Literature Review of Deep Learning in Medium Range Weather Forecasting

No Thumbnail Available

Files

URL

Journal Title

Journal ISSN

Volume Title

Perustieteiden korkeakoulu | Bachelor's thesis
Electronic archive copy is available locally at the Harald Herlin Learning Centre. The staff of Aalto University has access to the electronic bachelor's theses by logging into Aaltodoc with their personal Aalto user ID. Read more about the availability of the bachelor's theses.
Unless otherwise stated, all rights belong to the author. You may download, display and print this publication for Your own personal use. Commercial use is prohibited.

Authors

Date

2024-12-13

Department

Major/Subject

Data Science

Mcode

SCI3095

Degree programme

Aalto Bachelor’s Programme in Science and Technology

Language

en

Pages

21+4

Series

Abstract

Weather forecasting is essential for many industries, informing decisions that safeguard operations, resources, and public safety. For decades, Numerical Weather Prediction (NWP) models-sophisticated simulations based on atmospheric physics-have served as the principal instruments for producing forecasts. Nonetheless, these models need considerable computer resources and have intrinsic limits associated with the chaotic characteristics of the environment. Consequently, deep learning (DL) has arisen as a viable data-driven alternative, proficient at identifying intricate patterns from extensive datasets and occasionally attaining predictive accuracy comparable to traditional numerical weather prediction (NWP) techniques. This thesis offers a literature analysis of the utilization of deep learning models in medium-range weather forecasting (about three to ten days). The analysis examines their precision, computational efficiency, and applicability in comparison to conventional NWP models. Results indicate that some deep learning models can either surpass or equal the accuracy of conventional numerical weather prediction methods at particular lead times, provide lower computing expenses, and enable expedited inference. However, problems concerning the method remain, including data quality standards, interpretability concerns, difficulties in managing extreme occurrences, and model generalizability under fluctuating climate circumstances. The findings of this paper indicate that deep learning should be regarded as a supplementary method rather than a substitute for numerical weather prediction. The integration of deep learning approaches with physics-based modeling and enhanced data assimilation can increase forecasting capabilities. Future endeavors should concentrate on hybrid techniques that utilize the advantages of both methodologies, resulting in more resilient, accessible, and economical medium-range weather forecasting systems.

Description

Supervisor

Korpi-Lagg, Maarit

Thesis advisor

Heinonen, Markus

Keywords

deep learning, weather forecasting

Other note

Citation

Permanent link to this item

https://urn.fi/URN:NBN:fi:aalto-202502182438

Collections

[kand] Perustieteiden korkeakoulu / SCI