Deutsch
 
Datenschutzhinweis Impressum
  DetailsucheBrowse

Datensatz

DATENSATZ AKTIONENEXPORT

Freigegeben

Zeitschriftenartikel

Maschinelles Lernen verbessert Gefährdungs- und Risikoanalysen von Naturgefahren

Urheber*innen
/persons/resource/kreib

Kreibich,  H.
5.4 Hydrology, 5.0 Geoarchives, Departments, GFZ Publication Database, Deutsches GeoForschungsZentrum;
Vol. 8, Issue 1 (2018), GFZ Journal 2018, System Erde : GFZ Journal, Deutsches GeoForschungsZentrum;

/persons/resource/kschroet

Schröter,  Kai
5.4 Hydrology, 5.0 Geoarchives, Departments, GFZ Publication Database, Deutsches GeoForschungsZentrum;
Vol. 8, Issue 1 (2018), GFZ Journal 2018, System Erde : GFZ Journal, Deutsches GeoForschungsZentrum;

/persons/resource/yshprits

Shprits,  Yuri
2.8 Magnetospheric Physics, 2.0 Geophysics, Departments, GFZ Publication Database, Deutsches GeoForschungsZentrum;
Vol. 8, Issue 1 (2018), GFZ Journal 2018, System Erde : GFZ Journal, Deutsches GeoForschungsZentrum;

/persons/resource/jbed

Bedford,  Jonathan
4.1 Lithosphere Dynamics, 4.0 Geomaterials, Departments, GFZ Publication Database, Deutsches GeoForschungsZentrum;
Vol. 8, Issue 1 (2018), GFZ Journal 2018, System Erde : GFZ Journal, Deutsches GeoForschungsZentrum;

/persons/resource/tilmann

Tilmann,  F.
2.4 Seismology, 2.0 Geophysics, Departments, GFZ Publication Database, Deutsches GeoForschungsZentrum;
Vol. 8, Issue 1 (2018), GFZ Journal 2018, System Erde : GFZ Journal, Deutsches GeoForschungsZentrum;

Volltexte (frei zugänglich)

GFZ_syserde.08.01.2.pdf
(Verlagsversion), 626KB

Ergänzendes Material (frei zugänglich)
Es sind keine frei zugänglichen Ergänzenden Materialien verfügbar
Zitation

Kreibich, H., Schröter, K., Shprits, Y., Bedford, J., Tilmann, F. (2018): Maschinelles Lernen verbessert Gefährdungs- und Risikoanalysen von Naturgefahren. - System Erde, 8, 1, 10-17.
https://doi.org/10.2312/GFZ.syserde.08.01.2


Zitierlink: https://gfzpublic.gfz-potsdam.de/pubman/item/item_3539918
Zusammenfassung
Hundreds of thousands of people are killed every year by disasters resulting from natural hazards, and property damage has been doubling about every seven years over the past four decades. Thus, understanding the causes and consequences of natural hazards and contributing to mitigation solutions is one of the grand challenges facing the GFZ German Research Centre for Geosciences. Increasingly, machine learning methods are being adopted by geoscientists as the wider scientific community begins to realise both the success and potential in employing a data science strategy to their research. GFZ scientists are using these methods to generate insights from large and heterogeneous data sets that would have been virtually impossible a decade ago. Such data sets consist of continuous recordings of natural and laboratory natural hazard data, hazard maps as well as exposure and damage data. In the case of supervised machine learning, predictions are optimized based upon features mined from the data and, in some cases, probabilities for specific events can be calculated. For instance, machine learning is applied for both classification of earthquake events and in attempts to predict future seismic behaviour in a target region. Additionally, machine learning, and in particular neural networks, are used to predict the planetary Kp-index and for reconstructions of the global plasmasphere dynamics depending on the geomagnetic conditions. Random forests and Bayesian Networks are used to analyse and model the complex damage processes during floods. The analysis and modelling of multi-hazard and cascading events is still a challenge, one that can also be tackled with machine learning. Thus, within the Natural Hazards research unit at GFZ, expertise with machine learning techniques is shared across disciplines and harmonised approaches of machine learning for improved multi-risk assessments are developed.