We introduce a novel neural network architecture—spectral encoder for sensor independence (SEnSeI)—by which several multispectral instruments, each with different combinations of spectral bands, can be used to train a generalized deep learning model. We focus on the problem of cloud masking, using several preexisting datasets, and a new, freely available dataset for Sentinel-2. Our model is shown to achieve state-of-the-art performance on the satellites on which it was trained (Sentinel-2 and Landsat 8) and is able to extrapolate to sensors that it has not seen during training, such as Landsat 7, PerúSat-1, and Sentinel-3 Sea and Land Surface Temperature Radiometer (SLSTR). Model performance is shown to improve when multiple satellites are used in training, approaching, or surpassing the performance of specialized, single-sensor models. This work is motivated by the fact that the remote sensing community has access to data taken with a huge variety of sensors. This has inevitably led to labeling efforts being undertaken separately for different sensors, which limits the performance of deep learning models, given their need for huge training sets to perform optimally. Sensor independence can enable deep learning models to utilize multiple datasets for training simultaneously, boosting performance, and making them much more widely applicable. This may lead to deep learning approaches being used more frequently for onboard applications and in ground segment data processing, which generally requires models to be ready at launch or soon afterward.

This study proposed a new method of retrieving 10-m spectral surface albedo products. Three crucial components are incorporated into this high-resolution surface albedo generation system. Firstly, a deep learning system, CloudFCN based on the U-net paradigm has been developed. This produces the best available cloud detection of any algorithm published to date. Secondly, an advanced atmospheric correction model, the Sensor Invariant Atmospheric Correction (SIAC) is employed. The SIAC method considers the surface BRDF effects as these are usually ignored, because the atmosphere correction is a large signal and the largest uncertainty in converting top-of-atmosphere reflectance to top-of-canopy surface reflectance. Thirdly, an endmember-based new technology will be used to retrieve high-resolution albedo from high-resolution reflectance by combining downscaled MODIS BRDF. These methods are further described alongside results shown of the different stages and the final high resolution albedo.

We address the problem of automatically coregistering planetary images to a common baseline, introducing a novel generic technique that achieves an unprecedented robustness to different image inputs, thus making batch-mode coregistration achievable without requiring the usual parameter tweaking. We introduce a novel image matching technique, which boosts matching performance even under the most strenuous circumstances, and experimentally demonstrate validation through an extensive experimental multi-instrument setup that includes images from eight high-resolution data sets of the Mars and the Moon. The technique is further tested in a batch-mode processing, in which approximately 1.6% of all high-resolution Martian imagery is coregistered to a common baseline.

Data from Earth Observation (EO) satellites are increasingly used to monitor the environment, understand variability and change, inform evaluations of climate model simulations and forecasts and manage natural resources. Policy makers are progressively relying on information derived from EO data to make decisions on mitigating and adapting to climate change. These decisions should be evidence based, which requires complete confidence in satellite-derived products as well as the in situ measurements used to calibrate, validate or complement these data. This paper provides an overview of the components required to develop robust quality assurance procedures for terrestrial essential climate variable (ECV) data products derived from Earth Observation (EO) satellite datasets.

In this paper, we address the problem of fast and accurate extraction of points that correspond to the same location (named tie-points) from pairs of large-sized images. First, we conduct a theoretical analysis of the performance of the full-image matching approach, demonstrating its limitations when applied to large images. Subsequently, we introduce a novel technique to impose spatial constraints on the matching process without employing subsampled versions of the reference and the target image, which we name coupled image decomposition. This technique splits images into corresponding subimages through a process that is theoretically invariant to geometric transformations, additive noise, and global radiometric differences, as well as being robust to local changes. After presenting it, we demonstrate how coupled image decomposition can be used both for image registration and for automatic estimation of epipolar geometry. Finally, coupled image decomposition is tested on a data set consisting of several planetary images of different size, varying from less than one megapixel to several hundreds of megapixels. The reported experimental results, which includes comparison with full-image matching and state-of-the-art techniques, demonstrate the substantial computational cost reduction that can be achieved when matching large images through coupled decomposition, without at the same time compromising the overall matching accuracy.

The longevity and dispersion of smoke and associated chemical constituents released from wildfire events are dependent on several factors, crucially including the height at which the smoke is injected into the atmosphere. The aim here is to provide improved emission data for the initialization of chemical transport models in order to better predict aerosol and trace gas dispersion following injection into the free atmosphere. A new stereo-matching algorithm, named M6, which can effectively resolve smoke plume injection heights (SPIH), is presented here. M6 is extensively validated against two alternative spaceborne earth observation SPIH data sources and demonstrates good agreement. Further, due to the spectral and dual-view configuration of the Advanced Along-Track Scanning Radiometer imaging system, it is possible to automatically differentiate smoke from other atmospheric features effectively-a feat, which currently no other algorithm can achieve. Additionally, as the M6 algorithm shares a heritage with the other M-series matchers, it is here compared against one of its predecessors, M4, which, for the determination of SPIH, M6 is shown to substantially outperform.

The response of a variety of Earth observation satellite sensors from multiple platforms, as well as different sensors on the same platform, were cross-compared over Dome C in Antarctica. A single unified methodology was employed to remove geometric, temporal, and atmospheric biases between different sensors with a range of spatial resolutions, bandwidths, and overpass times. The result assumes no “correct” baseline value, only relative responses. The resulting responses of the sensors, namely, AATSR, MERIS, and VEGETATION 2, are here presented and, where overlap occurs, compared with the results from other groups. In addition to applying the Quality Assurance for Earth Observation protocols for the quality assurance of environmental monitoring satellites, this paper discusses the intercalibration of several sensors used to establish the technical basis of the ESA GlobAlbedo project.

An automated registration method for the correction of AATSR co-registration is presented. AATSR co-registration between the nadir and forward views is known to be poor, generally greater than 1 pixel. Here we apply the SURF algorithm to automatically generate a set of tie-points between the forward and nadir images. These tie-points are then used to create a set of warping parameters which effectively co-register the forward and nadir images. The warping parameters have been tested globally for a number of years of AATSR data and are shown to provide improved co-registration to less than a pixel in all cases. The same method will, in the future, be applied to the other ATSR instruments which share similar co-registration problems.

Envisat's Medium Resolution Imaging Spectrometer (MERIS) acquired multi-spectral imagery of the Earth in the optical domain over terrestrial surfaces for a decade at global scale. For the last ten years, scientists have used multi-spectral data or terrestrial geophysical products for characterizing the state of the global system and its variability. Our paper shows highlights of several achievements of the use of MERIS data over terrestrial surfaces but specifically focuses on regional to global scale applications. We first summarize daily operational biophysical parameters and present examples of their uses for the monitoring of land surface states and changes, especially related to ECVs. In addition, specific projects for deriving a series of land cover maps will be presented and we conclude on the MERIS data exploitation and highlights future applications.

This paper describes the algorithm underlying the ESA DUE globAlbedo product. The purpose of the project is to produce a global 8-day land surface albedo product with associated uncertainty on a 1 km grid with continuous spatial coverage using data from European sensors. The product covers the period 1999-2011.

In this paper, the response of 7 Earth observation satellite sensors from multiple platforms, as well as different sensors on the same platform, are cross-compared over the Dome C plateau, Antarctica, during the austral summer of 2008-2009. A single, unified methodology is employed to remove geometric, temporal and atmospheric biases between different sensors with a range of spatial resolutions, bandwidths and overpass times. The result assumes no “correct” baseline value, only relative responses. The method is derived from an in-depth sensitivity analysis, presented elsewhere [1], [2], that takes into account: BRDF effect, cloud cover, macro-scale hoar-frost change and variations in atmospheric pressure, temperature and composition, including water vapour, ozone and aerosol content variation with altitude. The resulting responses of the sensors: AATSR, AVNIR-2, ETM+ (Landsat 7), MERIS, NigeriaSat-1, VEGETATION 2 and UK-DMC-1 are presented, and where overlap occurs, compared with the results from other groups [3], [4], [5]. In addition to introducing the QA4EO protocols for the quality assurance of environmental monitoring satellites, this paper also discusses several sensors used to establish the technical basis of the ESA GlobAlbedo project: an open access, validated, 15 year long record of broadband land surface albedo (both snow and snow-free) from European sensors alone [6].

The Multi-angle Imaging SpectroRadiometer (MISR) instrument has been collecting global Earth data from NASA's Terra satellite since February 2000. With its nine along-track view angles, four visible/near-infrared spectral bands, intrinsic spatial resolution of 275 m, and stable radiometric and geometric calibration, no instrument that combines MISR's attributes has previously flown in space. The more than 10-year (and counting) MISR data record provides unprecedented opportunities for characterizing long-term trends in aerosol, cloud, and surface properties, and includes 3-D textural information conventionally thought to be accessible only to active sensors.

It is generally agreed that the optimal footprint size for a spaceborne lidar is 30 m. Over topography such a large footprint can blur the canopy and ground signal together preventing information extraction. Multi-spectral lidar waveforms have been simulated with Monte-Carlo ray tracing over explicit geometric forest models. A method for using multi-spectral waveform lidar to distinguish ground from canopy returns has been tested over a range of ground slopes. The results are promising, with an initial error of +/-5 m for a signal level of only 5,000 photons with noise; an easily achievable figure. The inversion algorithms completely dominate inversion errors for all cases above 10,000 signal photons.

A representative albedo is required by ESA for the improved retrieval of atmospheric products, such as water vapour, from the ENVISAT-MERIS instrument. This albedo is required at 13 of the 15 spectral channels of MERIS. An algorithm is described which has been applied to the generation of a global albedo for these 13 spectral bands of MERIS over 16-day time periods at a resolution of 0.05deg for the time period from June 2002 to December 2006. Inputs to this algorithm include improved cloud and snow detection, aerosol correction using the MERIS data itself and derived Spectral surface Directional Reflectances, SDRs (Schroder et ah, Preusker et al, see, http://envisat.esa. int/workshops/meris_aatsr2005/). Owing to the lack of sufficient directional samples for most of the Earth's land surface, even for a monthly compositing time period, direct inversion of BRDF parameters for MERIS are not possible worldwide. Instead the 16-day Collection 4 MODIS BRDFs at 0.05deg resolution (derived using either full inversions or magnitude inversions from the MOD43 values) were employed in a magnitude inversion scheme for the 4 common MERIS (490plusmn5 [b3], 560plusmn5 [b5], 665plusmn5 [b7], 865plusmn10 [bl3]) bands with MODIS (459-479 {b3}, 545-565 {b4}, 620-670 {bl}, 841-8766 {b2}). The impact of applying Terra-only compared to Terra+Aqua-combined are here compared and presented. In keeping with previously reported results, there are larger numbers of full MODIS inversions (and consequently better MERIS albedos) for the combined BRDF but the total number of MODIS BRDF retrievals remains stubbornly similar. Spectral interpolation to the remaining 9 MERIS bands and to 3 broadband regions (0.4-0.7 mum, 0.7-3 mum, 0.4-3 mum) is then performed using polynomial look-up tables derived by Dr Shunlin Liang (University of Maryland). As spectral albedo data is required on monthly time-steps for most applications, a simple- minded weighting function based on the fractional time-period of each 16-day time period within a month was adopted for creating monthly products from 16-day products at 10km and 0.1deg. Global products on a 16-day time-step are being generated for the whole time period using both Terra-only and, where available, Terra+Aqua-combined Collection 4 MODlS-BRDFs. These EO products have been compared against contemporaneous MODIS spectral albedos (over 16-days) which have been gap-filled (Moody et al., 2005) as well as Collection 5 products, where available. Satellite intercomparisons have also been performed for monthly products of the monthly MERIS spectral albedo products for common bands against M1SR level-3 and POLDER level-3 products as well as for a subset of M1SR level-2 instantaneous products and the results will be shown here. Finally, inter-comparisons are shown of these EO products against ground-based albedometer measurements for a site in Finland will be presented as representative of one of several worldwide sites available through the CA VE taking into account, where necessary, issues associated with land cover heterogeneity.

The sky-view factor is a commonly used parameter in urban climatology; but to date, no completely satisfactory technique has yet been developed for its calculation. This letter investigates the potential of a prototype thermal fisheye camera, originally developed for use in the CLOUDMAP2 project, to produce fisheye imagery suitable for sky-view factor analysis. The camera distinguishes between the fraction of cold-space background and warm obstructions (e.g., buildings, trees, and cloud) from which the sky-view factor can be calculated. Early results are promising with excellent segmentation results obtained under clear skies, although the techniques are currently less reliable when clouds are visible in the field of view. However, this is demonstrated to be an advantage as the camera can also be used to simultaneously measure other parameters such as solar-tracks and "local" cloud fractions. The preliminary results indicate potential for the use of the camera for modeling studies in urban environments

This paper describes the application of a stereomatching algorithm to the problem of geo-rectifying a set of declassified military air photos over Cat Tien National Park, Vietnam. A disparity model, related to topographic features, is first built and then co-registered to a geo-referenced elevation model to provide the transformation required to correct the air-photos. Accuracy of results is assessed through correlation coefficients between retrieved disparities and known elevations.

Annual sequences of the first reprocessed albedo, bidirectional reflectance distribution function (BRDF), and nadir BRDF-adjusted surface reflectance (NBAR) products are being evaluated. BRDF model parameters (or weights) are used to compute black sky albedo at local solar noon and white sky albedo and to compute surface reflectance at a common nadir geometry. In addition to these standard resolution albedo, BRDF, and NBAR products, which are provided in the integerized sinusoidal grid projection, the BRDF parameters, black sky albedos (at local solar noon), and white sky albedos are now also being operationally produced in a global geographic projection known as the Climate Modeling Grid (CMG). These are currently available at a 0.25 degree spatial resolution, although there is community interest in a 0.05 degree resolution. In addition to the operational CMGs, coarser 0.5 degree and 1 degree resolution versions of the reprocessed albedo data have also been produced for the ISLSCP-II initiative. The global distribution of these albedos (as they vary throughout the year) are presented, as well as discussions of the most recent evaluations of the quality of the standard products.

Due to its unique nine-angle configuration, the Multi-angle Imaging SpectroRadiometer (MISR) can retrieve cloud parameters such as cloud-motion vectors and cloud-top heights using a purely geometrical technique that involves locating the same cloud features at different viewing angles. The geometrical nature of this technique means that the retrievals are relatively insensitive to the absolute instrument calibration. Fast stereo-matching algorithms have been developed to perform this image matching automatically on an operational basis. Preliminary results are shown of the operational retrievals together with comparisons against other data. Cloud-top height is generally obtained on a 1.1-km grid with an accuracy of /spl plusmn/ 562 m, even over snow and ice. The limitations of the technique, resulting at times in height blunders, noisy retrievals, and discrete effects of wind correction, are discussed.

The Multi-angle Imaging SpectroRadiometer (MISR) instrument, launched in December 1999 on the NASA EOS Terra satellite, produces images in the red band at 275-m resolution, over a swath width of 360 km, for the nine camera angles 70.5/spl deg/, 60/spl deg/, 45.6/spl deg/, and 26.1/spl deg/ forward, nadir, and 26.1/spl deg/, 45.6/spl deg/, 60/spl deg/, and 70.5/spl deg/ aft. A set of accurate and fast algorithms was developed for automated stereo matching of cloud features to obtain cloud-top height and motion over the nominal six-year lifetime of the mission. Accuracy and speed requirements necessitated the use of a combination of area-based and feature-based stereo-matchers with only pixel-level acuity. Feature-based techniques are used for cloud motion retrieval with the off-nadir MISR camera views, and the motion is then used to provide a correction to the disparities used to measure cloud-top heights which are derived from the innermost three cameras. Intercomparison with a previously developed "superstereo" matcher shows that the results are very comparable in accuracy with much greater coverage and at ten times the speed. Intercomparison of feature-based and area-based techniques shows that the feature-based techniques are comparable in accuracy at a factor of eight times the speed. An assessment of the accuracy of the area-based matcher for cloud-free scenes demonstrates the accuracy and completeness of the stereo-matcher. This trade-off has resulted in the loss of a reliable quality metric to predict accuracy and a slightly high blunder rate. Examples are shown of the application of the MISR stereo-matchers on several difficult scenes which demonstrate the efficacy of the matching approach.

The Multi-angle Imaging SpectroRadiometer (MISR) instrument is scheduled for launch aboard the first of the Earth Observing System (EOS) spacecraft, EOS-AM1. MISR will provide global, radiometrically calibrated, georectified, and spatially coregistered imagery at nine discrete viewing angles and four visible/near-infrared spectral bands. Algorithms specifically developed to capitalize on this measurement strategy will be used to retrieve geophysical products for studies of clouds, aerosols, and surface radiation. This paper provides an overview of the as-built instrument characteristics and the application of MISR to remote sensing of the Earth.

The first Moderate Resolution Imaging Spectroradiometer (MODIS) instrument is planned for launch by NASA in 1998. This instrument will provide a new and improved capability for terrestrial satellite remote sensing aimed at meeting the needs of global change research. The MODIS standard products will provide new and improved tools for moderate resolution land surface monitoring. These higher order data products have been designed to remove the burden of certain common types of data processing from the user community and meet the more general needs of global-to-regional monitoring, modeling, and assessment. The near-daily coverage of moderate resolution data from MODIS, coupled with the planned increase in high-resolution sampling from Landsat 7, will provide a powerful combination of observations. The full potential of MODIS will be realized once a stable and well-calibrated time-series of multispectral data has been established. In this paper the proposed MODIS standard products for land applications are described along with the current plans for data quality assessment and product validation.

An accurate geometric correction algorithm of raw ATSR images is proposed in this paper. The proposed algorithm is divided into two modules: pre-correction and precision correction. In the pre-correction algorithm, the scanning geometry of the ATSR, ephemeris data and the Earth's shape and rotation were modeled mathematically. It corrected elliptically-distorted raw ATSR images and resulted in a mean collocation error of 1 pixel between the nadir and forward images. The proposed navigation model showed a better geolocation performance than ATSR image geometric correction algorithms currently in operation. The geometric errors in the pre-corrected images were reduced further by the precision correction algorithm. Coastlines were detected from the pre-corrected nadir and forward images and matched with map coastline data. The matched coastline pixels, GCPs, were combined with the pre-correction model to estimate the errors of the position, velocity and attitude of the satellite. An extended Kalman filter was used for the optimal estimation. The corrected images showed the collocation error of 0.4 pixel and the geolocation error of 0.7 pixel. The results were compared with those from a polynomial warping technique. It proved that the proposed precision correction algorithm gave better geometric correction performance than collocation algorithms based on a polynomial warping technique.