In the late 2000s, cloud computing emerged as a service model for the elastic – adaptable to dynamic demand – provision of computing resources in conjunction with a convenient payment model. Cloud offerings quickly diversified to provide multiple levels of computing abstraction – the hardware or infrastructure level (Infrastructure-as-a-Service (IaaS)), the middleware level (Platform-as-a-Service (PaaS)) and the application level (Software-as-a-Service (SaaS)) – enabling cloud applications of extraordinary complexity and scale [1]. There is now talk of the next generation of cloud computing as an evolving paradigm, known as serverless computing. This term has been increasingly used since the end of the 2010s and describes characteristics that differ from those of established cloud services: Serverless computing means the complete automation of the underlying resource management with fine-grained and usage-based billing. This makes it possible to run cloud applications without manual server management.
Processing and analysing earth observation data is often a big data problem due to its inherently high spatial and temporal extent and resolution. This places enormous demands on the computing infrastructure in terms of transmission volumes, memory consumption and computing capacity and is associated with the corresponding costs. For this reason, computing is traditionally carried out in specialised high-performance computing clusters, whereby data transport and location play a role that should not be underestimated. Access to such infrastructures and the associated analysis options were not available to the general public before cloud technologies.
Cloud-based platforms such as Google Earth Engine (GEE) [2], Code-DE and EO-Lab as part of Germany’s geoinformation strategy [3] as well as HPDA terrabyte [4] now allow access to petabyte-sized data catalogues and offer convenient analysis frameworks that can also be used free of charge to a certain extent. This is important for the democratisation of earth observation data and the acceleration of new and further developments.
One of the most frequently mentioned points of criticism is platform dependency, which is easy to underestimate. This means that a developed solution cannot be moved portably to another platform, or only with great effort.
Current challenges in the use of cloud technology for satellite data processing include the reproducibility of complex calculations for the transparent traceability of scientific results in this research domain, the standardisation and control of pre-processing steps for raw satellite data and the ability to estimate costs in advance of computing jobs on cloud resources.
Comparability with analogue phenomena
An analogue atlas, i.e. a collection of thematically, contextually or regionally related maps usually bound as a book, was a central source of information for many applications in the analogue age. Let’s remember, for example, road atlases, which were often carried on the back seat for navigation on car journeys, but very quickly became outdated within a few years. Digital navigation systems with up-to-date map material, enriched with real-time information on the traffic situation from a cloud, have probably largely replaced these road atlases. Digital enablers such as the high speed of computing processes, networking and data integration as well as the ability to generate, process and store large volumes of data have not only replaced road atlases with navigation systems. The broad spectrum of data and services, including in the context of cloud computing, has permanently changed satellite-based remote sensing, with a multitude of new possibilities and opportunities continuing to open up in this area.
Social relevance
Developing an in-depth understanding of the ongoing changes to our planet (based on remote sensing data), the effects of climate change and interactions between (eco)systems can be seen as a task for generations. This knowledge forms the basis for sustainable forms of living and economic activity. Cloud computing technologies have the potential to both democratise and accelerate this knowledge acquisition.
Sources
- Kounev, S. et al. (2023). Serverless Computing: What It Is, and What It Is Not? Commun. ACM 66, 9 (September 2023), 80–92. https://doi.org/10.1145/3587249
- Google Earth Engine.
- Code DE https://code-de.org/ und EO Lab https://eo-lab.org/ des Deutschen Zentrums für Luft- und Raumfahrt e.V. (DLR).
- HPDA terrabyte des DLR am Leibnitz Rechenzentrum LRZ in München.
