Astroparticle Physics is a young and evolving research area, where the amount of data continuously increase due to the modern instruments. Due to the geographical and experimental diversity a dedicated strategy for the digitalisation of the research eld is fundamental to astroparticle physics. For an e ective implementation with bene ts for science and society four strategic points were de ned: (i) Establishment of one or more global data centres. (ii) Development of methods for conservation of the measurement data. (iii) Development of applications of modern methods in data analysis. (iv) Expansion of courses for the training of young scientists in modern analytical methods. Aim of the initiative of an analysis and data centre in astroparticle physics is to develop and implement an interdisciplinary concept, which meets the needs of the digitization of the research eld and which is also attractive to society. The goal is to enable a more e cient analysis of the data that are recorded in di erent locations around the world (multi-messenger analyses), as well as modern training for Big Data Scientists in the synergy between basic research and the information society.
Understanding the high-energy Universe in the context of Astroparticle Physics
means rst and in foremost to answer the urgent question of the origin of
highenergy cosmic rays. This question could not be answered by roughly 100 years
of measurements of cosmic rays since their discovery in 1912 by a series of
balloon ights of Victor Hess [
Only charged particles can be accelerated in the source regions of the Universe. However, in these acceleration processes also secondary neutral products are generated, like gamma-rays and neutrinos. Once produced, these particles should travel straight on from the source to Earth. Neutrinos are very weakly interacting with material (detectors) and therefore di cult to measure. Gammaray measurements are among others motivated by studying them as tracers from charged cosmic-ray sources, but su er from absorption by the infrared and microwave background in our Universe. Due to all these di culties it became clear that only a combination of various measurements of the di erent tracers - not to forget the rare, but meanwhile detected, catastrophic events in our Universe generating Gravitational Waves - will bring us closer to an understanding of the astrophysics of the High-Energy Universe. This approach is called MultiMessenger Astroparticle Physics ( g. 1). 2
From the theory point of view multi-messenger astroparticle physics is to
connect the physics of the sources with the observations in gamma-rays, neutrinos,
and cosmic rays, and to use these observations to learn about the generation,
acceleration and propagation properties [
More events of parallel observations of the same source region will certainly
be provided in the next years, in particular by the new or enhanced
largescale observatories CTA [
With the help of modern Information Technology, Big Data Analytics and Research Data Management, the basic goal is to enter a new era of multi-messenger astroparticle physics. This can only be reached if several aspects of a coherent development are considered, where a dedicated analysis and data centre is a very important ingredient. It will not only provide the tools and environment to take the step into this new physics era, but also allows to take the digitized society along the path a must do in modern Big Data Science. Nowadays, basic research in the eld of particle physics, astroparticle physics, nuclear physics, astrophysics or astronomy is performed in large international collaborations with partly huge infrastructures producing large amounts of valuable scienti c data. To e ciently use all the information to solve the still mysterious question about the origin of matter and the Universe, a broad, simple and sustainable access to the scienti c data from these infrastructures has to be provided. In a general way, such a global data centre has to provide a vast of functionalities, at least covering the following pillars ( g. 2): { Data availability: All participating researchers of the individual experiments or facilities need a fast and simple access to the relevant data. { Analysis: A fast access to the Big Data from measurements and simulations is needed. { Simulations & Methods development: To prepare the analyses of the data the researchers need a mighty environment on computing power for the production of relevant simulations and the development of new methods, e.g. by deep machine learning. { Education in Data Science: The handling of the centre as well as the processing of the data needs specialized education in Big Data Science. { Open access: It becomes more and more important to provide the scienti c data not only to the internal research community, but also to the interested public: Public Data for Public Money! { Data archive: The valuable scienti c data need to be preserved for a later (re-)use.
All the present and future large-scale observatories mentioned above will provide their scienti c data via sophisticated infrastructures and data centres for internal and also external use. However, information from various experiments and various messengers like charged particles, gamma-rays or neutrinos, measured by di erent globally distributed large-scale facilities, have to be combined. For that a diverse set of astrophysical data is required to be made available and public as well as a framework for developing tools and methods to handle the (open and collaborative) data.
We aim to extend the current activities in the individual observatories on an
experiment-overarching, global and international level. This includes the
validation of both, providing public access to (an initial part of) the scienti c data
and using the computing environment by the involved researchers to perform
multi-messenger analyses. A further goal is to standardize the data, to make
the publication FAIR [
Whereas in Astronomy and Particle Physics data centres are already established,
which ful ll a part of the above mentioned requirements (although, not the same
parts), in Astroparticle Physics only rst attempts are presently under
development. For example, at the KASCADE experiment [
It is obvious that astroparticle physics has become a data intensive science with many terabytes of data and often with tens of measured parameters associated to each observation. Moreover new highly complex and massively large datasets are expected by novel and more complex scienti c instruments as well as simulated data needed for interpretation that will become available in the next decades, probably largely used by the community far beyond the next decade. Handling and exploring these new data volumes, and actually making unexpected scienti c discoveries, poses a considerable technical challenge that requires the adoption of new approaches in using heterogeneous computing and storage resources and in organizing scienti c collaborations, scienti c education and science communication, where sophisticated public data centres will play the key role.
Based on the experience gained with KCDC [
Here, a brief summary of the realisation of KCDC is given: When publishing data it is not enough to put some ASCII les on a plain webpage. To ensure that potential users can actually use the data, an extensive documentation on how the data has been obtained is needed. Depending on the kind of data, this meta-data is at least a description of the detector and the reconstruction procedures employed, but can also consist of parameters determined by the raw data as well as event identi ers like UUIDs. Since this information, in addition to a license agreement, is not expected to change often, creating some static pages is a viable option. Another important aspect is the user and access management. For KCDC it was the concept that only registered users get access to the data and to the detailed data shop. Therefore, it is su cient to check if an account is active and the user is authenticated. While there is already a basic implementation of a permission based access limitation, a useful categorization of the users into - possibly hierarchical - groups is needed (no administrator should manually manage privileges of single users) to e ectively use it. The heart of the data centre is the data shop. The design goals have to ensure an easy access to an user de ned subset of the whole dataset, a natural way to con gure additional detector components or observables without the need to change the code-base and a clear overview of previous requests with the possibility to use these selections as a template for future requests. For KCDC, the data are stored in a MongoDB. Its scheme-less design allows us to collect all available information of an event, although the available detector components may vary for each event. In principle KCDC can use any kind of input format, as these are implemented as plugins. Currently three output formats are supported, HDF5, ROOT and ASCII. These can be extended by adding additional plugins, too. The requests are processed using a Celery based task queue. The simultaneous processing of requests can be achieved by adding more worker processes, which can be distributed among several machines. Together with the realisation to run the MongoDB on a sharded cluster, a scaling of the needed processing power with the demand is achieved.
KCDC is a web portal where the KASCADE scienti c data is made available for
the interested public. In Russia, there is the operating TAIGA and Tunka-Rex
facilities [
The German part of the GRADCLI project [
Big Data Science Software: Advancement of Big Data Science: The data centre shall allow not only access to the data, but also provide the possibility of developing speci c analysis methods and perform corresponding simulations. The concept to reach this goal is the installation of a dedicated Data Life Cycle Lab.
To reach this goal rst, the basic software of the KCDC data centre needs to be improved. Then the data centre has to be moved to the Big Data environment provided by KIT-SCC. The software is being built as a modular, exible framework with a good scalability (e.g. to large computing centres). The con guration is hold to be simple and doable also via a web interface. The entire software will be based solely on Open Source Software (Python, Django, mongoDB, HDF5, ROOT, etc.). Having done this step, the result enables to install and publish a dedicated Data Life Cycle Lab for the astroparticle physics community. Dedicated access, storage, and interface software has to be developed. In addition, the appropriate hardware has to be installed and commissioned. The described concept and working plan is valid also for the next steps, i.e. to generalize and consolidate the software package of a global astroparticle data centre for public and scienti c use.
Multi-Messenger Data Analysis: Speci c analyses of the data provided by
the new data centre will be performed to test the entire concept giving
important contributions and con dence to the centre as a valuable scienti c tool.
For a detailed reliability test we aim for common data analyses using the
complementary information of independent experiments, in particular to apply deep
learning (machine learning) methods to multi-messenger analyses [
Go for the public: A coherent outreach of the project, including example applications for all level of users - from pupils to the directly involved scientists - with detailed tutorials and documentation is an important ingredient of any activity in publishing scienti c data.
In particular the documentation, i.e. the metadata, of any released dataset has to be prepared with reasonable care. The goal of having detailed tutorials, i.e. an education portal, is to provide the data also to a general public in the sense of a visible outreach of astroparticle physics. The target groups for the tutorials are teachers and pupils in high schools. A tutorial has to provide at least: i) a basic knowledge on the experiment, astrophysics and related topics; ii) the required software and KCDC data (preferably as a pre-selection); iii) a step by step explanation of a simple data analysis; iv) a modern programming language code example; v) the interpretation and discussion of the outcome. With increasing number of users also the user management system (e.g. Q&A sections, discussion blogs, etc.) have to be extended and maintained. This part includes also creating and designing outreach materials, conducting routine website maintenance, preparing text or images. In addition, web pages, press releases, distribution of the progress and news in social networks like Facebook, Twitter, etc. needs to be managed. 7
Several initiatives have been started towards a dedicated and global Analysis and Data Centre in Astroparticle Physics. Aim of these initiatives is to develop and implement an interdisciplinary framework, which meets the needs of the digitization of the research eld and which is also attractive to society. One of the immediate goals is to enable a more e cient analysis of the data that is recorded in di erent locations around the world for coherent multi-messenger studies in order to better understand the high-energy processes in our Universe.