Search

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 generalised deep learning model. We focus on the problem of cloud masking, using several pre-existing datasets, and a new, freely available dataset for Sentinel-2. Our model is shown to achieve state-of-the-art performance on the satellites it was trained on (Sentinel-2 and Landsat 8), and is able to extrapolate to sensors it has not seen during training such as Landsat 7, Per\'uSat-1, and Sentinel-3 SLSTR. Model performance is shown to improve when multiple satellites are used in training, approaching or surpassing the performance of specialised, single-sensor models. This work is motivated by the fact that the remote sensing community has access to data taken with a hugely variety of sensors. This has inevitably led to labelling 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 utilise 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 on-board applications and in ground segment data processing, which generally require models to be ready at launch or soon afterwards., Comment: 22 pages, 7 figures. This is an accepted version of work to be published in the IEEE Transactions on Geoscience and Remote Sensing

Ares is an extension of the TauREx 3 retrieval framework for the Martian atmosphere. Ares is a collection of new atmospheric parameters and forward models, designed for the European Space Agency's (ESA) Trace Gas Orbiter (TGO) Nadir and Occultation for MArs Discovery (NOMAD) instrument, Solar Occultation (SO) channel. Ares provides unique insights into the chemical composition of the Martian atmosphere by applying methods utilised in exoplanetary atmospheric retrievals, Waldmann et al. (2015), Al-Refaie et al. (2019). This insight may help unravel the true nature of $CH_{4}$ on Mars. The Ares model is here described. Subsequently, the components of Ares are defined, including; the NOMAD SO channel instrument function model, Martian atmospheric molecular absorption cross-sections, geometry models, and a NOMAD noise model. Ares atmospheric priors and forward models are presented, (i.e., simulated NOMAD observations), and are analysed, compared and validated against the Planetary Spectrum Generator, Villanueva et al. (2018).

CAR (Cloud Absorption Radiometer) is a multi-angular and multi-spectral airborne radiometer instrument, whose radiometric and geometric characteristics are well calibrated and adjusted before and after each flight campaign. CAR was built by NASA (National Aeronautics and Space Administration) in 1984. On 16 May 2008, a CAR flight campaign took place over the well-known calibration and validation site of Railroad Valley in Nevada (38.504 deg N, 115.692 deg W).The campaign coincided with the overpasses of several key EO (Earth Observation) satellites such as Landsat-7, Envisat and Terra. Thus, there are nearly simultaneous measurements from these satellites and the CAR airborne sensor over the same calibration site. The CAR spectral bands are close to those of most EO satellites. CAR has the ability to cover the whole range of azimuth view angles and a variety of zenith angles depending on altitude and, as a consequence, the biases seen between satellite and CAR measurements due to both unmatched spectral bands and unmatched angles can be significantly reduced. A comparison is presented here between CARs land surface reflectance (BRF or Bidirectional Reflectance Factor) with those derived from Terra/MODIS (MOD09 and MAIAC), Terra/MISR, Envisat/MERIS and Landsat-7. In this study, we utilized CAR data from low altitude flights (approx. 180 m above the surface) in order to minimize the effects of the atmosphere on these measurements and then obtain a valuable ground-truth data set of surface reflectance. Furthermore, this study shows that differences between measurements caused by surface heterogeneity can be tolerated, thanks to the high homogeneity of the study site on the one hand, and on the other hand, to the spatial sampling and the large number of CAR samples. These results demonstrate that satellite BRF measurements over this site are in good agreement with CAR with variable biases across different spectral bands. This is most likely due to residual aerosol effects in the EO derived reflectances.