. First, in order to increase the contrast of the clouds and background, a band ratioing function is applied to each sub-image. Second, the sub-images are segmented by MLP Neural Networks technique. The proposed approach was tested on 40 Landsat ETM+ sub-images of Gulf of Mexico and on 40 MSG SEVIRI sub-images over Italy. The same parameters were used in all tests. For the overall dataset, the average accuracy of 89 % was obtained for Landsat ETM+ images and the average accuracy of 85 % was obtained for MSG SEVIRI images. Our experimental results demonstrate that the proposed approach is robust and effective.

Clouds are one of the most important meteorological phenomena affecting the Earth radiation balance. The increasing development of whole-sky images enables temporal and spatial high-resolution sky observations and provides the possibility to understand and quantify cloud effects more accurately. In this letter, an attempt has been made to examine the machine learning [multilayer perceptron (MLP) neural networks and support vector machine (SVM)] capabilities for automatic cloud detection in whole-sky images. The approaches have been tested on a significant number of whole-sky images (containing a variety of cloud overages in different seasons and at different daytimes) from Vigna di Valle and Tor Vergata test sites, located near Rome. The pixel values of red, green, and blue bands of the images have been used as inputs of the mentioned models, while the outputs provided classified pixels in terms of cloud coverage or others (cloud-free pixels and sun). For the test data set, the overall accuracies of 95.07%, with a standard deviation of 3.37, and 93.66%, with a standard deviation of 4.45, have been obtained from MLP neural networks and SVM models, respectively. Although the two approaches generally generate similar accuracies, the MLP neural networks gave a better performance in some specific cases where the SVM generates poor accuracy.

The use of COSMO-SkyMed constellation has been tested for post-fire monitoring activities. A typical Mediterranean ecosystem, seriously damaged by a wildfire, has been selected as study area. The multitemporal and multipolarization capabilities of COSMO-SkyMed have been exploited for automatic burnt area detection and monitoring of vegetation regrowth. Different imaging configurations have been tested to define the proper monitoring strategy in Mediterranean areas. Results showed X-band suitability for Mediterranean maquis post-fire monitoring.

Concurrent availability of VHR (Very High Resolution) images at both optical and microwave bands opens new challenges in many applicative scenarios of Earth Observation (EO). In particular this is true for precision farming activities where the retrieval on the metric scale of biophysical parameters and of information regarding vegetation spatial distributions can be very effective in supporting farmers during the production cycles. However, the inversion problem giving the value of the desired variable from the measured electromagnetic quantities (the image data) can be very complex and the nonlinear relationships involved need to be handled by suitable algorithms. In this paper a complete processing scheme providing quantities of interest for precision viticulture from data provided by WorldView-2 (WV2) and COSMOSkyMed (CSK) space platforms is presented. Once the appropriate season time was selected, the satellite data have been acquired over the test area within a limited time window and concurrently with the collection of the groundtruth. The workflow, besides adequate pre-processing steps, includes two neural networks (NN) modules, one is dedicated to the extraction of a restricted number of nonlinear components from the WV2 data, the other one to the actual inversion problem. The obtained results seem to be satisfactory with respect to the requirements provided by the users.

Dark-spot detection is a critical step in oil-spill detection. In this paper, a novel approach for automated dark-spot detection using synthetic aperture radar imagery is presented. A new approach from the combination of Weibull multiplicative model (WMM) and pulse-coupled neural network (PCNN) techniques is proposed to differentiate between the dark spots and the background. First, the filter created based on WMM is applied to each subimage. Second, the subimage is segmented by PCNN techniques. As the last step, a very simple filtering process is used to eliminate the false targets. The proposed approach was tested on 60 Envisat and ERS2 images which contained dark spots. The same parameters were used in all tests. For the overall data set, an average accuracy of 93.66% was obtained. The average computational time for dark-spot detection with a 512 × 512 image is about 7 s using IDL software, which is the fastest one in this field at present. Our experimental results demonstrate that the proposed approach is very fast, robust, and effective. The proposed approach can be applied on any kind of synthetic aperture radar imagery.

The operational level reached by polarimetric data processing techniques has been demonstrated during the last decade. The next generation of spaceborne Synthetic Aperture Radar satellites will implement full- or dual- polarimetric capabilities. In few years a huge amount of data will have to be processed in a fast and reliable way, implementing polarimetric decompositions or accurate classifications. Two neural network approaches for fast and accurate processing of polarimetric data are presented. In the first approach a neural network based processing chain for fast model based polarimetric decomposition is developed, while in the second approach a Non-Linear Principal Component Analisys of polarimetric data has been performed using an Auto Associative Neural Network. The results show a considerable reduction of computational effort and a substantial data compression with a minimun loss of information.

LeanEO! is a 2-year Earth Observation education project funded by the European Space Agency (ESA) and developed by different European Institutions. Its main aim is to increase the understanding and knowledge of satellite data obtained from ESA missions and demonstrate how these can be used when faced with environmental problems in the real world. The project has developed hands-on training resources for use primarily (but not exclusively) by teachers and students at upper high school to university level. Each lesson comes complete with data, analysis tools and exhaustive background information necessary for the completion of the suggested activities and provides answers to the various study questions. Model answers are supplied for users working on their own or with limited specialist support. In this paper the aims and the opportunities provided by the project will be described in detail.

Recent X-band SAR missions, such as COSMO-SkyMed (CSK), which is able to provide very high spatial resolution images of an area of interest with a short revisit time, are expected to be quite useful sources of information for monitoring the terrestrial environment and its changes. On the other hand, the huge amount of data involved, as well as the need to promptly act in case of emergency, requires the development of automatic change detection tools. This paper reports on a novel automatic change detection algorithm combining multilayer perceptron neural networks (NNs) and pulse coupled NNs, which has been implemented and tested on pairs of Stripmap and Spotlight CSK images acquired on the Tor Vergata University area in the southeast outskirts of Rome, Italy, where a significant and continuous urbanization process is occurring.

The mapping of the coastline is a well known and tested procedure exploiting the capabilities of optical satellite sensor. Nevertheless, it is affected by several inherent limits like weather condition and revisit time of the areas. The recent availability of very high-resolution X-Band SAR data acquired by a constellation of satellites with frequent revisit capabilities has brought a potential alternative, or support, for this kind of application. To this purpose a new automatic algorithm, based on Pulse Coupled Neural Networks, has been developed to process COSMO-SkyMed products taken with different polarization, geometric configuration and measurements mode. The results have been validated through a GPS survey, also respect to a traditional C-band technique applied on X-band, with the final intent of an assessment of the real impact of the proposed procedure in the coastal mapping application.

The potential of SAR data in discriminating vegetation/forest types it is here explored using Neural Networks (NN) in an Alpine environment. Amplitude data from two SAR polarimetric sensors, namely RADARSAT2 Standard Quad Polarization (SQP) and ALOS PALSAR Fine Beam Dual (FBD), were used separately and in conjunction to discriminate four vegetation types: conifer forest, broadleaved forest, riparian vegetation, and dwarf pine and shrubs (mainly composed by Pinus mugo species). Results indicate successful separation of needle-leaved from broadleaved and/or riparian vegetation, but scarce ability to discriminate the other two types. ALOS PALSAR produced better results in separating vegetation types with respect to RADARSAT2 reaching in the best case a K Cohen's coefficient equal to 0.88. Results obtained from combination of the two SAR data were successful, but still in the range of those obtained by single scene usage.

Aim of this paper is to show how fusing SAR images having different characteristics can improve the classification accuracy, in spite of the geometrical problems rising in the fusion operation. To this end, we fused multiple-frequency (Cand X-band), multiple-polarization (HH, HV, VH and VV) and multi-resolution images. The classification has been carried out by a neural network algorithm (NN), in which the backscattering coefficients at each polarization for each image have been fused to form the input to the classifier. The evaluation of the classification accuracy has been performed in terms of overall accuracy, per class accuracy and Kappa coefficient. The obtained results show not only an enhancement of the classification accuracies, but also that more land cover classes can be better identified with respect to a single acquisition.

In the paper are presented some of the main outcomes of the projects, relevant to marine monitoring, developed in the framework of the first Announcement of Opportunity Cosmo-SkyMed. The principal subjects covered are: ship detection, oil spill monitoring, wind retrieval, sea surface currents estimation and coastline extraction.

The novel instruments of the COSMO-SkyMed (CSK) Earth Observation programme, offer an opportunity to explore at various resolutions the information content of X-band signal backscattered with different polarizations. In spite of their potential to render additional information about an area of interest, speckle noise and artifacts make X-band acquisitions difficult to interpret. This is a motivating scenario to explore what (semi-)automatic procedures might be able to offer. This paper is first attempt to process CSK Stripmap PingPong data using two well-known artificial neural network techniques: the supervised backpropagation multilayer perceptron and the unsupervised self-organizing map.

We have analysed the seismic source of the active fault generated Van Mw=7.1 earthquake occurred in Eastern Turkey the 23rdOctober 2011. To this aim the surface displacement field has been measured applying SAR Interferometry (InSAR) technique to the available dataset of coseismic COSMO-SkyMed image pairs. The seismic source model has been obtained by the use of a data inversion procedure based on the concurrent application of InSAR techniques and Neural Networks. The proposed approach elaborates the information on the coseismic deformation pattern stemming from available differential interferograms. The interferogram is the expression of the active fault at depth, thus its shape, size and its features somehow refer to the geometry and slip of the fault generating the seism. A Neural Network has been trained to recognize some fault parameters (Length, Width, Strike, Dip, Depth) from the unwrapped interferogram. The retrieval exercise consists in estimating these parameters from the coseismic interferogram exploiting Neural Networks.

The possibility offered by space-based sensors represents an irreplaceable resource for monitoring in near real time the eruption activities. The high revisit time of sensor like MODIS, seems to be the most effective way to mitigate the aviation hazard imaging the phenomenon evolution. In this work we propose a neural networks based approach to the volcanic ash mass retrieval. In comparison with the techniques based on radiative transfer models, the proposed algorithm has shown similar accuracy and faster computation. This issue can be of real interest to address the problems inherent the volcanic activity in short time. A set of MODIS images collected during the Eyjafjallajokull eruption, occurred from the 14thof April to the 23rdof May 2010, has been used to analyze the performance variations due to different selection of the algorithm inputs, i.e. the MODIS channels from visible to thermal infrared electromagnetic spectrum. The best wavelength sets for the retrieval of the ash mass, optical thickness and effective radius have been identified by means of neural network pruning algorithm.

The goal of this work is to investigate on the enhancement, in terms of accuracy and number of classes, of the vegetation mapping through the joint use of multi-sensors data. Several stacks of Spotlight COSMO-SkyMed, acquired both in HH and VV polarization, and Multispectral World-View2 images, taken in the same or different seasons, have been compounded and exploited to identify six types of natural surfaces by means of a Neural Network classifier. While the information content of the eight bands of the multispectral data may be sufficient to discriminate the classes of interest, the single polarization of each SAR image has to be integrated by extracting further features, such as textural parameters. The assessment of the provided vegetation maps has been carried out in terms of per class accuracy, overall accuracy and K coefficient. The achieved results demonstrate the improvement of the classifications obtained by fusing more information from multi-sensors acquisitions.

Spaceborne measurements may significantly support monitoring the concentration of atmospheric constituents affecting air quality, such as ozone. However, retrieving tropospheric ozone concentration information from nadir satellite data is an arduous task, given the weak sensitivity of the earth's radiance to ozone variations in the lower part of the atmosphere. We propose a new methodology, based on neural networks (NN), for retrieving the tropospheric ozone column from SCanning Imaging Absorption spectroMeter for Atmospheric CHartographY (SCIAMACHY) nadir UV/VIS measurements. The design of the NN algorithm is based on an analysis of the information content of measurements in both UV and VIS bands, carried out by a combined radiative transfer model and NN extended pruning procedure. The NN was trained and tested with simulated data and with matching World Ozone and Ultraviolet radiation Data Centre ozonesonde data sets and validated by independent data taken over two test sites. A significant improvement of the retrieval capabilities is observed when VIS wavelengths are included into the input vector. Finally, an example of tropospheric ozone map generated automatically by the methodology at a continental scale is provided and critically discussed.

The new generation of spaceborne instruments, capable of capturing a large amount of very-high resolution images within a short revisit time, is allowing remote sensing researchers and final users to receive huge amounts of data in rather short times. Such a scenario makes it mandatory the development of techniques, as much as possible automatic, for the understanding and the effective exploitation of the available information. This contribution deals with the features extraction from Spotlight Cosmo-SkyMed SAR imagery (1 m spatial resolution) by means Multi Layer Perceptron Neural Network (MLP-NN) algorithms. For a better pixel characterization, textural parameters have been also considered as additional information for the classification procedure.

In this work, fully polarimetric RADARSAT2 SAR images and advanced feature extraction strategies are investigated for improving the retrieval of soil moisture content in Alpine meadows and pastures. More in detail, standard Intensity & Phase polarimetric features, polarimetric decompositions and general purpose feature extraction strategies are exploited in combination with a sequential forward selection to increase the accuracy of the system. The capability of the system to provide spatially and temporally distributed estimates of soil moisture is also addressed by using different satellite acquisitions. The achieved results indicate that the polarimetric information in the SAR data, if properly exploited, allows one to improve the estimation of soil moisture content in the investigated mountain area. Concerning the mapping of the target variable, the analysis of the results suggest that the proposed estimation system is promising and effectively maps the soil moisture status both in time and space.

The great eruption of the Icelandic Eyjafjallajokull volcano that occurred from the 14thof April to the 23rdof May 2010 injected large and dense ash clouds into the atmosphere, causing major international air traffic disruption worldwide.

Purpose of this work is the study of cloud detection techniques. This work identifies the cloud cover of optical images acquired by the QuickBird satellite, comparing these with others of the same area, acquired by Landsat 7 in which there are no clouds. The images are combined using an early fusion technique [1]. The tool exploits the neighborhood model [2] for increasing the amount of information for the training set and the Singular Value Decomposition for carrying out the feature extraction [3]. In order to introduce these structures into thematic classification tasks by SVMs it was necessary develop a tree kernel function based on tree kernel function defined in SVM-LightTK. The aim of the tree kernel function is evaluate the similarity level between a generic couples of tree structures.In this paper we report the results obtained comparing the performance of different approaches in cloud classification problem. The final purpose is the production of cloud cover maps. Throughout such different experimental setups we measured the capabilities of each algorithm under different points of view. First of all, we considered the classification accuracy by computing traditional parameter such as overall accuracy. A second analysis regarded the efforts that are required in the design of optimal algorithms. Indeed, these techniques are characterized by different parameters that have to be appropriately tuned in order to obtain the best performance. Finally the robustness of the techniques has been also considered. In particular the classification accuracy has been evaluated also for images not considered in the training phase.

In this letter, a method to automatically retrieve building surface temperature maps using hyperspectral imagery is presented. The approach can be conceptually described by considering two different problems. The first consists in the design of an automatic procedure for the extraction of building surfaces from the hyperspectral image. Such an issue has been addressed using both unsupervised and supervised neural networks. The second problem deals with the retrieval of land surface temperature from the same image. The final step is the merging of the temperature map with the building mask. It is worthwhile to observe that the proposed approach aims at retrieving the temperature values by reducing the manual editing and the use of ancillary data to a minimum level. The obtained results show an accuracy in the building identification of 83.7% and a root-mean-square error (rmse) in the temperature retrieval of 1.59 K. The importance of this methodology has to be considered within the studies on urban heat islands, which is becoming an important issue in urban management politics.

Neural networks (NNs) are recognized as very effective techniques when facing complex retrieval tasks in remote sensing. In this paper, the potential of NNs has been applied in solving the unmixing problem in hyperspectral data. In its complete form, the processing scheme uses an NN architecture consisting of two stages: the first stage reduces the dimension of the input vector, while the second stage performs the mapping from the reduced input vector to the abundance percentages. The dimensionality reduction is performed by the so-called autoassociative NNs, which yield a nonlinear principal component analysis of the data. The evaluation of the whole performance is carried out for different sets of experimental data. The first one is provided by the Airborne Hyperspectral Scanner. The second set consists of images from the Compact High-Resolution Imaging Spectrometer on board the Project for On-Board Autonomy satellite, and it includes multiangle and multitemporal acquisitions. The third set is represented by Airborne Visible/InfraRed Imaging Spectrometer measurements. A quantitative performance analysis has been carried out in terms of effectiveness in the dimensionality reduction phase and in terms of the accuracy in the final estimation. The results obtained, when compared with those produced by appropriate benchmark techniques, show the advantages of this approach.

The basic idea of this paper relies on the concurrent exploitation of the capabilities of neural networks (NNs) and SAR interferometry (InSAR) for the characterization of a seismic source and the estimation of its geometric parameters. When a moderate-to-strong earthquake occurs, we can apply the InSAR technique to compute a differential interferogram. The earthquake is generated by an active seismogenic fault having its own specific geometry. The corresponding differential interferogram contains, in principle, information concerning the geometry of the seismic source that the earthquake comes from. To perform the inversion operation, a novel approach based on NNs is considered. This requires the generation of a statistically significant number of synthetic interferograms necessary for the network training phase. Each of them corresponds to a different combination of fault geometric parameters. After the training, the network is ready to perform, in real time, the inversion on new differential interferograms. This paper illustrates such a methodology and its validation on a set of experimental data.

The information enhancement of an hyperspectral image through building shadows removal is presented. Starting from the energy conservation model and considering the development of the simulated reflectance algorithm, the study demonstrates how it is possible to exploit the wide spectrum of an hyperspectral image in urban areas allowing the visualization of image details which were shaded by urban structures. Moreover, the described methodology positively affects land classification: the accuracy detection of four land cover classes (vegetation, buildings, asphalt and bare soil) has been improved using as input to a neural network classifier simulated reflectance image instead of the original one. The experimental data consisted of an Airborne Hyperspectral Scanner (AHS) image acquired over the city of Madrid.

Feature extraction for the dimensionality reduction of hyperspectral data is performed by means of Auto-Associative Neural Networks. The algorithm performance is compared to the Principal Component Analysis and the Maximum Noise Fraction ones. Results of land cover pixel-based maps yielded by the reduced vector and a dedicated neural network classification algorithm are also reported.

Information mining from heavy SAR images is considered from the point of view of the procedure automatization. Two schemes based on Neural Networks are evaluated, one based on the Self Organizing Map method exploiting polarimetric information and oriented to land cover classification, the other based on the Pulse-Coupled Neural Networks aiming at characterizing the imaged buildings.

A novel approach based on pulse-coupled neural networks (PCNNs) for image change detection is presented. PCNNs are based on the implementation of the mechanisms underlying the visual cortex of small mammals, and, with respect to more traditional NNs architectures, such as multilayer perceptron, own interesting advantages. In particular, they are unsupervised and context sensitive. This latter property may be particularly useful when very high resolution images are considered as, in this case, an object analysis might be more suitable than a pixel-based one. The qualitative and more quantitative results are reported. The performance of the algorithm has been evaluated on a pair of QuickBird images taken over the test area of Tor Vergata University, Rome.

In this paper we test an unsupervised neural network approach for extracting features from very high resolution X-band SAR images. The purpose of this study is buildings recognition in images of low density urban areas, acquired by COSMO-Skymed and TerraSAR-X satellites, by means of Pulse Coupled Neural Network (PCNN), a relatively novel unsupervised algorithm based on models of the visual cortex of small mammals. The features retrieved from geo-referenced SAR images are compared against the ground truth provided by corresponding optical images. The accuracy yielded by PCNN is quantitatively evaluated and critically discussed, also in comparison with commonly used feature extraction techniques.

In this paper Autoassociative Neural Networks (AANN) are used to implement Nonlinear Principal Component Analysis (NLPCA) for dimension reduction of hyperspectral data. The nonlinear components are then considered as inputs for a Multi-Layer Perceptron (MLP) network to perform pixel-based classification. The methodology has been applied considering the test area of Tor Vergata — Frascati, Italy, and the hyperspectral data provided by the CHRIS-PROBA mission. Comparative analysis with a similar procedure considering a more standard dimensionality reduction technique such as Principal Component Analysis (PCA) has been carried out.

The increasing number of satellite missions providing more and more data for updating land cover and land use maps requires to upgrade the level of automatism for the processing of remotely sensed imagery. In this paper we try to pursue the ambitious goal of designing a completely automatic (no human interaction) supervised scheme for the classification, in terms of land cover, of a multi-spectral image. An expert system, using appropriate spectral and textural features, drives the selection of suitable training pixels in the image. These are used for the learning of a neural network algorithm that successively performs the pixel-based land cover classification of the whole image. The processing scheme has been tested on a set of Landsat images taken on different European urban areas.

The basic idea of this paper relies on the concurrent exploitation of the capabilities of neural networks and SAR interferometry for the characterization of a seismic source and the estimation of its geometric parameters. When a moderate-to-strong earthquake occurs we can apply SAR Interferometry (InSAR) technique to compute a differential interferogram. The earthquake has been generated by an active, seismogenic, fault having its own specific geometry. Therefore each differential interferogram contains in principle information concerning the geometry of the seismic source the earthquake comes from. To perform the inversion operation an approach based on neural networks can be considered. The paper illustrates such a methodology and its assessment on experimental data.

Many algorithms for retrieving geophysical variables are based on optimal estimation approaches, which can be time consuming specially if a large amount of data is to be processed. On its part, neural networks provide results almost in real time, but their use is still not generalised for remote sensing applications. In this work, a set of neural networks was trained with simulations using numerical land emission models and tested using L-band radiometric data of bare soils acquired during the T-REX and MOUSE field experiments. Soil moisture retrieved by the neural networks was then compared to ground-truth data.

We discuss the development and application of a Complex-Valued Neural Network (CVNN) algorithm for retrieving forest biomass from polarimetric interferometric SAR data. After discussing some features of the net and of the training procedures, we analyze the performance of the algorithm in inverting combinations of simulated radar backscattering at different polarization states. The CVNN performance is compared with that of other retrieval algorithms.

This paper studies the impact of several learning issues in an image classification task with SVMs, such as rich feature-based representations, optimization and sensitivity to novelty in the test data sets. The employed imagery refers to the city of Rome, Italy and is acquired in different years and seasons by the European Remote Sensing Satellites ERS-1 and ERS-1/2 tandem mission. A comprehensive evaluation according to varying training conditions is reported, showing that SVMs provide robust and largely applicable tools.

Since a few months TerraSAR-X has been acquiring X-band SAR images of the earth surface from space. This contribution reports on a study carried out to understand the main textural features of the X-band radar return from various kinds of surfaces and in particular to assess the potential of images acquired by X-band space borne radars in mapping fire scars and in classifying suburban/agricultural land cover. To this end, a novel unsupervised neural network algorithm, the Textural Self-Organizing Map (TexSOM), based on the textural features of the radar image, has been worked out and tested on areas in Greece and Italy.

This paper reports the study of supervised neural network algorithm for classification purposes. SPOT 5 and TerraSAR-X dataset are analyzed. Classification results are critically discussed and compared to ground truth map and unsupervised neural classification of the same area. The aim is to demonstrate the capability of neural networks in managing heterogeneous dataset and the accuracy improvement obtained by the use of the textural object based layers fused with the optical and radar data.

In this paper we report on the different land cover classification performances, over the same urban and rural landscape, of Multi-spectral Landsat ETM and Hyperspectral CHRIS-PROBA instruments. A Landsat ETM+ acquisition of 19 August 2004 and a CHRIS Proba image of 19 August 2006 over the same test area of Frascati-Tor Vergata have been considered. Even though a time shift of two years is present, the images have been acquired in the same day of the year, which makes them sufficiently suitable for the comparison. For the classification task a neural network algorithm is considered for both type of images. Indeed neural algorithms in last years have demonstrated at the same time a particular ease on managing a large domain of inputs and an important effectiveness in performing the decision task for the classification of remotely sensed images. A quantitative analysis in terms of classification accuracy on the different types of surfaces is carried out and the results critically analysed.

A concurrent exploitation of the capabilities of neural networks and SAR interferometry is considered for the characterization of a seismic source and the estimation of its geometric parameters. This requires the generation of an appropriate number of synthetic interferograms necessary for the network training phase. After the training the network is ready to perform the inversion on new interferograms. The paper illustrates the assessment of such a methodology over a significant set of experimental data.

The paper aims at improving image automatic classification from remote sensing data using neural network algorithms (NN). NN have been found to have good generalization properties and their use is becoming increasingly prevalent in the field of remote sensing and in particular for image classification. However, the type of input to be considered for the algorithm in order to maximize the information available from the measurement is still an open issue. Using the mere spectral signature with no pre-processing is not an effective choice. Another point regards the use of textural features to improve the classification which involves taking decisions on how many and what specific features should be considered. More in general, minimizing the number of inputs of a neural network algorithm, avoiding significant loss of information, affects positively the NN mapping ability and computational efficiency. In this paper we propose a new methodology facing with the aforementioned problems and providing a solution to them.

The design and development of two neural network (NN) algorithms for retrieving atmospheric height-resolved temperature and nitrogen dioxide from UV/VIS satellite radiance data are presented. Sensitivity analyses, NNs optimization and results are discussed, based on radiative transfer simulations. The approach may lead to synergistic monitoring of different atmospheric parameters from the same UV/VIS radiance measurement.

In this paper we report on the design of Neural Networks algorithms to retrieve height resolved ozone information from Envisat SCIAMACHY and Aura OMI Level 1 data. We defined as input-output pairs the matching of (a) SCIAMACHY UV/VIS reflectances with ozonesondes concentrations, and(b) OMI UV/VIS reflectances with MLS concentrations. Design issues, as input vector dimensionality reduction, vertical resolution problems and topology selection are here analyzed. The inversion results are presented and discussed, with a special emphasis to retrievals at tropospheric height levels.

An active learning method is proposed for the semi-automatic selection of training sets in remote sensing image classification. The method adds iteratively to the current training set the unlabeled pixels for which the prediction of an ensemble of classifiers based on bagged training sets show maximum entropy. This way, the algorithm selects the pixels that are the most uncertain and that will improve the model if added in the training set. The user is asked to label such pixels at each iteration. Experiments using support vector machines (SVM) on an 8 classes QuickBird image show the excellent performances of the methods, that equals accuracies of both a model trained with ten times more pixels and a model whose training set has been built using a state-of-the-art SVM specific active learning method.

The 2007 Data Fusion Contest that was organized by the IEEE Geoscience and Remote Sensing Data Fusion Technical Committee was dealing with the extraction of a land use/land cover maps in and around an urban area, exploiting multitemporal and multisource coarse-resolution data sets. In particular, synthetic aperture radar and optical data from satellite sensors were considered. Excellent indicators for mapping accuracy were obtained by the top teams. The best algorithm is based on a neural classification enhanced by preprocessing and postprocessing steps.

Parametric and nonparametric approaches to evaluate land-cover change detection using very high resolution (VHR) satellite imagery are applied to the analysis of the demolition of the Rocky Flats nuclear weapons facility located near Denver, CO. Both maximum-likelihood and neural network classifiers are used to validate a new parallel architecture which improves the accuracy when applied to VHR satellite imagery for the study of land-cover change between sequential satellite acquisitions. An enhancement of about 14% was found between the single-step classification and the new parallel architecture, confirming the advantage and the robust improvement obtained with this architecture regardless of the classification algorithm used. In this paper, we demonstrate and document the demolition and removal of hundreds of buildings taken down to bare soil between 2003 and 2005 at the Rocky Flats site.

The small hyperspectral imager Compact High-Resolution Imaging Spectrometer (CHRIS) is the most important instrument for Earth observation included in the payload of the European Space Agency Third-Part Mission Project for On-Board Autonomy (PROBA)-1 satellite. This instrument has provided dozens of images in several target areas in the world, and a good number of acquisitions are available for the test site of Frascati and Tor Vergata, Italy. This paper reports several results concerning the generation of thematic maps obtained from CHRIS mode-3 imagery. The potential of the use of different configurations for the input vector exploiting multispectral, multiangular, and multitemporal measurements has been investigated, and the results have been evaluated and compared in terms of accuracy in the classification. The core of the decision task has been developed using the neural network methodology. Indeed, this approach is characterized by a particular ease in performing nonlinear mapping of a multidimensional set of inputs into the output one.

Polarimetric SAR interferometric data can provide estimates of forest biomass density. There are different approaches to deal with the inversion problem, such as neural networks and the traditional optimal estimation approach. This paper presents a study to evaluate their performance by means of quantitative indexes addressing both the computation time and the retrieval accuracy. Better forest parameters estimates have been obtained when neural networks algorithms were used.

Analysis of L-band polarimetric SAR data has not been extensively carried out for undulating, heterogeneous and fragmented landscapes, where classification can become quite challenging. This paper reports results of a study on the pixel-by- pixel unsupervised classification of very-high resolution polarimetric images by self-organizing neural networks.

The small hyperspectral imager CHRIS is the principal instrument for the remote sensing on board of the ESA Proba-1 satellite. The platform can tilt itself in the orbit allowing the acquisition of images with different angle over the same area and during the same overpass. In this work we report on the potentialities of such instrument in producing land cover maps. The images that we are using have been acquired over the test site of Frascati and Tor Vergata University campus, located in the surroundings of Rome, Italy, which is characterized by a mixing of several topologies of vegetated areas with residential and commercial buildings. The neural network approach has been used for the decision task. The results of the classification have been evaluated both by the visual interpretation and quantitative assessment. This latter demonstrated an accuracy of more than the 90%.

A comparative study on the complexity of the urban environments in SAR imagery at different spatial resolutions is presented. Two datasets have been considered, including the city of Rome, Italy imaged in decametric resolution and the Frascati area (Rome, Italy) acquired in very high spatial resolution (∼2 m) at L-band in a fully polarimetric mode. The different characteristics of the radar sensors require careful managing of the corresponding product capabilities to maximize the various pieces of information contained in the variety of scattering mechanisms.

The advent of very high spatial resolution optical satellite imagery has greatly increased our ability to monitor land cover changes in urban environments where the spatial resolution plays a key role related to the detection of fine-scale objects such as a single house or small structures. At the same time, very high spatial resolution imagery presents a new challenge over other satellite systems, in that a relatively large amount of data must be analyzed and corrected for registration and classification errors to identify the land cover changes, commonly resulting in a very extensive manual work. To improve on this situation we have developed a new method for land surface change detection that greatly reduces the human effort needed to remove the errors that occur with many methods applied to very high spatial resolution imagery. This change detection algorithm is based on Neural Networks and it is able to exploit in parallel both the multi-band and the multi-temporal data to discriminate between real changes and false alarms. In general the classification errors are reduced by a factor of 2–3 using this new method over a simple Post Classification Comparison based on a neural network classification of the same images.

In this paper we report on the design of a Neural Networks algorithm to retrieve tropospheric ozone information from satellite data. Following a combined radiative transfer model-extended pruning sensitivity analysis for input wavelengths selection, we first made an inversion exercise based on a syn thetically produced radiance-tropospheric ozone concentrations database. Starting from the encouraging obtained results, we tested the Net on ESA-ENVISAT SCIAMACHY Level lb data. A time series of Tropospheric Ozone Columns on some midlatitude sites has been retrieved from the satellite measurements and then compared with collocated and simultaneous ozonesondes reference columns. The inversion results are presented and critically discussed.

The effectiveness of multilayer perceptron (MLP) networks as a tool for the classification of remotely sensed images has been already proven in past years. However, most of the studies consider images characterized by high spatial resolution (around 15–30 m) while a detailed analysis of the performance of this type of classifier on very high resolution images (around 1–2 m) such as those provided by the Quickbird satellite is still lacking. Moreover, the classification problem is normally understood as the classification of a single image while the capabilities of a single network of performing automatic classification and feature extraction over a collection of archived images has not been explored so far. In this paper, besides assessing the performance of MLP for the classification of very high resolution images, we investigate on the generalization capabilities of this type of algorithms with the purpose of using them as a tool for fully automatic classification of collections of satellite images, either at very high or at high-resolution. In particular, applications to urban area monitoring have been addressed.

The advent of new high spatial resolution optical satellite imagery has greatly increased our ability to monitor land cover changes from space. Satellite observations are carried out regularly and continuously, and provide a great deal of insight into the temporal changes of land cover use. High spatial resolution imagery better resolves the details of these changes and makes it possible to overcome the "mixed-pixel" problem that is inherent with more moderate resolution satellite sensors. At the same time, high-resolution imagery presents a new challenge over other satellite systems, in that a relatively large amount of data must be analyzed and corrected for registration and classification errors to identify the land cover changes. To obtain the accuracies that are required by many applications to large areas, very extensive manual work is commonly required to remove the classification errors that are introduced by most methods. To improve on this situation, we have developed a new method for land surface change detection that greatly reduces the human effort that is needed to remove the errors that occur with many classification methods that are applied to high-resolution imagery. This change detection algorithm is based on neural networks, and it is able to exploit in parallel both the multiband and the multitemporal data to discriminate between real changes and false alarms. In general, the classification errors are reduced by a factor of 2-3 using our new method over a simple postclassification comparison based on a neural-network classification of the same images.

This contribution discusses the kind of information about land use in urban and sub-urban areas contained in multitemporal SAR data and suggests how it can be exploited for classifying a wide, complex and heterogenous landscape. Rome, Italy, is the test site. Multitemporal, coherence and textural features are obtained from a set of SAR images taken in winter, spring and summer by the ERS tandem mission. These features are used to identify areas belonging to diverse urban classes, including water surfaces, woodland and parks, and continuous high/low density residential areas. The decision-making process is performed by a classifier based on a neural network algorithm.

In continuation of the BACCHUS project, aimed at establishing a reference high quality geographic information system for vineyards, an airborne SAR survey has been carried out in fall 2005 in the Frascati area, near Rome (Italy) to demonstrate the potential of airborne radar remote sensing in vineyard characterization. This contribution reports on the polarimetric L-band and dual polarization C-band SAR data acquisition campaign supported by ESA and carried out on two dates in October 2005 (the first during the grape harvest and the other after the vintage completion).

In the recent years InSAR (Synthetic Aperture Radar Interferometry) technique showed its wide potentialities to detect the surface displacement field due to an earthquake. Of great interest and usefulness in this context is the solution of the inverse problem that means to recover the source parameters from the knowledge of InSAR surface displacement field. In this work a novel approach for the solution of such a problem is presented.

Synthetic Aperture Radar (SAR) images provided by satellite missions may provide a significant support for oil spill detection over the sea. In particular neural networks algorithms have recently demonstrated their potentialities for discrimination between oil spills and objects which resemble oil spills (called "look-alikes"). The main steps of the classification procedure are the identification of dark spots over the sea, the computing of a set of parameters (features) for each dark spot and the classification of the oil spill candidate using a trained neural network, where the network input is a vector containing the values of the features extracted. The features so far mainly consist of physical-geometrical characteristics of the dark spot. This study presents a new neural network algorithm for the oil spill detection. The results also report a sensitivity analysis of the classification performance on the quantities that are given as input to the neural network. Among the considered inputs the value of the local wind speed has been also included.

Several studies have reported the potentialities of high resolution multi-spectral imagery for classifying and monitoring urban areas [A. K. Shackelford et al. (2003)], [M. Pesaresi et al. (2000)], [G. Schiavon et al. (2003)]. In this paper we present the results obtained by processing high resolution multispectral QuickBird images of an urban area. The high resolution QuickBird data have been used for two different purposes: for an automatic image classification using neural network techniques and for a change detection analysis. In the first case, we have carried out a pixel-based classification procedure aimed at the discrimination among 4 main classes: buildings, roads, vegetated areas, bare soil; then we have examined the potentialities of Kohonen maps for discovering new subclasses within those already established: e.g. for the asphalt category, different subclasses such as highways pixels and the other different types of roads such as secondary street pixels have been identified. In the second case we have processed multitemporal QuickBird images for detecting major changes occurred over the selected test area, like news buildings not visible in the first image.

This paper describes an algorithm aimed at monitoring the soil moisture and the growth cycle of wheat fields using radar data. The algorithm is based on neural networks trained by model simulations and multitemporal ground data measured on fields taken as a reference. The backscatter of wheat canopies is modeled by a discrete approach, based on the radiative transfer theory and including multiple scattering effects. European Remote Sensing satellite synthetic aperture radar signatures and detailed ground truth, collected over wheat fields at the Great Driffield (U.K.) site, are used to test the model and train the networks. Multitemporal, multifrequency data collected by the Radiometer-Scatterometer (RASAM) instrument at the Central Plain site are used to test the retrieval algorithm.

We discuss the application of neural network algorithms (NNAs) for retrieving forest biomass from multifrequency (L- and P-band) multipolarization (hh, vv, and vv) backscattering. After discussing the training and pruning procedures, we examine the performances of neural algorithms in inverting combinations of radar backscattering coefficients at different frequencies and polarization states. The analysis includes an evaluation of the expected sensitivity of the algorithm to measurement noise stemming both from speckle and from fluctuations of vegetation and soil parameters. The NNA accomplishments are compared with those of linear regressions for the same channel combinations. The application of NNAs to invert actual multifrequency multipolarization measurements reported in literature is then considered. The NNA retrieval accuracy is now compared with those yielded by linear and nonlinear regressions and by a model-based technique. A direct analysis of the information content of the radar measurements is finally carried out through an extended pruning procedure of the net.

This paper describes an algorithm aimed at monitoring the soil moisture and the growth cycle of wheat fields using radar data. The algorithm is based on neural networks trained by an electromagnetic model and multitemporal ground data measured on fields taken as a reference. The retrieval procedure is tested using mutitemporal signatures collected at a test site.

This contribution is focussed onto clarifying some features and discussing the performances of Neural Network Algorithms (NNA's) for retrieving forest biomass density from multifrequency multipolarization backscattering.

We report on an investigation aimed at evaluating the performance of a neural-network based crop classification technique, which makes use of multi-polarization and/or multi-temporal back-scattering coefficients measured at C-band.

This paper reports on the results of the application of SAR interferometry to the study of a landslide disaster which happened in the Sarno area (Italy) in 1998. A DEM generated by means of an ERS-1/2 tandem pair has been compared with a topographic one allowing the identification of the areas mostly affected by the landslide.

The potentiality of high resolution multi-spectral imagery of QuickBird satellite for monitoring urban areas and the sinergy with multitemporal SAR data is presented in this paper. The high resolution imagery can be applied for two purposes: for the analysis of the different spectral behaviour of several types of surface present in the area under observation and as ground-truth for the validation of procedures using only ERS-SAR data as input. In the first case, pixel-based classification procedures can be designed, trying to discriminate among various urban classes, including buildings, roads and vegetated areas. The decision-making process is performed by a neural network algorithm. The synergy with ERS-SAR involves the often not trivial co-registration process. Once this is completed, we can use multitemporal radar imagery for detecting major changes in the urban manufactures like news buildings and structures not visible in the older images, and validate the results with the help of the high resolution imagery. In this paper we report on some results obtained by applying the high resolution imagery of QuickBird for the analysis of the Tor Vergata University campus urban test area, located in Italy, South-East of Rome, whose extension is of about 600 ha. The results are divided in two sections. In the first one only the use of Quickbird data is considered, in the second one we report on the benefic effects carried on by the synergy with SAR imagery.

This paper reports on an investigation aimed at evaluating the performance of a neural-network based crop classification technique, which makes use of backscattering coefficients measured in different C-band synthetic aperture radar (SAR) configurations (multipolarization/multitemporal). To this end, C-band AirSAR and European Remote Sensing Satellite (ERS) data collected on the Flevoland site, extracted from the European RAdar-Optical Research Assemblage (ERA-ORA) library, have been used. The results obtained in classifying seven types of crops are discussed on the basis of the computed confusion matrices. The effect of increasing the number of polarizations and/or measurements dates are discussed and a scheme of interyear dynamic classification of five crop types is considered.

An algorithm, based on an electromagnetic model and a neural network, aimed at monitoring the multitemporal evolution of wheat fields, is described. Three different sites are used to validate the model, provide reference ground data, and test the algorithm.

Considers the retrieval of cloud liquid from the brightness temperature measurements taken by a ground-based multifrequency microwave radiometer using a neural network-based inversion algorithm. Comparisons between the retrieved quantity and the atmospheric information provided by active instruments such as a 94-GHz radar and a lidar ceilometer, co-located on the RAL Chilbolton Observatory site (UK), are carried out.

The thermal structure of trace gases, their distribution in the atmosphere, and their circulation mechanisms result from a complex interplay between radiative, physical, and dynamical processes. Neural-network algorithms can be a useful tool to face such complexities in retrieval operations. In this paper, their potentialities have been exploited to design real-time procedures for the estimation of vertical profiles of ozone concentration from spectral radiances measured by GOME, the first instrument of the European Space Agency capable of monitoring global distribution of ozone and other trace gases.

Abstract—A neural network approach for semi-automatic de-tection of oil spills in European remote sensing satellite-synthetic aperture radar (ERS-SAR) imagery is presented. The network input is a vector containing the values of a set of features char-acterizing an oil spill candidate. The classification performance of the algorithm has been evaluated on a data set containing verified examples of oil spill and look-alike. A direct analysis of the information content of the calculated features has been also carried out through an extended pruning procedure of the net. Index Terms—ERS-synthetic aperture radar (SAR), neural net-works, oil spill detection. I.

A new neural network algorithm for the inversion of radiometric data to retrieve atmospheric profiles of temperature and vapor has been developed. The potentiality of the neural networks has been exploited not only for inversion purposes but also for data feature extraction and dimensionality reduction. In its complete form, the algorithm uses a neural network architecture consisting of three stages: 1) the input stage reduces the dimension of the input vector; 2) the middle stage performs the mapping from the reduced input vector to the reduced output vector; 3) the third stage brings the output of the middle stage to the desired actual dimension. The effectiveness of the algorithm has been evaluated comparing its performance to that obtainable with more traditional linear techniques.

We report experiments conducted to investigate the correlation between code coverage and software reliability. Black-, decision-, and all-use-coverage measures were used. Reliability was estimated to be the probability of no failure over the given input domain defined by an operational profile. Four of the five programs were selected from a set of Unix utilities. These utilities range in size from 121 to 8857 lines of code, artificial faults were seeded manually using a fault seeding algorithm. Test data was generated randomly using a variety of operational profiles for each program. One program was selected from a suite of outer space applications. Faults seeded into this program were obtained from the faults discovered during the integration testing phase of the application. Test cases were generated randomly using the operational profile for the space application. Data obtained was graphed and analyzed to observe the relationship between code coverage and reliability. In all cases it was observed that an increase in reliability is accompanied by an increase in at least one code coverage measure. It was also observed that a decrease in reliability is accompanied by a decrease in at least one code coverage measure. Statistical correlations between coverage and reliability were found to vary between -0.1 and 0.91 for the shortest two of the five programs considered; for the remaining three programs the correlations varied from 0.89 to 0.99.