The Satellite for Orbital Aerodynamics Research (SOAR) is a CubeSat mission, due to be launched in 2021, to investigate the interaction between different materials and the atmospheric flow regime in very low Earth orbits (VLEO). Improving knowledge of the gas-surface interactions at these altitudes and identification of novel materials that can minimise drag or improve aerodynamic control are important for the design of future spacecraft that can operate in lower altitude orbits. Such satellites may be smaller and cheaper to develop or can provide improved Earth observation data or communications link-budgets and latency. Using precise orbit and attitude determination information and the measured atmospheric flow characteristics the forces and torques experienced by the satellite in orbit can be studied and estimates of the aerodynamic coefficients calculated. This paper presents the scientific concept and design of the SOAR mission. The methodology for recovery of the aerodynamic coefficients from the measured orbit, attitude, and in-situ atmospheric data using a least-squares orbit determination and free-parameter fitting process is described and the experimental uncertainty of the resolved aerodynamic coefficients is estimated. The presented results indicate that the combination of the satellite design and experimental methodology are capable of clearly illustrating the variation of drag and lift coefficient for differing surface incidence angle. The lowest uncertainties for the drag coefficient measurement are found at approximately 300 km, whilst the measurement of lift coefficient improves for reducing orbital altitude to 200 km.

Soil moisture (SM) is a key state variable of the hydrological cycle, needed to monitor the effects of a changing climate on natural resources. Soil moisture is highly variable in space and time, presenting seasonalities, anomalies and long-term trends, but also, and important nonlinear behaviours. Here, we introduce a novel fast and nonlinear complex PCA method to analyze the spatio-temporal patterns of the Earth's surface SM. We use global SM estimates acquired during the period 2010-2017 by ESA's SMOS mission. Our approach unveils both time and space modes, trends and periodicities unlike standard PCA decompositions. Results show the distribution of the total SM variance among its different components, and indicate the dominant modes of temporal variability in surface soil moisture for different regions. The relationship of the derived SM spatio-temporal patterns with El Ni{\~n}o Southern Oscillation (ENSO) conditions is also explored.

Land-cover classification using remote sensing imagery is an important Earth observation task. Recently, land cover classification has benefited from the development of fully connected neural networks for semantic segmentation. The benchmark datasets available for training deep segmentation models in remote sensing imagery tend to be small, however, often consisting of only a handful of images from a single location with a single scale. This limits the models' ability to generalize to other datasets. Domain adaptation has been proposed to improve the models' generalization but we find these approaches are not effective for dealing with the scale variation commonly found between remote sensing image collections. We therefore propose a scale aware adversarial learning framework to perform joint cross-location and cross-scale land-cover classification. The framework has a dual discriminator architecture with a standard feature discriminator as well as a novel scale discriminator. We also introduce a scale attention module which produces scale-enhanced features. Experimental results show that the proposed framework outperforms state-of-the-art domain adaptation methods by a large margin.

Dealing with land cover classification of the new image sources has also turned to be a complex problem requiring large amount of memory and processing time. In order to cope with these problems, statistical learning has greatly helped in the last years to develop statistical retrieval and classification models that can ingest large amounts of Earth observation data. Kernel methods constitute a family of powerful machine learning algorithms, which have found wide use in remote sensing and geosciences. However, kernel methods are still not widely adopted because of the high computational cost when dealing with large scale problems, such as the inversion of radiative transfer models or the classification of high spatial-spectral-temporal resolution data. This paper introduces an efficient kernel method for fast statistical retrieval of bio-geo-physical parameters and image classification problems. The method allows to approximate a kernel matrix with a set of projections on random bases sampled from the Fourier domain. The method is simple, computationally very efficient in both memory and processing costs, and easily parallelizable. We show that kernel regression and classification is now possible for datasets with millions of examples and high dimensionality. Examples on atmospheric parameter retrieval from hyperspectral infrared sounders like IASI/Metop; large scale emulation and inversion of the familiar PROSAIL radiative transfer model on Sentinel-2 data; and the identification of clouds over landmarks in time series of MSG/Seviri images show the efficiency and effectiveness of the proposed technique.

Over land the vegetation canopy affects the microwave brightness temperature by emission, scattering and attenuation of surface soil emission. The questions addressed in this study are: 1) what is the transparency of the vegetation canopy for different biomes around the Globe at the low-frequency L-band?, 2) what is the seasonal amplitude of vegetation microwave optical depth for different biomes?, 3) what is the effective scattering at this frequency for different vegetation types?, 4) what is the impact of imprecise characterization of vegetation microwave properties on retrieval of soil surface conditions? These questions are addressed based on the recently completed one full annual cycle measurements by the NASA Soil Moisture Active Passive (SMAP) measurements.

An application of the Soil Moisture Agricultural Drought Index (SMADI) at the global scale is presented. The index integrates surface soil moisture from the SMOS mission with land surface temperature (LST) and Normalized Difference Vegetation Index (NDVI) from MODIS and allows for global drought monitoring at medium spatial scales (0.05 deg).. Biweekly maps of SMADI were obtained from year 2010 to 2015 over all agricultural areas on Earth. The SMADI time-series were compared with state-of-the-art drought indices over the Iberian Peninsula. Results show a good agreement between SMADI and the Crop Moisture Index (CMI) retrieved at five weather stations (with correlation coefficient, R from -0.64 to -0.79) and the Soil Water Deficit Index (SWDI) at the Soil Moisture Measurement Stations Network of the University of Salamanca (REMEDHUS) (R=-0.83). Some preliminary tests were also made over the continental United States using the Vegetation Drought Response Index (VegDRI), with very encouraging results regarding the spatial occurrence of droughts during summer seasons. Additionally, SMADI allowed to identify distinctive patterns of regional drought over the Indian Peninsula in spring of 2012. Overall results support the use of SMADI for monitoring agricultural drought events world-wide.

The GRACE and GRACE-FO missions have provided an unprecedented quantification of large-scale changes in the water cycle. However, it is still an open problem of how these missions' data sets can be referenced to a ground truth. Meanwhile, stationary optical clocks show fractional instabilities below $10^{-18}$ when averaged over an hour, and continue to be improved in terms of precision and accuracy, uptime, and transportability. The frequency of a clock is affected by the gravitational redshift, and thus depends on the local geopotential; a relative frequency change of $10^{-18}$ corresponds to a geoid height change of about $1$ cm. Here we suggest that this effect could be further exploited for sensing large-scale temporal geopotential changes via a network of clocks distributed at the Earth's surface. In fact, several projects have already proposed to create an ensemble of optical clocks connected across Europe via optical fibre links. Our hypothesis is that a clock network with collocated GNSS receivers spread over Europe - for which the physical infrastructure is already partly in place - would enable us to determine temporal variations of the Earth's gravity field at time scales of days and beyond, and thus provide a new means for validating satellite missions such as GRACE-FO or a future gravity mission. Here, we show through simulations how ice, hydrology and atmosphere variations over Europe could be observed with clock comparisons in a future network that follows current design concepts in the metrology community. We assume different scenarios for clock and GNSS uncertainties and find that even under conservative assumptions - a clock error of $10^{-18}$ and vertical height control error of $1.4$ mm for daily measurements - hydrological signals at the annual time scale and atmospheric signals down to the weekly time scale could be observed.

In recent years, BERT has made significant breakthroughs on many natural language processing tasks and attracted great attentions. Despite its accuracy gains, the BERT model generally involves a huge number of parameters and needs to be trained on massive datasets, so training such a model is computationally very challenging and time-consuming. Hence, training efficiency should be a critical issue. In this paper, we propose a novel coarse-refined training framework named CoRe to speed up the training of BERT. Specifically, we decompose the training process of BERT into two phases. In the first phase, by introducing fast attention mechanism and decomposing the large parameters in the feed-forward network sub-layer, we construct a relaxed BERT model which has much less parameters and much lower model complexity than the original BERT, so the relaxed model can be quickly trained. In the second phase, we transform the trained relaxed BERT model into the original BERT and further retrain the model. Thanks to the desired initialization provided by the relaxed model, the retraining phase requires much less training steps, compared with training an original BERT model from scratch with a random initialization. Experimental results show that the proposed CoRe framework can greatly reduce the training time without reducing the performance.

Time series data is prevalent in a wide variety of real-world applications and it calls for trustworthy and explainable models for people to understand and fully trust decisions made by AI solutions. We consider the problem of building explainable classifiers from multi-variate time series data. A key criterion to understand such predictive models involves elucidating and quantifying the contribution of time varying input variables to the classification. Hence, we introduce a novel, modular, convolution-based feature extraction and attention mechanism that simultaneously identifies the variables as well as time intervals which determine the classifier output. We present results of extensive experiments with several benchmark data sets that show that the proposed method outperforms the state-of-the-art baseline methods on multi-variate time series classification task. The results of our case studies demonstrate that the variables and time intervals identified by the proposed method make sense relative to available domain knowledge.

Polar mesocyclones (PMCs) and their intense subclass polar lows (PLs) are relatively small atmospheric vortices that form mostly over the ocean in high latitudes. PLs can strongly influence deep ocean water formation since they are associated with strong surface winds and heat fluxes. Detection and tracking of PLs are crucial for understanding the climatological dynamics of PLs and for the analysis of their impacts on other components of the climatic system. At the same time, visual tracking of PLs is a highly time-consuming procedure that requires expert knowledge and extensive examination of source data. There are known procedures involving deep convolutional neural networks (DCNNs) for the detection of large-scale atmospheric phenomena in reanalysis data that demonstrate a high quality of detection. However, one cannot apply these procedures to satellite data directly since, unlike reanalyses, satellite products register all the scales of atmospheric vortices. It is also known that DCNNs were originally designed to be scale-invariant. This leads to the problem of filtering the scale of detected phenomena. There are other problems to be solved, such as a low signal-to-noise ratio of satellite data and an unbalanced number of negative (without PLs) and positive (where a PL is presented) classes in a satellite dataset. In our study, we propose a deep learning approach for the detection of PLs and PMCs in remote sensing data, which addresses class imbalance and scale filtering problems. We also outline potential solutions for other problems, along with promising improvements to the presented approach.

Explainability and interpretability are two important concepts, the absence of which can and should impede the application of well-performing neural networks to real-world problems. At the same time, they are difficult to incorporate into the large, black-box models that achieve state-of-the-art results in a multitude of NLP tasks. Bidirectional Encoder Representations from Transformers (BERT) is one such black-box model. It has become a staple architecture to solve many different NLP tasks and has inspired a number of related Transformer models. Understanding how these models draw conclusions is crucial for both their improvement and application. We contribute to this challenge by presenting VisBERT, a tool for visualizing the contextual token representations within BERT for the task of (multi-hop) Question Answering. Instead of analyzing attention weights, we focus on the hidden states resulting from each encoder block within the BERT model. This way we can observe how the semantic representations are transformed throughout the layers of the model. VisBERT enables users to get insights about the model's internal state and to explore its inference steps or potential shortcomings. The tool allows us to identify distinct phases in BERT's transformations that are similar to a traditional NLP pipeline and offer insights during failed predictions.

BERT-enhanced neural machine translation (NMT) aims at leveraging BERT-encoded representations for translation tasks. A recently proposed approach uses attention mechanisms to fuse Transformer's encoder and decoder layers with BERT's last-layer representation and shows enhanced performance. However, their method doesn't allow for the flexible distribution of attention between the BERT representation and the encoder/decoder representation. In this work, we propose a novel BERT-enhanced NMT model called BERT-JAM which improves upon existing models from two aspects: 1) BERT-JAM uses joint-attention modules to allow the encoder/decoder layers to dynamically allocate attention between different representations, and 2) BERT-JAM allows the encoder/decoder layers to make use of BERT's intermediate representations by composing them using a gated linear unit (GLU). We train BERT-JAM with a novel three-phase optimization strategy that progressively unfreezes different components of BERT-JAM. Our experiments show that BERT-JAM achieves SOTA BLEU scores on multiple translation tasks.

Recent research on the multi-head attention mechanism, especially that in pre-trained models such as BERT, has shown us heuristics and clues in analyzing various aspects of the mechanism. As most of the research focus on probing tasks or hidden states, previous works have found some primitive patterns of attention head behavior by heuristic analytical methods, but a more systematic analysis specific on the attention patterns still remains primitive. In this work, we clearly cluster the attention heatmaps into significantly different patterns through unsupervised clustering on top of a set of proposed features, which corroborates with previous observations. We further study their corresponding functions through analytical study. In addition, our proposed features can be used to explain and calibrate different attention heads in Transformer models.

The assessment of microbial species biodiversity is essential in ecology and evolutionary biology (Reaka-Kudla et al. 1996), but especially challenging for communities of microorganisms found in the environment (Das et al. 2006, Hillebrand et al. 2018). Beyond providing a census of organisms in the ocean, assessing marine microbial biodiversity can reveal how microbes respond to environmental change (Salazar et al. 2017), clarify the ecological roles of community members (Hehemann et al. 2016), and lead to biotechnology discoveries (Das et al. 2006). Computational approaches to characterize taxonomic diversity and phylogeny based on the quality of available data for environmental sequence datasets is fundamental for advancing our understanding of the role of these organisms in the environment. Even more pressing is the need for comprehensive and consistent methods to assign taxonomy to environmentally-relevant microbial eukaryotes. Here, we present EUKulele, an open-source software tool designed to assign taxonomy to microeukaryotes detected in meta-omic samples, and complement analysis approaches in other domains by accommodating assembly output and providing concrete metrics reporting the taxonomic completeness of each sample.

Existing object detection frameworks are usually built on a single format of object/part representation, i.e., anchor/proposal rectangle boxes in RetinaNet and Faster R-CNN, center points in FCOS and RepPoints, and corner points in CornerNet. While these different representations usually drive the frameworks to perform well in different aspects, e.g., better classification or finer localization, it is in general difficult to combine these representations in a single framework to make good use of each strength, due to the heterogeneous or non-grid feature extraction by different representations. This paper presents an attention-based decoder module similar as that in Transformer~\cite{vaswani2017attention} to bridge other representations into a typical object detector built on a single representation format, in an end-to-end fashion. The other representations act as a set of \emph{key} instances to strengthen the main \emph{query} representation features in the vanilla detectors. Novel techniques are proposed towards efficient computation of the decoder module, including a \emph{key sampling} approach and a \emph{shared location embedding} approach. The proposed module is named \emph{bridging visual representations} (BVR). It can perform in-place and we demonstrate its broad effectiveness in bridging other representations into prevalent object detection frameworks, including RetinaNet, Faster R-CNN, FCOS and ATSS, where about $1.5\sim3.0$ AP improvements are achieved. In particular, we improve a state-of-the-art framework with a strong backbone by about $2.0$ AP, reaching $52.7$ AP on COCO test-dev. The resulting network is named RelationNet++. The code will be available at https://github.com/microsoft/RelationNet2.

A class of recent semi-supervised learning (SSL) methods heavily rely on domain-specific data augmentations. In contrast, generative SSL methods involve unsupervised learning based on generative models by either joint-training or pre-training, and are more appealing from the perspective of being domain-agnostic, since they do not inherently require data augmentations. Joint-training estimates the joint distribution of observations and labels, while pre-training is taken over observations only. Recently, energy-based models (EBMs) have achieved promising results for generative modeling. Joint-training via EBMs for SSL has been explored with encouraging results across different data modalities. In this paper, we make two contributions. First, we explore pre-training via EBMs for SSL and compare it to joint-training. Second, a suite of experiments are conducted over domains of image classification and natural language labeling to give a realistic whole picture of the performances of EBM based SSL methods. It is found that joint-training EBMs outperform pre-training EBMs marginally but nearly consistently.

The world-wide lockdown in response to the COVID-19 pandemic in year 2020 led to economic slowdown and large reduction of fossil fuel CO2 emissions, but it is unclear how much it would reduce atmospheric CO2 concentration, and whether it can be observed. We estimated that a 7.9% reduction in emissions for 4 months would result in a 0.25 ppm decrease in the Northern Hemisphere CO2, an increment that is within the capability of current CO2 analyzers, but is a few times smaller than natural CO2 variabilities caused by weather and the biosphere such as El Nino. We used a state-of-the-art atmospheric transport model to simulate CO2, driven by a new daily fossil fuel emissions dataset and hourly biospheric fluxes from a carbon cycle model forced with observed climate variability. Our results show a 0.13 ppm decrease in atmospheric column CO2 anomaly averaged over 50S-50N for the period February-April 2020 relative to a 10-year climatology. A similar decrease was observed by the carbon satellite GOSAT3. Using model sensitivity experiments, we further found that COVID, the biosphere and weather contributed 54%, 23%, and 23% respectively. This seemingly small change stands out as the largest sub-annual anomaly in the last 10 years. Measurements from global ground stations were analyzed. At city scale, on-road CO2 enhancement measured in Beijing shows reduction of 20-30 ppm, consistent with drastically reduced traffic during the lockdown. The ability of our current carbon monitoring systems in detecting the small and short-lasting COVID signal on the background of fossil fuel CO2 accumulated over the last two centuries is encouraging. The COVID-19 pandemic is an unintended experiment whose impact suggests that to keep atmospheric CO2 at a climate-safe level will require sustained effort of similar magnitude and improved accuracy and expanded spatiotemporal coverage of our monitoring systems.

While the Transformer architecture has become the de-facto standard for natural language processing tasks, its applications to computer vision remain limited. In vision, attention is either applied in conjunction with convolutional networks, or used to replace certain components of convolutional networks while keeping their overall structure in place. We show that this reliance on CNNs is not necessary and a pure transformer applied directly to sequences of image patches can perform very well on image classification tasks. When pre-trained on large amounts of data and transferred to multiple mid-sized or small image recognition benchmarks (ImageNet, CIFAR-100, VTAB, etc.), Vision Transformer (ViT) attains excellent results compared to state-of-the-art convolutional networks while requiring substantially fewer computational resources to train.

We present the CHAOS-7 model of the time-dependent near-Earth geomagnetic field between 1999 and 2020 based on magnetic field observations collected by the low-Earth orbit satellites {\it Swarm}, CryoSat-2, CHAMP, SAC-C and {\O}rsted, and on annual differences of monthly means of ground observatory measurements. The CHAOS-7 model consists of a time-dependent internal field up to spherical harmonic degree 20, a static internal field which merges to the LCS-1 lithospheric field model above degree 25, a model of the magnetospheric field and its induced counterpart, estimates of Euler angles describing the alignment of satellite vector magnetometers, and magnetometer calibration parameters for CryoSat-2. Only data from dark regions satisfying strict geomagnetic quiet-time criteria (including conditions on IMF $B_z$ and $B_y$ at all latitudes) were used in the field estimation. Model parameters were estimated using an iteratively-reweighted regularized least-squares procedure; regularization of the time-dependent internal field was relaxed at high spherical harmonic degree compared with previous versions of the CHAOS model. We use CHAOS-7 to investigate recent changes in the geomagnetic field, studying the evolution of the South Atlantic weak field anomaly and rapid field changes in the Pacific region since 2014. At Earth's surface a secondary minimum of the South Atlantic Anomaly is now evident to the south west of Africa. Green's functions relating the core-mantle boundary radial field to the surface intensity show this feature is connected with the movement and evolution of a reversed flux feature under South Africa. The continuing growth in size and weakening of the main anomaly is linked to the westward motion and gathering of reversed flux under South America.

Transformer has achieved great success in the NLP field by composing various advanced models like BERT and GPT. However, Transformer and its existing variants may not be optimal in capturing token distances because the position or distance embeddings used by these methods usually cannot keep the precise information of real distances, which may not be beneficial for modeling the orders and relations of contexts. In this paper, we propose DA-Transformer, which is a distance-aware Transformer that can exploit the real distance. We propose to incorporate the real distances between tokens to re-scale the raw self-attention weights, which are computed by the relevance between attention query and key. Concretely, in different self-attention heads the relative distance between each pair of tokens is weighted by different learnable parameters, which control the different preferences on long- or short-term information of these heads. Since the raw weighted real distances may not be optimal for adjusting self-attention weights, we propose a learnable sigmoid function to map them into re-scaled coefficients that have proper ranges. We first clip the raw self-attention weights via the ReLU function to keep non-negativity and introduce sparsity, and then multiply them with the re-scaled coefficients to encode real distance information into self-attention. Extensive experiments on five benchmark datasets show that DA-Transformer can effectively improve the performance of many tasks and outperform the vanilla Transformer and its several variants.

The recent advances in artificial intelligence namely in machine learning and deep learning, have boosted the performance of intelligent systems in several ways. This gave rise to human expectations, but also created the need for a deeper understanding of how intelligent systems think and decide. The concept of explainability appeared, in the extent of explaining the internal system mechanics in human terms. Recommendation systems are intelligent systems that support human decision making, and as such, they have to be explainable in order to increase user trust and improve the acceptance of recommendations. In this work, we focus on a context-aware recommendation system for energy efficiency and develop a mechanism for explainable and persuasive recommendations, which are personalized to user preferences and habits. The persuasive facts either emphasize on the economical saving prospects (Econ) or on a positive ecological impact (Eco) and explanations provide the reason for recommending an energy saving action. Based on a study conducted using a Telegram bot, different scenarios have been validated with actual data and human feedback. Current results show a total increase of 19\% on the recommendation acceptance ratio when both economical and ecological persuasive facts are employed. This revolutionary approach on recommendation systems, demonstrates how intelligent recommendations can effectively encourage energy saving behavior.

Relying on large pretrained language models such as Bidirectional Encoder Representations from Transformers (BERT) for encoding and adding a simple prediction layer has led to impressive performance in many clinical natural language processing (NLP) tasks. In this work, we present a novel extension to the Transformer architecture, by incorporating signature transform with the self-attention model. This architecture is added between embedding and prediction layers. Experiments on a new Swedish prescription data show the proposed architecture to be superior in two of the three information extraction tasks, comparing to baseline models. Finally, we evaluate two different embedding approaches between applying Multilingual BERT and translating the Swedish text to English then encode with a BERT model pretrained on clinical notes.

A homology cylinder of a surface induces an automorphism of the completed group ring of the fundamental group of the surface. We introduce a new method of computing the automorphism by using the Goldman Lie algebra of the surface or some skein algebra. In particular, we give a refinement of a formula by Kuno and Massuyeau.

The damage characteristics of a shallow buried tunnel under multiple explosive loads is an important research issue in the design and evaluation of protective engineering. It is of great significance to develop a method for early warning of the safety of the shallow buried features. The discrete element method is used to establish a mechanical model of the shallow buried tunnel. The South Load Equivalent Principle treats blast loads as a series of dynamic forces acting uniformly on the surface. Based on the discrete element method, the dynamic response after each blast load and the damage evolution process of the surrounding rock of the tunnel are obtained. The strength reduction method is used to obtain the surrounding rock of the tunnel. Introduce the theory of continuous homology, and use the mathematical method of continuous homology to quantitatively and qualitatively analyze the failure characteristics of the discrete element model under multiple explosive loads. The results show that the method of continuous homology can accurately reflect the topological characteristics of the surrounding rock of the tunnel The maximum one-dimensional bar code connection radius can effectively warn tunnel instability. This provides a new mathematical method for tunnel safety design and disaster prediction research.

Optical remote sensing imagery has been widely used in many fields due to its high resolution and stable geometric properties. However, remote sensing imagery is inevitably affected by climate, especially clouds. Removing the cloud in the high-resolution remote sensing satellite image is an indispensable pre-processing step before analyzing it. For the sake of large-scale training data, neural networks have been successful in many image processing tasks, but the use of neural networks to remove cloud in remote sensing imagery is still relatively small. We adopt generative adversarial network to solve this task and introduce the spatial attention mechanism into the remote sensing imagery cloud removal task, proposes a model named spatial attention generative adversarial network (SpA GAN), which imitates the human visual mechanism, and recognizes and focuses the cloud area with local-to-global spatial attention, thereby enhancing the information recovery of these areas and generating cloudless images with better quality...

Investigating molecular heterogeneity provides insights about tumor origin and metabolomics. The increasing amount of data gathered makes manual analyses infeasible - therefore, automated unsupervised learning approaches are utilized for discovering heterogeneity. However, automated unsupervised analyses require a lot of experience with setting their hyperparameters and usually an upfront knowledge about the number of expected substructures. Moreover, numerous measured molecules require an additional step of feature engineering to provide valuable results. In this work, we propose DiviK: a scalable stepwise algorithm with local data-driven feature space adaptation for the segmentation of high-dimensional datasets. The combination of three quality indices: Dice Index, Rand Index and EXIMS score are used to assess the quality of unsupervised analyses in 3D space. DiviK was validated on two separate high-throughput datasets acquired by Mass Spectrometry Imaging in 2D and 3D. DiviK could be one of the default choices to consider during the initial exploration of Mass Spectrometry Imaging data. It provides a trade-off between absolute heterogeneity detection and focus on biologically plausible structures, and does not require specifying the number of expected structures before the analysis. With its unique local feature space adaptation, it is robust against dominating global patterns when focusing on the detail. Finally, due to its simplicity, DiviK is easily generalizable to an even more flexible framework, useful for other '-omics' data, or tabular data in general (including medical images after appropriate embedding). A generic implementation is freely available under Apache 2.0 license at https://github.com/gmrukwa/divik.

Recent work in the field of cryo-seismology demonstrates that high frequency (>1 Hz) waves provide key constraints on a wide range of glacier processes such as basal friction, surface crevassing or subglacial water flow. Establishing quantitative links between the seismic signal and the processes of interest however requires detailed characterization of the wavefield, which at the high frequencies of interest necessitates the deployment of large and particularly dense seismic arrays. Although dense seismic array monitoring has recently become routine in geophysics, its application to glaciated environments has yet remained limited. Here we present a dense seismic array experiment made of 98 3-component seismic stations continuously recording during 35 days in early spring on the Argenti\`ere Glacier, French Alps. The seismic dataset is supplemented by a wide range of complementary observations obtained from ground penetrating radar, drone imagery, GNSS positioning and in-situ instrumentation of basal glacier sliding velocities and subglacial water discharge. Through applying multiple processing techniques including event detection from template matching and systematic beamforming, we demonstrate that the present dataset provides enhanced spatial resolution on basal stick slip and englacial fracturing sources as well as novel constraints on heterogeneous nature of the noise field generated by subglacial water flow and on the link between crevasse properties and englacial seismic velocities. We finally outline in which ways further work using this dataset could help tackle key remaining questions in the field.

We study the response of the Lagrangian sea ice model neXtSIM to the uncertainty in the sea surface wind and sea ice cohesion. The ice mechanics in neXtSIM is based on a brittle-like rheological framework. The study considers short-term ensemble forecasts of the Arctic sea ice from January to April 2008. Ensembles are generated by perturbing the wind inputs and ice cohesion field both separately and jointly. The resulting uncertainty in the probabilistic forecasts is evaluated statistically based on the analysis of Lagrangian sea ice trajectories as sampled by virtual drifters seeded in the model to cover the Arctic Ocean and using metrics borrowed from the search-and-rescue literature. The comparison among the different ensembles indicates that wind perturbations dominate the forecast uncertainty i.e. the absolute spread of the ensemble, while the inhomogeneities in the ice cohesion field significantly increase the degree of anisotropy in the spread i.e. trajectories drift differently in different directions. We suggest that in order to get a full flavor of uncertainties in a sea ice model with brittle-like rheologies, to predict sea ice drift and trajectories, one should consider using ensemble-based simulations where both wind forcing and sea ice cohesion are perturbed.

GATR (Globally-scalable Automated Target Recognition) is a Lockheed Martin software system for real-time object detection and classification in satellite imagery on a worldwide basis. GATR uses GPU-accelerated deep learning software to quickly search large geographic regions. On a single GPU it processes imagery at a rate of over 16 square km/sec (or more than 10 Mpixels/sec), and it requires only two hours to search the entire state of Pennsylvania for gas fracking wells. The search time scales linearly with the geographic area, and the processing rate scales linearly with the number of GPUs. GATR has a modular, cloud-based architecture that uses the Maxar GBDX platform and provides an ATR analytic as a service. Applications include broad area search, watch boxes for monitoring ports and airfields, and site characterization. ATR is performed by deep learning models including RetinaNet and Faster R-CNN. Results are presented for the detection of aircraft and fracking wells and show that the recalls exceed 90% even in geographic regions never seen before. GATR is extensible to new targets, such as cars and ships, and it also handles radar and infrared imagery.

Precipitation and dissolution are prime processes in carbonate rocks and being able to monitor them is of major importance for aquifer and reservoir exploitation or environmental studies. Electrical conductivity is a physical property sensitive both to transport phenomena of porous media and to dissolution and precipitation processes. However, its quantitative use depends on the effectiveness of the petrophysical relationship to relate the electrical conductivity to hydrological properties of interest. In this work, we develop a new physically-based model to estimate the electrical conductivity by upscaling a microstructural description of water-saturated fractal porous media. This model is successfully compared to published data from both unconsolidated and consolidated samples, or during precipitation and dissolution numerical experiments. For the latter, we show that the permeability can be linked to the predicted electrical conductivity.

The ROMS modeling system was applied to the California Upwelling System (CalUS) to understand the key hydrodynamic conditions and dynamics of the nitrogen-based ecosystem using the NPZD model proposed by Powell et al. (2006). A new type of sponge layer has been successfully implemented in the ROMS modelling system in order to stabilize the hydrodynamic part of the modeling system when using so-called reduced boundary conditions. The hydrodynamic performance of the model was examined using a tidal analysis based on tidal measurement data, a comparison of the modeled sea surface temperature (SST) with buoy and satellite data, and vertical sections of the currents along the coast and the water temperature. This validation process shows that the hydrodynamic module used in this study can reproduce the basic hydrodynamic and circulation characteristics within the CalUS. The results of the ecosystem model show the characteristic features of upwelling regions as well as the well-known spotty horizontal structures of the zooplankton community. The model thus provides a solid basis for the hydrodynamic and ecological characteristics of the CalUS and enables the ecological model to be expanded into a complex ecological model for investigating the effects of climate change on the ecological balance in the area investigated.

We investigate active learning in the context of deep neural network models for change detection and map updating. Active learning is a natural choice for a number of remote sensing tasks, including the detection of local surface changes: changes are on the one hand rare and on the other hand their appearance is varied and diffuse, making it hard to collect a representative training set in advance. In the active learning setting, one starts from a minimal set of training examples and progressively chooses informative samples that are annotated by a user and added to the training set. Hence, a core component of an active learning system is a mechanism to estimate model uncertainty, which is then used to pick uncertain, informative samples. We study different mechanisms to capture and quantify this uncertainty when working with deep networks, based on the variance or entropy across explicit or implicit model ensembles. We show that active learning successfully finds highly informative samples and automatically balances the training distribution, and reaches the same performance as a model supervised with a large, pre-annotated training set, with $\approx$99% fewer annotated samples.

Co-saliency detection aims to detect common salient objects from a group of relevant images. Some attempts have been made with the Fully Convolutional Network (FCN) framework and achieve satisfactory detection results. However, due to stacking convolution layers and pooling operation, the boundary details tend to be lost. In addition, existing models often utilize the extracted features without discrimination, leading to redundancy in representation since actually not all features are helpful to the final prediction and some even bring distraction. In this paper, we propose a co-attention module embedded FCN framework, called as Co-Attention FCN (CA-FCN). Specifically, the co-attention module is plugged into the high-level convolution layers of FCN, which can assign larger attention weights on the common salient objects and smaller ones on the background and uncommon distractors to boost final detection performance. Extensive experiments on three popular co-saliency benchmark datasets demonstrate the superiority of the proposed CA-FCN, which outperforms state-of-the-arts in most cases. Besides, the effectiveness of our new co-attention module is also validated with ablation studies.

Here, we provide preliminary experimental results of the geopotential determination based on time elapse comparisons between two remote atomic clocks located at Beijing and Wuhan, respectively. After synchronizing two hydrogen atomic clocks at Beijing 203 Institute Laboratory (BIL) for 20 days as zero-baseline calibration, we transport one clock to Luojiashan Time-Frequency Station (LTS), Wuhan, without stopping its running. Continuous comparisons between the two remote clocks were conducted for 65 days based on the Common View Satellite Time Transfer (CVSTT) technique. The ensemble empirical mode decomposition (EEMD) technique is applied to removing the uninteresting periodic signals contaminated in the original CVSTT observations to obtain the residual clocks-offsets series, from which the time elapse between the two remote clocks was determined. Based on the accumulated time elapse between these two clocks the geopotential difference between these two stations was determined. Given the orthometric height (OH) of BIL, the OH of the LTS was determined based on the determined geopotential difference. Comparisons show that the OH of the LTS determined by time elapse comparisons deviates from that determined by Earth gravity model EGM2008 by about 98 m. The results are consistent with the frequency stabilities of the hydrogen atomic clocks (at the level of $10^{-15}$/day) applied in our experiments. In addition, we used 85-days original observations to determine the geopotential difference between two remote stations based on the CVSTT technique. Using more precise atomic or optical clocks, the CVSTT method for geopotential determination could be applied effectively and extensively in geodesy in the future.

Here we provide an alternative approach to determine the Earth's external gravitational potential field based on low-orbit target satellite (TS), geostationary satellites (GS), and microwave signal links between them. By emitting and receiving frequency signals controlled by precise clocks between TS and GS, we can determine the gravitational potential (GP) at the TS orbit. We set the TS with polar orbits, altitude of around 500 km above ground, and three evenly distributed GSs with equatorial orbits, altitudes of around 35000 km from the Earth's center. In this case, at any time the TS can be observed via frequency signal links by at least one GS. In this way we may determine a potential distribution over the TS-defined sphere (TDS), which is a sphere that best fits the TS' orbits. Then, based on the potential distribution over the TDS, an Earth's external gravitational field can be determined. Simulation results show that the accuracy of the potential filed established based on 30-days observations can achieve decimeter level if optical atomic clocks with instability of $1\times 10^{-17}\tau^{-1/2}$ are available. The formulation proposed in this study may enrich the approachs for determining the Earth's external gravity field.

Object proposal generation is often the first step in many detection models. It is lucrative to train a good proposal model, that generalizes to unseen classes. This could help scaling detection models to larger number of classes with fewer annotations. Motivated by this, we study how a detection model trained on a small set of source classes can provide proposals that generalize to unseen classes. We systematically study the properties of the dataset - visual diversity and label space granularity - required for good generalization. We show the trade-off between using fine-grained labels and coarse labels. We introduce the idea of prototypical classes: a set of sufficient and necessary classes required to train a detection model to obtain generalized proposals in a more data-efficient way. On the Open Images V4 dataset, we show that only 25% of the classes can be selected to form such a prototypical set. The resulting proposals from a model trained with these classes is only 4.3% worse than using all the classes, in terms of average recall (AR). We also demonstrate that Faster R-CNN model leads to better generalization of proposals compared to a single-stage network like RetinaNet.

We develop a non-associated flow rule (NAFR) based elasto-viscoplastic (EVP) model for isotropic clays. For the model formulation, we introduce the critical state soil mechanics theory (CSSMT), the bounding surface theory and Perzyna's overstress theory. The NAFR based EVP model comprises three surfaces: the potential surface, the reference surface and the loading surface. Additionally, in the model formulation, assuming the potential surface and the reference surface are identical, we obtain the associated flow rule-based EVP model. Both EVP models require seven parameters and five of them are identical to the Modified Cam Clay model. The other two parameters are the surface shape parameter and the secondary compression index. Moreover, we introduce the shape parameter in the model formulation to control the surface shape and to account for the overconsolidation state of clay. Additionally, we incorporate the secondary compression index to introduce the viscosity of clay. Also, we validate the EVP model performances for the Shanghai clay,the San Francisco Bay Mud (SFBM) clay and the Kaolin clay. Furthermore, we use the EVP models to predict the long-term field monitoring measurement of the Nerang Broadbeach roadway embankment in Australia. From the comparison of model predictions, we find that the non-associated flow rule EVP model captures well a wide range of experimental results and field monitoring embankment data. Furthermore, we also observe that the natural clay exhibits the flow rule effect more compared to the reconstituted clay.

Vehicle detection in remote sensing images has attracted increasing interest in recent years. However, its detection ability is limited due to lack of well-annotated samples, especially in densely crowded scenes. Furthermore, since a list of remotely sensed data sources is available, efficient exploitation of useful information from multi-source data for better vehicle detection is challenging. To solve the above issues, a multi-source active fine-tuning vehicle detection (Ms-AFt) framework is proposed, which integrates transfer learning, segmentation, and active classification into a unified framework for auto-labeling and detection. The proposed Ms-AFt employs a fine-tuning network to firstly generate a vehicle training set from an unlabeled dataset. To cope with the diversity of vehicle categories, a multi-source based segmentation branch is then designed to construct additional candidate object sets. The separation of high quality vehicles is realized by a designed attentive classifications network. Finally, all three branches are combined to achieve vehicle detection. Extensive experimental results conducted on two open ISPRS benchmark datasets, namely the Vaihingen village and Potsdam city datasets, demonstrate the superiority and effectiveness of the proposed Ms-AFt for vehicle detection. In addition, the generalization ability of Ms-AFt in dense remote sensing scenes is further verified on stereo aerial imagery of a large camping site.

Drifters designed to mimic floating marine debris and small patches of pelagic \emph{Sargassum} were satellite tracked in four regions across the North Atlantic. Though subjected to the same initial conditions at each site, the tracks of different drifters quickly diverged after deployment. We explain the clustering of drifter types using a recent Maxey-Riley theory for surface ocean inertial particle dynamics applied on multidata-based mesoscale ocean currents and winds from reanalysis. Simulated trajectories of objects at the air-sea interface are significantly improved when represented as inertial (accounting for buoyancy and size), rather than as perfectly Lagrangian (fluid following) particles. Separation distances between simulated and observed trajectories were substantially smaller for debris-like drifters than for \emph{Sargassum}-like drifters, suggesting that additional consideration of its physical properties relative to fluid velocities may be useful. Our findings can be applied to model variability in movements and distribution of diverse objects floating at the ocean surface.

The increasing accessibility and precision of Earth observation satellite data offers considerable opportunities for industrial and state actors alike. This calls however for efficient methods able to process time-series on a global scale. Building on recent work employing multi-headed self-attention mechanisms to classify remote sensing time sequences, we propose a modification of the Temporal Attention Encoder. In our network, the channels of the temporal inputs are distributed among several compact attention heads operating in parallel. Each head extracts highly-specialized temporal features which are in turn concatenated into a single representation. Our approach outperforms other state-of-the-art time series classification algorithms on an open-access satellite image dataset, while using significantly fewer parameters and with a reduced computational complexity.

In order to mitigate the spread of COVID-19, Wuhan was the first city to implement strict lockdown policy in 2020. Even though numerous researches have discussed the travel restriction between cities and provinces, few studies focus on the effect of transportation control inside the city due to the lack of the measurement and available data in Wuhan. Since the public transports have been shut down in the beginning of city lockdown, the change of traffic density is a good indicator to reflect the intracity population flow. Therefore, in this paper, we collected time-series high-resolution remote sensing images with the resolution of 1m acquired before, during and after Wuhan lockdown by GF-2 satellite. Vehicles on the road were extracted and counted for the statistics of traffic density to reflect the changes of human transmissions in the whole period of Wuhan lockdown. Open Street Map was used to obtain observation road surfaces, and a vehicle detection method combing morphology filter and deep learning was utilized to extract vehicles with the accuracy of 62.56%. According to the experimental results, the traffic density of Wuhan dropped with the percentage higher than 80%, and even higher than 90% on main roads during city lockdown; after lockdown lift, the traffic density recovered to the normal rate. Traffic density distributions also show the obvious reduction and increase throughout the whole study area. The significant reduction and recovery of traffic density indicates that the lockdown policy in Wuhan show effectiveness in controlling human transmission inside the city, and the city returned to normal after lockdown lift.

In light of growing challenges in agriculture with ever growing food demand across the world, efficient crop management techniques are necessary to increase crop yield. Precision agriculture techniques allow the stakeholders to make effective and customized crop management decisions based on data gathered from monitoring crop environments. Plant phenotyping techniques play a major role in accurate crop monitoring. Advancements in deep learning have made previously difficult phenotyping tasks possible. This survey aims to introduce the reader to the state of the art research in deep plant phenotyping.

The Annulated Treeboa (Corallus annulatus) is one of nine currently recognized species in the boid genus Corallus. Its disjunct range extends from eastern Guatemala into northern Honduras, southeastern Nicaragua, northeastern Costa Rica, and southwestern Panama to northern Colombia west of the Andes. It is the only species of Corallus found on the Caribbean versant of Costa Rica, where it occurs at elevations to at least 650m and perhaps as high as 1,000m. Corallus annulatus occurs mostly in primary and secondary lowland tropical wet and moist rainforest and it appears to be genuinely rare. Besides C. cropanii and C. blombergi (the latter closely related to C. annulatus), it is the rarest member of the genus. Aside from information on habitat and activity, little is known regarding its natural history.

In computer vision, object detection is one of most important tasks, which underpins a few instance-level recognition tasks and many downstream applications. Recently one-stage methods have gained much attention over two-stage approaches due to their simpler design and competitive performance. Here we propose a fully convolutional one-stage object detector (FCOS) to solve object detection in a per-pixel prediction fashion, analogue to other dense prediction problems such as semantic segmentation. Almost all state-of-the-art object detectors such as RetinaNet, SSD, YOLOv3, and Faster R-CNN rely on pre-defined anchor boxes. In contrast, our proposed detector FCOS is anchor box free, as well as proposal free. By eliminating the pre-defined set of anchor boxes, FCOS completely avoids the complicated computation related to anchor boxes such as calculating the intersection over union (IoU) scores during training. More importantly, we also avoid all hyper-parameters related to anchor boxes, which are often sensitive to the final detection performance. With the only post-processing non-maximum suppression (NMS), we demonstrate a much simpler and flexible detection framework achieving improved detection accuracy. We hope that the proposed FCOS framework can serve as a simple and strong alternative for many other instance-level tasks. Code and pre-trained models are available at: https://git.io/AdelaiDet

The Australian bushfires around the turn of the year 2020 generated an unprecedented perturbation of stratospheric composition, dynamical circulation and radiative balance. Here we show from satellite observations that the resulting planetary-scale blocking of solar radiation by the smoke is larger than any previously documented wildfires and of the same order as the radiative forcing produced by moderate volcanic eruptions. A striking effect of the solar heating of an intense smoke patch was the generation of a self-maintained anticyclonic vortex measuring 1000 km in diameter and featuring its own ozone hole. The highly stable vortex persisted in the stratosphere for over 13 weeks, travelled 66,000 km and lifted a confined bubble of smoke and moisture to 35 km altitude. Its evolution was tracked by several satellite-based sensors and was successfully resolved by the European Centre for Medium-Range Weather Forecasts operational system, primarily based on satellite data. Because wildfires are expected to increase in frequency and strength in a changing climate, we suggest that extraordinary events of this type may contribute significantly to the global stratospheric composition in the coming decades.

Object detection is an important task in computer vision which serves a lot of real-world applications such as autonomous driving, surveillance and robotics. Along with the rapid thrive of large-scale data, numerous state-of-the-art generalized object detectors (e.g. Faster R-CNN, YOLO, SSD) were developed in the past decade. Despite continual efforts in model modification and improvement in training strategies to boost detection accuracy, there are still limitations in performance of detectors when it comes to specialized datasets with uneven object class distributions. This originates from the common usage of Cross Entropy loss function for object classification sub-task that simply ignores the frequency of appearance of object class during training, and thus results in lower accuracies for object classes with fewer number of samples. Class-imbalance in general machine learning has been widely studied, however, little attention has been paid on the subject of object detection. In this paper, we propose to explore and overcome such problem by application of several weighted variants of Cross Entropy loss, for examples Balanced Cross Entropy, Focal Loss and Class-Balanced Loss Based on Effective Number of Samples to our object detector. Experiments with BDD100K (a highly class-imbalanced driving database acquired from on-vehicle cameras capturing mostly Car-class objects and other minority object classes such as Bus, Person and Motor) have proven better class-wise performances of detector trained with the afore-mentioned loss functions.

Knowledge of the past behavior of Antarctic polynyas such as the Ross and Weddell Seas contributes to the understanding of biological productivity, sea ice production, katabatic and Southern Hemisphere Westerly (SHW) wind strength, Antarctic bottom water (ABW) formation, and marine CO2 sequestration. Previous studies link barium (Ba) marine sedimentation to polynya primary productivity (Bonn et al., 1998; McManus et al., 2002; Pirrung et al., 2008), polynya area to katabatic wind strength and proximal cyclones (Bromwich et al., 1998; Drucker et al., 2011), and highlight the influence of Ross Ice Shelf calving event effects on the Ross Sea Polynya (RSP) (Rhodes et al., 2009). Here we use the RICE ice core, located just 120 km from the Ross Ice Shelf front to capture 1150 years of RSP behavior. We link atmospheric Ba fluctuations to Ba marine sedimentation in the summer Ross Sea Polynya, creating the first deep ice core based RSP proxy. RSP area is currently the smallest ever observed over our 1150-year record, and varied throughout the Little Ice Age with fluctuations in Amundsen Sea Low (ASL) strength related with Ross Sea cyclones. Past RSP reconstructions allow us to predict future responses of RSP area to future climate change, which is of special interest considering the recent disappearance of the Weddell Sea Polynya in response to anthropogenic forcing (de Lavergne et al., 2014).

State-of-the-art object detection approaches such as Fast/Faster R-CNN, SSD, or YOLO have difficulties detecting dense, small targets with arbitrary orientation in large aerial images. The main reason is that using interpolation to align RoI features can result in a lack of accuracy or even loss of location information. We present the Local-aware Region Convolutional Neural Network (LR-CNN), a novel two-stage approach for vehicle detection in aerial imagery. We enhance translation invariance to detect dense vehicles and address the boundary quantization issue amongst dense vehicles by aggregating the high-precision RoIs' features. Moreover, we resample high-level semantic pooled features, making them regain location information from the features of a shallower convolutional block. This strengthens the local feature invariance for the resampled features and enables detecting vehicles in an arbitrary orientation. The local feature invariance enhances the learning ability of the focal loss function, and the focal loss further helps to focus on the hard examples. Taken together, our method better addresses the challenges of aerial imagery. We evaluate our approach on several challenging datasets (VEDAI, DOTA), demonstrating a significant improvement over state-of-the-art methods. We demonstrate the good generalization ability of our approach on the DLR 3K dataset.

The effect of anthropogenic aerosol on the reflectivity of stratocumulus cloud decks through changes in cloud amount is a major uncertainty in climate projections. The focus of this study is the frequently occurring non-precipitating stratocumulus. In this regime, cloud amount can decrease through aerosol-enhanced cloud-top mixing. The climatological relevance of this effect is debated because ship exhaust does not appear to generate significant change in the amount of these clouds. Through a novel analysis of detailed numerical simulations in comparison to satellite data, we show that results from ship-track studies cannot be generalized to estimate the climatological forcing of anthropogenic aerosol. We specifically find that the ship-track-derived sensitivity of the radiative effect of non-precipitating stratocumulus to aerosol overestimates their cooling effect by up to 200%. This offsetting warming effect needs to be taken into account if we are to constrain the aerosol-cloud radiative forcing of stratocumulus.

The earthquake occurred on October 15, 2016 (Ml=5.0) re-attracted attention to the tectonic activity of Western Black Sea Basin. The focal mechanism solution of this earthquake indicates reverse faulting, similar to the Bartin Earthquake of September 3, 1968 (MS=6.6), which is the strongest instrumentally recorded earthquake along the Turkish margin of Black Sea, and reveals another seismological evidence for the active thrusting in the region. In this study, the fault structures that considered to be formed by the effect of compressional tectonic regime and the structures formed by the activities of these faults beneath the shelf and slope areas between the region offshore Akcakoca-Cide at the southern part of the Western Black Sea Basin revealed by using marine seismic reflection data and the composite well log data. By means of the geological sections in previous geological studies, the land-offshore geological sections were prepared and findings about the continuation of the geological features from land to offshore in the study area were presented. These geological sections developed for the study area and the geological features recognized from the seismic migration sections, support the presence of the N-S directional compressional tectonic regime in Western Black Sea.