Remote-sensing image scene classification can provide significant value, ranging from forest fire monitoring to land-use and land-cover classification. Beginning with the first aerial photographs of the early 20th century to the satellite imagery of today, the amount of remote-sensing data has increased geometrically with a higher resolution. The need to analyze these modern digital data motivated research to accelerate remote-sensing image classification. Fortunately, great advances have been made by the computer vision community to classify natural images or photographs taken with an ordinary camera. Natural image datasets can range up to millions of samples and are, therefore, amenable to deep-learning techniques. Many fields of science, remote sensing included, were able to exploit the success of natural image classification by convolutional neural network models using a technique commonly called transfer learning. We provide a systematic review of transfer learning application for scene classification using different datasets and different deep-learning models. We evaluate how the specialization of convolutional neural network models affects the transfer learning process by splitting original models in different points. As expected, we find the choice of hyperparameters used to train the model has a significant influence on the final performance of the models. Curiously, we find transfer learning from models trained on larger, more generic natural images datasets outperformed transfer learning from models trained directly on smaller remotely sensed datasets. Nonetheless, results show that transfer learning provides a powerful tool for remote-sensing scene classification.

Artificial intelligence methods have a very wide range of applications. From speech recognition to self-driving cars, the development of modern deep-learning architectures is helping researchers to achieve new levels of accuracy in different fields. Although deep convolutional neural networks (CNNs) (a kind of deep-learning technique) have reached or surpassed human-level performance in image recognition tasks, little has been done to transport this new image classification technology to geoscientific problems. We have developed what we believe to be the first use of CNNs to identify lithofacies in cores. We use highly accurate models (trained with millions of images) and transfer learning to classify images of cored carbonate rocks. We found that different modern CNN architectures can achieve high levels of lithologic image classification accuracy (approximately 90%) and can aid in the core description task. This core image classification technique has the potential to greatly standardize and accelerate the description process. We also provide the community with a new set of labeled data that can be used for further geologic/data science studies.

We present the recent progress in upgrading a predictive model for megaelectron‐volt (MeV) electrons inside the Earth's outer Van Allen belt. This updated model, called PreMevE 2.0, provides improved forecasts, particularly at outer L‐shells, by adding upstream solar wind speeds to the model's input parameter list that originally includes precipitating electrons observed at low Earth orbits and MeV electron fluxes in situ measured by a geosynchronous satellite. Furthermore, based on several kinds of linear and artificial neural networks algorithms, a list of models was constructed, trained, validated, and tested with 42‐month MeV electron observations from Van Allen Probes. Out‐of‐sample test results from these models show that, with optimized model hyperparameters and input parameter combinations, the top performer from each category of models has the similar capability of making reliable 1‐day (2‐day) forecasts of 1‐MeV electron flux distributions with performance efficiency values ~0.87 (~0.82) averaged over the L‐shell range of 2.8–6.6, significantly outperforming the previous version of PreMevE particularly at L‐shells > ~4.5. Interestingly, the linear regression model is often the most successful when compared to other models, which suggests the relationship between dynamics of trapped 1‐MeV electrons and precipitating electrons is dominated by linear components. Results also show that PreMevE 2.0 can reasonably well predict the onsets of MeV electron events in 2‐day forecasts. PreMevE 2.0 is designed to be driven by observations from longstanding space infrastructure to make high‐fidelity forecasts for MeV electrons, thus can be an invaluable space weather forecasting tool for the future.

Accurate taxonomic classification of microfossils in thin-sections is an important biostratigraphic procedure. As paleontological expertise is typically restricted to specific taxonomic groups and experts are not present in all institutions, geoscience researchers often suffer from lack of quick access to critical taxonomic knowledge for biostratigraphic analyses. Moreover, diminishing emphasis on education and training in systematics poses a major challenge for the future of biostratigraphy, and on associated endeavors reliant on systematics. Here we present a machine learning approach to classify and organize fusulinids—microscopic index fossils for the late Paleozoic. The technique we employ has the potential to use such important taxonomic knowledge in models that can be applied to recognize and categorize fossil specimens. Our results demonstrate that, given adequate images and training, convolutional neural network models can correctly identify fusulinids with high levels of accuracy. Continued efforts in digitization of biological and paleontological collections at numerous museums and adoption of machine learning by paleontologists can enable the development of highly accurate and easy-to-use classification tools and, thus, facilitate biostratigraphic analyses by non-experts as well as allow for cross-validation of disparate collections around the world. Automation of classification work would also enable expert paleontologists and others to focus efforts on exploration of more complex interpretations and concepts.

Despite the growing use of deep learning in sea ice mapping with SAR imagery, the study of model uncertainty and segmentation results remains limited. Deep learning models often produce overconfident predictions, a concern in sea ice mapping where misclassification can impact marine navigation safety. We incorporate and compare dropout and model ensemble within a convolutional neural network segmentation architecture to highlight regions with prediction uncertainty, and explore the impact of loss function choice. We evaluate model generalization and uncertainty characterization by training and evaluating models on the AI4Arctic Sea Ice Challenge Dataset (primary). We further explore model uncertainty by testing the trained models on the Extreme Earth version 2 Dataset (secondary). The primary and secondary datasets vary in number of scenes as well as in the available data and preprocessing. We obtain test F1 results higher than 0.97 for the primary dataset. Although the F1 performance for the secondary dataset is reduced to 0.93, the generated sea ice maps are reasonable across several Sentinel-1 scenes, and our proposed strategy helps in identification of misclassified and uncertain regions for human quality control. Our models seem to be robust against banding noise in Sentinel-1 SAR, and the prediction uncertainty frequently highlights ice regions misclassified as water, indicating its potential for real-world applications. Our study advances the field of machine learning-based sea ice mapping and highlights the importance of uncertainty estimation and cross-dataset evaluation for model development and deployment. Our approach can be adopted for other remote sensing applications as well.

Convolutional neural networks (CNN) are currently the most widely used tool for the classification of images, especially if such images have large within- and small between- group variance. Thus, one of the main factors driving the development of CNN models is the creation of large, labelled computer vision datasets, some containing millions of images. Thanks to transfer learning, a technique that modifies a model trained on a primary task to execute a secondary task, the adaptation of CNN models trained on such large datasets has rapidly gained popularity in many fields of science, geosciences included. However, the trade-off between two main components of the transfer learning methodology for geoscience images is still unclear: the difference between the datasets used in the primary and secondary tasks; and the amount of available data for the primary task itself. We evaluate the performance of CNN models pretrained with different types of image datasets—specifically, dermatology, histology, and raw food—that are fine-tuned to the task of petrographic thin-section image classification. Results show that CNN models pretrained on ImageNet achieve higher accuracy due to the larger number of samples, as well as a larger variability in the samples in ImageNet compared to the other datasets evaluated.