Search

Offshore wind energy is at the advent of a massive global expansion. To investigate the development of the offshore wind energy sector, optimal offshore wind farm locations, or the impact of offshore wind farm projects, a freely accessible spatiotemporal data set of offshore wind energy infrastructure is necessary. With free and direct access to such data, it is more likely that all stakeholders who operate in marine and coastal environments will become involved in the upcoming massive expansion of offshore wind farms. To that end, we introduce the DeepOWT (Deep-learning-derived Offshore Wind Turbines) data set (available at https://doi.org/10.5281/zenodo.5933967, Hoeser and Kuenzer, 2022b), which provides 9941 offshore wind energy infrastructure locations along with their deployment stages on a global scale. DeepOWT is based on freely accessible Earth observation data from the Sentinel-1 radar mission. The offshore wind energy infrastructure locations were derived by applying deep-learning-based object detection with two cascading convolutional neural networks (CNNs) to search the entire Sentinel-1 archive on a global scale. The two successive CNNs have previously been optimised solely on synthetic training examples to detect the offshore wind energy infrastructures in real-world imagery. With subsequent temporal analysis of the radar signal at the detected locations, the DeepOWT data set reports the deployment stages of each infrastructure with a quarterly frequency from July 2016 until June 2021. The spatiotemporal information is compiled in a ready-to-use geographic information system (GIS) format to make the usability of the data set as accessible as possible.

In Earth observation (EO), large-scale land-surface dynamics are traditionally analyzed by investigating aggregated classes. The increase in data with a very high spatial resolution enables investigations on a fine-grained feature level which can help us to better understand the dynamics of land surfaces by taking object dynamics into account. To extract fine-grained features and objects, the most popular deep-learning model for image analysis is commonly used: the convolutional neural network (CNN). In this review, we provide a comprehensive overview of the impact of deep learning on EO applications by reviewing 429 studies on image segmentation and object detection with CNNs. We extensively examine the spatial distribution of study sites, employed sensors, used datasets and CNN architectures, and give a thorough overview of applications in EO which used CNNs. Our main finding is that CNNs are in an advanced transition phase from computer vision to EO. Upon this, we argue that in the near future, investigations which analyze object dynamics with CNNs will have a significant impact on EO research. With a focus on EO applications in this Part II, we complete the methodological review provided in Part I.

Deep learning (DL) has great influence on large parts of science and increasingly established itself as an adaptive method for new challenges in the field of Earth observation (EO). Nevertheless, the entry barriers for EO researchers are high due to the dense and rapidly developing field mainly driven by advances in computer vision (CV). To lower the barriers for researchers in EO, this review gives an overview of the evolution of DL with a focus on image segmentation and object detection in convolutional neural networks (CNN). The survey starts in 2012, when a CNN set new standards in image recognition, and lasts until late 2019. Thereby, we highlight the connections between the most important CNN architectures and cornerstones coming from CV in order to alleviate the evaluation of modern DL models. Furthermore, we briefly outline the evolution of the most popular DL frameworks and provide a summary of datasets in EO. By discussing well performing DL architectures on these datasets as well as reflecting on advances made in CV and their impact on future research in EO, we narrow the gap between the reviewed, theoretical concepts from CV and practical application in EO.

Rock glaciers in semiarid mountains contain large amounts of ice and might be important water stores aside from glaciers, lakes, and rivers. Yet whether and how rock glaciers interact with river channels in mountain valleys remains largely unresolved. We examine the potential for rock glaciers to block or disrupt river channels, using a new inventory of more than 2000 intact rock glaciers that we mapped from remotely sensed imagery in the Karakoram (KR), Tien Shan (TS), and Altai (ALT) mountains. We find that between 5% and 14% of the rock glaciers partly buried, blocked, diverted or constricted at least 95 km of mountain rivers in the entire study area. We use a Bayesian robust logistic regression with multiple topographic and climatic inputs to discern those rock glaciers disrupting mountain rivers from those with no obvious impacts. We identify elevation and potential incoming solar radiation (PISR), together with the size of feeder basins, as dominant predictors, so that lower-lying and larger rock glaciers from larger basins are more likely to disrupt river channels. Given that elevation and PISR are key inputs for modelling the regional distribution of mountain permafrost from the positions of rock-glacier toes, we infer that river-blocking rock glaciers may be diagnostic of non-equilibrated permafrost. Principal component analysis adds temperature evenness and wet-season precipitation to the controls that characterise rock glaciers impacting on rivers. Depending on the choice of predictors, the accuracy of our classification is moderate to good with median posterior area-under-the-curve values of 0.71-0.89. Clarifying whether rapidly advancing rock glaciers can physically impound rivers, or fortify existing dams instead, deserves future field investigation. We suspect that rock-glacier dams are conspicuous features that have a polygenetic history and encourage more research on the geomorphic coupling between permafrost lobes, river channels, and the sediment cascades of semiarid mountain belts. (c) 2018 John Wiley & Sons, Ltd.