Electromagnetic Ion Cyclotron (EMIC) waves can drive radiation belt depletion and Low-Earth Orbit (LEO) satellites can detect the resulting electron and proton precipitation. The ELFIN (Electron Losses and Fields InvestigatioN) CubeSats provide an excellent opportunity to study the properties of EMIC-driven electron precipitation with much higher energy and pitch-angle resolution than previously allowed. We collect EMIC-driven electron precipitation events from ELFIN observations and use POES (Polar Orbiting Environmental Satellites) to search for 10s-100s keV proton precipitation nearby as a proxy of EMIC wave activity. Electron precipitation mainly occurs on localized radial scales (0.3 L), over 15-24 MLT and 5-8 L shells, stronger at MeV energies and weaker down to 100-200 keV. Additionally, the observed loss cone pitch-angle distribution agrees with quasilinear predictions at >250 keV (more filled loss cone with increasing energy), while additional mechanisms are needed to explain the observed low-energy precipitation.

The Maasai Mara in Kenya, renowned for its biodiversity, is witnessing ecosystem degradation and species endangerment due to intensified human activities. Addressing this, we introduce a dynamic system harmonizing ecological and human priorities. Our agent-based model replicates the Maasai Mara savanna ecosystem, incorporating 71 animal species, 10 human classifications, and 2 natural resource types. The model employs the metabolic rate-mass relationship for animal energy dynamics, logistic curves for animal growth, individual interactions for food web simulation, and human intervention impacts. Algorithms like fitness proportional selection and particle swarm mimic organism preferences for resources. To guide preservation activities, we formulated 21 management strategies encompassing tourism, transportation, taxation, environmental conservation, research, diplomacy, and poaching, employing a game-theoretic framework. Using the TOPSIS method, we prioritized four key developmental indicators: environmental health, research advancement, economic growth, and security. The interplay of 16 factors determines these indicators, each influenced by our policies to varying degrees. By evaluating the policies' repercussions, we aim to mitigate adverse animal-human interactions and equitably address human concerns. We classified the policy impacts into three categories: Environmental Preservation, Economic Prosperity, and Holistic Development. By applying these policy groupings to our ecosystem model, we tracked the effects on the intricate animal-human-resource dynamics. Utilizing the entropy weight method, we assessed the efficacy of these policy clusters over a decade, identifying the optimal blend emphasizing both environmental conservation and economic progression.

Real-time semantic segmentation of remote sensing imagery is a challenging task that requires a trade-off between effectiveness and efficiency. It has many applications including tracking forest fires, detecting changes in land use and land cover, crop health monitoring, and so on. With the success of efficient deep learning methods (i.e., efficient deep neural networks) for real-time semantic segmentation in computer vision, researchers have adopted these efficient deep neural networks in remote sensing image analysis. This paper begins with a summary of the fundamental compression methods for designing efficient deep neural networks and provides a brief but comprehensive survey, outlining the recent developments in real-time semantic segmentation of remote sensing imagery. We examine several seminal efficient deep learning methods, placing them in a taxonomy based on the network architecture design approach. Furthermore, we evaluate the quality and efficiency of some existing efficient deep neural networks on a publicly available remote sensing semantic segmentation benchmark dataset, the OpenEarthMap. The experimental results of an extensive comparative study demonstrate that most of the existing efficient deep neural networks have good segmentation quality, but they suffer low inference speed (i.e., high latency rate), which may limit their capability of deployment in real-time applications of remote sensing image segmentation. We provide some insights into the current trend and future research directions for real-time semantic segmentation of remote sensing imagery.

The rapid advancement in high-resolution satellite remote sensing data acquisition, particularly those achieving submeter precision, has uncovered the potential for detailed extraction of surface architectural features. However, the diversity and complexity of surface distributions frequently lead to current methods focusing exclusively on localized information of surface features. This often results in significant intraclass variability in boundary recognition and between buildings. Therefore, the task of fine-grained extraction of surface features from high-resolution satellite imagery has emerged as a critical challenge in remote sensing image processing. In this work, we propose the Feature Aggregation Network (FANet), concentrating on extracting both global and local features, thereby enabling the refined extraction of landmark buildings from high-resolution satellite remote sensing imagery. The Pyramid Vision Transformer captures these global features, which are subsequently refined by the Feature Aggregation Module and merged into a cohesive representation by the Difference Elimination Module. In addition, to ensure a comprehensive feature map, we have incorporated the Receptive Field Block and Dual Attention Module, expanding the receptive field and intensifying attention across spatial and channel dimensions. Extensive experiments on multiple datasets have validated the outstanding capability of FANet in extracting features from high-resolution satellite images. This signifies a major breakthrough in the field of remote sensing image processing. We will release our code soon.

Unsupervised and self-supervised representation learning has become popular in recent years for learning useful features from unlabelled data. Representation learning has been mostly developed in the neural network literature, and other models for representation learning are surprisingly unexplored. In this work, we introduce and analyze several kernel-based representation learning approaches: Firstly, we define two kernel Self-Supervised Learning (SSL) models using contrastive loss functions and secondly, a Kernel Autoencoder (AE) model based on the idea of embedding and reconstructing data. We argue that the classical representer theorems for supervised kernel machines are not always applicable for (self-supervised) representation learning, and present new representer theorems, which show that the representations learned by our kernel models can be expressed in terms of kernel matrices. We further derive generalisation error bounds for representation learning with kernel SSL and AE, and empirically evaluate the performance of these methods in both small data regimes as well as in comparison with neural network based models.

Self-Supervised Learning (SSL) is an important paradigm for learning representations from unlabelled data, and SSL with neural networks has been highly successful in practice. However current theoretical analysis of SSL is mostly restricted to generalisation error bounds. In contrast, learning dynamics often provide a precise characterisation of the behaviour of neural networks based models but, so far, are mainly known in supervised settings. In this paper, we study the learning dynamics of SSL models, specifically representations obtained by minimising contrastive and non-contrastive losses. We show that a naive extension of the dymanics of multivariate regression to SSL leads to learning trivial scalar representations that demonstrates dimension collapse in SSL. Consequently, we formulate SSL objectives with orthogonality constraints on the weights, and derive the exact (network width independent) learning dynamics of the SSL models trained using gradient descent on the Grassmannian manifold. We also argue that the infinite width approximation of SSL models significantly deviate from the neural tangent kernel approximations of supervised models. We numerically illustrate the validity of our theoretical findings, and discuss how the presented results provide a framework for further theoretical analysis of contrastive and non-contrastive SSL.

The sea breeze is a phenomenon frequently impacting Long Island, New York, especially during the spring and early summer, when land surface temperatures can exceed ocean temperatures considerably. The sea breeze influences daily weather conditions by causing a shift in wind direction and speed, limiting the maximum temperature, and occasionally serving as a trigger for precipitation and thunderstorms. Advance prediction of the presence or absence of the sea breeze for a certain location on a given day would therefore be beneficial to weather forecasters. To forecast sea breeze occurrence based on the previous night's weather conditions, we used a novel algorithm called the $D$-Basis. We analyzed sea breeze data from a recent four year period (2017-2020) at a single weather station several miles inland from the coast. High or constant station pressure, high or constant dew point, and onshore wind from the previous night were found to be strong predictors of sea breeze formation the following day. The accuracy of the prediction was around 74\% for June 2020. Unlike other prediction methods which involve the comparison of sea surface and land surface temperatures in near real time, our prediction method is based on the parameters from the prior night, allowing it to potentially aid in advanced forecasting of the sea breeze.

Decentralized Federated Learning (DFL) enables collaborative model training without a central server but faces challenges in efficiency, stability, and trustworthiness due to communication and computational limitations among distributed nodes. To address these critical issues, we introduce a sparsity constraint on the shared model, leading to Sparse DFL (SDFL), and propose a novel algorithm, CEPS. The sparsity constraint facilitates the use of one-bit compressive sensing to transmit one-bit information between partially selected neighbour nodes at specific steps, thereby significantly improving communication efficiency. Moreover, we integrate differential privacy into the algorithm to ensure privacy preservation and bolster the trustworthiness of the learning process. Furthermore, CEPS is underpinned by theoretical guarantees regarding both convergence and privacy. Numerical experiments validate the effectiveness of the proposed algorithm in improving communication and computation efficiency while maintaining a high level of trustworthiness.

We quantitatively assess the impacts of Downgrading Protected Areas (PAD) on biodiversity in the U.S.. Results show that PAD events significantly reduce biodiversity. The proximity to PAD events decreases the biodiversity by 26.0% within 50 km compared with records of species further away from the PAD events. We observe an overall 32.3% decrease in abundance after those nearest PAD events are enacted. Abundance declines more in organisms living in contact with water and non-mammals. Species abundance is more sensitive to the negative impacts in areas where PAD events were later reversed, as well as in areas close to protected areas belonging to the International Union for Conservation of Nature (IUCN) category. The enacted PAD events between the period 1903 to 2018 in the U.S. lead to economic losses of approximately $689.95 million due to decrease in abundance. Our results contribute to the understanding on the impact of environmental interventions such as PAD events on biodiversity change and provide important implications on biodiversity conservation policies.

Self-supervised learning (SSL) is a growing torrent that has recently transformed machine learning and its many real world applications, by learning on massive amounts of unlabeled data via self-generated supervisory signals. Unsupervised anomaly detection (AD) has also capitalized on SSL, by self-generating pseudo-anomalies through various data augmentation functions or external data exposure. In this vision paper, we first underline the importance of the choice of SSL strategies on AD performance, by presenting evidences and studies from the AD literature. Equipped with the understanding that SSL incurs various hyperparameters (HPs) to carefully tune, we present recent developments on unsupervised model selection and augmentation tuning for SSL-based AD. We then highlight emerging challenges and future opportunities; on designing new pretext tasks and augmentation functions for different data modalities, creating novel model selection solutions for systematically tuning the SSL HPs, as well as on capitalizing on the potential of pretrained foundation models on AD through effective density estimation.

Avalanche disaster is a major natural disaster that seriously threatens the national infrastructure and personnel's life safety. For a long time, the research of avalanche disaster prediction in the world is insufficient, there are only some basic models and basic conditions of occurrence, and there is no long series and wide range of avalanche disaster prediction products. Based on 7 different bands and different types of multi-source remote sensing data,this study combined with existing avalanche occurrence models, field investigation and statistical data to analyze the causes of avalanche. The U-net convolutional neural network and threshold analysis were used to extract the distribution of long time series avalanch-prone areas in two study areas, Heiluogou in Sichuan Province and along the Zangpo River in Palong, Tibet Autonomous Region. In addition, the relationship between earthquake magnitude and spatial distribution and avalanche occurrence is also analyzed in this study. This study will also continue to build a prior knowledge base of avalanche occurrence conditions, improve the prediction accuracy of the two methods, and produce products in long time series interannual avalanch-prone areas in southwest China, including Sichuan Province, Yunnan Province, and Tibet Autonomous Region. The resulting products will provide high-precision avalanche prediction and safety assurance for engineering construction and mountaineering activities in Southwest China.

This decade is marked by the introduction of Vision Transformer, a radical paradigm shift in broad computer vision. A similar trend is followed in medical imaging, UNet, one of the most influential architectures, has been redesigned with transformers. Recently, the efficacy of convolutional models in vision is being reinvestigated by seminal works such as ConvNext, which elevates a ResNet to Swin Transformer level. Deriving inspiration from this, we aim to improve a purely convolutional UNet model so that it can be on par with the transformer-based models, e.g, Swin-Unet or UCTransNet. We examined several advantages of the transformer-based UNet models, primarily long-range dependencies and cross-level skip connections. We attempted to emulate them through convolution operations and thus propose, ACC-UNet, a completely convolutional UNet model that brings the best of both worlds, the inherent inductive biases of convnets with the design decisions of transformers. ACC-UNet was evaluated on 5 different medical image segmentation benchmarks and consistently outperformed convnets, transformers, and their hybrids. Notably, ACC-UNet outperforms state-of-the-art models Swin-Unet and UCTransNet by $2.64 \pm 2.54\%$ and $0.45 \pm 1.61\%$ in terms of dice score, respectively, while using a fraction of their parameters ($59.26\%$ and $24.24\%$). Our codes are available at https://github.com/kiharalab/ACC-UNet.

Deep learning has gradually become powerful in segmenting and classifying aerial images. However, in remote sensing applications, the lack of training datasets and the difficulty of accuracy assessment have always been challenges for the deep learning based classification. In recent years, interactive semantic segmentation proposed in computer vision has achieved an ideal state of human-computer interaction segmentation. It can provide expert experience and utilize deep learning for efficient segmentation. However, few papers discussed its application in remote sensing imagery. This study aims to bridge the gap between interactive segmentation and remote sensing analysis by conducting a benchmark study on various interactive segmentation models. We assessed the performance of five state-of-the-art interactive segmentation methods (Reviving Iterative Training with Mask Guidance for Interactive Segmentation (RITM), FocalClick, SimpleClick, Iterative Click Loss (ICL), and Segment Anything (SAM)) on two high-resolution aerial imagery datasets. The Cascade-Forward Refinement approach, an innovative inference strategy for interactive segmentation, was also introduced to enhance the segmentation results. We evaluated these methods on various land cover types, object sizes, and band combinations in the datasets. SimpleClick model consistently outperformed the other methods in our experiments. Conversely, the SAM performed less effectively than other models. Building upon these findings, we developed an online tool called RSISeg for interactive segmentation of remote sensing data. RSISeg incorporates a well-performing interactive model that is finetuned with remote sensing data. Compared to existing interactive segmentation tools, RSISeg offers robust interactivity, modifiability, and adaptability to remote sensing data.

While large volumes of unlabeled data are usually available, associated labels are often scarce. The unsupervised domain adaptation problem aims at exploiting labels from a source domain to classify data from a related, yet different, target domain. When time series are at stake, new difficulties arise as temporal shifts may appear in addition to the standard feature distribution shift. In this paper, we introduce the Match-And-Deform (MAD) approach that aims at finding correspondences between the source and target time series while allowing temporal distortions. The associated optimization problem simultaneously aligns the series thanks to an optimal transport loss and the time stamps through dynamic time warping. When embedded into a deep neural network, MAD helps learning new representations of time series that both align the domains and maximize the discriminative power of the network. Empirical studies on benchmark datasets and remote sensing data demonstrate that MAD makes meaningful sample-to-sample pairing and time shift estimation, reaching similar or better classification performance than state-of-the-art deep time series domain adaptation strategies.

Seismology has witnessed significant advancements in recent years with the application of deep learning methods to address a broad range of problems. These techniques have demonstrated their remarkable ability to effectively extract statistical properties from extensive datasets, surpassing the capabilities of traditional approaches to an extent. In this study, we present SAIPy, an open source Python package specifically developed for fast data processing by implementing deep learning. SAIPy offers solutions for multiple seismological tasks, including earthquake detection, magnitude estimation, seismic phase picking, and polarity identification. We introduce upgraded versions of previously published models such as CREIMERT capable of identifying earthquakes with an accuracy above 99.8 percent and a root mean squared error of 0.38 unit in magnitude estimation. These upgraded models outperform state of the art approaches like the Vision Transformer network. SAIPy provides an API that simplifies the integration of these advanced models, including CREIMERT, DynaPickerv2, and PolarCAP, along with benchmark datasets. The package has the potential to be used for real time earthquake monitoring to enable timely actions to mitigate the impact of seismic events. Ongoing development efforts aim to enhance the performance of SAIPy and incorporate additional features that enhance exploration efforts, and it also would be interesting to approach the retraining of the whole package as a multi-task learning problem.

Trace-wise noise is a type of noise often seen in seismic data, which is characterized by vertical coherency and horizontal incoherency. Using self-supervised deep learning to attenuate this type of noise, the conventional blind-trace deep learning trains a network to blindly reconstruct each trace in the data from its surrounding traces; it attenuates isolated trace-wise noise but causes signal leakage in clean and noisy traces and reconstruction errors next to each noisy trace. To reduce signal leakage and improve denoising, we propose a new loss function and masking procedure in semi-blind-trace deep learning. Our hybrid loss function has weighted active zones that cover masked and non-masked traces. Therefore, the network is not blinded to clean traces during their reconstruction. During training, we dynamically change the masks' characteristics. The goal is to train the network to learn the characteristics of the signal instead of noise. The proposed algorithm enables the designed U-net to detect and attenuate trace-wise noise without having prior information about the noise. A new hyperparameter of our method is the relative weight between the masked and non-masked traces' contribution to the loss function. Numerical experiments show that selecting a small value for this parameter is enough to significantly decrease signal leakage. The proposed algorithm is tested on synthetic and real off-shore and land datasets with different noises. The results show the superb ability of the method to attenuate trace-wise noise while preserving other events. An implementation of the proposed algorithm as a Python code is also made available.

Prediction of dynamic environmental variables in unmonitored sites remains a long-standing challenge for water resources science. The majority of the world's freshwater resources have inadequate monitoring of critical environmental variables needed for management. Yet, the need to have widespread predictions of hydrological variables such as river flow and water quality has become increasingly urgent due to climate and land use change over the past decades, and their associated impacts on water resources. Modern machine learning methods increasingly outperform their process-based and empirical model counterparts for hydrologic time series prediction with their ability to extract information from large, diverse data sets. We review relevant state-of-the art applications of machine learning for streamflow, water quality, and other water resources prediction and discuss opportunities to improve the use of machine learning with emerging methods for incorporating watershed characteristics into deep learning models, transfer learning, and incorporating process knowledge into machine learning models. The analysis here suggests most prior efforts have been focused on deep learning learning frameworks built on many sites for predictions at daily time scales in the United States, but that comparisons between different classes of machine learning methods are few and inadequate. We identify several open questions for time series predictions in unmonitored sites that include incorporating dynamic inputs and site characteristics, mechanistic understanding and spatial context, and explainable AI techniques in modern machine learning frameworks.

Semantic segmentation (classification) of Earth Observation imagery is a crucial task in remote sensing. This paper presents a comprehensive review of technical factors to consider when designing neural networks for this purpose. The review focuses on Convolutional Neural Networks (CNNs), Recurrent Neural Networks (RNNs), Generative Adversarial Networks (GANs), and transformer models, discussing prominent design patterns for these ANN families and their implications for semantic segmentation. Common pre-processing techniques for ensuring optimal data preparation are also covered. These include methods for image normalization and chipping, as well as strategies for addressing data imbalance in training samples, and techniques for overcoming limited data, including augmentation techniques, transfer learning, and domain adaptation. By encompassing both the technical aspects of neural network design and the data-related considerations, this review provides researchers and practitioners with a comprehensive and up-to-date understanding of the factors involved in designing effective neural networks for semantic segmentation of Earth Observation imagery.

The increasing frequency and scale of wildfires carry significant ecological, socioeconomic, and environmental implications, prompting the need for a deeper grasp of wildfire characteristics. Essential meteorological factors like temperature, humidity, and precipitation wield a crucial impact on fire behavior and the estimation of burned areas. This study aims to unravel the interconnections between meteorological conditions and fire attributes within the Salmon-Challis National Forest located in east-central Idaho, USA. Through the utilization of remotely sensed data from the Fire Monitoring, Mapping, and Modeling system (Fire M3) alongside meteorological variables recorded between 2010 and 2020, an exploration is conducted into varied meteorological patterns associated with wildfire events. By integrating the computed burned area into the clustering process, valuable insights are gained into the specific influences of fire weather conditions on the extent of burned areas. The Salmon-Challis National Forest, encompassing more than 4.3 million acres and encompassing the largest wilderness area in the Continental United States, emerges as a pivotal research site for wildfire investigations. This work elucidates the data attributes employed for clustering and visualization, along with the algorithms employed. Additionally, the study presents research findings and delineates potential future applications, ultimately contributing to the advancement of fire management and mitigation strategies in regions prone to wildfires.

Multispectral imagery is frequently incorporated into agricultural tasks, providing valuable support for applications such as image segmentation, crop monitoring, field robotics, and yield estimation. From an image segmentation perspective, multispectral cameras can provide rich spectral information, helping with noise reduction and feature extraction. As such, this paper concentrates on the use of fusion approaches to enhance the segmentation process in agricultural applications. More specifically, in this work, we compare different fusion approaches by combining RGB and NDVI as inputs for crop row detection, which can be useful in autonomous robots operating in the field. The inputs are used individually as well as combined at different times of the process (early and late fusion) to perform classical and DL-based semantic segmentation. In this study, two agriculture-related datasets are subjected to analysis using both deep learning (DL)-based and classical segmentation methodologies. The experiments reveal that classical segmentation methods, utilizing techniques such as edge detection and thresholding, can effectively compete with DL-based algorithms, particularly in tasks requiring precise foreground-background separation. This suggests that traditional methods retain their efficacy in certain specialized applications within the agricultural domain. Moreover, among the fusion strategies examined, late fusion emerges as the most robust approach, demonstrating superiority in adaptability and effectiveness across varying segmentation scenarios. The dataset and code is available at https://github.com/Cybonic/MISAgriculture.git.

Limited availability of labeled data for machine learning on multimodal time-series extensively hampers progress in the field. Self-supervised learning (SSL) is a promising approach to learning data representations without relying on labels. However, existing SSL methods require expensive computations of negative pairs and are typically designed for single modalities, which limits their versatility. We introduce CroSSL (Cross-modal SSL), which puts forward two novel concepts: masking intermediate embeddings produced by modality-specific encoders, and their aggregation into a global embedding through a cross-modal aggregator that can be fed to down-stream classifiers. CroSSL allows for handling missing modalities and end-to-end cross-modal learning without requiring prior data preprocessing for handling missing inputs or negative-pair sampling for contrastive learning. We evaluate our method on a wide range of data, including motion sensors such as accelerometers or gyroscopes and biosignals (heart rate, electroencephalograms, electromyograms, electrooculograms, and electrodermal) to investigate the impact of masking ratios and masking strategies for various data types and the robustness of the learned representations to missing data. Overall, CroSSL outperforms previous SSL and supervised benchmarks using minimal labeled data, and also sheds light on how latent masking can improve cross-modal learning. Our code is open-sourced at https://github.com/dr-bell/CroSSL.

Change detection based on remote sensing images has been a prominent area of interest in the field of remote sensing. Deep networks have demonstrated significant success in detecting changes in bi-temporal remote sensing images and have found applications in various fields. Given the degradation of natural environments and the frequent occurrence of natural disasters, accurately and swiftly identifying damaged buildings in disaster-stricken areas through remote sensing images holds immense significance. This paper aims to investigate change detection specifically for natural disasters. Considering that existing public datasets used in change detection research are registered, which does not align with the practical scenario where bi-temporal images are not matched, this paper introduces an unregistered end-to-end change detection synthetic dataset called xBD-E2ECD. Furthermore, we propose an end-to-end change detection network named E2ECDNet, which takes an unregistered bi-temporal image pair as input and simultaneously generates the flow field prediction result and the change detection prediction result. It is worth noting that our E2ECDNet also supports change detection for registered image pairs, as registration can be seen as a special case of non-registration. Additionally, this paper redefines the criteria for correctly predicting a positive case and introduces neighborhood-based change detection evaluation metrics. The experimental results have demonstrated significant improvements.

Semantic segmentation plays a vital role in computer vision tasks, enabling precise pixel-level understanding of images. In this paper, we present a comprehensive library for semantic segmentation, which contains implementations of popular segmentation models like SegNet, FCN, UNet, and PSPNet. We also evaluate and compare these models on several datasets, offering researchers and practitioners a powerful toolset for tackling diverse segmentation challenges.

Self-supervised learning (SSL) for clinical time series data has received significant attention in recent literature, since these data are highly rich and provide important information about a patient's physiological state. However, most existing SSL methods for clinical time series are limited in that they are designed for unimodal time series, such as a sequence of structured features (e.g., lab values and vitals signs) or an individual high-dimensional physiological signal (e.g., an electrocardiogram). These existing methods cannot be readily extended to model time series that exhibit multimodality, with structured features and high-dimensional data being recorded at each timestep in the sequence. In this work, we address this gap and propose a new SSL method -- Sequential Multi-Dimensional SSL -- where a SSL loss is applied both at the level of the entire sequence and at the level of the individual high-dimensional data points in the sequence in order to better capture information at both scales. Our strategy is agnostic to the specific form of loss function used at each level -- it can be contrastive, as in SimCLR, or non-contrastive, as in VICReg. We evaluate our method on two real-world clinical datasets, where the time series contains sequences of (1) high-frequency electrocardiograms and (2) structured data from lab values and vitals signs. Our experimental results indicate that pre-training with our method and then fine-tuning on downstream tasks improves performance over baselines on both datasets, and in several settings, can lead to improvements across different self-supervised loss functions.

Remote sensing observations of the Earth's surface are frequently stymied by clouds, water vapour, and aerosols in our atmosphere. These degrade or preclude the measurementof quantities critical to scientific and, hence, societal applications. In this study, we train a natural language processing (NLP) algorithm with high-fidelity ocean simulations in order to accurately reconstruct masked or missing data in sea surface temperature (SST)--i.e. one of 54 essential climate variables identified by the Global Climate Observing System. We demonstrate that the Enki model repeatedly outperforms previously adopted inpainting techniques by up to an order-of-magnitude in reconstruction error, while displaying high performance even in circumstances where the majority of pixels are masked. Furthermore, experiments on real infrared sensor data with masking fractions of at least 40% show reconstruction errors of less than the known sensor uncertainty (RMSE < ~0.1K). We attribute Enki's success to the attentive nature of NLP combined with realistic SST model outputs, an approach that may be extended to other remote sensing variables. This study demonstrates that systems built upon Enki--or other advanced systems like it--may therefore yield the optimal solution to accurate estimates of otherwise missing or masked parameters in climate-critical datasets sampling a rapidly changing Earth.

Transformer-based pre-trained language models, such as BERT, achieve great success in various natural language understanding tasks. Prior research found that BERT captures a rich hierarchy of linguistic information at different layers. However, the vanilla BERT uses the same self-attention mechanism for each layer to model the different contextual features. In this paper, we propose a HybridBERT model which combines self-attention and pooling networks to encode different contextual features in each layer. Additionally, we propose a simple DropMask method to address the mismatch between pre-training and fine-tuning caused by excessive use of special mask tokens during Masked Language Modeling pre-training. Experiments show that HybridBERT outperforms BERT in pre-training with lower loss, faster training speed (8% relative), lower memory cost (13% relative), and also in transfer learning with 1.5% relative higher accuracies on downstream tasks. Additionally, DropMask improves accuracies of BERT on downstream tasks across various masking rates.

The accurate mapping of crop production is crucial for ensuring food security, effective resource management, and sustainable agricultural practices. One way to achieve this is by analyzing high-resolution satellite imagery. Deep Learning has been successful in analyzing images, including remote sensing imagery. However, capturing intricate crop patterns is challenging due to their complexity and variability. In this paper, we propose a novel Deep learning approach that integrates HRNet with Separable Convolutional layers to capture spatial patterns and Self-attention to capture temporal patterns of the data. The HRNet model acts as a backbone and extracts high-resolution features from crop images. Spatially separable convolution in the shallow layers of the HRNet model captures intricate crop patterns more effectively while reducing the computational cost. The multi-head attention mechanism captures long-term temporal dependencies from the encoded vector representation of the images. Finally, a CNN decoder generates a crop map from the aggregated representation. Adaboost is used on top of this to further improve accuracy. The proposed algorithm achieves a high classification accuracy of 97.5\% and IoU of 55.2\% in generating crop maps. We evaluate the performance of our pipeline on the Zuericrop dataset and demonstrate that our results outperform state-of-the-art models such as U-Net++, ResNet50, VGG19, InceptionV3, DenseNet, and EfficientNet. This research showcases the potential of Deep Learning for Earth Observation Systems.

In recent years (2000-2021), human-space activities have been increasing faster than ever. More than 36000 Earth' orbiting objects, all larger than 10 cm, in orbit around the Earth, are currently tracked by the European Space Agency (ESA). Around 70\% of all cataloged objects are in Low-Earth Orbit (LEO). Aerodynamic drag provides one of the main sources of perturbations in this population, gradually decreasing the semi-major axis and period of the LEO satellites. Usually, an empirical atmosphere model as a function of solar radio flux and geomagnetic data is used to calculate the orbital decay and lifetimes of LEO satellites. In this respect, a good forecast for the space weather data could be a key tool to improve the model of drag. In this work, we propose using Time Series Forecasting Model to predict the future behavior of the solar flux and to calculate the atmospheric density, to improve the analytical models and reduce the drag uncertainty.

Collision avoidance is a topic of growing importance for any satellite orbiting Earth. Especially those satellites without thrusting capabilities face the problem of not being able to perform impulsive collision avoidance manoeuvres. For satellites in Low Earth Orbits, though, perturbing accelerations due to aerodynamic drag may be used to influence their trajectories, thus offering a possibility to avoid collisions without consuming propellant. Here, this manoeuvring option is investigated for the satellite Flying Laptop of the University of Stuttgart, which orbits the Earth at approximately 600 km. In a first step, the satellite is aerodynamically analysed making use of the tool ADBSat. By employing an analytic equation from literature, in-track separation distances can then be derived following a variation of the ballistic coefficient through a change in attitude. A further examination of the achievable separation distances proves the feasibility of aerodynamic collision avoidance manoeuvres for the Flying Laptop for moderate and high solar and geomagnetic activity. The predicted separation distances are further compared to flight data, where the principle effect of the manoeuvre on the satellite trajectory becomes visible. The results suggest an applicability of collision avoidance manoeuvres for all satellites in comparable and especially in lower orbits than the Flying Laptop, which are able to vary their ballistic coefficient.

Remote sensing image super-resolution (RSISR) plays a vital role in enhancing spatial detials and improving the quality of satellite imagery. Recently, Transformer-based models have shown competitive performance in RSISR. To mitigate the quadratic computational complexity resulting from global self-attention, various methods constrain attention to a local window, enhancing its efficiency. Consequently, the receptive fields in a single attention layer are inadequate, leading to insufficient context modeling. Furthermore, while most transform-based approaches reuse shallow features through skip connections, relying solely on these connections treats shallow and deep features equally, impeding the model's ability to characterize them. To address these issues, we propose a novel transformer architecture called Cross-Spatial Pixel Integration and Cross-Stage Feature Fusion Based Transformer Network (SPIFFNet) for RSISR. Our proposed model effectively enhances global cognition and understanding of the entire image, facilitating efficient integration of features cross-stages. The model incorporates cross-spatial pixel integration attention (CSPIA) to introduce contextual information into a local window, while cross-stage feature fusion attention (CSFFA) adaptively fuses features from the previous stage to improve feature expression in line with the requirements of the current stage. We conducted comprehensive experiments on multiple benchmark datasets, demonstrating the superior performance of our proposed SPIFFNet in terms of both quantitative metrics and visual quality when compared to state-of-the-art methods.

The North American Aerospace Defense Command (NORAD) tracking of the SpaceX Starlink satellite launch on February 03, 2022 is reviewed. Of the 49 Starlink satellites released into orbit, 38 were eventually lost. Thirty-two of the satellites were never tracked by NORAD. There have been three articles written proposing physical mechanisms to explain the satellite losses. It is argued that none of the proposed mechanisms can explain the immediate loss of 32 of the 49 satellites. The non-availability of telemetry data from the lost satellites has hindered the search for a physical mechanism to explain the density increase observed in a short time interval.

We provide an updated 5 $^\prime $ $\times $ 5 $^\prime $ global geoid model GGM2022, which is determined based on the shallow layer method (Shen method). First, we choose an inner surface $\Gamma$ below the EGM2008 global geoid by 15 m, and the layer bounded by the inner surface $\Gamma$ and the Earth's geographical surface $S$ is referred to as the shallow layer. The Earth's geographical surface $S$ is determined by the digital topographic model DTM2006.0 combining with the DNSC2008 mean sea surface. Second, we formulate the 3D shallow mass layer model using the refined 5 $^\prime $ $\times $ 5 $^\prime $ crust density model CRUST$_-$re , which { is an improved 5 $^\prime $ $\times $ 5 $^\prime $ density model of the CRUST2.0 or CRUST1.0 with taking into account the corrections of the areas covered by ice sheets and the land-ocean crossing regions. Third, based on the shallow mass layer model and the gravity field EGM2008 that is defined in the region outside the Earth's geographical surface $S$, we determine the gravity field model EGM2008s that is defined in the whole region outside the inner surface $\Gamma$, where the definition domain of the gravity field is extended from the domain outside $S$ to the domain outside $\Gamma$. Fourth, based on the gravity field EGM2008s and the geodetic equation $W(P)=W_0$ (where $W_0$ is the geopotential constant on the geoid and $P$ is the point on the geoid $G$), we determine a 5 $^\prime $ $\times $ 5 $^\prime $ global geoid, which is referred to as GGM2022. Comparisons show that the GGM2022 fits the globally available GPS/leveling data better than EGM2008 global geoid in the USA, Europe and the western part of China.

Remotely sensed nighttime lights (NTL) uniquely capture urban change processes that are important to human and ecological well-being, such as urbanization, socio-political conflicts and displacement, impacts from disasters, holidays, and changes in daily human patterns of movement. Though several NTL products are global in extent, intrinsic city-specific factors that affect lighting, such as development levels, and social, economic, and cultural characteristics, are unique to each city, making the urban processes embedded in NTL signatures difficult to characterize, and limiting the scalability of urban change analyses. In this study, we propose a data-driven approach to detect urban changes from daily satellite-derived NTL data records that is adaptive across cities and effective at learning city-specific temporal patterns. The proposed method learns to forecast NTL signatures from past data records using neural networks and allows the use of large volumes of unlabeled data, eliminating annotation effort. Urban changes are detected based on deviations of observed NTL from model forecasts using an anomaly detection approach. Comparing model forecasts with observed NTL also allows identifying the direction of change (positive or negative) and monitoring change severity for tracking recovery. In operationalizing the model, we consider ten urban areas from diverse geographic regions with dynamic NTL time-series and demonstrate the generalizability of the approach for detecting the change processes with different drivers and rates occurring within these urban areas based on NTL deviation. This scalable approach for monitoring changes from daily remote sensing observations efficiently utilizes large data volumes to support continuous monitoring and decision making.

Time series classification is a task which deals with temporal sequences, a prevalent data type common in domains such as human activity recognition, sports analytics and general sensing. In this area, interest in explainability has been growing as explanation is key to understand the data and the model better. Recently, a great variety of techniques have been proposed and adapted for time series to provide explanation in the form of saliency maps, where the importance of each data point in the time series is quantified with a numerical value. However, the saliency maps can and often disagree, so it is unclear which one to use. This paper provides a novel framework to quantitatively evaluate and rank explanation methods for time series classification. We show how to robustly evaluate the informativeness of a given explanation method (i.e., relevance for the classification task), and how to compare explanations side-by-side. The goal is to recommend the best explainer for a given time series classification dataset. We propose AMEE, a Model-Agnostic Explanation Evaluation framework, for recommending saliency-based explanations for time series classification. In this approach, data perturbation is added to the input time series guided by each explanation. Our results show that perturbing discriminative parts of the time series leads to significant changes in classification accuracy, which can be used to evaluate each explanation. To be robust to different types of perturbations and different types of classifiers, we aggregate the accuracy loss across perturbations and classifiers. This novel approach allows us to recommend the best explainer among a set of different explainers, including random and oracle explainers. We provide a quantitative and qualitative analysis for synthetic datasets, a variety of timeseries datasets, as well as a real-world case study with known expert ground truth.

Accurate seismic phase detection and onset picking are fundamental to seismological studies. Supervised deep-learning phase pickers have shown promise with excellent performance on land seismic data. Although it may be acceptable to apply them to OBS (Ocean Bottom Seismometers) data that are indispensable for studying ocean regions, they suffer from a significant performance drop. In this study, we develop a generalised transfer-learned OBS phase picker - OBSTransformer, based on automated labelling and transfer learning. First, we compile a comprehensive dataset of catalogued earthquakes recorded by 423 OBSs from 11 temporary deployments worldwide. Through automated processes, we label the P and S phases of these earthquakes by analysing the consistency of at least three arrivals from four widely used machine learning pickers (EqTransformer, PhaseNet, Generalized Phase Detection, and PickNet), as well as the AIC picker. This results in an inclusive OBS dataset containing ~36,000 earthquake samples. Subsequently, we employ this dataset for transfer learning and utilize a well-trained land machine learning model - EqTransformer as our base model. Moreover, we extract 25,000 OBS noise samples from the same OBS networks using the Kurtosis method, which are then used for model training alongside the labelled earthquake samples. Using three groups of test datasets at sub-global, regional, and local scales, we demonstrate that OBSTransformer outperforms EqTransformer. Particularly, the P and S recall rates at large distances (>200 km) are increased by 68% and 76%, respectively. Our extensive tests and comparisons demonstrate that OBSTransformer is less dependent on the detection/picking thresholds and is more robust to noise levels.

Solar power harbors immense potential in mitigating climate change by substantially reducing CO$_{2}$ emissions. Nonetheless, the inherent variability of solar irradiance poses a significant challenge for seamlessly integrating solar power into the electrical grid. While the majority of prior research has centered on employing purely time series-based methodologies for solar forecasting, only a limited number of studies have taken into account factors such as cloud cover or the surrounding physical context. In this paper, we put forth a deep learning architecture designed to harness spatio-temporal context using satellite data, to attain highly accurate \textit{day-ahead} time-series forecasting for any given station, with a particular emphasis on forecasting Global Horizontal Irradiance (GHI). We also suggest a methodology to extract a distribution for each time step prediction, which can serve as a very valuable measure of uncertainty attached to the forecast. When evaluating models, we propose a testing scheme in which we separate particularly difficult examples from easy ones, in order to capture the model performances in crucial situations, which in the case of this study are the days suffering from varying cloudy conditions. Furthermore, we present a new multi-modal dataset gathering satellite imagery over a large zone and time series for solar irradiance and other related physical variables from multiple geographically diverse solar stations. Our approach exhibits robust performance in solar irradiance forecasting, including zero-shot generalization tests at unobserved solar stations, and holds great promise in promoting the effective integration of solar power into the grid.

This paper presents SSSegmenation, which is an open source supervised semantic image segmentation toolbox based on PyTorch. The design of this toolbox is motivated by MMSegmentation while it is easier to use because of fewer dependencies and achieves superior segmentation performance under a comparable training and testing setup. Moreover, the toolbox also provides plenty of trained weights for popular and contemporary semantic segmentation methods, including Deeplab, PSPNet, OCRNet, MaskFormer, \emph{etc}. We expect that this toolbox can contribute to the future development of semantic segmentation. Codes and model zoos are available at \href{https://github.com/SegmentationBLWX/sssegmentation/}{SSSegmenation}.

Monitoring plant health is crucial for maintaining agricultural productivity and food safety. Disruptions in the plant's normal state, caused by diseases, often interfere with essential plant activities, and timely detection of these diseases can significantly mitigate crop loss. In this study, we propose a deep learning-based approach for efficient detection of plant diseases using drone-captured imagery. A comprehensive database of various plant species, exhibiting numerous diseases, was compiled from the Internet and utilized as the training and test dataset. A Convolutional Neural Network (CNN), renowned for its performance in image classification tasks, was employed as our primary predictive model. The CNN model, trained on this rich dataset, demonstrated superior proficiency in crop disease categorization and detection, even under challenging imaging conditions. For field implementation, we deployed a prototype drone model equipped with a high-resolution camera for live monitoring of extensive agricultural fields. The captured images served as the input for our trained model, enabling real-time identification of healthy and diseased plants. Our approach promises an efficient and scalable solution for improving crop health monitoring systems.

Although the Transformer model can effectively acquire context features via a self-attention mechanism, deeper syntactic knowledge is still not effectively modeled. To alleviate the above problem, we propose Syntactic knowledge via Graph attention with BERT (SGB) in Machine Translation (MT) scenarios. Graph Attention Network (GAT) and BERT jointly represent syntactic dependency feature as explicit knowledge of the source language to enrich source language representations and guide target language generation. Our experiments use gold syntax-annotation sentences and Quality Estimation (QE) model to obtain interpretability of translation quality improvement regarding syntactic knowledge without being limited to a BLEU score. Experiments show that the proposed SGB engines improve translation quality across the three MT tasks without sacrificing BLEU scores. We investigate what length of source sentences benefits the most and what dependencies are better identified by the SGB engines. We also find that learning of specific dependency relations by GAT can be reflected in the translation quality containing such relations and that syntax on the graph leads to new modeling of syntactic aspects of source sentences in the middle and bottom layers of BERT.

In this report we investigate whether and under what conditions virtual seismology via the acoustic Marchenko method can be applied to DAS data from a survey in the province of Groningen, The Netherlands. Virtual seismology allows to retrieve the band-limited Green's function between a virtual source at an arbitrary focal point in the subsurface, while accounting for all orders of multiples. The method requires the reflection response at the surface and an estimate of the traveltime between the surface and focal point. However, in order to successfully apply the method the reflection response needs to be free from surface waves and other direct waves, and properly scaled in order for the Marchenko scheme to converge. These limitations severely complicate the application of the Marchenko method to field data, especially seismic surveys on land. This report considers a full 2D geophone survey as well as a 1.5D approximation for a DAS survey, and compares the results of the virtual sources with an actual dynamite source. The results show that virtual seismology can be used to recreate the reflections recorded at the surface from the dynamite source using either geophone or DAS data.

This paper investigates the key role of Feed-Forward Networks (FFNs) in transformer models by utilizing the Parallel Attention and Feed-Forward Net Design (PAF) architecture, and comparing it to their Series Attention and Feed-Forward Net Design (SAF) counterparts. Central to the effectiveness of PAF are two main assumptions regarding the FFN block and the attention block within a layer: 1) the primary function of the FFN block is to maintain isotropy among token embeddings and prevent their degeneration, and 2) the residual norm computed in the attention block is substantially smaller than the input token embedding norm. To empirically validate these assumptions, we train PAF variants of two large language models (RoBERTa-large and bert-large-uncased). Our results demonstrate that both assumptions hold true in the PAF design. This study contributes to a deeper understanding of the roles and interactions between FFNs and self-attention mechanisms in transformer architectures.

The competitive exclusion principle (CEP) is a fundamental concept in the niche theory, which posits that the number of available resources constrains the coexistence of species. While the CEP offers an intuitive explanation on coexistence, it has been challenged by counterexamples observed in nature. One prominent counterexample is the phytoplankton community, known as the paradox of the plankton. Diverse phytoplankton species coexist in the ocean even though they demand a limited number of resources. To shed light on this remarkable biodiversity in large ecosystems quantitatively, we consider \textit{intraspecific suppression} into the generalized MacArthur's consumer-resource model and study the relative diversity, the number ratio between coexisting consumers and resource kinds. By employing the cavity method and generating functional analysis, we demonstrate that, under intraspecific suppression, the number of consumer species can surpass the available resources. This phenomenon stems from the fact that intraspecific suppression prevents the emergence of dominant species, thereby fostering high biodiversity. Furthermore, our study highlights that the impact of this competition on biodiversity is contingent upon environmental conditions. Our work presents a comprehensive framework that encompasses the CEP and its counterexamples by introducing intraspecific suppression.

The growing demand for precision agriculture necessitates efficient and accurate crop-weed recognition and classification systems. Current datasets often lack the sample size, diversity, and hierarchical structure needed to develop robust deep learning models for discriminating crops and weeds in agricultural fields. Moreover, the similar external structure and phenomics of crops and weeds complicate recognition tasks. To address these issues, we present the CWD30 dataset, a large-scale, diverse, holistic, and hierarchical dataset tailored for crop-weed recognition tasks in precision agriculture. CWD30 comprises over 219,770 high-resolution images of 20 weed species and 10 crop species, encompassing various growth stages, multiple viewing angles, and environmental conditions. The images were collected from diverse agricultural fields across different geographic locations and seasons, ensuring a representative dataset. The dataset's hierarchical taxonomy enables fine-grained classification and facilitates the development of more accurate, robust, and generalizable deep learning models. We conduct extensive baseline experiments to validate the efficacy of the CWD30 dataset. Our experiments reveal that the dataset poses significant challenges due to intra-class variations, inter-class similarities, and data imbalance. Additionally, we demonstrate that minor training modifications like using CWD30 pretrained backbones can significantly enhance model performance and reduce convergence time, saving training resources on several downstream tasks. These challenges provide valuable insights and opportunities for future research in crop-weed recognition. We believe that the CWD30 dataset will serve as a benchmark for evaluating crop-weed recognition algorithms, promoting advancements in precision agriculture, and fostering collaboration among researchers in the field.

Creep deformation in shale rocks is an important factor in many applications, such as the sustainability of geostructures, wellbore stability, evaluation of land subsidence, CO2 storage, toxic waste containment, and hydraulic fracturing. One mechanism leading to this time-dependent deformation under a constant load is the dissolution/formation processes accompanied by chemo-mechanical interactions with a reactive environment. When dissolution/formation processes occur within the material phases, the distribution of stress and strain within the material microstructure changes. In the case of the dissolution process, the stress carried by the dissolving phase is distributed into neighboring voxels, which leads to further deformation of the material. The aim of this study was to explore the relationship between the microstructural evolution and time-dependent creep behavior of rocks subjected to chemo-mechanical loading. This work uses the experimentally characterized microstructural and mechanical evolution of a shale rock induced by interactions with a reactive brine (CO2-rich brine) and a non-reactive brine (N2-rich brine) under high-pressure and high-temperature conditions to compute the resulting time-dependent deformation using a time-stepping finite-element-based modeling approach. Sample microstructure snapshots were obtained using segmented micro-CT images of the rock samples before and after the reactions. Coupled nanoindentation/EDS provided spatial alteration of the mechanical properties of individual material phases due to the dissolution and precipitation processes as a result of the chemo-mechanical loading of the samples. The time-dependent mechanically informed microstructures were then incorporated into a mechanical model to calculate the creep behavior caused by the dissolution/precipitation processes independent of the inherent viscous properties of the mineral phases.

The Schr\"odinger bridge problem (SBP) is gaining increasing attention in generative modeling and showing promising potential even in comparison with the score-based generative models (SGMs). SBP can be interpreted as an entropy-regularized optimal transport problem, which conducts projections onto every other marginal alternatingly. However, in practice, only approximated projections are accessible and their convergence is not well understood. To fill this gap, we present a first convergence analysis of the Schr\"odinger bridge algorithm based on approximated projections. As for its practical applications, we apply SBP to probabilistic time series imputation by generating missing values conditioned on observed data. We show that optimizing the transport cost improves the performance and the proposed algorithm achieves the state-of-the-art result in healthcare and environmental data while exhibiting the advantage of exploring both temporal and feature patterns in probabilistic time series imputation.

The ability of pollen to enable the glaciation of supercooled liquid water has been demonstrated in laboratory studies; however, the potential large-scale effect of trees and pollen on clouds, precipitation and climate is pressing knowledge to better understand and project clouds in the current and future climate. Combining ground-based measurements of pollen concentrations and satellite observations of cloud properties within the United States, we show that enhanced pollen concentrations during springtime lead to a higher cloud ice fraction. We further establish the link from the pollen-induced increase in cloud ice to a higher precipitation frequency. In light of anthropogenic climate change, the extended and strengthened pollen season and future alterations in biodiversity can introduce a localized climate forcing and a modification of the precipitation frequency and intensity.

The COVID-19 pandemic is considered as the most alarming global health calamity of this century. COVID-19 has been confirmed to be mutated from coronavirus family. As stated by the records of The World Health Organization (WHO at April 18 2020), the present epidemic of COVID-19, has influenced more than 2,164,111 persons and killed more than 146,198 folks in over 200 countries across the globe and billions had confronted impacts in lifestyle because of this virus outbreak. The ongoing overall outbreak of the COVID-19 opened up new difficulties to the research sectors. Artificial intelligence (AI) driven strategies can be valuable to predict the parameters, hazards, and impacts of such an epidemic in a cost-efficient manner. The fundamental difficulties of AI in this situation is the limited availability of information and the uncertain nature of the disease. Here in this article, we have tried to integrate AI to predict the infection outbreak and along with this, we have also tried to test whether AI with help deep learning can recognize COVID-19 infected chest X-Rays or not. The global outbreak of the virus posed enormous economic, ecological and societal challenges into the human population and with help of this paper, we have tried to give a message that AI can help us to identify certain features of the disease outbreak that could prove to be essential to protect the humanity from this deadly disease.

The overburden structures often can distort the responses of the target region in seismic data, especially in land datasets. Ideally, all effects of the overburden and underburden structures should be removed, leaving only the responses of the target region. This can be achieved using the Marchenko method. The Marchenko method is capable of estimating Green's functions between the surface of the Earth and arbitrary locations in the subsurface. These Green's functions can then be used to redatum wavefields to a level in the subsurface. As a result, the Marchenko method enables the isolation of the response of a specific layer or package of layers, free from the influence of the overburden and underburden. In this study, we apply the Marchenko-based isolation technique to land S-wave seismic data acquired in the Groningen province, the Netherlands. We apply the technique for combined removal of the overburden and underburden, which leaves the isolated response of the target region which is selected between 30 m and 270 m depth. Our results indicate that this approach enhances the resolution of reflection data. These enhanced reflections can be utilised for imaging and monitoring applications.

This work presents a novel domain adaption paradigm for studying contrastive self-supervised representation learning and knowledge transfer using remote sensing satellite data. Major state-of-the-art remote sensing visual domain efforts primarily focus on fully supervised learning approaches that rely entirely on human annotations. On the other hand, human annotations in remote sensing satellite imagery are always subject to limited quantity due to high costs and domain expertise, making transfer learning a viable alternative. The proposed approach investigates the knowledge transfer of selfsupervised representations across the distinct source and target data distributions in depth in the remote sensing data domain. In this arrangement, self-supervised contrastive learning-based pretraining is performed on the source dataset, and downstream tasks are performed on the target datasets in a round-robin fashion. Experiments are conducted on three publicly available datasets, UC Merced Landuse (UCMD), SIRI-WHU, and MLRSNet, for different downstream classification tasks versus label efficiency. In self-supervised knowledge transfer, the proposed approach achieves state-of-the-art performance with label efficiency labels and outperforms a fully supervised setting. A more in-depth qualitative examination reveals consistent evidence for explainable representation learning. The source code and trained models are published on GitHub.

We introduce a new methodology to study marine heat waves, extreme events in the sea surface temperature (SST) of the global ocean. Motivated by previously large and impactful marine heat waves and by theoretical expectation that the dominant heating processes coherently affect large regions of the ocean, we introduce a methodology from computer vision to construct marine heat wave systems (MWHSs) -- the collation of SST extrema in dimensions of area and time. We identify 649,475 MHWSs in the 37 year period (1983-2019) of daily SST records and find that the duration t_dur (days), maximum area A_max (km$^2$), and total ``volume'' N_vox (days km$^2$) for the majority of MHWSs are well-described by power-law distributions: t_dur^(-3), A_max^(-2) and N_vox^(-2). These characteristics confirm SST extrema exhibit strong spatial coherence that define the formation and evolution of marine heat waves. Furthermore, the most severe MHWSs deviate from these power-laws and are the dominant manifestation of marine heat waves: extrema in ocean heating are driven by the ~200 systems with largest area and duration. We further demonstrate that the previously purported rise in the incidence of marine heat wave events over the past decade is only significant in these severe systems. A change point analysis reveals a rapid increase in days under a severe MHW in most regions of the global ocean over the period of 2000-2005. Understanding the origin and impacts of marine heat waves in the current and future ocean, therefore, should focus on the production and evolution of the largest-scale and longest-duration heating phenomena.