Information about clouds is important for observing and predicting weather and climate as well as for generating and distributing solar power. Most existing approaches extract cloud information from satellite data by classifying individual pixels instead of using closely integrated spatial information, ignoring the fact that clouds are highly dynamic, spatially continuous entities. This paper proposes a novel cloud classification method based on deep learning. Relying on a Convolutional Neural Network (CNN) architecture for image segmentation, the presented Cloud Segmentation CNN (CS-CNN), classifies all pixels of a scene simultaneously rather than individually. We show that CS-CNN can successfully process multispectral satellite data to classify continuous phenomena such as highly dynamic clouds. The proposed approach produces excellent results on Meteosat Second Generation (MSG) satellite data in terms of quality, robustness, and runtime compared to other machine learning methods such as random forests. In particular, comparing CS-CNN with the CLAAS-2 cloud mask derived from MSG data shows high accuracy (0.94) and Heidke Skill Score (0.90) values. In contrast to a random forest, CS-CNN produces robust results and is insensitive to challenges created by coast lines and bright (sand) surface areas. Using GPU acceleration, CS-CNN requires only 25 ms of computation time for classification of images of Europe with $508\times 508$ pixels.

There is a growing demand to train Earth Observation (EO) data users in how to access and use existing and upcoming data. A promising tool for data-related training is computational notebooks, which are interactive web applications that combine text, code and computational output. Here, we present the Learning Tool for Python (LTPy), which is a training course (based on Jupyter notebooks) on atmospheric composition data. LTPy consists of more than 70 notebooks and has taught over 1000 EO data users so far, whose feedback is overall positive. We adapted five guiding principles from different fields (mainly scientific computing and Jupyter notebook research) to make the Jupyter notebooks more educational and reusable. The Jupyter notebooks developed (i) follow the literate programming paradigm by a text/code ratio of 3, (ii) use instructional design elements to improve navigation and user experience, (iii) modularize functions to follow best practices for scientific computing, (iv) leverage the wider Jupyter ecosystem to make content accessible and (v) aim for being reproducible. We see two areas for future developments: first, to collect feedback and evaluate whether the instructional design elements proposed meet their objective; and second, to develop tools that automatize the implementation of best practices.

Cloud-based services introduce a paradigm shift in how users access, process and analyse Big Earth data. A key challenge is to align the current state of how users access, process and analyse the data with trends and roadmaps large data organisations layout. In addition, due to the increased availability of open data, a more diverse user base wants to take advantage of Earth science data leading to new user requirements. We run a web-based survey among Big Earth data users to better understand the motivation to migrate to cloud-based services as well as the challenges and opportunities that might arise. Results show an overall interest in moving to cloud-based services but air an insufficient literacy in cloud systems and a lack of trust due to security concerns and opacity of emerging costs. These gaps demand efforts on three levels. First, cloud services shall be targeted at intermediate users instead of policy- and decision-makers and over-engineered systems with a high level of abstraction should be avoided. Second, more substantial capacity-building efforts are required to decrease the existing gap in cloud skills and uptake. Third, a cloud certification mechanism could help in building up overall trust in cloud-based services.

Adaptivity is a challenging open issue in data stream management. In this paper, we tackle the problem of memory adaptivity inside a system executing temporal sliding window queries over continuous data streams. Two different techniques to control the memory usage at runtime are proposed which refer to changes in window sizes and time granularities. Both techniques differ from standard load shedding approaches based on sampling as they ensure precise query answers for user-defined Quality of Service (QoS) specifications, even under query re-optimization.