Digitala Vetenskapliga Arkivet

Change search
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf
Probabilistic Weather Forecasting using Generative modeling
KTH, School of Engineering Sciences (SCI), Mathematics (Dept.).
2024 (English)Independent thesis Advanced level (degree of Master (Two Years)), 20 credits / 30 HE creditsStudent thesisAlternative title
Väderprognosensembler med hjälp av generativa modeller (Swedish)
Abstract [en]

This thesis explores the application of generative modeling techniques to probabilistic weather forecasting, specifically through the development of DEFfusion (Direct Ensemble Forecasting with Diffusion). DEFfusion aims to address limitations in current Machine Learning Weather Prediction (MLWP) models, which tend to produce blurry forecasts and rely heavily on iterative methods. By using a diffusion model, the thesis demonstrates the potential of direct forecasting to generate more accurate and computationally efficient weather predictions. The approach involves generating an ensemble of possible future weather states from a given initial condition, utilizing a Quasi-Geostrophic (QG) model for experimental feasibility. The methodology includes training a neural network to perform dimensionality reduction via an autoencoder and employing a diffusion model to handle the probabilistic aspects of forecasting. Evaluation metrics such as Root Mean Squared Error (RMSE), Continuous Ranked Probability Score (CRPS), and Skill/Spread ratio indicate that DEFfusion offers improved performance over iterative approaches. This research contributes to advancing MLWP models by enhancing their ability to quantify uncertainty and capture extreme events, paving the way for more reliable and efficient weather prediction systems.

Abstract [sv]

Denna masteruppsats utforskar tillämpningen av generativ modellering för sannolikhetsväderprognoser, specifikt genom utvecklingen av DEFfusion (Direct Ensemble Forecasting with Diffusion). DEFfusion syftar till att åtgärda begränsningarna i nuvarande maskinilärningsväderprognosmodeller, som tenderar att producera suddiga prognoser och använder iterativa metoder. Genom att använda en diffusionsmodell påvisar uppsatsen potentialen hos direkta prognoser för mer exakta och effektivare väderprognoser. Detta görs genom att generera en ensemble av möjliga framtida vädertillstånd från ett givet initialtillstånd för en kvasi-geostrofisk modell. Metoden inkluderar träning av ett neuralt nätverk för att utföra dimensionsreduktionen, och användningen av en diffusionsmodell för att hantera de probabilistiska aspekterna av prognoser. Evalueringsmått så som Root Mean Squared Error (RMSE), Continuous Ranked Probability Score (CRPS), och Skill/Spread ratio visar på att DEFfusion presterar bättre än traditionella iterativa metoder. Denna forskning bidrar till att utveckla maskinilärningsväderprognosmodeller genom att förbättra deras förmåga att kvantifiera osäkerhet och fånga extrema händelser, vilket banar väg för mer tillförlitliga och effektiva väderprognossystem.

Place, publisher, year, edition, pages
2024. , p. 71
Series
TRITA-SCI-GRU ; 2024:280
Keywords [en]
Weather Forecasting, Generative modeling, Uncertainty Estimation, Deep Learning, Stochastic Dynamical Systems, Diffusion
Keywords [sv]
Väderprognoser, generativa modeller, osäkerhetsestimering, djupinlärning, stokastiska dynamiska system, diffusion
National Category
Other Mathematics
Identifiers
URN: urn:nbn:se:kth:diva-361487OAI: oai:DiVA.org:kth-361487DiVA, id: diva2:1946045
External cooperation
SMHI
Subject / course
Mathematical Statistics
Educational program
Master of Science - Applied and Computational Mathematics
Supervisors
Examiners
Available from: 2025-03-20 Created: 2025-03-20 Last updated: 2025-03-20Bibliographically approved

Open Access in DiVA

fulltext(2042 kB)83 downloads
File information
File name FULLTEXT01.pdfFile size 2042 kBChecksum SHA-512
05aac03e452d7ab4828a7b8408623367f2536640ebc1b5574f971685ef83bc9ddbe6da336c4ee9722d7a59a35eacb173b3654fa001790e0a4d077f18b43cd5f6
Type fulltextMimetype application/pdf

By organisation
Mathematics (Dept.)
Other Mathematics

Search outside of DiVA

GoogleGoogle Scholar
Total: 83 downloads
The number of downloads is the sum of all downloads of full texts. It may include eg previous versions that are now no longer available

urn-nbn

Altmetric score

urn-nbn
Total: 295 hits
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf