Polar night jet characterization through artificial intelligence

Simple item page
dc.contributor.authorRodríguez Montes, María
dc.contributor.authorAyarzagüena Porras, Blanca
dc.contributor.authorGuijarro Mata-García, María
dc.date.accessioned2023-06-22T10:54:24Z
dc.date.available2023-06-22T10:54:24Z
dc.date.issued2022-07-08
dc.descriptionCRUE-CSIC (Acuerdos Transformativos 2022)
dc.description.abstractThe stratospheric polar vortex is a cyclonic circulation that forms over the winter pole, whose edge is characterized by a strong westerly jet (also called polar night jet, PNJ). The PNJ plays a key role in processes such as the distribution of atmospheric constituents in the polar stratosphere or the wave propagation. Further, variations in PNJ can also affect the troposphere, being behind the occurrence of extreme events near the Earth’s surface. Thus, it is important to correctly characterize the mean state of the PNJ and its variability. Already existing algorithms, although working, may present several issues. The simplest ones, those based on zonal mean wind, can miss important information. In contrast, the 2-dimensional ones usually involve multiple calculations with several fields, some of them not always included in typical datasets. In this study, we describe a new artificial intelligence technique to characterize the PNJ. The algorithm only requires data of zonal wind that is classified each time step with a decision trees algorithm with 95.5% accuracy, trained with images processed by a climate science researcher. The classifier is applied to JRA-55 reanalysis data and the output of simulations of three climate models and is found to perform reasonably well when validated against traditional zonal-mean methods. Indeed, it provides more information about the PNJ, as it offers in one step the PNJ region, averaged magnitudes and even identify if the PNJ is under perturbed conditions. We have explored two examples of potential applications of the classifier such as the study of the influence of climate change on the PNJ and the variability of the PNJ on monthly and daily scales. In both cases, our algorithm has produced coherent results with those produced with previous studies, but with more detail obtained at a single step.
dc.description.departmentDepto. de Física de la Tierra y Astrofísica
dc.description.departmentDepto. de Arquitectura de Computadores y Automática
dc.description.facultyFac. de Ciencias Físicas
dc.description.facultyFac. de Informática
dc.description.refereedTRUE
dc.description.sponsorshipMinisterio de Ciencia e Innovación (MICINN)/FEDER
dc.description.statuspub
dc.eprint.idhttps://eprints.ucm.es/id/eprint/74083
dc.identifier.doi10.1016/j.cageo.2022.105176
dc.identifier.issn0098-3004
dc.identifier.officialurlhttps://doi.org/10.1016/j.cageo.2022.105176
dc.identifier.urihttps://hdl.handle.net/20.500.14352/71868
dc.journal.titleComputers & Geosciences
dc.language.isoeng
dc.page.initial105176
dc.publisherElsevier
dc.relation.projectIDRTI2018-094902-B-C21; JeDiS (RTI2018- 096402-B-I00)
dc.rightsAtribución-NoComercial-SinDerivadas 3.0 España
dc.rights.accessRightsopen access
dc.rights.urihttps://creativecommons.org/licenses/by-nc-nd/3.0/es/
dc.subject.keywordStratosphere
dc.subject.keywordPolar night jet
dc.subject.keywordRegion growing
dc.subject.keywordMachine learning
dc.subject.keywordClimate change
dc.subject.keywordDecision trees
dc.subject.ucmFísica atmosférica
dc.subject.ucmInteligencia artificial (Informática)
dc.subject.unesco2501 Ciencias de la Atmósfera
dc.subject.unesco1203.04 Inteligencia Artificial
dc.titlePolar night jet characterization through artificial intelligence
dc.typejournal article
dc.volume.number166
dspace.entity.typePublication
relation.isAuthorOfPublicationafac4741-04ec-4805-9476-53451704e8de
relation.isAuthorOfPublicationd5518066-7ea8-448c-8e86-42673e11a8ee
relation.isAuthorOfPublication.latestForDiscoveryafac4741-04ec-4805-9476-53451704e8de
Download
Original bundle
Now showing 1 - 1 of 1
Thumbnail Image
Name:
1-s2.0-S0098300422001297-main.pdf
Size:
3.81 MB
Format:
Adobe Portable Document Format
downoload Download
Collections
Artículos